diff options
Diffstat (limited to 'src/backend')
95 files changed, 3106 insertions, 993 deletions
diff --git a/src/backend/access/common/toast_compression.c b/src/backend/access/common/toast_compression.c index 21f2f4af97e..926f1e4008a 100644 --- a/src/backend/access/common/toast_compression.c +++ b/src/backend/access/common/toast_compression.c @@ -25,11 +25,11 @@ /* GUC */ int default_toast_compression = TOAST_PGLZ_COMPRESSION; -#define NO_LZ4_SUPPORT() \ +#define NO_COMPRESSION_SUPPORT(method) \ ereport(ERROR, \ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), \ - errmsg("compression method lz4 not supported"), \ - errdetail("This functionality requires the server to be built with lz4 support."))) + errmsg("compression method %s not supported", method), \ + errdetail("This functionality requires the server to be built with %s support.", method))) /* * Compress a varlena using PGLZ. @@ -139,7 +139,7 @@ struct varlena * lz4_compress_datum(const struct varlena *value) { #ifndef USE_LZ4 - NO_LZ4_SUPPORT(); + NO_COMPRESSION_SUPPORT("lz4"); return NULL; /* keep compiler quiet */ #else int32 valsize; @@ -182,7 +182,7 @@ struct varlena * lz4_decompress_datum(const struct varlena *value) { #ifndef USE_LZ4 - NO_LZ4_SUPPORT(); + NO_COMPRESSION_SUPPORT("lz4"); return NULL; /* keep compiler quiet */ #else int32 rawsize; @@ -215,7 +215,7 @@ struct varlena * lz4_decompress_datum_slice(const struct varlena *value, int32 slicelength) { #ifndef USE_LZ4 - NO_LZ4_SUPPORT(); + NO_COMPRESSION_SUPPORT("lz4"); return NULL; /* keep compiler quiet */ #else int32 rawsize; @@ -289,7 +289,7 @@ CompressionNameToMethod(const char *compression) else if (strcmp(compression, "lz4") == 0) { #ifndef USE_LZ4 - NO_LZ4_SUPPORT(); + NO_COMPRESSION_SUPPORT("lz4"); #endif return TOAST_LZ4_COMPRESSION; } diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c index 745a04ef26e..8f918e00af7 100644 --- a/src/backend/access/heap/visibilitymap.c +++ b/src/backend/access/heap/visibilitymap.c @@ -364,7 +364,7 @@ visibilitymap_get_status(Relation rel, BlockNumber heapBlk, Buffer *vmbuf) { *vmbuf = vm_readbuf(rel, mapBlock, false); if (!BufferIsValid(*vmbuf)) - return false; + return (uint8) 0; } map = PageGetContents(BufferGetPage(*vmbuf)); diff --git a/src/backend/access/index/amapi.c b/src/backend/access/index/amapi.c index f0f4f974bce..60684c53422 100644 --- a/src/backend/access/index/amapi.c +++ b/src/backend/access/index/amapi.c @@ -42,6 +42,19 @@ GetIndexAmRoutine(Oid amhandler) elog(ERROR, "index access method handler function %u did not return an IndexAmRoutine struct", amhandler); + /* Assert that all required callbacks are present. */ + Assert(routine->ambuild != NULL); + Assert(routine->ambuildempty != NULL); + Assert(routine->aminsert != NULL); + Assert(routine->ambulkdelete != NULL); + Assert(routine->amvacuumcleanup != NULL); + Assert(routine->amcostestimate != NULL); + Assert(routine->amoptions != NULL); + Assert(routine->amvalidate != NULL); + Assert(routine->ambeginscan != NULL); + Assert(routine->amrescan != NULL); + Assert(routine->amendscan != NULL); + return routine; } diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c index 4af1ff1e9e5..d69798795b4 100644 --- a/src/backend/access/nbtree/nbtsearch.c +++ b/src/backend/access/nbtree/nbtsearch.c @@ -892,9 +892,9 @@ _bt_first(IndexScanDesc scan, ScanDirection dir) OffsetNumber offnum; BTScanInsertData inskey; ScanKey startKeys[INDEX_MAX_KEYS]; - ScanKeyData notnullkeys[INDEX_MAX_KEYS]; + ScanKeyData notnullkey; int keysz = 0; - StrategyNumber strat_total; + StrategyNumber strat_total = InvalidStrategy; BlockNumber blkno = InvalidBlockNumber, lastcurrblkno; @@ -1034,7 +1034,6 @@ _bt_first(IndexScanDesc scan, ScanDirection dir) * need to be kept in sync. *---------- */ - strat_total = BTEqualStrategyNumber; if (so->numberOfKeys > 0) { AttrNumber curattr; @@ -1122,16 +1121,15 @@ _bt_first(IndexScanDesc scan, ScanDirection dir) ScanDirectionIsForward(dir) : ScanDirectionIsBackward(dir))) { - /* Yes, so build the key in notnullkeys[keysz] */ - bkey = ¬nullkeys[keysz]; + /* Final startKeys[] entry will be deduced NOT NULL key */ + bkey = ¬nullkey; ScanKeyEntryInitialize(bkey, (SK_SEARCHNOTNULL | SK_ISNULL | (impliesNN->sk_flags & (SK_BT_DESC | SK_BT_NULLS_FIRST))), curattr, - ((impliesNN->sk_flags & SK_BT_NULLS_FIRST) ? - BTGreaterStrategyNumber : - BTLessStrategyNumber), + ScanDirectionIsForward(dir) ? + BTGreaterStrategyNumber : BTLessStrategyNumber, InvalidOid, InvalidOid, InvalidOid, diff --git a/src/backend/access/transam/twophase.c b/src/backend/access/transam/twophase.c index 85cbe397cb2..7918176fc58 100644 --- a/src/backend/access/transam/twophase.c +++ b/src/backend/access/transam/twophase.c @@ -1183,7 +1183,11 @@ EndPrepare(GlobalTransaction gxact) * starting immediately after the WAL record is inserted could complete * without fsync'ing our state file. (This is essentially the same kind * of race condition as the COMMIT-to-clog-write case that - * RecordTransactionCommit uses DELAY_CHKPT_START for; see notes there.) + * RecordTransactionCommit uses DELAY_CHKPT_IN_COMMIT for; see notes + * there.) Note that DELAY_CHKPT_IN_COMMIT is used to find transactions in + * the critical commit section. We need to know about such transactions + * for conflict detection in logical replication. See + * GetOldestActiveTransactionId(true, false) and its use. * * We save the PREPARE record's location in the gxact for later use by * CheckPointTwoPhase. @@ -2298,7 +2302,7 @@ ProcessTwoPhaseBuffer(FullTransactionId fxid, * RecordTransactionCommitPrepared * * This is basically the same as RecordTransactionCommit (q.v. if you change - * this function): in particular, we must set DELAY_CHKPT_START to avoid a + * this function): in particular, we must set DELAY_CHKPT_IN_COMMIT to avoid a * race condition. * * We know the transaction made at least one XLOG entry (its PREPARE), @@ -2318,7 +2322,7 @@ RecordTransactionCommitPrepared(TransactionId xid, const char *gid) { XLogRecPtr recptr; - TimestampTz committs = GetCurrentTimestamp(); + TimestampTz committs; bool replorigin; /* @@ -2331,8 +2335,24 @@ RecordTransactionCommitPrepared(TransactionId xid, START_CRIT_SECTION(); /* See notes in RecordTransactionCommit */ - Assert((MyProc->delayChkptFlags & DELAY_CHKPT_START) == 0); - MyProc->delayChkptFlags |= DELAY_CHKPT_START; + Assert((MyProc->delayChkptFlags & DELAY_CHKPT_IN_COMMIT) == 0); + MyProc->delayChkptFlags |= DELAY_CHKPT_IN_COMMIT; + + /* + * Ensures the DELAY_CHKPT_IN_COMMIT flag write is globally visible before + * commit time is written. + */ + pg_write_barrier(); + + /* + * Note it is important to set committs value after marking ourselves as + * in the commit critical section (DELAY_CHKPT_IN_COMMIT). This is because + * we want to ensure all transactions that have acquired commit timestamp + * are finished before we allow the logical replication client to advance + * its xid which is used to hold back dead rows for conflict detection. + * See comments atop worker.c. + */ + committs = GetCurrentTimestamp(); /* * Emit the XLOG commit record. Note that we mark 2PC commits as @@ -2381,7 +2401,7 @@ RecordTransactionCommitPrepared(TransactionId xid, TransactionIdCommitTree(xid, nchildren, children); /* Checkpoint can proceed now */ - MyProc->delayChkptFlags &= ~DELAY_CHKPT_START; + MyProc->delayChkptFlags &= ~DELAY_CHKPT_IN_COMMIT; END_CRIT_SECTION(); diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 41601fcb280..b46e7e9c2a6 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -1431,10 +1431,22 @@ RecordTransactionCommit(void) * without holding the ProcArrayLock, since we're the only one * modifying it. This makes checkpoint's determination of which xacts * are delaying the checkpoint a bit fuzzy, but it doesn't matter. + * + * Note, it is important to get the commit timestamp after marking the + * transaction in the commit critical section. See + * RecordTransactionCommitPrepared. */ - Assert((MyProc->delayChkptFlags & DELAY_CHKPT_START) == 0); + Assert((MyProc->delayChkptFlags & DELAY_CHKPT_IN_COMMIT) == 0); START_CRIT_SECTION(); - MyProc->delayChkptFlags |= DELAY_CHKPT_START; + MyProc->delayChkptFlags |= DELAY_CHKPT_IN_COMMIT; + + Assert(xactStopTimestamp == 0); + + /* + * Ensures the DELAY_CHKPT_IN_COMMIT flag write is globally visible + * before commit time is written. + */ + pg_write_barrier(); /* * Insert the commit XLOG record. @@ -1537,7 +1549,7 @@ RecordTransactionCommit(void) */ if (markXidCommitted) { - MyProc->delayChkptFlags &= ~DELAY_CHKPT_START; + MyProc->delayChkptFlags &= ~DELAY_CHKPT_IN_COMMIT; END_CRIT_SECTION(); } diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index 8e7827c6ed9..5553c20fee8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -96,6 +96,7 @@ #include "utils/guc_hooks.h" #include "utils/guc_tables.h" #include "utils/injection_point.h" +#include "utils/pgstat_internal.h" #include "utils/ps_status.h" #include "utils/relmapper.h" #include "utils/snapmgr.h" @@ -702,7 +703,7 @@ static void InitControlFile(uint64 sysidentifier, uint32 data_checksum_version); static void WriteControlFile(void); static void ReadControlFile(void); static void UpdateControlFile(void); -static char *str_time(pg_time_t tnow); +static char *str_time(pg_time_t tnow, char *buf, size_t bufsize); static int get_sync_bit(int method); @@ -1091,6 +1092,9 @@ XLogInsertRecord(XLogRecData *rdata, pgWalUsage.wal_bytes += rechdr->xl_tot_len; pgWalUsage.wal_records++; pgWalUsage.wal_fpi += num_fpi; + + /* Required for the flush of pending stats WAL data */ + pgstat_report_fixed = true; } return EndPos; @@ -2108,6 +2112,12 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, TimeLineID tli, bool opportunistic) LWLockRelease(WALWriteLock); pgWalUsage.wal_buffers_full++; TRACE_POSTGRESQL_WAL_BUFFER_WRITE_DIRTY_DONE(); + + /* + * Required for the flush of pending stats WAL data, per + * update of pgWalUsage. + */ + pgstat_report_fixed = true; } } } @@ -5361,11 +5371,9 @@ BootStrapXLOG(uint32 data_checksum_version) } static char * -str_time(pg_time_t tnow) +str_time(pg_time_t tnow, char *buf, size_t bufsize) { - char *buf = palloc(128); - - pg_strftime(buf, 128, + pg_strftime(buf, bufsize, "%Y-%m-%d %H:%M:%S %Z", pg_localtime(&tnow, log_timezone)); @@ -5608,6 +5616,7 @@ StartupXLOG(void) XLogRecPtr missingContrecPtr; TransactionId oldestActiveXID; bool promoted = false; + char timebuf[128]; /* * We should have an aux process resource owner to use, and we should not @@ -5636,25 +5645,29 @@ StartupXLOG(void) */ ereport(IsPostmasterEnvironment ? LOG : NOTICE, (errmsg("database system was shut down at %s", - str_time(ControlFile->time)))); + str_time(ControlFile->time, + timebuf, sizeof(timebuf))))); break; case DB_SHUTDOWNED_IN_RECOVERY: ereport(LOG, (errmsg("database system was shut down in recovery at %s", - str_time(ControlFile->time)))); + str_time(ControlFile->time, + timebuf, sizeof(timebuf))))); break; case DB_SHUTDOWNING: ereport(LOG, (errmsg("database system shutdown was interrupted; last known up at %s", - str_time(ControlFile->time)))); + str_time(ControlFile->time, + timebuf, sizeof(timebuf))))); break; case DB_IN_CRASH_RECOVERY: ereport(LOG, (errmsg("database system was interrupted while in recovery at %s", - str_time(ControlFile->time)), + str_time(ControlFile->time, + timebuf, sizeof(timebuf))), errhint("This probably means that some data is corrupted and" " you will have to use the last backup for recovery."))); break; @@ -5662,7 +5675,8 @@ StartupXLOG(void) case DB_IN_ARCHIVE_RECOVERY: ereport(LOG, (errmsg("database system was interrupted while in recovery at log time %s", - str_time(ControlFile->checkPointCopy.time)), + str_time(ControlFile->checkPointCopy.time, + timebuf, sizeof(timebuf))), errhint("If this has occurred more than once some data might be corrupted" " and you might need to choose an earlier recovery target."))); break; @@ -5670,7 +5684,8 @@ StartupXLOG(void) case DB_IN_PRODUCTION: ereport(LOG, (errmsg("database system was interrupted; last known up at %s", - str_time(ControlFile->time)))); + str_time(ControlFile->time, + timebuf, sizeof(timebuf))))); break; default: @@ -6315,6 +6330,12 @@ StartupXLOG(void) */ CompleteCommitTsInitialization(); + /* Clean up EndOfWalRecoveryInfo data to appease Valgrind leak checking */ + if (endOfRecoveryInfo->lastPage) + pfree(endOfRecoveryInfo->lastPage); + pfree(endOfRecoveryInfo->recoveryStopReason); + pfree(endOfRecoveryInfo); + /* * All done with end-of-recovery actions. * @@ -7121,7 +7142,7 @@ CreateCheckPoint(int flags) * starting snapshot of locks and transactions. */ if (!shutdown && XLogStandbyInfoActive()) - checkPoint.oldestActiveXid = GetOldestActiveTransactionId(); + checkPoint.oldestActiveXid = GetOldestActiveTransactionId(false, true); else checkPoint.oldestActiveXid = InvalidTransactionId; diff --git a/src/backend/access/transam/xlogreader.c b/src/backend/access/transam/xlogreader.c index ac1f801b1eb..dcc8d4f9c1b 100644 --- a/src/backend/access/transam/xlogreader.c +++ b/src/backend/access/transam/xlogreader.c @@ -723,11 +723,12 @@ restart: /* Calculate pointer to beginning of next page */ targetPagePtr += XLOG_BLCKSZ; - /* Wait for the next page to become available */ - readOff = ReadPageInternal(state, targetPagePtr, - Min(total_len - gotlen + SizeOfXLogShortPHD, - XLOG_BLCKSZ)); - + /* + * Read the page header before processing the record data, so we + * can handle the case where the previous record ended as being a + * partial one. + */ + readOff = ReadPageInternal(state, targetPagePtr, SizeOfXLogShortPHD); if (readOff == XLREAD_WOULDBLOCK) return XLREAD_WOULDBLOCK; else if (readOff < 0) @@ -776,6 +777,15 @@ restart: goto err; } + /* Wait for the next page to become available */ + readOff = ReadPageInternal(state, targetPagePtr, + Min(total_len - gotlen + SizeOfXLogShortPHD, + XLOG_BLCKSZ)); + if (readOff == XLREAD_WOULDBLOCK) + return XLREAD_WOULDBLOCK; + else if (readOff < 0) + goto err; + /* Append the continuation from this page to the buffer */ pageHeaderSize = XLogPageHeaderSize(pageHeader); diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index 23878b2dd91..f23ec8969c2 100644 --- a/src/backend/access/transam/xlogrecovery.c +++ b/src/backend/access/transam/xlogrecovery.c @@ -1626,6 +1626,7 @@ ShutdownWalRecovery(void) close(readFile); readFile = -1; } + pfree(xlogreader->private_data); XLogReaderFree(xlogreader); XLogPrefetcherFree(xlogprefetcher); @@ -4760,7 +4761,7 @@ bool check_primary_slot_name(char **newval, void **extra, GucSource source) { if (*newval && strcmp(*newval, "") != 0 && - !ReplicationSlotValidateName(*newval, WARNING)) + !ReplicationSlotValidateName(*newval, false, WARNING)) return false; return true; diff --git a/src/backend/catalog/pg_subscription.c b/src/backend/catalog/pg_subscription.c index 1395032413e..244acf52f36 100644 --- a/src/backend/catalog/pg_subscription.c +++ b/src/backend/catalog/pg_subscription.c @@ -103,6 +103,7 @@ GetSubscription(Oid subid, bool missing_ok) sub->passwordrequired = subform->subpasswordrequired; sub->runasowner = subform->subrunasowner; sub->failover = subform->subfailover; + sub->retaindeadtuples = subform->subretaindeadtuples; /* Get conninfo */ datum = SysCacheGetAttrNotNull(SUBSCRIPTIONOID, @@ -319,7 +320,7 @@ AddSubscriptionRelState(Oid subid, Oid relid, char state, */ void UpdateSubscriptionRelState(Oid subid, Oid relid, char state, - XLogRecPtr sublsn) + XLogRecPtr sublsn, bool already_locked) { Relation rel; HeapTuple tup; @@ -327,9 +328,24 @@ UpdateSubscriptionRelState(Oid subid, Oid relid, char state, Datum values[Natts_pg_subscription_rel]; bool replaces[Natts_pg_subscription_rel]; - LockSharedObject(SubscriptionRelationId, subid, 0, AccessShareLock); + if (already_locked) + { +#ifdef USE_ASSERT_CHECKING + LOCKTAG tag; - rel = table_open(SubscriptionRelRelationId, RowExclusiveLock); + Assert(CheckRelationOidLockedByMe(SubscriptionRelRelationId, + RowExclusiveLock, true)); + SET_LOCKTAG_OBJECT(tag, InvalidOid, SubscriptionRelationId, subid, 0); + Assert(LockHeldByMe(&tag, AccessShareLock, true)); +#endif + + rel = table_open(SubscriptionRelRelationId, NoLock); + } + else + { + LockSharedObject(SubscriptionRelationId, subid, 0, AccessShareLock); + rel = table_open(SubscriptionRelRelationId, RowExclusiveLock); + } /* Try finding existing mapping. */ tup = SearchSysCacheCopy2(SUBSCRIPTIONRELMAP, diff --git a/src/backend/catalog/system_views.sql b/src/backend/catalog/system_views.sql index b2d5332effc..f6eca09ee15 100644 --- a/src/backend/catalog/system_views.sql +++ b/src/backend/catalog/system_views.sql @@ -1386,7 +1386,8 @@ REVOKE ALL ON pg_subscription FROM public; GRANT SELECT (oid, subdbid, subskiplsn, subname, subowner, subenabled, subbinary, substream, subtwophasestate, subdisableonerr, subpasswordrequired, subrunasowner, subfailover, - subslotname, subsynccommit, subpublications, suborigin) + subretaindeadtuples, subslotname, subsynccommit, + subpublications, suborigin) ON pg_subscription TO public; CREATE VIEW pg_stat_subscription_stats AS diff --git a/src/backend/commands/explain.c b/src/backend/commands/explain.c index 7e2792ead71..8345bc0264b 100644 --- a/src/backend/commands/explain.c +++ b/src/backend/commands/explain.c @@ -3582,6 +3582,7 @@ static void show_memoize_info(MemoizeState *mstate, List *ancestors, ExplainState *es) { Plan *plan = ((PlanState *) mstate)->plan; + Memoize *mplan = (Memoize *) plan; ListCell *lc; List *context; StringInfoData keystr; @@ -3602,7 +3603,7 @@ show_memoize_info(MemoizeState *mstate, List *ancestors, ExplainState *es) plan, ancestors); - foreach(lc, ((Memoize *) plan)->param_exprs) + foreach(lc, mplan->param_exprs) { Node *expr = (Node *) lfirst(lc); @@ -3618,6 +3619,24 @@ show_memoize_info(MemoizeState *mstate, List *ancestors, ExplainState *es) pfree(keystr.data); + if (es->costs) + { + if (es->format == EXPLAIN_FORMAT_TEXT) + { + ExplainIndentText(es); + appendStringInfo(es->str, "Estimates: capacity=%u distinct keys=%.0f lookups=%.0f hit percent=%.2f%%\n", + mplan->est_entries, mplan->est_unique_keys, + mplan->est_calls, mplan->est_hit_ratio * 100.0); + } + else + { + ExplainPropertyUInteger("Estimated Capacity", NULL, mplan->est_entries, es); + ExplainPropertyFloat("Estimated Distinct Lookup Keys", NULL, mplan->est_unique_keys, 0, es); + ExplainPropertyFloat("Estimated Lookups", NULL, mplan->est_calls, 0, es); + ExplainPropertyFloat("Estimated Hit Percent", NULL, mplan->est_hit_ratio * 100.0, 2, es); + } + } + if (!es->analyze) return; diff --git a/src/backend/commands/foreigncmds.c b/src/backend/commands/foreigncmds.c index 8d2d7431544..77f8461f42e 100644 --- a/src/backend/commands/foreigncmds.c +++ b/src/backend/commands/foreigncmds.c @@ -1588,6 +1588,7 @@ ImportForeignSchema(ImportForeignSchemaStmt *stmt) pstmt->utilityStmt = (Node *) cstmt; pstmt->stmt_location = rs->stmt_location; pstmt->stmt_len = rs->stmt_len; + pstmt->planOrigin = PLAN_STMT_INTERNAL; /* Execute statement */ ProcessUtility(pstmt, cmd, false, diff --git a/src/backend/commands/schemacmds.c b/src/backend/commands/schemacmds.c index 546160f0941..0f03d9743d2 100644 --- a/src/backend/commands/schemacmds.c +++ b/src/backend/commands/schemacmds.c @@ -215,6 +215,7 @@ CreateSchemaCommand(CreateSchemaStmt *stmt, const char *queryString, wrapper->utilityStmt = stmt; wrapper->stmt_location = stmt_location; wrapper->stmt_len = stmt_len; + wrapper->planOrigin = PLAN_STMT_INTERNAL; /* do this step */ ProcessUtility(wrapper, diff --git a/src/backend/commands/subscriptioncmds.c b/src/backend/commands/subscriptioncmds.c index e23b0de7242..cd6c3684482 100644 --- a/src/backend/commands/subscriptioncmds.c +++ b/src/backend/commands/subscriptioncmds.c @@ -14,6 +14,7 @@ #include "postgres.h" +#include "access/commit_ts.h" #include "access/htup_details.h" #include "access/table.h" #include "access/twophase.h" @@ -71,8 +72,9 @@ #define SUBOPT_PASSWORD_REQUIRED 0x00000800 #define SUBOPT_RUN_AS_OWNER 0x00001000 #define SUBOPT_FAILOVER 0x00002000 -#define SUBOPT_LSN 0x00004000 -#define SUBOPT_ORIGIN 0x00008000 +#define SUBOPT_RETAIN_DEAD_TUPLES 0x00004000 +#define SUBOPT_LSN 0x00008000 +#define SUBOPT_ORIGIN 0x00010000 /* check if the 'val' has 'bits' set */ #define IsSet(val, bits) (((val) & (bits)) == (bits)) @@ -98,6 +100,7 @@ typedef struct SubOpts bool passwordrequired; bool runasowner; bool failover; + bool retaindeadtuples; char *origin; XLogRecPtr lsn; } SubOpts; @@ -105,8 +108,10 @@ typedef struct SubOpts static List *fetch_table_list(WalReceiverConn *wrconn, List *publications); static void check_publications_origin(WalReceiverConn *wrconn, List *publications, bool copydata, - char *origin, Oid *subrel_local_oids, - int subrel_count, char *subname); + bool retain_dead_tuples, char *origin, + Oid *subrel_local_oids, int subrel_count, + char *subname); +static void check_pub_dead_tuple_retention(WalReceiverConn *wrconn); static void check_duplicates_in_publist(List *publist, Datum *datums); static List *merge_publications(List *oldpublist, List *newpublist, bool addpub, const char *subname); static void ReportSlotConnectionError(List *rstates, Oid subid, char *slotname, char *err); @@ -162,6 +167,8 @@ parse_subscription_options(ParseState *pstate, List *stmt_options, opts->runasowner = false; if (IsSet(supported_opts, SUBOPT_FAILOVER)) opts->failover = false; + if (IsSet(supported_opts, SUBOPT_RETAIN_DEAD_TUPLES)) + opts->retaindeadtuples = false; if (IsSet(supported_opts, SUBOPT_ORIGIN)) opts->origin = pstrdup(LOGICALREP_ORIGIN_ANY); @@ -210,7 +217,7 @@ parse_subscription_options(ParseState *pstate, List *stmt_options, if (strcmp(opts->slot_name, "none") == 0) opts->slot_name = NULL; else - ReplicationSlotValidateName(opts->slot_name, ERROR); + ReplicationSlotValidateName(opts->slot_name, false, ERROR); } else if (IsSet(supported_opts, SUBOPT_COPY_DATA) && strcmp(defel->defname, "copy_data") == 0) @@ -307,6 +314,15 @@ parse_subscription_options(ParseState *pstate, List *stmt_options, opts->specified_opts |= SUBOPT_FAILOVER; opts->failover = defGetBoolean(defel); } + else if (IsSet(supported_opts, SUBOPT_RETAIN_DEAD_TUPLES) && + strcmp(defel->defname, "retain_dead_tuples") == 0) + { + if (IsSet(opts->specified_opts, SUBOPT_RETAIN_DEAD_TUPLES)) + errorConflictingDefElem(defel, pstate); + + opts->specified_opts |= SUBOPT_RETAIN_DEAD_TUPLES; + opts->retaindeadtuples = defGetBoolean(defel); + } else if (IsSet(supported_opts, SUBOPT_ORIGIN) && strcmp(defel->defname, "origin") == 0) { @@ -563,7 +579,8 @@ CreateSubscription(ParseState *pstate, CreateSubscriptionStmt *stmt, SUBOPT_SYNCHRONOUS_COMMIT | SUBOPT_BINARY | SUBOPT_STREAMING | SUBOPT_TWOPHASE_COMMIT | SUBOPT_DISABLE_ON_ERR | SUBOPT_PASSWORD_REQUIRED | - SUBOPT_RUN_AS_OWNER | SUBOPT_FAILOVER | SUBOPT_ORIGIN); + SUBOPT_RUN_AS_OWNER | SUBOPT_FAILOVER | + SUBOPT_RETAIN_DEAD_TUPLES | SUBOPT_ORIGIN); parse_subscription_options(pstate, stmt->options, supported_opts, &opts); /* @@ -630,6 +647,10 @@ CreateSubscription(ParseState *pstate, CreateSubscriptionStmt *stmt, stmt->subname))); } + /* Ensure that we can enable retain_dead_tuples */ + if (opts.retaindeadtuples) + CheckSubDeadTupleRetention(true, !opts.enabled, WARNING); + if (!IsSet(opts.specified_opts, SUBOPT_SLOT_NAME) && opts.slot_name == NULL) opts.slot_name = stmt->subname; @@ -670,6 +691,8 @@ CreateSubscription(ParseState *pstate, CreateSubscriptionStmt *stmt, values[Anum_pg_subscription_subpasswordrequired - 1] = BoolGetDatum(opts.passwordrequired); values[Anum_pg_subscription_subrunasowner - 1] = BoolGetDatum(opts.runasowner); values[Anum_pg_subscription_subfailover - 1] = BoolGetDatum(opts.failover); + values[Anum_pg_subscription_subretaindeadtuples - 1] = + BoolGetDatum(opts.retaindeadtuples); values[Anum_pg_subscription_subconninfo - 1] = CStringGetTextDatum(conninfo); if (opts.slot_name) @@ -722,7 +745,11 @@ CreateSubscription(ParseState *pstate, CreateSubscriptionStmt *stmt, { check_publications(wrconn, publications); check_publications_origin(wrconn, publications, opts.copy_data, - opts.origin, NULL, 0, stmt->subname); + opts.retaindeadtuples, opts.origin, + NULL, 0, stmt->subname); + + if (opts.retaindeadtuples) + check_pub_dead_tuple_retention(wrconn); /* * Set sync state based on if we were asked to do data copy or @@ -881,8 +908,8 @@ AlterSubscription_refresh(Subscription *sub, bool copy_data, sizeof(Oid), oid_cmp); check_publications_origin(wrconn, sub->publications, copy_data, - sub->origin, subrel_local_oids, - subrel_count, sub->name); + sub->retaindeadtuples, sub->origin, + subrel_local_oids, subrel_count, sub->name); /* * Rels that we want to remove from subscription and drop any slots @@ -1040,18 +1067,22 @@ AlterSubscription_refresh(Subscription *sub, bool copy_data, } /* - * Common checks for altering failover and two_phase options. + * Common checks for altering failover, two_phase, and retain_dead_tuples + * options. */ static void CheckAlterSubOption(Subscription *sub, const char *option, bool slot_needs_update, bool isTopLevel) { + Assert(strcmp(option, "failover") == 0 || + strcmp(option, "two_phase") == 0 || + strcmp(option, "retain_dead_tuples") == 0); + /* - * The checks in this function are required only for failover and - * two_phase options. + * Altering the retain_dead_tuples option does not update the slot on the + * publisher. */ - Assert(strcmp(option, "failover") == 0 || - strcmp(option, "two_phase") == 