2
2
mirror of https://github.com/octoleo/restic.git synced 2024-11-15 17:47:21 +00:00
restic/internal/repository/repository.go

845 lines
22 KiB
Go
Raw Normal View History

package repository
2014-12-21 16:02:49 +00:00
import (
"bufio"
"bytes"
2017-06-04 09:16:55 +00:00
"context"
"encoding/json"
"fmt"
"io"
2015-07-26 19:58:03 +00:00
"os"
"sort"
2020-10-10 19:31:40 +00:00
"sync"
2014-12-21 16:02:49 +00:00
"github.com/cenkalti/backoff/v4"
"github.com/restic/chunker"
"github.com/restic/restic/internal/backend/dryrun"
"github.com/restic/restic/internal/cache"
"github.com/restic/restic/internal/crypto"
"github.com/restic/restic/internal/debug"
2017-07-23 12:21:03 +00:00
"github.com/restic/restic/internal/errors"
"github.com/restic/restic/internal/pack"
"github.com/restic/restic/internal/restic"
2020-10-10 19:31:40 +00:00
"github.com/restic/restic/internal/ui/progress"
"golang.org/x/sync/errgroup"
2014-12-21 16:02:49 +00:00
)
const MaxStreamBufferSize = 4 * 1024 * 1024
2015-05-09 21:59:58 +00:00
// Repository is used to access a repository in a backend.
type Repository struct {
2016-08-31 18:29:54 +00:00
be restic.Backend
2016-08-31 20:39:36 +00:00
cfg restic.Config
key *crypto.Key
keyName string
2015-10-12 20:34:12 +00:00
idx *MasterIndex
Cache *cache.Cache
noAutoIndexUpdate bool
treePM *packerManager
dataPM *packerManager
2014-12-21 16:02:49 +00:00
}
// New returns a new repository with backend be.
2016-08-31 18:29:54 +00:00
func New(be restic.Backend) *Repository {
repo := &Repository{
be: be,
idx: NewMasterIndex(),
dataPM: newPackerManager(be, nil),
treePM: newPackerManager(be, nil),
}
2016-03-06 12:14:06 +00:00
return repo
2014-12-21 16:02:49 +00:00
}
// DisableAutoIndexUpdate deactives the automatic finalization and upload of new
// indexes once these are full
func (r *Repository) DisableAutoIndexUpdate() {
r.noAutoIndexUpdate = true
}
2016-08-31 20:51:35 +00:00
// Config returns the repository configuration.
2016-08-31 20:39:36 +00:00
func (r *Repository) Config() restic.Config {
return r.cfg
}
// UseCache replaces the backend with the wrapped cache.
func (r *Repository) UseCache(c *cache.Cache) {
if c == nil {
return
}
debug.Log("using cache")
r.Cache = c
r.be = c.Wrap(r.be)
}
// SetDryRun sets the repo backend into dry-run mode.
func (r *Repository) SetDryRun() {
r.be = dryrun.New(r.be)
}
2014-12-21 16:02:49 +00:00
// PrefixLength returns the number of bytes required so that all prefixes of
// all IDs of type t are unique.
func (r *Repository) PrefixLength(ctx context.Context, t restic.FileType) (int, error) {
return restic.PrefixLength(ctx, r.be, t)
2014-12-21 16:02:49 +00:00
}
// LoadUnpacked loads and decrypts the file with the given type and ID, using
// the supplied buffer (which must be empty). If the buffer is nil, a new
// buffer will be allocated and returned.
