aboutsummaryrefslogtreecommitdiffstats
path: root/vendor/github.com/syndtr/goleveldb/leveldb/db.go
blob: 3655418ad3630584278170ad23fce3dd5b0ca895 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
// Copyright (c) 2012, Suryandaru Triandana <syndtr@gmail.com>
// All rights reserved.
//
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

package leveldb

import (
    "container/list"
    "fmt"
    "io"
    "os"
    "runtime"
    "strings"
    "sync"
    "sync/atomic"
    "time"

    "github.com/syndtr/goleveldb/leveldb/errors"
    "github.com/syndtr/goleveldb/leveldb/iterator"
    "github.com/syndtr/goleveldb/leveldb/journal"
    "github.com/syndtr/goleveldb/leveldb/memdb"
    "github.com/syndtr/goleveldb/leveldb/opt"
    "github.com/syndtr/goleveldb/leveldb/storage"
    "github.com/syndtr/goleveldb/leveldb/table"
    "github.com/syndtr/goleveldb/leveldb/util"
)

// DB is a LevelDB database.
type DB struct {
    // Need 64-bit alignment.
    seq uint64

    // Stats. Need 64-bit alignment.
    cWriteDelay            int64 // The cumulative duration of write delays
    cWriteDelayN           int32 // The cumulative number of write delays
    aliveSnaps, aliveIters int32

    // Session.
    s *session

    // MemDB.
    memMu           sync.RWMutex
    memPool         chan *memdb.DB
    mem, frozenMem  *memDB
    journal         *journal.Writer
    journalWriter   storage.Writer
    journalFd       storage.FileDesc
    frozenJournalFd storage.FileDesc
    frozenSeq       uint64

    // Snapshot.
    snapsMu   sync.Mutex
    snapsList *list.List

    // Write.
    batchPool    sync.Pool
    writeMergeC  chan writeMerge
    writeMergedC chan bool
    writeLockC   chan struct{}
    writeAckC    chan error
    writeDelay   time.Duration
    writeDelayN  int
    tr           *Transaction

    // Compaction.
    compCommitLk     sync.Mutex
    tcompCmdC        chan cCmd
    tcompPauseC      chan chan<- struct{}
    mcompCmdC        chan cCmd
    compErrC         chan error
    compPerErrC      chan error
    compErrSetC      chan error
    compWriteLocking bool
    compStats        cStats
    memdbMaxLevel    int // For testing.

    // Close.
    closeW sync.WaitGroup
    closeC chan struct{}
    closed uint32
    closer io.Closer
}

func openDB(s *session) (*DB, error) {
    s.log("db@open opening")
    start := time.Now()
    db := &DB{
        s: s,
        // Initial sequence
        seq: s.stSeqNum,
        // MemDB
        memPool: make(chan *memdb.DB, 1),
        // Snapshot
        snapsList: list.New(),
        // Write
        batchPool:    sync.Pool{New: newBatch},
        writeMergeC:  make(chan writeMerge),
        writeMergedC: make(chan bool),
        writeLockC:   make(chan struct{}, 1),
        writeAckC:    make(chan error),
        // Compaction
        tcompCmdC:   make(chan cCmd),
        tcompPauseC: make(chan chan<- struct{}),
        mcompCmdC:   make(chan cCmd),
        compErrC:    make(chan error),
        compPerErrC: make(chan error),
        compErrSetC: make(chan error),
        // Close
        closeC: make(chan struct{}),
    }

    // Read-only mode.
    readOnly := s.o.GetReadOnly()

    if readOnly {
        // Recover journals (read-only mode).
        if err := db.recoverJournalRO(); err != nil {
            return nil, err
        }
    } else {
        // Recover journals.
        if err := db.recoverJournal(); err != nil {
            return nil, err
        }

        // Remove any obsolete files.
        if err := db.checkAndCleanFiles(); err != nil {
            // Close journal.
            if db.journal != nil {
                db.journal.Close()
                db.journalWriter.Close()
            }
            return nil, err
        }

