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
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
|
/*****************************************************************************
Copyright (c) 1996, 2017, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2016, 2020, MariaDB Corporation.
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1335 USA
*****************************************************************************/
/**************************************************//**
@file trx/trx0roll.cc
Transaction rollback
Created 3/26/1996 Heikki Tuuri
*******************************************************/
#include "trx0roll.h"
#include <my_service_manager.h>
#include <mysql/service_wsrep.h>
#include "fsp0fsp.h"
#include "lock0lock.h"
#include "mach0data.h"
#include "pars0pars.h"
#include "que0que.h"
#include "row0mysql.h"
#include "row0undo.h"
#include "srv0mon.h"
#include "srv0start.h"
#include "trx0rec.h"
#include "trx0rseg.h"
#include "trx0sys.h"
#include "trx0trx.h"
#include "trx0undo.h"
/** This many pages must be undone before a truncate is tried within
rollback */
static const ulint TRX_ROLL_TRUNC_THRESHOLD = 1;
/** true if trx_rollback_all_recovered() thread is active */
bool trx_rollback_is_active;
/** In crash recovery, the current trx to be rolled back; NULL otherwise */
const trx_t* trx_roll_crash_recv_trx;
/** Finish transaction rollback.
@param[in,out] trx transaction
@return whether the rollback was completed normally
@retval false if the rollback was aborted by shutdown */
static bool trx_rollback_finish(trx_t* trx)
{
trx->mod_tables.clear();
bool finished = trx->error_state == DB_SUCCESS;
if (UNIV_LIKELY(finished)) {
trx_commit(trx);
} else {
ut_a(trx->error_state == DB_INTERRUPTED);
ut_ad(!srv_is_being_started);
ut_a(!srv_undo_sources);
ut_ad(srv_fast_shutdown);
ut_d(trx->in_rollback = false);
if (trx_undo_t*& undo = trx->rsegs.m_redo.old_insert) {
UT_LIST_REMOVE(trx->rsegs.m_redo.rseg->old_insert_list,
undo);
ut_free(undo);
undo = NULL;
}
if (trx_undo_t*& undo = trx->rsegs.m_redo.undo) {
UT_LIST_REMOVE(trx->rsegs.m_redo.rseg->undo_list,
undo);
ut_free(undo);
undo = NULL;
}
if (trx_undo_t*& undo = trx->rsegs.m_noredo.undo) {
UT_LIST_REMOVE(trx->rsegs.m_noredo.rseg->undo_list,
undo);
ut_free(undo);
undo = NULL;
}
trx_commit_low(trx, NULL);
}
trx->lock.que_state = TRX_QUE_RUNNING;
return finished;
}
/*******************************************************************//**
Rollback a transaction used in MySQL. */
static
void
trx_rollback_to_savepoint_low(
/*==========================*/
trx_t* trx, /*!< in: transaction handle */
trx_savept_t* savept) /*!< in: pointer to savepoint undo number, if
partial rollback requested, or NULL for
complete rollback */
{
que_thr_t* thr;
mem_heap_t* heap;
roll_node_t* roll_node;
heap = mem_heap_create(512);
roll_node = roll_node_create(heap);
if (savept != NULL) {
roll_node->savept = savept;
check_trx_state(trx);
} else {
assert_trx_nonlocking_or_in_list(trx);
}
trx->error_state = DB_SUCCESS;
if (trx->has_logged_or_recovered()) {
ut_ad(trx->rsegs.m_redo.rseg != 0
|| trx->rsegs.m_noredo.rseg != 0);
thr = pars_complete_graph_for_exec(roll_node, trx, heap, NULL);
ut_a(thr == que_fork_start_command(
static_cast<que_fork_t*>(que_node_get_parent(thr))));
que_run_threads(thr);
ut_a(roll_node->undo_thr != NULL);
que_run_threads(roll_node->undo_thr);
/* Free the memory reserved by the undo graph. */
que_graph_free(static_cast<que_t*>(
roll_node->undo_thr->common.parent));
}
if (savept == NULL) {
trx_rollback_finish(trx);
MONITOR_INC(MONITOR_TRX_ROLLBACK);
} else {
ut_a(trx->error_state == DB_SUCCESS);
const undo_no_t limit = savept->least_undo_no;
for (trx_mod_tables_t::iterator i = trx->mod_tables.begin();
i != trx->mod_tables.end(); ) {
trx_mod_tables_t::iterator j = i++;
ut_ad(j->second.valid());
if (j->second.rollback(limit)) {
trx->mod_tables.erase(j);
}
}
trx->lock.que_state = TRX_QUE_RUNNING;
MONITOR_INC(MONITOR_TRX_ROLLBACK_SAVEPOINT);
}
mem_heap_free(heap);
/* There might be work for utility threads.*/
srv_active_wake_master_thread();
MONITOR_DEC(MONITOR_TRX_ACTIVE);
}
/*******************************************************************//**
Rollback a transaction to a given savepoint or do a complete rollback.
