Currently the maximum delay is always derived automatically from the
initial delay. This is fine in most cases, but for some use cases (large
files that take a long time to write) we need to be able to set a longer
max delay than the computed value (e.g., 15s delay with 10min timeout).
This adds a small package `geoip` which knows how to download and manage
the Maxmind GeoLite2 database we use. This removes the need for various
scripts to download and manage the geoip database, something that today
happens on Docker startup for the relay pool server and using various
hand written hacks for the usage reporting server.
The database is downloaded when needed and then refreshed on a
best-effort basis weekly.
### Purpose
Adds a new metric `syncthing_connections_active` which equals to the
amount of active connections per device.
Fixes#9527
<!--
Describe the purpose of this change. If there is an existing issue that
is
resolved by this pull request, ensure that the commit subject is on the
form
`Some short description (fixes#1234)` where 1234 is the issue number.
-->
### Testing
I've manually tested it by running syncthing with these changes locally
and examining the returned metrics from `/metrics`.
I've done the following things:
- Connect & disconnect a device
- Increase & decrease the number of connections and verify that the
value of the metric matches with the amount displayed in the GUI.
### Documentation
https://github.com/syncthing/docs/blob/main/includes/metrics-list.rst
needs to be regenerated with
[find-metrics.go](https://github.com/syncthing/docs/blob/main/_script/find-metrics/find-metrics.go)
## Authorship
Your name and email will be added automatically to the AUTHORS file
based on the commit metadata.
---------
Co-authored-by: Jakob Borg <jakob@kastelo.net>
### Purpose
Resend our indexes since we fixed that index-sending issue.
I made a new thing to only drop the non-local-device index IDs, i.e.,
those for other devices. This means we will see a mismatch and resend
all indexes, but they will not. This is somewhat cleaner as it avoids
resending everything twice when two devices are upgraded, and in any
case, we have no reason to force a resend of incoming indexes here.
### Testing
It happens on my computer...
This is an extract from PR #9175, which can be reviewed in isolation to
reduce the volume of changes to review all at once in #9175. There are
about to be several services and API handlers that read and set cookies
and session state, so this abstraction will prove helpful.
In particular a motivating cause for this is that with the current
architecture in PR #9175, in `api.go` the [`webauthnService` needs to
access the
session](https://github.com/syncthing/syncthing/pull/9175/files#diff-e2e14f22d818b8e635572ef0ee7718dee875c365e07225d760a6faae8be7772dR309-R310)
for authentication purposes but needs to be instantiated before the
`configMuxBuilder` for config purposes, because the WebAuthn additions
to config management need to perform WebAuthn registration ceremonies,
but currently the session management is embedded in the
`basicAuthAndSessionMiddleware` which is [instantiated much
later](https://github.com/syncthing/syncthing/pull/9175/files#diff-e2e14f22d818b8e635572ef0ee7718dee875c365e07225d760a6faae8be7772dL371-R380)
and only if authentication is enabled in `guiCfg`. This refactorization
extracts the session management out from `basicAuthAndSessionMiddleware`
so that `basicAuthAndSessionMiddleware` and `webauthnService` can both
use the same shared session management service to perform session
management logic.
### Testing
This is a refactorization intended to not change any externally
observable behaviour, so existing tests (e.g., `api_auth_test.go`)
should cover this where appropriate. I have manually verified that:
- Appending `+ "foo"` to the cookie name in `createSession` causes
`TestHtmlFormLogin/invalid_URL_returns_403_before_auth_and_404_after_auth`
and `TestHtmlFormLogin/UTF-8_auth_works` to fail
- Inverting the return value of `hasValidSession` cases a whole bunch of
tests in `TestHTTPLogin` and `TestHtmlFormLogin` to fail
- (Fixed) Changing the cookie to `MaxAge: 1000` in `destroySession` does
NOT cause any tests to fail!