    }

    // Doesn't need to be included in the wait group.
    go db.compactionError()
    go db.mpoolDrain()

    if readOnly {
        db.SetReadOnly()
    } else {
        db.closeW.Add(2)
        go db.tCompaction()
        go db.mCompaction()
        // go db.jWriter()
    }

    s.logf("db@open done T·%v", time.Since(start))

    runtime.SetFinalizer(db, (*DB).Close)
    return db, nil
}

// Open opens or creates a DB for the given storage.
// The DB will be created if not exist, unless ErrorIfMissing is true.
// Also, if ErrorIfExist is true and the DB exist Open will returns
// os.ErrExist error.
//
// Open will return an error with type of ErrCorrupted if corruption
// detected in the DB. Use errors.IsCorrupted to test whether an error is
// due to corruption. Corrupted DB can be recovered with Recover function.
//
// The returned DB instance is safe for concurrent use.
// The DB must be closed after use, by calling Close method.
func Open(stor storage.Storage, o *opt.Options) (db *DB, err error) {
    s, err := newSession(stor, o)
    if err != nil {
        return
    }
    defer func() {
        if err != nil {
            s.close()
            s.release()
        }
    }()

    err = s.recover()
    if err != nil {
        if !os.IsNotExist(err) || s.o.GetErrorIfMissing() {
            return
        }
        err = s.create()
        if err != nil {
            return
        }
    } else if s.o.GetErrorIfExist() {
        err = os.ErrExist
        return
    }

    return openDB(s)
}

// OpenFile opens or creates a DB for the given path.
// The DB will be created if not exist, unless ErrorIfMissing is true.
// Also, if ErrorIfExist is true and the DB exist OpenFile will returns
// os.ErrExist error.
//
// OpenFile uses standard file-system backed storage implementation as
// described in the leveldb/storage package.
//
// OpenFile will return an error with type of ErrCorrupted if corruption
// detected in the DB. Use errors.IsCorrupted to test whether an error is
// due to corruption. Corrupted DB can be recovered with Recover function.
//
// The returned DB instance is safe for concurrent use.
// The DB must be closed after use, by calling Close method.
func OpenFile(path string, o *opt.Options) (db *DB, err error) {
    stor, err := storage.OpenFile(path, o.GetReadOnly())
    if err != nil {
        return
    }
    db, err = Open(stor, o)
    if err != nil {
        stor.Close()
    } else {
        db.closer = stor
    }
    return
}

// Recover recovers and opens a DB with missing or corrupted manifest files
// for the given storage. It will ignore any manifest files, valid or not.
// The DB must already exist or it will returns an error.
// Also, Recover will ignore ErrorIfMissing and ErrorIfExist options.
//
// The returned DB instance is safe for concurrent use.
// The DB must be closed after use, by calling Close method.
func Recover(stor storage.Storage, o *opt.Options) (db *DB, err error) {
    s, err := newSession(stor, o)
    if err != nil {
        return
    }
    defer func() {
        if err != nil {
            s.close()
            s.release()
        }
    }()

    err = recoverTable(s, o)
    if err != nil {
        return
    }
    return openDB(s)
}

// RecoverFile recovers and opens a DB with missing or corrupted manifest files
// for the given path. It will ignore any manifest files, valid or not.
// The DB must already exist or it will returns an error.
// Also, Recover will ignore ErrorIfMissing and ErrorIfExist options.
//
// RecoverFile uses standard file-system backed storage implementation as described
// in the leveldb/storage package.
//
// The returned DB instance is safe for concurrent use.
// The DB must be closed after use, by calling Close method.
func RecoverFile(path string, o *opt.Options) (db *DB, err error) {
    stor, err := storage.OpenFile(path, false)
    if err != nil {
        return
    }
    db, err = Recover(stor, o)
    if err != nil {
        stor.Close()
    } else {
        db.closer = stor
    }
    return
}

func recoverTable(s *session, o *opt.Options) error {
    o = dupOptions(o)
    // Mask StrictReader, lets StrictRecovery doing its job.
    o.Strict &= ^opt.StrictReader