func (r *Repository) LoadUnpacked(ctx context.Context, buf []byte, t restic.FileType, id restic.ID) ([]byte, error) {
if len(buf) != 0 {
panic("buf is not empty")
}
2018-01-25 19:49:41 +00:00
debug.Log("load %v with id %v", t, id)
if t == restic.ConfigFile {
id = restic.ID{}
}
2016-09-01 19:19:30 +00:00
h := restic.Handle{Type: t, Name: id.String()}
err := r.be.Load(ctx, h, 0, 0, func(rd io.Reader) error {
// make sure this call is idempotent, in case an error occurs
wr := bytes.NewBuffer(buf[:0])
_, cerr := io.Copy(wr, rd)
if cerr != nil {
return cerr
}
buf = wr.Bytes()
return nil
})
2015-03-28 10:50:23 +00:00
if err != nil {
return nil, err
}
2016-08-31 18:29:54 +00:00
if t != restic.ConfigFile && !restic.Hash(buf).Equal(id) {
2017-02-11 13:28:15 +00:00
return nil, errors.Errorf("load %v: invalid data returned", h)
2015-03-28 10:50:23 +00:00
}
2017-10-29 10:33:57 +00:00
nonce, ciphertext := buf[:r.key.NonceSize()], buf[r.key.NonceSize():]
plaintext, err := r.key.Open(ciphertext[:0], nonce, ciphertext, nil)
if err != nil {
return nil, err
}
2017-10-29 10:33:57 +00:00
return plaintext, nil
}
type haver interface {
Has(restic.Handle) bool
}
// sortCachedPacksFirst moves all cached pack files to the front of blobs.
func sortCachedPacksFirst(cache haver, blobs []restic.PackedBlob) {
if cache == nil {
return
2017-07-16 19:06:43 +00:00
}
// no need to sort a list with one element
if len(blobs) == 1 {
return
}
cached := blobs[:0]
2017-07-16 19:06:43 +00:00
noncached := make([]restic.PackedBlob, 0, len(blobs)/2)
for _, blob := range blobs {
if cache.Has(restic.Handle{Type: restic.PackFile, Name: blob.PackID.String()}) {
2017-07-16 19:06:43 +00:00
cached = append(cached, blob)
continue
}
noncached = append(noncached, blob)
}
copy(blobs[len(cached):], noncached)
2017-07-16 19:06:43 +00:00
}
// LoadBlob loads a blob of type t from the repository.
// It may use all of buf[:cap(buf)] as scratch space.
func (r *Repository) LoadBlob(ctx context.Context, t restic.BlobType, id restic.ID, buf []byte) ([]byte, error) {
debug.Log("load %v with id %v (buf len %v, cap %d)", t, id, len(buf), cap(buf))
// lookup packs
2020-11-05 21:18:00 +00:00
blobs := r.idx.Lookup(restic.BlobHandle{ID: id, Type: t})
if len(blobs) == 0 {
2018-01-25 19:49:41 +00:00
debug.Log("id %v not found in index", id)
return nil, errors.Errorf("id %v not found in repository", id)
}
2017-07-16 19:06:43 +00:00
// try cached pack files first
sortCachedPacksFirst(r.Cache, blobs)
2017-07-16 19:06:43 +00:00
2016-08-28 20:18:02 +00:00
var lastError error
for _, blob := range blobs {
2018-01-25 19:49:41 +00:00
debug.Log("blob %v/%v found: %v", t, id, blob)
if blob.Type != t {
2016-09-27 20:35:08 +00:00
debug.Log("blob %v has wrong block type, want %v", blob, t)
}
// load blob from pack
2020-07-28 08:13:11 +00:00
bt := t
if r.idx.IsMixedPack(blob.PackID) {
bt = restic.InvalidBlob
}
h := restic.Handle{Type: restic.PackFile,
Name: blob.PackID.String(), ContainedBlobType: bt}
switch {
case cap(buf) < int(blob.Length):
buf = make([]byte, blob.Length)
case len(buf) != int(blob.Length):
buf = buf[:blob.Length]
}
n, err := restic.ReadAt(ctx, r.be, h, int64(blob.Offset), buf)
if err != nil {
2016-09-27 20:35:08 +00:00
debug.Log("error loading blob %v: %v", blob, err)
2016-08-28 20:18:02 +00:00
lastError = err
continue
}
if uint(n) != blob.Length {
2016-08-28 20:18:02 +00:00
lastError = errors.Errorf("error loading blob %v: wrong length returned, want %d, got %d",
id.Str(), blob.Length, uint(n))
2016-09-27 20:35:08 +00:00
debug.Log("lastError: %v", lastError)
continue
}
// decrypt
nonce, ciphertext := buf[:r.key.NonceSize()], buf[r.key.NonceSize():]
2017-10-29 10:33:57 +00:00
plaintext, err := r.key.Open(ciphertext[:0], nonce, ciphertext, nil)
if err != nil {
2016-08-28 20:18:02 +00:00
lastError = errors.Errorf("decrypting blob %v failed: %v", id, err)
continue
}
// check hash
2017-10-29 10:33:57 +00:00
if !restic.Hash(plaintext).Equal(id) {
2016-08-28 20:18:02 +00:00
lastError = errors.Errorf("blob %v returned invalid hash", id)
continue
}
// move decrypted data to the start of the buffer
copy(buf, plaintext)
return buf[:len(plaintext)], nil
}
2016-08-28 20:18:02 +00:00
if lastError != nil {
return nil, lastError
2016-08-28 20:18:02 +00:00
}
return nil, errors.Errorf("loading blob %v from %v packs failed", id.Str(), len(blobs))
}
2015-05-04 18:39:45 +00:00
// LoadJSONUnpacked decrypts the data and afterwards calls json.Unmarshal on
// the item.
