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
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
|
/*-
* Copyright (c) 2014-2018 MongoDB, Inc.
* Copyright (c) 2008-2014 WiredTiger, Inc.
* All rights reserved.
*
* See the file LICENSE for redistribution information.
*/
#include "wt_internal.h"
/*
* __snapsort_partition --
* Custom quick sort partitioning for snapshots.
*/
static uint32_t
__snapsort_partition(uint64_t *array, uint32_t f, uint32_t l, uint64_t pivot)
{
uint32_t i, j;
i = f - 1;
j = l + 1;
for (;;) {
while (pivot < array[--j])
;
while (array[++i] < pivot)
;
if (i < j) {
uint64_t tmp = array[i];
array[i] = array[j];
array[j] = tmp;
} else
return (j);
}
}
/*
* __snapsort_impl --
* Custom quick sort implementation for snapshots.
*/
static void
__snapsort_impl(uint64_t *array, uint32_t f, uint32_t l)
{
while (f + 16 < l) {
uint64_t v1 = array[f], v2 = array[l], v3 = array[(f + l)/2];
uint64_t median = v1 < v2 ?
(v3 < v1 ? v1 : WT_MIN(v2, v3)) :
(v3 < v2 ? v2 : WT_MIN(v1, v3));
uint32_t m = __snapsort_partition(array, f, l, median);
__snapsort_impl(array, f, m);
f = m + 1;
}
}
/*
* __snapsort --
* Sort an array of transaction IDs.
*/
static void
__snapsort(uint64_t *array, uint32_t size)
{
__snapsort_impl(array, 0, size - 1);
WT_INSERTION_SORT(array, size, uint64_t, WT_TXNID_LT);
}
/*
* __txn_remove_from_global_table --
* Remove the txn id from the global txn table.
*/
static inline void
__txn_remove_from_global_table(WT_SESSION_IMPL *session)
{
#ifdef HAVE_DIAGNOSTIC
WT_TXN *txn;
WT_TXN_GLOBAL *txn_global;
WT_TXN_STATE *txn_state;
txn = &session->txn;
txn_global = &S2C(session)->txn_global;
txn_state = WT_SESSION_TXN_STATE(session);
WT_ASSERT(session, !WT_TXNID_LT(txn->id, txn_global->last_running));
WT_ASSERT(session,
txn->id != WT_TXN_NONE && txn_state->id != WT_TXN_NONE);
#else
WT_TXN_STATE *txn_state;
txn_state = WT_SESSION_TXN_STATE(session);
#endif
WT_PUBLISH(txn_state->id, WT_TXN_NONE);
}
/*
* __txn_sort_snapshot --
* Sort a snapshot for faster searching and set the min/max bounds.
*/
static void
__txn_sort_snapshot(WT_SESSION_IMPL *session, uint32_t n, uint64_t snap_max)
{
WT_TXN *txn;
txn = &session->txn;
if (n > 1)
__snapsort(txn->snapshot, n);
txn->snapshot_count = n;
txn->snap_max = snap_max;
txn->snap_min = (n > 0 && WT_TXNID_LE(txn->snapshot[0], snap_max)) ?
txn->snapshot[0] : snap_max;
F_SET(txn, WT_TXN_HAS_SNAPSHOT);
WT_ASSERT(session, n == 0 || txn->snap_min != WT_TXN_NONE);
}
/*
* __wt_txn_release_snapshot --
* Release the snapshot in the current transaction.
*/
void
__wt_txn_release_snapshot(WT_SESSION_IMPL *session)
{
WT_TXN *txn;
WT_TXN_STATE *txn_state;
txn = &session->txn;
txn_state = WT_SESSION_TXN_STATE(session);
WT_ASSERT(session,
txn_state->pinned_id == WT_TXN_NONE ||
session->txn.isolation == WT_ISO_READ_UNCOMMITTED ||
!__wt_txn_visible_all(session, txn_state->pinned_id, NULL));
txn_state->metadata_pinned = txn_state->pinned_id = WT_TXN_NONE;
F_CLR(txn, WT_TXN_HAS_SNAPSHOT);
}
/*
* __wt_txn_get_snapshot --
* Allocate a snapshot.
*/
void
__wt_txn_get_snapshot(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
WT_TXN *txn;
WT_TXN_GLOBAL *txn_global;
WT_TXN_STATE *s, *txn_state;
uint64_t current_id, id;
uint64_t prev_oldest_id, pinned_id;
uint32_t i, n, session_cnt;
conn = S2C(session);
txn = &session->txn;
txn_global = &conn->txn_global;
txn_state = WT_SESSION_TXN_STATE(session);
n = 0;
/* We're going to scan the table: wait for the lock. */
__wt_readlock(session, &txn_global->rwlock);
current_id = pinned_id = txn_global->current;
prev_oldest_id = txn_global->oldest_id;
/*
* Include the checkpoint transaction, if one is running: we should
* ignore any uncommitted changes the checkpoint has written to the
* metadata. We don't have to keep the checkpoint's changes pinned so
* don't including it in the published pinned ID.
*/
if ((id = txn_global->checkpoint_state.id) != WT_TXN_NONE) {
txn->snapshot[n++] = id;
txn_state->metadata_pinned = id;
}
/* For pure read-only workloads, avoid scanning. */
if (prev_oldest_id == current_id) {
txn_state->pinned_id = current_id;
/* Check that the oldest ID has not moved in the meantime. */
WT_ASSERT(session, prev_oldest_id == txn_global->oldest_id);
goto done;
}
/* Walk the array of concurrent transactions. */
WT_ORDERED_READ(session_cnt, conn->session_cnt);
for (i = 0, s = txn_global->states; i < session_cnt; i++, s++) {
/*
* Build our snapshot of any concurrent transaction IDs.
*
* Ignore:
* - Our own ID: we always read our own updates.
* - The ID if it is older than the oldest ID we saw. This
* can happen if we race with a thread that is allocating
* an ID -- the ID will not be used because the thread will
* keep spinning until it gets a valid one.
*/
if (s != txn_state &&
(id = s->id) != WT_TXN_NONE &&
WT_TXNID_LE(prev_oldest_id, id)) {
txn->snapshot[n++] = id;
if (WT_TXNID_LT(id, pinned_id))
pinned_id = id;
}
}
/*
* If we got a new snapshot, update the published pinned ID for this
* session.
