-
Notifications
You must be signed in to change notification settings - Fork 209
/
rowlocks.c
4189 lines (3592 loc) · 130 KB
/
rowlocks.c
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
/*
Copyright 2015 Bloomberg Finance L.P.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
#include <segstr.h>
#include <list.h>
#include <plhash.h>
#include <fsnapf.h>
#include <net.h>
#include <bdb_int.h>
#include <locks.h>
#include <locks_wrap.h>
#include <ctrace.h>
#include <string.h>
#include <strings.h>
#include <assert.h>
#include <errno.h>
#include <alloca.h>
#include <build/db.h>
#include <build/db_int.h>
#include <dbinc/btree.h>
#include <dbinc/db_am.h>
#include <dbinc/db_page.h>
#include <dbinc/db_shash.h>
#include <dbinc/db_swap.h>
#include <dbinc/lock.h>
#include <dbinc/mp.h>
#include <dbinc/txn.h>
#include <dbinc_auto/db_auto.h>
#include <dbinc_auto/txn_auto.h>
#include <dbinc_auto/txn_ext.h>
#include <llog_auto.h>
#include <llog_ext.h>
#include <llog_handlers.h>
#include <bdb_api.h>
#include <bdb_osqllog.h>
#include <endian_core.h>
#include <printformats.h>
#include <logmsg.h>
#include <tohex.h>
#ifdef NEWSI_STAT
#include <time.h>
#include <sys/time.h>
#include <util.h>
#endif
extern int gbl_dispatch_rowlocks_bench;
/* TODO:
* Old blkseq stuff is not handled.
* Also I am assuming we are never turning on paulbit 67, 16
* to enable the old blkseq scheme. */
/* For compensating log records, remember the LSN we compensated so
we skip it (don't want to undelete a record twice) */
struct remembered_record {
DB_LSN lsn;
};
/* Set to 1 to debug compensation records */
static int debug_comp = 1;
static int release_locks_for_logical_transaction(bdb_state_type *bdb_state,
unsigned long long ltranid);
int logical_commit_replicant(bdb_state_type *bdb_state,
unsigned long long ltranid);
static int logical_release_transaction(bdb_state_type *bdb_state,
unsigned long long ltranid,
int repcommit);
static int undo_physical_transaction(bdb_state_type *bdb_state, tran_type *tran,
DBT *logdta, int *did_something,
DB_LSN *lsn, DB_LSN *prev_lsn_out);
static int undo_upd_dta_lk(bdb_state_type *bdb_state, tran_type *tran,
llog_undo_upd_dta_lk_args *upd_dta_lk,
DB_LSN *undolsn, DB_LSN *prev, int just_load_lsn);
static int undo_upd_ix_lk(bdb_state_type *bdb_state, tran_type *tran,
llog_undo_upd_ix_lk_args *upd_ix_lk, DB_LSN *undolsn,
DB_LSN *prev, int just_load_lsn);
static int print_log_records(bdb_state_type *bdb_state, DB_LSN *lsn)
{
DB_LOGC *cur = NULL;
DBT logent;
DB_LSN nextlsn = *lsn;
u_int32_t rectype;
int rc = 0;
rc = bdb_state->dbenv->log_cursor(bdb_state->dbenv, &cur, 0);
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cursor rc %d\n", __FILE__, __LINE__,
__func__, rc);
return rc;
}
bzero(&logent, sizeof(DBT));
while (nextlsn.file != 0) {
logent.flags = DB_DBT_MALLOC;
rc = cur->get(cur, &nextlsn, &logent, DB_SET);
if (rc == DB_NOTFOUND) {
rc = BDBERR_NO_LOG;
break;
} else if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cur->get(%u:%u) rc %d\n", __FILE__,
__LINE__, __func__, nextlsn.file, nextlsn.offset, rc);
break;
}
if (logent.size < sizeof(u_int32_t)) {
rc = -1;
logmsg(LOGMSG_ERROR,
"%s:%d %s lsn %u:%u unexpected log records size %d\n",
__FILE__, __LINE__, __func__, nextlsn.file, nextlsn.file, rc);
break;
}
LOGCOPY_32(&rectype, logent.data);
/* first 2 words are type + txnid */
logmsg(LOGMSG_USER, "%d\n", rectype);
LOGCOPY_TOLSN(&nextlsn,
(u_int8_t *)logent.data + 2 * sizeof(u_int32_t));
}
if (cur)
cur->close(cur, 0);
return rc;
}
/* These routines inspect the berkeley log starting at the LSN for a logical
transaction record and return
the original key and data that corresponds to the physical transaction that
LSN logs. We do this to avoid
logging large payloads twice every time for the the fairly rare case we need
to undo them. The rules of
the game are:
* berkeley logs an addrem record for the key/data DBTs - there are two
separate log records.
