s3fs-fuse/src/s3fs.cpp

4746 lines
137 KiB
C++
Raw Normal View History

/*
* s3fs - FUSE-based file system backed by Amazon S3
*
* Copyright 2007-2008 Randy Rizun <rrizun@gmail.com>
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <stdint.h>
#include <dirent.h>
#include <errno.h>
#include <sys/stat.h>
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
#include <sys/types.h>
#include <libxml/xpath.h>
#include <libxml/xpathInternals.h>
#include <libxml/tree.h>
#include <curl/curl.h>
#include <pwd.h>
#include <grp.h>
#include <getopt.h>
#include <sys/xattr.h>
#include <fstream>
#include <vector>
#include <algorithm>
#include <map>
#include <string>
#include <list>
#include "common.h"
#include "s3fs.h"
#include "curl.h"
#include "cache.h"
#include "string_util.h"
#include "s3fs_util.h"
#include "fdcache.h"
#include "s3fs_auth.h"
using namespace std;
//-------------------------------------------------------------------
// Define
//-------------------------------------------------------------------
#define DIRTYPE_UNKNOWN -1
#define DIRTYPE_NEW 0
#define DIRTYPE_OLD 1
#define DIRTYPE_FOLDER 2
#define DIRTYPE_NOOBJ 3
#define IS_REPLACEDIR(type) (DIRTYPE_OLD == type || DIRTYPE_FOLDER == type || DIRTYPE_NOOBJ == type)
#define IS_RMTYPEDIR(type) (DIRTYPE_OLD == type || DIRTYPE_FOLDER == type)
#if !defined(ENOATTR)
#define ENOATTR ENODATA
#endif
//-------------------------------------------------------------------
// Structs
//-------------------------------------------------------------------
typedef struct uncomplete_multipart_info{
string key;
string id;
string date;
}UNCOMP_MP_INFO;
typedef std::list<UNCOMP_MP_INFO> uncomp_mp_list_t;
//-------------------------------------------------------------------
// Global valiables
//-------------------------------------------------------------------
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
bool debug = false;
bool foreground = false;
bool foreground2 = false;
bool nomultipart = false;
bool pathrequeststyle = false;
bool is_specified_endpoint = false;
std::string program_name;
std::string service_path = "/";
std::string host = "http://s3.amazonaws.com";
std::string bucket = "";
std::string endpoint = "us-east-1";
//-------------------------------------------------------------------
// Static valiables
//-------------------------------------------------------------------
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
static uid_t mp_uid = 0; // owner of mount point(only not specified uid opt)
static gid_t mp_gid = 0; // group of mount point(only not specified gid opt)
static mode_t mp_mode = 0; // mode of mount point
static mode_t mp_umask = 0; // umask for mount point
static bool is_mp_umask = false;// default does not set.
static std::string mountpoint;
static std::string passwd_file = "";
static bool utility_mode = false;
static bool noxmlns = false;
static bool nocopyapi = false;
static bool norenameapi = false;
static bool nonempty = false;
static bool allow_other = false;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
static uid_t s3fs_uid = 0;
static gid_t s3fs_gid = 0;
static mode_t s3fs_umask = 0;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
static bool is_s3fs_uid = false;// default does not set.
static bool is_s3fs_gid = false;// default does not set.
static bool is_s3fs_umask = false;// default does not set.
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
static bool is_remove_cache = false;
static bool create_bucket = false;
static int64_t singlepart_copy_limit = FIVE_GB;
//-------------------------------------------------------------------
// Static functions : prototype
//-------------------------------------------------------------------
static bool is_special_name_folder_object(const char* path);
static int chk_dir_object_type(const char* path, string& newpath, string& nowpath, string& nowcache, headers_t* pmeta = NULL, int* pDirType = NULL);
static int get_object_attribute(const char* path, struct stat* pstbuf, headers_t* pmeta = NULL, bool overcheck = true, bool* pisforce = NULL);
static int check_object_access(const char* path, int mask, struct stat* pstbuf);
static int check_object_owner(const char* path, struct stat* pstbuf);
static int check_parent_object_access(const char* path, int mask);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
static FdEntity* get_local_fent(const char* path, bool is_load = false);
static bool multi_head_callback(S3fsCurl* s3fscurl);
static S3fsCurl* multi_head_retry_callback(S3fsCurl* s3fscurl);
static int readdir_multi_head(const char* path, S3ObjList& head, void* buf, fuse_fill_dir_t filler);
static int list_bucket(const char* path, S3ObjList& head, const char* delimiter, bool check_content_only = false);
static int directory_empty(const char* path);
static bool is_truncated(xmlDocPtr doc);;
static int append_objects_from_xml_ex(const char* path, xmlDocPtr doc, xmlXPathContextPtr ctx,
const char* ex_contents, const char* ex_key, const char* ex_etag, int isCPrefix, S3ObjList& head);
static int append_objects_from_xml(const char* path, xmlDocPtr doc, S3ObjList& head);
static bool GetXmlNsUrl(xmlDocPtr doc, string& nsurl);
static xmlChar* get_base_exp(xmlDocPtr doc, const char* exp);
static xmlChar* get_prefix(xmlDocPtr doc);
static xmlChar* get_next_marker(xmlDocPtr doc);
static char* get_object_name(xmlDocPtr doc, xmlNodePtr node, const char* path);
2014-08-26 17:11:10 +00:00
static int put_headers(const char* path, headers_t& meta, bool is_copy);
static int rename_large_object(const char* from, const char* to);
static int create_file_object(const char* path, mode_t mode, uid_t uid, gid_t gid);
static int create_directory_object(const char* path, mode_t mode, time_t time, uid_t uid, gid_t gid);
static int rename_object(const char* from, const char* to);
static int rename_object_nocopy(const char* from, const char* to);
static int clone_directory_object(const char* from, const char* to);
static int rename_directory(const char* from, const char* to);
static int remote_mountpath_exists(const char* path);
static xmlChar* get_exp_value_xml(xmlDocPtr doc, xmlXPathContextPtr ctx, const char* exp_key);
static void print_uncomp_mp_list(uncomp_mp_list_t& list);
static bool abort_uncomp_mp_list(uncomp_mp_list_t& list);
static bool get_uncomp_mp_list(xmlDocPtr doc, uncomp_mp_list_t& list);
2015-06-06 16:39:39 +00:00
static void free_xattrs(xattrs_t& xattrs);
static bool parse_xattr_keyval(const std::string& xattrpair, string& key, PXATTRVAL& pval);
static size_t parse_xattrs(const std::string& strxattrs, xattrs_t& xattrs);
static std::string build_xattrs(const xattrs_t& xattrs);
static int s3fs_utility_mode(void);
static int s3fs_check_service(void);
static int check_for_aws_format(void);
static int check_passwd_file_perms(void);
static int read_passwd_file(void);
static int get_access_keys(void);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
static int set_moutpoint_attribute(struct stat& mpst);
static int my_fuse_opt_proc(void* data, const char* arg, int key, struct fuse_args* outargs);
// fuse interface functions
static int s3fs_getattr(const char* path, struct stat* stbuf);
static int s3fs_readlink(const char* path, char* buf, size_t size);
static int s3fs_mknod(const char* path, mode_t mode, dev_t rdev);
static int s3fs_mkdir(const char* path, mode_t mode);
static int s3fs_unlink(const char* path);
static int s3fs_rmdir(const char* path);
static int s3fs_symlink(const char* from, const char* to);
static int s3fs_rename(const char* from, const char* to);
static int s3fs_link(const char* from, const char* to);
static int s3fs_chmod(const char* path, mode_t mode);
static int s3fs_chmod_nocopy(const char* path, mode_t mode);
static int s3fs_chown(const char* path, uid_t uid, gid_t gid);
static int s3fs_chown_nocopy(const char* path, uid_t uid, gid_t gid);
static int s3fs_utimens(const char* path, const struct timespec ts[2]);
static int s3fs_utimens_nocopy(const char* path, const struct timespec ts[2]);
static int s3fs_truncate(const char* path, off_t size);
static int s3fs_create(const char* path, mode_t mode, struct fuse_file_info* fi);
static int s3fs_open(const char* path, struct fuse_file_info* fi);
static int s3fs_read(const char* path, char* buf, size_t size, off_t offset, struct fuse_file_info* fi);
static int s3fs_write(const char* path, const char* buf, size_t size, off_t offset, struct fuse_file_info* fi);
static int s3fs_statfs(const char* path, struct statvfs* stbuf);
static int s3fs_flush(const char* path, struct fuse_file_info* fi);
2015-03-10 16:18:03 +00:00
static int s3fs_fsync(const char* path, int datasync, struct fuse_file_info* fi);
static int s3fs_release(const char* path, struct fuse_file_info* fi);
static int s3fs_opendir(const char* path, struct fuse_file_info* fi);
static int s3fs_readdir(const char* path, void* buf, fuse_fill_dir_t filler, off_t offset, struct fuse_file_info* fi);
static int s3fs_access(const char* path, int mask);
static void* s3fs_init(struct fuse_conn_info* conn);
static void s3fs_destroy(void*);
#if defined(__APPLE__)
static int s3fs_setxattr(const char* path, const char* name, const char* value, size_t size, int flags, uint32_t position);
static int s3fs_getxattr(const char* path, const char* name, char* value, size_t size, uint32_t position);
#else
static int s3fs_setxattr(const char* path, const char* name, const char* value, size_t size, int flags);
static int s3fs_getxattr(const char* path, const char* name, char* value, size_t size);
#endif
static int s3fs_listxattr(const char* path, char* list, size_t size);
static int s3fs_removexattr(const char* path, const char* name);
//-------------------------------------------------------------------
// Functions
//-------------------------------------------------------------------
static bool is_special_name_folder_object(const char* path)
{
if(!path || '\0' == path[0]){
return false;
}
string strpath = path;
headers_t header;
if(string::npos == strpath.find("_$folder$", 0)){
if('/' == strpath[strpath.length() - 1]){
strpath = strpath.substr(0, strpath.length() - 1);
}
strpath += "_$folder$";
}
S3fsCurl s3fscurl;
if(0 != s3fscurl.HeadRequest(strpath.c_str(), header)){
return false;
}
header.clear();
S3FS_MALLOCTRIM(0);
return true;
}
// [Detail]
// This function is complicated for checking directory object type.
// Arguments is used for deleting cache/path, and remake directory object.
// Please see the codes which calls this function.
//
// path: target path
// newpath: should be object path for making/putting/getting after checking
// nowpath: now object name for deleting after checking
// nowcache: now cache path for deleting after checking
// pmeta: headers map
// pDirType: directory object type
//
static int chk_dir_object_type(const char* path, string& newpath, string& nowpath, string& nowcache, headers_t* pmeta, int* pDirType)
{
int TypeTmp;
int result = -1;
bool isforce = false;
int* pType = pDirType ? pDirType : &TypeTmp;
// Normalize new path.
newpath = path;
if('/' != newpath[newpath.length() - 1]){
string::size_type Pos;
if(string::npos != (Pos = newpath.find("_$folder$", 0))){
newpath = newpath.substr(0, Pos);
}
newpath += "/";
}
// Alwayes check "dir/" at first.
if(0 == (result = get_object_attribute(newpath.c_str(), NULL, pmeta, false, &isforce))){
// Found "dir/" cache --> Check for "_$folder$", "no dir object"
nowcache = newpath;
if(is_special_name_folder_object(newpath.c_str())){
// "_$folder$" type.
(*pType) = DIRTYPE_FOLDER;
nowpath = newpath.substr(0, newpath.length() - 1) + "_$folder$"; // cut and add
}else if(isforce){
// "no dir object" type.
(*pType) = DIRTYPE_NOOBJ;
nowpath = "";
}else{
nowpath = path;
if(0 < nowpath.length() && '/' == nowpath[nowpath.length() - 1]){
// "dir/" type
(*pType) = DIRTYPE_NEW;
}else{
// "dir" type
(*pType) = DIRTYPE_OLD;
}
}
}else{
// Check "dir"
nowpath = newpath.substr(0, newpath.length() - 1);
if(0 == (result = get_object_attribute(nowpath.c_str(), NULL, pmeta, false, &isforce))){
// Found "dir" cache --> this case is only "dir" type.
// Because, if object is "_$folder$" or "no dir object", the cache is "dir/" type.
// (But "no dir objet" is checked here.)
nowcache = nowpath;
if(isforce){
(*pType) = DIRTYPE_NOOBJ;
nowpath = "";
}else{
(*pType) = DIRTYPE_OLD;
}
}else{
// Not found cache --> check for "_$folder$" and "no dir object".
nowcache = ""; // This case is no cahce.
nowpath += "_$folder$";
if(is_special_name_folder_object(nowpath.c_str())){
// "_$folder$" type.
(*pType) = DIRTYPE_FOLDER;
result = 0; // result is OK.
}else if(-ENOTEMPTY == directory_empty(newpath.c_str())){
// "no dir object" type.
(*pType) = DIRTYPE_NOOBJ;
nowpath = ""; // now path.
result = 0; // result is OK.
}else{
// Error: Unknown type.
(*pType) = DIRTYPE_UNKNOWN;
newpath = "";
nowpath = "";
}
}
}
return result;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
//
// Get object attributes with stat cache.
// This function is base for s3fs_getattr().
//
// [NOTICE]
// Checking order is changed following list because of reducing the number of the requests.
// 1) "dir"
// 2) "dir/"
// 3) "dir_$folder$"
//
static int get_object_attribute(const char* path, struct stat* pstbuf, headers_t* pmeta, bool overcheck, bool* pisforce)
{
int result = -1;
struct stat tmpstbuf;
struct stat* pstat = pstbuf ? pstbuf : &tmpstbuf;
headers_t tmpHead;
headers_t* pheader = pmeta ? pmeta : &tmpHead;
string strpath;
S3fsCurl s3fscurl;
bool forcedir = false;
string::size_type Pos;
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
FPRNINFO("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(!path || '\0' == path[0]){
return -ENOENT;
}
memset(pstat, 0, sizeof(struct stat));
if(0 == strcmp(path, "/") || 0 == strcmp(path, ".")){
pstat->st_nlink = 1; // see fuse faq
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
pstat->st_mode = mp_mode;
pstat->st_uid = is_s3fs_uid ? s3fs_uid : mp_uid;
pstat->st_gid = is_s3fs_gid ? s3fs_gid : mp_gid;
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return 0;
}
// Check cache.
strpath = path;
if(overcheck && string::npos != (Pos = strpath.find("_$folder$", 0))){
strpath = strpath.substr(0, Pos);
strpath += "/";
}
if(pisforce){
(*pisforce) = false;
}
if(StatCache::getStatCacheData()->GetStat(strpath, pstat, pheader, overcheck, pisforce)){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return 0;
}
if(StatCache::getStatCacheData()->IsNoObjectCache(strpath)){
// there is the path in the cache for no object, it is no object.
return -ENOENT;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
// At first, check path
strpath = path;
result = s3fscurl.HeadRequest(strpath.c_str(), (*pheader));
s3fscurl.DestroyCurlHandle();
// overcheck
if(overcheck && 0 != result){
if('/' != strpath[strpath.length() - 1] && string::npos == strpath.find("_$folder$", 0)){
// path is "object", check "object/" for overcheck
strpath += "/";
result = s3fscurl.HeadRequest(strpath.c_str(), (*pheader));
s3fscurl.DestroyCurlHandle();
}
if(0 != result){
// not found "object/", check "_$folder$"
strpath = path;
if(string::npos == strpath.find("_$folder$", 0)){
if('/' == strpath[strpath.length() - 1]){
strpath = strpath.substr(0, strpath.length() - 1);
}
strpath += "_$folder$";
result = s3fscurl.HeadRequest(strpath.c_str(), (*pheader));
s3fscurl.DestroyCurlHandle();
}
}
if(0 != result){
// not found "object/" and "object_$folder$", check no dir object.
strpath = path;
if(string::npos == strpath.find("_$folder$", 0)){
if('/' == strpath[strpath.length() - 1]){
strpath = strpath.substr(0, strpath.length() - 1);
}
if(-ENOTEMPTY == directory_empty(strpath.c_str())){
// found "no dir obejct".
strpath += "/";
forcedir = true;
if(pisforce){
(*pisforce) = true;
}
result = 0;
}
}
}
}else{
// found "path" object.
if('/' != strpath[strpath.length() - 1]){
// check a case of that "object" does not have attribute and "object" is possible to be directory.
if(is_need_check_obj_detail(*pheader)){
if(-ENOTEMPTY == directory_empty(strpath.c_str())){
strpath += "/";
forcedir = true;
if(pisforce){
(*pisforce) = true;
}
result = 0;
}
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
}
if(0 != result){
// finally, "path" object did not find. Add no object cache.
strpath = path; // reset original
StatCache::getStatCacheData()->AddNoObjectCache(strpath);
return result;
}
// if path has "_$folder$", need to cut it.
if(string::npos != (Pos = strpath.find("_$folder$", 0))){
strpath = strpath.substr(0, Pos);
strpath += "/";
}
// Set into cache
if(0 != StatCache::getStatCacheData()->GetCacheSize()){
// add into stat cache
if(!StatCache::getStatCacheData()->AddStat(strpath, (*pheader), forcedir)){
DPRN("failed adding stat cache [path=%s]", strpath.c_str());
return -ENOENT;
}
if(!StatCache::getStatCacheData()->GetStat(strpath, pstat, pheader, overcheck, pisforce)){
// There is not in cache.(why?) -> retry to convert.
if(!convert_header_to_stat(strpath.c_str(), (*pheader), pstat, forcedir)){
DPRN("failed convert headers to stat[path=%s]", strpath.c_str());
return -ENOENT;
}
}
}else{
// cache size is Zero -> only convert.
if(!convert_header_to_stat(strpath.c_str(), (*pheader), pstat, forcedir)){
DPRN("failed convert headers to stat[path=%s]", strpath.c_str());
return -ENOENT;
}
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return 0;
}
//
// Check the object uid and gid for write/read/execute.
// The param "mask" is as same as access() function.
// If there is not a target file, this function returns -ENOENT.
// If the target file can be accessed, the result always is 0.
//
// path: the target object path
// mask: bit field(F_OK, R_OK, W_OK, X_OK) like access().
// stat: NULL or the pointer of struct stat.
