Loading...
Searching...
No Matches
predicate.c File Reference
#include "postgres.h"#include "access/parallel.h"#include "access/slru.h"#include "access/transam.h"#include "access/twophase.h"#include "access/twophase_rmgr.h"#include "access/xact.h"#include "access/xlog.h"#include "miscadmin.h"#include "pgstat.h"#include "port/pg_lfind.h"#include "storage/predicate.h"#include "storage/predicate_internals.h"#include "storage/proc.h"#include "storage/procarray.h"#include "utils/guc_hooks.h"#include "utils/rel.h"#include "utils/snapmgr.h"
Include dependency graph for predicate.c:
Go to the source code of this file.
Data Structures | |
| struct | SerialControlData |
Typedefs | |
| typedef struct SerialControlData | SerialControlData |
| typedef struct SerialControlData * | SerialControl |
Macro Definition Documentation
◆ NPREDICATELOCKTARGETENTS
| #define NPREDICATELOCKTARGETENTS | ( | ) | mul_size(max_predicate_locks_per_xact, add_size(MaxBackends, max_prepared_xacts)) |
Definition at line 264 of file predicate.c.
336 : (page) + 1)
337
338#define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
339 (SerialSlruCtl->shared->page_buffer[slotno] + \
340 ((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
341
342#define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
343
345{
349} SerialControlData;
350
352
354
355/*
356 * When the oldest committed transaction on the "finished" list is moved to
357 * SLRU, its predicate locks will be moved to this "dummy" transaction,
358 * collapsing duplicate targets. When a duplicate is found, the later
359 * commitSeqNo is used.
360 */
362
363
364/*
365 * These configuration variables are used to set the predicate lock table size
366 * and to control promotion of predicate locks to coarser granularity in an
367 * attempt to degrade performance (mostly as false positive serialization
368 * failure) gracefully in the face of memory pressure.
369 */
373
374/*
375 * This provides a list of objects in order to track transactions
376 * participating in predicate locking. Entries in the list are fixed size,
377 * and reside in shared memory. The memory address of an entry must remain
378 * fixed during its lifetime. The list will be protected from concurrent
379 * update externally; no provision is made in this code to manage that. The
380 * number of entries in the list, and the size allowed for each entry is
381 * fixed upon creation.
382 */
384
385/*
386 * This provides a pool of RWConflict data elements to use in conflict lists
387 * between transactions.
388 */
390
391/*
392 * The predicate locking hash tables are in shared memory.
393 * Each backend keeps pointers to them.
394 */
399
400/*
401 * Tag for a dummy entry in PredicateLockTargetHash. By temporarily removing
402 * this entry, you can ensure that there's enough scratch space available for
403 * inserting one entry in the hash table. This is an otherwise-invalid tag.
404 */
408
409/*
410 * The local hash table used to determine when to combine multiple fine-
411 * grained locks into a single courser-grained lock.
412 */
414
415/*
416 * Keep a pointer to the currently-running serializable transaction (if any)
417 * for quick reference. Also, remember if we have written anything that could
418 * cause a rw-conflict.
419 */
422
423/*
424 * The SXACT_FLAG_RO_UNSAFE optimization might lead us to release
425 * MySerializableXact early. If that happens in a parallel query, the leader
426 * needs to defer the destruction of the SERIALIZABLEXACT until end of
427 * transaction, because the workers still have a reference to it. In that
428 * case, the leader stores it here.
429 */
431
432/* local functions */
433
436
442
448
457 PREDICATELOCKTARGETTAG *parent);
488
489
490/*------------------------------------------------------------------------*/
491
492/*
493 * Does this relation participate in predicate locking? Temporary and system
494 * relations are exempt.
495 */
496static inline bool
498{
500 RelationUsesLocalBuffers(relation));
501}
502
503/*
504 * When a public interface method is called for a read, this is the test to
505 * see if we should do a quick return.
506 *
507 * Note: this function has side-effects! If this transaction has been flagged
508 * as RO-safe since the last call, we release all predicate locks and reset
509 * MySerializableXact. That makes subsequent calls to return quickly.
510 *
511 * This is marked as 'inline' to eliminate the function call overhead in the
512 * common case that serialization is not needed.
513 */
514static inline bool
516{
517 /* Nothing to do if this is not a serializable transaction */
519 return false;
520
521 /*
522 * Don't acquire locks or conflict when scanning with a special snapshot.
523 * This excludes things like CLUSTER and REINDEX. They use the wholesale
524 * functions TransferPredicateLocksToHeapRelation() and
525 * CheckTableForSerializableConflictIn() to participate in serialization,
526 * but the scans involved don't need serialization.
527 */
529 return false;
530
531 /*
532 * Check if we have just become "RO-safe". If we have, immediately release
533 * all locks as they're not needed anymore. This also resets
534 * MySerializableXact, so that subsequent calls to this function can exit
535 * quickly.
536 *
537 * A transaction is flagged as RO_SAFE if all concurrent R/W transactions
538 * commit without having conflicts out to an earlier snapshot, thus
539 * ensuring that no conflicts are possible for this transaction.
540 */
542 {
544 return false;
545 }
546
547 /* Check if the relation doesn't participate in predicate locking */
549 return false;
550
551 return true; /* no excuse to skip predicate locking */
552}
553
554/*
555 * Like SerializationNeededForRead(), but called on writes.
556 * The logic is the same, but there is no snapshot and we can't be RO-safe.
557 */
558static inline bool
560{
561 /* Nothing to do if this is not a serializable transaction */
563 return false;
564
565 /* Check if the relation doesn't participate in predicate locking */
567 return false;
568
569 return true; /* no excuse to skip predicate locking */
570}
571
572
573/*------------------------------------------------------------------------*/
574
575/*
576 * These functions are a simple implementation of a list for this specific
577 * type of struct. If there is ever a generalized shared memory list, we
578 * should probably switch to that.
579 */
582{
584
587
592}
593
594static void
596{
598
601}
602
603/*------------------------------------------------------------------------*/
604
605/*
606 * These functions manage primitive access to the RWConflict pool and lists.
607 */
608static bool
610{
611 dlist_iter iter;
612
614
615 /* Check the ends of the purported conflict first. */
620 return false;
621
622 /*
623 * A conflict is possible; walk the list to find out.
624 *
625 * The unconstify is needed as we have no const version of
626 * dlist_foreach().
627 */
629 {
632
634 return true;
635 }
636
637 /* No conflict found. */
638 return false;
639}
640
641static void
643{
645
648
653 errhint("You might need to run fewer transactions at a time or increase \"max_connections\".")));
654
657
658 conflict->sxactOut = reader;
662}
663
664static void
667{
669
673
678 errhint("You might need to run fewer transactions at a time or increase \"max_connections\".")));
679
682
687}
688
689static void
691{
695}
696
697static void
699{
700 dlist_mutable_iter iter;
701
704
706
707 /*
708 * We know this isn't a safe snapshot, so we can stop looking for other
709 * potential conflicts.
710 */
712 {
715
718
720 }
721}
722
723/*------------------------------------------------------------------------*/
724
725/*
726 * Decide whether a Serial page number is "older" for truncation purposes.
727 * Analogous to CLOGPagePrecedes().
728 */
729static bool
731{
734
739
742}
743
744#ifdef USE_ASSERT_CHECKING
745static void
747{
749 offset = per_page / 2;
751 oldestPage,
752 headPage,
753 targetPage;
755 oldestXact;
756
757 /* GetNewTransactionId() has assigned the last XID it can safely use. */
761 oldestXact = newestXact + 1;
762 oldestXact -= 1U << 31;
764
765 /*
766 * In this scenario, the SLRU headPage pertains to the last ~1000 XIDs
767 * assigned. oldestXact finishes, ~2B XIDs having elapsed since it
768 * started. Further transactions cause us to summarize oldestXact to
769 * tailPage. Function must return false so SerialAdd() doesn't zero
770 * tailPage (which may contain entries for other old, recently-finished
771 * XIDs) and half the SLRU. Reaching this requires burning ~2B XIDs in
772 * single-user mode, a negligible possibility.
773 */
777
778 /*
779 * In this scenario, the SLRU headPage pertains to oldestXact. We're
780 * summarizing an XID near newestXact. (Assume few other XIDs used
781 * SERIALIZABLE, hence the minimal headPage advancement. Assume
782 * oldestXact was long-running and only recently reached the SLRU.)
783 * Function must return true to make SerialAdd() create targetPage.
784 *
785 * Today's implementation mishandles this case, but it doesn't matter
786 * enough to fix. Verify that the defect affects just one page by
787 * asserting correct treatment of its prior page. Reaching this case
788 * requires burning ~2B XIDs in single-user mode, a negligible
789 * possibility. Moreover, if it does happen, the consequence would be
790 * mild, namely a new transaction failing in SimpleLruReadPage().
791 */
795#if 0
797#endif
798}
799#endif
800
801/*
802 * Initialize for the tracking of old serializable committed xids.
803 */
804static void
806{
807 bool found;
808
809 /*
810 * Set up SLRU management of the pg_serial data.
811 */
817#ifdef USE_ASSERT_CHECKING
819#endif
821
822 /*
823 * Create or attach to the SerialControl structure.
824 */
827
829 if (!found)
830 {
831 /*
832 * Set control information to reflect empty SLRU.
833 */
839 }
840}
841
842/*
843 * GUC check_hook for serializable_buffers
844 */
845bool
847{
849}
850
851/*
852 * Record a committed read write serializable xid and the minimum
853 * commitSeqNo of any transactions to which this xid had a rw-conflict out.
854 * An invalid commitSeqNo means that there were no conflicts out from xid.
855 */
856static void
858{
864 LWLock *lock;
865
867
870
871 /*
872 * In this routine, we must hold both SerialControlLock and the SLRU bank
873 * lock simultaneously while making the SLRU data catch up with the new
874 * state that we determine.
875 */
877
878 /*
879 * If 'xid' is older than the global xmin (== tailXid), there's no need to
880 * store it, after all. This can happen if the oldest transaction holding
881 * back the global xmin just finished, making 'xid' uninteresting, but
882 * ClearOldPredicateLocks() has not yet run.
883 */
886 {
888 return;
889 }
890
891 /*
892 * If the SLRU is currently unused, zero out the whole active region from
893 * tailXid to headXid before taking it into use. Otherwise zero out only
894 * any new pages that enter the tailXid-headXid range as we advance
895 * headXid.
896 */
898 {
901 }
902 else
903 {
906 targetPage);
907 }
908
914
916 {
917 /* Initialize intervening pages; might involve trading locks */
918 for (;;)
919 {
924 break;
926 LWLockRelease(lock);
927 }
928 }
929 else
930 {
933 }
934
937
938 LWLockRelease(lock);
940}
941
942/*
943 * Get the minimum commitSeqNo for any conflict out for the given xid. For
944 * a transaction which exists but has no conflict out, InvalidSerCommitSeqNo
945 * will be returned.
946 */
949{
954
956
961
963 return 0;
964
966
969 return 0;
970
971 /*
972 * The following function must be called without holding SLRU bank lock,
973 * but will return with that lock held, which must then be released.
974 */
976 SerialPage(xid), xid);
980}
981
982/*
983 * Call this whenever there is a new xmin for active serializable
984 * transactions. We don't need to keep information on transactions which
985 * precede that. InvalidTransactionId means none active, so everything in
986 * the SLRU can be discarded.
987 */
988static void
990{
992
993 /*
994 * When no sxacts are active, nothing overlaps, set the xid values to
995 * invalid to show that there are no valid entries. Don't clear headPage,
996 * though. A new xmin might still land on that page, and we don't want to
997 * repeatedly zero out the same page.
998 */
1000 {
1004 return;
1005 }
1006
1007 /*
1008 * When we're recovering prepared transactions, the global xmin might move
1009 * backwards depending on the order they're recovered. Normally that's not
1010 * OK, but during recovery no serializable transactions will commit, so
1011 * the SLRU is empty and we can get away with it.
1012 */
1014 {
1018 {
1020 }
1022 return;
1023 }
1024
1027
1029
1031}
1032
1033/*
1034 * Perform a checkpoint --- either during shutdown, or on-the-fly
1035 *
1036 * We don't have any data that needs to survive a restart, but this is a
1037 * convenient place to truncate the SLRU.
1038 */
1039void
1041{
1043
1045
1046 /* Exit quickly if the SLRU is currently not in use. */
1048 {
1050 return;
1051 }
1052
1054 {
1056
1058
1059 /*
1060 * It is possible for the tailXid to be ahead of the headXid. This
1061 * occurs if we checkpoint while there are in-progress serializable
1062 * transaction(s) advancing the tail but we are yet to summarize the
1063 * transactions. In this case, we cutoff up to the headPage and the
1064 * next summary will advance the headXid.
1065 */
1067 {
1068 /* We can truncate the SLRU up to the page containing tailXid */
1070 }
1071 else
1073 }
1074 else
1075 {
1076 /*----------
1077 * The SLRU is no longer needed. Truncate to head before we set head
1078 * invalid.
1079 *
1080 * XXX: It's possible that the SLRU is not needed again until XID
1081 * wrap-around has happened, so that the segment containing headPage
1082 * that we leave behind will appear to be new again. In that case it
1083 * won't be removed until XID horizon advances enough to make it
1084 * current again.
1085 *
1086 * XXX: This should happen in vac_truncate_clog(), not in checkpoints.
1087 * Consider this scenario, starting from a system with no in-progress
1088 * transactions and VACUUM FREEZE having maximized oldestXact:
1089 * - Start a SERIALIZABLE transaction.
1090 * - Start, finish, and summarize a SERIALIZABLE transaction, creating
1091 * one SLRU page.
1092 * - Consume XIDs to reach xidStopLimit.
1093 * - Finish all transactions. Due to the long-running SERIALIZABLE
1094 * transaction, earlier checkpoints did not touch headPage. The
1095 * next checkpoint will change it, but that checkpoint happens after
1096 * the end of the scenario.
1097 * - VACUUM to advance XID limits.
1098 * - Consume ~2M XIDs, crossing the former xidWrapLimit.
1099 * - Start, finish, and summarize a SERIALIZABLE transaction.
1100 * SerialAdd() declines to create the targetPage, because headPage
1101 * is not regarded as in the past relative to that targetPage. The
1102 * transaction instigating the summarize fails in
1103 * SimpleLruReadPage().
1104 */
1107 }
1108
1110
1111 /*
1112 * Truncate away pages that are no longer required. Note that no
1113 * additional locking is required, because this is only called as part of
1114 * a checkpoint, and the validity limits have already been determined.
1115 */
1117
1118 /*
1119 * Write dirty SLRU pages to disk
1120 *
1121 * This is not actually necessary from a correctness point of view. We do
1122 * it merely as a debugging aid.
1123 *
1124 * We're doing this after the truncation to avoid writing pages right
1125 * before deleting the file in which they sit, which would be completely
1126 * pointless.
1127 */
1129}
1130
1131/*------------------------------------------------------------------------*/
1132
1133/*
1134 * PredicateLockShmemInit -- Initialize the predicate locking data structures.
1135 *
1136 * This is called from CreateSharedMemoryAndSemaphores(), which see for
1137 * more comments. In the normal postmaster case, the shared hash tables
1138 * are created here. Backends inherit the pointers
1139 * to the shared tables via fork(). In the EXEC_BACKEND case, each
1140 * backend re-executes this code to obtain pointers to the already existing
1141 * shared hash tables.
1142 */
1143void
1145{
1146 HASHCTL info;
1149 bool found;
1150
1151#ifndef EXEC_BACKEND
1153#endif
1154
1155 /*
1156 * Compute size of predicate lock target hashtable. Note these
1157 * calculations must agree with PredicateLockShmemSize!
1158 */
1160
1161 /*
1162 * Allocate hash table for PREDICATELOCKTARGET structs. This stores
1163 * per-predicate-lock-target information.
1164 */
1168
1170 max_table_size,
1171 max_table_size,
1172 &info,
1175
1176 /*
1177 * Reserve a dummy entry in the hash table; we use it to make sure there's
1178 * always one entry available when we need to split or combine a page,
1179 * because running out of space there could mean aborting a
1180 * non-serializable transaction.
1181 */
1183 {
1185 HASH_ENTER, &found);
1186 Assert(!found);
1187 }
1188
1189 /* Pre-calculate the hash and partition lock of the scratch entry */
1192
1193 /*
1194 * Allocate hash table for PREDICATELOCK structs. This stores per
1195 * xact-lock-of-a-target information.
1196 */
1201
1202 /* Assume an average of 2 xacts per target */
1203 max_table_size *= 2;
1204
1206 max_table_size,
1207 max_table_size,
1208 &info,
1211
1212 /*
1213 * Compute size for serializable transaction hashtable. Note these
1214 * calculations must agree with PredicateLockShmemSize!
1215 */
1217
1218 /*
1219 * Allocate a list to hold information on transactions participating in
1220 * predicate locking.
1221 *
1222 * Assume an average of 10 predicate locking transactions per backend.
1223 * This allows aggressive cleanup while detail is present before data must
1224 * be summarized for storage in SLRU and the "dummy" transaction.
1225 */
1226 max_table_size *= 10;
1227
1231
1233 requestSize,
1234 &found);
1236 if (!found)
1237 {
1239
1240 /* clean everything, both the header and the element */
1242
1253 /* Add all elements to available list, clean. */
1255 {
1259 }
1276 }
1277 /* This never changes, so let's keep a local copy. */
1279
1280 /*
1281 * Allocate hash table for SERIALIZABLEXID structs. This stores per-xid
1282 * information for serializable transactions which have accessed data.
1283 */
1286
1288 max_table_size,
1289 max_table_size,
1290 &info,
1292 HASH_FIXED_SIZE);
1293
1294 /*
1295 * Allocate space for tracking rw-conflicts in lists attached to the
1296 * transactions.
1297 *
1298 * Assume an average of 5 conflicts per transaction. Calculations suggest
1299 * that this will prevent resource exhaustion in even the most pessimal
1300 * loads up to max_connections = 200 with all 200 connections pounding the
1301 * database with serializable transactions. Beyond that, there may be
1302 * occasional transactions canceled when trying to flag conflicts. That's
1303 * probably OK.
1304 */
1305 max_table_size *= 5;
1306
1309 RWConflictDataSize);
1310
1312 requestSize,
1313 &found);
1315 if (!found)
1316 {
1318
1319 /* clean everything, including the elements */
1321
1324 RWConflictPoolHeaderDataSize);
1325 /* Add all elements to available list, clean. */
1327 {
1330 }
1331 }
1332
1333 /*
1334 * Create or attach to the header for the list of finished serializable
1335 * transactions.
1336 */
1340 &found);
1342 if (!found)
1344
1345 /*
1346 * Initialize the SLRU storage for old committed serializable
1347 * transactions.
1348 */
1349 SerialInit();
1350}
1351
1352/*
1353 * Estimate shared-memory space used for predicate lock table
1354 */
1355Size
1357{
1358 Size size = 0;
1360
1361 /* predicate lock target hash table */
1365
1366 /* predicate lock hash table */
1367 max_table_size *= 2;
1370
1371 /*
1372 * Since NPREDICATELOCKTARGETENTS is only an estimate, add 10% safety
1373 * margin.
1374 */
1375 size = add_size(size, size / 10);
1376
1377 /* transaction list */
1379 max_table_size *= 10;
1383
1384 /* transaction xid table */
1387
1388 /* rw-conflict pool */
1389 max_table_size *= 5;
1392 RWConflictDataSize));
1393
1394 /* Head for list of finished serializable transactions. */
1396
1397 /* Shared memory structures for SLRU tracking of old committed xids. */
1400
1401 return size;
1402}
1403
1404
1405/*
1406 * Compute the hash code associated with a PREDICATELOCKTAG.
1407 *
1408 * Because we want to use just one set of partition locks for both the
1409 * PREDICATELOCKTARGET and PREDICATELOCK hash tables, we have to make sure
1410 * that PREDICATELOCKs fall into the same partition number as their
1411 * associated PREDICATELOCKTARGETs. dynahash.c expects the partition number
1412 * to be the low-order bits of the hash code, and therefore a
1413 * PREDICATELOCKTAG's hash code must have the same low-order bits as the
1414 * associated PREDICATELOCKTARGETTAG's hash code. We achieve this with this
1415 * specialized hash function.
1416 */
1419{
1422
1424
1425 /* Look into the associated target object, and compute its hash code */
1427
1429}
1430
1431
1432/*
1433 * GetPredicateLockStatusData
1434 * Return a table containing the internal state of the predicate
1435 * lock manager for use in pg_lock_status.
1436 *
1437 * Like GetLockStatusData, this function tries to hold the partition LWLocks
1438 * for as short a time as possible by returning two arrays that simply
1439 * contain the PREDICATELOCKTARGETTAG and SERIALIZABLEXACT for each lock
1440 * table entry. Multiple copies of the same PREDICATELOCKTARGETTAG and
1441 * SERIALIZABLEXACT will likely appear.
1442 */
1443PredicateLockData *
1445{
1449 el;
1452
1454
1455 /*
1456 * To ensure consistency, take simultaneous locks on all partition locks
1457 * in ascending order, then SerializableXactHashLock.
1458 */
1462
1463 /* Get number of locks and allocate appropriately-sized arrays. */
1468
1469
1470 /* Scan through PredicateLockHash and copy contents */
1472
1473 el = 0;
1474
1476 {
1479 el++;
1480 }
1481
1483
1484 /* Release locks in reverse order */
1488
1490}
1491
1492/*
1493 * Free up shared memory structures by pushing the oldest sxact (the one at
1494 * the front of the SummarizeOldestCommittedSxact queue) into summary form.
1495 * Each call will free exactly one SERIALIZABLEXACT structure and may also
1496 * free one or more of these structures: SERIALIZABLEXID, PREDICATELOCK,
1497 * PREDICATELOCKTARGET, RWConflictData.
1498 */
1499static void
1501{
1503
1505
1506 /*
1507 * This function is only called if there are no sxact slots available.
1508 * Some of them must belong to old, already-finished transactions, so
1509 * there should be something in FinishedSerializableTransactions list that
1510 * we can summarize. However, there's a race condition: while we were not
1511 * holding any locks, a transaction might have ended and cleaned up all
1512 * the finished sxact entries already, freeing up their sxact slots. In
1513 * that case, we have nothing to do here. The caller will find one of the
1514 * slots released by the other backend when it retries.
1515 */
1517 {
1519 return;
1520 }
1521
1522 /*
1523 * Grab the first sxact off the finished list -- this will be the earliest
1524 * commit. Remove it from the list.
1525 */
1529
1530 /* Add to SLRU summary information. */
1534
1535 /* Summarize and release the detail. */
1537
1539}
1540
1541/*
1542 * GetSafeSnapshot
1543 * Obtain and register a snapshot for a READ ONLY DEFERRABLE
1544 * transaction. Ensures that the snapshot is "safe", i.e. a
1545 * read-only transaction running on it can execute serializably
1546 * without further checks. This requires waiting for concurrent
1547 * transactions to complete, and retrying with a new snapshot if
1548 * one of them could possibly create a conflict.
1549 *
1550 * As with GetSerializableTransactionSnapshot (which this is a subroutine
1551 * for), the passed-in Snapshot pointer should reference a static data
1552 * area that can safely be passed to GetSnapshotData.
1553 */
1556{
1557 Snapshot snapshot;
1558
1560
1561 while (true)
1562 {
1563 /*
1564 * GetSerializableTransactionSnapshotInt is going to call
1565 * GetSnapshotData, so we need to provide it the static snapshot area
1566 * our caller passed to us. The pointer returned is actually the same
1567 * one passed to it, but we avoid assuming that here.
1568 */
1571
1573 return snapshot; /* no concurrent r/w xacts; it's safe */
1574
1576
1577 /*
1578 * Wait for concurrent transactions to finish. Stop early if one of
1579 * them marked us as conflicted.
1580 */
1584 {
1588 }
1590
1592 {
1594 break; /* success */
1595 }
1596
1598
1599 /* else, need to retry... */
1604 }
1605
1606 /*
1607 * Now we have a safe snapshot, so we don't need to do any further checks.
1608 */
1611
1612 return snapshot;
1613}
1614
1615/*
1616 * GetSafeSnapshotBlockingPids
1617 * If the specified process is currently blocked in GetSafeSnapshot,
1618 * write the process IDs of all processes that it is blocked by
1619 * into the caller-supplied buffer output[]. The list is truncated at
1620 * output_size, and the number of PIDs written into the buffer is
1621 * returned. Returns zero if the given PID is not currently blocked
1622 * in GetSafeSnapshot.
1623 */
1624int
1626{
1628 dlist_iter iter;
1630
1632
1633 /* Find blocked_pid's SERIALIZABLEXACT by linear search. */
1635 {
1638
1640 {
1642 break;
1643 }
1644 }
1645
1646 /* Did we find it, and is it currently waiting in GetSafeSnapshot? */
1648 {
1649 /* Traverse the list of possible unsafe conflicts collecting PIDs. */
1651 {
1654
1656
1658 break;
1659 }
1660 }
1661
1663
1665}
1666
1667/*
1668 * Acquire a snapshot that can be used for the current transaction.
1669 *
1670 * Make sure we have a SERIALIZABLEXACT reference in MySerializableXact.
1671 * It should be current for this process and be contained in PredXact.
1672 *
1673 * The passed-in Snapshot pointer should reference a static data area that
1674 * can safely be passed to GetSnapshotData. The return value is actually
1675 * always this same pointer; no new snapshot data structure is allocated
1676 * within this function.
1677 */
1678Snapshot
1680{
1682
1683 /*
1684 * Can't use serializable mode while recovery is still active, as it is,
1685 * for example, on a hot standby. We could get here despite the check in
1686 * check_transaction_isolation() if default_transaction_isolation is set
1687 * to serializable, so phrase the hint accordingly.
1688 */
1694 errhint("You can use \"SET default_transaction_isolation = 'repeatable read'\" to change the default.")));
1695
1696 /*
1697 * A special optimization is available for SERIALIZABLE READ ONLY
1698 * DEFERRABLE transactions -- we can wait for a suitable snapshot and
1699 * thereby avoid all SSI overhead once it's running.
1700 */
1703
1706}
1707
1708/*
1709 * Import a snapshot to be used for the current transaction.
1710 *
1711 * This is nearly the same as GetSerializableTransactionSnapshot, except that
1712 * we don't take a new snapshot, but rather use the data we're handed.
1713 *
1714 * The caller must have verified that the snapshot came from a serializable
1715 * transaction; and if we're read-write, the source transaction must not be
1716 * read-only.