- Added tests `TestHtmlFormLogin/Logout_removes_the_session_cookie`,
`TestHTTPLogin/*/Logout_removes_the_session_cookie`,
`TestHtmlFormLogin/Session_cookie_is_invalid_after_logout` and
`TestHTTPLogin/200_path#01/Session_cookie_is_invalid_after_logout` to
cover this.
- Manually verified that these tests pass both before and after the
changes in this PR, and that changing the cookie to `MaxAge: 1000` or
not calling `m.tokens.Delete(cookie.Value)` in `destroySession` makes
the respective pair of tests fail.
Adds a bool flag to `scanIfItemChanged()` to indicate when the scan was initiated from a delete function, and if so, tell `IsEquivalentOptional()` to ignore Xattrs and Ownership regardless of the global setting.
I tested this with my sledgehammer and it seems to pass.
Somewhere along the way, the non-parallel test became parallel, and at
that point, timeouts occurred. Parallel is better, so increase the
timeout on the offending call a bit...
The new test has a flakiness factor on slow platforms, where the close
on the sending connection races with the last index message, potentially
messing up the count. This adds a wait to ensure that all sent messages
are received, or the test will eventually fail with a timeout.
Currently `IsTraced("xyz")` will return true for
any inclusion of "xyz" in string.
This change splits `STTRACE` using `','`, `' '` and `';'`
as delimiters. That makes facilities separation
more clear.
This makes a couple of small improvements to the folder summary
mechanism:
- The folder summary includes the local and remote sequence numbers in
clear text, rather than some odd sum that I'm not sure what it was
intended to represent.
- The folder summary event is generated when appropriate, regardless of
whether there is an event listener. We did this before because
generating it was expensive, and we wanted to avoid doing it
unnecessarily. Nowadays, however, it's mostly just reading out
pre-calculated metadata, and anyway, it's nice if it shows up reliably
when running with -verbose.
The point of all this is to make it easier to use these events to judge
when devices are, in fact, in sync. As-is, if I'm looking at two
devices, it's very difficult to reliably determine if they are in sync
or not. The reason is that while we can ask device A if it thinks it's
in sync, we can't see if the answer is "yes" because it has processed
all changes from B, or if it just doesn't know about the changes from B
yet. With proper sequence numbers in the event we can compare the two
and determine the truth. This makes testing a lot easier.
This is a refactor of the protocol/model interface to take the actual
message as the parameter, instead of the broken-out fields:
```diff
type Model interface {
// An index was received from the peer device
- Index(conn Connection, folder string, files []FileInfo) error
+ Index(conn Connection, idx *Index) error
// An index update was received from the peer device
- IndexUpdate(conn Connection, folder string, files []FileInfo) error
+ IndexUpdate(conn Connection, idxUp *IndexUpdate) error
// A request was made by the peer device
- Request(conn Connection, folder, name string, blockNo, size int32, offset int64, hash []byte, weakHash uint32, fromTemporary bool) (RequestResponse, error)
+ Request(conn Connection, req *Request) (RequestResponse, error)
// A cluster configuration message was received
- ClusterConfig(conn Connection, config ClusterConfig) error
+ ClusterConfig(conn Connection, config *ClusterConfig) error
// The peer device closed the connection or an error occurred
Closed(conn Connection, err error)
// The peer device sent progress updates for the files it is currently downloading
- DownloadProgress(conn Connection, folder string, updates []FileDownloadProgressUpdate) error
+ DownloadProgress(conn Connection, p *DownloadProgress) error
}
```
(and changing the `ClusterConfig` to `*ClusterConfig` for symmetry;
we'll be forced to use all pointers everywhere at some point anyway...)
The reason for this is that I have another thing cooking which is a
small troubleshooting change to check index consistency during transfer.
This required adding a field or two to the index/indexupdate messages,
and plumbing the extra parameters in umpteen changes is almost as big a
diff as this is. I figured let's do it once and avoid having to do that
in the future again...
The rest of the diff falls out of the change above, much of it being in
test code where we run these methods manually...