* the key is logged first, then the data
* the key or the data key can be logged as
a) for an add the dbt portion of the contains contains the record (key or
data)
b) for a delete, the dbt portion will be empty, and the hdr portion
contains either a B_KEYDATA or a B_OVERFLOW
structure
* if it's a B_KEYDATA, the payload (key or data) is the dta portion of the
BKEYDATA structure
* if it's a B_OVERFLOW it is followed by one or more 'big' records that
contain the payload
* we process things in reverse, so we'll see the addrem last, and the big
records in the reverse order
* we should only see 2 addrems unless there's been a page split or a reverse
split
* if there is a split, there will be an additional addrem to insert a
reference to a into its parent
page into a parent, which we ignore (it comes before the "real" addrems
that we are interested in,
so we see it later since we are going in reverse
* if there is a reverse split, we see it last, so we can't stop when we see 2
addrems
* a split/reverse split can percolate up the tree, creating more
addrem/split/reverse split records
So to summarize: on a delete we keep running until we hit the begin record,
and the 2 records (key/data pair) we are looking for will be the last 2
addrem records we see. On an add
(undoing_a_delete == false), we stop as soon as we find 2 addrem records.
For every record we reconstruct, we check the size and the genid payload to
make sure
we have sane results. Any error in this subsystem is fatal - if means we are
failing recovery and the
db is potentially corrupt.
*/
static int get_next_addrem_buffer(bdb_state_type *bdb_state, DB_LSN *lsn,
void *buf, int len, int *outlen, int *outpage,
int *outidx, int *have_record,
DB_LSN *nextlsn)
{
void *p = NULL;
int off = 0;
DBT logent;
u_int32_t rectype = 0;
int rc;
BOVERFLOW *ov;
BKEYDATA *kd;
DB_LOGC *cur = NULL;
DB_LSN prevlsn;
DB_LSN savedlsn = *lsn;
int last_was_pgfree = 0;
int n = 0;
__db_addrem_args *addrem_rec;
__db_big_args *big_rec;
rc = bdb_state->dbenv->log_cursor(bdb_state->dbenv, &cur, 0);
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cursor rc %d\n", __FILE__, __LINE__,
__func__, rc);
return rc;
}
bzero(&logent, sizeof(DBT));
rc = 0;
while (rc == 0 && lsn->file != 0) {
if (p) {
__os_free(bdb_state->dbenv, p);
p = NULL;
}
/* TODO: These can't be bigger than a max page size, I think. Should be
able to get
away with a single buffer. */
logent.flags = DB_DBT_MALLOC;
p = NULL;
if (logent.data) {
free(logent.data);
logent.data = NULL;
}
rc = cur->get(cur, lsn, &logent, DB_SET);
if (rc == DB_NOTFOUND) {
rc = BDBERR_NO_LOG;
goto done;
}
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s %d log_cur->get(%u:%u) rc %d\n", __FILE__,
__LINE__, __func__, n, lsn->file, lsn->offset, rc);
break;
}
if (logent.size < sizeof(u_int32_t)) {
rc = -1;
logmsg(LOGMSG_ERROR,
"%s:%d %s lsn %u:%u unexpected log records size %d\n",
__FILE__, __LINE__, __func__, lsn->file, lsn->file, rc);
break;
}
LOGCOPY_32(&rectype, logent.data);
/* first 2 words are type + txnid */
LOGCOPY_TOLSN(&prevlsn,
(u_int8_t *)logent.data + 2 * sizeof(u_int32_t));
*nextlsn = prevlsn;
normalize_rectype(&rectype);
if ((rectype < 10000) && (rectype > 1000)) {
rectype -= 1000;
}
if (rectype == DB___db_pg_free || rectype == DB___db_pg_freedata)
/* pg_free is generating an extra addrem that I don't understand.