1717 */
1718void
1722{
1724
1725 /*
1726 * If this is called by parallel.c in a parallel worker, we don't want to
1727 * create a SERIALIZABLEXACT just yet because the leader's
1728 * SERIALIZABLEXACT will be installed with AttachSerializableXact(). We
1729 * also don't want to reject SERIALIZABLE READ ONLY DEFERRABLE in this
1730 * case, because the leader has already determined that the snapshot it
1731 * has passed us is safe. So there is nothing for us to do.
1732 */
1734 return;
1735
1736 /*
1737 * We do not allow SERIALIZABLE READ ONLY DEFERRABLE transactions to
1738 * import snapshots, since there's no way to wait for a safe snapshot when
1739 * we're using the snap we're told to. (XXX instead of throwing an error,
1740 * we could just ignore the XactDeferrable flag?)
1741 */
1746
1748 sourcepid);
1749}
1750
1751/*
1752 * Guts of GetSerializableTransactionSnapshot
1753 *
1754 * If sourcevxid is valid, this is actually an import operation and we should
1755 * skip calling GetSnapshotData, because the snapshot contents are already
1756 * loaded up. HOWEVER: to avoid race conditions, we must check that the
1757 * source xact is still running after we acquire SerializableXactHashLock.
1758 * We do that by calling ProcArrayInstallImportedXmin.
1759 */
1764{
1765 PGPROC *proc;
1766 VirtualTransactionId vxid;
1768 *othersxact;
1769
1770 /* We only do this for serializable transactions. Once. */
1772
1774
1775 /*
1776 * Since all parts of a serializable transaction must use the same
1777 * snapshot, it is too late to establish one after a parallel operation
1778 * has begun.
1779 */
1782
1783 proc = MyProc;
1785 GET_VXID_FROM_PGPROC(vxid, *proc);
1786
1787 /*
1788 * First we get the sxact structure, which may involve looping and access
1789 * to the "finished" list to free a structure for use.
1790 *
1791 * We must hold SerializableXactHashLock when taking/checking the snapshot
1792 * to avoid race conditions, for much the same reasons that
1793 * GetSnapshotData takes the ProcArrayLock. Since we might have to
1794 * release SerializableXactHashLock to call SummarizeOldestCommittedSxact,
1795 * this means we have to create the sxact first, which is a bit annoying
1796 * (in particular, an elog(ERROR) in procarray.c would cause us to leak
1797 * the sxact). Consider refactoring to avoid this.
1798 */
1799#ifdef TEST_SUMMARIZE_SERIAL
1800 SummarizeOldestCommittedSxact();
1801#endif
1803 do
1804 {
1806 /* If null, push out committed sxact to SLRU summary & retry. */
1808 {
1810 SummarizeOldestCommittedSxact();
1812 }
1814
1815 /* Get the snapshot, or check that it's safe to use */
1817 snapshot = GetSnapshotData(snapshot);
1819 {
1826 sourcepid)));
1827 }
1828
1829 /*
1830 * If there are no serializable transactions which are not read-only, we
1831 * can "opt out" of predicate locking and conflict checking for a
1832 * read-only transaction.
1833 *
1834 * The reason this is safe is that a read-only transaction can only become
1835 * part of a dangerous structure if it overlaps a writable transaction
1836 * which in turn overlaps a writable transaction which committed before
1837 * the read-only transaction started. A new writable transaction can
1838 * overlap this one, but it can't meet the other condition of overlapping
1839 * a transaction which committed before this one started.
1840 */
1842 {
1845 return snapshot;
1846 }
1847
1848 /* Initialize the structure. */
1849 sxact->vxid = vxid;
1863 sxact->flags = 0;
1865 {
1866 dlist_iter iter;
1867
1869
1870 /*
1871 * Register all concurrent r/w transactions as possible conflicts; if
1872 * all of them commit without any outgoing conflicts to earlier
1873 * transactions then this snapshot can be deemed safe (and we can run
1874 * without tracking predicate locks).
1875 */
1877 {
1879
1883 {
1885 }
1886 }
1887
1888 /*
1889 * If we didn't find any possibly unsafe conflicts because every
1890 * uncommitted writable transaction turned out to be doomed, then we
1891 * can "opt out" immediately. See comments above the earlier check
1892 * for PredXact->WritableSxactCount == 0.
1893 */
1895 {
1898 return snapshot;
1899 }
1900 }
1901 else
1902 {
1906 }
1907
1908 /* Maintain serializable global xmin info. */
1910 {
1915 }
1917 {
1920 }
1921 else
1922 {
1924 }
1925
1928
1930
1931 CreateLocalPredicateLockHash();
1932
1933 return snapshot;
1934}
1935
1936static void
1938{
1940
1941 /* Initialize the backend-local hash table of parent locks */
1947 &hash_ctl,
1949}
1950
1951/*
1952 * Register the top level XID in SerializableXidHash.
1953 * Also store it for easy reference in MySerializableXact.
1954 */
1955void
1957{
1960 bool found;
1961
1962 /*
1963 * If we're not tracking predicate lock data for this transaction, we
1964 * should ignore the request and return quickly.
1965 */
1967 return;
1968
1969 /* We should have a valid XID and be at the top level. */
1971
1973
1974 /* This should only be done once per transaction. */
1976
1978
1979 sxidtag.xid = xid;
1981 &sxidtag,
1982 HASH_ENTER, &found);
1983 Assert(!found);
1984
1985 /* Initialize the structure. */
1988}
1989
1990
1991/*
1992 * Check whether there are any predicate locks held by any transaction
1993 * for the page at the given block number.
1994 *
1995 * Note that the transaction may be completed but not yet subject to
1996 * cleanup due to overlapping serializable transactions. This must
1997 * return valid information regardless of transaction isolation level.
1998 *
1999 * Also note that this doesn't check for a conflicting relation lock,
2000 * just a lock specifically on the given page.
2001 *
2002 * One use is to support proper behavior during GiST index vacuum.
2003 */
2004bool
2006{
2010 PREDICATELOCKTARGET *target;
2011
2014 relation->rd_id,
2015 blkno);
2016
2020 target = (PREDICATELOCKTARGET *)
2025
2027}
2028
2029
2030/*
2031 * Check whether a particular lock is held by this transaction.
2032 *
2033 * Important note: this function may return false even if the lock is
2034 * being held, because it uses the local lock table which is not
2035 * updated if another transaction modifies our lock list (e.g. to
2036 * split an index page). It can also return true when a coarser
2037 * granularity lock that covers this target is being held. Be careful
2038 * to only use this function in circumstances where such errors are
2039 * acceptable!
2040 */
2041static bool
2043{
2044 LOCALPREDICATELOCK *lock;
2045
2046 /* check local hash table */
2048 targettag,
2050
2051 if (!lock)
2052 return false;
2053
2054 /*
2055 * Found entry in the table, but still need to check whether it's actually
2056 * held -- it could just be a parent of some held lock.
2057 */
2059}
2060
2061/*
2062 * Return the parent lock tag in the lock hierarchy: the next coarser
2063 * lock that covers the provided tag.
2064 *
2065 * Returns true and sets *parent to the parent tag if one exists,
2066 * returns false if none exists.
2067 */
2068static bool
2070 PREDICATELOCKTARGETTAG *parent)
2071{
2073 {
2075 /* relation locks have no parent lock */
2076 return false;
2077
2079 /* parent lock is relation lock */
2080 SET_PREDICATELOCKTARGETTAG_RELATION(*parent,
2081 GET_PREDICATELOCKTARGETTAG_DB(*tag),
2082 GET_PREDICATELOCKTARGETTAG_RELATION(*tag));
2083
2084 return true;
2085
2087 /* parent lock is page lock */
2088 SET_PREDICATELOCKTARGETTAG_PAGE(*parent,
2089 GET_PREDICATELOCKTARGETTAG_DB(*tag),
2090 GET_PREDICATELOCKTARGETTAG_RELATION(*tag),
2091 GET_PREDICATELOCKTARGETTAG_PAGE(*tag));
2092 return true;
2093 }
2094
2095 /* not reachable */
2097 return false;
2098}
2099
2100/*
2101 * Check whether the lock we are considering is already covered by a
2102 * coarser lock for our transaction.
2103 *
2104 * Like PredicateLockExists, this function might return a false
2105 * negative, but it will never return a false positive.
2106 */
2107static bool
2109{
2111 parenttag;
2112
2114
2115 /* check parents iteratively until no more */
2117 {
2120 return true;
2121 }
2122
2123 /* no more parents to check; lock is not covered */
2124 return false;
2125}
2126
2127/*
2128 * Remove the dummy entry from the predicate lock target hash, to free up some
2129 * scratch space. The caller must be holding SerializablePredicateListLock,
2130 * and must restore the entry with RestoreScratchTarget() before releasing the
2131 * lock.
2132 *
2133 * If lockheld is true, the caller is already holding the partition lock
2134 * of the partition containing the scratch entry.
2135 */
2136static void
2138{
2139 bool found;
2140
2142
2146 &ScratchTargetTag,
2147 ScratchTargetTagHash,
2148 HASH_REMOVE, &found);
2149 Assert(found);
2152}
2153
2154/*
2155 * Re-insert the dummy entry in predicate lock target hash.
2156 */
2157static void
2159{
2160 bool found;
2161
2163
2167 &ScratchTargetTag,
2168 ScratchTargetTagHash,
2169 HASH_ENTER, &found);
2170 Assert(!found);
2173}
2174
2175/*
2176 * Check whether the list of related predicate locks is empty for a
2177 * predicate lock target, and remove the target if it is.
2178 */
2179static void
2181{
2183
2185
2186 /* Can't remove it until no locks at this target. */
2188 return;
2189
2190 /* Actually remove the target. */
2192 &target->tag,
2193 targettaghash,
2196}
2197
2198/*
2199 * Delete child target locks owned by this process.
2200 * This implementation is assuming that the usage of each target tag field
2201 * is uniform. No need to make this hard if we don't have to.
2202 *
2203 * We acquire an LWLock in the case of parallel mode, because worker
2204 * backends have access to the leader's SERIALIZABLEXACT. Otherwise,
2205 * we aren't acquiring LWLocks for the predicate lock or lock
2206 * target structures associated with this transaction unless we're going
2207 * to modify them, because no other process is permitted to modify our
2208 * locks.
2209 */
2210static void
2212{
2215 dlist_mutable_iter iter;
2216
2221
2223 {
2227
2229
2234
2236 {
2240
2243
2245
2249 (PredicateLockHash,
2250 &oldlocktag,
2252 oldtargettaghash),
2255
2257
2259
2261 }
2262 }
2266}
2267
2268/*
2269 * Returns the promotion limit for a given predicate lock target. This is the
2270 * max number of descendant locks allowed before promoting to the specified
2271 * tag. Note that the limit includes non-direct descendants (e.g., both tuples
2272 * and pages for a relation lock).
2273 *
2274 * Currently the default limit is 2 for a page lock, and half of the value of
2275 * max_pred_locks_per_transaction - 1 for a relation lock, to match behavior
2276 * of earlier releases when upgrading.
2277 *
2278 * TODO SSI: We should probably add additional GUCs to allow a maximum ratio
2279 * of page and tuple locks based on the pages in a relation, and the maximum
2280 * ratio of tuple locks to tuples in a page. This would provide more
2281 * generally "balanced" allocation of locks to where they are most useful,
2282 * while still allowing the absolute numbers to prevent one relation from
2283 * tying up all predicate lock resources.
2284 */
2285static int
2287{
2289 {
2292 ? (max_predicate_locks_per_xact
2293 / (-max_predicate_locks_per_relation)) - 1
2294 : max_predicate_locks_per_relation;
2295
2298
2300
2301 /*
2302 * not reachable: nothing is finer-granularity than a tuple, so we
2303 * should never try to promote to it.
2304 */
2306 return 0;
2307 }
2308
2309 /* not reachable */
2311 return 0;
2312}
2313
2314/*
2315 * For all ancestors of a newly-acquired predicate lock, increment
2316 * their child count in the parent hash table. If any of them have
2317 * more descendants than their promotion threshold, acquire the
2318 * coarsest such lock.
2319 *
2320 * Returns true if a parent lock was acquired and false otherwise.
2321 */
2322static bool
2324{
2326 nexttag,
2327 promotiontag;
2329 bool found,
2330 promote;
2331
2333
2335
2336 /* check parents iteratively */
2338 {
2341 &targettag,
2342 HASH_ENTER,
2343 &found);
2344 if (!found)
2345 {
2347 parentlock->childLocks = 1;
2348 }
2349 else
2350 parentlock->childLocks++;
2351
2354 {
2355 /*
2356 * We should promote to this parent lock. Continue to check its
2357 * ancestors, however, both to get their child counts right and to
2358 * check whether we should just go ahead and promote to one of
2359 * them.
2360 */
2363 }
2364 }
2365
2367 {
2368 /* acquire coarsest ancestor eligible for promotion */
2370 return true;
2371 }
2372 else
2373 return false;
2374}
2375
2376/*
2377 * When releasing a lock, decrement the child count on all ancestor
2378 * locks.
2379 *
2380 * This is called only when releasing a lock via
2381 * DeleteChildTargetLocks (i.e. when a lock becomes redundant because
2382 * we've acquired its parent, possibly due to promotion) or when a new
2383 * MVCC write lock makes the predicate lock unnecessary. There's no
2384 * point in calling it when locks are released at transaction end, as
2385 * this information is no longer needed.
2386 */
2387static void
2389{
2391 nexttag;
2392
2394
2396 {
2400
2407
2408 /*
2409 * There's a small chance the parent lock doesn't exist in the lock
2410 * table. This can happen if we prematurely removed it because an
2411 * index split caused the child refcount to be off.
2412 */
2414 continue;
2415
2416 parentlock->childLocks--;
2417
2418 /*
2419 * Under similar circumstances the parent lock's refcount might be
2420 * zero. This only happens if we're holding that lock (otherwise we
2421 * would have removed the entry).
2422 */
2424 {
2426 parentlock->childLocks = 0;
2427 }
2428
2430 {
2436 }
2437 }
2438}
2439
2440/*
2441 * Indicate that a predicate lock on the given target is held by the
2442 * specified transaction. Has no effect if the lock is already held.
2443 *
2444 * This updates the lock table and the sxact's lock list, and creates
2445 * the lock target if necessary, but does *not* do anything related to
2446 * granularity promotion or the local lock table. See
2447 * PredicateLockAcquire for that.
2448 */
2449static void
2453{
2454 PREDICATELOCKTARGET *target;
2455 PREDICATELOCKTAG locktag;
2456 PREDICATELOCK *lock;
2458 bool found;
2459
2461
2466
2467 /* Make sure that the target is represented. */
2468 target = (PREDICATELOCKTARGET *)
2471 HASH_ENTER_NULL, &found);
2472 if (!target)
2477 if (!found)
2479
2480 /* We've got the sxact and target, make sure they're joined. */
2481 locktag.myTarget = target;
2483 lock = (PREDICATELOCK *)
2486 HASH_ENTER_NULL, &found);
2487 if (!lock)
2492
2493 if (!found)
2494 {
2498 }
2499
2504}
2505
2506/*
2507 * Acquire a predicate lock on the specified target for the current
2508 * connection if not already held. This updates the local lock table
2509 * and uses it to implement granularity promotion. It will consolidate
2510 * multiple locks into a coarser lock if warranted, and will release
2511 * any finer-grained locks covered by the new one.
2512 */
2513static void
2515{
2517 bool found;
2519
2520 /* Do we have the lock already, or a covering lock? */
2522 return;
2523
2525 return;
2526
2527 /* the same hash and LW lock apply to the lock target and the local lock. */
2529
2530 /* Acquire lock in local table */
2534 HASH_ENTER, &found);
2536 if (!found)
2537 locallock->childLocks = 0;
2538
2539 /* Actually create the lock */
2541
2542 /*
2543 * Lock has been acquired. Check whether it should be promoted to a
2544 * coarser granularity, or whether there are finer-granularity locks to
2545 * clean up.
2546 */
2548 {
2549 /*
2550 * Lock request was promoted to a coarser-granularity lock, and that
2551 * lock was acquired. It will delete this lock and any of its
2552 * children, so we're done.
2553 */
2554 }
2555 else
2556 {
2557 /* Clean up any finer-granularity locks */
2560 }
2561}
2562
2563
2564/*
2565 * PredicateLockRelation
2566 *
2567 * Gets a predicate lock at the relation level.
2568 * Skip if not in full serializable transaction isolation level.
2569 * Skip if this is a temporary table.
2570 * Clear any finer-grained predicate locks this session has on the relation.
2571 */
2572void
2574{
2575 PREDICATELOCKTARGETTAG tag;
2576
2578 return;
2579
2580 SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2582 relation->rd_id);
2583 PredicateLockAcquire(&tag);
2584}
2585
2586/*
2587 * PredicateLockPage
2588 *
2589 * Gets a predicate lock at the page level.
2590 * Skip if not in full serializable transaction isolation level.
2591 * Skip if this is a temporary table.
2592 * Skip if a coarser predicate lock already covers this page.
2593 * Clear any finer-grained predicate locks this session has on the relation.
2594 */
2595void
2597{
2598 PREDICATELOCKTARGETTAG tag;
2599
2601 return;
2602
2603 SET_PREDICATELOCKTARGETTAG_PAGE(tag,
2605 relation->rd_id,
2606 blkno);
2607 PredicateLockAcquire(&tag);
2608}
2609
2610/*
2611 * PredicateLockTID
2612 *
2613 * Gets a predicate lock at the tuple level.
2614 * Skip if not in full serializable transaction isolation level.
2615 * Skip if this is a temporary table.
2616 */
2617void
2620{
2621 PREDICATELOCKTARGETTAG tag;
2622
2624 return;
2625
2626 /*
2627 * Return if this xact wrote it.
2628 */
2630 {
2631 /* If we wrote it; we already have a write lock. */
2633 return;
2634 }
2635
2636 /*
2637 * Do quick-but-not-definitive test for a relation lock first. This will
2638 * never cause a return when the relation is *not* locked, but will
2639 * occasionally let the check continue when there really *is* a relation
2640 * level lock.
2641 */
2642 SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2644 relation->rd_id);
2646 return;
2647
2648 SET_PREDICATELOCKTARGETTAG_TUPLE(tag,
2650 relation->rd_id,
2651 ItemPointerGetBlockNumber(tid),
2652 ItemPointerGetOffsetNumber(tid));
2653 PredicateLockAcquire(&tag);
2654}
2655
2656
2657/*
2658 * DeleteLockTarget
2659 *
2660 * Remove a predicate lock target along with any locks held for it.
2661 *
2662 * Caller must hold SerializablePredicateListLock and the
2663 * appropriate hash partition lock for the target.
2664 */
2665static void
2667{
2668 dlist_mutable_iter iter;
2669
2671 LW_EXCLUSIVE));
2673
2675
2677 {
2680 bool found;
2681
2684
2686 (PredicateLockHash,
2687 &predlock->tag,
2689 targettaghash),
2690 HASH_REMOVE, &found);
2691 Assert(found);
2692 }
2694
2695 /* Remove the target itself, if possible. */
2697}
2698
2699
2700/*
2701 * TransferPredicateLocksToNewTarget
2702 *
2703 * Move or copy all the predicate locks for a lock target, for use by
2704 * index page splits/combines and other things that create or replace
2705 * lock targets. If 'removeOld' is true, the old locks and the target
2706 * will be removed.
2707 *
2708 * Returns true on success, or false if we ran out of shared memory to
2709 * allocate the new target or locks. Guaranteed to always succeed if
2710 * removeOld is set (by using the scratch entry in PredicateLockTargetHash
2711 * for scratch space).
2712 *
2713 * Warning: the "removeOld" option should be used only with care,
2714 * because this function does not (indeed, can not) update other
2715 * backends' LocalPredicateLockHash. If we are only adding new
2716 * entries, this is not a problem: the local lock table is used only
2717 * as a hint, so missing entries for locks that are held are
2718 * OK. Having entries for locks that are no longer held, as can happen
2719 * when using "removeOld", is not in general OK. We can only use it
2720 * safely when replacing a lock with a coarser-granularity lock that
2721 * covers it, or if we are absolutely certain that no one will need to
2722 * refer to that lock in the future.
2723 *
2724 * Caller must hold SerializablePredicateListLock exclusively.
2725 */
2726static bool
2730{
2736 bool found;
2738
2740 LW_EXCLUSIVE));
2741
2746
2748 {
2749 /*
2750 * Remove the dummy entry to give us scratch space, so we know we'll
2751 * be able to create the new lock target.
2752 */
2754 }
2755
2756 /*
2757 * We must get the partition locks in ascending sequence to avoid
2758 * deadlocks. If old and new partitions are the same, we must request the
2759 * lock only once.
2760 */
2762 {
2766 }
2768 {
2772 }
2773 else
2775
2776 /*
2777 * Look for the old target. If not found, that's OK; no predicate locks
2778 * are affected, so we can just clean up and return. If it does exist,
2779 * walk its list of predicate locks and move or copy them to the new
2780 * target.
2781 */
2783 &oldtargettag,
2784 oldtargettaghash,
2786
2788 {
2791 dlist_mutable_iter iter;
2792
2794 &newtargettag,
2795 newtargettaghash,
2796 HASH_ENTER_NULL, &found);
2797
2799 {
2800 /* Failed to allocate due to insufficient shmem */
2803 }
2804
2805 /* If we created a new entry, initialize it */
2806 if (!found)
2808
2810
2811 /*
2812 * Loop through all the locks on the old target, replacing them with
2813 * locks on the new target.
2814 */
2816
2818 {
2823
2825
2827 {
2830
2832 (PredicateLockHash,
2833 &oldpredlock->tag,
2835 oldtargettaghash),
2836 HASH_REMOVE, &found);
2837 Assert(found);
2838 }
2839
2842 &newpredlocktag,
2844 newtargettaghash),
2845 HASH_ENTER_NULL,
2846 &found);
2848 {
2849 /* Out of shared memory. Undo what we've done so far. */
2854 }
2855 if (!found)
2856 {
2858 &(newpredlock->targetLink));
2860 &(newpredlock->xactLink));
2862 }
2863 else
2864 {
2867 }
2868
2872 }
2874
2876 {
2879 }
2880 }
2881
2882
2883exit:
2884 /* Release partition locks in reverse order of acquisition. */
2886 {
2889 }
2891 {
2894 }
2895 else
2897
2899 {
2900 /* We shouldn't run out of memory if we're moving locks */
2902
2903 /* Put the scratch entry back */
2905 }
2906
2908}
2909
2910/*
2911 * Drop all predicate locks of any granularity from the specified relation,
2912 * which can be a heap relation or an index relation. If 'transfer' is true,
2913 * acquire a relation lock on the heap for any transactions with any lock(s)
2914 * on the specified relation.
2915 *
2916 * This requires grabbing a lot of LW locks and scanning the entire lock
2917 * target table for matches. That makes this more expensive than most
2918 * predicate lock management functions, but it will only be called for DDL
2919 * type commands that are expensive anyway, and there are fast returns when
2920 * no serializable transactions are active or the relation is temporary.
2921 *
2922 * We don't use the TransferPredicateLocksToNewTarget function because it
2923 * acquires its own locks on the partitions of the two targets involved,
2924 * and we'll already be holding all partition locks.
2925 *
2926 * We can't throw an error from here, because the call could be from a
2927 * transaction which is not serializable.
2928 *
2929 * NOTE: This is currently only called with transfer set to true, but that may
2930 * change. If we decide to clean up the locks from a table on commit of a
2931 * transaction which executed DROP TABLE, the false condition will be useful.
2932 */
2933static void
2935{
2939 Oid dbId;
2940 Oid relId;
2944 bool found;
2946
2947 /*
2948 * Bail out quickly if there are no serializable transactions running.
2949 * It's safe to check this without taking locks because the caller is
2950 * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
2951 * would matter here can be acquired while that is held.
2952 */
2954 return;
2955
2957 return;
2958
2960 relId = relation->rd_id;
2962 {
2964 heapId = relId;
2965 }
2966 else
2967 {
2970 }
2973 * transfer */
2974
2975 /* Retrieve first time needed, then keep. */
2976 heaptargettaghash = 0;
2978
2979 /* Acquire locks on all lock partitions */
2984
2985 /*
2986 * Remove the dummy entry to give us scratch space, so we know we'll be
2987 * able to create the new lock target.
2988 */
2991
2992 /* Scan through target map */
2994
2996 {
2997 dlist_mutable_iter iter;
2998
2999 /*
3000 * Check whether this is a target which needs attention.
3001 */
3003 continue; /* wrong relation id */
3005 continue; /* wrong database id */
3008 continue; /* already the right lock */
3009
3010 /*
3011 * If we made it here, we have work to do. We make sure the heap
3012 * relation lock exists, then we walk the list of predicate locks for
3013 * the old target we found, moving all locks to the heap relation lock
3014 * -- unless they already hold that.
3015 */
3016
3017 /*
3018 * First make sure we have the heap relation target. We only need to
3019 * do this once.
3020 */
3022 {
3024
3028 &heaptargettag,
3029 heaptargettaghash,
3030 HASH_ENTER, &found);
3031 if (!found)
3033 }
3034
3035 /*
3036 * Loop through all the locks on the old target, replacing them with
3037 * locks on the new target.
3038 */
3040 {
3046
3047 /*
3048 * Remove the old lock first. This avoids the chance of running
3049 * out of lock structure entries for the hash table.
3050 */
3053
3055
3056 /*
3057 * No need for retail delete from oldtarget list, we're removing
3058 * the whole target anyway.
3059 */
3061 &oldpredlock->tag,
3062 HASH_REMOVE, &found);
3063 Assert(found);
3064
3066 {
3068
3073 &newpredlocktag,
3075 heaptargettaghash),
3076 HASH_ENTER,
3077 &found);
3078 if (!found)
3079 {
3081 &(newpredlock->targetLink));
3083 &(newpredlock->xactLink));
3085 }
3086 else
3087 {
3090 }
3091
3095 }
3096 }
3097
3099 &found);
3100 Assert(found);
3101 }
3102
3103 /* Put the scratch entry back */
3106
3107 /* Release locks in reverse order */
3112}
3113
3114/*
3115 * TransferPredicateLocksToHeapRelation
3116 * For all transactions, transfer all predicate locks for the given
3117 * relation to a single relation lock on the heap.
3118 */
3119void
3121{
3123}
3124
3125
3126/*
3127 * PredicateLockPageSplit
3128 *
3129 * Copies any predicate locks for the old page to the new page.
3130 * Skip if this is a temporary table or toast table.
3131 *
3132 * NOTE: A page split (or overflow) affects all serializable transactions,
3133 * even if it occurs in the context of another transaction isolation level.
3134 *
3135 * NOTE: This currently leaves the local copy of the locks without
3136 * information on the new lock which is in shared memory. This could cause
3137 * problems if enough page splits occur on locked pages without the processes
3138 * which hold the locks getting in and noticing.