2017-06-04 09:16:55 +00:00
func (r *Repository) LoadJSONUnpacked(ctx context.Context, t restic.FileType, id restic.ID, item interface{}) (err error) {
buf, err := r.LoadUnpacked(ctx, nil, t, id)
if err != nil {
return err
}
return json.Unmarshal(buf, item)
}
// LookupBlobSize returns the size of blob id.
func (r *Repository) LookupBlobSize(id restic.ID, tpe restic.BlobType) (uint, bool) {
2020-11-05 21:18:00 +00:00
return r.idx.LookupSize(restic.BlobHandle{ID: id, Type: tpe})
}
// saveAndEncrypt encrypts data and stores it to the backend as type t. If data
// is small enough, it will be packed together with other small blobs.
// The caller must ensure that the id matches the data.
func (r *Repository) saveAndEncrypt(ctx context.Context, t restic.BlobType, data []byte, id restic.ID) error {
2018-01-25 19:49:41 +00:00
debug.Log("save id %v (%v, %d bytes)", id, t, len(data))
2017-10-29 10:33:57 +00:00
nonce := crypto.NewRandomNonce()
ciphertext := make([]byte, 0, restic.CiphertextLength(len(data)))
2017-10-29 10:33:57 +00:00
ciphertext = append(ciphertext, nonce...)
// encrypt blob
2017-10-29 10:33:57 +00:00
ciphertext = r.key.Seal(ciphertext, nonce, data, nil)
// find suitable packer and add blob
var pm *packerManager
switch t {
case restic.TreeBlob:
pm = r.treePM
case restic.DataBlob:
pm = r.dataPM
default:
panic(fmt.Sprintf("invalid type: %v", t))
}
packer, err := pm.findPacker()
if err != nil {
return err
}
// save ciphertext
_, err = packer.Add(t, id, ciphertext)
2015-10-12 20:34:12 +00:00
if err != nil {
return err
2015-10-12 20:34:12 +00:00
}
// if the pack is not full enough, put back to the list
if packer.Size() < minPackSize {
2016-09-27 20:35:08 +00:00
debug.Log("pack is not full enough (%d bytes)", packer.Size())
pm.insertPacker(packer)
return nil
}
// else write the pack to the backend
return r.savePacker(ctx, t, packer)
}
// SaveJSONUnpacked serialises item as JSON and encrypts and saves it in the
// backend as type t, without a pack. It returns the storage hash.
2017-06-04 09:16:55 +00:00
func (r *Repository) SaveJSONUnpacked(ctx context.Context, t restic.FileType, item interface{}) (restic.ID, error) {
2016-09-27 20:35:08 +00:00
debug.Log("save new blob %v", t)
2016-01-24 17:50:41 +00:00
plaintext, err := json.Marshal(item)
if err != nil {
2016-08-31 18:29:54 +00:00
return restic.ID{}, errors.Wrap(err, "json.Marshal")
}
2017-06-04 09:16:55 +00:00
return r.SaveUnpacked(ctx, t, plaintext)
2016-01-24 17:52:11 +00:00
}
// SaveUnpacked encrypts data and stores it in the backend. Returned is the
// storage hash.