*/
WT_ASSERT(session, WT_TXNID_LE(prev_oldest_id, pinned_id));
WT_ASSERT(session, prev_oldest_id == txn_global->oldest_id);
txn_state->pinned_id = pinned_id;
done: __wt_readunlock(session, &txn_global->rwlock);
__txn_sort_snapshot(session, n, current_id);
}
/*
* __txn_oldest_scan --
* Sweep the running transactions to calculate the oldest ID required.
*/
static void
__txn_oldest_scan(WT_SESSION_IMPL *session,
uint64_t *oldest_idp, uint64_t *last_runningp, uint64_t *metadata_pinnedp,
WT_SESSION_IMPL **oldest_sessionp)
{
WT_CONNECTION_IMPL *conn;
WT_SESSION_IMPL *oldest_session;
WT_TXN_GLOBAL *txn_global;
WT_TXN_STATE *s;
uint64_t id, last_running, metadata_pinned, oldest_id, prev_oldest_id;
uint32_t i, session_cnt;
conn = S2C(session);
txn_global = &conn->txn_global;
oldest_session = NULL;
/* The oldest ID cannot change while we are holding the scan lock. */
prev_oldest_id = txn_global->oldest_id;
last_running = oldest_id = txn_global->current;
if ((metadata_pinned = txn_global->checkpoint_state.id) == WT_TXN_NONE)
metadata_pinned = oldest_id;
/* Walk the array of concurrent transactions. */
WT_ORDERED_READ(session_cnt, conn->session_cnt);
for (i = 0, s = txn_global->states; i < session_cnt; i++, s++) {
/* Update the last running transaction ID. */
if ((id = s->id) != WT_TXN_NONE &&
WT_TXNID_LE(prev_oldest_id, id) &&
WT_TXNID_LT(id, last_running))
last_running = id;
/* Update the metadata pinned ID. */
if ((id = s->metadata_pinned) != WT_TXN_NONE &&
WT_TXNID_LT(id, metadata_pinned))
metadata_pinned = id;
/*
* !!!
* Note: Don't ignore pinned ID values older than the previous
* oldest ID. Read-uncommitted operations publish pinned ID
* values without acquiring the scan lock to protect the global
* table. See the comment in __wt_txn_cursor_op for more
* details.
*/
if ((id = s->pinned_id) != WT_TXN_NONE &&
WT_TXNID_LT(id, oldest_id)) {
oldest_id = id;
oldest_session = &conn->sessions[i];
}
}
if (WT_TXNID_LT(last_running, oldest_id))
oldest_id = last_running;
/* The oldest ID can't move past any named snapshots. */
if ((id = txn_global->nsnap_oldest_id) != WT_TXN_NONE &&
WT_TXNID_LT(id, oldest_id))
oldest_id = id;
/* The metadata pinned ID can't move past the oldest ID. */
if (WT_TXNID_LT(oldest_id, metadata_pinned))
metadata_pinned = oldest_id;
*last_runningp = last_running;
*metadata_pinnedp = metadata_pinned;
*oldest_idp = oldest_id;
*oldest_sessionp = oldest_session;
}
/*
* __wt_txn_update_oldest --
* Sweep the running transactions to update the oldest ID required.
*/
int
__wt_txn_update_oldest(WT_SESSION_IMPL *session, uint32_t flags)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_SESSION_IMPL *oldest_session;
WT_TXN_GLOBAL *txn_global;
uint64_t current_id, last_running, metadata_pinned, oldest_id;
uint64_t prev_last_running, prev_metadata_pinned, prev_oldest_id;
bool strict, wait;
conn = S2C(session);
txn_global = &conn->txn_global;
strict = LF_ISSET(WT_TXN_OLDEST_STRICT);
wait = LF_ISSET(WT_TXN_OLDEST_WAIT);
current_id = last_running = metadata_pinned = txn_global->current;
prev_last_running = txn_global->last_running;
prev_metadata_pinned = txn_global->metadata_pinned;
prev_oldest_id = txn_global->oldest_id;
#ifdef HAVE_TIMESTAMPS
/* Try to move the pinned timestamp forward. */
if (strict)
WT_RET(__wt_txn_update_pinned_timestamp(session, false));
#endif
/*
* For pure read-only workloads, or if the update isn't forced and the
* oldest ID isn't too far behind, avoid scanning.
*/
if ((prev_oldest_id == current_id &&
prev_metadata_pinned == current_id) ||
(!strict && WT_TXNID_LT(current_id, prev_oldest_id + 100)))
return (0);
/* First do a read-only scan. */
if (wait)
__wt_readlock(session, &txn_global->rwlock);
else if ((ret =
__wt_try_readlock(session, &txn_global->rwlock)) != 0)
return (ret == EBUSY ? 0 : ret);
__txn_oldest_scan(session,
&oldest_id, &last_running, &metadata_pinned, &oldest_session);
__wt_readunlock(session, &txn_global->rwlock);
/*
* If the state hasn't changed (or hasn't moved far enough for
* non-forced updates), give up.
*/
if ((oldest_id == prev_oldest_id ||
(!strict && WT_TXNID_LT(oldest_id, prev_oldest_id + 100))) &&
((last_running == prev_last_running) ||
(!strict && WT_TXNID_LT(last_running, prev_last_running + 100))) &&
metadata_pinned == prev_metadata_pinned)
return (0);
/* It looks like an update is necessary, wait for exclusive access. */
if (wait)
__wt_writelock(session, &txn_global->rwlock);
else if ((ret =
__wt_try_writelock(session, &txn_global->rwlock)) != 0)
return (ret == EBUSY ? 0 : ret);
/*
* If the oldest ID has been updated while we waited, don't bother
* scanning.
*/
if (WT_TXNID_LE(oldest_id, txn_global->oldest_id) &&
WT_TXNID_LE(last_running, txn_global->last_running) &&
WT_TXNID_LE(metadata_pinned, txn_global->metadata_pinned))
goto done;
/*
* Re-scan now that we have exclusive access. This is necessary because
* threads get transaction snapshots with read locks, and we have to be
* sure that there isn't a thread that has got a snapshot locally but
* not yet published its snap_min.
*/
__txn_oldest_scan(session,
&oldest_id, &last_running, &metadata_pinned, &oldest_session);
#ifdef HAVE_DIAGNOSTIC
{
/*
* Make sure the ID doesn't move past any named snapshots.
