// selectBlobs splits the list of all blobs randomly into two lists. A blob
// will be contained in the firstone ith probability p.
func selectBlobs(t *testing.T, repo restic.Repository, p float32) (list1, list2 restic.BlobSet) {
done := make(chan struct{})
defer close(done)
list1 = restic.NewBlobSet()
list2 = restic.NewBlobSet()
blobs := restic.NewBlobSet()
for id := range repo.List(restic.DataFile, done) {
entries, _, err := repo.ListPack(id)
if err != nil {
t.Fatalf("error listing pack %v: %v", id, err)
}
for _, entry := range entries {
h := restic.BlobHandle{ID: entry.ID, Type: entry.Type}
if blobs.Has(h) {
t.Errorf("ignoring duplicate blob %v", h)
continue
}
blobs.Insert(h)
if rand.Float32() <= p {
list1.Insert(restic.BlobHandle{ID: entry.ID, Type: entry.Type})
} else {
list2.Insert(restic.BlobHandle{ID: entry.ID, Type: entry.Type})
}
}
}
return list1, list2
}
func validateIndex(t testing.TB, repo restic.Repository, idx *Index) {
for id := range repo.List(restic.DataFile, nil) {
if _, ok := idx.Packs[id]; !ok {
t.Errorf("pack %v missing from index", id.Str())
}
}
}
func packIDTester(repo restic.Repository, inChan <-chan restic.ID, errChan chan<- error, wg *sync.WaitGroup, done <-chan struct{}) {
debug.Log("worker start")
defer debug.Log("worker done")
defer wg.Done()
for id := range inChan {
ok, err := repo.Backend().Test(restic.DataFile, id.String())
if err != nil {
err = PackError{ID: id, Err: err}
} else {
if !ok {
err = PackError{ID: id, Err: errors.New("does not exist")}
}
}
if err != nil {
debug.Log("error checking for pack %s: %v", id.Str(), err)
select {
case <-done:
return
case errChan <- err:
}
continue
}
debug.Log("pack %s exists", id.Str())
}
}
func newDirFromSnapshot(repo restic.Repository, snapshot SnapshotWithId, ownerIsRoot bool) (*dir, error) {
debug.Log("new dir for snapshot %v (%v)", snapshot.ID.Str(), snapshot.Tree.Str())
tree, err := repo.LoadTree(*snapshot.Tree)
if err != nil {
debug.Log(" loadTree(%v) failed: %v", snapshot.ID.Str(), err)
return nil, err
}
items := make(map[string]*restic.Node)
for _, n := range tree.Nodes {
nodes, err := replaceSpecialNodes(repo, n)
if err != nil {
debug.Log(" replaceSpecialNodes(%v) failed: %v", n, err)
return nil, err
}
for _, node := range nodes {
items[node.Name] = node
}
}
return &dir{
repo: repo,
node: &restic.Node{
UID: uint32(os.Getuid()),
GID: uint32(os.Getgid()),
AccessTime: snapshot.Time,
ModTime: snapshot.Time,
ChangeTime: snapshot.Time,
Mode: os.ModeDir | 0555,
},
items: items,
inode: inodeFromBackendID(snapshot.ID),
ownerIsRoot: ownerIsRoot,
}, nil
}
// Save writes a new index containing the given packs.