@return error code or DB_SUCCESS */
dberr_t
trx_rollback_to_savepoint(
/*======================*/
trx_t* trx, /*!< in: transaction handle */
trx_savept_t* savept) /*!< in: pointer to savepoint undo number, if
partial rollback requested, or NULL for
complete rollback */
{
ut_ad(!trx_mutex_own(trx));
trx_start_if_not_started_xa(trx, true);
trx_rollback_to_savepoint_low(trx, savept);
return(trx->error_state);
}
/*******************************************************************//**
Rollback a transaction used in MySQL.
@return error code or DB_SUCCESS */
static
dberr_t
trx_rollback_for_mysql_low(
/*=======================*/
trx_t* trx) /*!< in/out: transaction */
{
trx->op_info = "rollback";
/* If we are doing the XA recovery of prepared transactions,
then the transaction object does not have an InnoDB session
object, and we set a dummy session that we use for all MySQL
transactions. */
trx_rollback_to_savepoint_low(trx, NULL);
trx->op_info = "";
return(trx->error_state);
}
/** Rollback a transaction used in MySQL
@param[in, out] trx transaction
@return error code or DB_SUCCESS */
dberr_t trx_rollback_for_mysql(trx_t* trx)
{
/* We are reading trx->state without holding trx_sys.mutex
here, because the rollback should be invoked for a running
active MySQL transaction (or recovered prepared transaction)
that is associated with the current thread. */
switch (trx->state) {
case TRX_STATE_NOT_STARTED:
trx->will_lock = 0;
ut_ad(trx->mysql_thd);
#ifdef WITH_WSREP
trx->wsrep = false;
#endif
return(DB_SUCCESS);
case TRX_STATE_ACTIVE:
ut_ad(trx->mysql_thd);
assert_trx_nonlocking_or_in_list(trx);
return(trx_rollback_for_mysql_low(trx));
case TRX_STATE_PREPARED:
case TRX_STATE_PREPARED_RECOVERED:
ut_ad(!trx_is_autocommit_non_locking(trx));
if (trx->rsegs.m_redo.undo || trx->rsegs.m_redo.old_insert) {
/* The XA ROLLBACK of a XA PREPARE transaction
will consist of multiple mini-transactions.
As the very first step of XA ROLLBACK, we must
change the undo log state back from
TRX_UNDO_PREPARED to TRX_UNDO_ACTIVE, in order
to ensure that recovery will complete the
rollback.
Failure to perform this step could cause a
situation where we would roll back part of
a XA PREPARE transaction, the server would be
killed, and finally, the transaction would be
recovered in XA PREPARE state, with some of
the actions already having been rolled back. */
ut_ad(!trx->rsegs.m_redo.undo
|| trx->rsegs.m_redo.undo->rseg
== trx->rsegs.m_redo.rseg);
ut_ad(!trx->rsegs.m_redo.old_insert
|| trx->rsegs.m_redo.old_insert->rseg
== trx->rsegs.m_redo.rseg);
mtr_t mtr;
mtr.start();
mutex_enter(&trx->rsegs.m_redo.rseg->mutex);
if (trx_undo_t* undo = trx->rsegs.m_redo.undo) {
trx_undo_set_state_at_prepare(trx, undo, true,
&mtr);
}
if (trx_undo_t* undo = trx->rsegs.m_redo.old_insert) {
trx_undo_set_state_at_prepare(trx, undo, true,
&mtr);
}
mutex_exit(&trx->rsegs.m_redo.rseg->mutex);
/* Write the redo log for the XA ROLLBACK
state change to the global buffer. It is
not necessary to flush the redo log. If
a durable log write of a later mini-transaction
takes place for whatever reason, then this state
change will be durable as well. */
mtr.commit();
ut_ad(mtr.commit_lsn() > 0);
}
return(trx_rollback_for_mysql_low(trx));
case TRX_STATE_COMMITTED_IN_MEMORY:
check_trx_state(trx);
break;
}
ut_error;
return(DB_CORRUPTION);
}
/*******************************************************************//**
Rollback the latest SQL statement for MySQL.