*
* Don't include the read/assignment in the assert statement. Coverity
* complains if there are assignments only done in diagnostic builds,
* and when the read is from a volatile.
*/
uint64_t id = txn_global->nsnap_oldest_id;
WT_ASSERT(session,
id == WT_TXN_NONE || !WT_TXNID_LT(id, oldest_id));
}
#endif
/* Update the public IDs. */
if (WT_TXNID_LT(txn_global->metadata_pinned, metadata_pinned))
txn_global->metadata_pinned = metadata_pinned;
if (WT_TXNID_LT(txn_global->oldest_id, oldest_id))
txn_global->oldest_id = oldest_id;
if (WT_TXNID_LT(txn_global->last_running, last_running)) {
txn_global->last_running = last_running;
/* Output a verbose message about long-running transactions,
* but only when some progress is being made. */
if (WT_VERBOSE_ISSET(session, WT_VERB_TRANSACTION) &&
current_id - oldest_id > 10000 && oldest_session != NULL) {
__wt_verbose(session, WT_VERB_TRANSACTION,
"old snapshot %" PRIu64
" pinned in session %" PRIu32 " [%s]"
" with snap_min %" PRIu64,
oldest_id, oldest_session->id,
oldest_session->lastop,
oldest_session->txn.snap_min);
}
}
done: __wt_writeunlock(session, &txn_global->rwlock);
return (ret);
}
/*
* __wt_txn_config --
* Configure a transaction.
*/
int
__wt_txn_config(WT_SESSION_IMPL *session, const char *cfg[])
{
WT_CONFIG_ITEM cval;
WT_TXN *txn;
txn = &session->txn;
WT_RET(__wt_config_gets_def(session, cfg, "isolation", 0, &cval));
if (cval.len != 0)
txn->isolation =
WT_STRING_MATCH("snapshot", cval.str, cval.len) ?
WT_ISO_SNAPSHOT :
WT_STRING_MATCH("read-committed", cval.str, cval.len) ?
WT_ISO_READ_COMMITTED : WT_ISO_READ_UNCOMMITTED;
/*
* The default sync setting is inherited from the connection, but can
* be overridden by an explicit "sync" setting for this transaction.
*
* We want to distinguish between inheriting implicitly and explicitly.
*/
F_CLR(txn, WT_TXN_SYNC_SET);
WT_RET(__wt_config_gets_def(
session, cfg, "sync", (int)UINT_MAX, &cval));
if (cval.val == 0 || cval.val == 1)
/*
* This is an explicit setting of sync. Set the flag so
* that we know not to overwrite it in commit_transaction.
*/
F_SET(txn, WT_TXN_SYNC_SET);
/*
* If sync is turned off explicitly, clear the transaction's sync field.
*/
if (cval.val == 0)
txn->txn_logsync = 0;
WT_RET(__wt_config_gets_def(session, cfg, "snapshot", 0, &cval));
if (cval.len > 0)
/*
* The layering here isn't ideal - the named snapshot get
* function does both validation and setup. Otherwise we'd
* need to walk the list of named snapshots twice during
* transaction open.
*/
WT_RET(__wt_txn_named_snapshot_get(session, &cval));
/* Check if prepared updates should be ignored during reads. */
WT_RET(__wt_config_gets_def(session, cfg, "ignore_prepare", 0, &cval));
if (cval.val)
F_SET(txn, WT_TXN_IGNORE_PREPARE);
WT_RET(__wt_txn_parse_read_timestamp(session, cfg));
return (0);
}
/*
* __wt_txn_reconfigure --
* WT_SESSION::reconfigure for transactions.
*/
int
__wt_txn_reconfigure(WT_SESSION_IMPL *session, const char *config)
{
WT_CONFIG_ITEM cval;
WT_DECL_RET;
WT_TXN *txn;
txn = &session->txn;
ret = __wt_config_getones(session, config, "isolation", &cval);
if (ret == 0 && cval.len != 0) {
session->isolation = txn->isolation =
WT_STRING_MATCH("snapshot", cval.str, cval.len) ?
WT_ISO_SNAPSHOT :
WT_STRING_MATCH("read-uncommitted", cval.str, cval.len) ?
WT_ISO_READ_UNCOMMITTED : WT_ISO_READ_COMMITTED;
}
WT_RET_NOTFOUND_OK(ret);
return (0);
}
/*
* __wt_txn_release --
* Release the resources associated with the current transaction.
*/
void
__wt_txn_release(WT_SESSION_IMPL *session)
{
WT_TXN *txn;
WT_TXN_GLOBAL *txn_global;
txn = &session->txn;
txn_global = &S2C(session)->txn_global;
WT_ASSERT(session, txn->mod_count == 0);
txn->notify = NULL;
/* Clear the transaction's ID from the global table. */
if (WT_SESSION_IS_CHECKPOINT(session)) {
WT_ASSERT(session,
WT_SESSION_TXN_STATE(session)->id == WT_TXN_NONE);
txn->id = txn_global->checkpoint_state.id =
txn_global->checkpoint_state.pinned_id = WT_TXN_NONE;
/*
* Be extra careful to cleanup everything for checkpoints: once
* the global checkpoint ID is cleared, we can no longer tell
* if this session is doing a checkpoint.
*/
txn_global->checkpoint_id = 0;
} else if (F_ISSET(txn, WT_TXN_HAS_ID)) {
/*
* If transaction is prepared, this would have been done in
* prepare.
*/
if (!F_ISSET(txn, WT_TXN_PREPARE))
__txn_remove_from_global_table(session);
txn->id = WT_TXN_NONE;
}
__wt_txn_clear_commit_timestamp(session);
__wt_txn_clear_read_timestamp(session);
/* Free the scratch buffer allocated for logging. */
__wt_logrec_free(session, &txn->logrec);
/* Discard any memory from the session's stash that we can. */
WT_ASSERT(session, __wt_session_gen(session, WT_GEN_SPLIT) == 0);
__wt_stash_discard(session);
/*
* Reset the transaction state to not running and release the snapshot.
*/
__wt_txn_release_snapshot(session);
txn->isolation = session->isolation;
txn->rollback_reason = NULL;
/* Ensure the transaction flags are cleared on exit */
txn->flags = 0;
}
#ifdef HAVE_TIMESTAMPS
/*
* __txn_commit_timestamp_validate --
* Validate that timestamp provided to commit is legal.