    // Get all tables and sort it by file number.
    fds, err := s.stor.List(storage.TypeTable)
    if err != nil {
        return err
    }
    sortFds(fds)

    var (
        maxSeq                                                            uint64
        recoveredKey, goodKey, corruptedKey, corruptedBlock, droppedTable int

        // We will drop corrupted table.
        strict = o.GetStrict(opt.StrictRecovery)
        noSync = o.GetNoSync()

        rec   = &sessionRecord{}
        bpool = util.NewBufferPool(o.GetBlockSize() + 5)
    )
    buildTable := func(iter iterator.Iterator) (tmpFd storage.FileDesc, size int64, err error) {
        tmpFd = s.newTemp()
        writer, err := s.stor.Create(tmpFd)
        if err != nil {
            return
        }
        defer func() {
            writer.Close()
            if err != nil {
                s.stor.Remove(tmpFd)
                tmpFd = storage.FileDesc{}
            }
        }()

        // Copy entries.
        tw := table.NewWriter(writer, o)
        for iter.Next() {
            key := iter.Key()
            if validInternalKey(key) {
                err = tw.Append(key, iter.Value())
                if err != nil {
                    return
                }
            }
        }
        err = iter.Error()
        if err != nil && !errors.IsCorrupted(err) {
            return
        }
        err = tw.Close()
        if err != nil {
            return
        }
        if !noSync {
            err = writer.Sync()
            if err != nil {
                return
            }
        }
        size = int64(tw.BytesLen())
        return
    }
    recoverTable := func(fd storage.FileDesc) error {
        s.logf("table@recovery recovering @%d", fd.Num)
        reader, err := s.stor.Open(fd)
        if err != nil {
            return err
        }
        var closed bool
        defer func() {
            if !closed {
                reader.Close()
            }
        }()

        // Get file size.
        size, err := reader.Seek(0, 2)
        if err != nil {
            return err
        }

        var (
            tSeq                                     uint64
            tgoodKey, tcorruptedKey, tcorruptedBlock int
            imin, imax                               []byte
        )
        tr, err := table.NewReader(reader, size, fd, nil, bpool, o)
        if err != nil {
            return err
        }
        iter := tr.NewIterator(nil, nil)
        if itererr, ok := iter.(iterator.ErrorCallbackSetter); ok {
            itererr.SetErrorCallback(func(err error) {
                if errors.IsCorrupted(err) {
                    s.logf("table@recovery block corruption @%d %q", fd.Num, err)
                    tcorruptedBlock++
                }
            })
        }

        // Scan the table.
        for iter.Next() {
            key := iter.Key()
            _, seq, _, kerr := parseInternalKey(key)
            if kerr != nil {
                tcorruptedKey++
                continue
            }
            tgoodKey++
            if seq > tSeq {
                tSeq = seq
            }
            if imin == nil {
                imin = append([]byte{}, key...)
            }
            imax = append(imax[:0], key...)
        }
        if err := iter.Error(); err != nil && !errors.IsCorrupted(err) {
            iter.Release()
            return err
        }
        iter.Release()

        goodKey += tgoodKey
        corruptedKey += tcorruptedKey
        corruptedBlock += tcorruptedBlock

        if strict && (tcorruptedKey > 0 || tcorruptedBlock > 0) {
            droppedTable++
            s.logf("table@recovery dropped @%d Gk·%d Ck·%d Cb·%d %d %d", fd.Num, tgoodKey, tcorruptedKey, tcorruptedBlock, size, tSeq)
            return nil
        }