3139 */
3140void
3143{
3147
3148 /*
3149 * Bail out quickly if there are no serializable transactions running.
3150 *
3151 * It's safe to do this check without taking any additional locks. Even if
3152 * a serializable transaction starts concurrently, we know it can't take
3153 * any SIREAD locks on the page being split because the caller is holding
3154 * the associated buffer page lock. Memory reordering isn't an issue; the
3155 * memory barrier in the LWLock acquisition guarantees that this read
3156 * occurs while the buffer page lock is held.
3157 */
3159 return;
3160
3162 return;
3163
3167
3170 relation->rd_id,
3171 oldblkno);
3174 relation->rd_id,
3175 newblkno);
3176
3178
3179 /*
3180 * Try copying the locks over to the new page's tag, creating it if
3181 * necessary.
3182 */
3184 newtargettag,
3185 false);
3186
3188 {
3189 /*
3190 * No more predicate lock entries are available. Failure isn't an
3191 * option here, so promote the page lock to a relation lock.
3192 */
3193
3194 /* Get the parent relation lock's lock tag */
3196 &newtargettag);
3198
3199 /*
3200 * Move the locks to the parent. This shouldn't fail.
3201 *
3202 * Note that here we are removing locks held by other backends,
3203 * leading to a possible inconsistency in their local lock hash table.
3204 * This is OK because we're replacing it with a lock that covers the
3205 * old one.
3206 */
3208 newtargettag,
3209 true);
3211 }
3212
3214}
3215
3216/*
3217 * PredicateLockPageCombine
3218 *
3219 * Combines predicate locks for two existing pages.
3220 * Skip if this is a temporary table or toast table.
3221 *
3222 * NOTE: A page combine affects all serializable transactions, even if it
3223 * occurs in the context of another transaction isolation level.
3224 */
3225void
3228{
3229 /*
3230 * Page combines differ from page splits in that we ought to be able to
3231 * remove the locks on the old page after transferring them to the new
3232 * page, instead of duplicating them. However, because we can't edit other
3233 * backends' local lock tables, removing the old lock would leave them
3234 * with an entry in their LocalPredicateLockHash for a lock they're not
3235 * holding, which isn't acceptable. So we wind up having to do the same
3236 * work as a page split, acquiring a lock on the new page and keeping the
3237 * old page locked too. That can lead to some false positives, but should
3238 * be rare in practice.
3239 */
3241}
3242
3243/*
3244 * Walk the list of in-progress serializable transactions and find the new
3245 * xmin.
3246 */
3247static void
3249{
3250 dlist_iter iter;
3251
3253
3256
3258 {
3261
3265 {
3270 {
3273 }
3277 }
3278 }
3279
3281}
3282
3283/*
3284 * ReleasePredicateLocks
3285 *
3286 * Releases predicate locks based on completion of the current transaction,
3287 * whether committed or rolled back. It can also be called for a read only
3288 * transaction when it becomes impossible for the transaction to become
3289 * part of a dangerous structure.
3290 *
3291 * We do nothing unless this is a serializable transaction.
3292 *
3293 * This method must ensure that shared memory hash tables are cleaned
3294 * up in some relatively timely fashion.
3295 *
3296 * If this transaction is committing and is holding any predicate locks,
3297 * it must be added to a list of completed serializable transactions still
3298 * holding locks.
3299 *
3300 * If isReadOnlySafe is true, then predicate locks are being released before
3301 * the end of the transaction because MySerializableXact has been determined
3302 * to be RO_SAFE. In non-parallel mode we can release it completely, but it
3303 * in parallel mode we partially release the SERIALIZABLEXACT and keep it
3304 * around until the end of the transaction, allowing each backend to clear its
3305 * MySerializableXact variable and benefit from the optimization in its own
3306 * time.
3307 */
3308void
3310{
3314 dlist_mutable_iter iter;
3315
3316 /*
3317 * We can't trust XactReadOnly here, because a transaction which started
3318 * as READ WRITE can show as READ ONLY later, e.g., within
3319 * subtransactions. We want to flag a transaction as READ ONLY if it
3320 * commits without writing so that de facto READ ONLY transactions get the
3321 * benefit of some RO optimizations, so we will use this local variable to
3322 * get some cleanup logic right which is based on whether the transaction
3323 * was declared READ ONLY at the top level.
3324 */
3326
3327 /* We can't be both committing and releasing early due to RO_SAFE. */
3329
3330 /* Are we at the end of a transaction, that is, a commit or abort? */
3332 {
3333 /*
3334 * Parallel workers mustn't release predicate locks at the end of
3335 * their transaction. The leader will do that at the end of its
3336 * transaction.
3337 */
3339 {
3340 ReleasePredicateLocksLocal();
3341 return;
3342 }
3343
3344 /*
3345 * By the time the leader in a parallel query reaches end of
3346 * transaction, it has waited for all workers to exit.
3347 */
3349
3350 /*
3351 * If the leader in a parallel query earlier stashed a partially
3352 * released SERIALIZABLEXACT for final clean-up at end of transaction
3353 * (because workers might still have been accessing it), then it's
3354 * time to restore it.
3355 */
3357 {
3362 }
3363 }
3364
3366 {
3368 return;
3369 }
3370
3372
3373 /*
3374 * If the transaction is committing, but it has been partially released
3375 * already, then treat this as a roll back. It was marked as rolled back.
3376 */
3379
3380 /*
3381 * If we're called in the middle of a transaction because we discovered
3382 * that the SXACT_FLAG_RO_SAFE flag was set, then we'll partially release
3383 * it (that is, release the predicate locks and conflicts, but not the
3384 * SERIALIZABLEXACT itself) if we're the first backend to have noticed.
3385 */
3387 {
3388 /*
3389 * The leader needs to stash a pointer to it, so that it can
3390 * completely release it at end-of-transaction.
3391 */
3394
3395 /*
3396 * The first backend to reach this condition will partially release
3397 * the SERIALIZABLEXACT. All others will just clear their
3398 * backend-local state so that they stop doing SSI checks for the rest
3399 * of the transaction.
3400 */
3402 {
3404 ReleasePredicateLocksLocal();
3405 return;
3406 }
3407 else
3408 {
3411 /* ... and proceed to perform the partial release below. */
3412 }
3413 }
3419
3420 /* may not be serializable during COMMIT/ROLLBACK PREPARED */
3422
3423 /* We'd better not already be on the cleanup list. */
3425
3427
3428 /*
3429 * We don't hold XidGenLock lock here, assuming that TransactionId is
3430 * atomic!
3431 *
3432 * If this value is changing, we don't care that much whether we get the
3433 * old or new value -- it is just used to determine how far
3434 * SxactGlobalXmin must advance before this transaction can be fully
3435 * cleaned up. The worst that could happen is we wait for one more
3436 * transaction to complete before freeing some RAM; correctness of visible
3437 * behavior is not affected.
3438 */
3440
3441 /*
3442 * If it's not a commit it's either a rollback or a read-only transaction
3443 * flagged SXACT_FLAG_RO_SAFE, and we can clear our locks immediately.
3444 */
3446 {
3449 /* Recognize implicit read-only transaction (commit without write). */
3452 }
3453 else
3454 {
3455 /*
3456 * The DOOMED flag indicates that we intend to roll back this
3457 * transaction and so it should not cause serialization failures for
3458 * other transactions that conflict with it. Note that this flag might
3459 * already be set, if another backend marked this transaction for
3460 * abort.
3461 *
3462 * The ROLLED_BACK flag further indicates that ReleasePredicateLocks
3463 * has been called, and so the SerializableXact is eligible for
3464 * cleanup. This means it should not be considered when calculating
3465 * SxactGlobalXmin.
3466 */
3469
3470 /*
3471 * If the transaction was previously prepared, but is now failing due
3472 * to a ROLLBACK PREPARED or (hopefully very rare) error after the
3473 * prepare, clear the prepared flag. This simplifies conflict
3474 * checking.
3475 */
3477 }
3478
3480 {
3483 {
3484 /*
3485 * Release predicate locks and rw-conflicts in for all committed
3486 * transactions. There are no longer any transactions which might
3487 * conflict with the locks and no chance for new transactions to
3488 * overlap. Similarly, existing conflicts in can't cause pivots,
3489 * and any conflicts in which could have completed a dangerous
3490 * structure would already have caused a rollback, so any
3491 * remaining ones must be benign.
3492 */
3494 }
3495 }
3496 else
3497 {
3498 /*
3499 * Read-only transactions: clear the list of transactions that might
3500 * make us unsafe. Note that we use 'inLink' for the iteration as
3501 * opposed to 'outLink' for the r/w xacts.
3502 */
3504 {
3507
3510
3512 }
3513 }
3514
3515 /* Check for conflict out to old committed transactions. */
3519 {
3520 /*
3521 * we don't know which old committed transaction we conflicted with,
3522 * so be conservative and use FirstNormalSerCommitSeqNo here
3523 */
3527 }
3528
3529 /*
3530 * Release all outConflicts to committed transactions. If we're rolling
3531 * back clear them all. Set SXACT_FLAG_CONFLICT_OUT if any point to
3532 * previously committed transactions.
3533 */
3535 {
3538
3542 {
3547 }
3548
3553 }
3554
3555 /*
3556 * Release all inConflicts from committed and read-only transactions. If
3557 * we're rolling back, clear them all.
3558 */
3560 {
3563
3568 }
3569
3571 {
3572 /*
3573 * Remove ourselves from the list of possible conflicts for concurrent
3574 * READ ONLY transactions, flagging them as unsafe if we have a
3575 * conflict out. If any are waiting DEFERRABLE transactions, wake them
3576 * up if they are known safe or known unsafe.
3577 */
3579 {
3582
3586
3587 /* Mark conflicted if necessary. */
3589 && MyXactDidWrite
3592 <= roXact->SeqNo.lastCommitBeforeSnapshot))
3593 {
3594 /*
3595 * This releases possibleUnsafeConflict (as well as all other
3596 * possible conflicts for roXact)
3597 */
3599 }
3600 else
3601 {
3603
3604 /*
3605 * If we were the last possible conflict, flag it safe. The
3606 * transaction can now safely release its predicate locks (but
3607 * that transaction's backend has to do that itself).
3608 */
3611 }
3612
3613 /*
3614 * Wake up the process for a waiting DEFERRABLE transaction if we
3615 * now know it's either safe or conflicted.
3616 */
3620 }
3621 }
3622
3623 /*
3624 * Check whether it's time to clean up old transactions. This can only be
3625 * done when the last serializable transaction with the oldest xmin among
3626 * serializable transactions completes. We then find the "new oldest"
3627 * xmin and purge any transactions which finished before this transaction
3628 * was launched.
3629 *
3630 * For parallel queries in read-only transactions, it might run twice. We
3631 * only release the reference on the first call.
3632 */
3638 {
3641 {
3642 SetNewSxactGlobalXmin();
3644 }
3645 }
3646
3648
3650
3651 /* Add this to the list of transactions to check for later cleanup. */
3655
3656 /*
3657 * If we're releasing a RO_SAFE transaction in parallel mode, we'll only
3658 * partially release it. That's necessary because other backends may have
3659 * a reference to it. The leader will release the SERIALIZABLEXACT itself
3660 * at the end of the transaction after workers have stopped running.
3661 */
3665 false);
3666
3668
3670 ClearOldPredicateLocks();
3671
3672 ReleasePredicateLocksLocal();
3673}
3674
3675static void
3677{
3680
3681 /* Delete per-transaction lock table */
3683 {
3686 }
3687}
3688
3689/*
3690 * Clear old predicate locks, belonging to committed transactions that are no
3691 * longer interesting to any in-progress transaction.
3692 */
3693static void
3695{
3696 dlist_mutable_iter iter;
3697
3698 /*
3699 * Loop through finished transactions. They are in commit order, so we can
3700 * stop as soon as we find one that's still interesting.
3701 */
3705 {
3708
3712 {
3713 /*
3714 * This transaction committed before any in-progress transaction
3715 * took its snapshot. It's no longer interesting.
3716 */
3721 }
3724 {
3725 /*
3726 * Any active transactions that took their snapshot before this
3727 * transaction committed are read-only, so we can clear part of
3728 * its state.
3729 */
3731
3733 {
3734 /* A read-only transaction can be removed entirely */
3737 }
3738 else
3739 {
3740 /*
3741 * A read-write transaction can only be partially cleared. We
3742 * need to keep the SERIALIZABLEXACT but can release the
3743 * SIREAD locks and conflicts in.
3744 */
3746 }
3747
3750 }
3751 else
3752 {
3753 /* Still interesting. */
3754 break;
3755 }
3756 }
3758
3759 /*
3760 * Loop through predicate locks on dummy transaction for summarized data.
3761 */
3764 {
3768
3774
3775 /*
3776 * If this lock originally belonged to an old enough transaction, we
3777 * can release it.
3778 */
3780 {
3781 PREDICATELOCKTAG tag;
3782 PREDICATELOCKTARGET *target;
3786
3787 tag = predlock->tag;
3788 target = tag.myTarget;
3792
3794
3797
3799 PredicateLockHashCodeFromTargetHashCode(&tag,
3800 targettaghash),
3803
3805 }
3806 }
3807
3810}
3811
3812/*
3813 * This is the normal way to delete anything from any of the predicate
3814 * locking hash tables. Given a transaction which we know can be deleted:
3815 * delete all predicate locks held by that transaction and any predicate
3816 * lock targets which are now unreferenced by a lock; delete all conflicts
3817 * for the transaction; delete all xid values for the transaction; then
3818 * delete the transaction.
3819 *
3820 * When the partial flag is set, we can release all predicate locks and
3821 * in-conflict information -- we've established that there are no longer
3822 * any overlapping read write transactions for which this transaction could
3823 * matter -- but keep the transaction entry itself and any outConflicts.
3824 *
3825 * When the summarize flag is set, we've run short of room for sxact data
3826 * and must summarize to the SLRU. Predicate locks are transferred to a
3827 * dummy "old" transaction, with duplicate locks on a single target
3828 * collapsing to a single lock with the "latest" commitSeqNo from among
3829 * the conflicting locks..
3830 */
3831static void
3834{
3836 dlist_mutable_iter iter;
3837
3842
3843 /*
3844 * First release all the predicate locks held by this xact (or transfer
3845 * them to OldCommittedSxact if summarize is true)
3846 */
3851 {
3854 PREDICATELOCKTAG tag;
3855 PREDICATELOCKTARGET *target;
3859
3860 tag = predlock->tag;
3861 target = tag.myTarget;
3865
3867
3869
3871 PredicateLockHashCodeFromTargetHashCode(&tag,
3872 targettaghash),
3875 {
3876 bool found;
3877
3878 /* Fold into dummy transaction list. */
3881 PredicateLockHashCodeFromTargetHashCode(&tag,
3882 targettaghash),
3883 HASH_ENTER_NULL, &found);
3889 if (found)
3890 {
3895 }
3896 else
3897 {
3899 &predlock->targetLink);
3901 &predlock->xactLink);
3903 }
3904 }
3905 else
3907
3909 }
3910
3911 /*
3912 * Rather than retail removal, just re-init the head after we've run
3913 * through the list.
3914 */
3916
3920
3923
3924 /* Release all outConflicts (unless 'partial' is true) */
3925 if (!partial)
3926 {
3928 {
3931
3935 }
3936 }
3937
3938 /* Release all inConflicts. */
3940 {
3943
3947 }
3948
3949 /* Finally, get rid of the xid and the record of the transaction itself. */
3950 if (!partial)
3951 {
3955 }
3956
3958}
3959
3960/*
3961 * Tests whether the given top level transaction is concurrent with
3962 * (overlaps) our current transaction.
3963 *
3964 * We need to identify the top level transaction for SSI, anyway, so pass
3965 * that to this function to save the overhead of checking the snapshot's
3966 * subxip array.
3967 */
3968static bool
3970{
3972
3975
3977
3979 return false;
3980
3982 return true;
3983
3985}
3986
3987bool
3989{
3991 return false;
3992
3993 /* Check if someone else has already decided that we need to die */
3995 {
3999 errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4001 }
4002
4003 return true;
4004}
4005
4006/*
4007 * CheckForSerializableConflictOut
4008 * A table AM is reading a tuple that has been modified. If it determines
4009 * that the tuple version it is reading is not visible to us, it should
4010 * pass in the top level xid of the transaction that created it.
4011 * Otherwise, if it determines that it is visible to us but it has been
4012 * deleted or there is a newer version available due to an update, it
4013 * should pass in the top level xid of the modifying transaction.
4014 *
4015 * This function will check for overlap with our own transaction. If the given
4016 * xid is also serializable and the transactions overlap (i.e., they cannot see
4017 * each other's writes), then we have a conflict out.
4018 */
4019void
4021{
4025
4027 return;
4028
4029 /* Check if someone else has already decided that we need to die */
4031 {
4035 errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4037 }
4039
4041 return;
4042
4043 /*
4044 * Find sxact or summarized info for the top level xid.
4045 */
4046 sxidtag.xid = xid;
4051 {
4052 /*
4053 * Transaction not found in "normal" SSI structures. Check whether it
4054 * got pushed out to SLRU storage for "old committed" transactions.
4055 */
4057
4060 {
4063 || conflictCommitSeqNo
4070
4076 errdetail_internal("Reason code: Canceled on identification as a pivot, with conflict out to old committed transaction %u.", xid),
4078
4080 }
4081
4082 /* It's not serializable or otherwise not important. */
4084 return;
4085 }
4089 {
4090 /* Can't conflict with ourself or a transaction that will roll back. */
4092 return;
4093 }
4094
4095 /*
4096 * We have a conflict out to a transaction which has a conflict out to a
4097 * summarized transaction. That summarized transaction must have
4098 * committed first, and we can't tell when it committed in relation to our
4099 * snapshot acquisition, so something needs to be canceled.
4100 */
4102 {
4104 {
4107 return;
4108 }
4109 else
4110 {
4117 }
4118 }
4119
4120 /*
4121 * If this is a read-only transaction and the writing transaction has
4122 * committed, and it doesn't have a rw-conflict to a transaction which
4123 * committed before it, no conflict.
4124 */
4129 || MySerializableXact->SeqNo.lastCommitBeforeSnapshot < sxact->SeqNo.earliestOutConflictCommit))
4130 {
4131 /* Read-only transaction will appear to run first. No conflict. */
4133 return;
4134 }
4135
4137 {
4138 /* This write was already in our snapshot; no conflict. */
4140 return;
4141 }
4142
4144 {
4145 /* We don't want duplicate conflict records in the list. */
4147 return;
4148 }
4149
4150 /*
4151 * Flag the conflict. But first, if this conflict creates a dangerous
4152 * structure, ereport an error.
4153 */
4156}
4157
4158/*
4159 * Check a particular target for rw-dependency conflict in. A subroutine of
4160 * CheckForSerializableConflictIn().
4161 */
4162static void
4164{
4167 PREDICATELOCKTARGET *target;
4170 dlist_mutable_iter iter;
4171
4173
4174 /*
4175 * The same hash and LW lock apply to the lock target and the lock itself.
4176 */
4180 target = (PREDICATELOCKTARGET *)
4184 if (!target)
4185 {
4186 /* Nothing has this target locked; we're done here. */
4188 return;
4189 }
4190
4191 /*
4192 * Each lock for an overlapping transaction represents a conflict: a
4193 * rw-dependency in to this transaction.
4194 */
4196
4198 {
4202
4204 {
4205 /*
4206 * If we're getting a write lock on a tuple, we don't need a
4207 * predicate (SIREAD) lock on the same tuple. We can safely remove
4208 * our SIREAD lock, but we'll defer doing so until after the loop
4209 * because that requires upgrading to an exclusive partition lock.
4210 *
4211 * We can't use this optimization within a subtransaction because
4212 * the subtransaction could roll back, and we would be left
4213 * without any lock at the top level.
4214 */
4217 {
4220 }
4221 }
4225 sxact->finishedBefore))
4227 {
4230
4231 /*
4232 * Re-check after getting exclusive lock because the other
4233 * transaction may have flagged a conflict.
4234 */
4238 sxact->finishedBefore))
4240 {
4242 }
4243
4246 }
4247 }
4250
4251 /*
4252 * If we found one of our own SIREAD locks to remove, remove it now.
4253 *
4254 * At this point our transaction already has a RowExclusiveLock on the
4255 * relation, so we are OK to drop the predicate lock on the tuple, if
4256 * found, without fearing that another write against the tuple will occur
4257 * before the MVCC information makes it to the buffer.
4258 */
4260 {
4263
4269
4270 /*
4271 * Remove the predicate lock from shared memory, if it wasn't removed
4272 * while the locks were released. One way that could happen is from
4273 * autovacuum cleaning up an index.
4274 */
4279 &mypredlocktag,
4280 predlockhashcode,
4283 {
4285
4288
4291 &mypredlocktag,
4292 predlockhashcode,
4295
4297 }
4298
4304
4306 {
4307 /*
4308 * Remove entry in local lock table if it exists. It's OK if it
4309 * doesn't exist; that means the lock was transferred to a new
4310 * target by a different backend.
4311 */
4315
4317 }
4318 }
4319}
4320
4321/*
4322 * CheckForSerializableConflictIn
4323 * We are writing the given tuple. If that indicates a rw-conflict
4324 * in from another serializable transaction, take appropriate action.
4325 *
4326 * Skip checking for any granularity for which a parameter is missing.
4327 *
4328 * A tuple update or delete is in conflict if we have a predicate lock
4329 * against the relation or page in which the tuple exists, or against the
4330 * tuple itself.
4331 */
4332void
4333CheckForSerializableConflictIn(Relation relation, const ItemPointerData *tid, BlockNumber blkno)
4334{
4336
4338 return;
4339
4340 /* Check if someone else has already decided that we need to die */
4345 errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict in checking."),
4347
4348 /*
4349 * We're doing a write which might cause rw-conflicts now or later.
4350 * Memorize that fact.
4351 */
4353
4354 /*
4355 * It is important that we check for locks from the finest granularity to
4356 * the coarsest granularity, so that granularity promotion doesn't cause
4357 * us to miss a lock. The new (coarser) lock will be acquired before the
4358 * old (finer) locks are released.
4359 *
4360 * It is not possible to take and hold a lock across the checks for all
4361 * granularities because each target could be in a separate partition.
4362 */
4364 {
4367 relation->rd_id,
4368 ItemPointerGetBlockNumber(tid),
4369 ItemPointerGetOffsetNumber(tid));
4371 }
4372
4374 {
4377 relation->rd_id,
4378 blkno);
4380 }
4381
4384 relation->rd_id);
4386}
4387
4388/*
4389 * CheckTableForSerializableConflictIn
4390 * The entire table is going through a DDL-style logical mass delete
4391 * like TRUNCATE or DROP TABLE. If that causes a rw-conflict in from
4392 * another serializable transaction, take appropriate action.
4393 *
4394 * While these operations do not operate entirely within the bounds of
4395 * snapshot isolation, they can occur inside a serializable transaction, and
4396 * will logically occur after any reads which saw rows which were destroyed
4397 * by these operations, so we do what we can to serialize properly under
4398 * SSI.
4399 *
4400 * The relation passed in must be a heap relation. Any predicate lock of any
4401 * granularity on the heap will cause a rw-conflict in to this transaction.
4402 * Predicate locks on indexes do not matter because they only exist to guard
4403 * against conflicting inserts into the index, and this is a mass *delete*.
4404 * When a table is truncated or dropped, the index will also be truncated
4405 * or dropped, and we'll deal with locks on the index when that happens.
4406 *
4407 * Dropping or truncating a table also needs to drop any existing predicate
4408 * locks on heap tuples or pages, because they're about to go away. This
4409 * should be done before altering the predicate locks because the transaction
4410 * could be rolled back because of a conflict, in which case the lock changes
4411 * are not needed. (At the moment, we don't actually bother to drop the
4412 * existing locks on a dropped or truncated table at the moment. That might
4413 * lead to some false positives, but it doesn't seem worth the trouble.)
4414 */
4415void
4417{
4419 PREDICATELOCKTARGET *target;
4420 Oid dbId;
4423
4424 /*
4425 * Bail out quickly if there are no serializable transactions running.
4426 * It's safe to check this without taking locks because the caller is
4427 * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
4428 * would matter here can be acquired while that is held.
4429 */
4431 return;
4432
4434 return;
4435
4436 /*
4437 * We're doing a write which might cause rw-conflicts now or later.
4438 * Memorize that fact.
4439 */
4441
4443
4446
4451
4452 /* Scan through target list */
4454
4456 {
4457 dlist_mutable_iter iter;
4458
4459 /*
4460 * Check whether this is a target which needs attention.
4461 */
4463 continue; /* wrong relation id */
4465 continue; /* wrong database id */
4466
4467 /*
4468 * Loop through locks for this target and flag conflicts.
4469 */
4471 {
4474
4477 {
4479 }
4480 }
4481 }
4482
4483 /* Release locks in reverse order */
4488}
4489
4490
4491/*
4492 * Flag a rw-dependency between two serializable transactions.
4493 *
4494 * The caller is responsible for ensuring that we have a LW lock on
4495 * the transaction hash table.
4496 */
4497static void
4499{
4501
4502 /* First, see if this conflict causes failure. */
4504
4505 /* Actually do the conflict flagging. */
4510 else
4512}
4513
4514/*----------------------------------------------------------------------------
4515 * We are about to add a RW-edge to the dependency graph - check that we don't
4516 * introduce a dangerous structure by doing so, and abort one of the
4517 * transactions if so.
4518 *
4519 * A serialization failure can only occur if there is a dangerous structure
4520 * in the dependency graph:
4521 *
4522 * Tin ------> Tpivot ------> Tout
4523 * rw rw
4524 *
4525 * Furthermore, Tout must commit first.
4526 *
4527 * One more optimization is that if Tin is declared READ ONLY (or commits
4528 * without writing), we can only have a problem if Tout committed before Tin
4529 * acquired its snapshot.
4530 *----------------------------------------------------------------------------
4531 */
4532static void
4535{
4537
4539
4541
4542 /*------------------------------------------------------------------------
4543 * Check for already-committed writer with rw-conflict out flagged
4544 * (conflict-flag on W means that T2 committed before W):
4545 *
4546 * R ------> W ------> T2
4547 * rw rw
4548 *
4549 * That is a dangerous structure, so we must abort. (Since the writer
4550 * has already committed, we must be the reader)
4551 *------------------------------------------------------------------------
4552 */
4556
4557 /*------------------------------------------------------------------------
4558 * Check whether the writer has become a pivot with an out-conflict
4559 * committed transaction (T2), and T2 committed first:
4560 *
4561 * R ------> W ------> T2
4562 * rw rw
4563 *
4564 * Because T2 must've committed first, there is no anomaly if:
4565 * - the reader committed before T2
4566 * - the writer committed before T2
4567 * - the reader is a READ ONLY transaction and the reader was concurrent
4568 * with T2 (= reader acquired its snapshot before T2 committed)
4569 *
4570 * We also handle the case that T2 is prepared but not yet committed
4571 * here. In that case T2 has already checked for conflicts, so if it
4572 * commits first, making the above conflict real, it's too late for it
4573 * to abort.