*/
static inline int
__txn_commit_timestamp_validate(WT_SESSION_IMPL *session)
{
WT_DECL_TIMESTAMP(op_timestamp)
WT_TXN *txn;
WT_TXN_OP *op;
WT_UPDATE *upd;
u_int i;
bool op_zero_ts, upd_zero_ts;
txn = &session->txn;
/*
* Debugging checks on timestamps, if user requested them.
*/
if (F_ISSET(txn, WT_TXN_TS_COMMIT_ALWAYS) &&
!F_ISSET(txn, WT_TXN_HAS_TS_COMMIT) &&
txn->mod_count != 0)
WT_RET_MSG(session, EINVAL, "commit_timestamp required and "
"none set on this transaction");
if (F_ISSET(txn, WT_TXN_TS_COMMIT_NEVER) &&
F_ISSET(txn, WT_TXN_HAS_TS_COMMIT) &&
txn->mod_count != 0)
WT_RET_MSG(session, EINVAL, "no commit_timestamp required and "
"timestamp set on this transaction");
/*
* If we're not doing any key consistency checking, we're done.
*/
if (!F_ISSET(txn, WT_TXN_TS_COMMIT_KEYS))
return (0);
/*
* Error on any valid update structures for the same key that
* are at a later timestamp or use timestamps inconsistently.
*/
for (i = 0, op = txn->mod; i < txn->mod_count; i++, op++)
if (op->type == WT_TXN_OP_BASIC) {
/*
* Skip over any aborted update structures or ones
* from our own transaction.
*/
upd = op->u.upd->next;
while (upd != NULL && (upd->txnid == WT_TXN_ABORTED ||
upd->txnid == txn->id))
upd = upd->next;
/*
* Check the timestamp on this update with the
* first valid update in the chain. They're in
* most recent order.
*/
if (upd == NULL)
continue;
/*
* Check for consistent per-key timestamp usage.
* If timestamps are or are not used originally then
* they should be used the same way always. For this
* transaction, timestamps are in use anytime the
* commit timestamp is set.
* Check timestamps are used in order.
*/
op_zero_ts = !F_ISSET(txn, WT_TXN_HAS_TS_COMMIT);
upd_zero_ts = __wt_timestamp_iszero(&upd->timestamp);
if (op_zero_ts != upd_zero_ts)
WT_RET_MSG(session, EINVAL,
"per-key timestamps used inconsistently");
/*
* If we aren't using timestamps for this transaction
* then we are done checking. Don't check the timestamp
* because the one in the transaction is not cleared.
*/
if (op_zero_ts)
continue;
op_timestamp = op->u.upd->timestamp;
/*
* Only if the update structure doesn't have a timestamp
* then use the one in the transaction structure.
*/
if (__wt_timestamp_iszero(&op->u.upd->timestamp))
op_timestamp = txn->commit_timestamp;
if (__wt_timestamp_cmp(&op_timestamp,
&upd->timestamp) < 0)
WT_RET_MSG(session, EINVAL,
"out of order timestamps");
}
return (0);
}
#endif
/*
* __wt_txn_commit --
* Commit the current transaction.
*/
int
__wt_txn_commit(WT_SESSION_IMPL *session, const char *cfg[])
{
WT_CONFIG_ITEM cval;
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_TXN *txn;
WT_TXN_GLOBAL *txn_global;
WT_TXN_OP *op;
WT_UPDATE *upd;
u_int i;
bool locked, readonly;
#ifdef HAVE_TIMESTAMPS
WT_REF *ref;
WT_UPDATE **updp;
wt_timestamp_t prev_commit_timestamp, ts;
uint32_t previous_state;
bool update_timestamp;
#endif
txn = &session->txn;
conn = S2C(session);
txn_global = &conn->txn_global;
locked = false;
WT_ASSERT(session, F_ISSET(txn, WT_TXN_RUNNING));
WT_ASSERT(session, !F_ISSET(txn, WT_TXN_ERROR) ||
txn->mod_count == 0);
readonly = txn->mod_count == 0;
/*
* Look for a commit timestamp.
*/
WT_ERR(
__wt_config_gets_def(session, cfg, "commit_timestamp", 0, &cval));
if (cval.len != 0) {
#ifdef HAVE_TIMESTAMPS
WT_ERR(__wt_txn_parse_timestamp(session, "commit", &ts, &cval));
WT_ERR(__wt_timestamp_validate(session,
"commit", &ts, &cval, true, true, true));
__wt_timestamp_set(&txn->commit_timestamp, &ts);
__wt_txn_set_commit_timestamp(session);
#else
WT_ERR_MSG(session, EINVAL, "commit_timestamp requires a "
"version of WiredTiger built with timestamp support");
#endif
} else if (F_ISSET(txn, WT_TXN_PREPARE))
WT_ERR_MSG(session, EINVAL,
"commit_timestamp is required for a prepared transaction");
#ifdef HAVE_TIMESTAMPS
WT_ERR(__txn_commit_timestamp_validate(session));
#endif
/*
* The default sync setting is inherited from the connection, but can
* be overridden by an explicit "sync" setting for this transaction.
*/
WT_ERR(__wt_config_gets_def(session, cfg, "sync", 0, &cval));
/*
* If the user chose the default setting, check whether sync is enabled
* for this transaction (either inherited or via begin_transaction).
* If sync is disabled, clear the field to avoid the log write being
* flushed.
*
* Otherwise check for specific settings. We don't need to check for
* "on" because that is the default inherited from the connection. If
* the user set anything in begin_transaction, we only override with an
* explicit setting.
*/
if (cval.len == 0) {
if (!FLD_ISSET(txn->txn_logsync, WT_LOG_SYNC_ENABLED) &&
!F_ISSET(txn, WT_TXN_SYNC_SET))
txn->txn_logsync = 0;
} else {
/*
* If the caller already set sync on begin_transaction then
* they should not be using sync on commit_transaction.
* Flag that as an error.
*/
if (F_ISSET(txn, WT_TXN_SYNC_SET))
WT_ERR_MSG(session, EINVAL,
"Sync already set during begin_transaction");
if (WT_STRING_MATCH("background", cval.str, cval.len))
txn->txn_logsync = WT_LOG_BACKGROUND;
else if (WT_STRING_MATCH("off", cval.str, cval.len))
txn->txn_logsync = 0;
/*
* We don't need to check for "on" here because that is the
* default to inherit from the connection setting.