        if tgoodKey > 0 {
            if tcorruptedKey > 0 || tcorruptedBlock > 0 {
                // Rebuild the table.
                s.logf("table@recovery rebuilding @%d", fd.Num)
                iter := tr.NewIterator(nil, nil)
                tmpFd, newSize, err := buildTable(iter)
                iter.Release()
                if err != nil {
                    return err
                }
                closed = true
                reader.Close()
                if err := s.stor.Rename(tmpFd, fd); err != nil {
                    return err
                }
                size = newSize
            }
            if tSeq > maxSeq {
                maxSeq = tSeq
            }
            recoveredKey += tgoodKey
            // Add table to level 0.
            rec.addTable(0, fd.Num, size, imin, imax)
            s.logf("table@recovery recovered @%d Gk·%d Ck·%d Cb·%d %d %d", fd.Num, tgoodKey, tcorruptedKey, tcorruptedBlock, size, tSeq)
        } else {
            droppedTable++
            s.logf("table@recovery unrecoverable @%d Ck·%d Cb·%d %d", fd.Num, tcorruptedKey, tcorruptedBlock, size)
        }

        return nil
    }

    // Recover all tables.
    if len(fds) > 0 {
        s.logf("table@recovery F·%d", len(fds))

        // Mark file number as used.
        s.markFileNum(fds[len(fds)-1].Num)

        for _, fd := range fds {
            if err := recoverTable(fd); err != nil {
                return err
            }
        }

        s.logf("table@recovery recovered F·%d %d Gk·%d Ck·%d %d", len(fds), recoveredKey, goodKey, corruptedKey, maxSeq)
    }

    // Set sequence number.
    rec.setSeqNum(maxSeq)

    // Create new manifest.
    if err := s.create(); err != nil {
        return err
    }

    // Commit.
    return s.commit(rec)
}

func (db *DB) recoverJournal() error {
    // Get all journals and sort it by file number.
    rawFds, err := db.s.stor.List(storage.TypeJournal)
    if err != nil {
        return err
    }
    sortFds(rawFds)

    // Journals that will be recovered.
    var fds []storage.FileDesc
    for _, fd := range rawFds {
        if fd.Num >= db.s.stJournalNum || fd.Num == db.s.stPrevJournalNum {
            fds = append(fds, fd)
        }
    }

    var (
        ofd storage.FileDesc // Obsolete file.
        rec = &sessionRecord{}
    )

    // Recover journals.
    if len(fds) > 0 {
        db.logf("journal@recovery F·%d", len(fds))

        // Mark file number as used.
        db.s.markFileNum(fds[len(fds)-1].Num)

        var (
            // Options.
            strict      = db.s.o.GetStrict(opt.StrictJournal)
            checksum    = db.s.o.GetStrict(opt.StrictJournalChecksum)
            writeBuffer = db.s.o.GetWriteBuffer()

            jr       *journal.Reader
            mdb      = memdb.New(db.s.icmp, writeBuffer)
            buf      = &util.Buffer{}
            batchSeq uint64
            batchLen int
        )

        for _, fd := range fds {
            db.logf("journal@recovery recovering @%d", fd.Num)

            fr, err := db.s.stor.Open(fd)
            if err != nil {
                return err
            }

            // Create or reset journal reader instance.
            if jr == nil {
                jr = journal.NewReader(fr, dropper{db.s, fd}, strict, checksum)
            } else {
                jr.Reset(fr, dropper{db.s, fd}, strict, checksum)
            }

            // Flush memdb and remove obsolete journal file.
            if !ofd.Zero() {
                if mdb.Len() > 0 {
                    if _, err := db.s.flushMemdb(rec, mdb, 0); err != nil {
                        fr.Close()
                        return err
                    }
                }

                rec.setJournalNum(fd.Num)
                rec.setSeqNum(db.seq)
                if err := db.s.commit(rec); err != nil {
                    fr.Close()
                    return err
                }
                rec.resetAddedTables()

                db.s.stor.Remove(ofd)
                ofd = storage.FileDesc{}
            }

            // Replay journal to memdb.
            mdb.Reset()
            for {
                r, err := jr.Next()
                if err != nil {
                    if err == io.EOF {
                        break
                    }