4574 *------------------------------------------------------------------------
4575 */
4579 {
4580 dlist_iter iter;
4581
4583 {
4587
4589 && (!SxactIsCommitted(reader)
4593 && (!SxactIsReadOnly(reader)
4595 {
4597 break;
4598 }
4599 }
4600 }
4601
4602 /*------------------------------------------------------------------------
4603 * Check whether the reader has become a pivot with a writer
4604 * that's committed (or prepared):
4605 *
4606 * T0 ------> R ------> W
4607 * rw rw
4608 *
4609 * Because W must've committed first for an anomaly to occur, there is no
4610 * anomaly if:
4611 * - T0 committed before the writer
4612 * - T0 is READ ONLY, and overlaps the writer
4613 *------------------------------------------------------------------------
4614 */
4616 {
4618 {
4620 }
4621 else
4622 {
4623 dlist_iter iter;
4624
4625 /*
4626 * The unconstify is needed as we have no const version of
4627 * dlist_foreach().
4628 */
4630 {
4634
4640 {
4642 break;
4643 }
4644 }
4645 }
4646 }
4647
4649 {
4650 /*
4651 * We have to kill a transaction to avoid a possible anomaly from
4652 * occurring. If the writer is us, we can just ereport() to cause a
4653 * transaction abort. Otherwise we flag the writer for termination,
4654 * causing it to abort when it tries to commit. However, if the writer
4655 * is a prepared transaction, already prepared, we can't abort it
4656 * anymore, so we have to kill the reader instead.
4657 */
4659 {
4666 }
4668 {
4670
4671 /* if we're not the writer, we have to be the reader */
4676 errdetail_internal("Reason code: Canceled on conflict out to pivot %u, during read.", writer->topXid),
4678 }
4680 }
4681}
4682
4683/*
4684 * PreCommit_CheckForSerializationFailure
4685 * Check for dangerous structures in a serializable transaction
4686 * at commit.
4687 *
4688 * We're checking for a dangerous structure as each conflict is recorded.
4689 * The only way we could have a problem at commit is if this is the "out"
4690 * side of a pivot, and neither the "in" side nor the pivot has yet
4691 * committed.
4692 *
4693 * If a dangerous structure is found, the pivot (the near conflict) is
4694 * marked for death, because rolling back another transaction might mean
4695 * that we fail without ever making progress. This transaction is
4696 * committing writes, so letting it commit ensures progress. If we
4697 * canceled the far conflict, it might immediately fail again on retry.
4698 */
4699void
4701{
4703
4705 return;
4706
4708
4710
4711 /*
4712 * Check if someone else has already decided that we need to die. Since
4713 * we set our own DOOMED flag when partially releasing, ignore in that
4714 * case.
4715 */
4718 {
4723 errdetail_internal("Reason code: Canceled on identification as a pivot, during commit attempt."),
4725 }
4726
4728 {
4731
4734 {
4736
4738 {
4741
4746 {
4747 /*
4748 * Normally, we kill the pivot transaction to make sure we
4749 * make progress if the failing transaction is retried.
4750 * However, we can't kill it if it's already prepared, so
4751 * in that case we commit suicide instead.
4752 */
4754 {
4759 errdetail_internal("Reason code: Canceled on commit attempt with conflict in from prepared pivot."),
4761 }
4763 break;
4764 }
4765 }
4766 }
4767 }
4768
4771
4773}
4774
4775/*------------------------------------------------------------------------*/
4776
4777/*
4778 * Two-phase commit support
4779 */
4780
4781/*
4782 * AtPrepare_Locks
4783 * Do the preparatory work for a PREPARE: make 2PC state file
4784 * records for all predicate locks currently held.
4785 */
4786void
4788{
4790 TwoPhasePredicateRecord record;
4791 TwoPhasePredicateXactRecord *xactRecord;
4792 TwoPhasePredicateLockRecord *lockRecord;
4793 dlist_iter iter;
4794
4798
4800 return;
4801
4802 /* Generate an xact record for our SERIALIZABLEXACT */
4806
4807 /*
4808 * Note that we don't include the list of conflicts in our out in the
4809 * statefile, because new conflicts can be added even after the
4810 * transaction prepares. We'll just make a conservative assumption during
4811 * recovery instead.
4812 */
4813
4815 &record, sizeof(record));
4816
4817 /*
4818 * Generate a lock record for each lock.
4819 *
4820 * To do this, we need to walk the predicate lock list in our sxact rather
4821 * than using the local predicate lock table because the latter is not
4822 * guaranteed to be accurate.
4823 */
4825
4826 /*
4827 * No need to take sxact->perXactPredicateListLock in parallel mode
4828 * because there cannot be any parallel workers running while we are
4829 * preparing a transaction.
4830 */
4832
4834 {
4837
4840
4842 &record, sizeof(record));
4843 }
4844
4846}
4847
4848/*
4849 * PostPrepare_Locks
4850 * Clean up after successful PREPARE. Unlike the non-predicate
4851 * lock manager, we do not need to transfer locks to a dummy
4852 * PGPROC because our SERIALIZABLEXACT will stay around
4853 * anyway. We only need to clean up our local state.
4854 */
4855void
4857{
4859 return;
4860
4862
4865
4868
4871}
4872
4873/*
4874 * PredicateLockTwoPhaseFinish
4875 * Release a prepared transaction's predicate locks once it
4876 * commits or aborts.
4877 */
4878void
4880{
4883
4885
4890
4891 /* xid will not be found if it wasn't a serializable transaction */
4893 return;
4894
4895 /* Release its locks */
4898 * something */
4900}
4901
4902/*
4903 * Re-acquire a predicate lock belonging to a transaction that was prepared.
4904 */
4905void
4908{
4909 TwoPhasePredicateRecord *record;
4911
4913
4915
4918
4920 {
4921 /* Per-transaction record. Set up a SERIALIZABLEXACT. */
4922 TwoPhasePredicateXactRecord *xactRecord;
4926 bool found;
4927
4929
4936
4937 /* vxid for a prepared xact is INVALID_PROC_NUMBER/xid; no pid */
4940 sxact->pid = 0;
4942
4943 /* a prepared xact hasn't committed yet */
4947
4949
4950 /*
4951 * Don't need to track this; no transactions running at the time the
4952 * recovered xact started are still active, except possibly other
4953 * prepared xacts and we don't care whether those are RO_SAFE or not.
4954 */
4956
4959
4960 sxact->topXid = xid;
4965 {
4969 }
4970
4971 /*
4972 * We don't know whether the transaction had any conflicts or not, so
4973 * we'll conservatively assume that it had both a conflict in and a
4974 * conflict out, and represent that with the summary conflict flags.
4975 */
4980
4981 /* Register the transaction's xid */
4982 sxidtag.xid = xid;
4984 &sxidtag,
4985 HASH_ENTER, &found);
4987 Assert(!found);
4989
4990 /*
4991 * Update global xmin. Note that this is a special case compared to
4992 * registering a normal transaction, because the global xmin might go
4993 * backwards. That's OK, because until recovery is over we're not
4994 * going to complete any transactions or create any non-prepared
4995 * transactions, so there's no danger of throwing away.
4996 */
4999 {
5003 }
5005 {
5008 }
5009
5011 }
5013 {
5014 /* Lock record. Recreate the PREDICATELOCK */
5015 TwoPhasePredicateLockRecord *lockRecord;
5020
5023
5025 sxidtag.xid = xid;
5029
5033
5035 }
5036}
5037
5038/*
5039 * Prepare to share the current SERIALIZABLEXACT with parallel workers.
5040 * Return a handle object that can be used by AttachSerializableXact() in a
5041 * parallel worker.
5042 */
5045{
5047}
5048
5049/*
5050 * Allow parallel workers to import the leader's SERIALIZABLEXACT.
5051 */
5052void
5054{
5055
5057
5060 CreateLocalPredicateLockHash();
5061}
void * hash_search(HTAB *hashp, const void *keyPtr, HASHACTION action, bool *foundPtr)
Definition dynahash.c:952
Size hash_estimate_size(int64 num_entries, Size entrysize)
Definition dynahash.c:783
HTAB * hash_create(const char *tabname, int64 nelem, const HASHCTL *info, int flags)
Definition dynahash.c:358
void * hash_search_with_hash_value(HTAB *hashp, const void *keyPtr, uint32 hashvalue, HASHACTION action, bool *foundPtr)
Definition dynahash.c:965
#define newval
static void dlist_push_tail(dlist_head *head, dlist_node *node)
Definition ilist.h:364
static OffsetNumber ItemPointerGetOffsetNumber(const ItemPointerData *pointer)
Definition itemptr.h:124
static BlockNumber ItemPointerGetBlockNumber(const ItemPointerData *pointer)
Definition itemptr.h:103
bool LWLockHeldByMeInMode(LWLock *lock, LWLockMode mode)
Definition lwlock.c:1955
static bool pg_lfind32(uint32 key, const uint32 *base, uint32 nelem)
Definition pg_lfind.h:153
PredicateLockData * GetPredicateLockStatusData(void)
Definition predicate.c:1445
void PredicateLockPageSplit(Relation relation, BlockNumber oldblkno, BlockNumber newblkno)
Definition predicate.c:3142
static void DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag)
Definition predicate.c:2389
static void SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact, SERIALIZABLEXACT *activeXact)
Definition predicate.c:666
#define PredicateLockTargetTagHashCode(predicatelocktargettag)
Definition predicate.c:303
void CheckForSerializableConflictIn(Relation relation, const ItemPointerData *tid, BlockNumber blkno)
Definition predicate.c:4334
void SetSerializableTransactionSnapshot(Snapshot snapshot, VirtualTransactionId *sourcevxid, int sourcepid)
Definition predicate.c:1720
static bool RWConflictExists(const SERIALIZABLEXACT *reader, const SERIALIZABLEXACT *writer)
Definition predicate.c:610
static bool PredicateLockingNeededForRelation(Relation relation)
Definition predicate.c:498
static bool SerializationNeededForRead(Relation relation, Snapshot snapshot)
Definition predicate.c:516
void PredicateLockPage(Relation relation, BlockNumber blkno, Snapshot snapshot)
Definition predicate.c:2597
static Snapshot GetSerializableTransactionSnapshotInt(Snapshot snapshot, VirtualTransactionId *sourcevxid, int sourcepid)
Definition predicate.c:1762
static void PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag)
Definition predicate.c:2515
static void SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
Definition predicate.c:643
int GetSafeSnapshotBlockingPids(int blocked_pid, int *output, int output_size)
Definition predicate.c:1626
static void RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target, uint32 targettaghash)
Definition predicate.c:2181
static uint32 predicatelock_hash(const void *key, Size keysize)
Definition predicate.c:1419
void CheckForSerializableConflictOut(Relation relation, TransactionId xid, Snapshot snapshot)
Definition predicate.c:4021
static void DropAllPredicateLocksFromTable(Relation relation, bool transfer)
Definition predicate.c:2935
bool PageIsPredicateLocked(Relation relation, BlockNumber blkno)
Definition predicate.c:2006
static void CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag, uint32 targettaghash, SERIALIZABLEXACT *sxact)
Definition predicate.c:2451
static void SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
Definition predicate.c:858
static bool GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag, PREDICATELOCKTARGETTAG *parent)
Definition predicate.c:2070
#define PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash)
Definition predicate.c:316
static void DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag)
Definition predicate.c:2212
static void DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash)
Definition predicate.c:2667
void PredicateLockTwoPhaseFinish(FullTransactionId fxid, bool isCommit)
Definition predicate.c:4880
void predicatelock_twophase_recover(FullTransactionId fxid, uint16 info, void *recdata, uint32 len)
Definition predicate.c:4907
static int MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag)
Definition predicate.c:2287
void CheckTableForSerializableConflictIn(Relation relation)
Definition predicate.c:4417
void AttachSerializableXact(SerializableXactHandle handle)
Definition predicate.c:5054
static bool PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag)
Definition predicate.c:2043
static void SerialSetActiveSerXmin(TransactionId xid)
Definition predicate.c:990
static dlist_head * FinishedSerializableTransactions
Definition predicate.c:399
static bool SerializationNeededForWrite(Relation relation)
Definition predicate.c:560
static bool CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag)
Definition predicate.c:2324
void PredicateLockPageCombine(Relation relation, BlockNumber oldblkno, BlockNumber newblkno)
Definition predicate.c:3227
static bool SerialPagePrecedesLogically(int64 page1, int64 page2)
Definition predicate.c:731
static void CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag)
Definition predicate.c:4164
void PreCommit_CheckForSerializationFailure(void)
Definition predicate.c:4701
void ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
Definition predicate.c:3310
static void FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
Definition predicate.c:4499
#define PredicateLockHashPartitionLockByIndex(i)
Definition predicate.c:261
static void OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
Definition predicate.c:4534
static bool CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag)
Definition predicate.c:2109
void PredicateLockRelation(Relation relation, Snapshot snapshot)
Definition predicate.c:2574
void PredicateLockTID(Relation relation, const ItemPointerData *tid, Snapshot snapshot, TransactionId tuple_xid)
Definition predicate.c:2619
static SerCommitSeqNo SerialGetMinConflictCommitSeqNo(TransactionId xid)
Definition predicate.c:949
static void SummarizeOldestCommittedSxact(void)
Definition predicate.c:1501
bool check_serial_buffers(int *newval, void **extra, GucSource source)
Definition predicate.c:847
void PostPrepare_PredicateLocks(FullTransactionId fxid)
Definition predicate.c:4857
#define TargetTagIsCoveredBy(covered_target, covering_target)
Definition predicate.c:233
static bool TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag, PREDICATELOCKTARGETTAG newtargettag, bool removeOld)
Definition predicate.c:2728
void RegisterPredicateLockingXid(TransactionId xid)
Definition predicate.c:1957
#define PredicateLockHashPartitionLock(hashcode)
Definition predicate.c:258
static void ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial, bool summarize)
Definition predicate.c:3833
bool CheckForSerializableConflictOutNeeded(Relation relation, Snapshot snapshot)
Definition predicate.c:3989
void TransferPredicateLocksToHeapRelation(Relation relation)
Definition predicate.c:3121
static void CreateLocalPredicateLockHash(void)
Definition predicate.c:1938
Snapshot GetSerializableTransactionSnapshot(Snapshot snapshot)
Definition predicate.c:1680
#define SXACT_FLAG_DEFERRABLE_WAITING
Definition predicate_internals.h:132
#define SXACT_FLAG_SUMMARY_CONFLICT_IN
Definition predicate_internals.h:135
#define GET_PREDICATELOCKTARGETTAG_DB(locktag)
Definition predicate_internals.h:406
#define GET_PREDICATELOCKTARGETTAG_RELATION(locktag)
Definition predicate_internals.h:408
#define RWConflictPoolHeaderDataSize
Definition predicate_internals.h:214
#define SET_PREDICATELOCKTARGETTAG_PAGE(locktag, dboid, reloid, blocknum)
Definition predicate_internals.h:394
#define GET_PREDICATELOCKTARGETTAG_TYPE(locktag)
Definition predicate_internals.h:414
#define SET_PREDICATELOCKTARGETTAG_RELATION(locktag, dboid, reloid)
Definition predicate_internals.h:388
#define SET_PREDICATELOCKTARGETTAG_TUPLE(locktag, dboid, reloid, blocknum, offnum)
Definition predicate_internals.h:400
#define SXACT_FLAG_PARTIALLY_RELEASED
Definition predicate_internals.h:142
#define GET_PREDICATELOCKTARGETTAG_PAGE(locktag)
Definition predicate_internals.h:410
#define SXACT_FLAG_SUMMARY_CONFLICT_OUT
Definition predicate_internals.h:136
#define GET_PREDICATELOCKTARGETTAG_OFFSET(locktag)
Definition predicate_internals.h:412
bool ProcArrayInstallImportedXmin(TransactionId xmin, VirtualTransactionId *sourcevxid)
Definition procarray.c:2482
HTAB * ShmemInitHash(const char *name, int64 init_size, int64 max_size, HASHCTL *infoP, int hash_flags)
Definition shmem.c:334
void * ShmemInitStruct(const char *name, Size size, bool *foundPtr)
Definition shmem.c:389
void SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns, const char *subdir, int buffer_tranche_id, int bank_tranche_id, SyncRequestHandler sync_handler, bool long_segment_names)
Definition slru.c:252
int SimpleLruReadPage_ReadOnly(SlruCtl ctl, int64 pageno, TransactionId xid)
Definition slru.c:630
int SimpleLruReadPage(SlruCtl ctl, int64 pageno, bool write_ok, TransactionId xid)
Definition slru.c:527
static LWLock * SimpleLruGetBankLock(SlruCtl ctl, int64 pageno)
Definition slru.h:160
Definition transam.h:66
Definition hsearch.h:66
Definition hsearch.h:121
Definition dynahash.c:222
Definition itemptr.h:37
Definition predicate_internals.h:348
Definition lwlock.h:42
Definition proc.h:179
Definition predicate_internals.h:303
Definition predicate_internals.h:285
Definition predicate_internals.h:318
Definition predicate_internals.h:145
SerCommitSeqNo LastSxactCommitSeqNo
Definition predicate_internals.h:161
SerCommitSeqNo CanPartialClearThrough
Definition predicate_internals.h:165
SERIALIZABLEXACT * OldCommittedSxact
Definition predicate_internals.h:172
SerCommitSeqNo HavePartialClearedThrough
Definition predicate_internals.h:170
Definition predicate_internals.h:376
Definition predicate_internals.h:194
Definition rel.h:56
Definition predicate_internals.h:59
LWLock perXactPredicateListLock
Definition predicate_internals.h:102
SerCommitSeqNo lastCommitBeforeSnapshot
Definition predicate_internals.h:85
dlist_head possibleUnsafeConflicts
Definition predicate_internals.h:108
union SERIALIZABLEXACT::@129 SeqNo
SerCommitSeqNo earliestOutConflictCommit
Definition predicate_internals.h:83
Definition predicate_internals.h:223
Definition predicate_internals.h:241
Definition predicate.c:346
Definition snapshot.h:139
PREDICATELOCKTARGETTAG target
Definition predicate_internals.h:450
union TwoPhasePredicateRecord::@130 data
TwoPhasePredicateLockRecord lockRecord
Definition predicate_internals.h:460
TwoPhasePredicateXactRecord xactRecord
Definition predicate_internals.h:459
Definition lock.h:62
Definition ilist.h:152
Definition ilist.h:178
Definition ilist.h:199
static bool TransactionIdFollows(TransactionId id1, TransactionId id2)
Definition transam.h:297
static bool TransactionIdPrecedesOrEquals(TransactionId id1, TransactionId id2)
Definition transam.h:282
static bool TransactionIdFollowsOrEquals(TransactionId id1, TransactionId id2)
Definition transam.h:312
static bool TransactionIdPrecedes(TransactionId id1, TransactionId id2)
Definition transam.h:263
void RegisterTwoPhaseRecord(TwoPhaseRmgrId rmid, uint16 info, const void *data, uint32 len)
Definition twophase.c:1271
bool TransactionIdIsCurrentTransactionId(TransactionId xid)
Definition xact.c:942
◆ PredicateLockHashCodeFromTargetHashCode
| #define PredicateLockHashCodeFromTargetHashCode | ( | predicatelocktag, | |
| targethash | |||
| ) |
Value:
Definition at line 316 of file predicate.c.
◆ PredicateLockHashPartition
| #define PredicateLockHashPartition | ( | hashcode | ) | ((hashcode) % NUM_PREDICATELOCK_PARTITIONS) |
Definition at line 256 of file predicate.c.
◆ PredicateLockHashPartitionLock
| #define PredicateLockHashPartitionLock | ( | hashcode | ) |
Value:
PredicateLockHashPartition(hashcode)].lock)
#define PREDICATELOCK_MANAGER_LWLOCK_OFFSET
Definition lwlock.h:105
Definition at line 258 of file predicate.c.
◆ PredicateLockHashPartitionLockByIndex
| #define PredicateLockHashPartitionLockByIndex | ( | i | ) | (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + (i)].lock) |
Definition at line 261 of file predicate.c.
◆ PredicateLockTargetTagHashCode
| #define PredicateLockTargetTagHashCode | ( | predicatelocktargettag | ) | get_hash_value(PredicateLockTargetHash, predicatelocktargettag) |
Definition at line 303 of file predicate.c.
◆ SERIAL_ENTRIESPERPAGE
| #define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE) |
Definition at line 330 of file predicate.c.
◆ SERIAL_ENTRYSIZE
| #define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo) |
Definition at line 329 of file predicate.c.
◆ SERIAL_MAX_PAGE
| #define SERIAL_MAX_PAGE (MaxTransactionId / SERIAL_ENTRIESPERPAGE) |
Definition at line 335 of file predicate.c.
◆ SERIAL_PAGESIZE
Definition at line 328 of file predicate.c.
◆ SerialNextPage
| #define SerialNextPage | ( | page | ) | (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1) |
Definition at line 337 of file predicate.c.
◆ SerialPage
| #define SerialPage | ( | xid | ) | (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE) |
Definition at line 343 of file predicate.c.
◆ SerialSlruCtl
| #define SerialSlruCtl (&SerialSlruCtlData) |
Definition at line 326 of file predicate.c.
◆ SerialValue
Value:
(*((SerCommitSeqNo *) \
(SerialSlruCtl->shared->page_buffer[slotno] + \
Definition at line 339 of file predicate.c.
◆ SxactHasConflictOut
| #define SxactHasConflictOut | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_CONFLICT_OUT) != 0) |
Definition at line 289 of file predicate.c.
◆ SxactHasSummaryConflictIn
| #define SxactHasSummaryConflictIn | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_IN) != 0) |
Definition at line 282 of file predicate.c.
◆ SxactHasSummaryConflictOut
| #define SxactHasSummaryConflictOut | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_OUT) != 0) |
Definition at line 283 of file predicate.c.
◆ SxactIsCommitted
| #define SxactIsCommitted | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_COMMITTED) != 0) |
Definition at line 277 of file predicate.c.
◆ SxactIsDeferrableWaiting
| #define SxactIsDeferrableWaiting | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_DEFERRABLE_WAITING) != 0) |
Definition at line 290 of file predicate.c.
◆ SxactIsDoomed
| #define SxactIsDoomed | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_DOOMED) != 0) |
Definition at line 280 of file predicate.c.
◆ SxactIsOnFinishedList
| #define SxactIsOnFinishedList | ( | sxact | ) | (!dlist_node_is_detached(&(sxact)->finishedLink)) |
Definition at line 267 of file predicate.c.
◆ SxactIsPartiallyReleased
| #define SxactIsPartiallyReleased | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_PARTIALLY_RELEASED) != 0) |
Definition at line 293 of file predicate.c.
◆ SxactIsPrepared
| #define SxactIsPrepared | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_PREPARED) != 0) |
Definition at line 278 of file predicate.c.
◆ SxactIsReadOnly
| #define SxactIsReadOnly | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_READ_ONLY) != 0) |
Definition at line 281 of file predicate.c.
◆ SxactIsRolledBack
| #define SxactIsRolledBack | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_ROLLED_BACK) != 0) |
Definition at line 279 of file predicate.c.
◆ SxactIsROSafe
| #define SxactIsROSafe | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_RO_SAFE) != 0) |
Definition at line 291 of file predicate.c.
◆ SxactIsROUnsafe
| #define SxactIsROUnsafe | ( | sxact | ) | (((sxact)->flags & SXACT_FLAG_RO_UNSAFE) != 0) |
Definition at line 292 of file predicate.c.
◆ TargetTagIsCoveredBy
| #define TargetTagIsCoveredBy | ( | covered_target, | |
| covering_target | |||
| ) |
Value:
InvalidOffsetNumber) /* (3) */ \
&& (((GET_PREDICATELOCKTARGETTAG_OFFSET(covered_target) != \
InvalidOffsetNumber) /* (4a) */ \
|| ((GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
InvalidBlockNumber) /* (4b) */ \
!= InvalidBlockNumber))) \
Definition at line 233 of file predicate.c.
Typedef Documentation
◆ SerialControl
Definition at line 352 of file predicate.c.
◆ SerialControlData
Function Documentation
◆ AtPrepare_PredicateLocks()
Definition at line 4788 of file predicate.c.
4789{
4791 TwoPhasePredicateRecord record;
4792 TwoPhasePredicateXactRecord *xactRecord;
4793 TwoPhasePredicateLockRecord *lockRecord;
4794 dlist_iter iter;
4795
4799
4801 return;
4802
4803 /* Generate an xact record for our SERIALIZABLEXACT */
4807
4808 /*
4809 * Note that we don't include the list of conflicts in our out in the
4810 * statefile, because new conflicts can be added even after the
4811 * transaction prepares. We'll just make a conservative assumption during
4812 * recovery instead.
4813 */
4814
4816 &record, sizeof(record));
4817
4818 /*
4819 * Generate a lock record for each lock.
4820 *
4821 * To do this, we need to walk the predicate lock list in our sxact rather
4822 * than using the local predicate lock table because the latter is not
4823 * guaranteed to be accurate.
4824 */
4826
4827 /*
4828 * No need to take sxact->perXactPredicateListLock in parallel mode
4829 * because there cannot be any parallel workers running while we are
4830 * preparing a transaction.
4831 */
4833
4835 {
4838
4841
4843 &record, sizeof(record));
4844 }
4845
4847}
References Assert, dlist_iter::cur, TwoPhasePredicateRecord::data, dlist_container, dlist_foreach, fb(), SERIALIZABLEXACT::flags, TwoPhasePredicateXactRecord::flags, InvalidSerializableXact, IsParallelWorker, TwoPhasePredicateRecord::lockRecord, LW_SHARED, LWLockAcquire(), LWLockRelease(), MySerializableXact, ParallelContextActive(), RegisterTwoPhaseRecord(), TwoPhasePredicateLockRecord::target, TWOPHASE_RM_PREDICATELOCK_ID, TWOPHASEPREDICATERECORD_LOCK, TWOPHASEPREDICATERECORD_XACT, TwoPhasePredicateRecord::type, TwoPhasePredicateRecord::xactRecord, SERIALIZABLEXACT::xmin, and TwoPhasePredicateXactRecord::xmin.
Referenced by PrepareTransaction().
◆ AttachSerializableXact()
| void AttachSerializableXact | ( | SerializableXactHandle | handle | ) |
Definition at line 5054 of file predicate.c.