* skip it */
last_was_pgfree = 1;
else if (rectype != DB___db_addrem && rectype != DB___db_debug)
last_was_pgfree = 0;
switch (rectype) {
case DB___db_addrem:
__db_addrem_read_int(bdb_state->dbenv, logent.data, 0, &addrem_rec);
p = addrem_rec;
/* skip if last record was a pg_free - this isn't one of our addrems
*/
if (!last_was_pgfree) {
last_was_pgfree = 0;
if (addrem_rec->hdr.data && addrem_rec->hdr.size > 0) {
kd = addrem_rec->hdr.data;
ov = addrem_rec->hdr.data;
} else {
kd = NULL;
ov = NULL;
}
if (addrem_rec->opcode == DB_ADD_DUP &&
(kd == NULL || B_TYPE(kd) == 0)) {
/* If it was an add, the buffer is going to be in the dbt
section of the record.
otherwise it'll refer to an existing page entry and it'll
be in the hdr
section of the record. */
*outlen = addrem_rec->dbt.size;
if (buf && addrem_rec->dbt.size > len) {
*have_record = 0;
} else {
*have_record = 1;
if (buf)
memcpy(buf, addrem_rec->dbt.data,
addrem_rec->dbt.size);
if (outpage)
*outpage = addrem_rec->pgno;
if (outidx)
*outidx = addrem_rec->indx;
}
goto done;
} else {
ov = addrem_rec->hdr.data;
kd = addrem_rec->hdr.data;
if (B_TYPE(kd) == B_OVERFLOW) {
/* Overflows get handles by __db_big type records, but
we need to figure out
the offset here - it's then checked by the __db_big
handler below. */
if (LOG_SWAPPED()) {
M_16_SWAP(ov->unused1);
M_32_SWAP(ov->pgno);
M_32_SWAP(ov->tlen);
}
*outlen = ov->tlen;
off = ov->tlen;
} else if (B_TYPE(kd) == B_KEYDATA) {
/* we don't have off-page duplicates, so this is a real
* entry */
*outlen = kd->len;
if (buf && (kd->len > len)) {
/* huh? */
logmsg(LOGMSG_ERROR, "no record: len %d expected %d %d\n",
kd->len, len, __LINE__);
fsnapf(stdout, kd->data, kd->len);
logmsg(LOGMSG_ERROR, "full:\n");
fsnapf(stdout, addrem_rec->hdr.data,
addrem_rec->hdr.size);
*have_record = 0;
bdb_dump_log(bdb_state->dbenv, &savedlsn);
abort();
} else {
*have_record = 1;
if (buf)
memcpy(buf, kd->data, kd->len);
if (outpage)
*outpage = addrem_rec->pgno;
if (outidx)
*outidx = addrem_rec->indx;
#if 0
int iii;
printf("Retrieving %d len bytes for %d op\n",
kd->len, addrem_rec->opcode);
for (iii=0; iii<kd->len; iii++)
{
printf("%02x ", kd->data[iii]);
}
printf("\n");
#endif
}
goto done;
} else {
logmsg(LOGMSG_ERROR, "Unexpected type %d\n", kd->type);
*have_record = 0;
}
}
}
break;
case DB___db_big:
__db_big_read_int(bdb_state->dbenv, logent.data, 0, &big_rec);
p = big_rec;
off -= big_rec->dbt.size;
if (off < 0) {
logmsg(LOGMSG_ERROR, "huh?\n");
*have_record = 0;
} else {
if (buf)
memcpy(((char *)buf) + off, big_rec->dbt.data,
big_rec->dbt.size);
if (off == 0) {
/* this was the last big record and we have a full buffer */
*have_record = 1;
goto done;
}
}
/* Page-order cursor doesn't apply to blobs. */
break;
default:
break;
}
*lsn = prevlsn;
n++;
}
done:
if (p)
__os_free(bdb_state->dbenv, p);
if (logent.data) {
free(logent.data);
logent.data = NULL;
}
/* if we stopped on an addrem, move cursor to the previous lsn */
if (cur && rc == 0 && (rectype == DB___db_addrem || rectype == DB___db_big))
*lsn = prevlsn;
if (cur)
cur->close(cur, 0);
if (!have_record)