@return error code or DB_SUCCESS */
dberr_t
trx_rollback_last_sql_stat_for_mysql(
/*=================================*/
trx_t* trx) /*!< in/out: transaction */
{
dberr_t err;
/* We are reading trx->state without holding trx_sys.mutex
here, because the statement rollback should be invoked for a
running active MySQL transaction that is associated with the
current thread. */
ut_ad(trx->mysql_thd);
switch (trx->state) {
case TRX_STATE_NOT_STARTED:
return(DB_SUCCESS);
case TRX_STATE_ACTIVE:
assert_trx_nonlocking_or_in_list(trx);
trx->op_info = "rollback of SQL statement";
err = trx_rollback_to_savepoint(
trx, &trx->last_sql_stat_start);
if (trx->fts_trx != NULL) {
fts_savepoint_rollback_last_stmt(trx);
}
/* The following call should not be needed,
but we play it safe: */
trx_mark_sql_stat_end(trx);
trx->op_info = "";
return(err);
case TRX_STATE_PREPARED:
case TRX_STATE_PREPARED_RECOVERED:
case TRX_STATE_COMMITTED_IN_MEMORY:
/* The statement rollback is only allowed on an ACTIVE
transaction, not a PREPARED or COMMITTED one. */
break;
}
ut_error;
return(DB_CORRUPTION);
}
/*******************************************************************//**
Search for a savepoint using name.
@return savepoint if found else NULL */
static
trx_named_savept_t*
trx_savepoint_find(
/*===============*/
trx_t* trx, /*!< in: transaction */
const char* name) /*!< in: savepoint name */
{
trx_named_savept_t* savep;
for (savep = UT_LIST_GET_FIRST(trx->trx_savepoints);
savep != NULL;
savep = UT_LIST_GET_NEXT(trx_savepoints, savep)) {
if (0 == ut_strcmp(savep->name, name)) {
return(savep);
}
}
return(NULL);
}
/*******************************************************************//**
Frees a single savepoint struct. */
static
void
trx_roll_savepoint_free(
/*=====================*/
trx_t* trx, /*!< in: transaction handle */
trx_named_savept_t* savep) /*!< in: savepoint to free */
{
UT_LIST_REMOVE(trx->trx_savepoints, savep);
ut_free(savep->name);
ut_free(savep);
}
/*******************************************************************//**
Frees savepoint structs starting from savep. */
void
trx_roll_savepoints_free(
/*=====================*/
trx_t* trx, /*!< in: transaction handle */
trx_named_savept_t* savep) /*!< in: free all savepoints starting
with this savepoint i*/
{
while (savep != NULL) {
trx_named_savept_t* next_savep;
next_savep = UT_LIST_GET_NEXT(trx_savepoints, savep);
trx_roll_savepoint_free(trx, savep);
savep = next_savep;
}
}
/*******************************************************************//**
Rolls back a transaction back to a named savepoint. Modifications after the
savepoint are undone but InnoDB does NOT release the corresponding locks
which are stored in memory. If a lock is 'implicit', that is, a new inserted
row holds a lock where the lock information is carried by the trx id stored in
the row, these locks are naturally released in the rollback. Savepoints which
were set after this savepoint are deleted.
@return if no savepoint of the name found then DB_NO_SAVEPOINT,
otherwise DB_SUCCESS */
static MY_ATTRIBUTE((nonnull, warn_unused_result))
dberr_t
trx_rollback_to_savepoint_for_mysql_low(
/*====================================*/
trx_t* trx, /*!< in/out: transaction */
trx_named_savept_t* savep, /*!< in/out: savepoint */
int64_t* mysql_binlog_cache_pos)
/*!< out: the MySQL binlog
cache position corresponding
to this savepoint; MySQL needs
this information to remove the
binlog entries of the queries
executed after the savepoint */
{
dberr_t err;
ut_ad(trx_state_eq(trx, TRX_STATE_ACTIVE));
ut_ad(trx->mysql_thd);
/* Free all savepoints strictly later than savep. */
trx_roll_savepoints_free(
trx, UT_LIST_GET_NEXT(trx_savepoints, savep));
*mysql_binlog_cache_pos = savep->mysql_binlog_cache_pos;
trx->op_info = "rollback to a savepoint";
err = trx_rollback_to_savepoint(trx, &savep->savept);
/* Store the current undo_no of the transaction so that
we know where to roll back if we have to roll back the
next SQL statement: */
trx_mark_sql_stat_end(trx);
trx->op_info = "";
#ifdef WITH_WSREP
if (trx->is_wsrep()) {
trx->lock.was_chosen_as_deadlock_victim = false;
}
#endif
return(err);
}
/*******************************************************************//**
Rolls back a transaction back to a named savepoint. Modifications after the
savepoint are undone but InnoDB does NOT release the corresponding locks
which are stored in memory. If a lock is 'implicit', that is, a new inserted
row holds a lock where the lock information is carried by the trx id stored in
the row, these locks are naturally released in the rollback. Savepoints which
were set after this savepoint are deleted.
