syncthing/lib/db/set.go

322 lines
10 KiB
Go
Raw Normal View History

2014-11-16 13:13:20 -07:00
// Copyright (C) 2014 The Syncthing Authors.
2014-09-29 12:43:32 -07:00
//
2015-03-07 13:36:35 -07:00
// This Source Code Form is subject to the terms of the Mozilla Public
// License, v. 2.0. If a copy of the MPL was not distributed with this file,
// You can obtain one at https://mozilla.org/MPL/2.0/.
2014-06-01 13:50:14 -07:00
// Package db provides a set type to track local/remote files with newness
// checks. We must do a certain amount of normalization in here. We will get
// fed paths with either native or wire-format separators and encodings
// depending on who calls us. We transform paths to wire-format (NFC and
// slashes) on the way to the database, and transform to native format
// (varying separator and encoding) on the way back out.
package db
import (
"os"
"time"
2015-10-20 06:58:18 -07:00
"github.com/syncthing/syncthing/lib/fs"
2015-08-06 02:29:25 -07:00
"github.com/syncthing/syncthing/lib/osutil"
2015-09-22 10:38:46 -07:00
"github.com/syncthing/syncthing/lib/protocol"
2015-08-06 02:29:25 -07:00
"github.com/syncthing/syncthing/lib/sync"
"github.com/syndtr/goleveldb/leveldb/util"
)
2015-01-12 06:52:24 -07:00
type FileSet struct {
folder string
fs fs.Filesystem
db *instance
meta *metadataTracker
updateMutex sync.Mutex // protects database updates and the corresponding metadata changes
}
// FileIntf is the set of methods implemented by both protocol.FileInfo and
// FileInfoTruncated.
type FileIntf interface {
FileSize() int64
FileName() string
FileLocalFlags() uint32
IsDeleted() bool
IsInvalid() bool
IsIgnored() bool
IsUnsupported() bool
MustRescan() bool
IsReceiveOnlyChanged() bool
IsDirectory() bool
IsSymlink() bool
ShouldConflict() bool
HasPermissionBits() bool
SequenceNo() int64
BlockSize() int
FileVersion() protocol.Vector
}
// The Iterator is called with either a protocol.FileInfo or a
// FileInfoTruncated (depending on the method) and returns true to
// continue iteration, false to stop.
type Iterator func(f FileIntf) bool
var databaseRecheckInterval = 30 * 24 * time.Hour
2015-10-21 00:10:26 -07:00
func init() {
if dur, err := time.ParseDuration(os.Getenv("STRECHECKDBEVERY")); err == nil {
databaseRecheckInterval = dur
2015-10-20 06:58:18 -07:00
}
}
func NewFileSet(folder string, fs fs.Filesystem, ll *Lowlevel) *FileSet {
db := newInstance(ll)
var s = FileSet{
folder: folder,
fs: fs,
db: db,
meta: newMetadataTracker(),
updateMutex: sync.NewMutex(),
2015-10-20 06:58:18 -07:00
}
2015-10-21 00:10:26 -07:00
if err := s.meta.fromDB(db, []byte(folder)); err != nil {
l.Infof("No stored folder metadata for %q: recalculating", folder)
s.recalcCounts()
} else if age := time.Since(s.meta.Created()); age > databaseRecheckInterval {
l.Infof("Stored folder metadata for %q is %v old; recalculating", folder, age)
s.recalcCounts()
2015-10-20 06:58:18 -07:00
}
return &s
}
func (s *FileSet) recalcCounts() {
s.meta = newMetadataTracker()
2015-10-20 06:58:18 -07:00
s.db.checkGlobals([]byte(s.folder), s.meta)
var deviceID protocol.DeviceID
s.db.withAllFolderTruncated([]byte(s.folder), func(device []byte, f FileInfoTruncated) bool {
copy(deviceID[:], device)
s.meta.addFile(deviceID, f)
return true
})
s.meta.SetCreated()
2019-02-02 04:16:27 -07:00
s.meta.toDB(s.db, []byte(s.folder))
}
func (s *FileSet) Drop(device protocol.DeviceID) {
l.Debugf("%s Drop(%v)", s.folder, device)
s.updateMutex.Lock()
defer s.updateMutex.Unlock()
s.db.dropDeviceFolder(device[:], []byte(s.folder), s.meta)
2014-10-07 14:15:01 -07:00
if device == protocol.LocalDeviceID {
s.meta.resetCounts(device)
// We deliberately do not reset the sequence number here. Dropping
// all files for the local device ID only happens in testing - which
// expects the sequence to be retained, like an old Replace() of all
// files would do. However, if we ever did it "in production" we
// would anyway want to retain the sequence for delta indexes to be
// happy.