                    fr.Close()
                    return errors.SetFd(err, fd)
                }

                buf.Reset()
                if _, err := buf.ReadFrom(r); err != nil {
                    if err == io.ErrUnexpectedEOF {
                        // This is error returned due to corruption, with strict == false.
                        continue
                    }

                    fr.Close()
                    return errors.SetFd(err, fd)
                }
                batchSeq, batchLen, err = decodeBatchToMem(buf.Bytes(), db.seq, mdb)
                if err != nil {
                    if !strict && errors.IsCorrupted(err) {
                        db.s.logf("journal error: %v (skipped)", err)
                        // We won't apply sequence number as it might be corrupted.
                        continue
                    }

                    fr.Close()
                    return errors.SetFd(err, fd)
                }

                // Save sequence number.
                db.seq = batchSeq + uint64(batchLen)

                // Flush it if large enough.
                if mdb.Size() >= writeBuffer {
                    if _, err := db.s.flushMemdb(rec, mdb, 0); err != nil {
                        fr.Close()
                        return err
                    }

                    mdb.Reset()
                }
            }

            fr.Close()
            ofd = fd
        }

        // Flush the last memdb.
        if mdb.Len() > 0 {
            if _, err := db.s.flushMemdb(rec, mdb, 0); err != nil {
                return err
            }
        }
    }

    // Create a new journal.
    if _, err := db.newMem(0); err != nil {
        return err
    }

    // Commit.
    rec.setJournalNum(db.journalFd.Num)
    rec.setSeqNum(db.seq)
    if err := db.s.commit(rec); err != nil {
        // Close journal on error.
        if db.journal != nil {
            db.journal.Close()
            db.journalWriter.Close()
        }
        return err
    }

    // Remove the last obsolete journal file.
    if !ofd.Zero() {
        db.s.stor.Remove(ofd)
    }

    return nil
}

func (db *DB) recoverJournalRO() error {
    // Get all journals and sort it by file number.
    rawFds, err := db.s.stor.List(storage.TypeJournal)
    if err != nil {
        return err
    }
    sortFds(rawFds)

    // Journals that will be recovered.
    var fds []storage.FileDesc
    for _, fd := range rawFds {
        if fd.Num >= db.s.stJournalNum || fd.Num == db.s.stPrevJournalNum {
            fds = append(fds, fd)
        }
    }

    var (
        // Options.
        strict      = db.s.o.GetStrict(opt.StrictJournal)
        checksum    = db.s.o.GetStrict(opt.StrictJournalChecksum)
        writeBuffer = db.s.o.GetWriteBuffer()

        mdb = memdb.New(db.s.icmp, writeBuffer)
    )

    // Recover journals.
    if len(fds) > 0 {
        db.logf("journal@recovery RO·Mode F·%d", len(fds))

        var (
            jr       *journal.Reader
            buf      = &util.Buffer{}
            batchSeq uint64
            batchLen int
        )

        for _, fd := range fds {
            db.logf("journal@recovery recovering @%d", fd.Num)

            fr, err := db.s.stor.Open(fd)
            if err != nil {
                return err
            }

            // Create or reset journal reader instance.
            if jr == nil {
                jr = journal.NewReader(fr, dropper{db.s, fd}, strict, checksum)
            } else {
                jr.Reset(fr, dropper{db.s, fd}, strict, checksum)
            }

            // Replay journal to memdb.
            for {
                r, err := jr.Next()
                if err != nil {
                    if err == io.EOF {
                        break
                    }

                    fr.Close()
                    return errors.SetFd(err, fd)
                }

                buf.Reset()
                if _, err := buf.ReadFrom(r); err != nil {
                    if err == io.ErrUnexpectedEOF {
                        // This is error returned due to corruption, with strict == false.
                        continue
                    }

                    fr.Close()
                    return errors.SetFd(err, fd)
                }
                batchSeq, batchLen, err = decodeBatchToMem(buf.Bytes(), db.seq, mdb)
                if err != nil {
                    if !strict && errors.IsCorrupted(err) {
                        db.s.logf("journal error: %v (skipped)", err)
                        // We won't apply sequence number as it might be corrupted.
                        continue
                    }