@return if no savepoint of the name found then DB_NO_SAVEPOINT,
otherwise DB_SUCCESS */
dberr_t
trx_rollback_to_savepoint_for_mysql(
/*================================*/
trx_t* trx, /*!< in: transaction handle */
const char* savepoint_name, /*!< in: savepoint name */
int64_t* mysql_binlog_cache_pos) /*!< out: the MySQL binlog cache
position corresponding to this
savepoint; MySQL needs this
information to remove the
binlog entries of the queries
executed after the savepoint */
{
trx_named_savept_t* savep;
/* We are reading trx->state without holding trx_sys.mutex
here, because the savepoint rollback should be invoked for a
running active MySQL transaction that is associated with the
current thread. */
ut_ad(trx->mysql_thd);
savep = trx_savepoint_find(trx, savepoint_name);
if (savep == NULL) {
return(DB_NO_SAVEPOINT);
}
switch (trx->state) {
case TRX_STATE_NOT_STARTED:
ib::error() << "Transaction has a savepoint "
<< savep->name
<< " though it is not started";
return(DB_ERROR);
case TRX_STATE_ACTIVE:
return(trx_rollback_to_savepoint_for_mysql_low(
trx, savep, mysql_binlog_cache_pos));
case TRX_STATE_PREPARED:
case TRX_STATE_PREPARED_RECOVERED:
case TRX_STATE_COMMITTED_IN_MEMORY:
/* The savepoint rollback is only allowed on an ACTIVE
transaction, not a PREPARED or COMMITTED one. */
break;
}
ut_error;
return(DB_CORRUPTION);
}
/*******************************************************************//**
Creates a named savepoint. If the transaction is not yet started, starts it.
If there is already a savepoint of the same name, this call erases that old
savepoint and replaces it with a new. Savepoints are deleted in a transaction
commit or rollback.
@return always DB_SUCCESS */
dberr_t
trx_savepoint_for_mysql(
/*====================*/
trx_t* trx, /*!< in: transaction handle */
const char* savepoint_name, /*!< in: savepoint name */
int64_t binlog_cache_pos) /*!< in: MySQL binlog cache
position corresponding to this
connection at the time of the
savepoint */
{
trx_named_savept_t* savep;
trx_start_if_not_started_xa(trx, false);
savep = trx_savepoint_find(trx, savepoint_name);
if (savep) {
/* There is a savepoint with the same name: free that */
UT_LIST_REMOVE(trx->trx_savepoints, savep);
ut_free(savep->name);
ut_free(savep);
}
/* Create a new savepoint and add it as the last in the list */
savep = static_cast<trx_named_savept_t*>(
ut_malloc_nokey(sizeof(*savep)));
savep->name = mem_strdup(savepoint_name);
savep->savept = trx_savept_take(trx);
savep->mysql_binlog_cache_pos = binlog_cache_pos;
UT_LIST_ADD_LAST(trx->trx_savepoints, savep);
return(DB_SUCCESS);
}
/*******************************************************************//**
Releases only the named savepoint. Savepoints which were set after this
savepoint are left as is.
@return if no savepoint of the name found then DB_NO_SAVEPOINT,
otherwise DB_SUCCESS */
dberr_t
trx_release_savepoint_for_mysql(
/*============================*/
trx_t* trx, /*!< in: transaction handle */
const char* savepoint_name) /*!< in: savepoint name */
{
trx_named_savept_t* savep;
ut_ad(trx_state_eq(trx, TRX_STATE_ACTIVE, true)
|| trx_state_eq(trx, TRX_STATE_PREPARED, true));
ut_ad(trx->mysql_thd);
savep = trx_savepoint_find(trx, savepoint_name);
if (savep != NULL) {
trx_roll_savepoint_free(trx, savep);
}
return(savep != NULL ? DB_SUCCESS : DB_NO_SAVEPOINT);
}
/*******************************************************************//**
Returns a transaction savepoint taken at this point in time.