func Save(repo restic.Repository, packs map[restic.ID][]restic.Blob, supersedes restic.IDs) (restic.ID, error) {
idx := &indexJSON{
Supersedes: supersedes,
Packs: make([]*packJSON, 0, len(packs)),
}
for packID, blobs := range packs {
b := make([]blobJSON, 0, len(blobs))
for _, blob := range blobs {
b = append(b, blobJSON{
ID: blob.ID,
Type: blob.Type,
Offset: blob.Offset,
Length: blob.Length,
})
}
p := &packJSON{
ID: packID,
Blobs: b,
}
idx.Packs = append(idx.Packs, p)
}
return repo.SaveJSONUnpacked(restic.IndexFile, idx)
}
func countPacks(repo restic.Repository, t restic.FileType) (n uint) {
for _ = range repo.Backend().List(t, nil) {
n++
}
return n
}
func loadBlob(t *testing.T, repo restic.Repository, id restic.ID, buf []byte) int {
n, err := repo.LoadBlob(restic.DataBlob, id, buf)
if err != nil {
t.Fatalf("LoadBlob(%v) returned error %v", id, err)
}
return n
}
func createFakeLock(repo restic.Repository, t time.Time, pid int) (restic.ID, error) {
hostname, err := os.Hostname()
if err != nil {
return restic.ID{}, err
}
newLock := &restic.Lock{Time: t, PID: pid, Hostname: hostname}
return repo.SaveJSONUnpacked(restic.LockFile, &newLock)
}
// SaveIndex saves an index in the repository.
func SaveIndex(repo restic.Repository, index *Index) (restic.ID, error) {
buf := bytes.NewBuffer(nil)
err := index.Finalize(buf)
if err != nil {
return restic.ID{}, err
}
return repo.SaveUnpacked(restic.IndexFile, buf.Bytes())
}
func loadIndexJSON(repo restic.Repository, id restic.ID) (*indexJSON, error) {
debug.Log("process index %v\n", id.Str())
var idx indexJSON
err := repo.LoadJSONUnpacked(restic.IndexFile, id, &idx)
if err != nil {
return nil, err
}
return &idx, nil
}
func listPacks(t *testing.T, repo restic.Repository) restic.IDSet {
done := make(chan struct{})
defer close(done)
list := restic.NewIDSet()
for id := range repo.List(restic.DataFile, done) {
list.Insert(id)
}
return list
}
// saveRandomDataBlobs generates random data blobs and saves them to the repository.
func saveRandomDataBlobs(t testing.TB, repo restic.Repository, num int, sizeMax int) {
for i := 0; i < num; i++ {
size := mrand.Int() % sizeMax
buf := make([]byte, size)
_, err := io.ReadFull(rand.Reader, buf)
OK(t, err)
_, err = repo.SaveBlob(restic.DataBlob, buf, restic.ID{})
OK(t, err)
}
}
// RebuildIndex lists all packs in the repo, writes a new index and removes all
// old indexes. This operation should only be done with an exclusive lock in
// place.
func RebuildIndex(repo restic.Repository) error {
debug.Log("start rebuilding index")
done := make(chan struct{})
defer close(done)
ch := make(chan worker.Job)
go list.AllPacks(repo, ch, done)
idx := NewIndex()
for job := range ch {
id := job.Data.(restic.ID)
if job.Error != nil {
fmt.Fprintf(os.Stderr, "error for pack %v: %v\n", id, job.Error)
continue
}
res := job.Result.(list.Result)
for _, entry := range res.Entries() {
pb := restic.PackedBlob{
Blob: entry,
PackID: res.PackID(),
}
idx.Store(pb)
}
}
oldIndexes := restic.NewIDSet()
for id := range repo.List(restic.IndexFile, done) {
idx.AddToSupersedes(id)
oldIndexes.Insert(id)
}
id, err := SaveIndex(repo, idx)
if err != nil {
debug.Log("error saving index: %v", err)
return err
}
debug.Log("new index saved as %v", id.Str())
for indexID := range oldIndexes {
err := repo.Backend().Remove(restic.IndexFile, indexID.String())
if err != nil {
fmt.Fprintf(os.Stderr, "unable to remove index %v: %v\n", indexID.Str(), err)
}
}
return nil
}
// replaceSpecialNodes replaces nodes with name "." and "/" by their contents.