                    fr.Close()
                    return errors.SetFd(err, fd)
                }

                // Save sequence number.
                db.seq = batchSeq + uint64(batchLen)
            }

            fr.Close()
        }
    }

    // Set memDB.
    db.mem = &memDB{db: db, DB: mdb, ref: 1}

    return nil
}

func memGet(mdb *memdb.DB, ikey internalKey, icmp *iComparer) (ok bool, mv []byte, err error) {
    mk, mv, err := mdb.Find(ikey)
    if err == nil {
        ukey, _, kt, kerr := parseInternalKey(mk)
        if kerr != nil {
            // Shouldn't have had happen.
            panic(kerr)
        }
        if icmp.uCompare(ukey, ikey.ukey()) == 0 {
            if kt == keyTypeDel {
                return true, nil, ErrNotFound
            }
            return true, mv, nil

        }
    } else if err != ErrNotFound {
        return true, nil, err
    }
    return
}

func (db *DB) get(auxm *memdb.DB, auxt tFiles, key []byte, seq uint64, ro *opt.ReadOptions) (value []byte, err error) {
    ikey := makeInternalKey(nil, key, seq, keyTypeSeek)

    if auxm != nil {
        if ok, mv, me := memGet(auxm, ikey, db.s.icmp); ok {
            return append([]byte{}, mv...), me
        }
    }

    em, fm := db.getMems()
    for _, m := range [...]*memDB{em, fm} {
        if m == nil {
            continue
        }
        defer m.decref()

        if ok, mv, me := memGet(m.DB, ikey, db.s.icmp); ok {
            return append([]byte{}, mv...), me
        }
    }

    v := db.s.version()
    value, cSched, err := v.get(auxt, ikey, ro, false)
    v.release()
    if cSched {
        // Trigger table compaction.
        db.compTrigger(db.tcompCmdC)
    }
    return
}

func nilIfNotFound(err error) error {
    if err == ErrNotFound {
        return nil
    }
    return err
}

func (db *DB) has(auxm *memdb.DB, auxt tFiles, key []byte, seq uint64, ro *opt.ReadOptions) (ret bool, err error) {
    ikey := makeInternalKey(nil, key, seq, keyTypeSeek)

    if auxm != nil {
        if ok, _, me := memGet(auxm, ikey, db.s.icmp); ok {
            return me == nil, nilIfNotFound(me)
        }
    }

    em, fm := db.getMems()
    for _, m := range [...]*memDB{em, fm} {
        if m == nil {
            continue
        }
        defer m.decref()

        if ok, _, me := memGet(m.DB, ikey, db.s.icmp); ok {
            return me == nil, nilIfNotFound(me)
        }
    }

    v := db.s.version()
    _, cSched, err := v.get(auxt, ikey, ro, true)
    v.release()
    if cSched {
        // Trigger table compaction.
        db.compTrigger(db.tcompCmdC)
    }
    if err == nil {
        ret = true
    } else if err == ErrNotFound {
        err = nil
    }
    return
}

// Get gets the value for the given key. It returns ErrNotFound if the
// DB does not contains the key.
//
// The returned slice is its own copy, it is safe to modify the contents
// of the returned slice.
// It is safe to modify the contents of the argument after Get returns.
func (db *DB) Get(key []byte, ro *opt.ReadOptions) (value []byte, err error) {
    err = db.ok()
    if err != nil {
        return
    }

    se := db.acquireSnapshot()
    defer db.releaseSnapshot(se)
    return db.get(nil, nil, key, se.seq, ro)
}

// Has returns true if the DB does contains the given key.
//
// It is safe to modify the contents of the argument after Has returns.
func (db *DB) Has(key []byte, ro *opt.ReadOptions) (ret bool, err error) {
    err = db.ok()
    if err != nil {
        return
    }

    se := db.acquireSnapshot()
    defer db.releaseSnapshot(se)
    return db.has(nil, nil, key, se.seq, ro)
}