0); + Assert(!slot_needs_update || strcmp(option, "retain_dead_tuples") != 0); /* * Do not allow changing the option if the subscription is enabled. This @@ -1063,6 +1094,39 @@ CheckAlterSubOption(Subscription *sub, const char *option, * the publisher by the existing walsender, so we could have allowed that * even when the subscription is enabled. But we kept this restriction for * the sake of consistency and simplicity. + * + * Additionally, do not allow changing the retain_dead_tuples option when + * the subscription is enabled to prevent race conditions arising from the + * new option value being acknowledged asynchronously by the launcher and + * apply workers. + * + * Without the restriction, a race condition may arise when a user + * disables and immediately re-enables the retain_dead_tuples option. In + * this case, the launcher might drop the slot upon noticing the disabled + * action, while the apply worker may keep maintaining + * oldest_nonremovable_xid without noticing the option change. During this + * period, a transaction ID wraparound could falsely make this ID appear + * as if it originates from the future w.r.t the transaction ID stored in + * the slot maintained by launcher. + * + * Similarly, if the user enables retain_dead_tuples concurrently with the + * launcher starting the worker, the apply worker may start calculating + * oldest_nonremovable_xid before the launcher notices the enable action. + * Consequently, the launcher may update slot.xmin to a newer value than + * that maintained by the worker. In subsequent cycles, upon integrating + * the worker's oldest_nonremovable_xid, the launcher might detect a + * retreat in the calculated xmin, necessitating additional handling. + * + * XXX To address the above race conditions, we can define + * oldest_nonremovable_xid as FullTransactionID and adds the check to + * disallow retreating the conflict slot's xmin. For now, we kept the + * implementation simple by disallowing change to the retain_dead_tuples, + * but in the future we can change this after some more analysis. + * + * Note that we could restrict only the enabling of retain_dead_tuples to + * avoid the race conditions described above, but we maintain the + * restriction for both enable and disable operations for the sake of + * consistency. */ if (sub->enabled) ereport(ERROR, @@ -1110,6 +1174,9 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, bool update_tuple = false; bool update_failover = false; bool update_two_phase = false; + bool check_pub_rdt = false; + bool retain_dead_tuples; + char *origin; Subscription *sub; Form_pg_subscription form; bits32 supported_opts; @@ -1137,6 +1204,9 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, sub = GetSubscription(subid, false); + retain_dead_tuples = sub->retaindeadtuples; + origin = sub->origin; + /* * Don't allow non-superuser modification of a subscription with * password_required=false. @@ -1165,7 +1235,7 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, SUBOPT_DISABLE_ON_ERR | SUBOPT_PASSWORD_REQUIRED | SUBOPT_RUN_AS_OWNER | SUBOPT_FAILOVER | - SUBOPT_ORIGIN); + SUBOPT_RETAIN_DEAD_TUPLES | SUBOPT_ORIGIN); parse_subscription_options(pstate, stmt->options, supported_opts, &opts); @@ -1325,11 +1395,62 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, replaces[Anum_pg_subscription_subfailover - 1] = true; } + if (IsSet(opts.specified_opts, SUBOPT_RETAIN_DEAD_TUPLES)) + { + values[Anum_pg_subscription_subretaindeadtuples - 1] = + BoolGetDatum(opts.retaindeadtuples); + replaces[Anum_pg_subscription_subretaindeadtuples - 1] = true; + + CheckAlterSubOption(sub, "retain_dead_tuples", false, isTopLevel); + + /* + * Workers may continue running even after the + * subscription has been disabled. + * + * To prevent race conditions (as described in + * CheckAlterSubOption()), ensure that all worker + * processes have already exited before proceeding. + */ + if (logicalrep_workers_find(subid, true, true)) + ereport(ERROR, + (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), + errmsg("cannot alter retain_dead_tuples when logical replication worker is still running"), + errhint("Try again after some time."))); + + /* + * Remind the user that enabling subscription will prevent + * the accumulation of dead tuples. + */ + if (opts.retaindeadtuples) + CheckSubDeadTupleRetention(true, !sub->enabled, NOTICE); + + /* + * Notify the launcher to manage the replication slot for + * conflict detection. This ensures that replication slot + * is efficiently handled (created, updated, or dropped) + * in response to any configuration changes. + */ + ApplyLauncherWakeupAtCommit(); + + check_pub_rdt = opts.retaindeadtuples; + retain_dead_tuples = opts.retaindeadtuples; + } + if (IsSet(opts.specified_opts, SUBOPT_ORIGIN)) { values[Anum_pg_subscription_suborigin - 1] = CStringGetTextDatum(opts.origin); replaces[Anum_pg_subscription_suborigin - 1] = true; + + /* + * Check if changes from different origins may be received + * from the publisher when the origin is changed to ANY + * and retain_dead_tuples is enabled. + */ + check_pub_rdt = retain_dead_tuples && + pg_strcasecmp(opts.origin, LOGICALREP_ORIGIN_ANY) == 0; + + origin = opts.origin; } update_tuple = true; @@ -1347,6 +1468,15 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), errmsg("cannot enable subscription that does not have a slot name"))); + /* + * Check track_commit_timestamp only when enabling the + * subscription in case it was disabled after creation. See + * comments atop CheckSubDeadTupleRetention() for details. + */ + if (sub->retaindeadtuples) + CheckSubDeadTupleRetention(opts.enabled, !opts.enabled, + WARNING); + values[Anum_pg_subscription_subenabled - 1] = BoolGetDatum(opts.enabled); replaces[Anum_pg_subscription_subenabled - 1] = true; @@ -1355,6 +1485,14 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, ApplyLauncherWakeupAtCommit(); update_tuple = true; + + /* + * The subscription might be initially created with + * connect=false and retain_dead_tuples=true, meaning the + * remote server's status may not be checked. Ensure this + * check is conducted now. + */ + check_pub_rdt = sub->retaindeadtuples && opts.enabled; break; } @@ -1369,6 +1507,13 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, CStringGetTextDatum(stmt->conninfo); replaces[Anum_pg_subscription_subconninfo - 1] = true; update_tuple = true; + + /* + * Since the remote server configuration might have changed, + * perform a check to ensure it permits enabling + * retain_dead_tuples. + */ + check_pub_rdt = sub->retaindeadtuples; break; case ALTER_SUBSCRIPTION_SET_PUBLICATION: @@ -1568,14 +1713,15 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, } /* - * Try to acquire the connection necessary for altering the slot, if - * needed. + * Try to acquire the connection necessary either for modifying the slot + * or for checking if the remote server permits enabling + * retain_dead_tuples. * * This has to be at the end because otherwise if there is an error while * doing the database operations we won't be able to rollback altered * slot. */ - if (update_failover || update_two_phase) + if (update_failover || update_two_phase || check_pub_rdt) { bool must_use_password; char *err; @@ -1584,10 +1730,14 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, /* Load the library providing us libpq calls. */ load_file("libpqwalreceiver", false); - /* Try to connect to the publisher. */ + /* + * Try to connect to the publisher, using the new connection string if + * available. + */ must_use_password = sub->passwordrequired && !sub->ownersuperuser; - wrconn = walrcv_connect(sub->conninfo, true, true, must_use_password, - sub->name, &err); + wrconn = walrcv_connect(stmt->conninfo ? stmt->conninfo : sub->conninfo, + true, true, must_use_password, sub->name, + &err); if (!wrconn) ereport(ERROR, (errcode(ERRCODE_CONNECTION_FAILURE), @@ -1596,9 +1746,17 @@ AlterSubscription(ParseState *pstate, AlterSubscriptionStmt *stmt, PG_TRY(); { - walrcv_alter_slot(wrconn, sub->slotname, - update_failover ? &opts.failover : NULL, - update_two_phase ? &opts.twophase : NULL); + if (retain_dead_tuples) + check_pub_dead_tuple_retention(wrconn); + + check_publications_origin(wrconn, sub->publications, false, + retain_dead_tuples, origin, NULL, 0, + sub->name); + + if (update_failover || update_two_phase) + walrcv_alter_slot(wrconn, sub->slotname, + update_failover ? &opts.failover : NULL, + update_two_phase ? &opts.twophase : NULL); } PG_FINALLY(); { @@ -2086,20 +2244,29 @@ AlterSubscriptionOwner_oid(Oid subid, Oid newOwnerId) * Check and log a warning if the publisher has subscribed to the same table, * its partition ancestors (if it's a partition), or its partition children (if * it's a partitioned table), from some other publishers. This check is - * required only if "copy_data = true" and "origin = none" for CREATE - * SUBSCRIPTION and ALTER SUBSCRIPTION ... REFRESH statements to notify the - * user that data having origin might have been copied. + * required in the following scenarios: * - * This check need not be performed on the tables that are already added - * because incremental sync for those tables will happen through WAL and the - * origin of the data can be identified from the WAL records. + * 1) For CREATE SUBSCRIPTION and ALTER SUBSCRIPTION ... REFRESH statements + * with "copy_data = true" and "origin = none": + * - Warn the user that data with an origin might have been copied. + * - This check is skipped for tables already added, as incremental sync via + * WAL allows origin tracking. The list of such tables is in + * subrel_local_oids. * - * subrel_local_oids contains the list of relation oids that are already - * present on the subscriber. + * 2) For CREATE SUBSCRIPTION and ALTER SUBSCRIPTION ... REFRESH statements + * with "retain_dead_tuples = true" and "origin = any", and for ALTER + * SUBSCRIPTION statements that modify retain_dead_tuples or origin, or + * when the publisher's status changes (e.g., due to a connection string + * update): + * - Warn the user that only conflict detection info for local changes on + * the publisher is retained. Data from other origins may lack sufficient + * details for reliable conflict detection. + * - See comments atop worker.c for more details. */ static void check_publications_origin(WalReceiverConn *wrconn, List *publications, - bool copydata, char *origin, Oid *subrel_local_oids, + bool copydata, bool retain_dead_tuples, + char *origin, Oid *subrel_local_oids, int subrel_count, char *subname) { WalRcvExecResult *res; @@ -2108,9 +2275,29 @@ check_publications_origin(WalReceiverConn *wrconn, List *publications, Oid tableRow[1] = {TEXTOID}; List *publist = NIL; int i; + bool check_rdt; + bool check_table_sync; + bool origin_none = origin && + pg_strcasecmp(origin, LOGICALREP_ORIGIN_NONE) == 0; + + /* + * Enable retain_dead_tuples checks only when origin is set to 'any', + * since with origin='none' only local changes are replicated to the + * subscriber. + */ + check_rdt = retain_dead_tuples && !origin_none; + + /* + * Enable table synchronization checks only when origin is 'none', to + * ensure that data from other origins is not inadvertently copied. + */ + check_table_sync = copydata && origin_none; - if (!copydata || !origin || - (pg_strcasecmp(origin, LOGICALREP_ORIGIN_NONE) != 0)) + /* retain_dead_tuples and table sync checks occur separately */ + Assert(!