//
static int check_object_access(const char* path, int mask, struct stat* pstbuf)
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
{
int result;
struct stat st;
struct stat* pst = (pstbuf ? pstbuf : &st);
struct fuse_context* pcxt;
FPRNINFO("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(NULL == (pcxt = fuse_get_context())){
return -EIO;
}
if(0 != (result = get_object_attribute(path, pst))){
// If there is not tha target file(object), reusult is -ENOENT.
return result;
}
if(0 == pcxt->uid){
// root is allowed all accessing.
return 0;
}
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(is_s3fs_uid && s3fs_uid == pcxt->uid){
// "uid" user is allowed all accessing.
return 0;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(F_OK == mask){
// if there is a file, always return allowed.
return 0;
}
// for "uid", "gid" option
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
uid_t obj_uid = (is_s3fs_uid ? s3fs_uid : pst->st_uid);
gid_t obj_gid = (is_s3fs_gid ? s3fs_gid : pst->st_gid);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
// compare file mode and uid/gid + mask.
mode_t mode;
mode_t base_mask = S_IRWXO;
if(is_s3fs_umask){
// If umask is set, all object attributes set ~umask.
mode = ((S_IRWXU | S_IRWXG | S_IRWXO) & ~s3fs_umask);
}else{
mode = pst->st_mode;
}
if(pcxt->uid == obj_uid){
base_mask |= S_IRWXU;
}
if(pcxt->gid == obj_gid){
base_mask |= S_IRWXG;
}
if(1 == is_uid_inculde_group(pcxt->uid, obj_gid)){
base_mask |= S_IRWXG;
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
}
mode &= base_mask;
if(X_OK == (mask & X_OK)){
if(0 == (mode & (S_IXUSR | S_IXGRP | S_IXOTH))){
return -EPERM;
}
}
if(W_OK == (mask & W_OK)){
if(0 == (mode & (S_IWUSR | S_IWGRP | S_IWOTH))){
return -EACCES;
}
}
if(R_OK == (mask & R_OK)){
if(0 == (mode & (S_IRUSR | S_IRGRP | S_IROTH))){
return -EACCES;
}
}
if(0 == mode){
return -EACCES;
}
return 0;
}
static int check_object_owner(const char* path, struct stat* pstbuf)
{
int result;
struct stat st;
struct stat* pst = (pstbuf ? pstbuf : &st);
struct fuse_context* pcxt;
FPRNINFO("[path=%s]", path);
if(NULL == (pcxt = fuse_get_context())){
return -EIO;
}
if(0 != (result = get_object_attribute(path, pst))){
// If there is not tha target file(object), reusult is -ENOENT.
return result;
}
// check owner
if(0 == pcxt->uid){
// root is allowed all accessing.
return 0;
}
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(is_s3fs_uid && s3fs_uid == pcxt->uid){
// "uid" user is allowed all accessing.
return 0;
}
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(pcxt->uid == pst->st_uid){
return 0;
}
return -EPERM;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
//
// Check accessing the parent directories of the object by uid and gid.
//
static int check_parent_object_access(const char* path, int mask)
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
{
string parent;
int result;
FPRNINFO("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 == strcmp(path, "/") || 0 == strcmp(path, ".")){
// path is mount point.
return 0;
}
if(X_OK == (mask & X_OK)){
2015-08-12 15:04:16 +00:00
for(parent = mydirname(path); 0 < parent.size(); parent = mydirname(parent)){
if(parent == "."){
parent = "/";
}
if(0 != (result = check_object_access(parent.c_str(), X_OK, NULL))){
return result;
}
if(parent == "/" || parent == "."){
break;
}
}
}
mask = (mask & ~X_OK);
if(0 != mask){
parent = mydirname(path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(parent == "."){
parent = "/";
}
if(0 != (result = check_object_access(parent.c_str(), mask, NULL))){
return result;
}
}
return 0;
}
2014-07-19 19:02:55 +00:00
//
// This function is global, is called fom curl class(GetObject).
//
char* get_object_sseckey_md5(const char* path)
{
if(!path){
return NULL;
}
headers_t meta;
if(0 != get_object_attribute(path, NULL, &meta)){
DPRNNN("Failed to get object(%s) headers", path);
return NULL;
}
for(headers_t::iterator iter = meta.begin(); iter != meta.end(); ++iter){
string key = (*iter).first;
if(0 == strcasecmp(key.c_str(), "x-amz-server-side-encryption-customer-key-md5")){
return strdup((*iter).second.c_str());
}
}
return NULL;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
static FdEntity* get_local_fent(const char* path, bool is_load)
{
struct stat stobj;
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
FPRNNN("[path=%s]", path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(0 != get_object_attribute(path, &stobj)){
return NULL;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// open
time_t mtime = (!S_ISREG(stobj.st_mode) || S_ISLNK(stobj.st_mode)) ? -1 : stobj.st_mtime;
bool force_tmpfile = S_ISREG(stobj.st_mode) ? false : true;
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(NULL == (ent = FdManager::get()->Open(path, stobj.st_size, mtime, force_tmpfile, true))){
DPRN("Coult not open file. errno(%d)", errno);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return NULL;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// load
if(is_load && !ent->LoadFull()){
DPRN("Coult not load file. errno(%d)", errno);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return NULL;
}
return ent;
}
/**
* create or update s3 meta
* ow_sse_flg is for over writing sse header by use_sse option.
* @return fuse return code
*/
2014-08-26 17:11:10 +00:00
static int put_headers(const char* path, headers_t& meta, bool is_copy)
{
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
int result;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
S3fsCurl s3fscurl(true);
struct stat buf;
FPRNNN("[path=%s]", path);
// files larger than 5GB must be modified via the multipart interface
// *** If there is not target object(a case of move command),
// get_object_attribute() returns error with initilizing buf.
(void)get_object_attribute(path, &buf);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(buf.st_size >= FIVE_GB){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// multipart
2014-08-26 17:11:10 +00:00
if(0 != (result = s3fscurl.MultipartHeadRequest(path, buf.st_size, meta, is_copy))){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}else{
2014-08-26 17:11:10 +00:00
if(0 != (result = s3fscurl.PutHeadRequest(path, meta, is_copy))){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return result;
}
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent = NULL;
if(NULL == (ent = FdManager::get()->ExistOpen(path))){
// no opened fd
if(FdManager::get()->IsCacheDir()){
// create cache file if be needed
ent = FdManager::get()->Open(path, buf.st_size, -1, false, true);
}
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(ent){
time_t mtime = get_mtime(meta);
ent->SetMtime(mtime);
FdManager::get()->Close(ent);
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return 0;
}
static int s3fs_getattr(const char* path, struct stat* stbuf)
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
{
int result;
FPRN("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
// check parent directory attribute.
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
if(0 != (result = check_object_access(path, F_OK, stbuf))){
return result;
}
// If has already opened fd, the st_size shuld be instead.
// (See: Issue 241)
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(stbuf){
FdEntity* ent;
if(NULL != (ent = FdManager::get()->ExistOpen(path))){
struct stat tmpstbuf;
if(ent->GetStats(tmpstbuf)){
stbuf->st_size = tmpstbuf.st_size;
}
FdManager::get()->Close(ent);
}
2014-01-16 16:04:41 +00:00
stbuf->st_blksize = 4096;
stbuf->st_blocks = get_blocks(stbuf->st_size);
}
FPRNINFO("[path=%s] uid=%u, gid=%u, mode=%04o", path, (unsigned int)(stbuf->st_uid), (unsigned int)(stbuf->st_gid), stbuf->st_mode);
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_readlink(const char* path, char* buf, size_t size)
{
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(!path || !buf || 0 >= size){
return 0;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// Open
FdEntity* ent;
if(NULL == (ent = get_local_fent(path))){
DPRN("could not get fent(file=%s)", path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
// Get size
off_t readsize;
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(!ent->GetSize(readsize)){
DPRN("could not get file size(file=%s)", path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return -EIO;
}
if(static_cast<off_t>(size) <= readsize){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
readsize = size - 1;
}
// Read
ssize_t ressize;
if(0 > (ressize = ent->Read(buf, 0, static_cast<size_t>(readsize)))){
DPRN("could not read file(file=%s, errno=%zd)", path, ressize);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return static_cast<int>(ressize);
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
buf[ressize] = '\0';
FdManager::get()->Close(ent);
S3FS_MALLOCTRIM(0);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return 0;
}
static int do_create_bucket(void)
{
FPRNNN("/");
headers_t meta;
S3fsCurl s3fscurl(true);
long res = s3fscurl.PutRequest("/", meta, -1);
if(res < 0){ // fd=-1 means for creating zero byte object.
long responseCode = s3fscurl.GetLastResponseCode();
if((responseCode == 400 || responseCode == 403) && S3fsCurl::IsSignatureV4()){
LOWSYSLOGPRINT(LOG_ERR, "Could not connect, so retry to connect by signature version 2.");
FPRN("Could not connect, so retry to connect by signature version 2.");
S3fsCurl::SetSignatureV4(false);
// retry to check
s3fscurl.DestroyCurlHandle();
res = s3fscurl.PutRequest("/", meta, -1);
}
}
return res;
}
// common function for creation of a plain object
static int create_file_object(const char* path, mode_t mode, uid_t uid, gid_t gid)
{
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
FPRNNN("[path=%s][mode=%04o]", path, mode);
headers_t meta;
meta["Content-Type"] = S3fsCurl::LookupMimeType(string(path));
meta["x-amz-meta-uid"] = str(uid);
meta["x-amz-meta-gid"] = str(gid);
meta["x-amz-meta-mode"] = str(mode);
meta["x-amz-meta-mtime"] = str(time(NULL));
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
S3fsCurl s3fscurl(true);
2014-08-26 17:11:10 +00:00
return s3fscurl.PutRequest(path, meta, -1); // fd=-1 means for creating zero byte object.
}
static int s3fs_mknod(const char *path, mode_t mode, dev_t rdev)
{
int result;
struct fuse_context* pcxt;
FPRN("[path=%s][mode=%04o][dev=%ju]", path, mode, (uintmax_t)rdev);
if(NULL == (pcxt = fuse_get_context())){
return -EIO;
}
if(0 != (result = create_file_object(path, mode, pcxt->uid, pcxt->gid))){
DPRN("could not create object for special file(result=%d)", result);
return result;
}
StatCache::getStatCacheData()->DelStat(path);
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_create(const char* path, mode_t mode, struct fuse_file_info* fi)
{
int result;
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
struct fuse_context* pcxt;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
FPRN("[path=%s][mode=%04o][flags=%d]", path, mode, fi->flags);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(NULL == (pcxt = fuse_get_context())){
return -EIO;
}
// check parent directory attribute.
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
result = check_object_access(path, W_OK, NULL);
if(-ENOENT == result){
if(0 != (result = check_parent_object_access(path, W_OK))){
return result;
}
}else if(0 != result){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
result = create_file_object(path, mode, pcxt->uid, pcxt->gid);
StatCache::getStatCacheData()->DelStat(path);
if(result != 0){
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
if(NULL == (ent = FdManager::get()->Open(path, 0, -1, false, true))){
return -EIO;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
fi->fh = ent->GetFd();
S3FS_MALLOCTRIM(0);
return 0;
}
static int create_directory_object(const char* path, mode_t mode, time_t time, uid_t uid, gid_t gid)
{
FPRNN("[path=%s][mode=%04o][time=%jd][uid=%u][gid=%u]", path, mode, (intmax_t)time, (unsigned int)uid, (unsigned int)gid);
if(!path || '\0' == path[0]){
return -1;
}
string tpath = path;
if('/' != tpath[tpath.length() - 1]){
tpath += "/";
}
headers_t meta;
meta["Content-Type"] = string("application/x-directory");
meta["x-amz-meta-uid"] = str(uid);
meta["x-amz-meta-gid"] = str(gid);
meta["x-amz-meta-mode"] = str(mode);
meta["x-amz-meta-mtime"] = str(time);
S3fsCurl s3fscurl;
2014-08-26 17:11:10 +00:00
return s3fscurl.PutRequest(tpath.c_str(), meta, -1); // fd=-1 means for creating zero byte object.
}
static int s3fs_mkdir(const char* path, mode_t mode)
{
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
int result;
struct fuse_context* pcxt;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
FPRN("[path=%s][mode=%04o]", path, mode);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(NULL == (pcxt = fuse_get_context())){
return -EIO;
}
// check parent directory attribute.
if(0 != (result = check_parent_object_access(path, W_OK | X_OK))){
return result;
}
if(-ENOENT != (result = check_object_access(path, F_OK, NULL))){
if(0 == result){
result = -EEXIST;
}
return result;
}
result = create_directory_object(path, mode, time(NULL), pcxt->uid, pcxt->gid);
StatCache::getStatCacheData()->DelStat(path);
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_unlink(const char* path)
{
int result;
FPRN("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(path, W_OK | X_OK))){
return result;
}
S3fsCurl s3fscurl;
result = s3fscurl.DeleteRequest(path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::DeleteCacheFile(path);
StatCache::getStatCacheData()->DelStat(path);
S3FS_MALLOCTRIM(0);
return result;
}
static int directory_empty(const char* path)
{
int result;
S3ObjList head;
if((result = list_bucket(path, head, "/", true)) != 0){
DPRNNN("list_bucket returns error.");
return result;
}
if(!head.IsEmpty()){
return -ENOTEMPTY;
}
return 0;
}
static int s3fs_rmdir(const char* path)
{
int result;
string strpath;
struct stat stbuf;
FPRN("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(path, W_OK | X_OK))){
return result;
}
// directory must be empty
if(directory_empty(path) != 0){
return -ENOTEMPTY;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
strpath = path;
if('/' != strpath[strpath.length() - 1]){
strpath += "/";
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
S3fsCurl s3fscurl;
result = s3fscurl.DeleteRequest(strpath.c_str());
s3fscurl.DestroyCurlHandle();
StatCache::getStatCacheData()->DelStat(strpath.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// double check for old version(before 1.63)
// The old version makes "dir" object, newer version makes "dir/".
// A case, there is only "dir", the first removing object is "dir/".
// Then "dir/" is not exists, but curl_delete returns 0.
// So need to check "dir" and should be removed it.
if('/' == strpath[strpath.length() - 1]){
strpath = strpath.substr(0, strpath.length() - 1);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
if(0 == get_object_attribute(strpath.c_str(), &stbuf, NULL, false)){
if(S_ISDIR(stbuf.st_mode)){
// Found "dir" object.
result = s3fscurl.DeleteRequest(strpath.c_str());
s3fscurl.DestroyCurlHandle();
StatCache::getStatCacheData()->DelStat(strpath.c_str());
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
// If there is no "dir" and "dir/" object(this case is made by s3cmd/s3sync),
// the cache key is "dir/". So we get error only onece(delete "dir/").
// check for "_$folder$" object.
// This processing is necessary for other S3 clients compatibility.
if(is_special_name_folder_object(strpath.c_str())){
strpath += "_$folder$";
result = s3fscurl.DeleteRequest(strpath.c_str());
}
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_symlink(const char* from, const char* to)
{
int result;
struct fuse_context* pcxt;
FPRN("[from=%s][to=%s]", from, to);
if(NULL == (pcxt = fuse_get_context())){
return -EIO;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(to, W_OK | X_OK))){
return result;
}
if(-ENOENT != (result = check_object_access(to, F_OK, NULL))){
if(0 == result){
result = -EEXIST;
}
return result;
}
headers_t headers;
headers["Content-Type"] = string("application/octet-stream"); // Static
headers["x-amz-meta-mode"] = str(S_IFLNK | S_IRWXU | S_IRWXG | S_IRWXO);
headers["x-amz-meta-mtime"] = str(time(NULL));
headers["x-amz-meta-uid"] = str(pcxt->uid);
headers["x-amz-meta-gid"] = str(pcxt->gid);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// open tmpfile
FdEntity* ent;
if(NULL == (ent = FdManager::get()->Open(to, 0, -1, true, true))){
DPRN("could not open tmpfile(errno=%d)", errno);
return -errno;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// write
ssize_t from_size = strlen(from);
if(from_size != ent->Write(from, 0, from_size)){
DPRN("could not write tmpfile(errno=%d)", errno);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return -errno;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// upload
2014-08-26 17:11:10 +00:00
if(0 != (result = ent->Flush(headers, true))){
DPRN("could not upload tmpfile(result=%d)", result);
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
StatCache::getStatCacheData()->DelStat(to);
S3FS_MALLOCTRIM(0);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return result;
}
static int rename_object(const char* from, const char* to)
{
int result;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
string s3_realpath;
headers_t meta;
FPRNN("[from=%s][to=%s]", from , to);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(to, W_OK | X_OK))){
// not permmit writing "to" object parent dir.
return result;
}
if(0 != (result = check_parent_object_access(from, W_OK | X_OK))){
// not permmit removing "from" object parent dir.
return result;
}
if(0 != (result = get_object_attribute(from, NULL, &meta))){
return result;
}
s3_realpath = get_realpath(from);
meta["x-amz-copy-source"] = urlEncode(service_path + bucket + s3_realpath);
meta["Content-Type"] = S3fsCurl::LookupMimeType(string(to));
meta["x-amz-metadata-directive"] = "REPLACE";
2014-08-26 17:11:10 +00:00
if(0 != (result = put_headers(to, meta, true))){
return result;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
FdManager::get()->Rename(from, to);
result = s3fs_unlink(from);
StatCache::getStatCacheData()->DelStat(to);
return result;
}
static int rename_object_nocopy(const char* from, const char* to)
{
int result;
headers_t meta;
FPRNN("[from=%s][to=%s]", from , to);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(to, W_OK | X_OK))){
// not permmit writing "to" object parent dir.
return result;
}
if(0 != (result = check_parent_object_access(from, W_OK | X_OK))){
// not permmit removing "from" object parent dir.
return result;
}
// Get attributes
if(0 != (result = get_object_attribute(from, NULL, &meta))){
return result;
}
// Set header
meta["Content-Type"] = S3fsCurl::LookupMimeType(string(to));
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// open & load
FdEntity* ent;
if(NULL == (ent = get_local_fent(from, true))){
DPRN("could not open and read file(%s)", from);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// upload
2014-08-26 17:11:10 +00:00
if(0 != (result = ent->RowFlush(to, meta, true))){
DPRN("could not upload file(%s): result=%d", to, result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
// Remove file
result = s3fs_unlink(from);
// Stats
StatCache::getStatCacheData()->DelStat(to);
StatCache::getStatCacheData()->DelStat(from);
return result;
}
static int rename_large_object(const char* from, const char* to)
{
int result;
struct stat buf;
headers_t meta;
FPRNN("[from=%s][to=%s]", from , to);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(to, W_OK | X_OK))){
// not permmit writing "to" object parent dir.
return result;
}
if(0 != (result = check_parent_object_access(from, W_OK | X_OK))){
// not permmit removing "from" object parent dir.
return result;
}
if(0 != (result = get_object_attribute(from, &buf, &meta, false))){
return result;
}
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
S3fsCurl s3fscurl(true);
if(0 != (result = s3fscurl.MultipartRenameRequest(from, to, meta, buf.st_size))){
return result;
}
s3fscurl.DestroyCurlHandle();
StatCache::getStatCacheData()->DelStat(to);
return s3fs_unlink(from);
}
static int clone_directory_object(const char* from, const char* to)
{
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
int result = -1;
struct stat stbuf;
FPRNN("[from=%s][to=%s]", from, to);
// get target's attributes
if(0 != (result = get_object_attribute(from, &stbuf))){
return result;
}
result = create_directory_object(to, stbuf.st_mode, stbuf.st_mtime, stbuf.st_uid, stbuf.st_gid);
StatCache::getStatCacheData()->DelStat(to);
return result;
}
static int rename_directory(const char* from, const char* to)
{
S3ObjList head;
s3obj_list_t headlist;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
string strfrom = from ? from : ""; // from is without "/".
string strto = to ? to : ""; // to is without "/" too.
string basepath = strfrom + "/";
string newpath; // should be from name(not used)
string nowcache; // now cache path(not used)
int DirType;
bool normdir;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
MVNODE* mn_head = NULL;
MVNODE* mn_tail = NULL;
MVNODE* mn_cur;
struct stat stbuf;
int result;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
bool is_dir;
FPRNN("[from=%s][to=%s]", from, to);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
//
// Initiate and Add base directory into MVNODE struct.
//
strto += "/";
if(0 == chk_dir_object_type(from, newpath, strfrom, nowcache, NULL, &DirType) && DIRTYPE_UNKNOWN != DirType){
if(DIRTYPE_NOOBJ != DirType){
normdir = false;
}else{
normdir = true;
strfrom = from; // from directory is not removed, but from directory attr is needed.
}
if(NULL == (add_mvnode(&mn_head, &mn_tail, strfrom.c_str(), strto.c_str(), true, normdir))){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -ENOMEM;
}
}else{
// Something wrong about "from" directory.
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
//
// get a list of all the objects
//
// No delimiter is specified, the result(head) is all object keys.
// (CommonPrefixes is empty, but all object is listed in Key.)
if(0 != (result = list_bucket(basepath.c_str(), head, NULL))){
DPRNNN("list_bucket returns error.");
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
head.GetNameList(headlist); // get name without "/".