} else {
// Here, on the other hand, we want to make sure that any file
// announced from the remote is newer than our current sequence
// number.
s.meta.resetAll(device)
2014-10-07 14:15:01 -07:00
}
2019-02-02 04:16:27 -07:00
s.meta.toDB(s.db, []byte(s.folder))
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) Update(device protocol.DeviceID, fs []protocol.FileInfo) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s Update(%v, [%d])", s.folder, device, len(fs))
// do not modify fs in place, it is still used in outer scope
fs = append([]protocol.FileInfo(nil), fs...)
normalizeFilenames(fs)
s.updateMutex.Lock()
defer s.updateMutex.Unlock()
2019-02-02 04:16:27 -07:00
defer s.meta.toDB(s.db, []byte(s.folder))
if device == protocol.LocalDeviceID {
// For the local device we have a bunch of metadata to track.
s.db.updateLocalFiles([]byte(s.folder), fs, s.meta)
return
}
// Easy case, just update the files and we're done.
s.db.updateRemoteFiles([]byte(s.folder), device[:], fs, s.meta)
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) WithNeed(device protocol.DeviceID, fn Iterator) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s WithNeed(%v)", s.folder, device)
s.db.withNeed([]byte(s.folder), device[:], false, nativeFileIterator(fn))
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) WithNeedTruncated(device protocol.DeviceID, fn Iterator) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s WithNeedTruncated(%v)", s.folder, device)
s.db.withNeed([]byte(s.folder), device[:], true, nativeFileIterator(fn))
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) WithHave(device protocol.DeviceID, fn Iterator) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s WithHave(%v)", s.folder, device)
s.db.withHave([]byte(s.folder), device[:], nil, false, nativeFileIterator(fn))
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) WithHaveTruncated(device protocol.DeviceID, fn Iterator) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s WithHaveTruncated(%v)", s.folder, device)
s.db.withHave([]byte(s.folder), device[:], nil, true, nativeFileIterator(fn))
}
func (s *FileSet) WithHaveSequence(startSeq int64, fn Iterator) {
l.Debugf("%s WithHaveSequence(%v)", s.folder, startSeq)
s.db.withHaveSequence([]byte(s.folder), startSeq, nativeFileIterator(fn))
}
// Except for an item with a path equal to prefix, only children of prefix are iterated.
// E.g. for prefix "dir", "dir/file" is iterated, but "dir.file" is not.
func (s *FileSet) WithPrefixedHaveTruncated(device protocol.DeviceID, prefix string, fn Iterator) {
l.Debugf(`%s WithPrefixedHaveTruncated(%v, "%v")`, s.folder, device, prefix)
s.db.withHave([]byte(s.folder), device[:], []byte(osutil.NormalizedFilename(prefix)), true, nativeFileIterator(fn))
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) WithGlobal(fn Iterator) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s WithGlobal()", s.folder)
s.db.withGlobal([]byte(s.folder), nil, false, nativeFileIterator(fn))
2014-08-12 07:17:28 -07:00
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) WithGlobalTruncated(fn Iterator) {
Implement facility based logger, debugging via REST API This implements a new debug/trace infrastructure based on a slightly hacked up logger. Instead of the traditional "if debug { ... }" I've rewritten the logger to have no-op Debugln and Debugf, unless debugging has been enabled for a given "facility". The "facility" is just a string, typically a package name. This will be slightly slower than before; but not that much as it's mostly a function call that returns immediately. For the cases where it matters (the Debugln takes a hex.Dump() of something for example, and it's not in a very occasional "if err != nil" branch) there is an l.ShouldDebug(facility) that is fast enough to be used like the old "if debug". The point of all this is that we can now toggle debugging for the various packages on and off at runtime. There's a new method /rest/system/debug that can be POSTed a set of facilities to enable and disable debug for, or GET from to get a list of facilities with descriptions and their current debug status. Similarly a /rest/system/log?since=... can grab the latest log entries, up to 250 of them (hardcoded constant in main.go) plus the initial few. Not implemented in this commit (but planned) is a simple debug GUI available on /debug that shows the current log in an easily pasteable format and has checkboxes to enable the various debug facilities. The debug instructions to a user then becomes "visit this URL, check these boxes, reproduce your problem, copy and paste the log". The actual log viewer on the hypothetical /debug URL can poll regularly for new log entries and this bypass the 250 line limit. The existing STTRACE=foo variable is still obeyed and just sets the start state of the system.
2015-10-03 08:25:21 -07:00
l.Debugf("%s WithGlobalTruncated()", s.folder)
s.db.withGlobal([]byte(s.folder), nil, true, nativeFileIterator(fn))
2015-02-07 03:52:42 -07:00
}
// Except for an item with a path equal to prefix, only children of prefix are iterated.
// E.g. for prefix "dir", "dir/file" is iterated, but "dir.file" is not.