(check_rdt && check_table_sync)); + + /* Return if no checks are required */ + if (!check_rdt && !check_table_sync) return; initStringInfo(&cmd); @@ -2129,16 +2316,23 @@ check_publications_origin(WalReceiverConn *wrconn, List *publications, /* * In case of ALTER SUBSCRIPTION ... REFRESH, subrel_local_oids contains * the list of relation oids that are already present on the subscriber. - * This check should be skipped for these tables. + * This check should be skipped for these tables if checking for table + * sync scenario. However, when handling the retain_dead_tuples scenario, + * ensure all tables are checked, as some existing tables may now include + * changes from other origins due to newly created subscriptions on the + * publisher. */ - for (i = 0; i < subrel_count; i++) + if (check_table_sync) { - Oid relid = subrel_local_oids[i]; - char *schemaname = get_namespace_name(get_rel_namespace(relid)); - char *tablename = get_rel_name(relid); + for (i = 0; i < subrel_count; i++) + { + Oid relid = subrel_local_oids[i]; + char *schemaname = get_namespace_name(get_rel_namespace(relid)); + char *tablename = get_rel_name(relid); - appendStringInfo(&cmd, "AND NOT (N.nspname = '%s' AND C.relname = '%s')\n", - schemaname, tablename); + appendStringInfo(&cmd, "AND NOT (N.nspname = '%s' AND C.relname = '%s')\n", + schemaname, tablename); + } } res = walrcv_exec(wrconn, cmd.data, 1, tableRow); @@ -2173,22 +2367,37 @@ check_publications_origin(WalReceiverConn *wrconn, List *publications, * XXX: For simplicity, we don't check whether the table has any data or * not. If the table doesn't have any data then we don't need to * distinguish between data having origin and data not having origin so we - * can avoid logging a warning in that case. + * can avoid logging a warning for table sync scenario. */ if (publist) { StringInfo pubnames = makeStringInfo(); + StringInfo err_msg = makeStringInfo(); + StringInfo err_hint = makeStringInfo(); /* Prepare the list of publication(s) for warning message. */ GetPublicationsStr(publist, pubnames, false); + + if (check_table_sync) + { + appendStringInfo(err_msg, _("subscription \"%s\" requested copy_data with origin = NONE but might copy data that had a different origin"), + subname); + appendStringInfoString(err_hint, _("Verify that initial data copied from the publisher tables did not come from other origins.")); + } + else + { + appendStringInfo(err_msg, _("subscription \"%s\" enabled retain_dead_tuples but might not reliably detect conflicts for changes from different origins"), + subname); + appendStringInfoString(err_hint, _("Consider using origin = NONE or disabling retain_dead_tuples.")); + } + ereport(WARNING, errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), - errmsg("subscription \"%s\" requested copy_data with origin = NONE but might copy data that had a different origin", - subname), - errdetail_plural("The subscription being created subscribes to a publication (%s) that contains tables that are written to by other subscriptions.", - "The subscription being created subscribes to publications (%s) that contain tables that are written to by other subscriptions.", + errmsg_internal("%s", err_msg->data), + errdetail_plural("The subscription subscribes to a publication (%s) that contains tables that are written to by other subscriptions.", + "The subscription subscribes to publications (%s) that contain tables that are written to by other subscriptions.", list_length(publist), pubnames->data), - errhint("Verify that initial data copied from the publisher tables did not come from other origins.")); + errhint_internal("%s", err_hint->data)); } ExecDropSingleTupleTableSlot(slot); @@ -2197,6 +2406,101 @@ check_publications_origin(WalReceiverConn *wrconn, List *publications, } /* + * Determine whether the retain_dead_tuples can be enabled based on the + * publisher's status. + * + * This option is disallowed if the publisher is running a version earlier + * than the PG19, or if the publisher is in recovery (i.e., it is a standby + * server). + * + * See comments atop worker.c for a detailed explanation. + */ +static void +check_pub_dead_tuple_retention(WalReceiverConn *wrconn) +{ + WalRcvExecResult *res; + Oid RecoveryRow[1] = {BOOLOID}; + TupleTableSlot *slot; + bool isnull; + bool remote_in_recovery; + + if (walrcv_server_version(wrconn) < 19000) + ereport(ERROR, + errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), + errmsg("cannot enable retain_dead_tuples if the publisher is running a version earlier than PostgreSQL 19")); + + res = walrcv_exec(wrconn, "SELECT pg_is_in_recovery()", 1, RecoveryRow); + + if (res->status != WALRCV_OK_TUPLES) + ereport(ERROR, + (errcode(ERRCODE_CONNECTION_FAILURE), + errmsg("could not obtain recovery progress from the publisher: %s", + res->err))); + + slot = MakeSingleTupleTableSlot(res->tupledesc, &TTSOpsMinimalTuple); + if (!tuplestore_gettupleslot(res->tuplestore, true, false, slot)) + elog(ERROR, "failed to fetch tuple for the recovery progress"); + + remote_in_recovery = DatumGetBool(slot_getattr(slot, 1, &isnull)); + + if (remote_in_recovery) + ereport(ERROR, + errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot enable retain_dead_tuples if the publisher is in recovery.")); + + ExecDropSingleTupleTableSlot(slot); + + walrcv_clear_result(res); +} + +/* + * Check if the subscriber's configuration is adequate to enable the + * retain_dead_tuples option. + * + * Issue an ERROR if the wal_level does not support the use of replication + * slots when check_guc is set to true. + * + * Issue a WARNING if track_commit_timestamp is not enabled when check_guc is + * set to true. This is only to highlight the importance of enabling + * track_commit_timestamp instead of catching all the misconfigurations, as + * this setting can be adjusted after subscription creation. Without it, the + * apply worker will simply skip conflict detection. + * + * Issue a WARNING or NOTICE if the subscription is disabled. Do not raise an + * ERROR since users can only modify retain_dead_tuples for disabled + * subscriptions. And as long as the subscription is enabled promptly, it will + * not pose issues. + */ +void +CheckSubDeadTupleRetention(bool check_guc, bool sub_disabled, + int elevel_for_sub_disabled) +{ + Assert(elevel_for_sub_disabled == NOTICE || + elevel_for_sub_disabled == WARNING); + + if (check_guc && wal_level < WAL_LEVEL_REPLICA) + ereport(ERROR, + errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), + errmsg("\"wal_level\" is insufficient to create the replication slot required by retain_dead_tuples"), + errhint("\"wal_level\" must be set to \"replica\" or \"logical\" at server start.")); + + if (check_guc && !track_commit_timestamp) + ereport(WARNING, + errcode(ERRCODE_INVALID_PARAMETER_VALUE), + errmsg("commit timestamp and origin data required for detecting conflicts won't be retained"), + errhint("Consider setting \"%s\" to true.", + "track_commit_timestamp")); + + if (sub_disabled) + ereport(elevel_for_sub_disabled, + errcode(ERRCODE_INVALID_PARAMETER_VALUE), + errmsg("deleted rows to detect conflicts would not be removed until the subscription is enabled"), + (elevel_for_sub_disabled > NOTICE) + ? errhint("Consider setting %s to false.", + "retain_dead_tuples") : 0); +} + +/* * Get the list of tables which belong to specified publications on the * publisher connection. * diff --git a/src/backend/commands/trigger.c b/src/backend/commands/trigger.c index 67f8e70f9c1..7dc121f73f1 100644 --- a/src/backend/commands/trigger.c +++ b/src/backend/commands/trigger.c @@ -80,6 +80,7 @@ static bool GetTupleForTrigger(EState *estate, ItemPointer tid, LockTupleMode lockmode, TupleTableSlot *oldslot, + bool do_epq_recheck, TupleTableSlot **epqslot, TM_Result *tmresultp, TM_FailureData *tmfdp); @@ -2693,7 +2694,8 @@ ExecBRDeleteTriggers(EState *estate, EPQState *epqstate, HeapTuple fdw_trigtuple, TupleTableSlot **epqslot, TM_Result *tmresult, - TM_FailureData *tmfd) + TM_FailureData *tmfd, + bool is_merge_delete) { TupleTableSlot *slot = ExecGetTriggerOldSlot(estate, relinfo); TriggerDesc *trigdesc = relinfo->ri_TrigDesc; @@ -2708,9 +2710,17 @@ ExecBRDeleteTriggers(EState *estate, EPQState *epqstate, { TupleTableSlot *epqslot_candidate = NULL; + /* + * Get a copy of the on-disk tuple we are planning to delete. In + * general, if the tuple has been concurrently updated, we should + * recheck it using EPQ. However, if this is a MERGE DELETE action, + * we skip this EPQ recheck and leave it to the caller (it must do + * additional rechecking, and might end up executing a different + * action entirely). + */ if (!GetTupleForTrigger(estate, epqstate, relinfo, tupleid, - LockTupleExclusive, slot, &epqslot_candidate, - tmresult, tmfd)) + LockTupleExclusive, slot, !is_merge_delete, + &epqslot_candidate, tmresult, tmfd)) return false; /* @@ -2800,6 +2810,7 @@ ExecARDeleteTriggers(EState *estate, tupleid, LockTupleExclusive, slot, + false, NULL, NULL, NULL); @@ -2944,7 +2955,8 @@ ExecBRUpdateTriggers(EState *estate, EPQState *epqstate, HeapTuple fdw_trigtuple, TupleTableSlot *newslot, TM_Result *tmresult, - TM_FailureData *tmfd) + TM_FailureData *tmfd, + bool is_merge_update) { TriggerDesc *trigdesc = relinfo->ri_TrigDesc; TupleTableSlot *oldslot = ExecGetTriggerOldSlot(estate, relinfo); @@ -2965,10 +2977,17 @@ ExecBRUpdateTriggers(EState *estate, EPQState *epqstate, { TupleTableSlot *epqslot_candidate = NULL; - /* get a copy of the on-disk tuple we are planning to update */ + /* + * Get a copy of the on-disk tuple we are planning to update. In + * general, if the tuple has been concurrently updated, we should + * recheck it using EPQ. However, if this is a MERGE UPDATE action, + * we skip this EPQ recheck and leave it to the caller (it must do + * additional rechecking, and might end up executing a different + * action entirely). + */ if (!GetTupleForTrigger(estate, epqstate, relinfo, tupleid, - lockmode, oldslot, &epqslot_candidate, - tmresult, tmfd)) + lockmode, oldslot, !is_merge_update, + &epqslot_candidate, tmresult, tmfd)) return false; /* cancel the update action */ /* @@ -3142,6 +3161,7 @@ ExecARUpdateTriggers(EState *estate, ResultRelInfo *relinfo, tupleid, LockTupleExclusive, oldslot, + false, NULL, NULL, NULL); @@ -3298,6 +3318,7 @@ GetTupleForTrigger(EState *estate, ItemPointer tid, LockTupleMode lockmode, TupleTableSlot *oldslot, + bool do_epq_recheck, TupleTableSlot **epqslot, TM_Result *tmresultp, TM_FailureData *tmfdp) @@ -3357,29 +3378,30 @@ GetTupleForTrigger(EState *estate, if (tmfd.traversed) { /* - * Recheck the tuple using EPQ. For MERGE, we leave this - * to the caller (it must do additional rechecking, and - * might end up executing a different action entirely). + * Recheck the tuple using EPQ, if requested. Otherwise, + * just return that it was concurrently updated. */ - if (estate->es_plannedstmt->commandType == CMD_MERGE) + if (do_epq_recheck) { - if (tmresultp) - *tmresultp = TM_Updated; - return false; + *epqslot = EvalPlanQual(epqstate, + relation, + relinfo->ri_RangeTableIndex, + oldslot); + + /* + * If PlanQual failed for updated tuple - we must not + * process this tuple! + */ + if (TupIsNull(*epqslot)) + { + *epqslot = NULL; + return false; + } } - - *epqslot = EvalPlanQual(epqstate, - relation, - relinfo->ri_RangeTableIndex, - oldslot); - - /* - * If PlanQual failed for updated tuple - we must not - * process this tuple! - */ - if (TupIsNull(*epqslot)) + else { - *epqslot = NULL; + if (tmresultp) + *tmresultp = TM_Updated; return false; } } diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c index 8a72b5e70a4..1a37737d4a2 100644 --- a/src/backend/executor/execExprInterp.c +++ b/src/backend/executor/execExprInterp.c @@ -5228,7 +5228,6 @@ ExecEvalJsonCoercionFinish(ExprState *state, ExprEvalStep *op) * JsonBehavior expression. */ jsestate->escontext.error_occurred = false; - jsestate->escontext.error_occurred = false; jsestate->escontext.details_wanted = true; } } diff --git a/src/backend/executor/execParallel.c b/src/backend/executor/execParallel.c index f3e77bda279..f098a5557cf 100644 --- a/src/backend/executor/execParallel.c +++ b/src/backend/executor/execParallel.c @@ -189,6 +189,7 @@ ExecSerializePlan(Plan *plan, EState *estate) pstmt->permInfos = estate->es_rteperminfos; pstmt->resultRelations = NIL; pstmt->appendRelations = NIL; + pstmt->planOrigin = PLAN_STMT_INTERNAL; /* * Transfer only parallel-safe subplans, leaving a NULL "hole" in the list diff --git a/src/backend/executor/execReplication.c b/src/backend/executor/execReplication.c index 53ddd25c42d..f262e7a66f7 100644 --- a/src/backend/executor/execReplication.c +++ b/src/backend/executor/execReplication.c @@ -670,7 +670,7 @@ ExecSimpleRelationUpdate(ResultRelInfo *resultRelInfo, resultRelInfo->ri_TrigDesc->trig_update_before_row) { if (!ExecBRUpdateTriggers(estate, epqstate, resultRelInfo, - tid, NULL, slot, NULL, NULL)) + tid, NULL, slot, NULL, NULL, false)) skip_tuple = true; /* "do nothing" */ } @@ -746,7 +746,7 @@ ExecSimpleRelationDelete(ResultRelInfo *resultRelInfo, resultRelInfo->ri_TrigDesc->trig_delete_before_row) { skip_tuple = !ExecBRDeleteTriggers(estate, epqstate, resultRelInfo, - tid, NULL, NULL, NULL, NULL); + tid, NULL, NULL, NULL, NULL, false); } if (!skip_tuple) diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c index 54da8e7995b..7c6c2c1f6e4 100644 --- a/src/backend/executor/nodeModifyTable.c +++ b/src/backend/executor/nodeModifyTable.c @@ -1474,7 +1474,8 @@ ExecDeletePrologue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, return ExecBRDeleteTriggers(context->estate, context->epqstate, resultRelInfo, tupleid, oldtuple, - epqreturnslot, result, &context->tmfd); + epqreturnslot, result, &context->tmfd, + context->mtstate->operation == CMD_MERGE); } return true; @@ -2117,7 +2118,8 @@ ExecUpdatePrologue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, return