S3ObjList::MakeHierarchizedList(headlist, false); // add hierarchized dir.
s3obj_list_t::const_iterator liter;
2015-08-12 15:04:16 +00:00
for(liter = headlist.begin(); headlist.end() != liter; ++liter){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// make "from" and "to" object name.
string from_name = basepath + (*liter);
string to_name = strto + (*liter);
string etag = head.GetETag((*liter).c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Check subdirectory.
StatCache::getStatCacheData()->HasStat(from_name, etag.c_str()); // Check ETag
if(0 != get_object_attribute(from_name.c_str(), &stbuf, NULL)){
DPRNNN("failed to get %s object attribute.", from_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
continue;
}
if(S_ISDIR(stbuf.st_mode)){
is_dir = true;
if(0 != chk_dir_object_type(from_name.c_str(), newpath, from_name, nowcache, NULL, &DirType) || DIRTYPE_UNKNOWN == DirType){
DPRNNN("failed to get %s%s object directory type.", basepath.c_str(), (*liter).c_str());
continue;
}
if(DIRTYPE_NOOBJ != DirType){
normdir = false;
}else{
normdir = true;
from_name = basepath + (*liter); // from directory is not removed, but from directory attr is needed.
}
}else{
is_dir = false;
normdir = false;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// push this one onto the stack
if(NULL == add_mvnode(&mn_head, &mn_tail, from_name.c_str(), to_name.c_str(), is_dir, normdir)){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -ENOMEM;
}
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
//
// rename
//
// rename directory objects.
for(mn_cur = mn_head; mn_cur; mn_cur = mn_cur->next){
if(mn_cur->is_dir && mn_cur->old_path && '\0' != mn_cur->old_path[0]){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(0 != (result = clone_directory_object(mn_cur->old_path, mn_cur->new_path))){
DPRN("clone_directory_object returned an error(%d)", result);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
free_mvnodes(mn_head);
return -EIO;
}
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
// iterate over the list - copy the files with rename_object
// does a safe copy - copies first and then deletes old
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
for(mn_cur = mn_head; mn_cur; mn_cur = mn_cur->next){
if(!mn_cur->is_dir){
if(!nocopyapi && !norenameapi){
result = rename_object(mn_cur->old_path, mn_cur->new_path);
}else{
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
result = rename_object_nocopy(mn_cur->old_path, mn_cur->new_path);
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(0 != result){
DPRN("rename_object returned an error(%d)", result);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
free_mvnodes(mn_head);
return -EIO;
}
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
// Iterate over old the directories, bottoms up and remove
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
for(mn_cur = mn_tail; mn_cur; mn_cur = mn_cur->prev){
if(mn_cur->is_dir && mn_cur->old_path && '\0' != mn_cur->old_path[0]){
if(!(mn_cur->is_normdir)){
if(0 != (result = s3fs_rmdir(mn_cur->old_path))){
DPRN("s3fs_rmdir returned an error(%d)", result);
free_mvnodes(mn_head);
return -EIO;
}
}else{
// cache clear.
StatCache::getStatCacheData()->DelStat(mn_cur->old_path);
}
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
free_mvnodes(mn_head);
return 0;
}
static int s3fs_rename(const char* from, const char* to)
{
struct stat buf;
int result;
FPRN("[from=%s][to=%s]", from, to);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_parent_object_access(to, W_OK | X_OK))){
// not permmit writing "to" object parent dir.
return result;
}
if(0 != (result = check_parent_object_access(from, W_OK | X_OK))){
// not permmit removing "from" object parent dir.
return result;
}
if(0 != (result = get_object_attribute(from, &buf, NULL))){
return result;
}
// files larger than 5GB must be modified via the multipart interface
if(S_ISDIR(buf.st_mode)){
result = rename_directory(from, to);
}else if(!nomultipart && buf.st_size >= singlepart_copy_limit){
result = rename_large_object(from, to);
}else{
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(!nocopyapi && !norenameapi){
result = rename_object(from, to);
}else{
result = rename_object_nocopy(from, to);
}
}
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_link(const char* from, const char* to)
{
FPRN("[from=%s][to=%s]", from, to);
return -EPERM;
}
static int s3fs_chmod(const char* path, mode_t mode)
{
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
FPRN("[path=%s][mode=%04o]", path, mode);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 == strcmp(path, "/")){
DPRNNN("Could not change mode for maount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
if(0 != (result = check_object_owner(path, &stbuf))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
if(0 != result){
return result;
}
if(S_ISDIR(stbuf.st_mode) && IS_REPLACEDIR(nDirType)){
// Should rebuild directory object(except new type)
// Need to remove old dir("dir" etc) and make new dir("dir/")
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
StatCache::getStatCacheData()->DelStat(nowcache);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), mode, stbuf.st_mtime, stbuf.st_uid, stbuf.st_gid))){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
}else{
// normal object or directory object of newer version
meta["x-amz-meta-mode"] = str(mode);
meta["x-amz-copy-source"] = urlEncode(service_path + bucket + get_realpath(strpath.c_str()));
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
meta["x-amz-metadata-directive"] = "REPLACE";
2014-08-26 17:11:10 +00:00
if(put_headers(strpath.c_str(), meta, true) != 0){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -EIO;
}
StatCache::getStatCacheData()->DelStat(nowcache);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
S3FS_MALLOCTRIM(0);
return 0;
}
static int s3fs_chmod_nocopy(const char* path, mode_t mode)
{
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
FPRNN("[path=%s][mode=%04o]", path, mode);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 == strcmp(path, "/")){
DPRNNN("Could not change mode for maount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
if(0 != (result = check_object_owner(path, &stbuf))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Get attributes
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(0 != result){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(S_ISDIR(stbuf.st_mode)){
// Should rebuild all directory object
// Need to remove old dir("dir" etc) and make new dir("dir/")
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
}
StatCache::getStatCacheData()->DelStat(nowcache);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), mode, stbuf.st_mtime, stbuf.st_uid, stbuf.st_gid))){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}else{
// normal object or directory object of newer version
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Change file mode
meta["x-amz-meta-mode"] = str(mode);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// open & load
FdEntity* ent;
if(NULL == (ent = get_local_fent(strpath.c_str(), true))){
DPRN("could not open and read file(%s)", strpath.c_str());
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// upload
2014-08-26 17:11:10 +00:00
if(0 != (result = ent->Flush(meta, true))){
DPRN("could not upload file(%s): result=%d", strpath.c_str(), result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
StatCache::getStatCacheData()->DelStat(nowcache);
}
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_chown(const char* path, uid_t uid, gid_t gid)
{
int result;
string strpath;
string newpath;
string nowcache;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
headers_t meta;
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
FPRN("[path=%s][uid=%u][gid=%u]", path, (unsigned int)uid, (unsigned int)gid);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 == strcmp(path, "/")){
DPRNNN("Could not change owner for maount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
if(0 != (result = check_object_owner(path, &stbuf))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
if((uid_t)(-1) == uid){
uid = stbuf.st_uid;
}
if((gid_t)(-1) == gid){
gid = stbuf.st_gid;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
if(0 != result){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
struct passwd* pwdata= getpwuid(uid);
struct group* grdata = getgrgid(gid);
if(pwdata){
uid = pwdata->pw_uid;
}
if(grdata){
gid = grdata->gr_gid;
}
if(S_ISDIR(stbuf.st_mode) && IS_REPLACEDIR(nDirType)){
// Should rebuild directory object(except new type)
// Need to remove old dir("dir" etc) and make new dir("dir/")
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
StatCache::getStatCacheData()->DelStat(nowcache);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), stbuf.st_mode, stbuf.st_mtime, uid, gid))){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
}else{
meta["x-amz-meta-uid"] = str(uid);
meta["x-amz-meta-gid"] = str(gid);
meta["x-amz-copy-source"] = urlEncode(service_path + bucket + get_realpath(strpath.c_str()));
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
meta["x-amz-metadata-directive"] = "REPLACE";
2014-08-26 17:11:10 +00:00
if(put_headers(strpath.c_str(), meta, true) != 0){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -EIO;
}
StatCache::getStatCacheData()->DelStat(nowcache);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
S3FS_MALLOCTRIM(0);
return 0;
}
static int s3fs_chown_nocopy(const char* path, uid_t uid, gid_t gid)
{
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
FPRNN("[path=%s][uid=%u][gid=%u]", path, (unsigned int)uid, (unsigned int)gid);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 == strcmp(path, "/")){
DPRNNN("Could not change owner for maount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
if(0 != (result = check_object_owner(path, &stbuf))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Get attributes
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(0 != result){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
struct passwd* pwdata= getpwuid(uid);
struct group* grdata = getgrgid(gid);
if(pwdata){
uid = pwdata->pw_uid;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(grdata){
gid = grdata->gr_gid;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(S_ISDIR(stbuf.st_mode)){
// Should rebuild all directory object
// Need to remove old dir("dir" etc) and make new dir("dir/")
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
}
StatCache::getStatCacheData()->DelStat(nowcache);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), stbuf.st_mode, stbuf.st_mtime, uid, gid))){
return result;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}else{
// normal object or directory object of newer version
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Change owner
meta["x-amz-meta-uid"] = str(uid);
meta["x-amz-meta-gid"] = str(gid);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// open & load
FdEntity* ent;
if(NULL == (ent = get_local_fent(strpath.c_str(), true))){
DPRN("could not open and read file(%s)", strpath.c_str());
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// upload
2014-08-26 17:11:10 +00:00
if(0 != (result = ent->Flush(meta, true))){
DPRN("could not upload file(%s): result=%d", strpath.c_str(), result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
StatCache::getStatCacheData()->DelStat(nowcache);
}
S3FS_MALLOCTRIM(0);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
static int s3fs_utimens(const char* path, const struct timespec ts[2])
{
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
FPRN("[path=%s][mtime=%jd]", path, (intmax_t)(ts[1].tv_sec));
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 == strcmp(path, "/")){
DPRNNN("Could not change mtime for maount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
if(0 != (result = check_object_access(path, W_OK, &stbuf))){
if(0 != check_object_owner(path, &stbuf)){
return result;
}
}
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
}
if(0 != result){
return result;
}
if(S_ISDIR(stbuf.st_mode) && IS_REPLACEDIR(nDirType)){
// Should rebuild directory object(except new type)
// Need to remove old dir("dir" etc) and make new dir("dir/")
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
return result;
}
}
StatCache::getStatCacheData()->DelStat(nowcache);
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), stbuf.st_mode, ts[1].tv_sec, stbuf.st_uid, stbuf.st_gid))){
return result;
}
}else{
meta["x-amz-meta-mtime"] = str(ts[1].tv_sec);
meta["x-amz-copy-source"] = urlEncode(service_path + bucket + get_realpath(strpath.c_str()));
meta["x-amz-metadata-directive"] = "REPLACE";
2014-08-26 17:11:10 +00:00
if(put_headers(strpath.c_str(), meta, true) != 0){
return -EIO;
}
StatCache::getStatCacheData()->DelStat(nowcache);
}
S3FS_MALLOCTRIM(0);
return 0;
}
static int s3fs_utimens_nocopy(const char* path, const struct timespec ts[2])
{
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
FPRNN("[path=%s][mtime=%s]", path, str(ts[1].tv_sec).c_str());
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 == strcmp(path, "/")){
2015-07-10 18:50:40 +00:00
DPRNNN("Could not change mtime for mount point.");
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
if(0 != (result = check_object_access(path, W_OK, &stbuf))){
if(0 != check_object_owner(path, &stbuf)){
return result;
}
}
// Get attributes
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
}
if(0 != result){
return result;
}
if(S_ISDIR(stbuf.st_mode)){
// Should rebuild all directory object
// Need to remove old dir("dir" etc) and make new dir("dir/")
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
return result;
}
}
StatCache::getStatCacheData()->DelStat(nowcache);
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), stbuf.st_mode, ts[1].tv_sec, stbuf.st_uid, stbuf.st_gid))){
return result;
}
}else{
// normal object or directory object of newer version
// Change date
meta["x-amz-meta-mtime"] = str(ts[1].tv_sec);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// open & load
FdEntity* ent;
if(NULL == (ent = get_local_fent(strpath.c_str(), true))){
DPRN("could not open and read file(%s)", strpath.c_str());
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// set mtime
if(0 != (result = ent->SetMtime(ts[1].tv_sec))){
DPRN("could not set mtime to file(%s): result=%d", strpath.c_str(), result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// upload
2014-08-26 17:11:10 +00:00
if(0 != (result = ent->Flush(meta, true))){
DPRN("could not upload file(%s): result=%d", strpath.c_str(), result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
StatCache::getStatCacheData()->DelStat(nowcache);
}
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_truncate(const char* path, off_t size)
{
int result;
headers_t meta;
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent = NULL;
FPRN("[path=%s][size=%jd]", path, (intmax_t)size);
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_object_access(path, W_OK, NULL))){
return result;
}
// Get file information
if(0 == (result = get_object_attribute(path, NULL, &meta))){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// Exists -> Get file(with size)
if(NULL == (ent = FdManager::get()->Open(path, size, -1, false, true))){
DPRN("could not open file(%s): errno=%d", path, errno);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
if(0 != (result = ent->Load(0, size))){
DPRN("could not download file(%s): result=%d", path, result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}else{
// Not found -> Make tmpfile(with size)
if(NULL == (ent = FdManager::get()->Open(path, size, -1, true, true))){
DPRN("could not open file(%s): errno=%d", path, errno);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// upload
2014-08-26 17:11:10 +00:00
if(0 != (result = ent->Flush(meta, true))){
DPRN("could not upload file(%s): result=%d", path, result);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
StatCache::getStatCacheData()->DelStat(path);
S3FS_MALLOCTRIM(0);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
static int s3fs_open(const char* path, struct fuse_file_info* fi)
{
int result;
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
struct stat st;
FPRN("[path=%s][flags=%d]", path, fi->flags);
// clear stat for reading fresh stat.
// (if object stat is changed, we refresh it. then s3fs gets always
// stat when s3fs open the object).
StatCache::getStatCacheData()->DelStat(path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
int mask = (O_RDONLY != (fi->flags & O_ACCMODE) ? W_OK : R_OK);
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
result = check_object_access(path, mask, &st);
if(-ENOENT == result){
if(0 != (result = check_parent_object_access(path, W_OK))){
return result;
}
}else if(0 != result){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
if((unsigned int)fi->flags & O_TRUNC){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
st.st_size = 0;
}
if(!S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)){
st.st_mtime = -1;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
if(NULL == (ent = FdManager::get()->Open(path, st.st_size, st.st_mtime, false, true))){
return -EIO;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
fi->fh = ent->GetFd();
S3FS_MALLOCTRIM(0);
return 0;
}
static int s3fs_read(const char* path, char* buf, size_t size, off_t offset, struct fuse_file_info* fi)
{
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
ssize_t res;
FPRNINFO("[path=%s][size=%zu][offset=%jd][fd=%llu]", path, size, (intmax_t)offset, (unsigned long long)(fi->fh));
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
if(NULL == (ent = FdManager::get()->ExistOpen(path, static_cast<int>(fi->fh)))){
DPRN("could not find opened fd(%s)", path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
if(ent->GetFd() != static_cast<int>(fi->fh)){
DPRNNN("Warning - different fd(%d - %llu)", ent->GetFd(), (unsigned long long)(fi->fh));
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}
// check real file size
off_t realsize = 0;
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(!ent->GetSize(realsize) || 0 >= realsize){
DPRNINFO("file size is 0, so break to read.");
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return 0;
}
if(0 > (res = ent->Read(buf, offset, size, false))){
DPRN("failed to read file(%s). result=%zd", path, res);
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
return static_cast<int>(res);
}
static int s3fs_write(const char* path, const char* buf, size_t size, off_t offset, struct fuse_file_info* fi)
{
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
ssize_t res;
FPRNINFO("[path=%s][size=%zu][offset=%jd][fd=%llu]", path, size, (intmax_t)offset, (unsigned long long)(fi->fh));
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
if(NULL == (ent = FdManager::get()->ExistOpen(path, static_cast<int>(fi->fh)))){
DPRN("could not find opened fd(%s)", path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
if(ent->GetFd() != static_cast<int>(fi->fh)){
DPRNNN("Warning - different fd(%d - %llu)", ent->GetFd(), (unsigned long long)(fi->fh));
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}
if(0 > (res = ent->Write(buf, offset, size))){
DPRN("failed to write file(%s). result=%zd", path, res);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}
FdManager::get()->Close(ent);
return static_cast<int>(res);
}
static int s3fs_statfs(const char* path, struct statvfs* stbuf)
{
// 256T
stbuf->f_bsize = 0X1000000;
stbuf->f_blocks = 0X1000000;
stbuf->f_bfree = 0x1000000;
stbuf->f_bavail = 0x1000000;
stbuf->f_namemax = NAME_MAX;
return 0;
}
static int s3fs_flush(const char* path, struct fuse_file_info* fi)
{
int result;
FPRN("[path=%s][fd=%llu]", path, (unsigned long long)(fi->fh));
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
int mask = (O_RDONLY != (fi->flags & O_ACCMODE) ? W_OK : R_OK);
if(0 != (result = check_parent_object_access(path, X_OK))){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
result = check_object_access(path, mask, NULL);
if(-ENOENT == result){
if(0 != (result = check_parent_object_access(path, W_OK))){
return result;
}
}else if(0 != result){
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
if(NULL != (ent = FdManager::get()->ExistOpen(path, static_cast<int>(fi->fh)))){
headers_t meta;
if(0 != (result = get_object_attribute(path, NULL, &meta))){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
// If both mtime are not same, force to change mtime based on fd.
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
time_t ent_mtime;
if(ent->GetMtime(ent_mtime)){
if(str(ent_mtime) != meta["x-amz-meta-mtime"]){
meta["x-amz-meta-mtime"] = str(ent_mtime);
}
}
2014-08-26 17:11:10 +00:00
result = ent->Flush(meta, false);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
}
S3FS_MALLOCTRIM(0);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return result;
}
2015-03-10 16:18:03 +00:00
// [NOTICE]
// Assumption is a valid fd.
//
static int s3fs_fsync(const char* path, int datasync, struct fuse_file_info* fi)
{
int result = 0;
FPRN("[path=%s][fd=%llu]", path, (unsigned long long)(fi->fh));
FdEntity* ent;
if(NULL != (ent = FdManager::get()->ExistOpen(path, static_cast<int>(fi->fh)))){
headers_t meta;
if(0 != (result = get_object_attribute(path, NULL, &meta))){
FdManager::get()->Close(ent);
return result;
}
// If datasync is not zero, only flush data without meta updating.
time_t ent_mtime;
if(ent->GetMtime(ent_mtime)){
if(0 == datasync && str(ent_mtime) != meta["x-amz-meta-mtime"]){
meta["x-amz-meta-mtime"] = str(ent_mtime);
}
}
result = ent->Flush(meta, false);
FdManager::get()->Close(ent);
}
S3FS_MALLOCTRIM(0);
return result;
}
static int s3fs_release(const char* path, struct fuse_file_info* fi)
{
FPRN("[path=%s][fd=%llu]", path, (unsigned long long)(fi->fh));
// [NOTICE]
// At first, we remove stats cache.
2014-10-01 10:42:39 +00:00
// Because fuse does not wait for response from "release" function. :-(
// And fuse runs next command before this function returns.
// Thus we call deleting stats function ASSAP.