2017-06-04 09:16:55 +00:00
func (r *Repository) SaveUnpacked(ctx context.Context, t restic.FileType, p []byte) (id restic.ID, err error) {
ciphertext := restic.NewBlobBuffer(len(p))
2017-10-29 10:33:57 +00:00
ciphertext = ciphertext[:0]
nonce := crypto.NewRandomNonce()
ciphertext = append(ciphertext, nonce...)
ciphertext = r.key.Seal(ciphertext, nonce, p, nil)
if t == restic.ConfigFile {
id = restic.ID{}
} else {
id = restic.Hash(ciphertext)
}
2016-09-01 19:19:30 +00:00
h := restic.Handle{Type: t, Name: id.String()}
err = r.be.Save(ctx, h, restic.NewByteReader(ciphertext, r.be.Hasher()))
if err != nil {
2016-09-27 20:35:08 +00:00
debug.Log("error saving blob %v: %v", h, err)
2016-08-31 18:29:54 +00:00
return restic.ID{}, err
}
2016-09-27 20:35:08 +00:00
debug.Log("blob %v saved", h)
2016-01-24 17:50:41 +00:00
return id, nil
}
// Flush saves all remaining packs and the index
func (r *Repository) Flush(ctx context.Context) error {
if err := r.FlushPacks(ctx); err != nil {
return err
}
// Save index after flushing only if noAutoIndexUpdate is not set
if r.noAutoIndexUpdate {
return nil
}
return r.SaveIndex(ctx)
}
// FlushPacks saves all remaining packs.
func (r *Repository) FlushPacks(ctx context.Context) error {
pms := []struct {
t restic.BlobType
pm *packerManager
}{
{restic.DataBlob, r.dataPM},
{restic.TreeBlob, r.treePM},
}
for _, p := range pms {
p.pm.pm.Lock()
debug.Log("manually flushing %d packs", len(p.pm.packers))
for _, packer := range p.pm.packers {
err := r.savePacker(ctx, p.t, packer)
if err != nil {
p.pm.pm.Unlock()
return err
}
}
p.pm.packers = p.pm.packers[:0]
p.pm.pm.Unlock()
}
return nil
}
// Backend returns the backend for the repository.
2016-08-31 18:29:54 +00:00
func (r *Repository) Backend() restic.Backend {
return r.be
}
2015-10-12 20:34:12 +00:00
// Index returns the currently used MasterIndex.
func (r *Repository) Index() restic.MasterIndex {
return r.idx
}
2015-05-09 11:25:52 +00:00
// SetIndex instructs the repository to use the given index.
func (r *Repository) SetIndex(i restic.MasterIndex) error {
2016-08-31 20:39:36 +00:00
r.idx = i.(*MasterIndex)
ids := restic.NewIDSet()
for _, idx := range r.idx.All() {
indexIDs, err := idx.IDs()
if err != nil {
debug.Log("not using index, ID() returned error %v", err)
continue
}
for _, id := range indexIDs {
ids.Insert(id)
}
}
return r.PrepareCache(ids)
}
2016-01-24 17:52:11 +00:00
// SaveIndex saves an index in the repository.
2017-06-04 09:16:55 +00:00
func SaveIndex(ctx context.Context, repo restic.Repository, index *Index) (restic.ID, error) {
2016-01-24 17:52:11 +00:00
buf := bytes.NewBuffer(nil)
2015-10-25 16:05:54 +00:00
err := index.Encode(buf)
2015-10-25 16:05:54 +00:00
if err != nil {
2016-08-31 18:29:54 +00:00
return restic.ID{}, err
2015-10-25 16:05:54 +00:00
}
2017-06-04 09:16:55 +00:00
return repo.SaveUnpacked(ctx, restic.IndexFile, buf.Bytes())
2015-10-25 16:05:54 +00:00
}
2015-10-12 21:59:17 +00:00
// saveIndex saves all indexes in the backend.