ExecBRUpdateTriggers(context->estate, context->epqstate, resultRelInfo, tupleid, oldtuple, slot, - result, &context->tmfd); + result, &context->tmfd, + context->mtstate->operation == CMD_MERGE); } return true; diff --git a/src/backend/jit/llvm/Makefile b/src/backend/jit/llvm/Makefile index e8c12060b93..68677ba42e1 100644 --- a/src/backend/jit/llvm/Makefile +++ b/src/backend/jit/llvm/Makefile @@ -31,7 +31,7 @@ endif # All files in this directory use LLVM. CFLAGS += $(LLVM_CFLAGS) CXXFLAGS += $(LLVM_CXXFLAGS) -override CPPFLAGS := $(LLVM_CPPFLAGS) $(CPPFLAGS) +override CPPFLAGS += $(LLVM_CPPFLAGS) SHLIB_LINK += $(LLVM_LIBS) # Because this module includes C++ files, we need to use a C++ diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c index 9f4d05ffbd4..4da46666439 100644 --- a/src/backend/libpq/auth.c +++ b/src/backend/libpq/auth.c @@ -94,8 +94,16 @@ static int auth_peer(hbaPort *port); #define PGSQL_PAM_SERVICE "postgresql" /* Service name passed to PAM */ +/* Work around original Solaris' lack of "const" in the conv_proc signature */ +#ifdef _PAM_LEGACY_NONCONST +#define PG_PAM_CONST +#else +#define PG_PAM_CONST const +#endif + static int CheckPAMAuth(Port *port, const char *user, const char *password); -static int pam_passwd_conv_proc(int num_msg, const struct pam_message **msg, +static int pam_passwd_conv_proc(int num_msg, + PG_PAM_CONST struct pam_message **msg, struct pam_response **resp, void *appdata_ptr); static struct pam_conv pam_passw_conv = { @@ -1917,7 +1925,7 @@ auth_peer(hbaPort *port) */ static int -pam_passwd_conv_proc(int num_msg, const struct pam_message **msg, +pam_passwd_conv_proc(int num_msg, PG_PAM_CONST struct pam_message **msg, struct pam_response **resp, void *appdata_ptr) { const char *passwd; diff --git a/src/backend/libpq/pg_ident.conf.sample b/src/backend/libpq/pg_ident.conf.sample index f5225f26cdf..8ee6c0ba315 100644 --- a/src/backend/libpq/pg_ident.conf.sample +++ b/src/backend/libpq/pg_ident.conf.sample @@ -13,25 +13,25 @@ # user names to their corresponding PostgreSQL user names. Records # are of the form: # -# MAPNAME SYSTEM-USERNAME PG-USERNAME +# MAPNAME SYSTEM-USERNAME DATABASE-USERNAME # # (The uppercase quantities must be replaced by actual values.) # # MAPNAME is the (otherwise freely chosen) map name that was used in # pg_hba.conf. SYSTEM-USERNAME is the detected user name of the -# client. PG-USERNAME is the requested PostgreSQL user name. The -# existence of a record specifies that SYSTEM-USERNAME may connect as -# PG-USERNAME. +# client. DATABASE-USERNAME is the requested PostgreSQL user name. +# The existence of a record specifies that SYSTEM-USERNAME may connect +# as DATABASE-USERNAME. # -# If SYSTEM-USERNAME starts with a slash (/), it will be treated as a -# regular expression. Optionally this can contain a capture (a -# parenthesized subexpression). The substring matching the capture -# will be substituted for \1 (backslash-one) if present in -# PG-USERNAME. +# If SYSTEM-USERNAME starts with a slash (/), the rest of it will be +# treated as a regular expression. Optionally this can contain a capture +# (a parenthesized subexpression). The substring matching the capture +# will be substituted for \1 (backslash-one) if that appears in +# DATABASE-USERNAME. # -# PG-USERNAME can be "all", a user name, a group name prefixed with "+", or -# a regular expression (if it starts with a slash (/)). If it is a regular -# expression, the substring matching with \1 has no effect. +# DATABASE-USERNAME can be "all", a user name, a group name prefixed with "+", +# or a regular expression (if it starts with a slash (/)). If it is a regular +# expression, no substitution for \1 will occur. # # Multiple maps may be specified in this file and used by pg_hba.conf. # @@ -69,4 +69,4 @@ # Put your actual configuration here # ---------------------------------- -# MAPNAME SYSTEM-USERNAME PG-USERNAME +# MAPNAME SYSTEM-USERNAME DATABASE-USERNAME diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c index e5171467de1..25f739a6a17 100644 --- a/src/backend/libpq/pqcomm.c +++ b/src/backend/libpq/pqcomm.c @@ -858,7 +858,6 @@ RemoveSocketFiles(void) (void) unlink(sock_path); } /* Since we're about to exit, no need to reclaim storage */ - sock_paths = NIL; } diff --git a/src/backend/libpq/pqmq.c b/src/backend/libpq/pqmq.c index f1a08bc32ca..5f39949a367 100644 --- a/src/backend/libpq/pqmq.c +++ b/src/backend/libpq/pqmq.c @@ -23,7 +23,7 @@ #include "tcop/tcopprot.h" #include "utils/builtins.h" -static shm_mq_handle *pq_mq_handle; +static shm_mq_handle *pq_mq_handle = NULL; static bool pq_mq_busy = false; static pid_t pq_mq_parallel_leader_pid = 0; static ProcNumber pq_mq_parallel_leader_proc_number = INVALID_PROC_NUMBER; @@ -66,7 +66,11 @@ pq_redirect_to_shm_mq(dsm_segment *seg, shm_mq_handle *mqh) static void pq_cleanup_redirect_to_shm_mq(dsm_segment *seg, Datum arg) { - pq_mq_handle = NULL; + if (pq_mq_handle != NULL) + { + pfree(pq_mq_handle); + pq_mq_handle = NULL; + } whereToSendOutput = DestNone; } @@ -131,8 +135,11 @@ mq_putmessage(char msgtype, const char *s, size_t len) if (pq_mq_busy) { if (pq_mq_handle != NULL) + { shm_mq_detach(pq_mq_handle); - pq_mq_handle = NULL; + pfree(pq_mq_handle); + pq_mq_handle = NULL; + } return EOF; } @@ -152,8 +159,6 @@ mq_putmessage(char msgtype, const char *s, size_t len) iov[1].data = s; iov[1].len = len; - Assert(pq_mq_handle != NULL); - for (;;) { /* @@ -161,6 +166,7 @@ mq_putmessage(char msgtype, const char *s, size_t len) * that the shared memory value is updated before we send the parallel * message signal right after this. */ + Assert(pq_mq_handle != NULL); result = shm_mq_sendv(pq_mq_handle, iov, 2, true, true); if (pq_mq_parallel_leader_pid != 0) diff --git a/src/backend/main/main.c b/src/backend/main/main.c index 7d63cf94a6b..bdcb5e4f261 100644 --- a/src/backend/main/main.c +++ b/src/backend/main/main.c @@ -125,13 +125,17 @@ main(int argc, char *argv[]) set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("postgres")); /* - * In the postmaster, absorb the environment values for LC_COLLATE and - * LC_CTYPE. Individual backends will change these later to settings - * taken from pg_database, but the postmaster cannot do that. If we leave - * these set to "C" then message localization might not work well in the - * postmaster. + * Collation is handled by pg_locale.c, and the behavior is dependent on + * the provider. strcoll(), etc., should not be called directly. + */ + init_locale("LC_COLLATE", LC_COLLATE, "C"); + + /* + * In the postmaster, absorb the environment value for LC_CTYPE. + * Individual backends will change it later to pg_database.datctype, but + * the postmaster cannot do that. If we leave it set to "C" then message + * localization might not work well in the postmaster. */ - init_locale("LC_COLLATE", LC_COLLATE, ""); init_locale("LC_CTYPE", LC_CTYPE, ""); /* diff --git a/src/backend/optimizer/path/costsize.c b/src/backend/optimizer/path/costsize.c index 1f04a2c182c..344a3188317 100644 --- a/src/backend/optimizer/path/costsize.c +++ b/src/backend/optimizer/path/costsize.c @@ -2572,13 +2572,13 @@ cost_memoize_rescan(PlannerInfo *root, MemoizePath *mpath, Cost input_startup_cost = mpath->subpath->startup_cost; Cost input_total_cost = mpath->subpath->total_cost; double tuples = mpath->subpath->rows; - double calls = mpath->calls; + Cardinality est_calls = mpath->est_calls; int width = mpath->subpath->pathtarget->width; double hash_mem_bytes; double est_entry_bytes; - double est_cache_entries; - double ndistinct; + Cardinality est_cache_entries; + Cardinality ndistinct; double evict_ratio; double hit_ratio; Cost startup_cost; @@ -2604,7 +2604,7 @@ cost_memoize_rescan(PlannerInfo *root, MemoizePath *mpath, est_cache_entries = floor(hash_mem_bytes / est_entry_bytes); /* estimate on the distinct number of parameter values */ - ndistinct = estimate_num_groups(root, mpath->param_exprs, calls, NULL, + ndistinct = estimate_num_groups(root, mpath->param_exprs, est_calls, NULL, &estinfo); /* @@ -2616,7 +2616,10 @@ cost_memoize_rescan(PlannerInfo *root, MemoizePath *mpath, * certainly mean a MemoizePath will never survive add_path(). */ if ((estinfo.flags & SELFLAG_USED_DEFAULT) != 0) - ndistinct = calls; + ndistinct = est_calls; + + /* Remember the ndistinct estimate for EXPLAIN */ + mpath->est_unique_keys = ndistinct; /* * Since we've already estimated the maximum number of entries we can @@ -2644,9 +2647,12 @@ cost_memoize_rescan(PlannerInfo *root, MemoizePath *mpath, * must look at how many scans are estimated in total for this node and * how many of those scans we expect to get a cache hit. */ - hit_ratio = ((calls - ndistinct) / calls) * + hit_ratio = ((est_calls - ndistinct) / est_calls) * (est_cache_entries / Max(ndistinct, est_cache_entries)); + /* Remember the hit ratio estimate for EXPLAIN */ + mpath->est_hit_ratio = hit_ratio; + Assert(hit_ratio >= 0 && hit_ratio <= 1.0); /* diff --git a/src/backend/optimizer/plan/createplan.c b/src/backend/optimizer/plan/createplan.c index 8a9f1d7a943..bfefc7dbea1 100644 --- a/src/backend/optimizer/plan/createplan.c +++ b/src/backend/optimizer/plan/createplan.c @@ -284,7 +284,10 @@ static Material *make_material(Plan *lefttree); static Memoize *make_memoize(Plan *lefttree, Oid *hashoperators, Oid *collations, List *param_exprs, bool singlerow, bool binary_mode, - uint32 est_entries, Bitmapset *keyparamids); + uint32 est_entries, Bitmapset *keyparamids, + Cardinality est_calls, + Cardinality est_unique_keys, + double est_hit_ratio); static WindowAgg *make_windowagg(List *tlist, WindowClause *wc, int partNumCols, AttrNumber *partColIdx, Oid *partOperators, Oid *partCollations, int ordNumCols, AttrNumber *ordColIdx, Oid *ordOperators, Oid *ordCollations, @@ -1753,7 +1756,8 @@ create_memoize_plan(PlannerInfo *root, MemoizePath *best_path, int flags) plan = make_memoize(subplan, operators, collations, param_exprs, best_path->singlerow, best_path->binary_mode, - best_path->est_entries, keyparamids); + best_path->est_entries, keyparamids, best_path->est_calls, + best_path->est_unique_keys, best_path->est_hit_ratio); copy_generic_path_info(&plan->plan, (Path *) best_path); @@ -6749,7 +6753,9 @@ materialize_finished_plan(Plan *subplan) static Memoize * make_memoize(Plan *lefttree, Oid *hashoperators, Oid *collations, List *param_exprs, bool singlerow, bool binary_mode, - uint32 est_entries, Bitmapset *keyparamids) + uint32 est_entries, Bitmapset *keyparamids, + Cardinality est_calls, Cardinality est_unique_keys, + double est_hit_ratio) { Memoize *node = makeNode(Memoize); Plan *plan = &node->plan; @@ -6767,6 +6773,9 @@ make_memoize(Plan *lefttree, Oid *hashoperators, Oid *collations, node->binary_mode = binary_mode; node->est_entries = est_entries; node->keyparamids = keyparamids; + node->est_calls = est_calls; + node->est_unique_keys = est_unique_keys; + node->est_hit_ratio = est_hit_ratio; return node; } diff --git a/src/backend/optimizer/plan/initsplan.c b/src/backend/optimizer/plan/initsplan.c index 01804b085b3..3e3fec89252 100644 --- a/src/backend/optimizer/plan/initsplan.c +++ b/src/backend/optimizer/plan/initsplan.c @@ -3048,36 +3048,16 @@ add_base_clause_to_rel(PlannerInfo *root, Index relid, * expr_is_nonnullable * Check to see if the Expr cannot be NULL * - * If the Expr is a simple Var that is defined NOT NULL and meanwhile is not - * nulled by any outer joins, then we can know that it cannot be NULL. + * Currently we only support simple Vars. */ static bool expr_is_nonnullable(PlannerInfo *root, Expr *expr) { - RelOptInfo *rel; - Var *var; - /* For now only check simple Vars */ if (!IsA(expr, Var)) return false; - var = (Var *) expr; - - /* could the Var be nulled by any outer joins? */ - if (!bms_is_empty(var->varnullingrels)) - return false; - - /* system columns cannot be NULL */ - if (var->varattno < 0) - return true; - - /* is the column defined NOT NULL? */ - rel = find_base_rel(root, var->varno); - if (var->varattno > 0 && - bms_is_member(var->varattno, rel->notnullattnums)) - return true; - - return false; + return var_is_nonnullable(root, (Var *) expr, true); } /* diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c index 549aedcfa99..d59d6e4c6a0 100644 --- a/src/backend/optimizer/plan/planner.c +++ b/src/backend/optimizer/plan/planner.c @@ -342,6 +342,7 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions, glob->transientPlan = false; glob->dependsOnRole = false; glob->partition_directory = NULL; + glob->rel_notnullatts_hash = NULL; /* * Assess whether it's feasible to use parallel mode for this query. We @@ -557,6 +558,7 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions, result->commandType = parse->commandType; result->queryId = parse->queryId; + result->planOrigin = PLAN_STMT_STANDARD; result->hasReturning = (parse->returningList != NIL); result->hasModifyingCTE = parse->hasModifyingCTE; result->canSetTag = parse->canSetTag; @@ -721,6 +723,18 @@ subquery_planner(PlannerGlobal *glob, Query *parse, PlannerInfo *parent_root, transform_MERGE_to_join(parse); /* + * Scan the rangetable for relation RTEs and retrieve the necessary + * catalog information for each relation. Using this information, clear + * the inh flag for any relation that has no children, collect not-null + * attribute numbers for any relation that has column not-null + * constraints, and expand virtual generated columns for any relation that + * contains them. Note that this step does not descend into sublinks and + * subqueries; if we pull up any sublinks or subqueries below, their + * relation RTEs are processed just before pulling them up. + */ + parse = root->parse = preprocess_relation_rtes(root); + + /* * If the FROM clause is empty, replace it with a dummy RTE_RESULT RTE, so * that we don't need so many special cases to deal with that situation. */ @@ -744,14 +758,6 @@ subquery_planner(PlannerGlobal *glob, Query *parse, PlannerInfo *parent_root, preprocess_function_rtes(root); /* - * Scan the rangetable for relations with virtual generated columns, and - * replace all Var nodes in the query that reference these columns with - * the generation expressions. Recursion issues here are handled in the - * same way as for SubLinks. - */ - parse = root->parse = expand_virtual_generated_columns(root); - - /* * Check to see if any subqueries in the jointree can be merged into this * query. */ @@ -787,23 +793,6 @@ subquery_planner(PlannerGlobal *glob, Query *parse, PlannerInfo *parent_root, switch (rte->rtekind) { - case RTE_RELATION: - if (rte->inh) - { - /* - * Check to see if the relation actually has any children; - * if not, clear the inh flag so we can treat it as a - * plain base relation. - * - * Note: this could give a false-positive result, if the - * rel once had children but no longer does. We used to - * be able to clear rte->inh later on when we discovered - * that, but no more; we have to handle such cases as - * full-fledged inheritance. - */ - rte->inh = has_subclass(rte->relid); - } - break; case RTE_JOIN: root->hasJoinRTEs = true; if (IS_OUTER_JOIN(rte->jointype)) diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c index e7cb3fede66..d71ed958e31 100644 --- a/src/backend/optimizer/plan/subselect.c +++ b/src/backend/optimizer/plan/subselect.c @@ -1454,6 +1454,7 @@ convert_EXISTS_sublink_to_join(PlannerInfo *root, SubLink *sublink, Query *parse = root->parse; Query *subselect = (Query *) sublink->subselect; Node *whereClause; + PlannerInfo subroot; int rtoffset; int varno; Relids clause_varnos; @@ -1516,6 +1517,35 @@ convert_EXISTS_sublink_to_join(PlannerInfo *root, SubLink *sublink, return NULL; /* + * Scan the rangetable for relation RTEs and retrieve the necessary + * catalog information for each relation. Using this information, clear + * the inh flag for any relation that has no children, collect not-null + * attribute numbers for any relation that has column not-null + * constraints, and expand virtual generated columns for any relation that + * contains them. + * + * Note: we construct up an entirely dummy PlannerInfo for use here. This + * is fine because only the "glob" and "parse" links will be used in this + * case. + * + * Note: we temporarily assign back the WHERE clause so that any virtual + * generated column references within it can be expanded. It should be + * separated out again afterward. + */ + MemSet(&subroot, 0, sizeof(subroot)); + subroot.type = T_PlannerInfo; + subroot.glob = root->glob; + subroot.parse = subselect; + subselect->jointree->quals = whereClause; + subselect = preprocess_relation_rtes(&subroot); + + /* + * Now separate out the WHERE clause again. + */ + whereClause = subselect->jointree->quals; + subselect->jointree->quals = NULL; + + /* * The subquery must have a nonempty jointree, but we can make it so. */ replace_empty_jointree(subselect); @@ -1732,6 +1762,7 @@ convert_EXISTS_to_ANY(PlannerInfo *root, Query *subselect, Node **testexpr, List **paramIds) { Node *whereClause; + PlannerInfo subroot; List *leftargs, *rightargs, *opids, @@ -1791,12 +1822,15 @@ convert_EXISTS_to_ANY(PlannerInfo *root, Query *subselect, * parent aliases were flattened already, and we're not going to pull any * child Vars (of any description) into the parent. * - * Note: passing the parent's root to eval_const_expressions is - * technically wrong, but we can get away with it since only the - * boundParams (if any) are used, and those would be the same in a - * subroot. - */ - whereClause = eval_const_expressions(root, whereClause); + * Note: we construct up an entirely dummy PlannerInfo to pass to + * eval_const_expressions. This is fine because only the "glob" and + * "parse" links are used by eval_const_expressions. + */ + MemSet(&subroot, 0, sizeof(subroot)); + subroot.type = T_PlannerInfo; + subroot.glob = root->glob; + subroot.parse = subselect; + whereClause = eval_const_expressions(&subroot, whereClause); whereClause = (Node *) canonicalize_qual((Expr *) whereClause, false); whereClause = (Node *) make_ands_implicit((Expr *) whereClause); diff --git a/src/backend/optimizer/prep/prepjointree.c b/src/backend/optimizer/prep/prepjointree.c index 87dc6f56b57..35e8d3c183b 100644 --- a/src/backend/optimizer/prep/prepjointree.c +++ b/src/backend/optimizer/prep/prepjointree.c @@ -4,10 +4,10 @@ * Planner preprocessing for subqueries and join tree manipulation. * * NOTE: the intended sequence for invoking these operations is + * preprocess_relation_rtes * replace_empty_jointree * pull_up_sublinks * preprocess_function_rtes - * expand_virtual_generated_columns * pull_up_subqueries * flatten_simple_union_all * do expression preprocessing (including flattening JOIN alias vars) @@ -36,6 +36,7 @@ #include "optimizer/clauses.h" #include "optimizer/optimizer.h" #include "optimizer/placeholder.h" +#include "optimizer/plancat.h" #include "optimizer/prep.h" #include "optimizer/subselect.h" #include "optimizer/tlist.h" @@ -102,6 +103,9 @@ typedef struct reduce_outer_joins_partial_state Relids unreduced_side; /* relids in its still-nullable side */ } reduce_outer_joins_partial_state; +static Query *expand_virtual_generated_columns(PlannerInfo *root, Query *parse, + RangeTblEntry *rte, int rt_index, + Relation relation); static Node *pull_up_sublinks_jointree_recurse(PlannerInfo *root, Node *jtnode, Relids *relids); static Node *pull_up_sublinks_qual_recurse(PlannerInfo *root, Node *node, @@ -393,6 +397,181 @@ transform_MERGE_to_join(Query *parse) } /* + * preprocess_relation_rtes + * Do the preprocessing work for any relation RTEs in the FROM clause. + * + * This scans the rangetable for relation RTEs and retrieves the necessary + * catalog information for each relation. Using this information, it clears + * the inh flag for any relation that has no children, collects not-null + * attribute numbers for any relation that has column not-null constraints, and + * expands virtual generated columns for any relation that contains them. + * + * Note that expanding virtual generated columns may cause the query tree to + * have new copies of rangetable entries. Therefore, we have to use list_nth + * instead of foreach when iterating over the query's rangetable. + * + * Returns a modified copy of the query tree, if any relations with virtual + * generated columns are present. + */ +Query * +preprocess_relation_rtes(PlannerInfo *root) +{ + Query *parse = root->parse; + int rtable_size; + int rt_index; + + rtable_size = list_length(parse->rtable); + + for (rt_index = 0; rt_index < rtable_size; rt_index++) + { + RangeTblEntry *rte = rt_fetch(rt_index + 1, parse->rtable); + Relation relation; + + /* We only care about relation RTEs. */ + if (rte->rtekind != RTE_RELATION) + continue; + + /* + * We need not lock the relation since it was already locked by the + * rewriter. + */ + relation = table_open(rte->relid, NoLock); + + /* + * Check to see if the relation actually has any children; if not, + * clear the inh flag so we can treat it as a plain base relation. + * + * Note: this could give a false-positive result, if the rel once had + * children but no longer does. We used to be able to clear rte->inh + * later on when we discovered that, but no more; we have to handle + * such cases as full-fledged inheritance. + */ + if (rte->inh) + rte->inh = relation->rd_rel->relhassubclass; + + /* + * Check to see if the relation has any column not-null constraints; + * if so, retrieve the constraint information and store it in a + * relation OID based hash table. + */ + get_relation_notnullatts(root, relation); + + /* + * Check to see if the relation has any virtual generated columns; if + * so, replace all Var nodes in the query that reference these columns + * with the generation expressions. + */ + parse = expand_virtual_generated_columns(root, parse, + rte, rt_index + 1, + relation); + + table_close(relation, NoLock); + } + + return parse; +} + +/* + * expand_virtual_generated_columns + * Expand virtual generated columns for the given relation. + * + * This checks whether the given relation has any virtual generated columns, + * and if so, replaces all Var nodes in the query that reference those columns + * with their generation expressions. + * + * Returns a modified copy of the query tree if the relation contains virtual + * generated columns. + */ +static Query * +expand_virtual_generated_columns(PlannerInfo *root, Query *parse, + RangeTblEntry *rte, int rt_index, + Relation relation) +{ + TupleDesc tupdesc; + + /* Only normal relations can have virtual generated columns */ + Assert(rte->rtekind == RTE_RELATION); + + tupdesc = RelationGetDescr(relation); + if (tupdesc->constr && tupdesc->constr->has_generated_virtual) + { + List *tlist = NIL; + pullup_replace_vars_context rvcontext; + + for (int i = 0; i < tupdesc->natts; i++) + { + Form_pg_attribute attr = TupleDescAttr(tupdesc, i); + TargetEntry *tle; + + if (attr->attgenerated == ATTRIBUTE_GENERATED_VIRTUAL) + { + Node *defexpr; + + defexpr = build_generation_expression(relation, i + 1); + ChangeVarNodes(defexpr, 1, rt_index, 0); + + tle = makeTargetEntry((Expr *) defexpr, i + 1, 0, false); + tlist = lappend(tlist, tle); + } + else + { + Var *var; + + var = makeVar(rt_index, + i + 1, + attr->atttypid, + attr->atttypmod, + attr->attcollation, + 0); + + tle = makeTargetEntry((Expr *) var, i + 1, 0, false); + tlist = lappend(tlist, tle); + } + } + + Assert(list_length(tlist) > 0); + Assert(!rte->lateral); + + /* + * The relation's targetlist items are now in the appropriate form to + * insert into the query, except that we may need to wrap them in + * PlaceHolderVars. Set up required context data for + * pullup_replace_vars. + */ + rvcontext.root = root; + rvcontext.targetlist = tlist; + rvcontext.target_rte = rte; + rvcontext.result_relation = parse->resultRelation; + /* won't need these values */ + rvcontext.relids = NULL; + rvcontext.nullinfo = NULL; + /* pass NULL for outer_hasSubLinks */ + rvcontext.outer_hasSubLinks = NULL; + rvcontext.varno = rt_index; + /* this flag will be set below, if needed */ + rvcontext.wrap_option = REPLACE_WRAP_NONE; + /* initialize cache array with indexes 0 .. length(tlist) */ + rvcontext.rv_cache = palloc0((list_length(tlist) + 1) * + sizeof(Node *)); + + /* + * If the query uses grouping sets, we need a PlaceHolderVar for each + * expression of the relation's targetlist items. (See comments in + * pull_up_simple_subquery().) + */ + if (parse->groupingSets) + rvcontext.wrap_option = REPLACE_WRAP_ALL; + + /* + * Apply pullup variable replacement throughout the query tree. + */ + parse = (Query *) pullup_replace_vars((Node *) parse, &rvcontext); + } + + return parse; +} + +/* * replace_empty_jointree * If the Query's jointree is empty, replace it with a dummy RTE_RESULT * relation. @@ -950,128 +1129,6 @@ preprocess_function_rtes(PlannerInfo *root) } /* - * expand_virtual_generated_columns - * Expand all virtual generated column references in a query. - * - * This scans the rangetable for relations with virtual generated columns, and - * replaces all Var nodes in the query that reference these columns with the - * generation expressions. Note that we do not descend into subqueries; that - * is taken care of when the subqueries are planned. - * - * This has to be done after we have pulled up any SubLinks within the query's - * quals; otherwise any virtual generated column references within the SubLinks - * that should be transformed into joins wouldn't get expanded. - * - * Returns a modified copy of the query tree, if any