logmsg(LOGMSG_ERROR, "no record\n");
return rc;
}
int bdb_reconstruct_add(bdb_state_type *bdb_state, DB_LSN *startlsn, void *key,
int keylen, void *data, int datalen, int *p_dtalen,
int *p_keylen)
{
int rc;
int have_record = 0;
int outlen;
DB_LSN nextlsn = {1, 0};
DB_LSN lsn = *startlsn;
if (!p_dtalen)
p_dtalen = &outlen;
if (!p_keylen)
p_keylen = &outlen;
rc = get_next_addrem_buffer(bdb_state, &lsn, data, datalen, p_dtalen, NULL,
NULL, &have_record, &nextlsn);
if (rc == BDBERR_NO_LOG)
return rc;
if (rc)
return -1;
rc = get_next_addrem_buffer(bdb_state, &nextlsn, key, keylen, p_keylen,
NULL, NULL, &have_record, &nextlsn);
if (rc == BDBERR_NO_LOG)
return rc;
if (rc)
return -1;
return 0;
}
int bdb_dump_log(DB_ENV *dbenv, DB_LSN *startlsn)
{
DB_LOGC *cur;
int rc;
DBT dbt;
DB_LSN lsn;
struct generic_record {
u_int32_t type;
DB_TXN *txnid;
DB_LSN prev_lsn;
} * generic_record;
/* debug: dump all the events for this record */
logmsg(LOGMSG_USER, "========\n");
logmsg(LOGMSG_USER, "lsn %u:%u\n", startlsn->file, startlsn->offset);
rc = dbenv->log_cursor(dbenv, &cur, 0);
if (rc) {
logmsg(LOGMSG_USER, "log_cursor rc %d\n", rc);
return rc;
}
bzero(&dbt, sizeof(DBT));
dbt.flags = DB_DBT_MALLOC;
lsn = *startlsn;
rc = cur->get(cur, &lsn, &dbt, DB_SET);
while (rc == 0) {
if (dbt.data) {
generic_record = dbt.data;
logmsg(LOGMSG_USER, " %u:%u type %u\n", lsn.file, lsn.offset,
generic_record->type);
lsn = generic_record->prev_lsn;
if (generic_record->type == DB___db_addrem) {
BKEYDATA *kd;
__db_addrem_args *addrem;
__db_addrem_read_int(dbenv, dbt.data, 0, &addrem);
if (addrem->hdr.data)
kd = addrem->hdr.data;
else
kd = NULL;
if (addrem->opcode == DB_ADD_DUP &&
(kd == NULL || B_TYPE(kd) == 0)) {
fsnapf(stdout, addrem->dbt.data, addrem->dbt.size);
} else {
fsnapf(stdout, kd->data, kd->len);
}
free(addrem);
} else if (generic_record->type == DB___db_debug) {
__db_debug_args *debug;
__db_debug_read_int(dbenv, dbt.data, 0, &debug);
logmsg(LOGMSG_USER, "debug: %s", (char *)debug->key.data);
free(debug);
}
free(dbt.data);
} else {
logmsg(LOGMSG_FATAL, "no log record returned?\n");
abort();
}
if (lsn.file == 0 && lsn.offset == 0)
break;
rc = cur->get(cur, &lsn, &dbt, DB_SET);
}
if (rc) {
logmsg(LOGMSG_USER, "log get rc %d at lsn %u:%u\n", rc, lsn.file, lsn.offset);
return -1;
}
logmsg(LOGMSG_USER, "========\n");
return -1;
}
int bdb_reconstruct_delete(bdb_state_type *bdb_state, DB_LSN *startlsn,
int *page, int *index, void *key, int keylen,
void *data, int datalen, int *outdatalen)
{
int outlen[2];
int rc;
unsigned char *buf[2];
int haveit[2] = {0};
int pages[2] = {0};
int indexes[2] = {0};
int alloclen;
int i = 0;
DB_LSN nextlsn;
DB_LSN lsn = *startlsn;
if (keylen > datalen)
alloclen = keylen;
else
alloclen = datalen;
if (alloclen <= 0) {
buf[0] = buf[1] = NULL;
} else if (alloclen <= 1024) {
buf[0] = alloca(alloclen);
buf[1] = alloca(alloclen);
} else {
buf[0] = malloc(alloclen);
buf[1] = malloc(alloclen);
}
/* look for the last two records (ie first in log order) */
do {
i++;
rc = get_next_addrem_buffer(bdb_state, &lsn, buf[i % 2], alloclen,