*/
}
/* Commit notification. */
if (txn->notify != NULL)
WT_ERR(txn->notify->notify(txn->notify,
(WT_SESSION *)session, txn->id, 1));
/*
* We are about to release the snapshot: copy values into any
* positioned cursors so they don't point to updates that could be
* freed once we don't have a snapshot.
* If this transaction is prepared, then copying values would have been
* done during prepare.
*/
if (session->ncursors > 0 && !F_ISSET(txn, WT_TXN_PREPARE)) {
WT_DIAGNOSTIC_YIELD;
WT_ERR(__wt_session_copy_values(session));
}
/* If we are logging, write a commit log record. */
if (txn->logrec != NULL &&
FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED) &&
!F_ISSET(session, WT_SESSION_NO_LOGGING)) {
/*
* We are about to block on I/O writing the log.
* Release our snapshot in case it is keeping data pinned.
* This is particularly important for checkpoints.
*/
__wt_txn_release_snapshot(session);
/*
* We hold the visibility lock for reading from the time
* we write our log record until the time we release our
* transaction so that the LSN any checkpoint gets will
* always reflect visible data.
*/
__wt_readlock(session, &txn_global->visibility_rwlock);
locked = true;
WT_ERR(__wt_txn_log_commit(session, cfg));
}
/* Note: we're going to commit: nothing can fail after this point. */
/* Process and free updates. */
for (i = 0, op = txn->mod; i < txn->mod_count; i++, op++) {
switch (op->type) {
case WT_TXN_OP_NONE:
break;
case WT_TXN_OP_BASIC:
case WT_TXN_OP_INMEM:
upd = op->u.upd;
/*
* Switch reserved operations to abort to
* simplify obsolete update list truncation.
*/
if (upd->type == WT_UPDATE_RESERVE) {
upd->txnid = WT_TXN_ABORTED;
break;
}
/*
* Writes to the lookaside file can be evicted as soon
* as they commit.
*/
if (conn->cache->las_fileid != 0 &&
op->fileid == conn->cache->las_fileid) {
upd->txnid = WT_TXN_NONE;
break;
}
#ifdef HAVE_TIMESTAMPS
if (!__wt_txn_update_needs_timestamp(session, op))
break;
if (F_ISSET(txn, WT_TXN_PREPARE)) {
/*
* In case of a prepared transaction, the order
* of modification of the prepare timestamp to
* the commit timestamp in the update chain will
* not affect the data visibility, a reader will
* encounter a prepared update resulting in
* prepare conflict.
*
* As updating timestamp might not be an atomic
* operation, we will manage using state.
*/
upd->state = WT_UPDATE_STATE_LOCKED;
__wt_timestamp_set(
&upd->timestamp, &txn->commit_timestamp);
upd->state = WT_UPDATE_STATE_READY;
} else
__wt_timestamp_set(
&upd->timestamp, &txn->commit_timestamp);
#endif
break;
case WT_TXN_OP_REF_DELETE:
#ifdef HAVE_TIMESTAMPS
if (!__wt_txn_update_needs_timestamp(session, op))
break;
ref = op->u.ref;
__wt_timestamp_set(
&ref->page_del->timestamp, &txn->commit_timestamp);
/*
* The page-deleted list can be discarded by eviction,
* lock the WT_REF to ensure we don't race.
*/
if (ref->page_del->update_list == NULL)
break;
for (;;) {
previous_state = ref->state;
if (__wt_atomic_casv32(
&ref->state, previous_state, WT_REF_LOCKED))
break;
}
if ((updp = ref->page_del->update_list) != NULL)
for (; *updp != NULL; ++updp)
__wt_timestamp_set(
&(*updp)->timestamp,
&txn->commit_timestamp);
/*
* Publish to ensure we don't let the page be evicted
* and the updates discarded before being written.
*/
WT_PUBLISH(ref->state, previous_state);
#endif
break;
case WT_TXN_OP_TRUNCATE_COL:
case WT_TXN_OP_TRUNCATE_ROW:
/* Other operations don't need timestamps. */
break;
}
__wt_txn_op_free(session, op);
}
txn->mod_count = 0;
#ifdef HAVE_TIMESTAMPS
/*
* Track the largest commit timestamp we have seen.
*
* We don't actually clear the local commit timestamp, just the flag.
* That said, we can't update the global commit timestamp until this
* transaction is visible, which happens when we release it.
*/
update_timestamp = F_ISSET(txn, WT_TXN_HAS_TS_COMMIT);
#endif
__wt_txn_release(session);
if (locked)
__wt_readunlock(session, &txn_global->visibility_rwlock);
#ifdef HAVE_TIMESTAMPS
/* First check if we've already committed something in the future. */
if (update_timestamp) {
WT_WITH_TIMESTAMP_READLOCK(session, &txn_global->rwlock,
__wt_timestamp_set(
&prev_commit_timestamp, &txn_global->commit_timestamp));
update_timestamp = __wt_timestamp_cmp(
&txn->commit_timestamp, &prev_commit_timestamp) > 0;
}
/*
* If it looks like we need to move the global commit timestamp,
* write lock and re-check.
*/
if (update_timestamp) {
#if WT_TIMESTAMP_SIZE == 8
while (__wt_timestamp_cmp(
&txn->commit_timestamp, &prev_commit_timestamp) > 0) {
if (__wt_atomic_cas64(
&txn_global->commit_timestamp.val,
prev_commit_timestamp.val,
txn->commit_timestamp.val)) {
txn_global->has_commit_timestamp = true;
break;
}
__wt_timestamp_set(
&prev_commit_timestamp, &txn_global->commit_timestamp);
}
#else
__wt_writelock(session, &txn_global->rwlock);
if (__wt_timestamp_cmp(&txn->commit_timestamp,
&txn_global->commit_timestamp) > 0) {
__wt_timestamp_set(&txn_global->commit_timestamp,
&txn->commit_timestamp);
txn_global->has_commit_timestamp = true;
}
__wt_writeunlock(session, &txn_global->rwlock);
#endif
}
#endif
/*
* We're between transactions, if we need to block for eviction, it's
* a good time to do so. Note that we must ignore any error return
* because the user's data is committed.
*/
if (!readonly)
(void)__wt_cache_eviction_check(session, false, false, NULL);
return (0);
err: /*
* If anything went wrong, roll back.
*
* !!!
* Nothing can fail after this point.