//
if((fi->flags & O_RDWR) || (fi->flags & O_WRONLY)){
StatCache::getStatCacheData()->DelStat(path);
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdEntity* ent;
if(NULL == (ent = FdManager::get()->GetFdEntity(path, static_cast<int>(fi->fh)))){
DPRN("could not find fd(file=%s)", path);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return -EIO;
}
if(ent->GetFd() != static_cast<int>(fi->fh)){
DPRNNN("Warning - different fd(%d - %llu)", ent->GetFd(), (unsigned long long)(fi->fh));
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::get()->Close(ent);
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// check - for debug
if(debug){
if(NULL != (ent = FdManager::get()->GetFdEntity(path, static_cast<int>(fi->fh)))){
DPRNNN("Warning - file(%s),fd(%d) is still opened.", path, ent->GetFd());
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}
}
S3FS_MALLOCTRIM(0);
return 0;
}
static int s3fs_opendir(const char* path, struct fuse_file_info* fi)
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
{
int result;
int mask = (O_RDONLY != (fi->flags & O_ACCMODE) ? W_OK : R_OK) | X_OK;
FPRN("[path=%s][flags=%d]", path, fi->flags);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 == (result = check_object_access(path, mask, NULL))){
result = check_parent_object_access(path, mask);
}
S3FS_MALLOCTRIM(0);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return result;
}
static bool multi_head_callback(S3fsCurl* s3fscurl)
{
if(!s3fscurl){
return false;
}
string saved_path = s3fscurl->GetSpacialSavedPath();
if(!StatCache::getStatCacheData()->AddStat(saved_path, *(s3fscurl->GetResponseHeaders()))){
DPRN("failed adding stat cache [path=%s]", saved_path.c_str());
return false;
}
return true;
}
static S3fsCurl* multi_head_retry_callback(S3fsCurl* s3fscurl)
{
if(!s3fscurl){
return NULL;
}
2014-07-19 19:02:55 +00:00
int ssec_key_pos = s3fscurl->GetLastPreHeadSeecKeyPos();
int next_retry_count = s3fscurl->GetMultipartRetryCount() + 1;
if(s3fscurl->IsOverMultipartRetryCount()){
2014-07-19 19:02:55 +00:00
if(S3fsCurl::IsSseCustomMode()){
// If sse-c mode, start check not sse-c(ssec_key_pos = -1).
// do increment ssec_key_pos for checking all sse-c key.
next_retry_count = 0;
ssec_key_pos++;
if(S3fsCurl::GetSseKeyCount() <= ssec_key_pos){
DPRN("Over retry count(%d) limit(%s).", s3fscurl->GetMultipartRetryCount(), s3fscurl->GetSpacialSavedPath().c_str());
return NULL;
}
}else{
DPRN("Over retry count(%d) limit(%s).", s3fscurl->GetMultipartRetryCount(), s3fscurl->GetSpacialSavedPath().c_str());
return NULL;
}
}
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
S3fsCurl* newcurl = new S3fsCurl(s3fscurl->IsUseAhbe());
string path = s3fscurl->GetPath();
string base_path = s3fscurl->GetBasePath();
string saved_path = s3fscurl->GetSpacialSavedPath();
2014-07-19 19:02:55 +00:00
if(!newcurl->PreHeadRequest(path, base_path, saved_path, ssec_key_pos)){
DPRN("Could not duplicate curl object(%s).", saved_path.c_str());
delete newcurl;
return NULL;
}
2014-07-19 19:02:55 +00:00
newcurl->SetMultipartRetryCount(next_retry_count);
return newcurl;
}
static int readdir_multi_head(const char* path, S3ObjList& head, void* buf, fuse_fill_dir_t filler)
{
S3fsMultiCurl curlmulti;
s3obj_list_t headlist;
s3obj_list_t fillerlist;
int result = 0;
FPRNN("[path=%s][list=%zu]", path, headlist.size());
// Make base path list.
head.GetNameList(headlist, true, false); // get name with "/".
// Initialize S3fsMultiCurl
curlmulti.SetSuccessCallback(multi_head_callback);
curlmulti.SetRetryCallback(multi_head_retry_callback);
// Loop
2015-08-12 15:04:16 +00:00
while(!headlist.empty()){
s3obj_list_t::iterator iter;
long cnt;
fillerlist.clear();
// Make single head request(with max).
for(iter = headlist.begin(), cnt = 0; headlist.end() != iter && cnt < S3fsMultiCurl::GetMaxMultiRequest(); iter = headlist.erase(iter)){
string disppath = path + (*iter);
string etag = head.GetETag((*iter).c_str());
string fillpath = disppath;
if('/' == disppath[disppath.length() - 1]){
fillpath = fillpath.substr(0, fillpath.length() -1);
}
fillerlist.push_back(fillpath);
if(StatCache::getStatCacheData()->HasStat(disppath, etag.c_str())){
continue;
}
2014-07-19 19:02:55 +00:00
// First check for directory, start checking "not sse-c".
// If checking failed, retry to check with "sse-c" by retry callback func when sse-c mode.
S3fsCurl* s3fscurl = new S3fsCurl();
if(!s3fscurl->PreHeadRequest(disppath, (*iter), disppath)){ // target path = cache key path.(ex "dir/")
DPRNNN("Could not make curl object for head request(%s).", disppath.c_str());
delete s3fscurl;
continue;
}
if(!curlmulti.SetS3fsCurlObject(s3fscurl)){
DPRNNN("Could not make curl object into multi curl(%s).", disppath.c_str());
delete s3fscurl;
continue;
}
cnt++; // max request count within S3fsMultiCurl::GetMaxMultiRequest()
}
// Multi request
if(0 != (result = curlmulti.Request())){
DPRN("error occuered in multi request(errno=%d).", result);
break;
}
// populate fuse buffer
// here is best posision, because a case is cache size < files in directory
//
2015-08-12 15:04:16 +00:00
for(iter = fillerlist.begin(); fillerlist.end() != iter; ++iter){
struct stat st;
string bpath = mybasename((*iter));
if(StatCache::getStatCacheData()->GetStat((*iter), &st)){
filler(buf, bpath.c_str(), &st, 0);
}else{
FPRNNN("Could not find %s file in stat cache.", (*iter).c_str());
filler(buf, bpath.c_str(), 0, 0);
}
}
// reinit for loop.
curlmulti.Clear();
}
return result;
}
static int s3fs_readdir(const char* path, void* buf, fuse_fill_dir_t filler, off_t offset, struct fuse_file_info* fi)
{
S3ObjList head;
int result;
FPRN("[path=%s]", path);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
if(0 != (result = check_object_access(path, X_OK, NULL))){
return result;
}
// get a list of all the objects
if((result = list_bucket(path, head, "/")) != 0){
DPRN("list_bucket returns error(%d).", result);
return result;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
// force to add "." and ".." name.
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
filler(buf, ".", 0, 0);
filler(buf, "..", 0, 0);
if(head.IsEmpty()){
return 0;
}
// Send multi head request for stats caching.
string strpath = path;
if(strcmp(path, "/") != 0){
strpath += "/";
}
if(0 != (result = readdir_multi_head(strpath.c_str(), head, buf, filler))){
DPRN("readdir_multi_head returns error(%d).", result);
}
S3FS_MALLOCTRIM(0);
return result;
}
static int list_bucket(const char* path, S3ObjList& head, const char* delimiter, bool check_content_only)
{
int result;
string s3_realpath;
string query_delimiter;;
string query_prefix;;
string query_maxkey;;
string next_marker = "";
bool truncated = true;
S3fsCurl s3fscurl;
xmlDocPtr doc;
BodyData* body;
FPRNN("[path=%s]", path);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(delimiter && 0 < strlen(delimiter)){
query_delimiter += "delimiter=";
query_delimiter += delimiter;
query_delimiter += "&";
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
query_prefix += "&prefix=";
s3_realpath = get_realpath(path);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(0 == s3_realpath.length() || '/' != s3_realpath[s3_realpath.length() - 1]){
// last word must be "/"
query_prefix += urlEncode(s3_realpath.substr(1) + "/");
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}else{
query_prefix += urlEncode(s3_realpath.substr(1));
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
if (check_content_only){
query_maxkey += "max-keys=1";
}else{
query_maxkey += "max-keys=1000";
}
while(truncated){
string each_query = query_delimiter;
if(next_marker != ""){
each_query += "marker=" + urlEncode(next_marker) + "&";
next_marker = "";
}
each_query += query_maxkey;
each_query += query_prefix;
// request
if(0 != (result = s3fscurl.ListBucketRequest(path, each_query.c_str()))){
DPRN("ListBucketRequest returns with error.");
return result;
}
body = s3fscurl.GetBodyData();
// xmlDocPtr
if(NULL == (doc = xmlReadMemory(body->str(), static_cast<int>(body->size()), "", NULL, 0))){
DPRN("xmlReadMemory returns with error.");
return -1;
}
if(0 != append_objects_from_xml(path, doc, head)){
DPRN("append_objects_from_xml returns with error.");
xmlFreeDoc(doc);
return -1;
}
if(true == (truncated = is_truncated(doc))){
xmlChar* tmpch = get_next_marker(doc);
if(tmpch){
next_marker = (char*)tmpch;
xmlFree(tmpch);
}else{
2014-06-28 17:36:35 +00:00
// If did not specify "delimiter", s3 did not return "NextMarker".
// On this case, can use lastest name for next marker.
//
string lastname;
if(!head.GetLastName(lastname)){
DPRN("Could not find next marker, thus break loop.");
truncated = false;
}else{
next_marker = s3_realpath.substr(1);
if(0 == s3_realpath.length() || '/' != s3_realpath[s3_realpath.length() - 1]){
next_marker += "/";
}
next_marker += lastname;
}
}
}
S3FS_XMLFREEDOC(doc);
// reset(initialize) curl object
s3fscurl.DestroyCurlHandle();
if (check_content_only)
break;
}
S3FS_MALLOCTRIM(0);
return 0;
}
const char* c_strErrorObjectName = "FILE or SUBDIR in DIR";
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
static int append_objects_from_xml_ex(const char* path, xmlDocPtr doc, xmlXPathContextPtr ctx,
const char* ex_contents, const char* ex_key, const char* ex_etag, int isCPrefix, S3ObjList& head)
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
{
xmlXPathObjectPtr contents_xp;
xmlNodeSetPtr content_nodes;
if(NULL == (contents_xp = xmlXPathEvalExpression((xmlChar*)ex_contents, ctx))){
DPRNNN("xmlXPathEvalExpression returns null.");
return -1;
}
if(xmlXPathNodeSetIsEmpty(contents_xp->nodesetval)){
DPRNNN("contents_xp->nodesetval is empty.");
S3FS_XMLXPATHFREEOBJECT(contents_xp);
return 0;
}
content_nodes = contents_xp->nodesetval;
bool is_dir;
string stretag;
int i;
for(i = 0; i < content_nodes->nodeNr; i++){
ctx->node = content_nodes->nodeTab[i];
// object name
xmlXPathObjectPtr key;
if(NULL == (key = xmlXPathEvalExpression((xmlChar*)ex_key, ctx))){
DPRNNN("key is null. but continue.");
continue;
}
if(xmlXPathNodeSetIsEmpty(key->nodesetval)){
DPRNNN("node is empty. but continue.");
xmlXPathFreeObject(key);
continue;
}
xmlNodeSetPtr key_nodes = key->nodesetval;
char* name = get_object_name(doc, key_nodes->nodeTab[0]->xmlChildrenNode, path);
if(!name){
DPRNNN("name is something wrong. but continue.");
}else if((const char*)name != c_strErrorObjectName){
is_dir = isCPrefix ? true : false;
stretag = "";
if(!isCPrefix && ex_etag){
// Get ETag
xmlXPathObjectPtr ETag;
if(NULL != (ETag = xmlXPathEvalExpression((xmlChar*)ex_etag, ctx))){
if(xmlXPathNodeSetIsEmpty(ETag->nodesetval)){
DPRNNN("ETag->nodesetval is empty.");
}else{
xmlNodeSetPtr etag_nodes = ETag->nodesetval;
xmlChar* petag = xmlNodeListGetString(doc, etag_nodes->nodeTab[0]->xmlChildrenNode, 1);
if(petag){
stretag = (char*)petag;
xmlFree(petag);
}
}
xmlXPathFreeObject(ETag);
}
}
if(!head.insert(name, (0 < stretag.length() ? stretag.c_str() : NULL), is_dir)){
DPRN("insert_object returns with error.");
xmlXPathFreeObject(key);
xmlXPathFreeObject(contents_xp);
free(name);
S3FS_MALLOCTRIM(0);
return -1;
}
free(name);
}else{
DPRNINFO("name is file or subdir in dir. but continue.");
}
xmlXPathFreeObject(key);
}
S3FS_XMLXPATHFREEOBJECT(contents_xp);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return 0;
}
static bool GetXmlNsUrl(xmlDocPtr doc, string& nsurl)
{
static time_t tmLast = 0; // cache for 60 sec.
static string strNs("");
bool result = false;
if(!doc){
return result;
}
if((tmLast + 60) < time(NULL)){
// refresh
tmLast = time(NULL);
strNs = "";
xmlNodePtr pRootNode = xmlDocGetRootElement(doc);
if(pRootNode){
xmlNsPtr* nslist = xmlGetNsList(doc, pRootNode);
if(nslist){
if(nslist[0] && nslist[0]->href){
strNs = (const char*)(nslist[0]->href);
}
S3FS_XMLFREE(nslist);
}
}
}
if(0 < strNs.size()){
nsurl = strNs;
result = true;
}
return result;
}
static int append_objects_from_xml(const char* path, xmlDocPtr doc, S3ObjList& head)
{
string xmlnsurl;
string ex_contents = "//";
string ex_key = "";
string ex_cprefix = "//";
string ex_prefix = "";
string ex_etag = "";
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(!doc){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
// If there is not <Prefix>, use path instead of it.
xmlChar* pprefix = get_prefix(doc);
string prefix = (pprefix ? (char*)pprefix : path ? path : "");
if(pprefix){
xmlFree(pprefix);
}
xmlXPathContextPtr ctx = xmlXPathNewContext(doc);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(!noxmlns && GetXmlNsUrl(doc, xmlnsurl)){
xmlXPathRegisterNs(ctx, (xmlChar*)"s3", (xmlChar*)xmlnsurl.c_str());
ex_contents+= "s3:";
ex_key += "s3:";
ex_cprefix += "s3:";
ex_prefix += "s3:";
ex_etag += "s3:";
}
ex_contents+= "Contents";
ex_key += "Key";
ex_cprefix += "CommonPrefixes";
ex_prefix += "Prefix";
ex_etag += "ETag";
if(-1 == append_objects_from_xml_ex(prefix.c_str(), doc, ctx, ex_contents.c_str(), ex_key.c_str(), ex_etag.c_str(), 0, head) ||
-1 == append_objects_from_xml_ex(prefix.c_str(), doc, ctx, ex_cprefix.c_str(), ex_prefix.c_str(), NULL, 1, head) )
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
{
DPRN("append_objects_from_xml_ex returns with error.");
S3FS_XMLXPATHFREECONTEXT(ctx);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
S3FS_XMLXPATHFREECONTEXT(ctx);
return 0;
}
static xmlChar* get_base_exp(xmlDocPtr doc, const char* exp)
{
xmlXPathObjectPtr marker_xp;
string xmlnsurl;
string exp_string = "//";
if(!doc){
return NULL;
}
xmlXPathContextPtr ctx = xmlXPathNewContext(doc);
if(!noxmlns && GetXmlNsUrl(doc, xmlnsurl)){
xmlXPathRegisterNs(ctx, (xmlChar*)"s3", (xmlChar*)xmlnsurl.c_str());
exp_string += "s3:";
}
exp_string += exp;
if(NULL == (marker_xp = xmlXPathEvalExpression((xmlChar *)exp_string.c_str(), ctx))){
xmlXPathFreeContext(ctx);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return NULL;
}
if(xmlXPathNodeSetIsEmpty(marker_xp->nodesetval)){
DPRNNN("marker_xp->nodesetval is empty.");
xmlXPathFreeObject(marker_xp);
xmlXPathFreeContext(ctx);
return NULL;
}
xmlNodeSetPtr nodes = marker_xp->nodesetval;
xmlChar* result = xmlNodeListGetString(doc, nodes->nodeTab[0]->xmlChildrenNode, 1);
xmlXPathFreeObject(marker_xp);
xmlXPathFreeContext(ctx);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return result;
}
static xmlChar* get_prefix(xmlDocPtr doc)
{
return get_base_exp(doc, "Prefix");
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}
static xmlChar* get_next_marker(xmlDocPtr doc)
{
return get_base_exp(doc, "NextMarker");
}
static bool is_truncated(xmlDocPtr doc)
{
bool result = false;
xmlChar* strTruncate = get_base_exp(doc, "IsTruncated");
if(!strTruncate){
return result;
}
if(0 == strcasecmp((const char*)strTruncate, "true")){
result = true;
}
xmlFree(strTruncate);
return result;
}
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
// return: the pointer to object name on allocated memory.