2017-06-04 09:16:55 +00:00
func (r *Repository) saveIndex(ctx context.Context, indexes ...*Index) error {
2015-10-12 21:59:17 +00:00
for i, idx := range indexes {
2016-09-27 20:35:08 +00:00
debug.Log("Saving index %d", i)
2017-06-04 09:16:55 +00:00
sid, err := SaveIndex(ctx, r, idx)
2015-10-12 20:34:12 +00:00
if err != nil {
return err
}
2015-02-15 23:24:58 +00:00
2018-01-25 19:49:41 +00:00
debug.Log("Saved index %d as %v", i, sid)
2015-10-12 20:34:12 +00:00
}
2021-01-30 17:49:30 +00:00
return r.idx.MergeFinalIndexes()
}
2015-10-12 21:59:17 +00:00
// SaveIndex saves all new indexes in the backend.
2017-06-04 09:16:55 +00:00
func (r *Repository) SaveIndex(ctx context.Context) error {
return r.saveIndex(ctx, r.idx.FinalizeNotFinalIndexes()...)
2015-10-12 21:59:17 +00:00
}
// SaveFullIndex saves all full indexes in the backend.
2017-06-04 09:16:55 +00:00
func (r *Repository) SaveFullIndex(ctx context.Context) error {
return r.saveIndex(ctx, r.idx.FinalizeFullIndexes()...)
2015-10-12 21:59:17 +00:00
}
2015-10-12 20:34:12 +00:00
// LoadIndex loads all index files from the backend in parallel and stores them
// in the master index. The first error that occurred is returned.
2017-06-04 09:16:55 +00:00
func (r *Repository) LoadIndex(ctx context.Context) error {
2016-09-27 20:35:08 +00:00
debug.Log("Loading index")
validIndex := restic.NewIDSet()
err := ForAllIndexes(ctx, r, func(id restic.ID, idx *Index, oldFormat bool, err error) error {
if err != nil {
return err
2015-07-04 16:38:32 +00:00
}
ids, err := idx.IDs()
if err != nil {
return err
}
for _, id := range ids {
validIndex.Insert(id)
2017-09-24 09:25:45 +00:00
}
r.idx.Insert(idx)
return nil
})
if err != nil {
return errors.Fatal(err.Error())
2015-07-04 16:38:32 +00:00
}
err = r.idx.MergeFinalIndexes()
if err != nil {
return err
}
// remove index files from the cache which have been removed in the repo
2020-12-11 08:41:59 +00:00
return r.PrepareCache(validIndex)
}
2017-07-18 21:16:50 +00:00
2020-10-10 19:31:40 +00:00
const listPackParallelism = 10
// CreateIndexFromPacks creates a new index by reading all given pack files (with sizes).
// The index is added to the MasterIndex but not marked as finalized.
// Returned is the list of pack files which could not be read.
func (r *Repository) CreateIndexFromPacks(ctx context.Context, packsize map[restic.ID]int64, p *progress.Counter) (invalid restic.IDs, err error) {
var m sync.Mutex
debug.Log("Loading index from pack files")
// track spawned goroutines using wg, create a new context which is
// cancelled as soon as an error occurs.
wg, ctx := errgroup.WithContext(ctx)
type FileInfo struct {
restic.ID
Size int64
}
ch := make(chan FileInfo)
// send list of pack files through ch, which is closed afterwards
wg.Go(func() error {
defer close(ch)
for id, size := range packsize {
select {
case <-ctx.Done():
return nil
case ch <- FileInfo{id, size}:
}
}
return nil
})
idx := NewIndex()
// a worker receives an pack ID from ch, reads the pack contents, and adds them to idx
worker := func() error {
for fi := range ch {
entries, _, err := r.ListPack(ctx, fi.ID, fi.Size)
if err != nil {
debug.Log("unable to list pack file %v", fi.ID.Str())
m.Lock()
invalid = append(invalid, fi.ID)
m.Unlock()
}
idx.StorePack(fi.ID, entries)
p.Add(1)
}
return nil
}
// run workers on ch
wg.Go(func() error {
return RunWorkers(listPackParallelism, worker)
})
err = wg.Wait()
if err != nil {
return invalid, errors.Fatal(err.Error())
}
// Add idx to MasterIndex
r.idx.Insert(idx)
return invalid, nil
}
// PrepareCache initializes the local cache. indexIDs is the list of IDs of
// index files still present in the repo.