// Otherwise, the node is returned.
func replaceSpecialNodes(repo restic.Repository, node *restic.Node) ([]*restic.Node, error) {
if node.Type != "dir" || node.Subtree == nil {
return []*restic.Node{node}, nil
}
if node.Name != "." && node.Name != "/" {
return []*restic.Node{node}, nil
}
tree, err := repo.LoadTree(*node.Subtree)
if err != nil {
return nil, err
}
return tree.Nodes, nil
}
func findPacksForBlobs(t *testing.T, repo restic.Repository, blobs restic.BlobSet) restic.IDSet {
packs := restic.NewIDSet()
idx := repo.Index()
for h := range blobs {
list, err := idx.Lookup(h.ID, h.Type)
if err != nil {
t.Fatal(err)
}
for _, pb := range list {
packs.Insert(pb.PackID)
}
}
return packs
}
// LoadIndexWithDecoder loads the index and decodes it with fn.
func LoadIndexWithDecoder(repo restic.Repository, id restic.ID, fn func(io.Reader) (*Index, error)) (idx *Index, err error) {
debug.Log("Loading index %v", id.Str())
buf, err := repo.LoadAndDecrypt(restic.IndexFile, id)
if err != nil {
return nil, err
}
idx, err = fn(bytes.NewReader(buf))
if err != nil {
debug.Log("error while decoding index %v: %v", id, err)
return nil, err
}
idx.id = id
return idx, nil
}
func checkSavedFile(t *testing.T, repo restic.Repository, treeID restic.ID, name string, rd io.Reader) {
tree, err := repo.LoadTree(treeID)
if err != nil {
t.Fatalf("LoadTree() returned error %v", err)
}
if len(tree.Nodes) != 1 {
t.Fatalf("wrong number of nodes for tree, want %v, got %v", 1, len(tree.Nodes))
}
node := tree.Nodes[0]
if node.Name != "fakefile" {
t.Fatalf("wrong filename, want %v, got %v", "fakefile", node.Name)
}
if len(node.Content) == 0 {
t.Fatalf("node.Content has length 0")
}
// check blobs
for i, id := range node.Content {
size, err := repo.LookupBlobSize(id, restic.DataBlob)
if err != nil {
t.Fatal(err)
}
buf := make([]byte, int(size))
n := loadBlob(t, repo, id, buf)
if n != len(buf) {
t.Errorf("wrong number of bytes read, want %d, got %d", len(buf), n)
}
buf2 := make([]byte, int(size))
_, err = io.ReadFull(rd, buf2)
if err != nil {
t.Fatal(err)
}
if !bytes.Equal(buf, buf2) {
t.Fatalf("blob %d (%v) is wrong", i, id.Str())
}
}
}
// loadSnapshotTreeIDs loads all snapshots from backend and returns the tree IDs.
func loadSnapshotTreeIDs(repo restic.Repository) (restic.IDs, []error) {
var trees struct {
IDs restic.IDs
sync.Mutex
}
var errs struct {
errs []error
sync.Mutex
}
snapshotWorker := func(strID string, done <-chan struct{}) error {
id, err := restic.ParseID(strID)
if err != nil {
return err
}
debug.Log("load snapshot %v", id.Str())
treeID, err := loadTreeFromSnapshot(repo, id)
if err != nil {
errs.Lock()
errs.errs = append(errs.errs, err)
errs.Unlock()
return nil
}
debug.Log("snapshot %v has tree %v", id.Str(), treeID.Str())
trees.Lock()
trees.IDs = append(trees.IDs, treeID)
trees.Unlock()
return nil
}
err := repository.FilesInParallel(repo.Backend(), restic.SnapshotFile, defaultParallelism, snapshotWorker)
if err != nil {
errs.errs = append(errs.errs, err)
}
return trees.IDs, errs.errs
}
func dumpIndexes(repo restic.Repository) error {
done := make(chan struct{})
defer close(done)
for id := range repo.List(restic.IndexFile, done) {
fmt.Printf("index_id: %v\n", id)
idx, err := repository.LoadIndex(repo, id)
if err != nil {
return err
}
err = idx.Dump(os.Stdout)
if err != nil {
return err
}
}
return nil
}
func newDir(repo restic.Repository, node *restic.Node, ownerIsRoot bool) (*dir, error) {
debug.Log("new dir for %v (%v)", node.Name, node.Subtree.Str())
tree, err := repo.LoadTree(*node.Subtree)
if err != nil {
debug.Log(" error loading tree %v: %v", node.Subtree.Str(), err)
return nil, err
}
items := make(map[string]*restic.Node)
for _, node := range tree.Nodes {
items[node.Name] = node
}
return &dir{
repo: repo,
node: node,
items: items,
inode: node.Inode,
ownerIsRoot: ownerIsRoot,
}, nil
}
// loadTreeWorker loads trees from repo and sends them to out.