// NewIterator returns an iterator for the latest snapshot of the
// underlying DB.
// The returned iterator is not safe for concurrent use, but it is safe to use
// multiple iterators concurrently, with each in a dedicated goroutine.
// It is also safe to use an iterator concurrently with modifying its
// underlying DB. The resultant key/value pairs are guaranteed to be
// consistent.
//
// Slice allows slicing the iterator to only contains keys in the given
// range. A nil Range.Start is treated as a key before all keys in the
// DB. And a nil Range.Limit is treated as a key after all keys in
// the DB.
//
// The iterator must be released after use, by calling Release method.
//
// Also read Iterator documentation of the leveldb/iterator package.
func (db *DB) NewIterator(slice *util.Range, ro *opt.ReadOptions) iterator.Iterator {
    if err := db.ok(); err != nil {
        return iterator.NewEmptyIterator(err)
    }

    se := db.acquireSnapshot()
    defer db.releaseSnapshot(se)
    // Iterator holds 'version' lock, 'version' is immutable so snapshot
    // can be released after iterator created.
    return db.newIterator(nil, nil, se.seq, slice, ro)
}

// GetSnapshot returns a latest snapshot of the underlying DB. A snapshot
// is a frozen snapshot of a DB state at a particular point in time. The
// content of snapshot are guaranteed to be consistent.
//
// The snapshot must be released after use, by calling Release method.
func (db *DB) GetSnapshot() (*Snapshot, error) {
    if err := db.ok(); err != nil {
        return nil, err
    }

    return db.newSnapshot(), nil
}

// GetProperty returns value of the given property name.
//
// Property names:
//  leveldb.num-files-at-level{n}
//      Returns the number of files at level 'n'.
//  leveldb.stats
//      Returns statistics of the underlying DB.
//  leveldb.iostats
//      Returns statistics of effective disk read and write.
//  leveldb.writedelay
//      Returns cumulative write delay caused by compaction.
//  leveldb.sstables
//      Returns sstables list for each level.
//  leveldb.blockpool
//      Returns block pool stats.
//  leveldb.cachedblock
//      Returns size of cached block.
//  leveldb.openedtables
//      Returns number of opened tables.
//  leveldb.alivesnaps
//      Returns number of alive snapshots.
//  leveldb.aliveiters
//      Returns number of alive iterators.
func (db *DB) GetProperty(name string) (value string, err error) {
    err = db.ok()
    if err != nil {
        return
    }

    const prefix = "leveldb."
    if !strings.HasPrefix(name, prefix) {
        return "", ErrNotFound
    }
    p := name[len(prefix):]

    v := db.s.version()
    defer v.release()

    numFilesPrefix := "num-files-at-level"
    switch {
    case strings.HasPrefix(p, numFilesPrefix):
        var level uint
        var rest string
        n, _ := fmt.Sscanf(p[len(numFilesPrefix):], "%d%s", &level, &rest)
        if n != 1 {
            err = ErrNotFound
        } else {
            value = fmt.Sprint(v.tLen(int(level)))
        }
    case p == "stats":
        value = "Compactions\n" +
            " Level |   Tables   |    Size(MB)   |    Time(sec)  |    Read(MB)   |   Write(MB)\n" +
            "-------+------------+---------------+---------------+---------------+---------------\n"
        for level, tables := range v.levels {
            duration, read, write := db.compStats.getStat(level)
            if len(tables) == 0 && duration == 0 {
                continue
            }
            value += fmt.Sprintf(" %3d   | %10d | %13.5f | %13.5f | %13.5f | %13.5f\n",
                level, len(tables), float64(tables.size())/1048576.0, duration.Seconds(),
                float64(read)/1048576.0, float64(write)/1048576.0)
        }
    case p == "iostats":
        value = fmt.Sprintf("Read(MB):%.5f Write(MB):%.5f",
            float64(db.s.stor.reads())/1048576.0,
            float64(db.s.stor.writes())/1048576.0)
    case p == "writedelay":
        writeDelayN, writeDelay := atomic.LoadInt32(&db.cWriteDelayN), time.Duration(atomic.LoadInt64(&db.cWriteDelay))
        value = fmt.Sprintf("DelayN:%d Delay:%s", writeDelayN, writeDelay)
    case p == "sstables":
        for level, tables := range v.levels {
            value += fmt.Sprintf("--- level %d ---\n", level)
            for _, t := range tables {
                value += fmt.Sprintf("%d:%d[%q .. %q]\n", t.fd.Num, t.size, t.imin, t.imax)
            }
        }
    case p == "blockpool":
        value = fmt.Sprintf("%v", db.s.tops.bpool)
    case p == "cachedblock":
        if db.s.tops.bcache != nil {
            value = fmt.Sprintf("%d", db.s.tops.bcache.Size())
        } else {
            value = "<nil>"
        }
    case p == "openedtables":
        value = fmt.Sprintf("%d", db.s.tops.cache.Size())
    case p == "alivesnaps":
        value = fmt.Sprintf("%d", atomic.LoadInt32(&db.aliveSnaps))
    case p == "aliveiters":
        value = fmt.Sprintf("%d", atomic.LoadInt32(&db.aliveIters))
    default:
        err = ErrNotFound
    }