@return savepoint */
trx_savept_t
trx_savept_take(
/*============*/
trx_t* trx) /*!< in: transaction */
{
trx_savept_t savept;
savept.least_undo_no = trx->undo_no;
return(savept);
}
/*******************************************************************//**
Roll back an active transaction. */
static
void
trx_rollback_active(
/*================*/
trx_t* trx) /*!< in/out: transaction */
{
mem_heap_t* heap;
que_fork_t* fork;
que_thr_t* thr;
roll_node_t* roll_node;
const trx_id_t trx_id = trx->id;
ut_ad(trx_id);
heap = mem_heap_create(512);
fork = que_fork_create(NULL, NULL, QUE_FORK_RECOVERY, heap);
fork->trx = trx;
thr = que_thr_create(fork, heap, NULL);
roll_node = roll_node_create(heap);
thr->child = roll_node;
roll_node->common.parent = thr;
trx->graph = fork;
ut_a(thr == que_fork_start_command(fork));
trx_roll_crash_recv_trx = trx;
const bool dictionary_locked = trx_get_dict_operation(trx)
!= TRX_DICT_OP_NONE;
if (dictionary_locked) {
row_mysql_lock_data_dictionary(trx);
}
que_run_threads(thr);
ut_a(roll_node->undo_thr != NULL);
que_run_threads(roll_node->undo_thr);
que_graph_free(
static_cast<que_t*>(roll_node->undo_thr->common.parent));
if (UNIV_UNLIKELY(!trx_rollback_finish(trx))) {
ut_ad(!dictionary_locked);
goto func_exit;
}
ut_a(trx->lock.que_state == TRX_QUE_RUNNING);
if (!dictionary_locked || !trx->table_id) {
} else if (dict_table_t* table = dict_table_open_on_id(
trx->table_id, TRUE, DICT_TABLE_OP_NORMAL)) {
ib::info() << "Dropping table " << table->name
<< ", with id " << trx->table_id
<< " in recovery";
dict_table_close_and_drop(trx, table);
trx_commit_for_mysql(trx);
}
ib::info() << "Rolled back recovered transaction " << trx_id;
func_exit:
if (dictionary_locked) {
row_mysql_unlock_data_dictionary(trx);
}
mem_heap_free(heap);
trx_roll_crash_recv_trx = NULL;
}
struct trx_roll_count_callback_arg
{
uint32_t n_trx;
uint64_t n_rows;
trx_roll_count_callback_arg(): n_trx(0), n_rows(0) {}
};
static my_bool trx_roll_count_callback(rw_trx_hash_element_t *element,
trx_roll_count_callback_arg *arg)
{
mutex_enter(&element->mutex);
if (trx_t *trx= element->trx)
{
if (trx->is_recovered && trx_state_eq(trx, TRX_STATE_ACTIVE))
{
arg->n_trx++;
arg->n_rows+= trx->undo_no;
}
}
mutex_exit(&element->mutex);
return 0;
}
/** Report progress when rolling back a row of a recovered transaction. */
void trx_roll_report_progress()
{
time_t now = time(NULL);
mutex_enter(&recv_sys->mutex);
bool report = recv_sys->report(now);
mutex_exit(&recv_sys->mutex);
if (report) {
trx_roll_count_callback_arg arg;
/* Get number of recovered active transactions and number of
rows they modified. Numbers must be accurate, because only this
thread is allowed to touch recovered transactions. */
trx_sys.rw_trx_hash.iterate_no_dups(
reinterpret_cast<my_hash_walk_action>
(trx_roll_count_callback), &arg);
if (arg.n_rows > 0) {
service_manager_extend_timeout(
INNODB_EXTEND_TIMEOUT_INTERVAL,
"To roll back: " UINT32PF " transactions, "
UINT64PF " rows", arg.n_trx, arg.n_rows);
}
ib::info() << "To roll back: " << arg.n_trx
<< " transactions, " << arg.n_rows << " rows";
}
}
static my_bool trx_rollback_recovered_callback(rw_trx_hash_element_t *element,
std::vector<trx_t*> *trx_list)
{
mutex_enter(&element->mutex);
if (trx_t *trx= element->trx)
{
mutex_enter(&trx->mutex);
if (trx_state_eq(trx, TRX_STATE_ACTIVE) && trx->is_recovered)
trx_list->push_back(trx);
mutex_exit(&trx->mutex);
}
mutex_exit(&element->mutex);
return 0;
}
/**
Rollback any incomplete transactions which were encountered in crash recovery.