References Assert, CreateLocalPredicateLockHash(), InvalidSerializableXact, and MySerializableXact.
Referenced by ParallelWorkerMain().
◆ check_serial_buffers()
Definition at line 847 of file predicate.c.
References check_slru_buffers(), and newval.
◆ CheckAndPromotePredicateLockRequest()
|
static |
Definition at line 2324 of file predicate.c.
2325{
2327 nexttag,
2328 promotiontag;
2330 bool found,
2331 promote;
2332
2334
2336
2337 /* check parents iteratively */
2339 {
2342 &targettag,
2343 HASH_ENTER,
2344 &found);
2345 if (!found)
2346 {
2348 parentlock->childLocks = 1;
2349 }
2350 else
2351 parentlock->childLocks++;
2352
2355 {
2356 /*
2357 * We should promote to this parent lock. Continue to check its
2358 * ancestors, however, both to get their child counts right and to
2359 * check whether we should just go ahead and promote to one of
2360 * them.
2361 */
2364 }
2365 }
2366
2368 {
2369 /* acquire coarsest ancestor eligible for promotion */
2371 return true;
2372 }
2373 else
2374 return false;
2375}
References fb(), GetParentPredicateLockTag(), HASH_ENTER, hash_search(), LOCALPREDICATELOCK::held, LocalPredicateLockHash, MaxPredicateChildLocks(), and PredicateLockAcquire().
Referenced by PredicateLockAcquire().
◆ CheckForSerializableConflictIn()
| void CheckForSerializableConflictIn | ( | Relation | relation, |
| const ItemPointerData * | tid, | ||
| BlockNumber | blkno | ||
| ) |
Definition at line 4334 of file predicate.c.
4335{
4337
4339 return;
4340
4341 /* Check if someone else has already decided that we need to die */
4346 errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict in checking."),
4348
4349 /*
4350 * We're doing a write which might cause rw-conflicts now or later.
4351 * Memorize that fact.
4352 */
4354
4355 /*
4356 * It is important that we check for locks from the finest granularity to
4357 * the coarsest granularity, so that granularity promotion doesn't cause
4358 * us to miss a lock. The new (coarser) lock will be acquired before the
4359 * old (finer) locks are released.
4360 *
4361 * It is not possible to take and hold a lock across the checks for all
4362 * granularities because each target could be in a separate partition.
4363 */
4365 {
4368 relation->rd_id,
4369 ItemPointerGetBlockNumber(tid),
4370 ItemPointerGetOffsetNumber(tid));
4372 }
4373
4375 {
4378 relation->rd_id,
4379 blkno);
4381 }
4382
4385 relation->rd_id);
4387}
References CheckTargetForConflictsIn(), RelFileLocator::dbOid, ereport, errcode(), ERRCODE_T_R_SERIALIZATION_FAILURE, errdetail_internal(), errhint(), errmsg(), ERROR, fb(), InvalidBlockNumber, ItemPointerGetBlockNumber(), ItemPointerGetOffsetNumber(), MySerializableXact, MyXactDidWrite, RelationData::rd_id, RelationData::rd_locator, SerializationNeededForWrite(), SET_PREDICATELOCKTARGETTAG_PAGE, SET_PREDICATELOCKTARGETTAG_RELATION, SET_PREDICATELOCKTARGETTAG_TUPLE, and SxactIsDoomed.
Referenced by _bt_check_unique(), _bt_doinsert(), _hash_doinsert(), ginEntryInsert(), ginFindLeafPage(), ginHeapTupleFastInsert(), gistinserttuples(), heap_delete(), heap_insert(), heap_multi_insert(), heap_update(), and index_insert().
◆ CheckForSerializableConflictOut()
| void CheckForSerializableConflictOut | ( | Relation | relation, |
| TransactionId | xid, | ||
| Snapshot | snapshot | ||
| ) |
Definition at line 4021 of file predicate.c.
4022{
4026
4028 return;
4029
4030 /* Check if someone else has already decided that we need to die */
4032 {
4036 errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4038 }
4040
4042 return;
4043
4044 /*
4045 * Find sxact or summarized info for the top level xid.
4046 */
4047 sxidtag.xid = xid;
4052 {
4053 /*
4054 * Transaction not found in "normal" SSI structures. Check whether it
4055 * got pushed out to SLRU storage for "old committed" transactions.
4056 */
4058
4061 {
4064 || conflictCommitSeqNo
4071
4077 errdetail_internal("Reason code: Canceled on identification as a pivot, with conflict out to old committed transaction %u.", xid),
4079
4081 }
4082
4083 /* It's not serializable or otherwise not important. */
4085 return;
4086 }
4090 {
4091 /* Can't conflict with ourself or a transaction that will roll back. */
4093 return;
4094 }
4095
4096 /*
4097 * We have a conflict out to a transaction which has a conflict out to a
4098 * summarized transaction. That summarized transaction must have
4099 * committed first, and we can't tell when it committed in relation to our
4100 * snapshot acquisition, so something needs to be canceled.
4101 */
4103 {
4105 {
4108 return;
4109 }
4110 else
4111 {
4118 }
4119 }
4120
4121 /*
4122 * If this is a read-only transaction and the writing transaction has
4123 * committed, and it doesn't have a rw-conflict to a transaction which
4124 * committed before it, no conflict.
4125 */
4130 || MySerializableXact->SeqNo.lastCommitBeforeSnapshot < sxact->SeqNo.earliestOutConflictCommit))
4131 {
4132 /* Read-only transaction will appear to run first. No conflict. */
4134 return;
4135 }
4136
4138 {
4139 /* This write was already in our snapshot; no conflict. */
4141 return;
4142 }
4143
4145 {
4146 /* We don't want duplicate conflict records in the list. */
4148 return;
4149 }
4150
4151 /*
4152 * Flag the conflict. But first, if this conflict creates a dangerous
4153 * structure, ereport an error.
4154 */
4157}
References Assert, dlist_is_empty(), ereport, errcode(), ERRCODE_T_R_SERIALIZATION_FAILURE, errdetail_internal(), errhint(), errmsg(), ERROR, fb(), FlagRWConflict(), SERIALIZABLEXACT::flags, GetTopTransactionIdIfAny(), HASH_FIND, hash_search(), SERIALIZABLEXACT::inConflicts, InvalidSerCommitSeqNo, SERIALIZABLEXACT::lastCommitBeforeSnapshot, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MySerializableXact, RWConflictExists(), SERIALIZABLEXACT::SeqNo, SerialGetMinConflictCommitSeqNo(), SerializableXidHash, SerializationNeededForRead(), SXACT_FLAG_DOOMED, SXACT_FLAG_SUMMARY_CONFLICT_OUT, SxactHasConflictOut, SxactHasSummaryConflictIn, SxactHasSummaryConflictOut, SxactIsCommitted, SxactIsDoomed, SxactIsPrepared, SxactIsReadOnly, TransactionIdEquals, TransactionIdIsValid, and XidIsConcurrent().
Referenced by HeapCheckForSerializableConflictOut().
◆ CheckForSerializableConflictOutNeeded()
Definition at line 3989 of file predicate.c.
3990{
3992 return false;
3993
3994 /* Check if someone else has already decided that we need to die */
3996 {
4000 errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4002 }
4003
4004 return true;
4005}
References ereport, errcode(), ERRCODE_T_R_SERIALIZATION_FAILURE, errdetail_internal(), errhint(), errmsg(), ERROR, MySerializableXact, SerializationNeededForRead(), and SxactIsDoomed.
Referenced by heap_prepare_pagescan(), and HeapCheckForSerializableConflictOut().
◆ CheckPointPredicate()
Definition at line 1041 of file predicate.c.
1042{
1044
1046
1047 /* Exit quickly if the SLRU is currently not in use. */
1049 {
1051 return;
1052 }
1053
1055 {
1057
1059
1060 /*
1061 * It is possible for the tailXid to be ahead of the headXid. This
1062 * occurs if we checkpoint while there are in-progress serializable
1063 * transaction(s) advancing the tail but we are yet to summarize the
1064 * transactions. In this case, we cutoff up to the headPage and the
1065 * next summary will advance the headXid.
1066 */
1068 {
1069 /* We can truncate the SLRU up to the page containing tailXid */
1071 }
1072 else
1074 }
1075 else
1076 {
1077 /*----------
1078 * The SLRU is no longer needed. Truncate to head before we set head
1079 * invalid.
1080 *
1081 * XXX: It's possible that the SLRU is not needed again until XID
1082 * wrap-around has happened, so that the segment containing headPage
1083 * that we leave behind will appear to be new again. In that case it
1084 * won't be removed until XID horizon advances enough to make it
1085 * current again.
1086 *
1087 * XXX: This should happen in vac_truncate_clog(), not in checkpoints.
1088 * Consider this scenario, starting from a system with no in-progress
1089 * transactions and VACUUM FREEZE having maximized oldestXact:
1090 * - Start a SERIALIZABLE transaction.
1091 * - Start, finish, and summarize a SERIALIZABLE transaction, creating
1092 * one SLRU page.
1093 * - Consume XIDs to reach xidStopLimit.
1094 * - Finish all transactions. Due to the long-running SERIALIZABLE
1095 * transaction, earlier checkpoints did not touch headPage. The
1096 * next checkpoint will change it, but that checkpoint happens after
1097 * the end of the scenario.
1098 * - VACUUM to advance XID limits.
1099 * - Consume ~2M XIDs, crossing the former xidWrapLimit.
1100 * - Start, finish, and summarize a SERIALIZABLE transaction.
1101 * SerialAdd() declines to create the targetPage, because headPage
1102 * is not regarded as in the past relative to that targetPage. The
1103 * transaction instigating the summarize fails in
1104 * SimpleLruReadPage().
1105 */
1108 }
1109
1111
1112 /*
1113 * Truncate away pages that are no longer required. Note that no
1114 * additional locking is required, because this is only called as part of
1115 * a checkpoint, and the validity limits have already been determined.
1116 */
1118
1119 /*
1120 * Write dirty SLRU pages to disk
1121 *
1122 * This is not actually necessary from a correctness point of view. We do
1123 * it merely as a debugging aid.
1124 *
1125 * We're doing this after the truncation to avoid writing pages right
1126 * before deleting the file in which they sit, which would be completely
1127 * pointless.
1128 */
1130}
References fb(), SerialControlData::headPage, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), serialControl, SerialPage, SerialPagePrecedesLogically(), SerialSlruCtl, SimpleLruTruncate(), SimpleLruWriteAll(), SerialControlData::tailXid, and TransactionIdIsValid.
Referenced by CheckPointGuts().
◆ CheckTableForSerializableConflictIn()
Definition at line 4417 of file predicate.c.
4418{
4420 PREDICATELOCKTARGET *target;
4421 Oid dbId;
4424
4425 /*
4426 * Bail out quickly if there are no serializable transactions running.
4427 * It's safe to check this without taking locks because the caller is
4428 * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
4429 * would matter here can be acquired while that is held.
4430 */
4432 return;
4433
4435 return;
4436
4437 /*
4438 * We're doing a write which might cause rw-conflicts now or later.
4439 * Memorize that fact.
4440 */
4442
4444
4447
4452
4453 /* Scan through target list */
4455
4457 {
4458 dlist_mutable_iter iter;
4459
4460 /*
4461 * Check whether this is a target which needs attention.
4462 */
4464 continue; /* wrong relation id */
4466 continue; /* wrong database id */
4467
4468 /*
4469 * Loop through locks for this target and flag conflicts.
4470 */
4472 {
4475
4478 {
4480 }
4481 }
4482 }
4483
4484 /* Release locks in reverse order */
4489}
References Assert, dlist_mutable_iter::cur, RelFileLocator::dbOid, dlist_container, dlist_foreach_modify, fb(), FlagRWConflict(), GET_PREDICATELOCKTARGETTAG_DB, GET_PREDICATELOCKTARGETTAG_RELATION, hash_seq_init(), hash_seq_search(), i, LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), MySerializableXact, MyXactDidWrite, NUM_PREDICATELOCK_PARTITIONS, PredicateLockHashPartitionLockByIndex, PREDICATELOCKTARGET::predicateLocks, PredicateLockTargetHash, PredXact, RelationData::rd_id, RelationData::rd_index, RelationData::rd_locator, RWConflictExists(), SerializationNeededForWrite(), PredXactListData::SxactGlobalXmin, PREDICATELOCKTARGET::tag, and TransactionIdIsValid.
Referenced by ExecuteTruncateGuts(), and heap_drop_with_catalog().
◆ CheckTargetForConflictsIn()
|
static |
Definition at line 4164 of file predicate.c.
4165{
4168 PREDICATELOCKTARGET *target;
4171 dlist_mutable_iter iter;
4172
4174
4175 /*
4176 * The same hash and LW lock apply to the lock target and the lock itself.
4177 */
4181 target = (PREDICATELOCKTARGET *)
4185 if (!target)
4186 {
4187 /* Nothing has this target locked; we're done here. */
4189 return;
4190 }
4191
4192 /*
4193 * Each lock for an overlapping transaction represents a conflict: a
4194 * rw-dependency in to this transaction.
4195 */
4197
4199 {
4203
4205 {
4206 /*
4207 * If we're getting a write lock on a tuple, we don't need a
4208 * predicate (SIREAD) lock on the same tuple. We can safely remove
4209 * our SIREAD lock, but we'll defer doing so until after the loop
4210 * because that requires upgrading to an exclusive partition lock.
4211 *
4212 * We can't use this optimization within a subtransaction because
4213 * the subtransaction could roll back, and we would be left
4214 * without any lock at the top level.
4215 */
4218 {
4221 }
4222 }
4226 sxact->finishedBefore))
4228 {
4231
4232 /*
4233 * Re-check after getting exclusive lock because the other
4234 * transaction may have flagged a conflict.
4235 */
4239 sxact->finishedBefore))
4241 {
4243 }
4244
4247 }
4248 }
4251
4252 /*
4253 * If we found one of our own SIREAD locks to remove, remove it now.
4254 *
4255 * At this point our transaction already has a RowExclusiveLock on the
4256 * relation, so we are OK to drop the predicate lock on the tuple, if
4257 * found, without fearing that another write against the tuple will occur
4258 * before the MVCC information makes it to the buffer.
4259 */
4261 {
4264
4270
4271 /*
4272 * Remove the predicate lock from shared memory, if it wasn't removed
4273 * while the locks were released. One way that could happen is from
4274 * autovacuum cleaning up an index.
4275 */
4280 &mypredlocktag,
4281 predlockhashcode,
4284 {
4286
4289
4292 &mypredlocktag,
4293 predlockhashcode,
4296
4298 }
4299
4305
4307 {
4308 /*
4309 * Remove entry in local lock table if it exists. It's OK if it
4310 * doesn't exist; that means the lock was transferred to a new
4311 * target by a different backend.
4312 */
4316
4318 }
4319 }
4320}
References Assert, dlist_mutable_iter::cur, DecrementParentLocks(), dlist_container, dlist_delete(), dlist_foreach_modify, fb(), FlagRWConflict(), GET_PREDICATELOCKTARGETTAG_OFFSET, GetTransactionSnapshot(), HASH_FIND, HASH_REMOVE, hash_search_with_hash_value(), InvalidSerializableXact, IsInParallelMode(), IsSubTransaction(), LocalPredicateLockHash, LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), MySerializableXact, SERIALIZABLEXACT::perXactPredicateListLock, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, PREDICATELOCKTARGET::predicateLocks, PredicateLockTargetHash, PredicateLockTargetTagHashCode, RemoveTargetIfNoLongerUsed(), RWConflictExists(), SxactIsCommitted, SxactIsDoomed, and TransactionIdPrecedes().
Referenced by CheckForSerializableConflictIn().
◆ ClearOldPredicateLocks()
Definition at line 3695 of file predicate.c.
3696{
3697 dlist_mutable_iter iter;
3698
3699 /*
3700 * Loop through finished transactions. They are in commit order, so we can
3701 * stop as soon as we find one that's still interesting.
3702 */
3706 {
3709
3713 {
3714 /*
3715 * This transaction committed before any in-progress transaction
3716 * took its snapshot. It's no longer interesting.
3717 */
3722 }
3725 {
3726 /*
3727 * Any active transactions that took their snapshot before this
3728 * transaction committed are read-only, so we can clear part of
3729 * its state.
3730 */
3732
3734 {
3735 /* A read-only transaction can be removed entirely */
3738 }
3739 else
3740 {
3741 /*
3742 * A read-write transaction can only be partially cleared. We
3743 * need to keep the SERIALIZABLEXACT but can release the
3744 * SIREAD locks and conflicts in.
3745 */
3747 }
3748
3751 }
3752 else
3753 {
3754 /* Still interesting. */
3755 break;
3756 }
3757 }
3759
3760 /*
3761 * Loop through predicate locks on dummy transaction for summarized data.
3762 */
3765 {
3769
3775
3776 /*
3777 * If this lock originally belonged to an old enough transaction, we
3778 * can release it.
3779 */
3781 {
3782 PREDICATELOCKTAG tag;
3783 PREDICATELOCKTARGET *target;
3787
3788 tag = predlock->tag;
3789 target = tag.myTarget;
3793
3795
3798
3800 PredicateLockHashCodeFromTargetHashCode(&tag,
3801 targettaghash),
3804
3806 }
3807 }
3808
3811}
References Assert, PredXactListData::CanPartialClearThrough, dlist_mutable_iter::cur, dlist_container, dlist_delete(), dlist_delete_thoroughly(), dlist_foreach_modify, fb(), FinishedSerializableTransactions, HASH_REMOVE, hash_search_with_hash_value(), PredXactListData::HavePartialClearedThrough, InvalidSerCommitSeqNo, LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), PREDICATELOCKTAG::myTarget, OldCommittedSxact, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, SERIALIZABLEXACT::predicateLocks, PredicateLockTargetTagHashCode, PredXact, ReleaseOneSerializableXact(), RemoveTargetIfNoLongerUsed(), PredXactListData::SxactGlobalXmin, SxactIsReadOnly, PREDICATELOCKTARGET::tag, TransactionIdIsValid, and TransactionIdPrecedesOrEquals().
Referenced by ReleasePredicateLocks().
◆ CoarserLockCovers()
|
static |
Definition at line 2109 of file predicate.c.
2110{
2112 parenttag;
2113
2115
2116 /* check parents iteratively until no more */
2118 {
2121 return true;
2122 }
2123
2124 /* no more parents to check; lock is not covered */
2125 return false;
2126}
References fb(), GetParentPredicateLockTag(), and PredicateLockExists().
Referenced by PredicateLockAcquire().
◆ CreateLocalPredicateLockHash()
Definition at line 1938 of file predicate.c.
References Assert, fb(), HASH_BLOBS, hash_create(), HASH_ELEM, LocalPredicateLockHash, and max_predicate_locks_per_xact.
Referenced by AttachSerializableXact(), and GetSerializableTransactionSnapshotInt().
◆ CreatePredicateLock()
|
static |
Definition at line 2451 of file predicate.c.
2454{
2455 PREDICATELOCKTARGET *target;
2456 PREDICATELOCKTAG locktag;
2457 PREDICATELOCK *lock;
2459 bool found;
2460
2462
2467
2468 /* Make sure that the target is represented. */
2469 target = (PREDICATELOCKTARGET *)
2472 HASH_ENTER_NULL, &found);
2473 if (!target)
2478 if (!found)
2480
2481 /* We've got the sxact and target, make sure they're joined. */
2482 locktag.myTarget = target;
2484 lock = (PREDICATELOCK *)
2487 HASH_ENTER_NULL, &found);
2488 if (!lock)
2493
2494 if (!found)
2495 {
2499 }
2500
2505}
References PREDICATELOCK::commitSeqNo, dlist_init(), dlist_push_tail(), ereport, errcode(), errhint(), errmsg(), ERROR, fb(), HASH_ENTER_NULL, hash_search_with_hash_value(), InvalidSerCommitSeqNo, IsInParallelMode(), LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), PREDICATELOCKTAG::myTarget, PREDICATELOCKTAG::myXact, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, PREDICATELOCKTARGET::predicateLocks, PredicateLockTargetHash, PREDICATELOCK::targetLink, and PREDICATELOCK::xactLink.
Referenced by predicatelock_twophase_recover(), and PredicateLockAcquire().
◆ CreatePredXact()
|
static |
Definition at line 582 of file predicate.c.
583{
585
588
593}
References PredXactListData::activeList, PredXactListData::availableList, dlist_container, dlist_is_empty(), dlist_pop_head_node(), dlist_push_tail(), fb(), and PredXact.
Referenced by GetSerializableTransactionSnapshotInt(), predicatelock_twophase_recover(), and PredicateLockShmemInit().
◆ DecrementParentLocks()
|
static |
Definition at line 2389 of file predicate.c.
2390{
2392 nexttag;
2393
2395
2397 {
2401
2408
2409 /*
2410 * There's a small chance the parent lock doesn't exist in the lock
2411 * table. This can happen if we prematurely removed it because an
2412 * index split caused the child refcount to be off.
2413 */
2415 continue;
2416
2417 parentlock->childLocks--;
2418
2419 /*
2420 * Under similar circumstances the parent lock's refcount might be
2421 * zero. This only happens if we're holding that lock (otherwise we
2422 * would have removed the entry).
2423 */
2425 {
2427 parentlock->childLocks = 0;
2428 }
2429
2431 {
2437 }
2438 }
2439}
References Assert, fb(), GetParentPredicateLockTag(), HASH_FIND, HASH_REMOVE, hash_search_with_hash_value(), LocalPredicateLockHash, PG_USED_FOR_ASSERTS_ONLY, and PredicateLockTargetTagHashCode.
Referenced by CheckTargetForConflictsIn(), and DeleteChildTargetLocks().
◆ DeleteChildTargetLocks()
|
static |
Definition at line 2212 of file predicate.c.
2213{
2216 dlist_mutable_iter iter;
2217
2222
2224 {
2228
2230
2235
2237 {
2241
2244
2246
2250 (PredicateLockHash,
2251 &oldlocktag,
2253 oldtargettaghash),
2256
2258
2260
2262 }
2263 }
2267}
References Assert, dlist_mutable_iter::cur, DecrementParentLocks(), dlist_container, dlist_delete(), dlist_foreach_modify, fb(), HASH_REMOVE, hash_search_with_hash_value(), IsInParallelMode(), LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), MySerializableXact, PG_USED_FOR_ASSERTS_ONLY, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, PredicateLockTargetTagHashCode, RemoveTargetIfNoLongerUsed(), and TargetTagIsCoveredBy.
Referenced by PredicateLockAcquire().
◆ DeleteLockTarget()
|
static |
Definition at line 2667 of file predicate.c.
2668{
2669 dlist_mutable_iter iter;
2670
2672 LW_EXCLUSIVE));
2674
2676
2678 {
2681 bool found;
2682
2685
2687 (PredicateLockHash,
2688 &predlock->tag,
2690 targettaghash),
2691 HASH_REMOVE, &found);
2692 Assert(found);
2693 }
2695
2696 /* Remove the target itself, if possible. */
2698}
References Assert, dlist_mutable_iter::cur, dlist_container, dlist_delete(), dlist_foreach_modify, fb(), HASH_REMOVE, hash_search_with_hash_value(), LW_EXCLUSIVE, LWLockAcquire(), LWLockHeldByMe(), LWLockHeldByMeInMode(), LWLockRelease(), PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, PREDICATELOCKTARGET::predicateLocks, and RemoveTargetIfNoLongerUsed().
Referenced by TransferPredicateLocksToNewTarget().
◆ DropAllPredicateLocksFromTable()
Definition at line 2935 of file predicate.c.
2936{
2940 Oid dbId;
2941 Oid relId;
2945 bool found;
2947
2948 /*
2949 * Bail out quickly if there are no serializable transactions running.
2950 * It's safe to check this without taking locks because the caller is
2951 * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
2952 * would matter here can be acquired while that is held.
2953 */
2955 return;
2956
2958 return;
2959
2961 relId = relation->rd_id;
2963 {
2965 heapId = relId;
2966 }
2967 else
2968 {
2971 }
2974 * transfer */
2975
2976 /* Retrieve first time needed, then keep. */
2977 heaptargettaghash = 0;
2979
2980 /* Acquire locks on all lock partitions */
2985
2986 /*
2987 * Remove the dummy entry to give us scratch space, so we know we'll be
2988 * able to create the new lock target.
2989 */
2992
2993 /* Scan through target map */
2995
2997 {
2998 dlist_mutable_iter iter;
2999
3000 /*
3001 * Check whether this is a target which needs attention.
3002 */
3004 continue; /* wrong relation id */
3006 continue; /* wrong database id */
3009 continue; /* already the right lock */
3010
3011 /*
3012 * If we made it here, we have work to do. We make sure the heap
3013 * relation lock exists, then we walk the list of predicate locks for
3014 * the old target we found, moving all locks to the heap relation lock
3015 * -- unless they already hold that.
3016 */
3017
3018 /*
3019 * First make sure we have the heap relation target. We only need to
3020 * do this once.
3021 */
3023 {
3025
3029 &heaptargettag,
3030 heaptargettaghash,
3031 HASH_ENTER, &found);
3032 if (!found)
3034 }
3035
3036 /*
3037 * Loop through all the locks on the old target, replacing them with
3038 * locks on the new target.
3039 */
3041 {
3047
3048 /*
3049 * Remove the old lock first. This avoids the chance of running
3050 * out of lock structure entries for the hash table.
3051 */
3054
3056
3057 /*
3058 * No need for retail delete from oldtarget list, we're removing
3059 * the whole target anyway.
3060 */
3062 &oldpredlock->tag,
3063 HASH_REMOVE, &found);
3064 Assert(found);
3065
3067 {
3069
3074 &newpredlocktag,
3076 heaptargettaghash),
3077 HASH_ENTER,
3078 &found);
3079 if (!found)
3080 {
3082 &(newpredlock->targetLink));
3084 &(newpredlock->xactLink));
3086 }
3087 else
3088 {
3091 }
3092
3096 }
3097 }
3098
3100 &found);
3101 Assert(found);
3102 }
3103
3104 /* Put the scratch entry back */
3107
3108 /* Release locks in reverse order */
3113}
References Assert, SERIALIZABLEXACT::commitSeqNo, dlist_mutable_iter::cur, RelFileLocator::dbOid, dlist_container, dlist_delete(), dlist_foreach_modify, dlist_init(), dlist_push_tail(), fb(), GET_PREDICATELOCKTARGETTAG_DB, GET_PREDICATELOCKTARGETTAG_RELATION, GET_PREDICATELOCKTARGETTAG_TYPE, HASH_ENTER, HASH_REMOVE, hash_search(), hash_search_with_hash_value(), hash_seq_init(), hash_seq_search(), i, InvalidOid, InvalidSerCommitSeqNo, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), PREDICATELOCKTAG::myTarget, NUM_PREDICATELOCK_PARTITIONS, OldCommittedSxact, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLockByIndex, PredicateLockingNeededForRelation(), PredicateLockTargetHash, PredicateLockTargetTagHashCode, PREDLOCKTAG_RELATION, PredXact, RelationData::rd_id, RelationData::rd_index, RelationData::rd_locator, RemoveScratchTarget(), RestoreScratchTarget(), SET_PREDICATELOCKTARGETTAG_RELATION, PredXactListData::SxactGlobalXmin, and TransactionIdIsValid.