func (r *Repository) PrepareCache(indexIDs restic.IDSet) error {
if r.Cache == nil {
return nil
}
debug.Log("prepare cache with %d index files", len(indexIDs))
// clear old index files
err := r.Cache.Clear(restic.IndexFile, indexIDs)
if err != nil {
fmt.Fprintf(os.Stderr, "error clearing index files in cache: %v\n", err)
}
packs := restic.NewIDSet()
for _, idx := range r.idx.All() {
for id := range idx.Packs() {
packs.Insert(id)
2017-07-18 21:16:50 +00:00
}
}
// clear old packs
err = r.Cache.Clear(restic.PackFile, packs)
if err != nil {
fmt.Fprintf(os.Stderr, "error clearing pack files in cache: %v\n", err)
}
return nil
}
2015-05-04 18:39:45 +00:00
// SearchKey finds a key with the supplied password, afterwards the config is
// read and parsed. It tries at most maxKeys key files in the repo.
2018-11-25 14:10:45 +00:00
func (r *Repository) SearchKey(ctx context.Context, password string, maxKeys int, keyHint string) error {
key, err := SearchKey(ctx, r, password, maxKeys, keyHint)
2014-12-21 17:10:19 +00:00
if err != nil {
return err
}
r.key = key.master
r.dataPM.key = key.master
r.treePM.key = key.master
r.keyName = key.Name()
2017-06-04 09:16:55 +00:00
r.cfg, err = restic.LoadConfig(ctx, r)
if err == crypto.ErrUnauthenticated {
return errors.Fatalf("config or key %v is damaged: %v", key.Name(), err)
} else if err != nil {
return errors.Fatalf("config cannot be loaded: %v", err)
}
return nil
2015-05-03 14:36:52 +00:00
}
2014-12-21 17:10:19 +00:00
2015-07-02 20:36:31 +00:00
// Init creates a new master key with the supplied password, initializes and
// saves the repository config.
func (r *Repository) Init(ctx context.Context, password string, chunkerPolynomial *chunker.Pol) error {
2017-06-04 09:16:55 +00:00
has, err := r.be.Test(ctx, restic.Handle{Type: restic.ConfigFile})
2015-05-03 15:46:18 +00:00
if err != nil {
return err
}
if has {
return errors.New("repository master key and config already initialized")
}
2016-08-31 20:39:36 +00:00
cfg, err := restic.CreateConfig()
if err != nil {
return err
}
if chunkerPolynomial != nil {
cfg.ChunkerPolynomial = *chunkerPolynomial
}
2017-06-04 09:16:55 +00:00
return r.init(ctx, password, cfg)
}
// init creates a new master key with the supplied password and uses it to save
// the config into the repo.
2017-06-04 09:16:55 +00:00
func (r *Repository) init(ctx context.Context, password string, cfg restic.Config) error {
key, err := createMasterKey(ctx, r, password)
2015-05-03 14:36:52 +00:00
if err != nil {
return err
}
r.key = key.master
r.dataPM.key = key.master
r.treePM.key = key.master
r.keyName = key.Name()
2016-08-31 20:39:36 +00:00
r.cfg = cfg
2017-06-04 09:16:55 +00:00
_, err = r.SaveJSONUnpacked(ctx, restic.ConfigFile, cfg)
2015-07-02 20:36:31 +00:00
return err
2014-12-21 17:10:19 +00:00
}
// Key returns the current master key.
func (r *Repository) Key() *crypto.Key {
return r.key
2014-12-21 17:10:19 +00:00
}
// KeyName returns the name of the current key in the backend.