&outlen[i % 2], &pages[i % 2],
&indexes[i % 2], &haveit[i % 2], &nextlsn);
if (rc == BDBERR_NO_LOG)
return rc;
} while (nextlsn.file != 0 && (!haveit[0] || !haveit[1]));
if (haveit[0] && haveit[1]) {
if (key) {
if (keylen < outlen[i % 2]) {
logmsg(LOGMSG_DEBUG,
"%s: asking for different keylength=%d, replicated=%d\n",
__func__, keylen, outlen[i % 2]);
}
memcpy(key, buf[i % 2], keylen);
}
if (data) {
if (datalen < outlen[(i + 1) % 2]) {
logmsg(LOGMSG_DEBUG,
"%s: asking for different datalength=%d, replicated=%d\n",
__func__, datalen, outlen[(i + 1) % 2]);
}
memcpy(data, buf[(i + 1) % 2], datalen);
if (outdatalen) {
*outdatalen = datalen;
}
}
if (page)
*page = pages[i % 2];
if (index)
*index = indexes[i % 2];
if (datalen > 1024) {
free(buf[0]);
free(buf[1]);
}
return 0;
} else {
if (datalen > 1024) {
free(buf[0]);
free(buf[1]);
}
return 1;
}
}
enum {
PREVIOUS_DATA_IX = 0,
PREVIOUS_KEY_IX = 1,
NEW_DATA_IX = 2,
NEW_KEY_IX = 3
};
int bdb_reconstruct_update(bdb_state_type *bdb_state, DB_LSN *startlsn,
int *page, int *index, void *prevkey,
int *prevkeylen, void *prevdata, int *prevdatalen,
void *newkey, int *newkeylen, void *newdata,
int *newdatalen)
{
int outlen;
int rc;
DB_LSN nextlsn;
DB_LSN lsn = *startlsn;
int tmp_page, tmp_index, haveit;
int pvlen;
/* Get prev data */
again1:
pvlen = prevdatalen ? *prevdatalen : 0;
do {
rc = get_next_addrem_buffer(bdb_state, &lsn, pvlen ? prevdata : NULL,
pvlen, &outlen, &tmp_page, &tmp_index,
&haveit, &nextlsn);
} while (rc == 0 && nextlsn.file != 0 && haveit == 0);
if (rc == BDBERR_NO_LOG || nextlsn.file == 0 || haveit == 0)
return BDBERR_NO_LOG;
if (prevdatalen)
*prevdatalen = outlen;
/* Get prev index */
pvlen = prevkeylen ? *prevkeylen : 0;
rc = get_next_addrem_buffer(bdb_state, &lsn, pvlen ? prevkey : NULL, pvlen,
&outlen, &tmp_page, &tmp_index, &haveit,
&nextlsn);
if (rc == BDBERR_NO_LOG || nextlsn.file == 0)
return rc;
if (!haveit)
goto again1;
if (prevkeylen)
*prevkeylen = outlen;
if (page)
*page = tmp_page;
if (index)
*index = tmp_index;
if (!newkey && !newdata)
return 0;
again2:
pvlen = newdatalen ? *newdatalen : 0;
do {
rc = get_next_addrem_buffer(bdb_state, &lsn, pvlen ? newdata : NULL,
pvlen, &outlen, &tmp_page, &tmp_index,
&haveit, &nextlsn);
} while (rc == 0 && nextlsn.file != 0 && haveit == 0);
if (rc == BDBERR_NO_LOG || nextlsn.file == 0 || haveit == 0)
return BDBERR_NO_LOG;
if (newdatalen)
*newdatalen = outlen;
/* Get newdata */
pvlen = newkeylen ? *newkeylen : 0;
rc = get_next_addrem_buffer(bdb_state, &lsn, pvlen ? newkey : NULL, pvlen,
&outlen, &tmp_page, &tmp_index, &haveit,
&nextlsn);
if (rc == BDBERR_NO_LOG || nextlsn.file == 0)
return rc;
if (!haveit)
goto again2;
if (newkeylen)
*newkeylen = outlen;
return 0;
}
int bdb_reconstruct_key_update(bdb_state_type *bdb_state, DB_LSN *startlsn,
void **diff, int *offset, int *difflen)
{
DB_LSN lsn;
int rc, found_repl = 0;
DBT logent = {0};
__bam_repl_args *repl = NULL;
DB_LOGC *cur = NULL;
u_int32_t rectype;
DB_LSN prevlsn;
/* Key updates don't use addrems unless someone manages to set up 512-byte