// the pointer to "c_strErrorObjectName".(not allocated)
2014-10-01 10:42:39 +00:00
// NULL(a case of something error occurred)
static char* get_object_name(xmlDocPtr doc, xmlNodePtr node, const char* path)
{
// Get full path
xmlChar* fullpath = xmlNodeListGetString(doc, node, 1);
if(!fullpath){
DPRN("could not get object full path name..");
return NULL;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// basepath(path) is as same as fullpath.
if(0 == strcmp((char*)fullpath, path)){
xmlFree(fullpath);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return (char*)c_strErrorObjectName;
}
// Make dir path and filename
string strdirpath = mydirname(string((char*)fullpath));
string strmybpath = mybasename(string((char*)fullpath));
const char* dirpath = strdirpath.c_str();
const char* mybname = strmybpath.c_str();
const char* basepath= (!path || '\0' == path[0] || '/' != path[0] ? path : &path[1]);
xmlFree(fullpath);
if(!mybname || '\0' == mybname[0]){
return NULL;
}
// check subdir & file in subdir
if(dirpath && 0 < strlen(dirpath)){
// case of "/"
if(0 == strcmp(mybname, "/") && 0 == strcmp(dirpath, "/")){
return (char*)c_strErrorObjectName;
}
// case of "."
if(0 == strcmp(mybname, ".") && 0 == strcmp(dirpath, ".")){
return (char*)c_strErrorObjectName;
}
// case of ".."
if(0 == strcmp(mybname, "..") && 0 == strcmp(dirpath, ".")){
return (char*)c_strErrorObjectName;
}
// case of "name"
if(0 == strcmp(dirpath, ".")){
// OK
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return strdup(mybname);
}else{
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(basepath && 0 == strcmp(dirpath, basepath)){
// OK
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return strdup(mybname);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
}else if(basepath && 0 < strlen(basepath) && '/' == basepath[strlen(basepath) - 1] && 0 == strncmp(dirpath, basepath, strlen(basepath) - 1)){
string withdirname = "";
if(strlen(dirpath) > strlen(basepath)){
withdirname = &dirpath[strlen(basepath)];
}
if(0 < withdirname.length() && '/' != withdirname[withdirname.length() - 1]){
withdirname += "/";
}
withdirname += mybname;
return strdup(withdirname.c_str());
}
}
}
// case of something wrong
return (char*)c_strErrorObjectName;
}
static int remote_mountpath_exists(const char* path)
{
struct stat stbuf;
FPRNN("[path=%s]", path);
// getattr will prefix the path with the remote mountpoint
if(0 != get_object_attribute("/", &stbuf, NULL)){
return -1;
}
if(!S_ISDIR(stbuf.st_mode)){
return -1;
}
return 0;
}
2015-06-06 16:39:39 +00:00
static void free_xattrs(xattrs_t& xattrs)
{
for(xattrs_t::iterator iter = xattrs.begin(); iter != xattrs.end(); xattrs.erase(iter++)){
if(iter->second){
delete iter->second;
}
}
}
static bool parse_xattr_keyval(const std::string& xattrpair, string& key, PXATTRVAL& pval)
{
// parse key and value
size_t pos;
2015-06-06 16:39:39 +00:00
string tmpval;
if(string::npos == (pos = xattrpair.find_first_of(":"))){
DPRNNN("one of xattr pair(%s) is wrong format.", xattrpair.c_str());
return false;
}
2015-06-06 16:39:39 +00:00
key = xattrpair.substr(0, pos);
tmpval = xattrpair.substr(pos + 1);
2015-06-06 16:39:39 +00:00
if(!takeout_str_dquart(key) || !takeout_str_dquart(tmpval)){
DPRNNN("one of xattr pair(%s) is wrong format.", xattrpair.c_str());
return false;
}
2015-06-06 16:39:39 +00:00
pval = new XATTRVAL;
pval->length = 0;
pval->pvalue = s3fs_decode64(tmpval.c_str(), &pval->length);
return true;
}
static size_t parse_xattrs(const std::string& strxattrs, xattrs_t& xattrs)
{
xattrs.clear();
2015-06-06 16:39:39 +00:00
// decode
string jsonxattrs = urlDecode(strxattrs);
// get from "{" to "}"
string restxattrs;
{
size_t startpos = string::npos;
size_t endpos = string::npos;
2015-06-06 16:39:39 +00:00
if(string::npos != (startpos = jsonxattrs.find_first_of("{"))){
endpos = jsonxattrs.find_last_of("}");
}
if(startpos == string::npos || endpos == string::npos || endpos <= startpos){
2015-06-06 16:39:39 +00:00
DPRNNN("xattr header(%s) is not json format.", jsonxattrs.c_str());
return 0;
}
2015-06-06 16:39:39 +00:00
restxattrs = jsonxattrs.substr(startpos + 1, endpos - (startpos + 1));
}
// parse each key:val
for(size_t pair_nextpos = restxattrs.find_first_of(","); 0 < restxattrs.length(); restxattrs = (pair_nextpos != string::npos ? restxattrs.substr(pair_nextpos + 1) : string("")), pair_nextpos = restxattrs.find_first_of(",")){
string pair = pair_nextpos != string::npos ? restxattrs.substr(0, pair_nextpos) : restxattrs;
2015-06-06 16:39:39 +00:00
string key = "";
PXATTRVAL pval = NULL;
if(!parse_xattr_keyval(pair, key, pval)){
// something format error, so skip this.
continue;
}
2015-06-06 16:39:39 +00:00
xattrs[key] = pval;
}
return xattrs.size();
}
static std::string build_xattrs(const xattrs_t& xattrs)
{
string strxattrs("{");
bool is_set = false;
for(xattrs_t::const_iterator iter = xattrs.begin(); iter != xattrs.end(); ++iter){
if(is_set){
strxattrs += ',';
}else{
is_set = true;
}
strxattrs += '\"';
strxattrs += iter->first;
strxattrs += "\":\"";
2015-06-06 16:39:39 +00:00
if(iter->second){
char* base64val = s3fs_base64((iter->second)->pvalue, (iter->second)->length);
if(base64val){
strxattrs += base64val;
free(base64val);
}
}
strxattrs += '\"';
}
strxattrs += '}';
2015-06-06 16:39:39 +00:00
strxattrs = urlEncode(strxattrs);
return strxattrs;
}
static int set_xattrs_to_header(headers_t& meta, const char* name, const char* value, size_t size, int flags)
{
string strxattrs;
xattrs_t xattrs;
headers_t::iterator iter;
if(meta.end() == (iter = meta.find("x-amz-meta-xattr"))){
if(XATTR_REPLACE == (flags & XATTR_REPLACE)){
// there is no xattr header but flags is replace, so failure.
return -ENOATTR;
}
}else{
if(XATTR_CREATE == (flags & XATTR_CREATE)){
// found xattr header but flags is only creating, so failure.
return -EEXIST;
}
strxattrs = iter->second;
}
// get map as xattrs_t
parse_xattrs(strxattrs, xattrs);
// add name(do not care overwrite and empty name/value)
xattrs_t::iterator xiter;
if(xattrs.end() != (xiter = xattrs.find(string(name)))){
2015-06-06 16:39:39 +00:00
// found same head. free value.
delete xiter->second;
2015-06-06 16:39:39 +00:00
}
PXATTRVAL pval = new XATTRVAL;
pval->length = size;
if(0 < size){
if(NULL == (pval->pvalue = (unsigned char*)malloc(size))){
delete pval;
free_xattrs(xattrs);
return -ENOMEM;
}
memcpy(pval->pvalue, value, size);
}else{
pval->pvalue = NULL;
}
xattrs[string(name)] = pval;
// build new strxattrs(not encoded) and set it to headers_t
meta["x-amz-meta-xattr"] = build_xattrs(xattrs);
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return 0;
}
#if defined(__APPLE__)
static int s3fs_setxattr(const char* path, const char* name, const char* value, size_t size, int flags, uint32_t position)
#else
static int s3fs_setxattr(const char* path, const char* name, const char* value, size_t size, int flags)
#endif
{
2015-06-06 16:39:39 +00:00
FPRN("[path=%s][name=%s][value=%p][size=%zu][flags=%d]", path, name, value, size, flags);
if((value && 0 == size) || (!value && 0 < size)){
DPRN("Wrong parameter: value(%p), size(%zu)", value, size);
return 0;
}
#if defined(__APPLE__)
if (position != 0) {
// No resource fork support
return -EINVAL;
}
#endif
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
if(0 == strcmp(path, "/")){
DPRNNN("Could not change mode for mount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
if(0 != (result = check_object_owner(path, &stbuf))){
return result;
}
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
}
if(0 != result){
return result;
}
// make new header_t
if(0 != (result = set_xattrs_to_header(meta, name, value, size, flags))){
return result;
}
if(S_ISDIR(stbuf.st_mode) && IS_REPLACEDIR(nDirType)){
// Should rebuild directory object(except new type)
// Need to remove old dir("dir" etc) and make new dir("dir/")
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
return result;
}
}
StatCache::getStatCacheData()->DelStat(nowcache);
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), stbuf.st_mode, stbuf.st_mtime, stbuf.st_uid, stbuf.st_gid))){
return result;
}
// need to set xattr header for directory.
strpath = newpath;
nowcache = strpath;
}
// set xattr all object
meta["x-amz-copy-source"] = urlEncode(service_path + bucket + get_realpath(strpath.c_str()));
meta["x-amz-metadata-directive"] = "REPLACE";
if(0 != put_headers(strpath.c_str(), meta, true)){
return -EIO;
}
StatCache::getStatCacheData()->DelStat(nowcache);
return 0;
}
#if defined(__APPLE__)
static int s3fs_getxattr(const char* path, const char* name, char* value, size_t size, uint32_t position)
#else
static int s3fs_getxattr(const char* path, const char* name, char* value, size_t size)
#endif
{
FPRN("[path=%s][name=%s][value=%p][size=%zu]", path, name, value, size);
if(!path || !name){
return -EIO;
}
#if (__APPLE__)
if (position != 0) {
// No resource fork support
return -EINVAL;
}
#endif
int result;
headers_t meta;
xattrs_t xattrs;
// check parent directory attribute.
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
// get headders
if(0 != (result = get_object_attribute(path, NULL, &meta))){
return result;
}
// get xattrs
2015-06-13 03:27:07 +00:00
headers_t::iterator hiter = meta.find("x-amz-meta-xattr");
if(meta.end() == hiter){
// object does not have xattrs
return -ENOATTR;
}
2015-06-13 03:27:07 +00:00
string strxattrs = hiter->second;
2015-06-06 16:39:39 +00:00
parse_xattrs(strxattrs, xattrs);
// search name
2015-06-13 03:27:07 +00:00
string strname = name;
xattrs_t::iterator xiter = xattrs.find(strname);
if(xattrs.end() == xiter){
// not found name in xattrs
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return -ENOATTR;
}
2015-06-06 16:39:39 +00:00
// decode
size_t length = 0;
unsigned char* pvalue = NULL;
2015-06-13 03:27:07 +00:00
if(NULL != xiter->second){
length = xiter->second->length;
pvalue = xiter->second->pvalue;
}
2015-06-06 16:39:39 +00:00
if(0 < size){
if(static_cast<size_t>(size) < length){
// over buffer size
free_xattrs(xattrs);
return -ERANGE;
}
if(pvalue){
memcpy(value, pvalue, length);
}
}
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
2015-06-06 16:39:39 +00:00
return static_cast<int>(length);
}
static int s3fs_listxattr(const char* path, char* list, size_t size)
{
FPRN("[path=%s][list=%p][size=%zu]", path, list, size);
if(!path){
return -EIO;
}
int result;
headers_t meta;
xattrs_t xattrs;
// check parent directory attribute.
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
// get headders
if(0 != (result = get_object_attribute(path, NULL, &meta))){
return result;
}
// get xattrs
headers_t::iterator iter;
if(meta.end() == (iter = meta.find("x-amz-meta-xattr"))){
// object does not have xattrs
return 0;
}
string strxattrs = iter->second;
2015-06-06 16:39:39 +00:00
parse_xattrs(strxattrs, xattrs);
// calculate total name length
size_t total = 0;
for(xattrs_t::const_iterator iter = xattrs.begin(); iter != xattrs.end(); ++iter){
if(0 < iter->first.length()){
total += iter->first.length() + 1;
}
}
2015-06-06 16:39:39 +00:00
if(0 == total){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return 0;
}
// check parameters
if(size <= 0){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return total;
}
if(!list || size < total){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return -ERANGE;
}
// copy to list
char* setpos = list;
for(xattrs_t::const_iterator iter = xattrs.begin(); iter != xattrs.end(); ++iter){
if(0 < iter->first.length()){
strcpy(setpos, iter->first.c_str());
setpos = &setpos[strlen(setpos) + 1];
}
}
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return total;
}
static int s3fs_removexattr(const char* path, const char* name)
{
FPRN("[path=%s][name=%s]", path, name);
if(!path || !name){
return -EIO;
}
int result;
string strpath;
string newpath;
string nowcache;
headers_t meta;
xattrs_t xattrs;
struct stat stbuf;
int nDirType = DIRTYPE_UNKNOWN;
if(0 == strcmp(path, "/")){
DPRNNN("Could not change mode for mount point.");
return -EIO;
}
if(0 != (result = check_parent_object_access(path, X_OK))){
return result;
}
if(0 != (result = check_object_owner(path, &stbuf))){
return result;
}
if(S_ISDIR(stbuf.st_mode)){
result = chk_dir_object_type(path, newpath, strpath, nowcache, &meta, &nDirType);
}else{
strpath = path;
nowcache = strpath;
result = get_object_attribute(strpath.c_str(), NULL, &meta);
}
if(0 != result){
return result;
}
// get xattrs
2015-06-13 03:08:56 +00:00
headers_t::iterator hiter = meta.find("x-amz-meta-xattr");
if(meta.end() == hiter){
// object does not have xattrs
return -ENOATTR;
}
2015-06-13 03:08:56 +00:00
string strxattrs = hiter->second;
2015-06-06 16:39:39 +00:00
parse_xattrs(strxattrs, xattrs);
// check name xattrs
2015-06-13 03:08:56 +00:00
string strname = name;
xattrs_t::iterator xiter = xattrs.find(strname);
if(xattrs.end() == xiter){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return -ENOATTR;
}
// make new header_t after deleting name xattr
2015-06-13 03:08:56 +00:00
if(xiter->second){
delete xiter->second;
2015-06-06 16:39:39 +00:00
}
xattrs.erase(xiter);
2015-06-06 16:39:39 +00:00
// build new xattr
2015-06-13 03:27:07 +00:00
if(!xattrs.empty()){
2015-06-06 16:39:39 +00:00
meta["x-amz-meta-xattr"] = build_xattrs(xattrs);
}else{
meta.erase("x-amz-meta-xattr");
}
if(S_ISDIR(stbuf.st_mode) && IS_REPLACEDIR(nDirType)){
// Should rebuild directory object(except new type)
// Need to remove old dir("dir" etc) and make new dir("dir/")
// At first, remove directory old object
if(IS_RMTYPEDIR(nDirType)){
S3fsCurl s3fscurl;
if(0 != (result = s3fscurl.DeleteRequest(strpath.c_str()))){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return result;
}
}
StatCache::getStatCacheData()->DelStat(nowcache);
// Make new directory object("dir/")
if(0 != (result = create_directory_object(newpath.c_str(), stbuf.st_mode, stbuf.st_mtime, stbuf.st_uid, stbuf.st_gid))){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return result;
}
// need to set xattr header for directory.
strpath = newpath;
nowcache = strpath;
}
// set xattr all object
meta["x-amz-copy-source"] = urlEncode(service_path + bucket + get_realpath(strpath.c_str()));
meta["x-amz-metadata-directive"] = "REPLACE";
if(0 != put_headers(strpath.c_str(), meta, true)){
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return -EIO;
}
StatCache::getStatCacheData()->DelStat(nowcache);
2015-06-06 16:39:39 +00:00
free_xattrs(xattrs);
return 0;
}
static void* s3fs_init(struct fuse_conn_info* conn)
{
FPRN("init v%s(commit:%s) with %s", VERSION, COMMIT_HASH_VAL, s3fs_crypt_lib_name());
LOWSYSLOGPRINT(LOG_ERR, "init v%s(commit:%s) with %s", VERSION, COMMIT_HASH_VAL, s3fs_crypt_lib_name());
// ssl init
if(!s3fs_init_global_ssl()){
fprintf(stderr, "%s: could not initialize for ssl libraries.\n", program_name.c_str());
exit(EXIT_FAILURE);
}
// init curl
if(!S3fsCurl::InitS3fsCurl("/etc/mime.types")){
fprintf(stderr, "%s: Could not initiate curl library.\n", program_name.c_str());
LOWSYSLOGPRINT(LOG_ERR, "Could not initiate curl library.");
exit(EXIT_FAILURE);
}
if (create_bucket){
do_create_bucket();
}
// Check Bucket
// If the network is up, check for valid credentials and if the bucket
// exists. skip check if mounting a public bucket
if(!S3fsCurl::IsPublicBucket()){
int result;
if(EXIT_SUCCESS != (result = s3fs_check_service())){
exit(result);
}
}
// Investigate system capabilities
#ifndef __APPLE__
if((unsigned int)conn->capable & FUSE_CAP_ATOMIC_O_TRUNC){
conn->want |= FUSE_CAP_ATOMIC_O_TRUNC;
}
#endif
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// cache
if(is_remove_cache && !FdManager::DeleteCacheDirectory()){
DPRNINFO("Could not inilialize cache directory.");
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}
return NULL;
}
static void s3fs_destroy(void*)
{
DPRN("destroy");
2014-10-01 10:42:39 +00:00
// Destroy curl
if(!S3fsCurl::DestroyS3fsCurl()){
DPRN("Could not release curl library.");
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// cache
if(is_remove_cache && !FdManager::DeleteCacheDirectory()){
DPRN("Could not remove cache directory.");
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
}
// ssl
s3fs_destroy_global_ssl();
}
static int s3fs_access(const char* path, int mask)
{
FPRN("[path=%s][mask=%s%s%s%s]", path,
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
((mask & R_OK) == R_OK) ? "R_OK " : "",
((mask & W_OK) == W_OK) ? "W_OK " : "",
((mask & X_OK) == X_OK) ? "X_OK " : "",
(mask == F_OK) ? "F_OK" : "");
int result = check_object_access(path, mask, NULL);
S3FS_MALLOCTRIM(0);
return result;
}
static xmlChar* get_exp_value_xml(xmlDocPtr doc, xmlXPathContextPtr ctx, const char* exp_key)
{
if(!doc || !ctx || !exp_key){
return NULL;
}
xmlXPathObjectPtr exp;
xmlNodeSetPtr exp_nodes;
xmlChar* exp_value;
// search exp_key tag
if(NULL == (exp = xmlXPathEvalExpression((xmlChar*)exp_key, ctx))){
DPRNNN("Could not find key(%s).", exp_key);
return NULL;
}
if(xmlXPathNodeSetIsEmpty(exp->nodesetval)){
DPRNNN("Key(%s) node is empty.", exp_key);
S3FS_XMLXPATHFREEOBJECT(exp);
return NULL;
}
// get exp_key value & set in struct
exp_nodes = exp->nodesetval;
if(NULL == (exp_value = xmlNodeListGetString(doc, exp_nodes->nodeTab[0]->xmlChildrenNode, 1))){
DPRNNN("Key(%s) value is empty.", exp_key);
S3FS_XMLXPATHFREEOBJECT(exp);
return NULL;
}
S3FS_XMLXPATHFREEOBJECT(exp);
return exp_value;
}
static void print_uncomp_mp_list(uncomp_mp_list_t& list)
{
printf("\n");
printf("Lists the parts that have been uploaded for a specific multipart upload.\n");
printf("\n");
2015-08-12 15:04:16 +00:00
if(!list.empty()){
printf("---------------------------------------------------------------\n");
int cnt = 0;
2015-08-12 15:04:16 +00:00
for(uncomp_mp_list_t::iterator iter = list.begin(); iter != list.end(); ++iter, ++cnt){
printf(" Path : %s\n", (*iter).key.c_str());
printf(" UploadId : %s\n", (*iter).id.c_str());
printf(" Date : %s\n", (*iter).date.c_str());
printf("\n");
}
printf("---------------------------------------------------------------\n");
}else{
printf("There is no list.\n");
}
}
static bool abort_uncomp_mp_list(uncomp_mp_list_t& list)
{
char buff[1024];
2015-08-12 15:04:16 +00:00
if(list.empty()){
return true;
}
memset(buff, 0, sizeof(buff));
// confirm
while(true){
printf("Would you remove all objects? [Y/N]\n");
if(NULL != fgets(buff, sizeof(buff), stdin)){
if(0 == strcasecmp(buff, "Y\n") || 0 == strcasecmp(buff, "YES\n")){
break;
}else if(0 == strcasecmp(buff, "N\n") || 0 == strcasecmp(buff, "NO\n")){
return true;
}
printf("*** please put Y(yes) or N(no).\n");
}
}
// do removing their.