    return
}

// SizeOf calculates approximate sizes of the given key ranges.
// The length of the returned sizes are equal with the length of the given
// ranges. The returned sizes measure storage space usage, so if the user
// data compresses by a factor of ten, the returned sizes will be one-tenth
// the size of the corresponding user data size.
// The results may not include the sizes of recently written data.
func (db *DB) SizeOf(ranges []util.Range) (Sizes, error) {
    if err := db.ok(); err != nil {
        return nil, err
    }

    v := db.s.version()
    defer v.release()

    sizes := make(Sizes, 0, len(ranges))
    for _, r := range ranges {
        imin := makeInternalKey(nil, r.Start, keyMaxSeq, keyTypeSeek)
        imax := makeInternalKey(nil, r.Limit, keyMaxSeq, keyTypeSeek)
        start, err := v.offsetOf(imin)
        if err != nil {
            return nil, err
        }
        limit, err := v.offsetOf(imax)
        if err != nil {
            return nil, err
        }
        var size int64
        if limit >= start {
            size = limit - start
        }
        sizes = append(sizes, size)
    }

    return sizes, nil
}

// Close closes the DB. This will also releases any outstanding snapshot,
// abort any in-flight compaction and discard open transaction.
//
// It is not safe to close a DB until all outstanding iterators are released.
// It is valid to call Close multiple times. Other methods should not be
// called after the DB has been closed.
func (db *DB) Close() error {
    if !db.setClosed() {
        return ErrClosed
    }

    start := time.Now()
    db.log("db@close closing")

    // Clear the finalizer.
    runtime.SetFinalizer(db, nil)

    // Get compaction error.
    var err error
    select {
    case err = <-db.compErrC:
        if err == ErrReadOnly {
            err = nil
        }
    default:
    }

    // Signal all goroutines.
    close(db.closeC)

    // Discard open transaction.
    if db.tr != nil {
        db.tr.Discard()
    }

    // Acquire writer lock.
    db.writeLockC <- struct{}{}

    // Wait for all gorotines to exit.
    db.closeW.Wait()

    // Closes journal.
    if db.journal != nil {
        db.journal.Close()
        db.journalWriter.Close()
        db.journal = nil
        db.journalWriter = nil
    }

    if db.writeDelayN > 0 {
        db.logf("db@write was delayed N·%d %v", db.writeDelayN, db.writeDelay)
    }

    // Close session.
    db.s.close()
    db.logf("db@close done T·%v", time.Since(start))
    db.s.release()

    if db.closer != nil {
        if err1 := db.closer.Close(); err == nil {
            err = err1
        }
        db.closer = nil
    }

    // Clear memdbs.
    db.clearMems()

    return err
}