*/
if (locked)
__wt_readunlock(session, &txn_global->visibility_rwlock);
WT_TRET(__wt_txn_rollback(session, cfg));
return (ret);
}
/*
* __wt_txn_prepare --
* Prepare the current transaction.
*/
int
__wt_txn_prepare(WT_SESSION_IMPL *session, const char *cfg[])
{
#ifdef HAVE_TIMESTAMPS
WT_CONFIG_ITEM cval;
WT_TXN *txn;
WT_TXN_OP *op;
WT_UPDATE *upd;
wt_timestamp_t ts;
u_int i;
txn = &session->txn;
WT_ASSERT(session, F_ISSET(txn, WT_TXN_RUNNING));
/* Transaction should not have a commit timestamp set. */
WT_ASSERT(session, !F_ISSET(txn, WT_TXN_HAS_TS_COMMIT));
WT_ASSERT(session, !F_ISSET(txn, WT_TXN_ERROR) || txn->mod_count == 0);
/* Transaction should not have updated any of the logged tables. */
WT_ASSERT(session, txn->logrec == NULL);
WT_RET(__wt_txn_context_check(session, true));
/* Look for a prepare timestamp. */
WT_RET(
__wt_config_gets_def(session, cfg, "prepare_timestamp", 0, &cval));
if (cval.len == 0)
WT_RET_MSG(session, EINVAL, "prepare timestamp is required");
/* TODO : Validate prepare timestamp. */
WT_RET(__wt_txn_parse_timestamp(session, "prepare", &ts, &cval));
__wt_timestamp_set(&txn->prepare_timestamp, &ts);
/*
* We are about to release the snapshot: copy values into any
* positioned cursors so they don't point to updates that could be
* freed once we don't have a snapshot.
*/
if (session->ncursors > 0) {
WT_DIAGNOSTIC_YIELD;
WT_RET(__wt_session_copy_values(session));
}
/* Prepare updates. */
for (i = 0, op = txn->mod; i < txn->mod_count; i++, op++) {
/* Assert it's not an update to the lookaside file. */
WT_ASSERT(session,
S2C(session)->cache->las_fileid == 0 ||
op->fileid != S2C(session)->cache->las_fileid);
/* Metadata updates are never prepared. */
if (op->fileid == WT_METAFILE_ID)
continue;
upd = op->u.upd;
switch (op->type) {
case WT_TXN_OP_NONE:
break;
case WT_TXN_OP_BASIC:
case WT_TXN_OP_INMEM:
/*
* Switch reserved operation to abort to simplify
* obsolete update list truncation. Clear the
* operation type so we don't try to visit this update
* again: it can now be evicted.
*/
if (upd->type == WT_UPDATE_RESERVE) {
upd->txnid = WT_TXN_ABORTED;
op->type = WT_TXN_OP_NONE;
break;
}
/* Set prepare timestamp. */
__wt_timestamp_set(&upd->timestamp, &ts);
upd->state = WT_UPDATE_STATE_PREPARED;
break;
case WT_TXN_OP_REF_DELETE:
__wt_timestamp_set(
&op->u.ref->page_del->timestamp, &ts);
break;
case WT_TXN_OP_TRUNCATE_COL:
case WT_TXN_OP_TRUNCATE_ROW:
/* Other operations don't need timestamps. */
break;
}
}
/* Set transaction state to prepare. */
F_SET(&session->txn, WT_TXN_PREPARE);
/* Release our snapshot in case it is keeping data pinned. */
__wt_txn_release_snapshot(session);
/*
* Clear the transaction's ID from the global table, to facilitate
* prepared data visibility, but not from local txn structure.
*/
if (F_ISSET(txn, WT_TXN_HAS_ID))
__txn_remove_from_global_table(session);
return (0);
#else
WT_UNUSED(cfg);
WT_RET_MSG(session, ENOTSUP, "prepare_transaction requires a version "
"of WiredTiger built with timestamp support");
#endif
}
/*
* __wt_txn_rollback --
* Roll back the current transaction.
*/
int
__wt_txn_rollback(WT_SESSION_IMPL *session, const char *cfg[])
{
WT_DECL_RET;
WT_TXN *txn;
WT_TXN_OP *op;
WT_UPDATE *upd;
u_int i;
bool readonly;
WT_UNUSED(cfg);
txn = &session->txn;
readonly = txn->mod_count == 0;
WT_ASSERT(session, F_ISSET(txn, WT_TXN_RUNNING));
/* Rollback notification. */
if (txn->notify != NULL)
WT_TRET(txn->notify->notify(txn->notify, (WT_SESSION *)session,
txn->id, 0));
/* Rollback updates. */
for (i = 0, op = txn->mod; i < txn->mod_count; i++, op++) {
/* Assert it's not an update to the lookaside file. */
WT_ASSERT(session,
S2C(session)->cache->las_fileid == 0 ||
op->fileid != S2C(session)->cache->las_fileid);
/* Metadata updates are never rolled back. */
if (op->fileid == WT_METAFILE_ID)
continue;
upd = op->u.upd;
switch (op->type) {
case WT_TXN_OP_NONE:
break;
case WT_TXN_OP_BASIC:
case WT_TXN_OP_INMEM:
WT_ASSERT(session,
upd->txnid == txn->id ||
upd->txnid == WT_TXN_ABORTED);
upd->txnid = WT_TXN_ABORTED;
break;
case WT_TXN_OP_REF_DELETE:
WT_TRET(__wt_delete_page_rollback(session, op->u.ref));
break;
case WT_TXN_OP_TRUNCATE_COL:
case WT_TXN_OP_TRUNCATE_ROW:
/*
* Nothing to do: these operations are only logged for
* recovery. The in-memory changes will be rolled back
* with a combination of WT_TXN_OP_REF_DELETE and
* WT_TXN_OP_INMEM operations.
*/
break;
}
/* Free any memory allocated for the operation. */
__wt_txn_op_free(session, op);
}
txn->mod_count = 0;
__wt_txn_release(session);
/*
* We're between transactions, if we need to block for eviction, it's
* a good time to do so. Note that we must ignore any error return
* because the user's data is committed.
*/
if (!readonly)
(void)__wt_cache_eviction_check(session, false, false, NULL);
return (ret);
}
/*
* __wt_txn_rollback_required --
* Prepare to log a reason if the user attempts to use the transaction to
* do anything other than rollback.
*/
int
__wt_txn_rollback_required(WT_SESSION_IMPL *session, const char *reason)
{
session->txn.rollback_reason = reason;
return (WT_ROLLBACK);
}
/*
* __wt_txn_init --
* Initialize a session's transaction data.