Referenced by TransferPredicateLocksToHeapRelation().
◆ FlagRWConflict()
|
static |
Definition at line 4499 of file predicate.c.
4500{
4502
4503 /* First, see if this conflict causes failure. */
4505
4506 /* Actually do the conflict flagging. */
4511 else
4513}
References Assert, fb(), SERIALIZABLEXACT::flags, OldCommittedSxact, OnConflict_CheckForSerializationFailure(), SetRWConflict(), SXACT_FLAG_SUMMARY_CONFLICT_IN, and SXACT_FLAG_SUMMARY_CONFLICT_OUT.
Referenced by CheckForSerializableConflictOut(), CheckTableForSerializableConflictIn(), and CheckTargetForConflictsIn().
◆ FlagSxactUnsafe()
|
static |
Definition at line 699 of file predicate.c.
700{
701 dlist_mutable_iter iter;
702
705
707
708 /*
709 * We know this isn't a safe snapshot, so we can stop looking for other
710 * potential conflicts.
711 */
713 {
716
719
721 }
722}
References Assert, dlist_mutable_iter::cur, dlist_container, dlist_foreach_modify, fb(), ReleaseRWConflict(), SXACT_FLAG_RO_UNSAFE, SxactIsReadOnly, and SxactIsROSafe.
Referenced by ReleasePredicateLocks().
◆ GetParentPredicateLockTag()
|
static |
Definition at line 2070 of file predicate.c.
2072{
2074 {
2076 /* relation locks have no parent lock */
2077 return false;
2078
2080 /* parent lock is relation lock */
2081 SET_PREDICATELOCKTARGETTAG_RELATION(*parent,
2082 GET_PREDICATELOCKTARGETTAG_DB(*tag),
2083 GET_PREDICATELOCKTARGETTAG_RELATION(*tag));
2084
2085 return true;
2086
2088 /* parent lock is page lock */
2089 SET_PREDICATELOCKTARGETTAG_PAGE(*parent,
2090 GET_PREDICATELOCKTARGETTAG_DB(*tag),
2091 GET_PREDICATELOCKTARGETTAG_RELATION(*tag),
2092 GET_PREDICATELOCKTARGETTAG_PAGE(*tag));
2093 return true;
2094 }
2095
2096 /* not reachable */
2098 return false;
2099}
References Assert, GET_PREDICATELOCKTARGETTAG_DB, GET_PREDICATELOCKTARGETTAG_PAGE, GET_PREDICATELOCKTARGETTAG_RELATION, GET_PREDICATELOCKTARGETTAG_TYPE, PREDLOCKTAG_PAGE, PREDLOCKTAG_RELATION, PREDLOCKTAG_TUPLE, SET_PREDICATELOCKTARGETTAG_PAGE, and SET_PREDICATELOCKTARGETTAG_RELATION.
Referenced by CheckAndPromotePredicateLockRequest(), CoarserLockCovers(), DecrementParentLocks(), and PredicateLockPageSplit().
◆ GetPredicateLockStatusData()
| PredicateLockData * GetPredicateLockStatusData | ( | void | ) |
Definition at line 1445 of file predicate.c.
1446{
1450 el;
1453
1455
1456 /*
1457 * To ensure consistency, take simultaneous locks on all partition locks
1458 * in ascending order, then SerializableXactHashLock.
1459 */
1463
1464 /* Get number of locks and allocate appropriately-sized arrays. */
1469
1470
1471 /* Scan through PredicateLockHash and copy contents */
1473
1474 el = 0;
1475
1477 {
1480 el++;
1481 }
1482
1484
1485 /* Release locks in reverse order */
1489
1491}
References Assert, data, fb(), hash_get_num_entries(), hash_seq_init(), hash_seq_search(), i, LW_SHARED, LWLockAcquire(), LWLockRelease(), NUM_PREDICATELOCK_PARTITIONS, palloc_array, palloc_object, PredicateLockHash, and PredicateLockHashPartitionLockByIndex.
Referenced by pg_lock_status().
◆ GetSafeSnapshot()
Definition at line 1556 of file predicate.c.
1557{
1558 Snapshot snapshot;
1559
1561
1562 while (true)
1563 {
1564 /*
1565 * GetSerializableTransactionSnapshotInt is going to call
1566 * GetSnapshotData, so we need to provide it the static snapshot area
1567 * our caller passed to us. The pointer returned is actually the same
1568 * one passed to it, but we avoid assuming that here.
1569 */
1572
1574 return snapshot; /* no concurrent r/w xacts; it's safe */
1575
1577
1578 /*
1579 * Wait for concurrent transactions to finish. Stop early if one of
1580 * them marked us as conflicted.
1581 */
1585 {
1589 }
1591
1593 {
1595 break; /* success */
1596 }
1597
1599
1600 /* else, need to retry... */
1605 }
1606
1607 /*
1608 * Now we have a safe snapshot, so we don't need to do any further checks.
1609 */
1612
1613 return snapshot;
1614}
References Assert, DEBUG2, dlist_is_empty(), ereport, errcode(), ERRCODE_T_R_SERIALIZATION_FAILURE, errmsg_internal(), fb(), SERIALIZABLEXACT::flags, GetSerializableTransactionSnapshotInt(), InvalidPid, InvalidSerializableXact, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MySerializableXact, SERIALIZABLEXACT::possibleUnsafeConflicts, ProcWaitForSignal(), ReleasePredicateLocks(), SXACT_FLAG_DEFERRABLE_WAITING, SxactIsROSafe, SxactIsROUnsafe, XactDeferrable, and XactReadOnly.
Referenced by GetSerializableTransactionSnapshot().
◆ GetSafeSnapshotBlockingPids()
Definition at line 1626 of file predicate.c.
1627{
1629 dlist_iter iter;
1631
1633
1634 /* Find blocked_pid's SERIALIZABLEXACT by linear search. */
1636 {
1639
1641 {
1643 break;
1644 }
1645 }
1646
1647 /* Did we find it, and is it currently waiting in GetSafeSnapshot? */
1649 {
1650 /* Traverse the list of possible unsafe conflicts collecting PIDs. */
1652 {
1655
1657
1659 break;
1660 }
1661 }
1662
1664
1666}
References PredXactListData::activeList, dlist_iter::cur, dlist_container, dlist_foreach, fb(), LW_SHARED, LWLockAcquire(), LWLockRelease(), output, PredXact, and SxactIsDeferrableWaiting.
Referenced by pg_isolation_test_session_is_blocked(), and pg_safe_snapshot_blocking_pids().
◆ GetSerializableTransactionSnapshot()
Definition at line 1680 of file predicate.c.
1681{
1683
1684 /*
1685 * Can't use serializable mode while recovery is still active, as it is,
1686 * for example, on a hot standby. We could get here despite the check in
1687 * check_transaction_isolation() if default_transaction_isolation is set
1688 * to serializable, so phrase the hint accordingly.
1689 */
1695 errhint("You can use \"SET default_transaction_isolation = 'repeatable read'\" to change the default.")));
1696
1697 /*
1698 * A special optimization is available for SERIALIZABLE READ ONLY
1699 * DEFERRABLE transactions -- we can wait for a suitable snapshot and
1700 * thereby avoid all SSI overhead once it's running.
1701 */
1704
1707}
References Assert, ereport, errcode(), errdetail(), errhint(), errmsg(), ERROR, fb(), GetSafeSnapshot(), GetSerializableTransactionSnapshotInt(), InvalidPid, IsolationIsSerializable, RecoveryInProgress(), XactDeferrable, and XactReadOnly.
Referenced by GetTransactionSnapshot().
◆ GetSerializableTransactionSnapshotInt()
|
static |
Definition at line 1762 of file predicate.c.
1765{
1766 PGPROC *proc;
1767 VirtualTransactionId vxid;
1769 *othersxact;
1770
1771 /* We only do this for serializable transactions. Once. */
1773
1775
1776 /*
1777 * Since all parts of a serializable transaction must use the same
1778 * snapshot, it is too late to establish one after a parallel operation
1779 * has begun.
1780 */
1783
1784 proc = MyProc;
1786 GET_VXID_FROM_PGPROC(vxid, *proc);
1787
1788 /*
1789 * First we get the sxact structure, which may involve looping and access
1790 * to the "finished" list to free a structure for use.
1791 *
1792 * We must hold SerializableXactHashLock when taking/checking the snapshot
1793 * to avoid race conditions, for much the same reasons that
1794 * GetSnapshotData takes the ProcArrayLock. Since we might have to
1795 * release SerializableXactHashLock to call SummarizeOldestCommittedSxact,
1796 * this means we have to create the sxact first, which is a bit annoying
1797 * (in particular, an elog(ERROR) in procarray.c would cause us to leak
1798 * the sxact). Consider refactoring to avoid this.
1799 */
1800#ifdef TEST_SUMMARIZE_SERIAL
1801 SummarizeOldestCommittedSxact();
1802#endif
1804 do
1805 {
1807 /* If null, push out committed sxact to SLRU summary & retry. */
1809 {
1811 SummarizeOldestCommittedSxact();
1813 }
1815
1816 /* Get the snapshot, or check that it's safe to use */
1818 snapshot = GetSnapshotData(snapshot);
1820 {
1827 sourcepid)));
1828 }
1829
1830 /*
1831 * If there are no serializable transactions which are not read-only, we
1832 * can "opt out" of predicate locking and conflict checking for a
1833 * read-only transaction.
1834 *
1835 * The reason this is safe is that a read-only transaction can only become
1836 * part of a dangerous structure if it overlaps a writable transaction
1837 * which in turn overlaps a writable transaction which committed before
1838 * the read-only transaction started. A new writable transaction can
1839 * overlap this one, but it can't meet the other condition of overlapping
1840 * a transaction which committed before this one started.
1841 */
1843 {
1846 return snapshot;
1847 }
1848
1849 /* Initialize the structure. */
1850 sxact->vxid = vxid;
1864 sxact->flags = 0;
1866 {
1867 dlist_iter iter;
1868
1870
1871 /*
1872 * Register all concurrent r/w transactions as possible conflicts; if
1873 * all of them commit without any outgoing conflicts to earlier
1874 * transactions then this snapshot can be deemed safe (and we can run
1875 * without tracking predicate locks).
1876 */
1878 {
1880
1884 {
1886 }
1887 }
1888
1889 /*
1890 * If we didn't find any possibly unsafe conflicts because every
1891 * uncommitted writable transaction turned out to be doomed, then we
1892 * can "opt out" immediately. See comments above the earlier check
1893 * for PredXact->WritableSxactCount == 0.
1894 */
1896 {
1899 return snapshot;
1900 }
1901 }
1902 else
1903 {
1907 }
1908
1909 /* Maintain serializable global xmin info. */
1911 {
1916 }
1918 {
1921 }
1922 else
1923 {
1925 }
1926
1929
1931
1932 CreateLocalPredicateLockHash();
1933
1934 return snapshot;
1935}
References PredXactListData::activeList, Assert, CreateLocalPredicateLockHash(), CreatePredXact(), dlist_iter::cur, dlist_container, dlist_foreach, dlist_init(), dlist_is_empty(), dlist_node_init(), elog, ereport, errcode(), errdetail(), errmsg(), ERROR, fb(), GET_VXID_FROM_PGPROC, GetSnapshotData(), GetTopTransactionIdIfAny(), InvalidSerCommitSeqNo, InvalidSerializableXact, InvalidTransactionId, IsInParallelMode(), PredXactListData::LastSxactCommitSeqNo, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), max_prepared_xacts, MaxBackends, MyProc, MyProcNumber, MyProcPid, MySerializableXact, MyXactDidWrite, PredXact, ProcArrayInstallImportedXmin(), RecoveryInProgress(), ReleasePredXact(), SerialSetActiveSerXmin(), SetPossibleUnsafeConflict(), SummarizeOldestCommittedSxact(), SXACT_FLAG_READ_ONLY, PredXactListData::SxactGlobalXmin, PredXactListData::SxactGlobalXminCount, SxactIsCommitted, SxactIsDoomed, SxactIsReadOnly, TransactionIdEquals, TransactionIdFollows(), TransactionIdIsValid, PredXactListData::WritableSxactCount, XactReadOnly, and SnapshotData::xmin.
Referenced by GetSafeSnapshot(), GetSerializableTransactionSnapshot(), and SetSerializableTransactionSnapshot().
◆ MaxPredicateChildLocks()
|
static |
Definition at line 2287 of file predicate.c.
2288{
2290 {
2293 ? (max_predicate_locks_per_xact
2294 / (-max_predicate_locks_per_relation)) - 1
2295 : max_predicate_locks_per_relation;
2296
2299
2301
2302 /*
2303 * not reachable: nothing is finer-granularity than a tuple, so we
2304 * should never try to promote to it.
2305 */
2307 return 0;
2308 }
2309
2310 /* not reachable */
2312 return 0;
2313}
References Assert, GET_PREDICATELOCKTARGETTAG_TYPE, max_predicate_locks_per_page, max_predicate_locks_per_relation, max_predicate_locks_per_xact, PREDLOCKTAG_PAGE, PREDLOCKTAG_RELATION, and PREDLOCKTAG_TUPLE.
Referenced by CheckAndPromotePredicateLockRequest().
◆ OnConflict_CheckForSerializationFailure()
|
static |
Definition at line 4534 of file predicate.c.
4536{
4538
4540
4542
4543 /*------------------------------------------------------------------------
4544 * Check for already-committed writer with rw-conflict out flagged
4545 * (conflict-flag on W means that T2 committed before W):
4546 *
4547 * R ------> W ------> T2
4548 * rw rw
4549 *
4550 * That is a dangerous structure, so we must abort. (Since the writer
4551 * has already committed, we must be the reader)
4552 *------------------------------------------------------------------------
4553 */
4557
4558 /*------------------------------------------------------------------------
4559 * Check whether the writer has become a pivot with an out-conflict
4560 * committed transaction (T2), and T2 committed first:
4561 *
4562 * R ------> W ------> T2
4563 * rw rw
4564 *
4565 * Because T2 must've committed first, there is no anomaly if:
4566 * - the reader committed before T2
4567 * - the writer committed before T2
4568 * - the reader is a READ ONLY transaction and the reader was concurrent
4569 * with T2 (= reader acquired its snapshot before T2 committed)
4570 *
4571 * We also handle the case that T2 is prepared but not yet committed
4572 * here. In that case T2 has already checked for conflicts, so if it
4573 * commits first, making the above conflict real, it's too late for it
4574 * to abort.
4575 *------------------------------------------------------------------------
4576 */
4580 {
4581 dlist_iter iter;
4582
4584 {
4588
4590 && (!SxactIsCommitted(reader)
4594 && (!SxactIsReadOnly(reader)
4596 {
4598 break;
4599 }
4600 }
4601 }
4602
4603 /*------------------------------------------------------------------------
4604 * Check whether the reader has become a pivot with a writer
4605 * that's committed (or prepared):
4606 *
4607 * T0 ------> R ------> W
4608 * rw rw
4609 *
4610 * Because W must've committed first for an anomaly to occur, there is no
4611 * anomaly if:
4612 * - T0 committed before the writer
4613 * - T0 is READ ONLY, and overlaps the writer
4614 *------------------------------------------------------------------------
4615 */
4617 {
4619 {
4621 }
4622 else
4623 {
4624 dlist_iter iter;
4625
4626 /*
4627 * The unconstify is needed as we have no const version of
4628 * dlist_foreach().
4629 */
4631 {
4635
4641 {
4643 break;
4644 }
4645 }
4646 }
4647 }
4648
4650 {
4651 /*
4652 * We have to kill a transaction to avoid a possible anomaly from
4653 * occurring. If the writer is us, we can just ereport() to cause a
4654 * transaction abort. Otherwise we flag the writer for termination,
4655 * causing it to abort when it tries to commit. However, if the writer
4656 * is a prepared transaction, already prepared, we can't abort it
4657 * anymore, so we have to kill the reader instead.
4658 */
4660 {
4667 }
4669 {
4671
4672 /* if we're not the writer, we have to be the reader */
4677 errdetail_internal("Reason code: Canceled on conflict out to pivot %u, during read.", writer->topXid),
4679 }
4681 }
4682}
References Assert, SERIALIZABLEXACT::commitSeqNo, dlist_iter::cur, dlist_container, dlist_foreach, ereport, errcode(), ERRCODE_T_R_SERIALIZATION_FAILURE, errdetail_internal(), errhint(), errmsg(), ERROR, fb(), SERIALIZABLEXACT::lastCommitBeforeSnapshot, LWLockHeldByMe(), LWLockRelease(), MySerializableXact, SERIALIZABLEXACT::SeqNo, SXACT_FLAG_DOOMED, SxactHasConflictOut, SxactHasSummaryConflictIn, SxactHasSummaryConflictOut, SxactIsCommitted, SxactIsDoomed, SxactIsPrepared, SxactIsReadOnly, and unconstify.
Referenced by FlagRWConflict().
◆ PageIsPredicateLocked()
| bool PageIsPredicateLocked | ( | Relation | relation, |
| BlockNumber | blkno | ||
| ) |
Definition at line 2006 of file predicate.c.
2007{
2011 PREDICATELOCKTARGET *target;
2012
2015 relation->rd_id,
2016 blkno);
2017
2021 target = (PREDICATELOCKTARGET *)
2026
2028}
References RelFileLocator::dbOid, fb(), HASH_FIND, hash_search_with_hash_value(), LW_SHARED, LWLockAcquire(), LWLockRelease(), PredicateLockHashPartitionLock, PredicateLockTargetHash, PredicateLockTargetTagHashCode, RelationData::rd_id, RelationData::rd_locator, and SET_PREDICATELOCKTARGETTAG_PAGE.
◆ PostPrepare_PredicateLocks()
| void PostPrepare_PredicateLocks | ( | FullTransactionId | fxid | ) |
Definition at line 4857 of file predicate.c.
4858{
4860 return;
4861
4863
4866
4869
4872}
References Assert, fb(), hash_destroy(), INVALID_PROC_NUMBER, InvalidSerializableXact, LocalPredicateLockHash, MySerializableXact, MyXactDidWrite, SERIALIZABLEXACT::pgprocno, SERIALIZABLEXACT::pid, and SxactIsPrepared.
Referenced by PrepareTransaction().
◆ PreCommit_CheckForSerializationFailure()
Definition at line 4701 of file predicate.c.
4702{
4704
4706 return;
4707
4709
4711
4712 /*
4713 * Check if someone else has already decided that we need to die. Since
4714 * we set our own DOOMED flag when partially releasing, ignore in that
4715 * case.
4716 */
4719 {
4724 errdetail_internal("Reason code: Canceled on identification as a pivot, during commit attempt."),
4726 }
4727
4729 {
4732
4735 {
4737
4739 {
4742
4747 {
4748 /*
4749 * Normally, we kill the pivot transaction to make sure we
4750 * make progress if the failing transaction is retried.
4751 * However, we can't kill it if it's already prepared, so
4752 * in that case we commit suicide instead.
4753 */
4755 {
4760 errdetail_internal("Reason code: Canceled on commit attempt with conflict in from prepared pivot."),
4762 }
4764 break;
4765 }
4766 }
4767 }
4768 }
4769
4772
4774}
References Assert, dlist_container, dlist_foreach, ereport, errcode(), ERRCODE_T_R_SERIALIZATION_FAILURE, errdetail_internal(), errhint(), errmsg(), ERROR, fb(), SERIALIZABLEXACT::flags, SERIALIZABLEXACT::inConflicts, InvalidSerializableXact, IsolationIsSerializable, PredXactListData::LastSxactCommitSeqNo, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MySerializableXact, PredXact, SERIALIZABLEXACT::prepareSeqNo, SXACT_FLAG_DOOMED, SXACT_FLAG_PREPARED, SxactIsCommitted, SxactIsDoomed, SxactIsPartiallyReleased, SxactIsPrepared, and SxactIsReadOnly.
Referenced by CommitTransaction(), and PrepareTransaction().
◆ predicatelock_hash()
Definition at line 1419 of file predicate.c.
1420{
1423
1425
1426 /* Look into the associated target object, and compute its hash code */
1428
1430}
References Assert, fb(), PredicateLockHashCodeFromTargetHashCode, and PredicateLockTargetTagHashCode.
Referenced by PredicateLockShmemInit().
◆ predicatelock_twophase_recover()
| void predicatelock_twophase_recover | ( | FullTransactionId | fxid, |
| uint16 | info, | ||
| void * | recdata, | ||
| uint32 | len | ||
| ) |
Definition at line 4907 of file predicate.c.
4909{
4910 TwoPhasePredicateRecord *record;
4912
4914
4916
4919
4921 {
4922 /* Per-transaction record. Set up a SERIALIZABLEXACT. */
4923 TwoPhasePredicateXactRecord *xactRecord;
4927 bool found;
4928
4930
4937
4938 /* vxid for a prepared xact is INVALID_PROC_NUMBER/xid; no pid */
4941 sxact->pid = 0;
4943
4944 /* a prepared xact hasn't committed yet */
4948
4950
4951 /*
4952 * Don't need to track this; no transactions running at the time the
4953 * recovered xact started are still active, except possibly other
4954 * prepared xacts and we don't care whether those are RO_SAFE or not.
4955 */
4957
4960
4961 sxact->topXid = xid;
4962 sxact->xmin = xactRecord->xmin;
4963 sxact->flags = xactRecord->flags;
4966 {
4970 }
4971
4972 /*
4973 * We don't know whether the transaction had any conflicts or not, so
4974 * we'll conservatively assume that it had both a conflict in and a
4975 * conflict out, and represent that with the summary conflict flags.
4976 */
4981
4982 /* Register the transaction's xid */
4983 sxidtag.xid = xid;
4985 &sxidtag,
4986 HASH_ENTER, &found);
4988 Assert(!found);
4990
4991 /*
4992 * Update global xmin. Note that this is a special case compared to
4993 * registering a normal transaction, because the global xmin might go
4994 * backwards. That's OK, because until recovery is over we're not
4995 * going to complete any transactions or create any non-prepared
4996 * transactions, so there's no danger of throwing away.
4997 */
5000 {
5004 }
5006 {
5009 }
5010
5012 }
5014 {
5015 /* Lock record. Recreate the PREDICATELOCK */
5016 TwoPhasePredicateLockRecord *lockRecord;
5021
5024
5026 sxidtag.xid = xid;
5030
5034
5036 }
5037}
References Assert, CreatePredicateLock(), CreatePredXact(), TwoPhasePredicateRecord::data, dlist_init(), dlist_node_init(), ereport, errcode(), errmsg(), ERROR, fb(), TwoPhasePredicateXactRecord::flags, HASH_ENTER, HASH_FIND, hash_search(), INVALID_PROC_NUMBER, InvalidSerCommitSeqNo, InvalidSerializableXact, InvalidTransactionId, len, TwoPhasePredicateRecord::lockRecord, LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockRelease(), max_prepared_xacts, MaxBackends, PredicateLockTargetTagHashCode, PredXact, RecoverySerCommitSeqNo, SerializableXidHash, SerialSetActiveSerXmin(), SXACT_FLAG_SUMMARY_CONFLICT_IN, SXACT_FLAG_SUMMARY_CONFLICT_OUT, PredXactListData::SxactGlobalXmin, PredXactListData::SxactGlobalXminCount, SxactIsPrepared, SxactIsReadOnly, TwoPhasePredicateLockRecord::target, TransactionIdEquals, TransactionIdFollows(), TransactionIdIsValid, TWOPHASEPREDICATERECORD_LOCK, TWOPHASEPREDICATERECORD_XACT, TwoPhasePredicateRecord::type, PredXactListData::WritableSxactCount, TwoPhasePredicateRecord::xactRecord, XidFromFullTransactionId, and TwoPhasePredicateXactRecord::xmin.
◆ PredicateLockAcquire()
|
static |
Definition at line 2515 of file predicate.c.
2516{
2518 bool found;
2520
2521 /* Do we have the lock already, or a covering lock? */
2523 return;
2524
2526 return;
2527
2528 /* the same hash and LW lock apply to the lock target and the local lock. */
2530
2531 /* Acquire lock in local table */
2535 HASH_ENTER, &found);
2537 if (!found)
2538 locallock->childLocks = 0;
2539
2540 /* Actually create the lock */
2542
2543 /*
2544 * Lock has been acquired. Check whether it should be promoted to a
2545 * coarser granularity, or whether there are finer-granularity locks to
2546 * clean up.
2547 */
2549 {
2550 /*
2551 * Lock request was promoted to a coarser-granularity lock, and that
2552 * lock was acquired. It will delete this lock and any of its
2553 * children, so we're done.
2554 */
2555 }
2556 else
2557 {
2558 /* Clean up any finer-granularity locks */
2561 }
2562}
References CheckAndPromotePredicateLockRequest(), CoarserLockCovers(), CreatePredicateLock(), DeleteChildTargetLocks(), fb(), GET_PREDICATELOCKTARGETTAG_TYPE, HASH_ENTER, hash_search_with_hash_value(), LocalPredicateLockHash, MySerializableXact, PredicateLockExists(), PredicateLockTargetTagHashCode, and PREDLOCKTAG_TUPLE.
Referenced by CheckAndPromotePredicateLockRequest(), PredicateLockPage(), PredicateLockRelation(), and PredicateLockTID().
◆ PredicateLockExists()
|
static |
Definition at line 2043 of file predicate.c.
2044{
2045 LOCALPREDICATELOCK *lock;
2046
2047 /* check local hash table */
2049 targettag,
2051
2052 if (!lock)
2053 return false;
2054
2055 /*
2056 * Found entry in the table, but still need to check whether it's actually
2057 * held -- it could just be a parent of some held lock.
2058 */
2060}
References fb(), HASH_FIND, hash_search(), LOCALPREDICATELOCK::held, and LocalPredicateLockHash.
Referenced by CoarserLockCovers(), PredicateLockAcquire(), and PredicateLockTID().
◆ PredicateLockingNeededForRelation()
Definition at line 498 of file predicate.c.
References FirstUnpinnedObjectId, RelationData::rd_id, and RelationUsesLocalBuffers.
Referenced by DropAllPredicateLocksFromTable(), PredicateLockPageSplit(), SerializationNeededForRead(), and SerializationNeededForWrite().