func (r *Repository) KeyName() string {
return r.keyName
}
// List runs fn for all files of type t in the repo.
func (r *Repository) List(ctx context.Context, t restic.FileType, fn func(restic.ID, int64) error) error {
return r.be.List(ctx, t, func(fi restic.FileInfo) error {
id, err := restic.ParseID(fi.Name)
if err != nil {
debug.Log("unable to parse %v as an ID", fi.Name)
return nil
2017-03-06 21:19:38 +00:00
}
return fn(id, fi.Size)
})
2014-12-21 16:02:49 +00:00
}
2016-08-07 19:56:42 +00:00
// ListPack returns the list of blobs saved in the pack id and the length of
2020-11-16 03:03:45 +00:00
// the the pack header.
func (r *Repository) ListPack(ctx context.Context, id restic.ID, size int64) ([]restic.Blob, uint32, error) {
h := restic.Handle{Type: restic.PackFile, Name: id.String()}
2016-08-07 19:56:42 +00:00
2020-11-16 03:03:45 +00:00
return pack.List(r.Key(), restic.ReaderAt(ctx, r.Backend(), h), size)
2016-05-08 11:51:21 +00:00
}
// Delete calls backend.Delete() if implemented, and returns an error
// otherwise.
2017-06-04 09:16:55 +00:00
func (r *Repository) Delete(ctx context.Context) error {
2017-10-14 13:56:38 +00:00
return r.be.Delete(ctx)
2014-12-21 16:02:49 +00:00
}
// Close closes the repository by closing the backend.
func (r *Repository) Close() error {
return r.be.Close()
2015-03-28 10:50:23 +00:00
}
// SaveBlob saves a blob of type t into the repository.
// It takes care that no duplicates are saved; this can be overwritten
// by setting storeDuplicate to true.
// If id is the null id, it will be computed and returned.
// Also returns if the blob was already known before
func (r *Repository) SaveBlob(ctx context.Context, t restic.BlobType, buf []byte, id restic.ID, storeDuplicate bool) (newID restic.ID, known bool, err error) {
// compute plaintext hash if not already set
if id.IsNull() {
newID = restic.Hash(buf)
} else {
newID = id
}
// first try to add to pending blobs; if not successful, this blob is already known
2020-11-05 21:18:00 +00:00
known = !r.idx.addPending(restic.BlobHandle{ID: newID, Type: t})
2020-10-05 21:13:38 +00:00
// only save when needed or explicitly told
if !known || storeDuplicate {
err = r.saveAndEncrypt(ctx, t, buf, newID)
2016-09-03 18:55:22 +00:00
}
return newID, known, err
2016-09-03 18:55:22 +00:00
}
// LoadTree loads a tree from the repository.
2017-06-04 09:16:55 +00:00
func (r *Repository) LoadTree(ctx context.Context, id restic.ID) (*restic.Tree, error) {
2018-01-25 19:49:41 +00:00
debug.Log("load tree %v", id)
buf, err := r.LoadBlob(ctx, restic.TreeBlob, id, nil)
if err != nil {
return nil, err
}
t := &restic.Tree{}
err = json.Unmarshal(buf, t)
if err != nil {
return nil, err
}
return t, nil
}
2016-09-03 18:55:22 +00:00
// SaveTree stores a tree into the repository and returns the ID. The ID is
// checked against the index. The tree is only stored when the index does not
// contain the ID.
2017-06-04 09:16:55 +00:00
func (r *Repository) SaveTree(ctx context.Context, t *restic.Tree) (restic.ID, error) {
2016-09-03 18:55:22 +00:00
buf, err := json.Marshal(t)
if err != nil {
2016-09-03 18:55:22 +00:00
return restic.ID{}, errors.Wrap(err, "MarshalJSON")
}
2016-09-03 18:55:22 +00:00
// append a newline so that the data is always consistent (json.Encoder
// adds a newline after each object)
buf = append(buf, '\n')
id, _, err := r.SaveBlob(ctx, restic.TreeBlob, buf, restic.ID{}, false)
2016-09-03 18:55:22 +00:00
return id, err
}
// Loader allows loading data from a backend.