*/
int
__wt_txn_init(WT_SESSION_IMPL *session, WT_SESSION_IMPL *session_ret)
{
WT_TXN *txn;
txn = &session_ret->txn;
txn->id = WT_TXN_NONE;
WT_RET(__wt_calloc_def(session,
S2C(session_ret)->session_size, &txn->snapshot));
#ifdef HAVE_DIAGNOSTIC
if (S2C(session_ret)->txn_global.states != NULL) {
WT_TXN_STATE *txn_state;
txn_state = WT_SESSION_TXN_STATE(session_ret);
WT_ASSERT(session, txn_state->pinned_id == WT_TXN_NONE);
}
#endif
/*
* Take care to clean these out in case we are reusing the transaction
* for eviction.
*/
txn->mod = NULL;
txn->isolation = session_ret->isolation;
return (0);
}
/*
* __wt_txn_stats_update --
* Update the transaction statistics for return to the application.
*/
void
__wt_txn_stats_update(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
WT_CONNECTION_STATS **stats;
WT_TXN_GLOBAL *txn_global;
uint64_t checkpoint_pinned, snapshot_pinned;
conn = S2C(session);
txn_global = &conn->txn_global;
stats = conn->stats;
checkpoint_pinned = txn_global->checkpoint_state.pinned_id;
snapshot_pinned = txn_global->nsnap_oldest_id;
WT_STAT_SET(session, stats, txn_pinned_range,
txn_global->current - txn_global->oldest_id);
#if WT_TIMESTAMP_SIZE == 8
WT_STAT_SET(session, stats, txn_pinned_timestamp,
txn_global->commit_timestamp.val -
txn_global->pinned_timestamp.val);
WT_STAT_SET(session, stats, txn_pinned_timestamp_oldest,
txn_global->commit_timestamp.val -
txn_global->oldest_timestamp.val);
#endif
WT_STAT_SET(session, stats, txn_pinned_snapshot_range,
snapshot_pinned == WT_TXN_NONE ?
0 : txn_global->current - snapshot_pinned);
WT_STAT_SET(session, stats, txn_pinned_checkpoint_range,
checkpoint_pinned == WT_TXN_NONE ?
0 : txn_global->current - checkpoint_pinned);
WT_STAT_SET(
session, stats, txn_checkpoint_time_max, conn->ckpt_time_max);
WT_STAT_SET(
session, stats, txn_checkpoint_time_min, conn->ckpt_time_min);
WT_STAT_SET(
session, stats, txn_checkpoint_time_recent, conn->ckpt_time_recent);
WT_STAT_SET(
session, stats, txn_checkpoint_time_total, conn->ckpt_time_total);
WT_STAT_SET(session,
stats, txn_commit_queue_len, txn_global->commit_timestampq_len);
WT_STAT_SET(session,
stats, txn_read_queue_len, txn_global->read_timestampq_len);
}
/*
* __wt_txn_destroy --
* Destroy a session's transaction data.
*/
void
__wt_txn_destroy(WT_SESSION_IMPL *session)
{
WT_TXN *txn;
txn = &session->txn;
__wt_free(session, txn->mod);
__wt_free(session, txn->snapshot);
}
/*
* __wt_txn_global_init --
* Initialize the global transaction state.
*/
int
__wt_txn_global_init(WT_SESSION_IMPL *session, const char *cfg[])
{
WT_CONNECTION_IMPL *conn;
WT_TXN_GLOBAL *txn_global;
WT_TXN_STATE *s;
u_int i;
WT_UNUSED(cfg);
conn = S2C(session);
txn_global = &conn->txn_global;
txn_global->current = txn_global->last_running =
txn_global->metadata_pinned = txn_global->oldest_id = WT_TXN_FIRST;
WT_RET(__wt_spin_init(
session, &txn_global->id_lock, "transaction id lock"));
WT_RWLOCK_INIT_TRACKED(session, &txn_global->rwlock, txn_global);
WT_RET(__wt_rwlock_init(session, &txn_global->visibility_rwlock));
WT_RWLOCK_INIT_TRACKED(session,
&txn_global->commit_timestamp_rwlock, commit_timestamp);
TAILQ_INIT(&txn_global->commit_timestamph);
WT_RWLOCK_INIT_TRACKED(session,
&txn_global->read_timestamp_rwlock, read_timestamp);
TAILQ_INIT(&txn_global->read_timestamph);
WT_RET(__wt_rwlock_init(session, &txn_global->nsnap_rwlock));
txn_global->nsnap_oldest_id = WT_TXN_NONE;
TAILQ_INIT(&txn_global->nsnaph);
WT_RET(__wt_calloc_def(
session, conn->session_size, &txn_global->states));
for (i = 0, s = txn_global->states; i < conn->session_size; i++, s++)
s->id = s->metadata_pinned = s->pinned_id = WT_TXN_NONE;
return (0);
}
/*
* __wt_txn_global_destroy --
* Destroy the global transaction state.
*/
void
__wt_txn_global_destroy(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
WT_TXN_GLOBAL *txn_global;
conn = S2C(session);
txn_global = &conn->txn_global;
if (txn_global == NULL)
return;
__wt_spin_destroy(session, &txn_global->id_lock);
__wt_rwlock_destroy(session, &txn_global->rwlock);
__wt_rwlock_destroy(session, &txn_global->commit_timestamp_rwlock);
__wt_rwlock_destroy(session, &txn_global->read_timestamp_rwlock);
__wt_rwlock_destroy(session, &txn_global->nsnap_rwlock);
__wt_rwlock_destroy(session, &txn_global->visibility_rwlock);
__wt_free(session, txn_global->states);
}
/*
* __wt_txn_global_shutdown --
* Shut down the global transaction state.
*/
int
__wt_txn_global_shutdown(WT_SESSION_IMPL *session)
{
bool txn_active;
/*
* We're shutting down. Make sure everything gets freed.
*
* It's possible that the eviction server is in the middle of a long
* operation, with a transaction ID pinned. In that case, we will loop
* here until the transaction ID is released, when the oldest
* transaction ID will catch up with the current ID.
*/
for (;;) {
WT_RET(__wt_txn_activity_check(session, &txn_active));
if (!txn_active)
break;
WT_STAT_CONN_INCR(session, txn_release_blocked);
__wt_yield();
}
#ifdef HAVE_TIMESTAMPS
/*
* Now that all transactions have completed, no timestamps should be
* pinned.