S3fsCurl s3fscurl;
bool result = true;
2015-08-12 15:04:16 +00:00
for(uncomp_mp_list_t::iterator iter = list.begin(); iter != list.end(); ++iter){
const char* tpath = (*iter).key.c_str();
string upload_id = (*iter).id;
if(0 != s3fscurl.AbortMultipartUpload(tpath, upload_id)){
fprintf(stderr, "Failed to remove %s multipart uploading object.\n", tpath);
result = false;
}else{
printf("Succeed to remove %s multipart uploading object.\n", tpath);
}
// reset(initialize) curl object
s3fscurl.DestroyCurlHandle();
}
return result;
}
static bool get_uncomp_mp_list(xmlDocPtr doc, uncomp_mp_list_t& list)
{
if(!doc){
return false;
}
xmlXPathContextPtr ctx = xmlXPathNewContext(doc);;
string xmlnsurl;
string ex_upload = "//";
string ex_key = "";
string ex_id = "";
string ex_date = "";
if(!noxmlns && GetXmlNsUrl(doc, xmlnsurl)){
xmlXPathRegisterNs(ctx, (xmlChar*)"s3", (xmlChar*)xmlnsurl.c_str());
ex_upload += "s3:";
ex_key += "s3:";
ex_id += "s3:";
ex_date += "s3:";
}
ex_upload += "Upload";
ex_key += "Key";
ex_id += "UploadId";
ex_date += "Initiated";
// get "Upload" Tags
xmlXPathObjectPtr upload_xp;
if(NULL == (upload_xp = xmlXPathEvalExpression((xmlChar*)ex_upload.c_str(), ctx))){
DPRNNN("xmlXPathEvalExpression returns null.");
return false;
}
if(xmlXPathNodeSetIsEmpty(upload_xp->nodesetval)){
DPRNNN("upload_xp->nodesetval is empty.");
S3FS_XMLXPATHFREEOBJECT(upload_xp);
S3FS_XMLXPATHFREECONTEXT(ctx);
return true;
}
// Make list
int cnt;
xmlNodeSetPtr upload_nodes;
list.clear();
for(cnt = 0, upload_nodes = upload_xp->nodesetval; cnt < upload_nodes->nodeNr; cnt++){
ctx->node = upload_nodes->nodeTab[cnt];
UNCOMP_MP_INFO part;
xmlChar* ex_value;
// search "Key" tag
if(NULL == (ex_value = get_exp_value_xml(doc, ctx, ex_key.c_str()))){
continue;
}
if('/' != *((char*)ex_value)){
part.key = "/";
}else{
part.key = "";
}
part.key += (char*)ex_value;
S3FS_XMLFREE(ex_value);
// search "UploadId" tag
if(NULL == (ex_value = get_exp_value_xml(doc, ctx, ex_id.c_str()))){
continue;
}
part.id = (char*)ex_value;
S3FS_XMLFREE(ex_value);
// search "Initiated" tag
if(NULL == (ex_value = get_exp_value_xml(doc, ctx, ex_date.c_str()))){
continue;
}
part.date = (char*)ex_value;
S3FS_XMLFREE(ex_value);
list.push_back(part);
}
S3FS_XMLXPATHFREEOBJECT(upload_xp);
S3FS_XMLXPATHFREECONTEXT(ctx);
return true;
}
static int s3fs_utility_mode(void)
{
if(!utility_mode){
return EXIT_FAILURE;
}
// ssl init
if(!s3fs_init_global_ssl()){
fprintf(stderr, "%s: could not initialize for ssl libraries.\n", program_name.c_str());
return EXIT_FAILURE;
}
// init curl
if(!S3fsCurl::InitS3fsCurl("/etc/mime.types")){
fprintf(stderr, "%s: Could not initiate curl library.\n", program_name.c_str());
LOWSYSLOGPRINT(LOG_ERR, "Could not initiate curl library.");
s3fs_destroy_global_ssl();
return EXIT_FAILURE;
}
printf("Utility Mode\n");
S3fsCurl s3fscurl;
string body;
int result = EXIT_SUCCESS;
if(0 != s3fscurl.MultipartListRequest(body)){
fprintf(stderr, "%s: Could not get list multipart upload.\n", program_name.c_str());
result = EXIT_FAILURE;
}else{
// perse result(uncomplete multipart upload information)
FPRNINFO("response body = {\n%s\n}", body.c_str());
xmlDocPtr doc;
if(NULL == (doc = xmlReadMemory(body.c_str(), static_cast<int>(body.size()), "", NULL, 0))){
2014-10-01 10:42:39 +00:00
DPRN("xmlReadMemory exited with error.");
result = EXIT_FAILURE;
}else{
// make working uploads list
uncomp_mp_list_t list;
if(!get_uncomp_mp_list(doc, list)){
2014-10-01 10:42:39 +00:00
DPRN("get_uncomp_mp_list exited with error.");
result = EXIT_FAILURE;
}else{
// print list
print_uncomp_mp_list(list);
// remove
if(!abort_uncomp_mp_list(list)){
2014-10-01 10:42:39 +00:00
DPRN("an error occurred during removal process.");
result = EXIT_FAILURE;
}
}
S3FS_XMLFREEDOC(doc);
}
}
2014-10-01 10:42:39 +00:00
// Destroy curl
if(!S3fsCurl::DestroyS3fsCurl()){
DPRN("Could not release curl library.");
}
// ssl
s3fs_destroy_global_ssl();
return result;
}
//
// If calling with wrong region, s3fs gets following error body as 400 erro code.
// "<Error><Code>AuthorizationHeaderMalformed</Code><Message>The authorization header is
// malformed; the region 'us-east-1' is wrong; expecting 'ap-northeast-1'</Message>
// <Region>ap-northeast-1</Region><RequestId>...</RequestId><HostId>...</HostId>
// </Error>"
//
// So this is cheep codes but s3fs should get correct reagion automatically.
//
static bool check_region_error(const char* pbody, string& expectregion)
{
if(!pbody){
return false;
}
const char* region;
const char* regionend;
if(NULL == (region = strcasestr(pbody, "<Message>The authorization header is malformed; the region "))){
return false;
}
if(NULL == (region = strcasestr(region, "expecting \'"))){
return false;
}
region += strlen("expecting \'");
if(NULL == (regionend = strchr(region, '\''))){
return false;
}
string strtmp(region, (regionend - region));
if(0 == strtmp.length()){
return false;
}
expectregion = strtmp;
return true;
}
static int s3fs_check_service(void)
{
FPRN("check services.");
// At first time for access S3, we check IAM role if it sets.
if(!S3fsCurl::CheckIAMCredentialUpdate()){
fprintf(stderr, "%s: Failed to check IAM role name(%s).\n", program_name.c_str(), S3fsCurl::GetIAMRole());
return EXIT_FAILURE;
}
S3fsCurl s3fscurl;
int res;
if(0 > (res = s3fscurl.CheckBucket())){
// get response code
long responseCode = s3fscurl.GetLastResponseCode();
// check wrong endpoint, and automatically switch endpoint
if(responseCode == 400 && !is_specified_endpoint){
// check region error
BodyData* body = s3fscurl.GetBodyData();
string expectregion;
if(check_region_error(body->str(), expectregion)){
// not specified endpoint, so try to connect to expected region.
LOWSYSLOGPRINT(LOG_ERR, "Could not connect wrong region %s, so retry to connect region %s.", endpoint.c_str(), expectregion.c_str());
FPRN("Could not connect wrong region %s, so retry to connect region %s.", endpoint.c_str(), expectregion.c_str());
endpoint = expectregion;
if(S3fsCurl::IsSignatureV4()){
if(host == "http://s3.amazonaws.com"){
host = "http://s3-" + endpoint + ".amazonaws.com";
}else if(host == "https://s3.amazonaws.com"){
host = "https://s3-" + endpoint + ".amazonaws.com";
}
}
// retry to check with new endpoint
s3fscurl.DestroyCurlHandle();
res = s3fscurl.CheckBucket();
responseCode = s3fscurl.GetLastResponseCode();
}
}
// try signature v2
if(0 > res && (responseCode == 400 || responseCode == 403) && S3fsCurl::IsSignatureV4()){
// switch sigv2
LOWSYSLOGPRINT(LOG_ERR, "Could not connect, so retry to connect by signature version 2.");
FPRN("Could not connect, so retry to connect by signature version 2.");
S3fsCurl::SetSignatureV4(false);
// retry to check with sigv2
s3fscurl.DestroyCurlHandle();
res = s3fscurl.CheckBucket();
responseCode = s3fscurl.GetLastResponseCode();
}
// check errors(after retrying)
if(0 > res && responseCode != 200 && responseCode != 301){
if(responseCode == 400){
fprintf(stderr, "%s: Bad Request\n", program_name.c_str());
return EXIT_FAILURE;
}
if(responseCode == 403){
fprintf(stderr, "%s: invalid credentials\n", program_name.c_str());
return EXIT_FAILURE;
}
if(responseCode == 404){
fprintf(stderr, "%s: bucket not found\n", program_name.c_str());
return EXIT_FAILURE;
}
// unable to connect
if(responseCode == CURLE_OPERATION_TIMEDOUT){
fprintf(stderr, "%s: unable to connect bucket and timeout\n", program_name.c_str());
return EXIT_FAILURE;
}
// another error
fprintf(stderr, "%s: unable to connect\n", program_name.c_str());
return EXIT_FAILURE;
}
}
// make sure remote mountpath exists and is a directory
if(mount_prefix.size() > 0){
if(remote_mountpath_exists(mount_prefix.c_str()) != 0){
fprintf(stderr, "%s: remote mountpath %s not found.\n",
program_name.c_str(), mount_prefix.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
}
S3FS_MALLOCTRIM(0);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
// Return: 1 - OK(could read and set accesskey etc.)
// 0 - NG(could not read)
// -1 - Should shoutdown immidiatly
static int check_for_aws_format(void)
{
size_t first_pos = string::npos;
string line;
bool got_access_key_id_line = 0;
bool got_secret_key_line = 0;
string str1 ("AWSAccessKeyId=");
string str2 ("AWSSecretKey=");
size_t found;
string AccessKeyId;
string SecretAccesskey;
ifstream PF(passwd_file.c_str());
if(PF.good()){
while (getline(PF, line)){
if(line[0]=='#'){
continue;
}
if(line.size() == 0){
continue;
}
if('\r' == line[line.size() - 1]){
line = line.substr(0, line.size() - 1);
if(line.size() == 0){
continue;
}
}
first_pos = line.find_first_of(" \t");
if(first_pos != string::npos){
printf ("%s: invalid line in passwd file, found whitespace character\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
first_pos = line.find_first_of("[");
if(first_pos != string::npos && first_pos == 0){
printf ("%s: invalid line in passwd file, found a bracket \"[\" character\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
found = line.find(str1);
if(found != string::npos){
first_pos = line.find_first_of("=");
AccessKeyId = line.substr(first_pos + 1, string::npos);
got_access_key_id_line = 1;
continue;
}
found = line.find(str2);
if(found != string::npos){
first_pos = line.find_first_of("=");
SecretAccesskey = line.substr(first_pos + 1, string::npos);
got_secret_key_line = 1;
continue;
}
}
}
if(got_access_key_id_line && got_secret_key_line){
if(!S3fsCurl::SetAccessKey(AccessKeyId.c_str(), SecretAccesskey.c_str())){
fprintf(stderr, "%s: if one access key is specified, both keys need to be specified\n", program_name.c_str());
return 0;
}
return 1;
}else{
return 0;
}
}
//
// check_passwd_file_perms
//
// expect that global passwd_file variable contains
// a non-empty value and is readable by the current user
//
// Check for too permissive access to the file
// help save users from themselves via a security hole
//
// only two options: return or error out
//
static int check_passwd_file_perms(void)
{
struct stat info;
// let's get the file info
if(stat(passwd_file.c_str(), &info) != 0){
fprintf (stderr, "%s: unexpected error from stat(%s, ) \n",
program_name.c_str(), passwd_file.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
// return error if any file has others permissions
if( (info.st_mode & S_IROTH) ||
(info.st_mode & S_IWOTH) ||
(info.st_mode & S_IXOTH)) {
fprintf (stderr, "%s: credentials file %s should not have others permissions\n",
program_name.c_str(), passwd_file.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
// Any local file should not have any group permissions
// /etc/passwd-s3fs can have group permissions
if(passwd_file != "/etc/passwd-s3fs"){
if( (info.st_mode & S_IRGRP) ||
(info.st_mode & S_IWGRP) ||
(info.st_mode & S_IXGRP)) {
fprintf (stderr, "%s: credentials file %s should not have group permissions\n",
program_name.c_str(), passwd_file.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
}else{
// "/etc/passwd-s3fs" does not allow group write.
if((info.st_mode & S_IWGRP)){
fprintf (stderr, "%s: credentials file %s should not have group writable permissions\n",
program_name.c_str(), passwd_file.c_str());
return EXIT_FAILURE;
}
}
if((info.st_mode & S_IXUSR) || (info.st_mode & S_IXGRP)){
fprintf (stderr, "%s: credentials file %s should not have executable permissions\n",
program_name.c_str(), passwd_file.c_str());
return EXIT_FAILURE;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
//
// read_passwd_file
//
// Support for per bucket credentials
//
// Format for the credentials file:
// [bucket:]AccessKeyId:SecretAccessKey
//
// Lines beginning with # are considered comments
// and ignored, as are empty lines
//
// Uncommented lines without the ":" character are flagged as
// an error, so are lines with spaces or tabs
//
// only one default key pair is allowed, but not required
//
static int read_passwd_file(void)
{
string line;
string field1, field2, field3;
size_t first_pos = string::npos;
size_t last_pos = string::npos;
bool default_found = 0;
2014-10-17 06:32:41 +00:00
int aws_format;
// if you got here, the password file
// exists and is readable by the
// current user, check for permissions
if(EXIT_SUCCESS != check_passwd_file_perms()){
return EXIT_FAILURE;
}
aws_format = check_for_aws_format();
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(1 == aws_format){
return EXIT_SUCCESS;
}else if(-1 == aws_format){
return EXIT_FAILURE;
}
ifstream PF(passwd_file.c_str());
if(PF.good()){
while (getline(PF, line)){
if(line[0]=='#'){
continue;
}
if(line.size() == 0){
continue;
}
if('\r' == line[line.size() - 1]){
line = line.substr(0, line.size() - 1);
if(line.size() == 0){
continue;
}
}
first_pos = line.find_first_of(" \t");
if(first_pos != string::npos){
printf ("%s: invalid line in passwd file, found whitespace character\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
first_pos = line.find_first_of("[");
if(first_pos != string::npos && first_pos == 0){
printf ("%s: invalid line in passwd file, found a bracket \"[\" character\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
first_pos = line.find_first_of(":");
if(first_pos == string::npos){
printf ("%s: invalid line in passwd file, no \":\" separator found\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
last_pos = line.find_last_of(":");
if(first_pos != last_pos){
// bucket specified
field1 = line.substr(0,first_pos);
field2 = line.substr(first_pos + 1, last_pos - first_pos - 1);
field3 = line.substr(last_pos + 1, string::npos);
}else{
// no bucket specified - original style - found default key
if(default_found == 1){
printf ("%s: more than one default key pair found in passwd file\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
default_found = 1;
field1.assign("");
field2 = line.substr(0,first_pos);
field3 = line.substr(first_pos + 1, string::npos);
if(!S3fsCurl::SetAccessKey(field2.c_str(), field3.c_str())){
fprintf(stderr, "%s: if one access key is specified, both keys need to be specified\n", program_name.c_str());
return EXIT_FAILURE;
}
}
// does the bucket we are mounting match this passwd file entry?
// if so, use that key pair, otherwise use the default key, if found,
// will be used
if(field1.size() != 0 && field1 == bucket){
if(!S3fsCurl::SetAccessKey(field2.c_str(), field3.c_str())){
fprintf(stderr, "%s: if one access key is specified, both keys need to be specified\n", program_name.c_str());
return EXIT_FAILURE;
}
break;
}
}
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
//
// get_access_keys
//
// called only when were are not mounting a
// public bucket
//
// Here is the order precedence for getting the
// keys:
//
// 1 - from the command line (security risk)
// 2 - from a password file specified on the command line
// 3 - from environment variables
// 4 - from the users ~/.passwd-s3fs
// 5 - from /etc/passwd-s3fs
//
static int get_access_keys(void)
{
// should be redundant
if(S3fsCurl::IsPublicBucket()){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
// 1 - keys specified on the command line
if(S3fsCurl::IsSetAccessKeyId()){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
// 2 - was specified on the command line
if(passwd_file.size() > 0){
ifstream PF(passwd_file.c_str());
if(PF.good()){
PF.close();
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return read_passwd_file();
}else{
fprintf(stderr, "%s: specified passwd_file is not readable\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
}
// 3 - environment variables
char* AWSACCESSKEYID = getenv("AWSACCESSKEYID");
char* AWSSECRETACCESSKEY = getenv("AWSSECRETACCESSKEY");
if(AWSACCESSKEYID != NULL || AWSSECRETACCESSKEY != NULL){
if( (AWSACCESSKEYID == NULL && AWSSECRETACCESSKEY != NULL) ||
(AWSACCESSKEYID != NULL && AWSSECRETACCESSKEY == NULL) ){
fprintf(stderr, "%s: if environment variable AWSACCESSKEYID is set then AWSSECRETACCESSKEY must be set too\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
if(!S3fsCurl::SetAccessKey(AWSACCESSKEYID, AWSSECRETACCESSKEY)){
fprintf(stderr, "%s: if one access key is specified, both keys need to be specified\n", program_name.c_str());
return EXIT_FAILURE;
}
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
// 3a - from the AWS_CREDENTIAL_FILE environment variable
char * AWS_CREDENTIAL_FILE;
AWS_CREDENTIAL_FILE = getenv("AWS_CREDENTIAL_FILE");
if(AWS_CREDENTIAL_FILE != NULL){
passwd_file.assign(AWS_CREDENTIAL_FILE);
if(passwd_file.size() > 0){
ifstream PF(passwd_file.c_str());
if(PF.good()){
PF.close();
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return read_passwd_file();
}else{
fprintf(stderr, "%s: AWS_CREDENTIAL_FILE: \"%s\" is not readable\n",
program_name.c_str(), passwd_file.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
}
}
// 4 - from the default location in the users home directory
char * HOME;
HOME = getenv ("HOME");
if(HOME != NULL){
passwd_file.assign(HOME);
passwd_file.append("/.passwd-s3fs");
ifstream PF(passwd_file.c_str());
if(PF.good()){
PF.close();
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(EXIT_SUCCESS != read_passwd_file()){
return EXIT_FAILURE;
}
// It is possible that the user's file was there but
// contained no key pairs i.e. commented out
// in that case, go look in the final location
if(S3fsCurl::IsSetAccessKeyId()){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_SUCCESS;
}
}
}
// 5 - from the system default location
passwd_file.assign("/etc/passwd-s3fs");
ifstream PF(passwd_file.c_str());
if(PF.good()){
PF.close();
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return read_passwd_file();
}
fprintf(stderr, "%s: could not determine how to establish security credentials\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return EXIT_FAILURE;
}
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
//
// Check & Set attributes for mount point.