type Loader interface {
Load(ctx context.Context, h restic.Handle, length int, offset int64, fn func(rd io.Reader) error) error
}
type BackendLoadFn func(ctx context.Context, h restic.Handle, length int, offset int64, fn func(rd io.Reader) error) error
// StreamPack loads the listed blobs from the specified pack file. The plaintext blob is passed to
// the handleBlobFn callback or an error if decryption failed or the blob hash does not match. In
// case of download errors handleBlobFn might be called multiple times for the same blob. If the
// callback returns an error, then StreamPack will abort and not retry it.
func StreamPack(ctx context.Context, beLoad BackendLoadFn, key *crypto.Key, packID restic.ID, blobs []restic.Blob, handleBlobFn func(blob restic.BlobHandle, buf []byte, err error) error) error {
if len(blobs) == 0 {
// nothing to do
return nil
}
sort.Slice(blobs, func(i, j int) bool {
return blobs[i].Offset < blobs[j].Offset
})
2021-08-20 10:12:38 +00:00
h := restic.Handle{Type: restic.PackFile, Name: packID.String(), ContainedBlobType: restic.DataBlob}
dataStart := blobs[0].Offset
dataEnd := blobs[len(blobs)-1].Offset + blobs[len(blobs)-1].Length
debug.Log("streaming pack %v (%d to %d bytes), blobs: %v", packID, dataStart, dataEnd, len(blobs))
ctx, cancel := context.WithCancel(ctx)
// stream blobs in pack
err := beLoad(ctx, h, int(dataEnd-dataStart), int64(dataStart), func(rd io.Reader) error {
// prevent callbacks after cancelation
if ctx.Err() != nil {
return ctx.Err()
}
bufferSize := int(dataEnd - dataStart)
if bufferSize > MaxStreamBufferSize {
bufferSize = MaxStreamBufferSize
}
2021-08-22 12:38:27 +00:00
// create reader here to allow reusing the buffered reader from checker.checkData
bufRd := bufio.NewReaderSize(rd, bufferSize)
currentBlobEnd := dataStart
var buf []byte
for _, entry := range blobs {
skipBytes := int(entry.Offset - currentBlobEnd)
if skipBytes < 0 {
return errors.Errorf("overlapping blobs in pack %v", packID)
}
_, err := bufRd.Discard(skipBytes)
if err != nil {
return err
}
h := restic.BlobHandle{ID: entry.ID, Type: entry.Type}
debug.Log(" process blob %v, skipped %d, %v", h, skipBytes, entry)
if uint(cap(buf)) < entry.Length {
buf = make([]byte, entry.Length)
}
buf = buf[:entry.Length]
n, err := io.ReadFull(bufRd, buf)
if err != nil {
debug.Log(" read error %v", err)
return errors.Wrap(err, "ReadFull")
}
if n != len(buf) {
return errors.Errorf("read blob %v from %v: not enough bytes read, want %v, got %v",
h, packID.Str(), len(buf), n)
}
currentBlobEnd = entry.Offset + entry.Length
if int(entry.Length) <= key.NonceSize() {
debug.Log("%v", blobs)
return errors.Errorf("invalid blob length %v", entry)
}
// decryption errors are likely permanent, give the caller a chance to skip them
nonce, ciphertext := buf[:key.NonceSize()], buf[key.NonceSize():]
plaintext, err := key.Open(ciphertext[:0], nonce, ciphertext, nil)
if err == nil {
id := restic.Hash(plaintext)
if !id.Equal(entry.ID) {
debug.Log("read blob %v/%v from %v: wrong data returned, hash is %v",
h.Type, h.ID, packID.Str(), id)
err = errors.Errorf("read blob %v from %v: wrong data returned, hash is %v",
h, packID.Str(), id)
}
}
err = handleBlobFn(entry.BlobHandle, plaintext, err)
if err != nil {
cancel()
return backoff.Permanent(err)
}
}
return nil
})
return errors.Wrap(err, "StreamPack")
}