This improves the ignore handling so that directories can be fully
ignored (skipped in the watcher) in more cases. Specifically, where the
previous rule was that any complex `!`-pattern would disable skipping
directories, the new rule is that only matches on patterns *after* such
a `!`-pattern disable skipping. That is, the following now does the
intuitive thing:
```
/foo
/bar
!whatever
*
```
- `/foo/**` and `/bar/**` are completely skipped, since there is no
chance anything underneath them could ever be not-ignored
- `!whatever` toggles the "can't skip directories any more" flag
- Anything that matches `*` can't skip directories, because it's
possible we can have `whatever` match something deeper.
To enable this, some refactoring was necessary:
- The "can skip dirs" flag is now a property of the match result, not of
the pattern set as a whole.
- That meant returning a boolean is not good enough, we need to actually
return the entire `Result` (or, like, two booleans but that seemed
uglier and more annoying to use)
- `ShouldIgnore(string) boolean` went away with
`Match(string).IsIgnored()` being the obvious replacement (API
simplification!)
- The watcher then needed to import the `ignore` package (for the
`Result` type), but `fs` imports the watcher and `ignore` imports `fs`.
That's a cycle, so I broke out `Result` into a package of its own so
that it can be safely imported everywhere in things like `type Matcher
interface { Match(string) result.Result }`. There's a fair amount of
stuttering in `result.Result` and maybe we should go with something like
`ignoreresult.R` or so, leaving this open for discussion.
Tests refactored to suit, I think this change is in fact quite well
covered by the existing ones...
Also some noise because a few of the changed files were quite old and
got the `gofumpt` treatment by my editor. Sorry not sorry.
---------
Co-authored-by: Simon Frei <freisim93@gmail.com>
### Purpose
Fix#9241 by expanding tildes in version paths.
When creating the versioner file system, first try to expand any leading
tildes to the user's home directory before handling relative paths. This
makes a version path `"~/p"` expand to `"$HOME/p"` instead of
`"/folder/~/p"`.
### Testing
Added a test to lib/versioner that exercises this code path. Also
manually tested with local syncthing instances.
### Purpose
This PR changes behaviour of syncthing related to `receive-only`
folders, which I believe to be a bug since I wouldn't expect the current
behaviour. With the current syncthing codebase, a file of a
`receive-only` folder that is only modified locally can cause the
creation of a `.sync-conflict` file.
### Testing
Consider this szenario: Setup two paired clients that sync a folder with
a given file (e.g. `Test.txt`). One of the clients configures the folder
to be `receive-only`. Now, change the contents of the file for the
receive-only client **_twice_**.
With the current syncthing codebase, this leads to the creation of a
`.sync-conflict` file that contains the modified contents, while the
regular `Test.txt` file is reset to the cluster's provided contents.
This is due to a `protocol.FileInfo#ShouldConflict` check, that is
succeeding on the locally modified file.
This PR changes this behaviour to not reset the file and not cause the
creation of a `.sync-conflict`. Instead, the second content update is
treated the same as the first content update.
This PR also contains a test that fails on the current codebase and
succeeds with the changes introduced in this PR.
### Screenshots
This is not a GUI change
### Documentation
This is not a user visible change.
## Authorship
Your name and email will be added automatically to the AUTHORS file
based on the commit metadata.
#### Thanks to all the syncthing folks for this awesome piece of
software!
This adds a "token manager" which handles storing and checking expired
tokens, used for both sessions and CSRF tokens. It removes the old,
corresponding functionality for CSRFs which saved things in a file. The
result is less crap in the state directory, and active login sessions
now survive a Syncthing restart (this really annoyed me).
It also adds a boolean on login to create a longer-lived session cookie,
which is now possible and useful. Thus we can remain logged in over
browser restarts, which was also annoying... :)
<img width="1001" alt="Screenshot 2023-12-12 at 09 56 34"
src="https://github.com/syncthing/syncthing/assets/125426/55cb20c8-78fc-453e-825d-655b94c8623b">
Best viewed with whitespace-insensitive diff, as a bunch of the auth
functions became methods instead of closures which changed indentation.