func loadTreeWorker(repo restic.Repository,
in <-chan restic.ID, out chan<- treeJob,
done <-chan struct{}, wg *sync.WaitGroup) {
defer func() {
debug.Log("exiting")
wg.Done()
}()
var (
inCh = in
outCh = out
job treeJob
)
outCh = nil
for {
select {
case <-done:
return
case treeID, ok := <-inCh:
if !ok {
return
}
debug.Log("load tree %v", treeID.Str())
tree, err := repo.LoadTree(treeID)
debug.Log("load tree %v (%v) returned err: %v", tree, treeID.Str(), err)
job = treeJob{ID: treeID, error: err, Tree: tree}
outCh = out
inCh = nil
case outCh <- job:
debug.Log("sent tree %v", job.ID.Str())
outCh = nil
inCh = in
}
}
}
func createRandomBlobs(t testing.TB, repo restic.Repository, blobs int, pData float32) {
for i := 0; i < blobs; i++ {
var (
tpe restic.BlobType
length int
)
if rand.Float32() < pData {
tpe = restic.DataBlob
length = randomSize(10*1024, 1024*1024) // 10KiB to 1MiB of data
} else {
tpe = restic.TreeBlob
length = randomSize(1*1024, 20*1024) // 1KiB to 20KiB
}
buf := random(t, length)
id := restic.Hash(buf)
if repo.Index().Has(id, restic.DataBlob) {
t.Errorf("duplicate blob %v/%v ignored", id, restic.DataBlob)
continue
}
_, err := repo.SaveBlob(tpe, buf, id)
if err != nil {
t.Fatalf("SaveFrom() error %v", err)
}
if rand.Float32() < 0.2 {
if err = repo.Flush(); err != nil {
t.Fatalf("repo.Flush() returned error %v", err)
}
}
}
if err := repo.Flush(); err != nil {
t.Fatalf("repo.Flush() returned error %v", err)
}
}
// checkPack reads a pack and checks the integrity of all blobs.
func checkPack(r restic.Repository, id restic.ID) error {
debug.Log("checking pack %v", id.Str())
h := restic.Handle{Type: restic.DataFile, Name: id.String()}
buf, err := backend.LoadAll(r.Backend(), h, nil)
if err != nil {
return err
}
hash := restic.Hash(buf)
if !hash.Equal(id) {
debug.Log("Pack ID does not match, want %v, got %v", id.Str(), hash.Str())
return errors.Errorf("Pack ID does not match, want %v, got %v", id.Str(), hash.Str())
}
blobs, err := pack.List(r.Key(), bytes.NewReader(buf), int64(len(buf)))
if err != nil {
return err
}
var errs []error
for i, blob := range blobs {
debug.Log(" check blob %d: %v", i, blob.ID.Str())
plainBuf := make([]byte, blob.Length)
n, err := crypto.Decrypt(r.Key(), plainBuf, buf[blob.Offset:blob.Offset+blob.Length])
if err != nil {
debug.Log(" error decrypting blob %v: %v", blob.ID.Str(), err)
errs = append(errs, errors.Errorf("blob %v: %v", i, err))
continue
}
plainBuf = plainBuf[:n]
hash := restic.Hash(plainBuf)
if !hash.Equal(blob.ID) {
debug.Log(" Blob ID does not match, want %v, got %v", blob.ID.Str(), hash.Str())
errs = append(errs, errors.Errorf("Blob ID does not match, want %v, got %v", blob.ID.Str(), hash.Str()))
continue
}
}
if len(errs) > 0 {
return errors.Errorf("pack %v contains %v errors: %v", id.Str(), len(errs), errs)
}
return nil
}
func removeLock(repo restic.Repository, id restic.ID) error {
return repo.Backend().Remove(restic.LockFile, id.String())
}
// Load creates an index by loading all index files from the repo.