key pages. We should probably
disallow that. So just go through the log until you see a repl. */
logent.flags = DB_DBT_MALLOC;
rc = bdb_state->dbenv->log_cursor(bdb_state->dbenv, &cur, 0);
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cursor rc %d\n", __FILE__, __LINE__,
__func__, rc);
return rc;
}
rc = 0;
lsn = *startlsn;
/* printf("%s:%d lsn=[%d:%d]\n", __FILE__,__LINE__, lsn.file,
* lsn.offset);*/
while (rc == 0 && lsn.file != 0) {
if (logent.data) {
free(logent.data);
logent.data = NULL;
}
rc = cur->get(cur, &lsn, &logent, DB_SET);
if (rc == DB_NOTFOUND) {
rc = BDBERR_NO_LOG;
goto done;
}
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cur->get rc %d lsn=[%d:%d]\n",
__FILE__, __LINE__, __func__, rc, lsn.file, lsn.offset);
break;
}
LOGCOPY_32(&rectype, logent.data);
LOGCOPY_TOLSN(&prevlsn,
(u_int8_t *)logent.data + 2 * sizeof(u_int32_t));
normalize_rectype(&rectype);
if ((rectype < 10000) && (rectype > 1000)) {
rectype -= 1000;
}
if (rectype == DB___bam_repl) {
rc = __bam_repl_read_int(bdb_state->dbenv, logent.data, 0, &repl);
if (rc) {
logmsg(LOGMSG_ERROR, "__bam_repl_read rc %d\n", rc);
goto done;
}
*diff = malloc(repl->orig.size);
memcpy(*diff, repl->orig.data, repl->orig.size);
*difflen = repl->orig.size;
*offset = repl->prefix;
found_repl = 1;
break;
}
lsn = prevlsn;
/*printf("%s:%d lsn=[%d:%d]\n", __FILE__,__LINE__, lsn.file,
* lsn.offset);*/
}
done:
if (logent.data) {
free(logent.data);
logent.data = NULL;
}
if (cur)
cur->close(cur, 0);
if (repl)
__os_free(bdb_state->dbenv, repl);
/* Abort if I didn't find it */
assert(found_repl);
(void)found_repl;
return rc;
}
/* Unlike reconstruct-key update, the memory has already been allocated. */
int bdb_reconstruct_inplace_update(bdb_state_type *bdb_state, DB_LSN *startlsn,
void *origd, int *origd_sz, void *newd,
int *newd_sz, int *offset, int *outpage,
int *outidx)
{
int rc = 0, foundit = 0, off = 0;
DB_LSN lsn = *startlsn, prevlsn;
DBT logent = {0};
BOVERFLOW *ov = NULL;
BKEYDATA *kd = NULL;
void *ovcur = NULL;
int *ovlen;
__db_addrem_args *addrem_rec;
__db_big_args *big_rec;
u_int32_t rectype;
__bam_repl_args *repl = NULL;
DB_LOGC *cur = NULL;
logent.flags = DB_DBT_MALLOC;
if (offset)
*offset = 0;
rc = bdb_state->dbenv->log_cursor(bdb_state->dbenv, &cur, 0);
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cursor rc %d\n", __FILE__, __LINE__,
__func__, rc);
return rc;
}
/* Search log records loop. */
while (rc == 0 && lsn.file != 0) {
/* Free previous log record. */
if (logent.data) {
free(logent.data);
logent.data = NULL;
}
/* Grab a log entry. */
rc = cur->get(cur, &lsn, &logent, DB_SET);
if (rc == DB_NOTFOUND) {
rc = BDBERR_NO_LOG;
goto done;
}
if (rc) {
logmsg(LOGMSG_ERROR, "%s:%d %s log_cur->get rc %d lsn=[%d:%d]\n",
__FILE__, __LINE__, __func__, rc, lsn.file, lsn.offset);
break;
}
/* Copy it's type. */
LOGCOPY_32(&rectype, logent.data);
/* Copy it's previous LSN. */
LOGCOPY_TOLSN(&prevlsn,
(u_int8_t *)logent.data + 2 * sizeof(u_int32_t));
normalize_rectype(&rectype);
if ((rectype < 10000) && (rectype > 1000)) {
rectype -= 1000;
}