If syncOwnership is enabled and the remote uses for example a dockerized
Syncthing it can't fetch the ownername and groupname of the local
instance. Without this patch this led to an endless cycle of detected
changes on the remote and failing re-sync attempts.
This patch skips comparing the ownername and groupname if they zare empty
on one side.
See https://github.com/syncthing/syncthing/issues/9039 for details.
### Testing
Proposed by @calmh in
https://github.com/syncthing/syncthing/issues/9039#issuecomment-1870584783
and tested locally in my setup,
Setup PC 1:
- Syncthing is run in Docker as user `root` and has none of the users
configured that synchronize their files
Setup PC 2:
- this PC has all users locally setup
- Syncthing runs as `systemd` service as user `syncthing` and has
multiple capabilities set to set the correct owner and permissions
Setup PC 3:
- same as PC 2
Handling:
- `PC 1` is send & receive and uses just the `UID` and `GID` identifiers
to store the files
- `PC 2` and `PC 3` synchronize their files over `PC 1` but not directly
to each other
Outcome:
- `PC 2` and `PC 3` should send and receive their files with the correct
ownership and groups from `PC 1`
### Purpose
Instead of hardcoding `SigningKey` as text use `go:embed`. Fixes#9247.
### Testing
* Building syncthing
* Trying to upgrade (signature verification)
On kqueue-systems, folders listen for folder summaries to (be able to)
warn for potential high resource usage. However, it listened for any
folder summary and not for the summary which matches the folder it's
about. This could cause that an unwatched folder causes a folder summary
containing more files than the threshold (10k), and the listening folder
(with the watcher enabled) triggers the warning.
This makes sure that only the folder summaries which are relevant to the
specific folder are being handled.
### Testing
- Fire up some kqueue-system (freebsd, I used).
- add folder A, disable the watcher, add 10001 files
- add folder B with the watcher enabled, no files are needed here
Before the change:
- add an item to folder A, trigger a rescan to speed up the process
- wait some seconds...warning triggered by folder B's
summarySubscription
After the change:
- Only a warning is triggered if the received folder summary matches the
folder which listens for the summaries
Cleanup after #9275.
This renames `fmut` -> `mut`, removes the deadlock detector and
associated plumbing, renames some things from `...PRLocked` to
`...RLocked` and similar, and updates comments.
Apart from the removal of the deadlock detection machinery, no
functional code changes... i.e. almost 100% diff noise, have fun
reviewing.
I'm tired of the fmut/pmut shenanigans. This consolidates both under one
lock; I'm not convinced there are any significant performance
differences with this approach since we're literally just protecting map
juggling...
- The locking goes away when we were already under an appropriate fmut
lock.
- Where we had fmut.RLock()+pmut.Lock() it gets upgraded to an
fmut.Lock().
- Otherwise s/pmut/fmut/.
In order to avoid diff noise for an important change I did not do the
following cleanups, which will be filed in a PR after this one, if
accepted:
- Renaming fmut to just mut
- Renaming methods that refer to being "PRLocked" and stuff like that
- Removing the no longer relevant deadlock detector
- Comments referring to pmut and locking sequences...
This pull request allows syncthing to request an IPv6
[pinhole](https://en.wikipedia.org/wiki/Firewall_pinhole), addressing
issue #7406. This helps users who prefer to use IPv6 for hosting their
services or are forced to do so because of
[CGNAT](https://en.wikipedia.org/wiki/Carrier-grade_NAT). Otherwise,
such users would have to configure their firewall manually to allow
syncthing traffic to pass through while IPv4 users can use UPnP to take
care of network configuration already.
### Testing
I have tested this in a virtual machine setup with miniupnpd running on
the virtualized router. It successfully added an IPv6 pinhole when used
with IPv6 only, an IPv4 port mapping when used with IPv4 only and both
when dual-stack (IPv4 and IPv6) is used.
Automated tests could be added for SOAP responses from the router but
automatically testing this with a real network is likely infeasible.