func Load(repo restic.Repository, p *restic.Progress) (*Index, error) {
debug.Log("loading indexes")
p.Start()
defer p.Done()
done := make(chan struct{})
defer close(done)
supersedes := make(map[restic.ID]restic.IDSet)
results := make(map[restic.ID]map[restic.ID]Pack)
index := newIndex()
for id := range repo.List(restic.IndexFile, done) {
p.Report(restic.Stat{Blobs: 1})
debug.Log("Load index %v", id.Str())
idx, err := loadIndexJSON(repo, id)
if err != nil {
return nil, err
}
res := make(map[restic.ID]Pack)
supersedes[id] = restic.NewIDSet()
for _, sid := range idx.Supersedes {
debug.Log(" index %v supersedes %v", id.Str(), sid)
supersedes[id].Insert(sid)
}
for _, jpack := range idx.Packs {
entries := make([]restic.Blob, 0, len(jpack.Blobs))
for _, blob := range jpack.Blobs {
entry := restic.Blob{
ID: blob.ID,
Type: blob.Type,
Offset: blob.Offset,
Length: blob.Length,
}
entries = append(entries, entry)
}
if err = index.AddPack(jpack.ID, 0, entries); err != nil {
return nil, err
}
}
results[id] = res
index.IndexIDs.Insert(id)
}
for superID, list := range supersedes {
for indexID := range list {
if _, ok := results[indexID]; !ok {
continue
}
debug.Log(" removing index %v, superseded by %v", indexID.Str(), superID.Str())
fmt.Fprintf(os.Stderr, "index %v can be removed, superseded by index %v\n", indexID.Str(), superID.Str())
delete(results, indexID)
}
}
return index, nil
}
// Repack takes a list of packs together with a list of blobs contained in
// these packs. Each pack is loaded and the blobs listed in keepBlobs is saved
// into a new pack. Afterwards, the packs are removed. This operation requires
// an exclusive lock on the repo.
func Repack(repo restic.Repository, packs restic.IDSet, keepBlobs restic.BlobSet) (err error) {
debug.Log("repacking %d packs while keeping %d blobs", len(packs), len(keepBlobs))
buf := make([]byte, 0, maxPackSize)
for packID := range packs {
// load the complete pack
h := restic.Handle{Type: restic.DataFile, Name: packID.String()}
l, err := repo.Backend().Load(h, buf[:cap(buf)], 0)
if errors.Cause(err) == io.ErrUnexpectedEOF {
err = nil
buf = buf[:l]
}
if err != nil {
return err
}
debug.Log("pack %v loaded (%d bytes)", packID.Str(), len(buf))
blobs, err := pack.List(repo.Key(), bytes.NewReader(buf), int64(len(buf)))
if err != nil {
return err
}
debug.Log("processing pack %v, blobs: %v", packID.Str(), len(blobs))
var plaintext []byte
for _, entry := range blobs {
h := restic.BlobHandle{ID: entry.ID, Type: entry.Type}
if !keepBlobs.Has(h) {
continue
}
debug.Log(" process blob %v", h)
ciphertext := buf[entry.Offset : entry.Offset+entry.Length]
plaintext = plaintext[:len(plaintext)]
if len(plaintext) < len(ciphertext) {
plaintext = make([]byte, len(ciphertext))
}
debug.Log(" ciphertext %d, plaintext %d", len(plaintext), len(ciphertext))