relations with virtual - * generated columns are present. - */ -Query * -expand_virtual_generated_columns(PlannerInfo *root) -{ - Query *parse = root->parse; - int rt_index; - ListCell *lc; - - rt_index = 0; - foreach(lc, parse->rtable) - { - RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc); - Relation rel; - TupleDesc tupdesc; - - ++rt_index; - - /* - * Only normal relations can have virtual generated columns. - */ - if (rte->rtekind != RTE_RELATION) - continue; - - rel = table_open(rte->relid, NoLock); - - tupdesc = RelationGetDescr(rel); - if (tupdesc->constr && tupdesc->constr->has_generated_virtual) - { - List *tlist = NIL; - pullup_replace_vars_context rvcontext; - - for (int i = 0; i < tupdesc->natts; i++) - { - Form_pg_attribute attr = TupleDescAttr(tupdesc, i); - TargetEntry *tle; - - if (attr->attgenerated == ATTRIBUTE_GENERATED_VIRTUAL) - { - Node *defexpr; - - defexpr = build_generation_expression(rel, i + 1); - ChangeVarNodes(defexpr, 1, rt_index, 0); - - tle = makeTargetEntry((Expr *) defexpr, i + 1, 0, false); - tlist = lappend(tlist, tle); - } - else - { - Var *var; - - var = makeVar(rt_index, - i + 1, - attr->atttypid, - attr->atttypmod, - attr->attcollation, - 0); - - tle = makeTargetEntry((Expr *) var, i + 1, 0, false); - tlist = lappend(tlist, tle); - } - } - - Assert(list_length(tlist) > 0); - Assert(!rte->lateral); - - /* - * The relation's targetlist items are now in the appropriate form - * to insert into the query, except that we may need to wrap them - * in PlaceHolderVars. Set up required context data for - * pullup_replace_vars. - */ - rvcontext.root = root; - rvcontext.targetlist = tlist; - rvcontext.target_rte = rte; - rvcontext.result_relation = parse->resultRelation; - /* won't need these values */ - rvcontext.relids = NULL; - rvcontext.nullinfo = NULL; - /* pass NULL for outer_hasSubLinks */ - rvcontext.outer_hasSubLinks = NULL; - rvcontext.varno = rt_index; - /* this flag will be set below, if needed */ - rvcontext.wrap_option = REPLACE_WRAP_NONE; - /* initialize cache array with indexes 0 .. length(tlist) */ - rvcontext.rv_cache = palloc0((list_length(tlist) + 1) * - sizeof(Node *)); - - /* - * If the query uses grouping sets, we need a PlaceHolderVar for - * each expression of the relation's targetlist items. (See - * comments in pull_up_simple_subquery().) - */ - if (parse->groupingSets) - rvcontext.wrap_option = REPLACE_WRAP_ALL; - - /* - * Apply pullup variable replacement throughout the query tree. - */ - parse = (Query *) pullup_replace_vars((Node *) parse, &rvcontext); - } - - table_close(rel, NoLock); - } - - return parse; -} - -/* * pull_up_subqueries * Look for subqueries in the rangetable that can be pulled up into * the parent query. If the subquery has no special features like @@ -1334,6 +1391,16 @@ pull_up_simple_subquery(PlannerInfo *root, Node *jtnode, RangeTblEntry *rte, Assert(subquery->cteList == NIL); /* + * Scan the rangetable for relation RTEs and retrieve the necessary + * catalog information for each relation. Using this information, clear + * the inh flag for any relation that has no children, collect not-null + * attribute numbers for any relation that has column not-null + * constraints, and expand virtual generated columns for any relation that + * contains them. + */ + subquery = subroot->parse = preprocess_relation_rtes(subroot); + + /* * If the FROM clause is empty, replace it with a dummy RTE_RESULT RTE, so * that we don't need so many special cases to deal with that situation. */ @@ -1353,13 +1420,6 @@ pull_up_simple_subquery(PlannerInfo *root, Node *jtnode, RangeTblEntry *rte, preprocess_function_rtes(subroot); /* - * Scan the rangetable for relations with virtual generated columns, and - * replace all Var nodes in the query that reference these columns with - * the generation expressions. - */ - subquery = subroot->parse = expand_virtual_generated_columns(subroot); - - /* * Recursively pull up the subquery's subqueries, so that * pull_up_subqueries' processing is complete for its jointree and * rangetable. diff --git a/src/backend/optimizer/util/clauses.c b/src/backend/optimizer/util/clauses.c index f45131c34c5..6f0b338d2cd 100644 --- a/src/backend/optimizer/util/clauses.c +++ b/src/backend/optimizer/util/clauses.c @@ -20,6 +20,7 @@ #include "postgres.h" #include "access/htup_details.h" +#include "catalog/pg_class.h" #include "catalog/pg_language.h" #include "catalog/pg_operator.h" #include "catalog/pg_proc.h" @@ -36,6 +37,7 @@ #include "optimizer/clauses.h" #include "optimizer/cost.h" #include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" #include "optimizer/plancat.h" #include "optimizer/planmain.h" #include "parser/analyze.h" @@ -43,6 +45,7 @@ #include "parser/parse_collate.h" #include "parser/parse_func.h" #include "parser/parse_oper.h" +#include "parser/parsetree.h" #include "rewrite/rewriteHandler.h" #include "rewrite/rewriteManip.h" #include "tcop/tcopprot.h" @@ -2242,7 +2245,8 @@ rowtype_field_matches(Oid rowtypeid, int fieldnum, * only operators and functions that are reasonable to try to execute. * * NOTE: "root" can be passed as NULL if the caller never wants to do any - * Param substitutions nor receive info about inlined functions. + * Param substitutions nor receive info about inlined functions nor reduce + * NullTest for Vars to constant true or constant false. * * NOTE: the planner assumes that this will always flatten nested AND and * OR clauses into N-argument form. See comments in prepqual.c. @@ -3544,6 +3548,31 @@ eval_const_expressions_mutator(Node *node, return makeBoolConst(result, false); } + if (!ntest->argisrow && arg && IsA(arg, Var) && context->root) + { + Var *varg = (Var *) arg; + bool result; + + if (var_is_nonnullable(context->root, varg, false)) + { + switch (ntest->nulltesttype) + { + case IS_NULL: + result = false; + break; + case IS_NOT_NULL: + result = true; + break; + default: + elog(ERROR, "unrecognized nulltesttype: %d", + (int) ntest->nulltesttype); + result = false; /* keep compiler quiet */ + break; + } + + return makeBoolConst(result, false); + } + } newntest = makeNode(NullTest); newntest->arg = (Expr *) arg; @@ -4163,6 +4192,67 @@ simplify_function(Oid funcid, Oid result_type, int32 result_typmod, } /* + * var_is_nonnullable: check to see if the Var cannot be NULL + * + * If the Var is defined NOT NULL and meanwhile is not nulled by any outer + * joins or grouping sets, then we can know that it cannot be NULL. + * + * use_rel_info indicates whether the corresponding RelOptInfo is available for + * use. + */ +bool +var_is_nonnullable(PlannerInfo *root, Var *var, bool use_rel_info) +{ + Relids notnullattnums = NULL; + + Assert(IsA(var, Var)); + + /* skip upper-level Vars */ + if (var->varlevelsup != 0) + return false; + + /* could the Var be nulled by any outer joins or grouping sets? */ + if (!bms_is_empty(var->varnullingrels)) + return false; + + /* system columns cannot be NULL */ + if (var->varattno < 0) + return true; + + /* + * Check if the Var is defined as NOT NULL. We retrieve the column NOT + * NULL constraint information from the corresponding RelOptInfo if it is + * available; otherwise, we search the hash table for this information. + */ + if (use_rel_info) + { + RelOptInfo *rel = find_base_rel(root, var->varno); + + notnullattnums = rel->notnullattnums; + } + else + { + RangeTblEntry *rte = planner_rt_fetch(var->varno, root); + + /* + * We must skip inheritance parent tables, as some child tables may + * have a NOT NULL constraint for a column while others may not. This + * cannot happen with partitioned tables, though. + */ + if (rte->inh && rte->relkind != RELKIND_PARTITIONED_TABLE) + return false; + + notnullattnums = find_relation_notnullatts(root, rte->relid); + } + + if (var->varattno > 0 && + bms_is_member(var->varattno, notnullattnums)) + return true; + + return false; +} + +/* * expand_function_arguments: convert named-notation args to positional args * and/or insert default args, as needed * diff --git a/src/backend/optimizer/util/inherit.c b/src/backend/optimizer/util/inherit.c index 17e51cd75d7..30d158069e3 100644 --- a/src/backend/optimizer/util/inherit.c +++ b/src/backend/optimizer/util/inherit.c @@ -466,8 +466,7 @@ expand_single_inheritance_child(PlannerInfo *root, RangeTblEntry *parentrte, Index *childRTindex_p) { Query *parse = root->parse; - Oid parentOID PG_USED_FOR_ASSERTS_ONLY = - RelationGetRelid(parentrel); + Oid parentOID = RelationGetRelid(parentrel); Oid childOID = RelationGetRelid(childrel); RangeTblEntry *childrte; Index childRTindex; @@ -514,6 +513,13 @@ expand_single_inheritance_child(PlannerInfo *root, RangeTblEntry *parentrte, *childRTindex_p = childRTindex; /* + * Retrieve column not-null constraint information for the child relation + * if its relation OID is different from the parent's. + */ + if (childOID != parentOID) + get_relation_notnullatts(root, childrel); + + /* * Build an AppendRelInfo struct for each parent/child pair. */ appinfo = make_append_rel_info(parentrel, childrel, diff --git a/src/backend/optimizer/util/pathnode.c b/src/backend/optimizer/util/pathnode.c index 9cc602788ea..a4c5867cdcb 100644 --- a/src/backend/optimizer/util/pathnode.c +++ b/src/backend/optimizer/util/pathnode.c @@ -1689,7 +1689,7 @@ create_material_path(RelOptInfo *rel, Path *subpath) MemoizePath * create_memoize_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, List *param_exprs, List *hash_operators, - bool singlerow, bool binary_mode, double calls) + bool singlerow, bool binary_mode, Cardinality est_calls) { MemoizePath *pathnode = makeNode(MemoizePath); @@ -1710,7 +1710,6 @@ create_memoize_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, pathnode->param_exprs = param_exprs; pathnode->singlerow = singlerow; pathnode->binary_mode = binary_mode; - pathnode->calls = clamp_row_est(calls); /* * For now we set est_entries to 0. cost_memoize_rescan() does all the @@ -1720,6 +1719,12 @@ create_memoize_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, */ pathnode->est_entries = 0; + pathnode->est_calls = clamp_row_est(est_calls); + + /* These will also be set later in cost_memoize_rescan() */ + pathnode->est_unique_keys = 0.0; + pathnode->est_hit_ratio = 0.0; + /* we should not generate this path type when enable_memoize=false */ Assert(enable_memoize); pathnode->path.disabled_nodes = subpath->disabled_nodes; @@ -4259,7 +4264,7 @@ reparameterize_path(PlannerInfo *root, Path *path, mpath->hash_operators, mpath->singlerow, mpath->binary_mode, - mpath->calls); + mpath->est_calls); } default: break; diff --git a/src/backend/optimizer/util/plancat.c b/src/backend/optimizer/util/plancat.c index 59233b64730..c6a58afc5e5 100644 --- a/src/backend/optimizer/util/plancat.c +++ b/src/backend/optimizer/util/plancat.c @@ -59,6 +59,12 @@ int constraint_exclusion = CONSTRAINT_EXCLUSION_PARTITION; /* Hook for plugins to get control in get_relation_info() */ get_relation_info_hook_type get_relation_info_hook = NULL; +typedef struct NotnullHashEntry +{ + Oid relid; /* OID of the relation */ + Relids notnullattnums; /* attnums of NOT NULL columns */ +} NotnullHashEntry; + static void get_relation_foreign_keys(PlannerInfo *root, RelOptInfo *rel, Relation relation, bool inhparent); @@ -172,27 +178,7 @@ get_relation_info(PlannerInfo *root, Oid relationObjectId, bool inhparent, * RangeTblEntry does get populated. */ if (!inhparent || relation->rd_rel->relkind == RELKIND_PARTITIONED_TABLE) - { - for (int i = 0; i < relation->rd_att->natts; i++) - { - CompactAttribute *attr = TupleDescCompactAttr(relation->rd_att, i); - - Assert(attr->attnullability != ATTNULLABLE_UNKNOWN); - - if (attr->attnullability == ATTNULLABLE_VALID) - { - rel->notnullattnums = bms_add_member(rel->notnullattnums, - i + 1); - - /* - * Per RemoveAttributeById(), dropped columns will have their - * attnotnull unset, so we needn't check for dropped columns - * in the above condition. - */ - Assert(!attr->attisdropped); - } - } - } + rel->notnullattnums = find_relation_notnullatts(root, relationObjectId); /* * Estimate relation size --- unless it's an inheritance parent, in which @@ -684,6 +670,105 @@ get_relation_foreign_keys(PlannerInfo *root, RelOptInfo *rel, } /* + * get_relation_notnullatts - + * Retrieves column not-null constraint information for a given relation. + * + * We do this while we have the relcache entry open, and store the column + * not-null constraint information in a hash table based on the relation OID. + */ +void +get_relation_notnullatts(PlannerInfo *root, Relation relation) +{ + Oid relid = RelationGetRelid(relation); + NotnullHashEntry *hentry; + bool found; + Relids notnullattnums = NULL; + + /* bail out if the relation has no