◆ PredicateLockPage()
| void PredicateLockPage | ( | Relation | relation, |
| BlockNumber | blkno, | ||
| Snapshot | snapshot | ||
| ) |
Definition at line 2597 of file predicate.c.
2598{
2599 PREDICATELOCKTARGETTAG tag;
2600
2602 return;
2603
2604 SET_PREDICATELOCKTARGETTAG_PAGE(tag,
2606 relation->rd_id,
2607 blkno);
2608 PredicateLockAcquire(&tag);
2609}
References RelFileLocator::dbOid, PredicateLockAcquire(), RelationData::rd_id, RelationData::rd_locator, SerializationNeededForRead(), and SET_PREDICATELOCKTARGETTAG_PAGE.
Referenced by _bt_readpage(), _hash_first(), _hash_readnext(), collectMatchBitmap(), gistScanPage(), IndexOnlyNext(), moveRightIfItNeeded(), scanPendingInsert(), and startScanEntry().
◆ PredicateLockPageCombine()
| void PredicateLockPageCombine | ( | Relation | relation, |
| BlockNumber | oldblkno, | ||
| BlockNumber | newblkno | ||
| ) |
Definition at line 3227 of file predicate.c.
3229{
3230 /*
3231 * Page combines differ from page splits in that we ought to be able to
3232 * remove the locks on the old page after transferring them to the new
3233 * page, instead of duplicating them. However, because we can't edit other
3234 * backends' local lock tables, removing the old lock would leave them
3235 * with an entry in their LocalPredicateLockHash for a lock they're not
3236 * holding, which isn't acceptable. So we wind up having to do the same
3237 * work as a page split, acquiring a lock on the new page and keeping the
3238 * old page locked too. That can lead to some false positives, but should
3239 * be rare in practice.
3240 */
3242}
References fb(), and PredicateLockPageSplit().
Referenced by _bt_mark_page_halfdead(), and ginDeletePage().
◆ PredicateLockPageSplit()
| void PredicateLockPageSplit | ( | Relation | relation, |
| BlockNumber | oldblkno, | ||
| BlockNumber | newblkno | ||
| ) |
Definition at line 3142 of file predicate.c.
3144{
3148
3149 /*
3150 * Bail out quickly if there are no serializable transactions running.
3151 *
3152 * It's safe to do this check without taking any additional locks. Even if
3153 * a serializable transaction starts concurrently, we know it can't take
3154 * any SIREAD locks on the page being split because the caller is holding
3155 * the associated buffer page lock. Memory reordering isn't an issue; the
3156 * memory barrier in the LWLock acquisition guarantees that this read
3157 * occurs while the buffer page lock is held.
3158 */
3160 return;
3161
3163 return;
3164
3168
3171 relation->rd_id,
3172 oldblkno);
3175 relation->rd_id,
3176 newblkno);
3177
3179
3180 /*
3181 * Try copying the locks over to the new page's tag, creating it if
3182 * necessary.
3183 */
3185 newtargettag,
3186 false);
3187
3189 {
3190 /*
3191 * No more predicate lock entries are available. Failure isn't an
3192 * option here, so promote the page lock to a relation lock.
3193 */
3194
3195 /* Get the parent relation lock's lock tag */
3197 &newtargettag);
3199
3200 /*
3201 * Move the locks to the parent. This shouldn't fail.
3202 *
3203 * Note that here we are removing locks held by other backends,
3204 * leading to a possible inconsistency in their local lock hash table.
3205 * This is OK because we're replacing it with a lock that covers the
3206 * old one.
3207 */
3209 newtargettag,
3210 true);
3212 }
3213
3215}
References Assert, BlockNumberIsValid(), RelFileLocator::dbOid, fb(), GetParentPredicateLockTag(), LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), PredicateLockingNeededForRelation(), PredXact, RelationData::rd_id, RelationData::rd_locator, SET_PREDICATELOCKTARGETTAG_PAGE, success, PredXactListData::SxactGlobalXmin, TransactionIdIsValid, and TransferPredicateLocksToNewTarget().
Referenced by _bt_insertonpg(), _hash_splitbucket(), createPostingTree(), ginPlaceToPage(), gistplacetopage(), and PredicateLockPageCombine().
◆ PredicateLockRelation()
Definition at line 2574 of file predicate.c.
2575{
2576 PREDICATELOCKTARGETTAG tag;
2577
2579 return;
2580
2581 SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2583 relation->rd_id);
2584 PredicateLockAcquire(&tag);
2585}
References RelFileLocator::dbOid, PredicateLockAcquire(), RelationData::rd_id, RelationData::rd_locator, SerializationNeededForRead(), and SET_PREDICATELOCKTARGETTAG_RELATION.
Referenced by _bt_endpoint(), _bt_first(), heap_beginscan(), and index_beginscan_internal().
◆ PredicateLockShmemInit()
Definition at line 1145 of file predicate.c.
1146{
1147 HASHCTL info;
1150 bool found;
1151
1152#ifndef EXEC_BACKEND
1154#endif
1155
1156 /*
1157 * Compute size of predicate lock target hashtable. Note these
1158 * calculations must agree with PredicateLockShmemSize!
1159 */
1161
1162 /*
1163 * Allocate hash table for PREDICATELOCKTARGET structs. This stores
1164 * per-predicate-lock-target information.
1165 */
1169
1171 max_table_size,
1172 max_table_size,
1173 &info,
1176
1177 /*
1178 * Reserve a dummy entry in the hash table; we use it to make sure there's
1179 * always one entry available when we need to split or combine a page,
1180 * because running out of space there could mean aborting a
1181 * non-serializable transaction.
1182 */
1184 {
1186 HASH_ENTER, &found);
1187 Assert(!found);
1188 }
1189
1190 /* Pre-calculate the hash and partition lock of the scratch entry */
1193
1194 /*
1195 * Allocate hash table for PREDICATELOCK structs. This stores per
1196 * xact-lock-of-a-target information.
1197 */
1202
1203 /* Assume an average of 2 xacts per target */
1204 max_table_size *= 2;
1205
1207 max_table_size,
1208 max_table_size,
1209 &info,
1212
1213 /*
1214 * Compute size for serializable transaction hashtable. Note these
1215 * calculations must agree with PredicateLockShmemSize!
1216 */
1218
1219 /*
1220 * Allocate a list to hold information on transactions participating in
1221 * predicate locking.
1222 *
1223 * Assume an average of 10 predicate locking transactions per backend.
1224 * This allows aggressive cleanup while detail is present before data must
1225 * be summarized for storage in SLRU and the "dummy" transaction.
1226 */
1227 max_table_size *= 10;
1228
1232
1234 requestSize,
1235 &found);
1237 if (!found)
1238 {
1240
1241 /* clean everything, both the header and the element */
1243
1254 /* Add all elements to available list, clean. */
1256 {
1260 }
1277 }
1278 /* This never changes, so let's keep a local copy. */
1280
1281 /*
1282 * Allocate hash table for SERIALIZABLEXID structs. This stores per-xid
1283 * information for serializable transactions which have accessed data.
1284 */
1287
1289 max_table_size,
1290 max_table_size,
1291 &info,
1293 HASH_FIXED_SIZE);
1294
1295 /*
1296 * Allocate space for tracking rw-conflicts in lists attached to the
1297 * transactions.
1298 *
1299 * Assume an average of 5 conflicts per transaction. Calculations suggest
1300 * that this will prevent resource exhaustion in even the most pessimal
1301 * loads up to max_connections = 200 with all 200 connections pounding the
1302 * database with serializable transactions. Beyond that, there may be
1303 * occasional transactions canceled when trying to flag conflicts. That's
1304 * probably OK.
1305 */
1306 max_table_size *= 5;
1307
1310 RWConflictDataSize);
1311
1313 requestSize,
1314 &found);
1316 if (!found)
1317 {
1319
1320 /* clean everything, including the elements */
1322
1325 RWConflictPoolHeaderDataSize);
1326 /* Add all elements to available list, clean. */
1328 {
1331 }
1332 }
1333
1334 /*
1335 * Create or attach to the header for the list of finished serializable
1336 * transactions.
1337 */
1341 &found);
1343 if (!found)
1345
1346 /*
1347 * Initialize the SLRU storage for old committed serializable
1348 * transactions.
1349 */
1350 SerialInit();
1351}
References PredXactListData::activeList, add_size(), Assert, PredXactListData::availableList, RWConflictPoolHeaderData::availableList, PredXactListData::CanPartialClearThrough, SERIALIZABLEXACT::commitSeqNo, CreatePredXact(), dlist_init(), dlist_node_init(), dlist_push_tail(), PredXactListData::element, RWConflictPoolHeaderData::element, HASHCTL::entrysize, fb(), SERIALIZABLEXACT::finishedBefore, SERIALIZABLEXACT::finishedLink, FinishedSerializableTransactions, FirstNormalSerCommitSeqNo, SERIALIZABLEXACT::flags, HASHCTL::hash, HASH_BLOBS, HASH_ELEM, HASH_ENTER, HASH_FIXED_SIZE, HASH_FUNCTION, HASH_PARTITION, hash_search(), PredXactListData::HavePartialClearedThrough, i, SERIALIZABLEXACT::inConflicts, INVALID_PROC_NUMBER, InvalidTransactionId, IsUnderPostmaster, HASHCTL::keysize, SERIALIZABLEXACT::lastCommitBeforeSnapshot, PredXactListData::LastSxactCommitSeqNo, LWLockInitialize(), max_prepared_xacts, MaxBackends, mul_size(), NPREDICATELOCKTARGETENTS, HASHCTL::num_partitions, NUM_PREDICATELOCK_PARTITIONS, OldCommittedSxact, PredXactListData::OldCommittedSxact, SERIALIZABLEXACT::outConflicts, RWConflictData::outLink, SERIALIZABLEXACT::perXactPredicateListLock, SERIALIZABLEXACT::pgprocno, SERIALIZABLEXACT::pid, SERIALIZABLEXACT::possibleUnsafeConflicts, predicatelock_hash(), PredicateLockHash, PredicateLockHashPartitionLock, SERIALIZABLEXACT::predicateLocks, PredicateLockTargetHash, PredicateLockTargetTagHashCode, PredXact, PredXactListDataSize, SERIALIZABLEXACT::prepareSeqNo, RWConflictDataSize, RWConflictPool, RWConflictPoolHeaderDataSize, ScratchPartitionLock, ScratchTargetTag, ScratchTargetTagHash, SERIALIZABLEXACT::SeqNo, SerialInit(), SerializableXidHash, SetInvalidVirtualTransactionId, ShmemInitHash(), ShmemInitStruct(), SXACT_FLAG_COMMITTED, PredXactListData::SxactGlobalXmin, PredXactListData::SxactGlobalXminCount, SERIALIZABLEXACT::topXid, SERIALIZABLEXACT::vxid, PredXactListData::WritableSxactCount, SERIALIZABLEXACT::xactLink, and SERIALIZABLEXACT::xmin.
Referenced by CreateOrAttachShmemStructs().
◆ PredicateLockShmemSize()
Definition at line 1357 of file predicate.c.
1358{
1359 Size size = 0;
1361
1362 /* predicate lock target hash table */
1366
1367 /* predicate lock hash table */
1368 max_table_size *= 2;
1371
1372 /*
1373 * Since NPREDICATELOCKTARGETENTS is only an estimate, add 10% safety
1374 * margin.
1375 */
1376 size = add_size(size, size / 10);
1377
1378 /* transaction list */
1380 max_table_size *= 10;
1384
1385 /* transaction xid table */
1388
1389 /* rw-conflict pool */
1390 max_table_size *= 5;
1393 RWConflictDataSize));
1394
1395 /* Head for list of finished serializable transactions. */
1397
1398 /* Shared memory structures for SLRU tracking of old committed xids. */
1401
1402 return size;
1403}
References add_size(), fb(), hash_estimate_size(), max_prepared_xacts, MaxBackends, mul_size(), NPREDICATELOCKTARGETENTS, PredXactListDataSize, RWConflictDataSize, RWConflictPoolHeaderDataSize, serializable_buffers, and SimpleLruShmemSize().
Referenced by CalculateShmemSize().
◆ PredicateLockTID()
| void PredicateLockTID | ( | Relation | relation, |
| const ItemPointerData * | tid, | ||
| Snapshot | snapshot, | ||
| TransactionId | tuple_xid | ||
| ) |
Definition at line 2619 of file predicate.c.
2621{
2622 PREDICATELOCKTARGETTAG tag;
2623
2625 return;
2626
2627 /*
2628 * Return if this xact wrote it.
2629 */
2631 {
2632 /* If we wrote it; we already have a write lock. */
2634 return;
2635 }
2636
2637 /*
2638 * Do quick-but-not-definitive test for a relation lock first. This will
2639 * never cause a return when the relation is *not* locked, but will
2640 * occasionally let the check continue when there really *is* a relation
2641 * level lock.
2642 */
2643 SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2645 relation->rd_id);
2647 return;
2648
2649 SET_PREDICATELOCKTARGETTAG_TUPLE(tag,
2651 relation->rd_id,
2652 ItemPointerGetBlockNumber(tid),
2653 ItemPointerGetOffsetNumber(tid));
2654 PredicateLockAcquire(&tag);
2655}
References RelFileLocator::dbOid, fb(), ItemPointerGetBlockNumber(), ItemPointerGetOffsetNumber(), PredicateLockAcquire(), PredicateLockExists(), RelationData::rd_id, RelationData::rd_index, RelationData::rd_locator, SerializationNeededForRead(), SET_PREDICATELOCKTARGETTAG_RELATION, SET_PREDICATELOCKTARGETTAG_TUPLE, and TransactionIdIsCurrentTransactionId().
Referenced by BitmapHeapScanNextBlock(), heap_fetch(), and heap_hot_search_buffer().
◆ PredicateLockTwoPhaseFinish()
| void PredicateLockTwoPhaseFinish | ( | FullTransactionId | fxid, |
| bool | isCommit | ||
| ) |
Definition at line 4880 of file predicate.c.
4881{
4884
4886
4891
4892 /* xid will not be found if it wasn't a serializable transaction */
4894 return;
4895
4896 /* Release its locks */
4899 * something */
4901}
References fb(), HASH_FIND, hash_search(), LW_SHARED, LWLockAcquire(), LWLockRelease(), MySerializableXact, MyXactDidWrite, ReleasePredicateLocks(), SerializableXidHash, SERIALIZABLEXIDTAG::xid, and XidFromFullTransactionId.
Referenced by FinishPreparedTransaction().
◆ RegisterPredicateLockingXid()
| void RegisterPredicateLockingXid | ( | TransactionId | xid | ) |
Definition at line 1957 of file predicate.c.
1958{
1961 bool found;
1962
1963 /*
1964 * If we're not tracking predicate lock data for this transaction, we
1965 * should ignore the request and return quickly.
1966 */
1968 return;
1969
1970 /* We should have a valid XID and be at the top level. */
1972
1974
1975 /* This should only be done once per transaction. */
1977
1979
1980 sxidtag.xid = xid;
1982 &sxidtag,
1983 HASH_ENTER, &found);
1984 Assert(!found);
1985
1986 /* Initialize the structure. */
1989}
References Assert, fb(), HASH_ENTER, hash_search(), InvalidSerializableXact, InvalidTransactionId, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MySerializableXact, SerializableXidHash, SERIALIZABLEXACT::topXid, and TransactionIdIsValid.
Referenced by AssignTransactionId().
◆ ReleaseOneSerializableXact()
|
static |
Definition at line 3833 of file predicate.c.
3835{
3837 dlist_mutable_iter iter;
3838
3843
3844 /*
3845 * First release all the predicate locks held by this xact (or transfer
3846 * them to OldCommittedSxact if summarize is true)
3847 */
3852 {
3855 PREDICATELOCKTAG tag;
3856 PREDICATELOCKTARGET *target;
3860
3861 tag = predlock->tag;
3862 target = tag.myTarget;
3866
3868
3870
3872 PredicateLockHashCodeFromTargetHashCode(&tag,
3873 targettaghash),
3876 {
3877 bool found;
3878
3879 /* Fold into dummy transaction list. */
3882 PredicateLockHashCodeFromTargetHashCode(&tag,
3883 targettaghash),
3884 HASH_ENTER_NULL, &found);
3890 if (found)
3891 {
3896 }
3897 else
3898 {
3900 &predlock->targetLink);
3902 &predlock->xactLink);
3904 }
3905 }
3906 else
3908
3910 }
3911
3912 /*
3913 * Rather than retail removal, just re-init the head after we've run
3914 * through the list.
3915 */
3917
3921
3924
3925 /* Release all outConflicts (unless 'partial' is true) */
3926 if (!partial)
3927 {
3929 {
3932
3936 }
3937 }
3938
3939 /* Release all inConflicts. */
3941 {
3944
3948 }
3949
3950 /* Finally, get rid of the xid and the record of the transaction itself. */
3951 if (!partial)
3952 {
3956 }
3957
3959}
References Assert, dlist_mutable_iter::cur, dlist_container, dlist_delete(), dlist_foreach_modify, dlist_init(), dlist_push_tail(), ereport, errcode(), errhint(), errmsg(), ERROR, fb(), HASH_ENTER_NULL, HASH_REMOVE, hash_search(), hash_search_with_hash_value(), InvalidSerCommitSeqNo, InvalidTransactionId, IsInParallelMode(), LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockHeldByMe(), LWLockRelease(), PREDICATELOCKTAG::myTarget, PREDICATELOCKTAG::myXact, OldCommittedSxact, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, SERIALIZABLEXACT::predicateLocks, PREDICATELOCKTARGET::predicateLocks, PredicateLockTargetTagHashCode, ReleasePredXact(), ReleaseRWConflict(), RemoveTargetIfNoLongerUsed(), SerializableXidHash, SXACT_FLAG_SUMMARY_CONFLICT_IN, SXACT_FLAG_SUMMARY_CONFLICT_OUT, SxactIsCommitted, SxactIsOnFinishedList, SxactIsRolledBack, and PREDICATELOCKTARGET::tag.
Referenced by ClearOldPredicateLocks(), ReleasePredicateLocks(), and SummarizeOldestCommittedSxact().
◆ ReleasePredicateLocks()
Definition at line 3310 of file predicate.c.
3311{
3315 dlist_mutable_iter iter;
3316
3317 /*
3318 * We can't trust XactReadOnly here, because a transaction which started
3319 * as READ WRITE can show as READ ONLY later, e.g., within
3320 * subtransactions. We want to flag a transaction as READ ONLY if it
3321 * commits without writing so that de facto READ ONLY transactions get the
3322 * benefit of some RO optimizations, so we will use this local variable to
3323 * get some cleanup logic right which is based on whether the transaction
3324 * was declared READ ONLY at the top level.
3325 */
3327
3328 /* We can't be both committing and releasing early due to RO_SAFE. */
3330
3331 /* Are we at the end of a transaction, that is, a commit or abort? */
3333 {
3334 /*
3335 * Parallel workers mustn't release predicate locks at the end of
3336 * their transaction. The leader will do that at the end of its
3337 * transaction.
3338 */
3340 {
3341 ReleasePredicateLocksLocal();
3342 return;
3343 }
3344
3345 /*
3346 * By the time the leader in a parallel query reaches end of
3347 * transaction, it has waited for all workers to exit.
3348 */
3350
3351 /*
3352 * If the leader in a parallel query earlier stashed a partially
3353 * released SERIALIZABLEXACT for final clean-up at end of transaction
3354 * (because workers might still have been accessing it), then it's
3355 * time to restore it.
3356 */
3358 {
3363 }
3364 }
3365
3367 {
3369 return;
3370 }
3371
3373
3374 /*
3375 * If the transaction is committing, but it has been partially released
3376 * already, then treat this as a roll back. It was marked as rolled back.
3377 */
3380
3381 /*
3382 * If we're called in the middle of a transaction because we discovered
3383 * that the SXACT_FLAG_RO_SAFE flag was set, then we'll partially release
3384 * it (that is, release the predicate locks and conflicts, but not the
3385 * SERIALIZABLEXACT itself) if we're the first backend to have noticed.
3386 */
3388 {
3389 /*
3390 * The leader needs to stash a pointer to it, so that it can
3391 * completely release it at end-of-transaction.
3392 */
3395
3396 /*
3397 * The first backend to reach this condition will partially release
3398 * the SERIALIZABLEXACT. All others will just clear their
3399 * backend-local state so that they stop doing SSI checks for the rest
3400 * of the transaction.
3401 */
3403 {
3405 ReleasePredicateLocksLocal();
3406 return;
3407 }
3408 else
3409 {
3412 /* ... and proceed to perform the partial release below. */
3413 }
3414 }
3420
3421 /* may not be serializable during COMMIT/ROLLBACK PREPARED */
3423
3424 /* We'd better not already be on the cleanup list. */
3426
3428
3429 /*
3430 * We don't hold XidGenLock lock here, assuming that TransactionId is
3431 * atomic!
3432 *
3433 * If this value is changing, we don't care that much whether we get the
3434 * old or new value -- it is just used to determine how far
3435 * SxactGlobalXmin must advance before this transaction can be fully
3436 * cleaned up. The worst that could happen is we wait for one more
3437 * transaction to complete before freeing some RAM; correctness of visible
3438 * behavior is not affected.
3439 */
3441
3442 /*
3443 * If it's not a commit it's either a rollback or a read-only transaction
3444 * flagged SXACT_FLAG_RO_SAFE, and we can clear our locks immediately.
3445 */
3447 {
3450 /* Recognize implicit read-only transaction (commit without write). */
3453 }
3454 else
3455 {
3456 /*
3457 * The DOOMED flag indicates that we intend to roll back this
3458 * transaction and so it should not cause serialization failures for
3459 * other transactions that conflict with it. Note that this flag might
3460 * already be set, if another backend marked this transaction for
3461 * abort.
3462 *
3463 * The ROLLED_BACK flag further indicates that ReleasePredicateLocks
3464 * has been called, and so the SerializableXact is eligible for
3465 * cleanup. This means it should not be considered when calculating
3466 * SxactGlobalXmin.
3467 */
3470
3471 /*
3472 * If the transaction was previously prepared, but is now failing due
3473 * to a ROLLBACK PREPARED or (hopefully very rare) error after the
3474 * prepare, clear the prepared flag. This simplifies conflict
3475 * checking.
3476 */
3478 }
3479
3481 {
3484 {
3485 /*
3486 * Release predicate locks and rw-conflicts in for all committed
3487 * transactions. There are no longer any transactions which might
3488 * conflict with the locks and no chance for new transactions to
3489 * overlap. Similarly, existing conflicts in can't cause pivots,
3490 * and any conflicts in which could have completed a dangerous
3491 * structure would already have caused a rollback, so any
3492 * remaining ones must be benign.
3493 */
3495 }
3496 }
3497 else
3498 {
3499 /*
3500 * Read-only transactions: clear the list of transactions that might
3501 * make us unsafe. Note that we use 'inLink' for the iteration as
3502 * opposed to 'outLink' for the r/w xacts.
3503 */
3505 {
3508
3511
3513 }
3514 }
3515
3516 /* Check for conflict out to old committed transactions. */
3520 {
3521 /*
3522 * we don't know which old committed transaction we conflicted with,
3523 * so be conservative and use FirstNormalSerCommitSeqNo here
3524 */
3528 }
3529
3530 /*
3531 * Release all outConflicts to committed transactions. If we're rolling
3532 * back clear them all. Set SXACT_FLAG_CONFLICT_OUT if any point to
3533 * previously committed transactions.
3534 */
3536 {
3539
3543 {
3548 }
3549
3554 }
3555
3556 /*
3557 * Release all inConflicts from committed and read-only transactions. If
3558 * we're rolling back, clear them all.
3559 */
3561 {
3564
3569 }
3570
3572 {
3573 /*
3574 * Remove ourselves from the list of possible conflicts for concurrent
3575 * READ ONLY transactions, flagging them as unsafe if we have a
3576 * conflict out. If any are waiting DEFERRABLE transactions, wake them
3577 * up if they are known safe or known unsafe.
3578 */
3580 {
3583
3587
3588 /* Mark conflicted if necessary. */
3590 && MyXactDidWrite
3593 <= roXact->SeqNo.lastCommitBeforeSnapshot))
3594 {
3595 /*
3596 * This releases possibleUnsafeConflict (as well as all other
3597 * possible conflicts for roXact)
3598 */
3600 }
3601 else
3602 {
3604
3605 /*
3606 * If we were the last possible conflict, flag it safe. The
3607 * transaction can now safely release its predicate locks (but
3608 * that transaction's backend has to do that itself).
3609 */
3612 }
3613
3614 /*
3615 * Wake up the process for a waiting DEFERRABLE transaction if we
3616 * now know it's either safe or conflicted.
3617 */
3621 }
3622 }
3623
3624 /*
3625 * Check whether it's time to clean up old transactions. This can only be
3626 * done when the last serializable transaction with the oldest xmin among
3627 * serializable transactions completes. We then find the "new oldest"
3628 * xmin and purge any transactions which finished before this transaction
3629 * was launched.
3630 *
3631 * For parallel queries in read-only transactions, it might run twice. We
3632 * only release the reference on the first call.
3633 */
3639 {
3642 {
3643 SetNewSxactGlobalXmin();
3645 }
3646 }
3647
3649
3651
3652 /* Add this to the list of transactions to check for later cleanup. */
3656
3657 /*
3658 * If we're releasing a RO_SAFE transaction in parallel mode, we'll only
3659 * partially release it. That's necessary because other backends may have
3660 * a reference to it. The leader will release the SERIALIZABLEXACT itself
3661 * at the end of the transaction after workers have stopped running.