If the transaction already was committed, then we clean up a possible insert
undo log. If the transaction was not yet committed, then we roll it back.
Note: For XA recovered transactions, we rely on MySQL to
do rollback. They will be in TRX_STATE_PREPARED state. If the server
is shutdown and they are still lingering in trx_sys_t::trx_list
then the shutdown will hang.
@param[in] all true=roll back all recovered active transactions;
false=roll back any incomplete dictionary transaction
*/
void trx_rollback_recovered(bool all)
{
std::vector<trx_t*> trx_list;
ut_a(srv_force_recovery < SRV_FORCE_NO_TRX_UNDO);
/*
Collect list of recovered ACTIVE transaction ids first. Once collected, no
other thread is allowed to modify or remove these transactions from
rw_trx_hash.
*/
trx_sys.rw_trx_hash.iterate_no_dups(reinterpret_cast<my_hash_walk_action>
(trx_rollback_recovered_callback),
&trx_list);
while (!trx_list.empty())
{
trx_t *trx= trx_list.back();
trx_list.pop_back();
ut_ad(trx);
ut_d(trx_mutex_enter(trx));
ut_ad(trx->is_recovered);
ut_ad(trx_state_eq(trx, TRX_STATE_ACTIVE));
ut_d(trx_mutex_exit(trx));
if (!srv_is_being_started && !srv_undo_sources && srv_fast_shutdown)
goto discard;
if (all || trx_get_dict_operation(trx) != TRX_DICT_OP_NONE)
{
trx_rollback_active(trx);
if (trx->error_state != DB_SUCCESS)
{
ut_ad(trx->error_state == DB_INTERRUPTED);
trx->error_state= DB_SUCCESS;
ut_ad(!srv_undo_sources);
ut_ad(srv_fast_shutdown);
discard:
/* Note: before kill_server() invoked innobase_end() via
unireg_end(), it invoked close_connections(), which should initiate
the rollback of any user transactions via THD::cleanup() in the
connection threads, and wait for all THD::cleanup() to complete.
So, no active user transactions should exist at this point.
srv_undo_sources=false was cleared early in innobase_end().
Generally, the server guarantees that all connections using
InnoDB must be disconnected by the time we are reaching this code,
be it during shutdown or UNINSTALL PLUGIN.
Because there is no possible race condition with any
concurrent user transaction, we do not have to invoke
trx->commit_state() or wait for !trx->is_referenced()
before trx_sys.deregister_rw(trx). */
trx_sys.deregister_rw(trx);
trx_free_at_shutdown(trx);
}
else
trx_free(trx);
}
}
}
/*******************************************************************//**
Rollback or clean up any incomplete transactions which were
encountered in crash recovery. If the transaction already was
committed, then we clean up a possible insert undo log. If the
transaction was not yet committed, then we roll it back.
Note: this is done in a background thread.
@return a dummy parameter */
extern "C"
os_thread_ret_t
DECLARE_THREAD(trx_rollback_all_recovered)(void*)
{
my_thread_init();
ut_ad(!srv_read_only_mode);
#ifdef UNIV_PFS_THREAD
pfs_register_thread(trx_rollback_clean_thread_key);
#endif /* UNIV_PFS_THREAD */
if (trx_sys.rw_trx_hash.size()) {
ib::info() << "Starting in background the rollback of"
" recovered transactions";
trx_rollback_recovered(true);
ib::info() << "Rollback of non-prepared transactions"
" completed";
}
trx_rollback_is_active = false;
my_thread_end();
/* We count the number of threads in os_thread_exit(). A created
thread should always use that to exit and not use return() to exit. */
os_thread_exit();
OS_THREAD_DUMMY_RETURN;
}
/** Try to truncate the undo logs.