### Documentation
https://docs.syncthing.net/users/firewall.html could be updated to
mention the fact that UPnP now works with IPv6, although this change is
more "behind the scenes".
---------
Co-authored-by: Simon Frei <freisim93@gmail.com>
Co-authored-by: bt90 <btom1990@googlemail.com>
Co-authored-by: André Colomb <github.com@andre.colomb.de>
Before introducing the service map and using it for folder runners, the
entries in folderCfgs and folderRunners for the same key/folder were
removed under a single lock. Stopping the folder happens separately
before that with just the read lock. Now with the service map stopping
the folder and removing it from the map is a single operation. And that
still happens with just a read-lock. However even with a full lock it’s
still problematic: After the folder stopped, the runner isn’t present
anymore while the folder-config still is and sais the folder isn't
paused.
The index handler in turn looks at the folder config that is not paused,
thus assumes the runner has to be present -> nil deref on the runner.
A better solution might be to push most of these fmut maps into the
folder - they anyway are needed in there. Then there's just a single
map/source of info that's necessarily consistent. That's quite a bit of
work though, and probably/likely there will be corner cases there too.
This reduces allocations, in number and in size, while getting extended
attributes. This is mostly noticable when there is a large number of new
files to scan and we're running with the default scanProgressInterval --
then a queue of files is built in-memory, and this queue includes
extended attributes as part of file metadata. (Arguable it shouldn't,
but that's a more difficult and involved change.)
With 1M files to scan, each with one extended attribute, current peak
memory usage looks like this:
Showing nodes accounting for 1425.30MB, 98.19% of 1451.64MB total
Dropped 1435 nodes (cum <= 7.26MB)
Showing top 10 nodes out of 54
flat flat% sum% cum cum%
976.56MB 67.27% 67.27% 976.56MB 67.27%
github.com/syncthing/syncthing/lib/fs.getXattr
305.44MB 21.04% 88.31% 305.44MB 21.04%
github.com/syncthing/syncthing/lib/scanner.(*walker).walk.func1
45.78MB 3.15% 91.47% 1045.23MB 72.00%
github.com/syncthing/syncthing/lib/fs.(*BasicFilesystem).GetXattr
22.89MB 1.58% 93.04% 22.89MB 1.58%
github.com/syncthing/syncthing/lib/fs.listXattr
22.89MB 1.58% 94.62% 22.89MB 1.58%
github.com/syncthing/syncthing/lib/protocol.(*PlatformData).SetXattrs
16MB 1.10% 95.72% 16.01MB 1.10%
github.com/syndtr/goleveldb/leveldb/memdb.New
After the change, it's this:
Showing nodes accounting for 502.32MB, 95.70% of 524.88MB total
Dropped 1400 nodes (cum <= 2.62MB)
Showing top 10 nodes out of 91
flat flat% sum% cum cum%
305.43MB 58.19% 58.19% 305.43MB 58.19%
github.com/syncthing/syncthing/lib/scanner.(*walker).walk.func1
45.79MB 8.72% 66.91% 68.68MB 13.09%
github.com/syncthing/syncthing/lib/fs.(*BasicFilesystem).GetXattr
32MB 6.10% 73.01% 32.01MB 6.10%
github.com/syndtr/goleveldb/leveldb/memdb.New
22.89MB 4.36% 77.37% 22.89MB 4.36%
github.com/syncthing/syncthing/lib/fs.listXattr
22.89MB 4.36% 81.73% 22.89MB 4.36%
github.com/syncthing/syncthing/lib/protocol.(*PlatformData).SetXattrs
15.35MB 2.92% 84.66% 15.36MB 2.93%
github.com/syndtr/goleveldb/leveldb/util.(*BufferPool).Get
15.28MB 2.91% 87.57% 15.28MB 2.91% strings.(*Builder).grow
(The usage for xattrs is reduced from 976 MB to 68 MB)
LastSeen for a device was only updated when they connected. This now
updates it when they disconnect, so that we remember the last time we
actually saw them. When asking the API for current stats, currently
connected devices get a last seen value of the current time.