//
static int set_moutpoint_attribute(struct stat& mpst)
{
mp_uid = geteuid();
mp_gid = getegid();
mp_mode = S_IFDIR | (allow_other ? (is_mp_umask ? (~mp_umask & (S_IRWXU | S_IRWXG | S_IRWXO)) : (S_IRWXU | S_IRWXG | S_IRWXO)) : S_IRWXU);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
FPRNNN("PROC(uid=%u, gid=%u) - MountPoint(uid=%u, gid=%u, mode=%04o)",
(unsigned int)mp_uid, (unsigned int)mp_gid, (unsigned int)(mpst.st_uid), (unsigned int)(mpst.st_gid), mpst.st_mode);
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
// check owner
if(0 == mp_uid || mpst.st_uid == mp_uid){
return true;
}
// check group permission
if(mpst.st_gid == mp_gid || 1 == is_uid_inculde_group(mp_uid, mpst.st_gid)){
if(S_IRWXG == (mpst.st_mode & S_IRWXG)){
return true;
}
}
// check other permission
if(S_IRWXO == (mpst.st_mode & S_IRWXO)){
return true;
}
return false;
}
// This is repeatedly called by the fuse option parser
// if the key is equal to FUSE_OPT_KEY_OPT, it's an option passed in prefixed by
// '-' or '--' e.g.: -f -d -ousecache=/tmp
//
// if the key is equal to FUSE_OPT_KEY_NONOPT, it's either the bucket name
// or the mountpoint. The bucket name will always come before the mountpoint
static int my_fuse_opt_proc(void* data, const char* arg, int key, struct fuse_args* outargs)
{
if(key == FUSE_OPT_KEY_NONOPT){
// the first NONOPT option is the bucket name
if(bucket.size() == 0){
// extract remote mount path
char *bucket_name = (char*)arg;
if(strstr(arg, ":")){
bucket = strtok(bucket_name, ":");
char* pmount_prefix = strtok(NULL, ":");
if(pmount_prefix){
if(0 == strlen(pmount_prefix) || '/' != pmount_prefix[0]){
fprintf(stderr, "%s: path(%s) must be prefix \"/\".\n", program_name.c_str(), pmount_prefix);
return -1;
}
mount_prefix = pmount_prefix;
// remove trailing slash
if(mount_prefix.at(mount_prefix.size() - 1) == '/'){
mount_prefix = mount_prefix.substr(0, mount_prefix.size() - 1);
}
}
}else{
bucket = arg;
}
return 0;
}
// the second NONPOT option is the mountpoint(not utility mode)
if(0 == mountpoint.size() && 0 == utility_mode){
// save the mountpoint and do some basic error checking
mountpoint = arg;
struct stat stbuf;
if(stat(arg, &stbuf) == -1){
fprintf(stderr, "%s: unable to access MOUNTPOINT %s: %s\n",
program_name.c_str(), mountpoint.c_str(), strerror(errno));
return -1;
}
if(!(S_ISDIR(stbuf.st_mode))){
fprintf(stderr, "%s: MOUNTPOINT: %s is not a directory\n",
program_name.c_str(), mountpoint.c_str());
return -1;
}
if(!set_moutpoint_attribute(stbuf)){
fprintf(stderr, "%s: MOUNTPOINT: %s permission denied.\n",
program_name.c_str(), mountpoint.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
if(!nonempty){
struct dirent *ent;
DIR *dp = opendir(mountpoint.c_str());
if(dp == NULL){
fprintf(stderr, "%s: failed to open MOUNTPOINT: %s: %s\n",
program_name.c_str(), mountpoint.c_str(), strerror(errno));
return -1;
}
while((ent = readdir(dp)) != NULL){
if(strcmp(ent->d_name, ".") != 0 && strcmp(ent->d_name, "..") != 0){
closedir(dp);
fprintf(stderr, "%s: MOUNTPOINT directory %s is not empty.\n"
"%s: if you are sure this is safe, can use the 'nonempty' mount option.\n",
program_name.c_str(), mountpoint.c_str(), program_name.c_str());
return -1;
}
}
closedir(dp);
}
return 1;
}
// Unknow option
if(0 == utility_mode){
fprintf(stderr, "%s: specified unknown third optioni(%s).\n", program_name.c_str(), arg);
}else{
fprintf(stderr, "%s: specified unknown second optioni(%s).\n"
"%s: you don't need to specify second option(mountpoint) for utility mode(-u).\n",
program_name.c_str(), arg, program_name.c_str());
}
return -1;
}else if(key == FUSE_OPT_KEY_OPT){
if(0 == STR2NCMP(arg, "uid=")){
s3fs_uid = get_uid(strchr(arg, '=') + sizeof(char));
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 != geteuid() && 0 == s3fs_uid){
fprintf(stderr, "%s: root user can only specify uid=0.\n", program_name.c_str());
return -1;
}
is_s3fs_uid = true;
return 1; // continue for fuse option
}
if(0 == STR2NCMP(arg, "gid=")){
s3fs_gid = get_gid(strchr(arg, '=') + sizeof(char));
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
if(0 != getegid() && 0 == s3fs_gid){
fprintf(stderr, "%s: root user can only specify gid=0.\n", program_name.c_str());
return -1;
}
is_s3fs_gid = true;
return 1; // continue for fuse option
}
if(0 == STR2NCMP(arg, "umask=")){
s3fs_umask = strtol(strchr(arg, '=') + sizeof(char), NULL, 0);
s3fs_umask &= (S_IRWXU | S_IRWXG | S_IRWXO);
is_s3fs_umask = true;
return 1; // continue for fuse option
}
if(0 == strcmp(arg, "allow_other")){
allow_other = true;
return 1; // continue for fuse option
}
if(0 == STR2NCMP(arg, "mp_umask=")){
mp_umask = strtol(strchr(arg, '=') + sizeof(char), NULL, 0);
mp_umask &= (S_IRWXU | S_IRWXG | S_IRWXO);
is_mp_umask = true;
return 0;
}
if(0 == STR2NCMP(arg, "default_acl=")){
const char* acl = strchr(arg, '=') + sizeof(char);
S3fsCurl::SetDefaultAcl(acl);
return 0;
}
if(0 == STR2NCMP(arg, "retries=")){
S3fsCurl::SetRetries(static_cast<int>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char))));
return 0;
}
if(0 == STR2NCMP(arg, "use_cache=")){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
FdManager::SetCacheDir(strchr(arg, '=') + sizeof(char));
return 0;
}
if(0 == strcmp(arg, "del_cache")){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
is_remove_cache = true;
return 0;
}
if(0 == STR2NCMP(arg, "multireq_max=")){
long maxreq = static_cast<long>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
S3fsMultiCurl::SetMaxMultiRequest(maxreq);
return 0;
}
if(0 == strcmp(arg, "nonempty")){
nonempty = true;
return 1; // need to continue for fuse.
}
if(0 == strcmp(arg, "nomultipart")){
nomultipart = true;
return 0;
}
// old format for storage_class
if(0 == strcmp(arg, "use_rrs") || 0 == STR2NCMP(arg, "use_rrs=")){
off_t rrs = 1;
// for an old format.
if(0 == STR2NCMP(arg, "use_rrs=")){
rrs = s3fs_strtoofft(strchr(arg, '=') + sizeof(char));
}
if(0 == rrs){
S3fsCurl::SetStorageClass(STANDARD);
}else if(1 == rrs){
if(S3fsCurl::GetUseSse()){
fprintf(stderr, "%s: use_rrs option could not be specified with use_sse.\n", program_name.c_str());
return -1;
}
S3fsCurl::SetStorageClass(REDUCED_REDUNDANCY);
}else{
fprintf(stderr, "%s: poorly formed argument to option: use_rrs\n", program_name.c_str());
return -1;
}
return 0;
}
if(0 == STR2NCMP(arg, "storage_class=")){
const char *storage_class = strchr(arg, '=') + sizeof(char);
if(0 == strcmp(storage_class, "standard")){
S3fsCurl::SetStorageClass(STANDARD);
}else if(0 == strcmp(storage_class, "standard_ia")){
S3fsCurl::SetStorageClass(STANDARD_IA);
}else if(0 == strcmp(storage_class, "reduced_redundancy")){
if(S3fsCurl::GetUseSse()){
fprintf(stderr, "%s: storage class reduced_redundancy option could not be specified with use_sse.\n", program_name.c_str());
return -1;
}
S3fsCurl::SetStorageClass(REDUCED_REDUNDANCY);
}else{
fprintf(stderr, "%s: unknown value for storage_class: %s\n", program_name.c_str(), storage_class);
return -1;
}
return 0;
}
if(0 == strcmp(arg, "use_sse") || 0 == STR2NCMP(arg, "use_sse=")){
if(0 == STR2NCMP(arg, "use_sse=")){
if(REDUCED_REDUNDANCY == S3fsCurl::GetStorageClass()){
fprintf(stderr, "%s: use_sse option could not be specified with storage class reduced_redundancy.\n", program_name.c_str());
return -1;
}
2014-07-19 19:02:55 +00:00
const char* ssecfile = &arg[strlen("use_sse=")];
if(0 == strcmp(ssecfile, "1")){
2014-08-26 17:11:10 +00:00
if(S3fsCurl::IsSseCustomMode()){
fprintf (stderr, "%s: already set SSE-C key by environment, and confrict use_sse option.\n", program_name.c_str());
return -1;
}
2014-07-19 19:02:55 +00:00
S3fsCurl::SetUseSse(true);
}else{
// testing sse-c, try to load AES256 keys
struct stat st;
if(0 != stat(ssecfile, &st)){
fprintf (stderr, "%s: could not open use_sse keys file(%s)\n", program_name.c_str(), ssecfile);
return -1;
}
if(st.st_mode & (S_IXUSR | S_IRWXG | S_IRWXO)){
fprintf (stderr, "%s: use_sse keys file %s should be 0600 permissions\n", program_name.c_str(), ssecfile);
return -1;
}
if(!S3fsCurl::SetSseKeys(ssecfile)){
fprintf (stderr, "%s: failed to load use_sse keys file %s\n", program_name.c_str(), ssecfile);
return -1;
}
}
}else{
if(REDUCED_REDUNDANCY == S3fsCurl::GetStorageClass()){
fprintf(stderr, "%s: use_sse option could not be specified with storage class reduced_redundancy.\n", program_name.c_str());
2014-07-19 19:02:55 +00:00
return -1;
}
2014-08-26 17:11:10 +00:00
if(S3fsCurl::IsSseCustomMode()){
fprintf (stderr, "%s: already set SSE-C key by environment, and confrict use_sse option.\n", program_name.c_str());
return -1;
}
2014-07-19 19:02:55 +00:00
S3fsCurl::SetUseSse(true);
}
return 0;
}
if(0 == STR2NCMP(arg, "ssl_verify_hostname=")){
long sslvh = static_cast<long>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
if(-1 == S3fsCurl::SetSslVerifyHostname(sslvh)){
fprintf(stderr, "%s: poorly formed argument to option: ssl_verify_hostname\n",
program_name.c_str());
return -1;
}
return 0;
}
if(0 == STR2NCMP(arg, "passwd_file=")){
passwd_file = strchr(arg, '=') + sizeof(char);
return 0;
}
if(0 == STR2NCMP(arg, "iam_role=")){
const char* role = strchr(arg, '=') + sizeof(char);
S3fsCurl::SetIAMRole(role);
return 0;
}
if(0 == STR2NCMP(arg, "public_bucket=")){
off_t pubbucket = s3fs_strtoofft(strchr(arg, '=') + sizeof(char));
if(1 == pubbucket){
S3fsCurl::SetPublicBucket(true);
}else if(0 == pubbucket){
S3fsCurl::SetPublicBucket(false);
}else{
fprintf(stderr, "%s: poorly formed argument to option: public_bucket\n",
program_name.c_str());
return -1;
}
return 0;
}
if(0 == STR2NCMP(arg, "host=")){
host = strchr(arg, '=') + sizeof(char);
return 0;
}
if(0 == STR2NCMP(arg, "servicepath=")){
service_path = strchr(arg, '=') + sizeof(char);
return 0;
}
if(0 == strcmp(arg, "no_check_certificate")){
S3fsCurl::SetCheckCertificate(false);
return 0;
}
if(0 == STR2NCMP(arg, "connect_timeout=")){
long contimeout = static_cast<long>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
S3fsCurl::SetConnectTimeout(contimeout);
return 0;
}
if(0 == STR2NCMP(arg, "readwrite_timeout=")){
time_t rwtimeout = static_cast<time_t>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
S3fsCurl::SetReadwriteTimeout(rwtimeout);
return 0;
}
if(0 == STR2NCMP(arg, "max_stat_cache_size=")){
unsigned long cache_size = static_cast<unsigned long>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
StatCache::getStatCacheData()->SetCacheSize(cache_size);
return 0;
}
if(0 == STR2NCMP(arg, "stat_cache_expire=")){
time_t expr_time = static_cast<time_t>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
StatCache::getStatCacheData()->SetExpireTime(expr_time);
Summary of Changes(1.62 -> 1.63) 1) Lifetime for the stats cache Added the new option "stat_cache_expire". This option which is specified by seconds means the lifetime for each stats cache entry. If this option is not specified, the stats cache is kept in s3fs process until the stats cache grown to maximum size. (default) If this option is specified, the stats cache entry is out from the memory when the entry expires time. 2) Enable file permission s3fs before 1.62 did not consider the file access permission. s3fs after this version can consider it. For access permission, the s3fs_getattr() function was divided into sub function which can check the file access permission. It is like access() function. And the function calling the s3fs_getattr() calls this new sub function instead of s3fs_getattr(). Last the s3fs_opendir() function which is called by FUSE was added for checking directory access permission when listing the files in directory. 3) UID/GUID When a file or a directory was created, the s3fs could not set the UID/GID as the user who executed a command. (Almost the UID/GID are root, because the s3fs run by root.) After this version, the s3fs set correct UID/GID as the user who executes the commond. 4) About the mtime If the object does not have "x-amz-meta-mtime" meta, the s3fs uses the "Last-Modified" header instead of it. But the s3fs had a bug in this code, and this version fixed this bug. When user modified the file, the s3fs did not update the mtime of the file. This version fixed this bug. In the get_local_fd() function, the local file's mtime was changed only when s3fs run with "use_cache" option. This version always updates the mtime whether the local cache file is used or not. And s3fs_flush ( ) function set the mtime of local cache file from S3 object mtime, but it was wrong . This version is that the s3fs_flush ( ) changes the mtime of S3 object from the local cache file or the tmpfile . The s3fs cuts some requests, because the s3fs can always check mtime whether the s3fs uses or does not use the local cache file. 5) A case of no "x-amz-meta-mode" If the object did not have "x-amz-meta-mtime" mete, the s3fs recognized the file as not regular file. After this version, the s3fs recognizes the file as regular file. 6) "." and ".." directory The s3fs_readdir() did not return "X" and "XX" directory name. After this version, the s3fs is changed that it returns "X" and "XX". Example, the result of "ls" lists "X" and "XX" directory. 7) Fixed a bug The insert_object() had a bug, and it is fixed. git-svn-id: http://s3fs.googlecode.com/svn/trunk@390 df820570-a93a-0410-bd06-b72b767a4274
2013-02-24 08:58:54 +00:00
return 0;
}
if(0 == strcmp(arg, "enable_noobj_cache")){
StatCache::getStatCacheData()->EnableCacheNoObject();
return 0;
}
if(0 == strcmp(arg, "nodnscache")){
S3fsCurl::SetDnsCache(false);
return 0;
}
if(0 == strcmp(arg, "nosscache")){
S3fsCurl::SetSslSessionCache(false);
return 0;
}
if(0 == STR2NCMP(arg, "parallel_count=") || 0 == STR2NCMP(arg, "parallel_upload=")){
int maxpara = static_cast<int>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
if(0 >= maxpara){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
fprintf(stderr, "%s: argument should be over 1: parallel_count\n",
program_name.c_str());
return -1;
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
S3fsCurl::SetMaxParallelCount(maxpara);
2014-03-30 07:53:41 +00:00
if(FdManager::GetPageSize() < static_cast<size_t>(S3fsCurl::GetMultipartSize() * S3fsCurl::GetMaxParallelCount())){
FdManager::SetPageSize(static_cast<size_t>(S3fsCurl::GetMultipartSize() * S3fsCurl::GetMaxParallelCount()));
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
return 0;
}
if(0 == STR2NCMP(arg, "fd_page_size=")){
size_t pagesize = static_cast<size_t>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
2014-03-30 07:53:41 +00:00
if(pagesize < static_cast<size_t>(S3fsCurl::GetMultipartSize() * S3fsCurl::GetMaxParallelCount())){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
fprintf(stderr, "%s: argument should be over 1MB: fd_page_size\n",
program_name.c_str());
return -1;
}
FdManager::SetPageSize(pagesize);
return 0;
}
2014-03-30 07:53:41 +00:00
if(0 == STR2NCMP(arg, "multipart_size=")){
off_t size = static_cast<off_t>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char)));
if(!S3fsCurl::SetMultipartSize(size)){
fprintf(stderr, "%s: multipart_size option must be at least 10 MB\n", program_name.c_str());
2014-03-30 07:53:41 +00:00
return -1;
}
if(FdManager::GetPageSize() < static_cast<size_t>(S3fsCurl::GetMultipartSize() * S3fsCurl::GetMaxParallelCount())){
FdManager::SetPageSize(static_cast<size_t>(S3fsCurl::GetMultipartSize() * S3fsCurl::GetMaxParallelCount()));
}
return 0;
}
if(0 == STR2NCMP(arg, "singlepart_copy_limit=")){
singlepart_copy_limit = static_cast<int64_t>(s3fs_strtoofft(strchr(arg, '=') + sizeof(char))) * 1024;
return 0;
}
if(0 == STR2NCMP(arg, "ahbe_conf=")){
Fixed Issue 229 and Changes codes 1) Set metadata "Content-Encoding" automatically(Issue 292) For this issue, s3fs is added new option "ahbe_conf". New option means the configuration file path, and this file specifies additional HTTP header by file(object) extension. Thus you can specify any HTTP header for each object by extension. * ahbe_conf file format: ----------- line = [file suffix] HTTP-header [HTTP-header-values] file suffix = file(object) suffix, if this field is empty, it means "*"(all object). HTTP-header = additional HTTP header name HTTP-header-values = additional HTTP header value ----------- * Example: ----------- .gz Content-Encoding gzip .Z Content-Encoding compress X-S3FS-MYHTTPHEAD myvalue ----------- A sample configuration file is uploaded in "test" directory. If ahbe_conf parameter is specified, s3fs loads it's configuration and compares extension(suffix) of object(file) when uploading (PUT/POST) it. If the extension is same, s3fs adds/sends specified HTTP header and value. A case of sample configuration file, if a object(it's extension is ".gz") which already has Content-Encoding HTTP header is renamed to ".txt" extension, s3fs does not set Content-Encoding. Because ".txt" is not match any line in configuration file. So, s3fs matches the extension by each PUT/POST action. * Please take care about "Content-Encoding". This new option allows setting ANY HTTP header by object extension. For example, you can specify "Content-Encoding" for ".gz"/etc extension in configuration. But this means that S3 always returns "Content-Encoding: gzip" when a client requests with other "Accept-Encoding:" header. It SHOULD NOT be good. Please see RFC 2616. 2) Changes about allow_other/uid/gid option for mount point I reviewed about mount point permission and allow_other/uid/gid options, and found bugs about these. s3fs is fixed bugs and changed to the following specifications. * s3fs only allows uid(gid) options as 0(root), when the effective user is zero(root). * A mount point(directory) must have a permission to allow accessing by effective user/group. * If allow_other option is specified, the mount point permission is set 0777(all users allow all access). In another case, the mount point is set 0700(only allows effective user). * When uid/gid option is specified, the mount point owner/group is set uid/gid option value. If uid/gid is not set, it is set effective user/group id. This changes maybe fixes some issue(321, 338). 3) Changes a logic about (Issue 229) The chmod command returns -EIO when changing the mount point. It is correct, s3fs can not changed owner/group/mtime for the mount point, but s3fs sends a request for changing the bucket. This revision does not send the request, and returns EIO as soon as possible. git-svn-id: http://s3fs.googlecode.com/svn/trunk@465 df820570-a93a-0410-bd06-b72b767a4274
2013-08-16 19:24:01 +00:00
string ahbe_conf = strchr(arg, '=') + sizeof(char);
if(!AdditionalHeader::get()->Load(ahbe_conf.c_str())){
fprintf(stderr, "%s: failed to load ahbe_conf file(%s).\n",
program_name.c_str(), ahbe_conf.c_str());
return -1;
}
AdditionalHeader::get()->Dump();
return 0;
}
if(0 == strcmp(arg, "noxmlns")){
noxmlns = true;
return 0;
}
if(0 == strcmp(arg, "nocopyapi")){
nocopyapi = true;
return 0;
}
if(0 == strcmp(arg, "norenameapi")){
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
norenameapi = true;
return 0;
}
if(0 == strcmp(arg, "enable_content_md5")){
S3fsCurl::SetContentMd5(true);
return 0;
}
if(0 == STR2NCMP(arg, "url=")){
host = strchr(arg, '=') + sizeof(char);
// strip the trailing '/', if any, off the end of the host
// string
size_t found, length;
found = host.find_last_of('/');
length = host.length();
while(found == (length - 1) && length > 0){
host.erase(found);
found = host.find_last_of('/');
length = host.length();
}
return 0;
}
if(0 == strcmp(arg, "sigv2")){
S3fsCurl::SetSignatureV4(false);
return 0;
}
if(0 == strcmp(arg, "createbucket")){
create_bucket = true;
return 0;
}
if(0 == STR2NCMP(arg, "endpoint=")){
endpoint = strchr(arg, '=') + sizeof(char);
is_specified_endpoint = true;
return 0;
}
if(0 == strcmp(arg, "use_path_request_style")){
pathrequeststyle = true;
return 0;
}
// debug option
//
// The first -d (or --debug) enables s3fs debug
// the second -d option is passed to fuse to turn on its
// debug output
if(0 == strcmp(arg, "-d") || 0 == strcmp(arg, "--debug")){
if(!debug){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
debug = true;
return 0;
}else{
// fuse doesn't understand "--debug", but it
// understands -d, but we can't pass -d back