@param[in,out] trx transaction */
static
void
trx_roll_try_truncate(trx_t* trx)
{
trx->pages_undone = 0;
undo_no_t undo_no = trx->undo_no;
if (trx_undo_t* undo = trx->rsegs.m_redo.undo) {
ut_ad(undo->rseg == trx->rsegs.m_redo.rseg);
mutex_enter(&undo->rseg->mutex);
trx_undo_truncate_end(undo, undo_no, false);
mutex_exit(&undo->rseg->mutex);
}
if (trx_undo_t* undo = trx->rsegs.m_noredo.undo) {
ut_ad(undo->rseg == trx->rsegs.m_noredo.rseg);
mutex_enter(&undo->rseg->mutex);
trx_undo_truncate_end(undo, undo_no, true);
mutex_exit(&undo->rseg->mutex);
}
}
/***********************************************************************//**
Pops the topmost undo log record in a single undo log and updates the info
about the topmost record in the undo log memory struct.
@return undo log record, the page s-latched */
static
trx_undo_rec_t*
trx_roll_pop_top_rec(
/*=================*/
trx_t* trx, /*!< in: transaction */
trx_undo_t* undo, /*!< in: undo log */
mtr_t* mtr) /*!< in: mtr */
{
page_t* undo_page = trx_undo_page_get_s_latched(
page_id_t(undo->rseg->space->id, undo->top_page_no), mtr);
ulint offset = undo->top_offset;
trx_undo_rec_t* prev_rec = trx_undo_get_prev_rec(
undo_page + offset, undo->hdr_page_no, undo->hdr_offset,
true, mtr);
if (prev_rec == NULL) {
undo->top_undo_no = IB_ID_MAX;
ut_ad(undo->empty());
} else {
page_t* prev_rec_page = page_align(prev_rec);
if (prev_rec_page != undo_page) {
trx->pages_undone++;
}
undo->top_page_no = page_get_page_no(prev_rec_page);
undo->top_offset = ulint(prev_rec - prev_rec_page);
undo->top_undo_no = trx_undo_rec_get_undo_no(prev_rec);
ut_ad(!undo->empty());
}
return(undo_page + offset);
}
/** Get the last undo log record of a transaction (for rollback).
@param[in,out] trx transaction
@param[out] roll_ptr DB_ROLL_PTR to the undo record
@param[in,out] heap memory heap for allocation
@return undo log record copied to heap
@retval NULL if none left or the roll_limit (savepoint) was reached */
trx_undo_rec_t*
trx_roll_pop_top_rec_of_trx(trx_t* trx, roll_ptr_t* roll_ptr, mem_heap_t* heap)
{
if (trx->pages_undone >= TRX_ROLL_TRUNC_THRESHOLD) {
trx_roll_try_truncate(trx);
}
trx_undo_t* undo = NULL;
trx_undo_t* insert = trx->rsegs.m_redo.old_insert;
trx_undo_t* update = trx->rsegs.m_redo.undo;
trx_undo_t* temp = trx->rsegs.m_noredo.undo;
const undo_no_t limit = trx->roll_limit;
ut_ad(!insert || !update || insert->empty() || update->empty()
|| insert->top_undo_no != update->top_undo_no);
ut_ad(!insert || !temp || insert->empty() || temp->empty()
|| insert->top_undo_no != temp->top_undo_no);
ut_ad(!update || !temp || update->empty() || temp->empty()
|| update->top_undo_no != temp->top_undo_no);
if (UNIV_LIKELY_NULL(insert)
&& !insert->empty() && limit <= insert->top_undo_no) {
undo = insert;
}
if (update && !update->empty() && update->top_undo_no >= limit) {
if (!undo) {
undo = update;
} else if (undo->top_undo_no < update->top_undo_no) {
undo = update;
}
}
if (temp && !temp->empty() && temp->top_undo_no >= limit) {
if (!undo) {
undo = temp;
} else if (undo->top_undo_no < temp->top_undo_no) {
undo = temp;
}
}
if (undo == NULL) {
trx_roll_try_truncate(trx);
/* Mark any ROLLBACK TO SAVEPOINT completed, so that
if the transaction object is committed and reused
later, we will default to a full ROLLBACK. */
trx->roll_limit = 0;
trx->in_rollback = false;
return(NULL);
}
ut_ad(!undo->empty());
ut_ad(limit <= undo->top_undo_no);
*roll_ptr = trx_undo_build_roll_ptr(
false, undo->rseg->id, undo->top_page_no, undo->top_offset);
mtr_t mtr;
mtr.start();
trx_undo_rec_t* undo_rec = trx_roll_pop_top_rec(trx, undo, &mtr);
const undo_no_t undo_no = trx_undo_rec_get_undo_no(undo_rec);
switch (trx_undo_rec_get_type(undo_rec)) {
case TRX_UNDO_INSERT_METADATA:
/* This record type was introduced in MDEV-11369
instant ADD COLUMN, which was implemented after
MDEV-12288 removed the insert_undo log. There is no
instant ADD COLUMN for temporary tables. Therefore,
this record can only be present in the main undo log. */
ut_ad(undo == update);
/* fall through */
case TRX_UNDO_RENAME_TABLE:
ut_ad(undo == insert || undo == update);
/* fall through */
case TRX_UNDO_INSERT_REC:
ut_ad(undo == insert || undo == update || undo == temp);
*roll_ptr |= 1ULL << ROLL_PTR_INSERT_FLAG_POS;
break;
default:
ut_ad(undo == update || undo == temp);
break;
}
trx->undo_no = undo_no;
trx_undo_rec_t* undo_rec_copy = trx_undo_rec_copy(undo_rec, heap);
mtr.commit();
return(undo_rec_copy);
}
/****************************************************************//**
Builds an undo 'query' graph for a transaction. The actual rollback is
performed by executing this query graph like a query subprocedure call.