2015-02-07 03:52:42 -07:00
func (s *FileSet) WithPrefixedGlobalTruncated(prefix string, fn Iterator) {
l.Debugf(`%s WithPrefixedGlobalTruncated("%v")`, s.folder, prefix)
s.db.withGlobal([]byte(s.folder), []byte(osutil.NormalizedFilename(prefix)), true, nativeFileIterator(fn))
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) Get(device protocol.DeviceID, file string) (protocol.FileInfo, bool) {
f, ok := s.db.getFileDirty([]byte(s.folder), device[:], []byte(osutil.NormalizedFilename(file)))
2014-11-05 16:41:51 -07:00
f.Name = osutil.NativeFilename(f.Name)
return f, ok
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) GetGlobal(file string) (protocol.FileInfo, bool) {
fi, ok := s.db.getGlobalDirty([]byte(s.folder), []byte(osutil.NormalizedFilename(file)), false)
2015-01-09 00:41:02 -07:00
if !ok {
return protocol.FileInfo{}, false
}
f := fi.(protocol.FileInfo)
2014-11-05 16:41:51 -07:00
f.Name = osutil.NativeFilename(f.Name)
2015-01-09 00:41:02 -07:00
return f, true
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) GetGlobalTruncated(file string) (FileInfoTruncated, bool) {
fi, ok := s.db.getGlobalDirty([]byte(s.folder), []byte(osutil.NormalizedFilename(file)), true)
2015-01-09 00:41:02 -07:00
if !ok {
return FileInfoTruncated{}, false
}
f := fi.(FileInfoTruncated)
f.Name = osutil.NativeFilename(f.Name)
return f, true
}
2015-01-12 06:52:24 -07:00
func (s *FileSet) Availability(file string) []protocol.DeviceID {
return s.db.availability([]byte(s.folder), []byte(osutil.NormalizedFilename(file)))
}
func (s *FileSet) Sequence(device protocol.DeviceID) int64 {
return s.meta.Sequence(device)
}
func (s *FileSet) LocalSize() Counts {
local := s.meta.Counts(protocol.LocalDeviceID, 0)
recvOnlyChanged := s.meta.Counts(protocol.LocalDeviceID, protocol.FlagLocalReceiveOnly)
return local.Add(recvOnlyChanged)
}
func (s *FileSet) ReceiveOnlyChangedSize() Counts {
return s.meta.Counts(protocol.LocalDeviceID, protocol.FlagLocalReceiveOnly)
2015-10-21 00:10:26 -07:00
}
func (s *FileSet) GlobalSize() Counts {
global := s.meta.Counts(protocol.GlobalDeviceID, 0)
recvOnlyChanged := s.meta.Counts(protocol.GlobalDeviceID, protocol.FlagLocalReceiveOnly)
return global.Add(recvOnlyChanged)
2015-10-21 00:10:26 -07:00
}
func (s *FileSet) IndexID(device protocol.DeviceID) protocol.IndexID {
id := s.db.getIndexID(device[:], []byte(s.folder))
if id == 0 && device == protocol.LocalDeviceID {
// No index ID set yet. We create one now.
id = protocol.NewIndexID()
s.db.setIndexID(device[:], []byte(s.folder), id)
}
return id
}
func (s *FileSet) SetIndexID(device protocol.DeviceID, id protocol.IndexID) {
if device == protocol.LocalDeviceID {
panic("do not explicitly set index ID for local device")
}
s.db.setIndexID(device[:], []byte(s.folder), id)
}
func (s *FileSet) MtimeFS() *fs.MtimeFS {
prefix := s.db.keyer.GenerateMtimesKey(nil, []byte(s.folder))
kv := NewNamespacedKV(s.db.Lowlevel, string(prefix))
return fs.NewMtimeFS(s.fs, kv)
}
func (s *FileSet) ListDevices() []protocol.DeviceID {
return s.meta.devices()
}
// DropFolder clears out all information related to the given folder from the
// database.
func DropFolder(ll *Lowlevel, folder string) {
db := newInstance(ll)
db.dropFolder([]byte(folder))
db.dropMtimes([]byte(folder))
db.dropFolderMeta([]byte(folder))
// Also clean out the folder ID mapping.
db.folderIdx.Delete([]byte(folder))
}
// DropDeltaIndexIDs removes all delta index IDs from the database.
// This will cause a full index transmission on the next connection.
func DropDeltaIndexIDs(db *Lowlevel) {
dbi := db.NewIterator(util.BytesPrefix([]byte{KeyTypeIndexID}), nil)
defer dbi.Release()
for dbi.Next() {
2019-02-02 04:16:27 -07:00
db.Delete(dbi.Key(), nil)
}
}
func normalizeFilenames(fs []protocol.FileInfo) {
for i := range fs {
2014-11-05 16:41:51 -07:00
fs[i].Name = osutil.NormalizedFilename(fs[i].Name)
}
}
func nativeFileIterator(fn Iterator) Iterator {
return func(fi FileIntf) bool {
switch f := fi.(type) {
case protocol.FileInfo:
2014-11-05 16:41:51 -07:00
f.Name = osutil.NativeFilename(f.Name)
return fn(f)
case FileInfoTruncated:
2014-11-05 16:41:51 -07:00
f.Name = osutil.NativeFilename(f.Name)
return fn(f)
default:
panic("unknown interface type")
}
}
}