3662 */
3666 false);
3667
3669
3671 ClearOldPredicateLocks();
3672
3673 ReleasePredicateLocksLocal();
3674}
References Assert, PredXactListData::CanPartialClearThrough, ClearOldPredicateLocks(), SERIALIZABLEXACT::commitSeqNo, dlist_mutable_iter::cur, dlist_container, dlist_foreach_modify, dlist_is_empty(), dlist_push_tail(), SERIALIZABLEXACT::earliestOutConflictCommit, fb(), SERIALIZABLEXACT::finishedBefore, SERIALIZABLEXACT::finishedLink, FinishedSerializableTransactions, FirstNormalSerCommitSeqNo, SERIALIZABLEXACT::flags, FlagSxactUnsafe(), SERIALIZABLEXACT::inConflicts, InvalidSerializableXact, IsInParallelMode(), IsolationIsSerializable, IsParallelWorker, PredXactListData::LastSxactCommitSeqNo, LocalPredicateLockHash, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), MySerializableXact, MyXactDidWrite, TransamVariablesData::nextXid, SERIALIZABLEXACT::outConflicts, ParallelContextActive(), SERIALIZABLEXACT::pid, SERIALIZABLEXACT::possibleUnsafeConflicts, PredXact, ProcSendSignal(), ReleaseOneSerializableXact(), ReleasePredicateLocksLocal(), ReleaseRWConflict(), SavedSerializableXact, SERIALIZABLEXACT::SeqNo, SetNewSxactGlobalXmin(), SXACT_FLAG_COMMITTED, SXACT_FLAG_CONFLICT_OUT, SXACT_FLAG_DOOMED, SXACT_FLAG_PARTIALLY_RELEASED, SXACT_FLAG_READ_ONLY, SXACT_FLAG_RO_SAFE, SXACT_FLAG_ROLLED_BACK, PredXactListData::SxactGlobalXmin, PredXactListData::SxactGlobalXminCount, SxactHasConflictOut, SxactHasSummaryConflictOut, SxactIsCommitted, SxactIsDeferrableWaiting, SxactIsDoomed, SxactIsOnFinishedList, SxactIsPartiallyReleased, SxactIsPrepared, SxactIsReadOnly, SxactIsRolledBack, SxactIsROSafe, SxactIsROUnsafe, TransactionIdEquals, TransamVariables, PredXactListData::WritableSxactCount, XidFromFullTransactionId, and SERIALIZABLEXACT::xmin.
Referenced by GetSafeSnapshot(), PredicateLockTwoPhaseFinish(), ResourceOwnerReleaseInternal(), and SerializationNeededForRead().
◆ ReleasePredicateLocksLocal()
Definition at line 3677 of file predicate.c.
3678{
3681
3682 /* Delete per-transaction lock table */
3684 {
3687 }
3688}
References fb(), hash_destroy(), InvalidSerializableXact, LocalPredicateLockHash, MySerializableXact, and MyXactDidWrite.
Referenced by ReleasePredicateLocks().
◆ ReleasePredXact()
|
static |
Definition at line 596 of file predicate.c.
597{
599
602}
References Assert, PredXactListData::availableList, dlist_delete(), dlist_push_tail(), fb(), PredXact, and ShmemAddrIsValid().
Referenced by GetSerializableTransactionSnapshotInt(), and ReleaseOneSerializableXact().
◆ ReleaseRWConflict()
|
static |
Definition at line 691 of file predicate.c.
692{
696}
References RWConflictPoolHeaderData::availableList, dlist_delete(), dlist_push_tail(), fb(), and RWConflictPool.
Referenced by FlagSxactUnsafe(), ReleaseOneSerializableXact(), and ReleasePredicateLocks().
◆ RemoveScratchTarget()
Definition at line 2138 of file predicate.c.
2139{
2140 bool found;
2141
2143
2147 &ScratchTargetTag,
2148 ScratchTargetTagHash,
2149 HASH_REMOVE, &found);
2150 Assert(found);
2153}
References Assert, fb(), HASH_REMOVE, hash_search_with_hash_value(), LW_EXCLUSIVE, LWLockAcquire(), LWLockHeldByMe(), LWLockRelease(), PredicateLockTargetHash, ScratchPartitionLock, ScratchTargetTag, and ScratchTargetTagHash.
Referenced by DropAllPredicateLocksFromTable(), and TransferPredicateLocksToNewTarget().
◆ RemoveTargetIfNoLongerUsed()
|
static |
Definition at line 2181 of file predicate.c.
2182{
2184
2186
2187 /* Can't remove it until no locks at this target. */
2189 return;
2190
2191 /* Actually remove the target. */
2193 &target->tag,
2194 targettaghash,
2197}
References Assert, dlist_is_empty(), fb(), HASH_REMOVE, hash_search_with_hash_value(), LWLockHeldByMe(), PG_USED_FOR_ASSERTS_ONLY, PREDICATELOCKTARGET::predicateLocks, PredicateLockTargetHash, and PREDICATELOCKTARGET::tag.
Referenced by CheckTargetForConflictsIn(), ClearOldPredicateLocks(), DeleteChildTargetLocks(), DeleteLockTarget(), ReleaseOneSerializableXact(), and TransferPredicateLocksToNewTarget().
◆ RestoreScratchTarget()
Definition at line 2159 of file predicate.c.
2160{
2161 bool found;
2162
2164
2168 &ScratchTargetTag,
2169 ScratchTargetTagHash,
2170 HASH_ENTER, &found);
2171 Assert(!found);
2174}
References Assert, fb(), HASH_ENTER, hash_search_with_hash_value(), LW_EXCLUSIVE, LWLockAcquire(), LWLockHeldByMe(), LWLockRelease(), PredicateLockTargetHash, ScratchPartitionLock, ScratchTargetTag, and ScratchTargetTagHash.
Referenced by DropAllPredicateLocksFromTable(), and TransferPredicateLocksToNewTarget().
◆ RWConflictExists()
|
static |
Definition at line 610 of file predicate.c.
611{
612 dlist_iter iter;
613
615
616 /* Check the ends of the purported conflict first. */
621 return false;
622
623 /*
624 * A conflict is possible; walk the list to find out.
625 *
626 * The unconstify is needed as we have no const version of
627 * dlist_foreach().
628 */
630 {
633
635 return true;
636 }
637
638 /* No conflict found. */
639 return false;
640}
References Assert, dlist_iter::cur, dlist_container, dlist_foreach, dlist_is_empty(), fb(), SERIALIZABLEXACT::outConflicts, SxactIsDoomed, and unconstify.
Referenced by CheckForSerializableConflictOut(), CheckTableForSerializableConflictIn(), CheckTargetForConflictsIn(), and SetRWConflict().
◆ SerialAdd()
|
static |
Definition at line 858 of file predicate.c.
859{
860 TransactionId tailXid;
865 LWLock *lock;
866
868
871
872 /*
873 * In this routine, we must hold both SerialControlLock and the SLRU bank
874 * lock simultaneously while making the SLRU data catch up with the new
875 * state that we determine.
876 */
878
879 /*
880 * If 'xid' is older than the global xmin (== tailXid), there's no need to
881 * store it, after all. This can happen if the oldest transaction holding
882 * back the global xmin just finished, making 'xid' uninteresting, but
883 * ClearOldPredicateLocks() has not yet run.
884 */
887 {
889 return;
890 }
891
892 /*
893 * If the SLRU is currently unused, zero out the whole active region from
894 * tailXid to headXid before taking it into use. Otherwise zero out only
895 * any new pages that enter the tailXid-headXid range as we advance
896 * headXid.
897 */
899 {
902 }
903 else
904 {
907 targetPage);
908 }
909
915
917 {
918 /* Initialize intervening pages; might involve trading locks */
919 for (;;)
920 {
925 break;
927 LWLockRelease(lock);
928 }
929 }
930 else
931 {
934 }
935
938
939 LWLockRelease(lock);
941}
References Assert, fb(), SerialControlData::headPage, SerialControlData::headXid, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), serialControl, SerialNextPage, SerialPage, SerialPagePrecedesLogically(), SerialSlruCtl, SerialValue, SimpleLruGetBankLock(), SimpleLruReadPage(), SimpleLruZeroPage(), SerialControlData::tailXid, TransactionIdFollows(), TransactionIdIsValid, and TransactionIdPrecedes().
Referenced by SummarizeOldestCommittedSxact().
◆ SerialGetMinConflictCommitSeqNo()
|
static |
Definition at line 949 of file predicate.c.
950{
951 TransactionId headXid;
952 TransactionId tailXid;
955
957
962
964 return 0;
965
967
969 || TransactionIdFollows(xid, headXid))
970 return 0;
971
972 /*
973 * The following function must be called without holding SLRU bank lock,
974 * but will return with that lock held, which must then be released.
975 */
977 SerialPage(xid), xid);
981}
References Assert, fb(), SerialControlData::headXid, LW_SHARED, LWLockAcquire(), LWLockRelease(), serialControl, SerialPage, SerialSlruCtl, SerialValue, SimpleLruGetBankLock(), SimpleLruReadPage_ReadOnly(), SerialControlData::tailXid, TransactionIdFollows(), TransactionIdIsValid, TransactionIdPrecedes(), and val.
Referenced by CheckForSerializableConflictOut().
◆ SerialInit()
Definition at line 806 of file predicate.c.
807{
808 bool found;
809
810 /*
811 * Set up SLRU management of the pg_serial data.
812 */
818#ifdef USE_ASSERT_CHECKING
820#endif
822
823 /*
824 * Create or attach to the SerialControl structure.
825 */
828
830 if (!found)
831 {
832 /*
833 * Set control information to reflect empty SLRU.
834 */
840 }
841}
References Assert, fb(), SerialControlData::headPage, SerialControlData::headXid, InvalidTransactionId, IsUnderPostmaster, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), SERIAL_ENTRIESPERPAGE, serialControl, serializable_buffers, SerialPagePrecedesLogically(), SerialSlruCtl, ShmemInitStruct(), SimpleLruInit(), SlruPagePrecedesUnitTests, SYNC_HANDLER_NONE, and SerialControlData::tailXid.
Referenced by PredicateLockShmemInit().
◆ SerializationNeededForRead()
Definition at line 516 of file predicate.c.
517{
518 /* Nothing to do if this is not a serializable transaction */
520 return false;
521
522 /*
523 * Don't acquire locks or conflict when scanning with a special snapshot.
524 * This excludes things like CLUSTER and REINDEX. They use the wholesale
525 * functions TransferPredicateLocksToHeapRelation() and
526 * CheckTableForSerializableConflictIn() to participate in serialization,
527 * but the scans involved don't need serialization.
528 */
530 return false;
531
532 /*
533 * Check if we have just become "RO-safe". If we have, immediately release
534 * all locks as they're not needed anymore. This also resets
535 * MySerializableXact, so that subsequent calls to this function can exit
536 * quickly.
537 *
538 * A transaction is flagged as RO_SAFE if all concurrent R/W transactions
539 * commit without having conflicts out to an earlier snapshot, thus
540 * ensuring that no conflicts are possible for this transaction.
541 */
543 {
545 return false;
546 }
547
548 /* Check if the relation doesn't participate in predicate locking */
550 return false;
551
552 return true; /* no excuse to skip predicate locking */
553}
References InvalidSerializableXact, IsMVCCSnapshot, MySerializableXact, PredicateLockingNeededForRelation(), ReleasePredicateLocks(), and SxactIsROSafe.
Referenced by CheckForSerializableConflictOut(), CheckForSerializableConflictOutNeeded(), PredicateLockPage(), PredicateLockRelation(), and PredicateLockTID().
◆ SerializationNeededForWrite()
Definition at line 560 of file predicate.c.
561{
562 /* Nothing to do if this is not a serializable transaction */
564 return false;
565
566 /* Check if the relation doesn't participate in predicate locking */
568 return false;
569
570 return true; /* no excuse to skip predicate locking */
571}
References InvalidSerializableXact, MySerializableXact, and PredicateLockingNeededForRelation().
Referenced by CheckForSerializableConflictIn(), and CheckTableForSerializableConflictIn().
◆ SerialPagePrecedesLogically()
Definition at line 731 of file predicate.c.
732{
735
740
743}
References fb(), FirstNormalTransactionId, SERIAL_ENTRIESPERPAGE, and TransactionIdPrecedes().
Referenced by CheckPointPredicate(), SerialAdd(), and SerialInit().
◆ SerialSetActiveSerXmin()
|
static |
Definition at line 990 of file predicate.c.
991{
993
994 /*
995 * When no sxacts are active, nothing overlaps, set the xid values to
996 * invalid to show that there are no valid entries. Don't clear headPage,
997 * though. A new xmin might still land on that page, and we don't want to
998 * repeatedly zero out the same page.
999 */
1001 {
1005 return;
1006 }
1007
1008 /*
1009 * When we're recovering prepared transactions, the global xmin might move
1010 * backwards depending on the order they're recovered. Normally that's not
1011 * OK, but during recovery no serializable transactions will commit, so
1012 * the SLRU is empty and we can get away with it.
1013 */
1015 {
1019 {
1021 }
1023 return;
1024 }
1025
1028
1030
1032}
References Assert, fb(), SerialControlData::headPage, SerialControlData::headXid, InvalidTransactionId, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), RecoveryInProgress(), serialControl, SerialControlData::tailXid, TransactionIdFollows(), TransactionIdIsValid, and TransactionIdPrecedes().
Referenced by GetSerializableTransactionSnapshotInt(), predicatelock_twophase_recover(), and SetNewSxactGlobalXmin().
◆ SetNewSxactGlobalXmin()
Definition at line 3249 of file predicate.c.
3250{
3251 dlist_iter iter;
3252
3254
3257
3259 {
3262
3266 {
3271 {
3274 }
3278 }
3279 }
3280
3282}
References PredXactListData::activeList, Assert, dlist_iter::cur, dlist_container, dlist_foreach, fb(), InvalidTransactionId, LWLockHeldByMe(), OldCommittedSxact, PredXact, SerialSetActiveSerXmin(), PredXactListData::SxactGlobalXmin, PredXactListData::SxactGlobalXminCount, SxactIsCommitted, SxactIsRolledBack, TransactionIdEquals, TransactionIdIsValid, and TransactionIdPrecedes().
Referenced by ReleasePredicateLocks().
◆ SetPossibleUnsafeConflict()
|
static |
Definition at line 666 of file predicate.c.
668{
670
674
679 errhint("You might need to run fewer transactions at a time or increase \"max_connections\".")));
680
683
688}
References Assert, RWConflictPoolHeaderData::availableList, dlist_delete(), dlist_head_element, dlist_is_empty(), dlist_push_tail(), ereport, errcode(), errhint(), errmsg(), ERROR, fb(), RWConflictPool, and SxactIsReadOnly.
Referenced by GetSerializableTransactionSnapshotInt().
◆ SetRWConflict()
|
static |
Definition at line 643 of file predicate.c.
644{
646
649
654 errhint("You might need to run fewer transactions at a time or increase \"max_connections\".")));
655
658
659 conflict->sxactOut = reader;
663}
References Assert, RWConflictPoolHeaderData::availableList, dlist_delete(), dlist_head_element, dlist_is_empty(), dlist_push_tail(), ereport, errcode(), errhint(), errmsg(), ERROR, fb(), SERIALIZABLEXACT::outConflicts, RWConflictExists(), and RWConflictPool.
Referenced by FlagRWConflict().
◆ SetSerializableTransactionSnapshot()
| void SetSerializableTransactionSnapshot | ( | Snapshot | snapshot, |
| VirtualTransactionId * | sourcevxid, | ||
| int | sourcepid | ||
| ) |
Definition at line 1720 of file predicate.c.
1723{
1725
1726 /*
1727 * If this is called by parallel.c in a parallel worker, we don't want to
1728 * create a SERIALIZABLEXACT just yet because the leader's
1729 * SERIALIZABLEXACT will be installed with AttachSerializableXact(). We
1730 * also don't want to reject SERIALIZABLE READ ONLY DEFERRABLE in this
1731 * case, because the leader has already determined that the snapshot it
1732 * has passed us is safe. So there is nothing for us to do.
1733 */
1735 return;
1736
1737 /*
1738 * We do not allow SERIALIZABLE READ ONLY DEFERRABLE transactions to
1739 * import snapshots, since there's no way to wait for a safe snapshot when
1740 * we're using the snap we're told to. (XXX instead of throwing an error,
1741 * we could just ignore the XactDeferrable flag?)
1742 */
1747
1749 sourcepid);
1750}
References Assert, ereport, errcode(), errmsg(), ERROR, fb(), GetSerializableTransactionSnapshotInt(), IsolationIsSerializable, IsParallelWorker, XactDeferrable, and XactReadOnly.
Referenced by SetTransactionSnapshot().
◆ ShareSerializableXact()
| SerializableXactHandle ShareSerializableXact | ( | void | ) |
Definition at line 5045 of file predicate.c.
References MySerializableXact.
Referenced by InitializeParallelDSM().
◆ SummarizeOldestCommittedSxact()
Definition at line 1501 of file predicate.c.
1502{
1504
1506
1507 /*
1508 * This function is only called if there are no sxact slots available.
1509 * Some of them must belong to old, already-finished transactions, so
1510 * there should be something in FinishedSerializableTransactions list that
1511 * we can summarize. However, there's a race condition: while we were not
1512 * holding any locks, a transaction might have ended and cleaned up all
1513 * the finished sxact entries already, freeing up their sxact slots. In
1514 * that case, we have nothing to do here. The caller will find one of the
1515 * slots released by the other backend when it retries.
1516 */
1518 {
1520 return;
1521 }
1522
1523 /*
1524 * Grab the first sxact off the finished list -- this will be the earliest
1525 * commit. Remove it from the list.
1526 */
1530
1531 /* Add to SLRU summary information. */
1535
1536 /* Summarize and release the detail. */
1538
1540}
References dlist_delete_thoroughly(), dlist_head_element, dlist_is_empty(), fb(), FinishedSerializableTransactions, InvalidSerCommitSeqNo, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), ReleaseOneSerializableXact(), SerialAdd(), SxactHasConflictOut, SxactIsReadOnly, and TransactionIdIsValid.
Referenced by GetSerializableTransactionSnapshotInt().
◆ TransferPredicateLocksToHeapRelation()
Definition at line 3121 of file predicate.c.
References DropAllPredicateLocksFromTable().
Referenced by ATRewriteTable(), cluster_rel(), index_concurrently_set_dead(), index_drop(), and reindex_index().
◆ TransferPredicateLocksToNewTarget()
|
static |
Definition at line 2728 of file predicate.c.
2731{
2737 bool found;
2739
2741 LW_EXCLUSIVE));
2742
2747
2749 {
2750 /*
2751 * Remove the dummy entry to give us scratch space, so we know we'll
2752 * be able to create the new lock target.
2753 */
2755 }
2756
2757 /*
2758 * We must get the partition locks in ascending sequence to avoid
2759 * deadlocks. If old and new partitions are the same, we must request the
2760 * lock only once.
2761 */
2763 {
2767 }
2769 {
2773 }
2774 else
2776
2777 /*
2778 * Look for the old target. If not found, that's OK; no predicate locks
2779 * are affected, so we can just clean up and return. If it does exist,
2780 * walk its list of predicate locks and move or copy them to the new
2781 * target.
2782 */
2784 &oldtargettag,
2785 oldtargettaghash,
2787
2789 {
2792 dlist_mutable_iter iter;
2793
2795 &newtargettag,
2796 newtargettaghash,
2797 HASH_ENTER_NULL, &found);
2798
2800 {
2801 /* Failed to allocate due to insufficient shmem */
2804 }
2805
2806 /* If we created a new entry, initialize it */
2807 if (!found)
2809
2811
2812 /*
2813 * Loop through all the locks on the old target, replacing them with
2814 * locks on the new target.
2815 */
2817
2819 {
2824
2826
2828 {
2831
2833 (PredicateLockHash,
2834 &oldpredlock->tag,
2836 oldtargettaghash),
2837 HASH_REMOVE, &found);
2838 Assert(found);
2839 }
2840
2843 &newpredlocktag,
2845 newtargettaghash),
2846 HASH_ENTER_NULL,
2847 &found);
2849 {
2850 /* Out of shared memory. Undo what we've done so far. */
2855 }
2856 if (!found)
2857 {
2859 &(newpredlock->targetLink));
2861 &(newpredlock->xactLink));
2863 }
2864 else
2865 {
2868 }
2869
2873 }
2875
2877 {
2880 }
2881 }
2882
2883
2884exit:
2885 /* Release partition locks in reverse order of acquisition. */
2887 {
2890 }
2892 {
2895 }
2896 else
2898
2900 {
2901 /* We shouldn't run out of memory if we're moving locks */
2903
2904 /* Put the scratch entry back */
2906 }
2907
2909}
References Assert, PREDICATELOCK::commitSeqNo, dlist_mutable_iter::cur, DeleteLockTarget(), dlist_container, dlist_delete(), dlist_foreach_modify, dlist_init(), dlist_is_empty(), dlist_push_tail(), fb(), HASH_ENTER_NULL, HASH_FIND, HASH_REMOVE, hash_search_with_hash_value(), InvalidSerCommitSeqNo, LW_EXCLUSIVE, LW_SHARED, LWLockAcquire(), LWLockHeldByMeInMode(), LWLockRelease(), OldCommittedSxact, PredicateLockHash, PredicateLockHashCodeFromTargetHashCode, PredicateLockHashPartitionLock, PredicateLockTargetHash, PredicateLockTargetTagHashCode, RemoveScratchTarget(), RemoveTargetIfNoLongerUsed(), and RestoreScratchTarget().
Referenced by PredicateLockPageSplit().
◆ XidIsConcurrent()
|
static |
Definition at line 3970 of file predicate.c.
3971{
3973
3976
3978
3980 return false;
3981
3983 return true;
3984
3986}
References Assert, fb(), GetTopTransactionIdIfAny(), GetTransactionSnapshot(), pg_lfind32(), TransactionIdEquals, TransactionIdFollowsOrEquals(), TransactionIdIsValid, and TransactionIdPrecedes().
Referenced by CheckForSerializableConflictOut().
Variable Documentation
◆ FinishedSerializableTransactions
|
static |
Definition at line 399 of file predicate.c.
Referenced by ClearOldPredicateLocks(), PredicateLockShmemInit(), ReleasePredicateLocks(), and SummarizeOldestCommittedSxact().
◆ LocalPredicateLockHash
Definition at line 414 of file predicate.c.
Referenced by CheckAndPromotePredicateLockRequest(), CheckTargetForConflictsIn(), CreateLocalPredicateLockHash(), DecrementParentLocks(), PostPrepare_PredicateLocks(), PredicateLockAcquire(), PredicateLockExists(), ReleasePredicateLocks(), and ReleasePredicateLocksLocal().
◆ max_predicate_locks_per_page
| int max_predicate_locks_per_page |
Definition at line 373 of file predicate.c.
Referenced by MaxPredicateChildLocks().
◆ max_predicate_locks_per_relation
| int max_predicate_locks_per_relation |
Definition at line 372 of file predicate.c.
Referenced by MaxPredicateChildLocks().
◆ max_predicate_locks_per_xact
| int max_predicate_locks_per_xact |
Definition at line 371 of file predicate.c.
Referenced by CreateLocalPredicateLockHash(), and MaxPredicateChildLocks().
◆ MySerializableXact
|
static |
Definition at line 421 of file predicate.c.
Referenced by AtPrepare_PredicateLocks(), AttachSerializableXact(), CheckForSerializableConflictIn(), CheckForSerializableConflictOut(), CheckForSerializableConflictOutNeeded(), CheckTableForSerializableConflictIn(), CheckTargetForConflictsIn(), DeleteChildTargetLocks(), GetSafeSnapshot(), GetSerializableTransactionSnapshotInt(), OnConflict_CheckForSerializationFailure(), PostPrepare_PredicateLocks(), PreCommit_CheckForSerializationFailure(), PredicateLockAcquire(), PredicateLockTwoPhaseFinish(), RegisterPredicateLockingXid(), ReleasePredicateLocks(), ReleasePredicateLocksLocal(), SerializationNeededForRead(), SerializationNeededForWrite(), and ShareSerializableXact().
◆ MyXactDidWrite
Definition at line 422 of file predicate.c.
Referenced by CheckForSerializableConflictIn(), CheckTableForSerializableConflictIn(), GetSerializableTransactionSnapshotInt(), PostPrepare_PredicateLocks(), PredicateLockTwoPhaseFinish(), ReleasePredicateLocks(), and ReleasePredicateLocksLocal().
◆ OldCommittedSxact
|
static |
Definition at line 362 of file predicate.c.
Referenced by ClearOldPredicateLocks(), DropAllPredicateLocksFromTable(), FlagRWConflict(), PredicateLockShmemInit(), ReleaseOneSerializableXact(), SetNewSxactGlobalXmin(), and TransferPredicateLocksToNewTarget().
◆ PredicateLockHash
|
static |
Definition at line 398 of file predicate.c.
Referenced by CheckTargetForConflictsIn(), ClearOldPredicateLocks(), CreatePredicateLock(), DeleteChildTargetLocks(), DeleteLockTarget(), DropAllPredicateLocksFromTable(), GetPredicateLockStatusData(), PredicateLockShmemInit(), ReleaseOneSerializableXact(), and TransferPredicateLocksToNewTarget().
◆ PredicateLockTargetHash
|
static |
Definition at line 397 of file predicate.c.
Referenced by CheckTableForSerializableConflictIn(), CheckTargetForConflictsIn(), CreatePredicateLock(), DropAllPredicateLocksFromTable(), PageIsPredicateLocked(), PredicateLockShmemInit(), RemoveScratchTarget(), RemoveTargetIfNoLongerUsed(), RestoreScratchTarget(), and TransferPredicateLocksToNewTarget().
◆ PredXact
|
static |
Definition at line 384 of file predicate.c.
Referenced by CheckTableForSerializableConflictIn(), ClearOldPredicateLocks(), CreatePredXact(), DropAllPredicateLocksFromTable(), GetSafeSnapshotBlockingPids(), GetSerializableTransactionSnapshotInt(), PreCommit_CheckForSerializationFailure(), predicatelock_twophase_recover(), PredicateLockPageSplit(), PredicateLockShmemInit(), ReleasePredicateLocks(), ReleasePredXact(), and SetNewSxactGlobalXmin().
◆ RWConflictPool
|
static |
Definition at line 390 of file predicate.c.
Referenced by PredicateLockShmemInit(), ReleaseRWConflict(), SetPossibleUnsafeConflict(), and SetRWConflict().
◆ SavedSerializableXact
|
static |
Definition at line 431 of file predicate.c.
Referenced by ReleasePredicateLocks().
◆ ScratchPartitionLock
|
static |
Definition at line 408 of file predicate.c.
Referenced by PredicateLockShmemInit(), RemoveScratchTarget(), and RestoreScratchTarget().
◆ ScratchTargetTag
|
static |
Definition at line 406 of file predicate.c.
406{0, 0, 0, 0};
Referenced by PredicateLockShmemInit(), RemoveScratchTarget(), and RestoreScratchTarget().
◆ ScratchTargetTagHash
|
static |
Definition at line 407 of file predicate.c.
Referenced by PredicateLockShmemInit(), RemoveScratchTarget(), and RestoreScratchTarget().
◆ serialControl
|
static |
Definition at line 354 of file predicate.c.
Referenced by CheckPointPredicate(), SerialAdd(), SerialGetMinConflictCommitSeqNo(), SerialInit(), and SerialSetActiveSerXmin().
◆ SerializableXidHash
|
static |
Definition at line 396 of file predicate.c.
Referenced by CheckForSerializableConflictOut(), predicatelock_twophase_recover(), PredicateLockShmemInit(), PredicateLockTwoPhaseFinish(), RegisterPredicateLockingXid(), and ReleaseOneSerializableXact().
◆ SerialSlruCtlData
|
static |
Definition at line 324 of file predicate.c.