The reply about the completion of the rollback will be sent by this
graph.
@return own: the query graph */
static
que_t*
trx_roll_graph_build(
/*=================*/
trx_t* trx) /*!< in/out: transaction */
{
mem_heap_t* heap;
que_fork_t* fork;
que_thr_t* thr;
ut_ad(trx_mutex_own(trx));
heap = mem_heap_create(512);
fork = que_fork_create(NULL, NULL, QUE_FORK_ROLLBACK, heap);
fork->trx = trx;
thr = que_thr_create(fork, heap, NULL);
thr->child = row_undo_node_create(trx, thr, heap);
return(fork);
}
/*********************************************************************//**
Starts a rollback operation, creates the UNDO graph that will do the
actual undo operation.
@return query graph thread that will perform the UNDO operations. */
static
que_thr_t*
trx_rollback_start(
/*===============*/
trx_t* trx, /*!< in: transaction */
undo_no_t roll_limit) /*!< in: rollback to undo no (for
partial undo), 0 if we are rolling back
the entire transaction */
{
ut_ad(trx_mutex_own(trx));
/* Initialize the rollback field in the transaction */
ut_ad(!trx->roll_limit);
ut_ad(!trx->in_rollback);
trx->roll_limit = roll_limit;
trx->in_rollback = true;
ut_a(trx->roll_limit <= trx->undo_no);
trx->pages_undone = 0;
/* Build a 'query' graph which will perform the undo operations */
que_t* roll_graph = trx_roll_graph_build(trx);
trx->graph = roll_graph;
trx->lock.que_state = TRX_QUE_ROLLING_BACK;
return(que_fork_start_command(roll_graph));
}
/*********************************************************************//**
Creates a rollback command node struct.
@return own: rollback node struct */
roll_node_t*
roll_node_create(
/*=============*/
mem_heap_t* heap) /*!< in: mem heap where created */
{
roll_node_t* node;
node = static_cast<roll_node_t*>(mem_heap_zalloc(heap, sizeof(*node)));
node->state = ROLL_NODE_SEND;
node->common.type = QUE_NODE_ROLLBACK;
return(node);
}
/***********************************************************//**
Performs an execution step for a rollback command node in a query graph.
@return query thread to run next, or NULL */
que_thr_t*
trx_rollback_step(
/*==============*/
que_thr_t* thr) /*!< in: query thread */
{
roll_node_t* node;
node = static_cast<roll_node_t*>(thr->run_node);
ut_ad(que_node_get_type(node) == QUE_NODE_ROLLBACK);
if (thr->prev_node == que_node_get_parent(node)) {
node->state = ROLL_NODE_SEND;
}
if (node->state == ROLL_NODE_SEND) {
trx_t* trx;
ib_id_t roll_limit;
trx = thr_get_trx(thr);
trx_mutex_enter(trx);
node->state = ROLL_NODE_WAIT;
ut_a(node->undo_thr == NULL);
roll_limit = node->savept ? node->savept->least_undo_no : 0;
trx_commit_or_rollback_prepare(trx);
node->undo_thr = trx_rollback_start(trx, roll_limit);
trx_mutex_exit(trx);
} else {
ut_ad(node->state == ROLL_NODE_WAIT);
thr->run_node = que_node_get_parent(node);
}
return(thr);
}
|