/* Find a btree-replace log record. */
if (rectype == DB___bam_repl) {
/* Invoke berkeley's 'read-repl-log' routine. */
if (0 != (rc = __bam_repl_read_int(bdb_state->dbenv, logent.data, 0,
&repl))) {
logmsg(LOGMSG_ERROR, "__bam_repl_read rc %d\n", rc);
goto done;
}
/* Copy the reconstructed record. */
if (origd) {
/* Make sure that our sizes agree. */
if (origd_sz && *origd_sz > 0 && *origd_sz != repl->orig.size) {
logmsg(LOGMSG_FATAL,
"%s line %d origd-sz %d != orig-sz %d\n", __func__,
__LINE__, *origd_sz, repl->orig.size);
abort();
}
/* Copy the original data. */
memcpy(origd, repl->orig.data, repl->orig.size);
if (origd_sz)
*origd_sz = repl->orig.size;
}
if (newd) {
if (newd_sz && *newd_sz > 0 && *newd_sz != repl->repl.size) {
logmsg(LOGMSG_FATAL,
"%s line %d newd-sz %d != repl-sz %d\n", __func__,
__LINE__, *newd_sz, repl->repl.size);
abort();
}
memcpy(newd, repl->repl.data, repl->repl.size);
if (newd_sz)
*newd_sz = repl->repl.size;
}
/* Latch the original offset (which should be 0). */
if (offset)
*offset = repl->prefix;
/* Grab the original page number- this won't change on an update. */
if (outpage)
*outpage = repl->pgno;
/* Grab the original page index- this won't change on an update. */
if (outidx)
*outidx = repl->indx;
foundit = 1;
break;
}
/* For big-record updates, we want the addrem that is deleting the
* payload */
if (rectype == DB___db_addrem) {
__db_addrem_read_int(bdb_state->dbenv, logent.data, 0, &addrem_rec);
if (IS_REM_OPCODE(addrem_rec->opcode)) {
ovcur = origd;
ovlen = origd_sz;
} else {
ovcur = newd;
ovlen = newd_sz;
}
kd = NULL;
ov = NULL;
if (addrem_rec->hdr.data && addrem_rec->hdr.size > 0) {
kd = addrem_rec->hdr.data;
ov = addrem_rec->hdr.data;
}
if (addrem_rec->opcode == DB_ADD_DUP &&
(kd == NULL || B_TYPE(kd) == 0)) {
if (ovcur && ovlen && *ovlen > 0)
assert(*ovlen >= addrem_rec->dbt.size);
if (ovcur)
memcpy(ovcur, addrem_rec->dbt.data, addrem_rec->dbt.size);
if (ovlen)
*ovlen = addrem_rec->dbt.size;
if (outpage)
*outpage = addrem_rec->pgno;
if (outidx)
*outidx = addrem_rec->indx;
off = 0;
foundit = 1;
__os_free(bdb_state->dbenv, addrem_rec);
break;
} else {
if (B_TYPE(kd) == B_OVERFLOW) {
if (LOG_SWAPPED()) {
M_16_SWAP(ov->unused1);
M_32_SWAP(ov->pgno);
M_32_SWAP(ov->tlen);
}
off = ov->tlen;
if (ovlen)
*ovlen = ov->tlen;
} else if (B_TYPE(kd) == B_KEYDATA) {
if (ovcur && ovlen && *ovlen > 0)
assert(*ovlen >= kd->len);
if (ovcur)
memcpy(ovcur, kd->data, kd->len);
if (ovlen)
*ovlen = kd->len;
if (outpage)
*outpage = addrem_rec->pgno;
if (outidx)
*outidx = addrem_rec->indx;
off = 0;
foundit = 1;
__os_free(bdb_state->dbenv, addrem_rec);
break;
}
}
}
if (rectype == DB___db_big) {
__db_big_read_int(bdb_state->dbenv, logent.data, 0, &big_rec);
if (big_rec->opcode == DB_REM_BIG) {
off -= big_rec->dbt.size;
assert(off >= 0);
if (ovcur)
memcpy(((char *)ovcur) + off, big_rec->dbt.data,
big_rec->dbt.size);
if (off == 0) {
foundit = 1;
__os_free(bdb_state->dbenv, big_rec);
big_rec = NULL;
if (ovcur == origd && newd != NULL) {
ovlen = NULL;
ovcur = NULL;
} else {
break;
}
}
}
__os_free(bdb_state->dbenv, big_rec);
}