*/
__wt_timestamp_set_inf(&S2C(session)->txn_global.pinned_timestamp);
#endif
return (0);
}
/*
* __wt_verbose_dump_txn_one --
* Output diagnostic information about a transaction structure.
*/
int
__wt_verbose_dump_txn_one(WT_SESSION_IMPL *session, WT_TXN *txn)
{
#ifdef HAVE_TIMESTAMPS
char hex_timestamp[3][2 * WT_TIMESTAMP_SIZE + 1];
#endif
const char *iso_tag;
WT_NOT_READ(iso_tag, "INVALID");
switch (txn->isolation) {
case WT_ISO_READ_COMMITTED:
iso_tag = "WT_ISO_READ_COMMITTED";
break;
case WT_ISO_READ_UNCOMMITTED:
iso_tag = "WT_ISO_READ_UNCOMMITTED";
break;
case WT_ISO_SNAPSHOT:
iso_tag = "WT_ISO_SNAPSHOT";
break;
}
#ifdef HAVE_TIMESTAMPS
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[0], &txn->commit_timestamp));
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[1], &txn->first_commit_timestamp));
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[2], &txn->read_timestamp));
WT_RET(__wt_msg(session,
"mod count: %u"
", snap min: %" PRIu64
", snap max: %" PRIu64
", commit_timestamp: %s"
", first_commit_timestamp: %s"
", read_timestamp: %s"
", flags: 0x%08" PRIx32
", isolation: %s",
txn->mod_count,
txn->snap_min,
txn->snap_max,
hex_timestamp[0],
hex_timestamp[1],
hex_timestamp[2],
txn->flags,
iso_tag));
#else
WT_RET(__wt_msg(session,
"mod count: %u"
", snap min: %" PRIu64
", snap max: %" PRIu64
", flags: 0x%08" PRIx32
", isolation: %s",
txn->mod_count,
txn->snap_min,
txn->snap_max,
txn->flags,
iso_tag));
#endif
return (0);
}
/*
* __wt_verbose_dump_txn --
* Output diagnostic information about the global transaction state.
*/
int
__wt_verbose_dump_txn(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
WT_SESSION_IMPL *sess;
WT_TXN_GLOBAL *txn_global;
WT_TXN_STATE *s;
uint64_t id;
uint32_t i, session_cnt;
#ifdef HAVE_TIMESTAMPS
char hex_timestamp[3][2 * WT_TIMESTAMP_SIZE + 1];
#endif
conn = S2C(session);
txn_global = &conn->txn_global;
WT_RET(__wt_msg(session, "%s", WT_DIVIDER));
WT_RET(__wt_msg(session, "transaction state dump"));
WT_RET(__wt_msg(session, "current ID: %" PRIu64, txn_global->current));
WT_RET(__wt_msg(session,
"last running ID: %" PRIu64, txn_global->last_running));
WT_RET(__wt_msg(session, "oldest ID: %" PRIu64, txn_global->oldest_id));
#ifdef HAVE_TIMESTAMPS
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[0], &txn_global->commit_timestamp));
WT_RET(__wt_msg(session, "commit timestamp: %s", hex_timestamp[0]));
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[0], &txn_global->oldest_timestamp));
WT_RET(__wt_msg(session, "oldest timestamp: %s", hex_timestamp[0]));
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[0], &txn_global->pinned_timestamp));
WT_RET(__wt_msg(session, "pinned timestamp: %s", hex_timestamp[0]));
WT_RET(__wt_timestamp_to_hex_string(
session, hex_timestamp[0], &txn_global->stable_timestamp));
WT_RET(__wt_msg(session, "stable timestamp: %s", hex_timestamp[0]));
WT_RET(__wt_msg(session, "has_commit_timestamp: %s",
txn_global->has_commit_timestamp ? "yes" : "no"));
WT_RET(__wt_msg(session, "has_oldest_timestamp: %s",
txn_global->has_oldest_timestamp ? "yes" : "no"));
WT_RET(__wt_msg(session, "has_pinned_timestamp: %s",
txn_global->has_pinned_timestamp ? "yes" : "no"));
WT_RET(__wt_msg(session, "has_stable_timestamp: %s",
txn_global->has_stable_timestamp ? "yes" : "no"));
WT_RET(__wt_msg(session, "oldest_is_pinned: %s",
txn_global->oldest_is_pinned ? "yes" : "no"));
WT_RET(__wt_msg(session, "stable_is_pinned: %s",
txn_global->stable_is_pinned ? "yes" : "no"));
#endif
WT_RET(__wt_msg(session, "checkpoint running: %s",
txn_global->checkpoint_running ? "yes" : "no"));
WT_RET(__wt_msg(session, "checkpoint generation: %" PRIu64,
__wt_gen(session, WT_GEN_CHECKPOINT)));
WT_RET(__wt_msg(session, "checkpoint pinned ID: %" PRIu64,
txn_global->checkpoint_state.pinned_id));
WT_RET(__wt_msg(session, "checkpoint txn ID: %" PRIu64,
txn_global->checkpoint_state.id));
WT_RET(__wt_msg(session,
"oldest named snapshot ID: %" PRIu64, txn_global->nsnap_oldest_id));
WT_ORDERED_READ(session_cnt, conn->session_cnt);
WT_RET(__wt_msg(session, "session count: %" PRIu32, session_cnt));
WT_RET(__wt_msg(session, "Transaction state of active sessions:"));
/*
* Walk each session transaction state and dump information. Accessing
* the content of session handles is not thread safe, so some
* information may change while traversing if other threads are active
* at the same time, which is OK since this is diagnostic code.
*/
for (i = 0, s = txn_global->states; i < session_cnt; i++, s++) {
/* Skip sessions with no active transaction */
if ((id = s->id) == WT_TXN_NONE && s->pinned_id == WT_TXN_NONE)
continue;
sess = &conn->sessions[i];
WT_RET(__wt_msg(session,
"ID: %" PRIu64
", pinned ID: %" PRIu64
", metadata pinned ID: %" PRIu64
", name: %s",
id, s->pinned_id, s->metadata_pinned,
sess->name == NULL ?
"EMPTY" : sess->name));
WT_RET(__wt_verbose_dump_txn_one(sess, &sess->txn));
}
return (0);
}
|