n, err := crypto.Decrypt(repo.Key(), plaintext, ciphertext)
if err != nil {
return err
}
plaintext = plaintext[:n]
_, err = repo.SaveBlob(entry.Type, plaintext, entry.ID)
if err != nil {
return err
}
debug.Log(" saved blob %v", entry.ID.Str())
keepBlobs.Delete(h)
}
}
if err := repo.Flush(); err != nil {
return err
}
for packID := range packs {
err := repo.Backend().Remove(restic.DataFile, packID.String())
if err != nil {
debug.Log("error removing pack %v: %v", packID.Str(), err)
return err
}
debug.Log("removed pack %v", packID.Str())
}
return nil
}
func reloadIndex(t *testing.T, repo restic.Repository) {
repo.SetIndex(repository.NewMasterIndex())
if err := repo.LoadIndex(); err != nil {
t.Fatalf("error loading new index: %v", err)
}
}
func saveIndex(t *testing.T, repo restic.Repository) {
if err := repo.SaveIndex(); err != nil {
t.Fatalf("repo.SaveIndex() %v", err)
}
}
// ArchiveReader reads from the reader and archives the data. Returned is the
// resulting snapshot and its ID.
func ArchiveReader(repo restic.Repository, p *restic.Progress, rd io.Reader, name string, tags []string) (*restic.Snapshot, restic.ID, error) {
debug.Log("start archiving %s", name)
sn, err := restic.NewSnapshot([]string{name}, tags)
if err != nil {
return nil, restic.ID{}, err
}
p.Start()
defer p.Done()
chnker := chunker.New(rd, repo.Config().ChunkerPolynomial)
var ids restic.IDs
var fileSize uint64
for {
chunk, err := chnker.Next(getBuf())
if errors.Cause(err) == io.EOF {
break
}
if err != nil {
return nil, restic.ID{}, errors.Wrap(err, "chunker.Next()")
}
id := restic.Hash(chunk.Data)
if !repo.Index().Has(id, restic.DataBlob) {
_, err := repo.SaveBlob(restic.DataBlob, chunk.Data, id)
if err != nil {
return nil, restic.ID{}, err
}
debug.Log("saved blob %v (%d bytes)\n", id.Str(), chunk.Length)
} else {
debug.Log("blob %v already saved in the repo\n", id.Str())
}
freeBuf(chunk.Data)
ids = append(ids, id)
p.Report(restic.Stat{Bytes: uint64(chunk.Length)})
fileSize += uint64(chunk.Length)
}
tree := &restic.Tree{
Nodes: []*restic.Node{
&restic.Node{
Name: name,
AccessTime: time.Now(),
ModTime: time.Now(),
Type: "file",
Mode: 0644,
Size: fileSize,
UID: sn.UID,
GID: sn.GID,
User: sn.Username,
Content: ids,
},
},
}
treeID, err := repo.SaveTree(tree)
if err != nil {
return nil, restic.ID{}, err
}
sn.Tree = &treeID
debug.Log("tree saved as %v", treeID.Str())
id, err := repo.SaveJSONUnpacked(restic.SnapshotFile, sn)
if err != nil {
return nil, restic.ID{}, err
}
debug.Log("snapshot saved as %v", id.Str())
err = repo.Flush()
if err != nil {
return nil, restic.ID{}, err
}
err = repo.SaveIndex()
if err != nil {
return nil, restic.ID{}, err
}
return sn, id, nil
}
func lockExists(repo restic.Repository, t testing.TB, id restic.ID) bool {
exists, err := repo.Backend().Test(restic.LockFile, id.String())
OK(t, err)
return exists
}