// to fuse, in this case just ignore the
// second --debug if is was provided. If we
// do not ignore this, fuse emits an error
if(strcmp(arg, "--debug") == 0){
return 0;
}
}
}
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
// for deep debugging message
if(0 == strcmp(arg, "f2")){
Changes codes for performance(part 3) * Summay This revision includes big change about temporary file and local cache file. By this big change, s3fs works with good performance when s3fs opens/ closes/syncs/reads object. I made a big change about the handling about temporary file and local cache file to do this implementation. * Detail 1) About temporary file(local file) s3fs uses a temporary file on local file system when s3fs does download/ upload/open/seek object on S3. After this revision, s3fs calls ftruncate() function when s3fs makes the temporary file. In this way s3fs can set a file size of precisely length without downloading. (Notice - ftruncate function is for XSI-compliant systems, so that possibly you have a problem on non-XSI-compliant systems.) By this change, s3fs can download a part of a object by requesting with "Range" http header. It seems like downloading by each block unit. The default block(part) size is 50MB, it is caused the result which is default parallel requests count(5) by default multipart upload size(10MB). If you need to change this block size, you can change by new option "fd_page_size". This option can take from 1MB(1024 * 1024) to any bytes. So that, you have to take care about that fdcache.cpp(and fdcache.h) were changed a lot. 2) About local cache Local cache files which are in directory specified by "use_cache" option do not have always all of object data. This cause is that s3fs uses ftruncate function and reads(writes) each block unit of a temporary file. s3fs manages each block unit's status which are "downloaded area" or "not". For this status, s3fs makes new temporary file in cache directory which is specified by "use_cache" option. This status files is in a directory which is named "<use_cache sirectory>/.<bucket_name>/". When s3fs opens this status file, s3fs locks this file for exclusive control by calling flock function. You need to take care about this, the status files can not be laid on network drive(like NFS). This revision changes about file open mode, s3fs always opens a local cache file and each status file with writable mode. Last, this revision adds new option "del_cache", this option means that s3fs deletes all local cache file when s3fs starts and exits. 3) Uploading When s3fs writes data to file descriptor through FUSE request, old s3fs revision downloads all of the object. But new revision does not download all, it downloads only small percial area(some block units) including writing data area. And when s3fs closes or flushes the file descriptor, s3fs downloads other area which is not downloaded from server. After that, s3fs uploads all of data. Already r456 revision has parallel upload function, then this revision with r456 and r457 are very big change for performance. 4) Downloading By changing a temporary file and a local cache file, when s3fs downloads a object, it downloads only the required range(some block units). And s3fs downloads units by parallel GET request, it is same as a case of uploading. (Maximum parallel request count and each download size are specified same parameters for uploading.) In the new revision, when s3fs opens file, s3fs returns file descriptor soon. Because s3fs only opens(makes) the file descriptor with no downloading data. And when s3fs reads a data, s3fs downloads only some block unit including specified area. This result is good for performance. 5) Changes option name The option "parallel_upload" which added at r456 is changed to new option name as "parallel_count". This reason is this option value is not only used by uploading object, but a uploading object also uses this option. (For a while, you can use old option name "parallel_upload" for compatibility.) git-svn-id: http://s3fs.googlecode.com/svn/trunk@458 df820570-a93a-0410-bd06-b72b767a4274
2013-07-23 16:01:48 +00:00
foreground2 = true;
return 0;
}
if(0 == strcmp(arg, "curldbg")){
S3fsCurl::SetVerbose(true);
return 0;
}
if(0 == STR2NCMP(arg, "accessKeyId=")){
fprintf(stderr, "%s: option accessKeyId is no longer supported\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
if(0 == STR2NCMP(arg, "secretAccessKey=")){
fprintf(stderr, "%s: option secretAccessKey is no longer supported\n",
program_name.c_str());
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
return -1;
}
}
return 1;
}
int main(int argc, char* argv[])
{
int ch;
int fuse_res;
int option_index = 0;
struct fuse_operations s3fs_oper;
static const struct option long_opts[] = {
{"help", no_argument, NULL, 'h'},
{"version", no_argument, 0, 0},
{"debug", no_argument, NULL, 'd'},
{0, 0, 0, 0}
};
// init xml2
xmlInitParser();
LIBXML_TEST_VERSION
// get progam name - emulate basename
size_t found = string::npos;
program_name.assign(argv[0]);
found = program_name.find_last_of("/");
if(found != string::npos){
program_name.replace(0, found+1, "");
}
while((ch = getopt_long(argc, argv, "dho:fsu", long_opts, &option_index)) != -1){
switch(ch){
case 0:
if(strcmp(long_opts[option_index].name, "version") == 0){
show_version();
exit(EXIT_SUCCESS);
}
break;
case 'h':
show_help();
exit(EXIT_SUCCESS);
case 'o':
break;
case 'd':
break;
case 'f':
foreground = true;
break;
case 's':
break;
case 'u':
utility_mode = 1;
break;
default:
exit(EXIT_FAILURE);
}
}
2014-08-26 17:11:10 +00:00
// Load SSE-C Key from env
S3fsCurl::LoadEnvSseKeys();
// clear this structure
memset(&s3fs_oper, 0, sizeof(s3fs_oper));
// This is the fuse-style parser for the arguments
// after which the bucket name and mountpoint names
// should have been set
struct fuse_args custom_args = FUSE_ARGS_INIT(argc, argv);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
if(0 != fuse_opt_parse(&custom_args, NULL, NULL, my_fuse_opt_proc)){
exit(EXIT_FAILURE);
}
// The first plain argument is the bucket
if(bucket.size() == 0){
fprintf(stderr, "%s: missing BUCKET argument\n", program_name.c_str());
show_usage();
exit(EXIT_FAILURE);
}
// bucket names cannot contain upper case characters in virtual-hosted style
if((!pathrequeststyle) && (lower(bucket) != bucket)){
fprintf(stderr, "%s: BUCKET %s, name not compatible with virtual-hosted style\n",
program_name.c_str(), bucket.c_str());
exit(EXIT_FAILURE);
}
// check bucket name for illegal characters
found = bucket.find_first_of("/:\\;!@#$%^&*?|+=");
if(found != string::npos){
fprintf(stderr, "%s: BUCKET %s -- bucket name contains an illegal character\n",
program_name.c_str(), bucket.c_str());
exit(EXIT_FAILURE);
}
// The second plain argument is the mountpoint
// if the option was given, we all ready checked for a
// readable, non-empty directory, this checks determines
// if the mountpoint option was ever supplied
if(utility_mode == 0){
if(mountpoint.size() == 0){
fprintf(stderr, "%s: missing MOUNTPOINT argument\n", program_name.c_str());
show_usage();
exit(EXIT_FAILURE);
}
}
2014-10-01 10:42:39 +00:00
// error checking of command line arguments for compatibility
if(S3fsCurl::IsPublicBucket() && S3fsCurl::IsSetAccessKeyId()){
fprintf(stderr, "%s: specifying both public_bucket and the access keys options is invalid\n",
program_name.c_str());
exit(EXIT_FAILURE);
}
if(passwd_file.size() > 0 && S3fsCurl::IsSetAccessKeyId()){
fprintf(stderr, "%s: specifying both passwd_file and the access keys options is invalid\n",
program_name.c_str());
exit(EXIT_FAILURE);
}
if(!S3fsCurl::IsPublicBucket()){
if(EXIT_SUCCESS != get_access_keys()){
exit(EXIT_FAILURE);
}
if(!S3fsCurl::IsSetAccessKeyId()){
fprintf(stderr, "%s: could not establish security credentials, check documentation\n",
program_name.c_str());
exit(EXIT_FAILURE);
}
// More error checking on the access key pair can be done
// like checking for appropriate lengths and characters
}
// check cache dir permission
if(!FdManager::CheckCacheTopDir() || !CacheFileStat::CheckCacheFileStatTopDir()){
fprintf(stderr, "%s: could not allow cache directory permission, check permission of cache directories.\n",
program_name.c_str());
exit(EXIT_FAILURE);
}
// There's room for more command line error checking
// Check to see if the bucket name contains periods and https (SSL) is
// being used. This is a known limitation:
// http://docs.amazonwebservices.com/AmazonS3/latest/dev/
// The Developers Guide suggests that either use HTTP of for us to write
// our own certificate verification logic.
// For now, this will be unsupported unless we get a request for it to
// be supported. In that case, we have a couple of options:
// - implement a command line option that bypasses the verify host
// but doesn't bypass verifying the certificate
// - write our own host verification (this might be complex)
// See issue #128strncasecmp
/*
if(1 == S3fsCurl::GetSslVerifyHostname()){
found = bucket.find_first_of(".");
if(found != string::npos){
found = host.find("https:");
if(found != string::npos){
fprintf(stderr, "%s: Using https and a bucket name with periods is unsupported.\n",
program_name.c_str());
exit(1);
}
}
}
*/
if(utility_mode){
exit(s3fs_utility_mode());
}
s3fs_oper.getattr = s3fs_getattr;
s3fs_oper.readlink = s3fs_readlink;
s3fs_oper.mknod = s3fs_mknod;
s3fs_oper.mkdir = s3fs_mkdir;
s3fs_oper.unlink = s3fs_unlink;
s3fs_oper.rmdir = s3fs_rmdir;
s3fs_oper.symlink = s3fs_symlink;
s3fs_oper.rename = s3fs_rename;
s3fs_oper.link = s3fs_link;
if(!nocopyapi){
s3fs_oper.chmod = s3fs_chmod;
s3fs_oper.chown = s3fs_chown;
s3fs_oper.utimens = s3fs_utimens;
}else{
s3fs_oper.chmod = s3fs_chmod_nocopy;
s3fs_oper.chown = s3fs_chown_nocopy;
s3fs_oper.utimens = s3fs_utimens_nocopy;
}
s3fs_oper.truncate = s3fs_truncate;
s3fs_oper.open = s3fs_open;
s3fs_oper.read = s3fs_read;
s3fs_oper.write = s3fs_write;
s3fs_oper.statfs = s3fs_statfs;
s3fs_oper.flush = s3fs_flush;
2015-03-10 16:18:03 +00:00
s3fs_oper.fsync = s3fs_fsync;
s3fs_oper.release = s3fs_release;
s3fs_oper.opendir = s3fs_opendir;
s3fs_oper.readdir = s3fs_readdir;
s3fs_oper.init = s3fs_init;
s3fs_oper.destroy = s3fs_destroy;
s3fs_oper.access = s3fs_access;
s3fs_oper.create = s3fs_create;
// extended attributes
s3fs_oper.setxattr = s3fs_setxattr;
s3fs_oper.getxattr = s3fs_getxattr;
s3fs_oper.listxattr = s3fs_listxattr;
s3fs_oper.removexattr = s3fs_removexattr;
if(!s3fs_init_global_ssl()){
fprintf(stderr, "%s: could not initialize for ssl libraries.\n", program_name.c_str());
exit(EXIT_FAILURE);
}
// now passing things off to fuse, fuse will finish evaluating the command line args
fuse_res = fuse_main(custom_args.argc, custom_args.argv, &s3fs_oper, NULL);
fuse_opt_free_args(&custom_args);
s3fs_destroy_global_ssl();
// cleanup xml2
xmlCleanupParser();
S3FS_MALLOCTRIM(0);
Summary of Changes(1.63 -> 1.64) * This new version was made for fixing big issue about directory object. Please be careful and review new s3fs. ========================== List of Changes ========================== 1) Fixed bugs Fixed some memory leak and un-freed curl handle. Fixed codes with a bug which is not found yet. Fixed a bug that the s3fs could not update object's mtime when the s3fs had a opened file descriptor. Please let us know a bug, when you find new bug of a memory leak. 2) Changed codes Changed codes of s3fs_readdir() and list_bucket() etc. Changed codes so that the get_realpath() function returned std::string. Changed codes about exit() function. Because the exit() function is called from many fuse callback function directly, these function called fuse_exit() function and retuned with error. Changed codes so that the case of the characters for the "x-amz-meta" response header is ignored. 3) Added a option Added the norenameapi option for the storage compatible with S3 without copy API. This option is subset of nocopyapi option. Please read man page or call with --help option. 4) Object for directory This is very big and important change. The object of directory is changed "dir/" instead of "dir" for being compatible with other S3 client applications. And this version understands the object of directory which is made by old version. If the new s3fs changes the attributes or owner/group or mtime of the directory object, the s3fs automatically changes the object from old object name("dir") to new("dir/"). If you need to change old object name("dir") to new("dir/") manually, you can use shell script(mergedir.sh) in test directory. * About the directory object name AWS S3 allows the object name as both "dir" and "dir/". The s3fs before this version understood only "dir" as directory object name, but old version did not understand the "dir/" object name. The new version understands both of "dir" and "dir/" object name. The s3fs user needs to be care for the special situation that I mentioned later. The new version deletes old "dir" object and makes new "dir/" object, when the user operates the directory object for changing the permission or owner/group or mtime. This operation does on background and automatically. If you need to merge manually, you can use shell script which is mergedir.sh in test directory. This script runs chmod/chown/touch commands after finding a directory. Other S3 client application makes a directory object("dir/") without meta information which is needed to understand by the s3fs, this script can add meta information for a directory object. If this script function is insufficient for you, you can read and modify the codes by yourself. Please use the shell script carefully because of changing the object. If you find a bug in this script, please let me know. * Details ** The directory object made by old version The directory object made by old version is not understood by other S3 client application. New s3fs version was updated for keeping compatibility with other clients. You can use the mergedir.sh in test directory for merging from old directory object("dir") to new("dir/"). The directory object name is changed from "dir" to "dir/" after the mergedir.sh is run, this changed "dir/" object is understood by other S3 clients. This script runs chmod/chown/chgrp/touch/etc commands against the old directory object("dir"), then new s3fs merges that directory automatically. If you need to change directory object from old to new manually, you can do it by running these commands which change the directory attributes(mode/owner/group/mtime). ** The directory object made by new version The directory object name made by new version is "dir/". Because the name includes "/", other S3 client applications understand it as the directory. I tested new directory by s3cmd/tntDrive/DragonDisk/Gladinet as other S3 clients, the result was good compatibility. You need to know that the compatibility has small problem by the difference in specifications between clients. And you need to be careful about that the old s3fs can not understand the directory object which made by new s3fs. You should change all s3fs which accesses same bucket. ** The directory object made by other S3 client application Because the object is determined as a directory by the s3fs, the s3fs makes and uses special meta information which is "x-amz-meta-***" and "Content-Type" as HTTP header. The s3fs sets and uses HTTP headers for the directory object, those headers are listed below. Content-Type: application/x-directory x-amz-meta-mode: <mode> x-amz-meta-uid: <UID> x-amz-meta-gid <GID> x-amz-meta-mtime: <unix time of modified file> Other S3 client application builds the directory object without attributes which is needed by the s3fs. When the "ls" command is run on the s3fs-fuse file system which has directories/files made by other S3 clients, this result is shown below. d--------- 1 root root 0 Feb 27 11:21 dir ---------- 1 root root 1024 Mar 14 02:15 file Because the objects don't have meta information("x-amz-meta-mode"), it means mode=0000. In this case, the directory object is shown only "d", because the s3fs determines the object as a directory when the object is the name with "/" or has "Content-type: application/x-directory" header. (The s3fs sets "Content-Type: application/x-directory" to the directory object, but other S3 clients set "binary/octet-stream".) In that result, nobody without root is allowed to operate the object. The owner and group are "root"(UID=0) because the object doesn't have "x-amz-meta-uid/gid". If the object doesn't have "x-amz-meta-mtime", the s3fs uses "Last-Modified" HTTP header. Therefore the object's mtime is "Last-Modified" value.(This logic is same as old version) It has been already explained, if you need to change the object attributes, you can do it by manually operation or mergedir.sh. * Example of the compatibility with s3cmd etc ** Case A) Only "dir/file" object One of case, there is only "dir/file" object without "dir/" object, that object is made by s3cmd or etc. In this case, the response of REST API(list bucket) with "delimiter=/" parameter has "CommonPrefixes", and the "dir/" is listed in "CommonPrefixes/Prefix", but the "dir/" object is not real object. The s3fs needs to determine this object as directory, however there is no real directory object("dir" or "dir/"). But both new s3fs and old one does NOT understand this "dir/" in "CommonPrefixes", because the s3fs fails to get meta information from "dir" or "dir/". On this case, the result of "ls" command is shown below. ??????????? ? ? ? ? ? dir This "dir" is not operated by anyone and any process, because the s3fs does not understand this object permission. And "dir/file" object can not be shown and operated too. Some other S3 clients(tntDrive/Gladinet/etc) can not understand this object as same as the s3fs. If you need to operate "dir/file" object, you need to make the "dir/" object as a directory. To make the "dir/" directory object, you need to do below. Because there is already the "dir" object which is not real object, you can not make "dir/" directory. (s3cmd does not make "dir/" object because the object name has "/".). You should make another name directory(ex: "dir2/"), and move the "dir/file" objects to in new directory. Last, you can rename the directory name from "dir2/" to "dir/". ** Case B) Both "dir" and "dir/file" object This case is that there are "dir" and "dir/file" objects which were made by s3cmd/etc. s3cmd and s3fs understand the "dir" object as normal(file) object because this object does not have meta information and a name with "/". But the result of REST API(list bucket) has "dir/" name in "CommonPrefixes/Prefix". The s3fs checks "dir/" and "dir" as a directory, but the "dir" object is not directory object. (Because the new s3fs need to compatible old version, the s3fs checks a directory object in order of "dir/", "dir") In this case, the result of "ls" command is shown below. ---------- 1 root root 0 Feb 27 02:48 dir As a result, the "dir/file" can not be shown and operated because the "dir" object is a file. If you determine the "dir" as a directory, you need to add mete information to the "dir" object by s3cmd. ** Case C) Both "dir" and "dir/" object Last case is that there are "dir" and "dir/" objects which were made by other S3 clients. (example: At first you upload a object "dir/" as a directory by new 3sfs, and you upload a object "dir" by s3cmd.) New s3fs determines "dir/" as a directory, because the s3fs searches in oder of "dir/", "dir". As a result, the "dir" object can not be shown and operated. ** Compatibility between S3 clients Both new and old s3fs do not understand both "dir" and "dir/" at the same time, tntDrive and Galdinet are same as the s3fs. If there are "dir/" and "dir" objects, the s3fs gives priority to "dir/". But s3cmd and DragonDisk understand both objects. git-svn-id: http://s3fs.googlecode.com/svn/trunk@392 df820570-a93a-0410-bd06-b72b767a4274
2013-03-23 14:04:07 +00:00
exit(fuse_res);
}
2014-09-07 15:08:27 +00:00
/*
* Local variables:
* tab-width: 4
* c-basic-offset: 4
* End:
* vim600: noet sw=4 ts=4 fdm=marker
* vim<600: noet sw=4 ts=4
*/