not-null constraints */ + if (relation->rd_att->constr == NULL || + !relation->rd_att->constr->has_not_null) + return; + + /* create the hash table if it hasn't been created yet */ + if (root->glob->rel_notnullatts_hash == NULL) + { + HTAB *hashtab; + HASHCTL hash_ctl; + + hash_ctl.keysize = sizeof(Oid); + hash_ctl.entrysize = sizeof(NotnullHashEntry); + hash_ctl.hcxt = CurrentMemoryContext; + + hashtab = hash_create("Relation NOT NULL attnums", + 64L, /* arbitrary initial size */ + &hash_ctl, + HASH_ELEM | HASH_BLOBS | HASH_CONTEXT); + + root->glob->rel_notnullatts_hash = hashtab; + } + + /* + * Create a hash entry for this relation OID, if we don't have one + * already. + */ + hentry = (NotnullHashEntry *) hash_search(root->glob->rel_notnullatts_hash, + &relid, + HASH_ENTER, + &found); + + /* bail out if a hash entry already exists for this relation OID */ + if (found) + return; + + /* collect the column not-null constraint information for this relation */ + for (int i = 0; i < relation->rd_att->natts; i++) + { + CompactAttribute *attr = TupleDescCompactAttr(relation->rd_att, i); + + Assert(attr->attnullability != ATTNULLABLE_UNKNOWN); + + if (attr->attnullability == ATTNULLABLE_VALID) + { + notnullattnums = bms_add_member(notnullattnums, i + 1); + + /* + * Per RemoveAttributeById(), dropped columns will have their + * attnotnull unset, so we needn't check for dropped columns in + * the above condition. + */ + Assert(!attr->attisdropped); + } + } + + /* ... and initialize the new hash entry */ + hentry->notnullattnums = notnullattnums; +} + +/* + * find_relation_notnullatts - + * Searches the hash table and returns the column not-null constraint + * information for a given relation. + */ +Relids +find_relation_notnullatts(PlannerInfo *root, Oid relid) +{ + NotnullHashEntry *hentry; + bool found; + + if (root->glob->rel_notnullatts_hash == NULL) + return NULL; + + hentry = (NotnullHashEntry *) hash_search(root->glob->rel_notnullatts_hash, + &relid, + HASH_FIND, + &found); + if (!found) + return NULL; + + return hentry->notnullattnums; +} + +/* * infer_arbiter_indexes - * Determine the unique indexes used to arbitrate speculative insertion. * diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y index 73345bb3c70..db43034b9db 100644 --- a/src/backend/parser/gram.y +++ b/src/backend/parser/gram.y @@ -318,6 +318,11 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query); %type <list> opt_qualified_name %type <boolean> opt_concurrently %type <dbehavior> opt_drop_behavior +%type <list> opt_utility_option_list +%type <list> utility_option_list +%type <defelt> utility_option_elem +%type <str> utility_option_name +%type <node> utility_option_arg %type <node> alter_column_default opclass_item opclass_drop alter_using %type <ival> add_drop opt_asc_desc opt_nulls_order @@ -338,10 +343,6 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query); create_extension_opt_item alter_extension_opt_item %type <ival> opt_lock lock_type cast_context -%type <str> utility_option_name -%type <defelt> utility_option_elem -%type <list> utility_option_list -%type <node> utility_option_arg %type <defelt> drop_option %type <boolean> opt_or_replace opt_no opt_grant_grant_option @@ -556,7 +557,6 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query); %type <list> generic_option_list alter_generic_option_list %type <ival> reindex_target_relation reindex_target_all -%type <list> opt_reindex_option_list %type <node> copy_generic_opt_arg copy_generic_opt_arg_list_item %type <defelt> copy_generic_opt_elem @@ -1141,6 +1141,41 @@ opt_drop_behavior: | /* EMPTY */ { $$ = DROP_RESTRICT; /* default */ } ; +opt_utility_option_list: + '(' utility_option_list ')' { $$ = $2; } + | /* EMPTY */ { $$ = NULL; } + ; + +utility_option_list: + utility_option_elem + { + $$ = list_make1($1); + } + | utility_option_list ',' utility_option_elem + { + $$ = lappend($1, $3); + } + ; + +utility_option_elem: + utility_option_name utility_option_arg + { + $$ = makeDefElem($1, $2, @1); + } + ; + +utility_option_name: + NonReservedWord { $$ = $1; } + | analyze_keyword { $$ = "analyze"; } + | FORMAT_LA { $$ = "format"; } + ; + +utility_option_arg: + opt_boolean_or_string { $$ = (Node *) makeString($1); } + | NumericOnly { $$ = (Node *) $1; } + | /* EMPTY */ { $$ = NULL; } + ; + /***************************************************************************** * * CALL statement @@ -2028,18 +2063,12 @@ constraints_set_mode: * Checkpoint statement */ CheckPointStmt: - CHECKPOINT + CHECKPOINT opt_utility_option_list { CheckPointStmt *n = makeNode(CheckPointStmt); $$ = (Node *) n; - } - | CHECKPOINT '(' utility_option_list ')' - { - CheckPointStmt *n = makeNode(CheckPointStmt); - - $$ = (Node *) n; - n->options = $3; + n->options = $2; } ; @@ -9354,7 +9383,7 @@ DropTransformStmt: DROP TRANSFORM opt_if_exists FOR Typename LANGUAGE name opt_d *****************************************************************************/ ReindexStmt: - REINDEX opt_reindex_option_list reindex_target_relation opt_concurrently qualified_name + REINDEX opt_utility_option_list reindex_target_relation opt_concurrently qualified_name { ReindexStmt *n = makeNode(ReindexStmt); @@ -9367,7 +9396,7 @@ ReindexStmt: makeDefElem("concurrently", NULL, @4)); $$ = (Node *) n; } - | REINDEX opt_reindex_option_list SCHEMA opt_concurrently name + | REINDEX opt_utility_option_list SCHEMA opt_concurrently name { ReindexStmt *n = makeNode(ReindexStmt); @@ -9380,7 +9409,7 @@ ReindexStmt: makeDefElem("concurrently", NULL, @4)); $$ = (Node *) n; } - | REINDEX opt_reindex_option_list reindex_target_all opt_concurrently opt_single_name + | REINDEX opt_utility_option_list reindex_target_all opt_concurrently opt_single_name { ReindexStmt *n = makeNode(ReindexStmt); @@ -9402,10 +9431,6 @@ reindex_target_all: SYSTEM_P { $$ = REINDEX_OBJECT_SYSTEM; } | DATABASE { $$ = REINDEX_OBJECT_DATABASE; } ; -opt_reindex_option_list: - '(' utility_option_list ')' { $$ = $2; } - | /* EMPTY */ { $$ = NULL; } - ; /***************************************************************************** * @@ -11903,13 +11928,13 @@ ClusterStmt: n->params = $3; $$ = (Node *) n; } - | CLUSTER '(' utility_option_list ')' + | CLUSTER opt_utility_option_list { ClusterStmt *n = makeNode(ClusterStmt); n->relation = NULL; n->indexname = NULL; - n->params = $3; + n->params = $2; $$ = (Node *) n; } /* unparenthesized VERBOSE kept for pre-14 compatibility */ @@ -11919,21 +11944,18 @@ ClusterStmt: n->relation = $3; n->indexname = $4; - n->params = NIL; if ($2) - n->params = lappend(n->params, makeDefElem("verbose", NULL, @2)); + n->params = list_make1(makeDefElem("verbose", NULL, @2)); $$ = (Node *) n; } /* unparenthesized VERBOSE kept for pre-17 compatibility */ - | CLUSTER opt_verbose + | CLUSTER VERBOSE { ClusterStmt *n = makeNode(ClusterStmt); n->relation = NULL; n->indexname = NULL; - n->params = NIL; - if ($2) - n->params = lappend(n->params, makeDefElem("verbose", NULL, @2)); + n->params = list_make1(makeDefElem("verbose", NULL, @2)); $$ = (Node *) n; } /* kept for pre-8.3 compatibility */ @@ -11943,9 +11965,8 @@ ClusterStmt: n->relation = $5; n->indexname = $3; - n->params = NIL; if ($2) - n->params = lappend(n->params, makeDefElem("verbose", NULL, @2)); + n->params = list_make1(makeDefElem("verbose", NULL, @2)); $$ = (Node *) n; } ; @@ -11996,64 +12017,31 @@ VacuumStmt: VACUUM opt_full opt_freeze opt_verbose opt_analyze opt_vacuum_relati } ; -AnalyzeStmt: analyze_keyword opt_verbose opt_vacuum_relation_list +AnalyzeStmt: analyze_keyword opt_utility_option_list opt_vacuum_relation_list { VacuumStmt *n = makeNode(VacuumStmt); - n->options = NIL; - if ($2) - n->options = lappend(n->options, - makeDefElem("verbose", NULL, @2)); + n->options = $2; n->rels = $3; n->is_vacuumcmd = false; $$ = (Node *) n; } - | analyze_keyword '(' utility_option_list ')' opt_vacuum_relation_list + | analyze_keyword VERBOSE opt_vacuum_relation_list { VacuumStmt *n = makeNode(VacuumStmt); - n->options = $3; - n->rels = $5; + n->options = list_make1(makeDefElem("verbose", NULL, @2)); + n->rels = $3; n->is_vacuumcmd = false; $$ = (Node *) n; } ; -utility_option_list: - utility_option_elem - { - $$ = list_make1($1); - } - | utility_option_list ',' utility_option_elem - { - $$ = lappend($1, $3); - } - ; - analyze_keyword: ANALYZE | ANALYSE /* British */ ; -utility_option_elem: - utility_option_name utility_option_arg - { - $$ = makeDefElem($1, $2, @1); - } - ; - -utility_option_name: - NonReservedWord { $$ = $1; } - | analyze_keyword { $$ = "analyze"; } - | FORMAT_LA { $$ = "format"; } - ; - -utility_option_arg: - opt_boolean_or_string { $$ = (Node *) makeString($1); } - | NumericOnly { $$ = (Node *) $1; } - | /* EMPTY */ { $$ = NULL; } - ; - opt_analyze: analyze_keyword { $$ = true; } | /*EMPTY*/ { $$ = false; } diff --git a/src/backend/partitioning/partbounds.c b/src/backend/partitioning/partbounds.c index 4bdc2941efb..822cf4ec451 100644 --- a/src/backend/partitioning/partbounds.c +++ b/src/backend/partitioning/partbounds.c @@ -1007,9 +1007,6 @@ partition_bounds_copy(PartitionBoundInfo src, int ndatums; int nindexes; int partnatts; - bool hash_part; - int natts; - Datum *boundDatums; dest = (PartitionBoundInfo) palloc(sizeof(PartitionBoundInfoData)); @@ -1023,7 +1020,7 @@ partition_bounds_copy(PartitionBoundInfo src, dest->datums = (Datum **) palloc(sizeof(Datum *) * ndatums); - if (src->kind != NULL) + if (src->kind != NULL && ndatums > 0) { PartitionRangeDatumKind *boundKinds; @@ -1058,36 +1055,40 @@ partition_bounds_copy(PartitionBoundInfo src, * For hash partitioning, datums array will have two elements - modulus * and remainder. */ - hash_part = (key->strategy == PARTITION_STRATEGY_HASH); - natts = hash_part ? 2 : partnatts; - boundDatums = palloc(ndatums * natts * sizeof(Datum)); - - for (i = 0; i < ndatums; i++) + if (ndatums > 0) { - int j; - - dest->datums[i] = &boundDatums[i * natts]; + bool hash_part = (key->strategy == PARTITION_STRATEGY_HASH); + int natts = hash_part ? 2 : partnatts; + Datum *boundDatums = palloc(ndatums * natts * sizeof(Datum)); - for (j = 0; j < natts; j++) + for (i = 0; i < ndatums; i++) { - bool byval; - int typlen; + int j; - if (hash_part) - { - typlen = sizeof(int32); /* Always int4 */ - byval = true; /* int4 is pass-by-value */ - } - else + dest->datums[i] = &boundDatums[i * natts]; + + for (j = 0; j < natts; j++) { - byval = key->parttypbyval[j]; - typlen = key->parttyplen[j]; - } + if (dest->kind == NULL || + dest->kind[i][j] == PARTITION_RANGE_DATUM_VALUE) + { + bool byval; + int typlen; - if (dest->kind == NULL || - dest->kind[i][j] == PARTITION_RANGE_DATUM_VALUE) - dest->datums[i][j] = datumCopy(src->datums[i][j], - byval, typlen); + if (hash_part) + { + typlen = sizeof(int32); /* Always int4 */ + byval = true; /* int4 is pass-by-value */ + } + else + { + byval = key->parttypbyval[j]; + typlen = key->parttyplen[j]; + } + dest->datums[i][j] = datumCopy(src->datums[i][j], + byval, typlen); + } + } } } diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c index 9474095f271..ff96b36d710 100644 --- a/src/backend/postmaster/autovacuum.c +++ b/src/backend/postmaster/autovacuum.c @@ -310,6 +310,16 @@ static AutoVacuumShmemStruct *AutoVacuumShmem; static dlist_head DatabaseList = DLIST_STATIC_INIT(DatabaseList); static MemoryContext DatabaseListCxt = NULL; +/* + * Dummy pointer to persuade Valgrind that we've not leaked the array of + * avl_dbase structs. Make it global to ensure the compiler doesn't + * optimize it away. + */ +#ifdef USE_VALGRIND +extern avl_dbase *avl_dbase_array; +avl_dbase *avl_dbase_array; +#endif + /* Pointer to my own WorkerInfo, valid on each worker */ static WorkerInfo MyWorkerInfo = NULL; @@ -562,10 +572,10 @@ AutoVacLauncherMain(const void *startup_data, size_t startup_data_len) /* * Create the initial database list. The invariant we want this list to - * keep is that it's ordered by decreasing next_time. As soon as an entry - * is updated to a higher time, it will be moved to the front (which is - * correct because the only operation is to add autovacuum_naptime to the - * entry, and time always increases). + * keep is that it's ordered by decreasing next_worker. As soon as an + * entry is updated to a higher time, it will be moved to the front (which + * is correct because the only operation is to add autovacuum_naptime to + * the entry, and time always increases). */ rebuild_database_list(InvalidOid); @@ -1020,6 +1030,10 @@ rebuild_database_list(Oid newdb) /* put all the hash elements into an array */ dbary = palloc(nelems * sizeof(avl_dbase)); + /* keep Valgrind quiet */ +#ifdef USE_VALGRIND + avl_dbase_array = dbary; +#endif i = 0; hash_seq_init(&seq, dbhash); @@ -2565,8 +2579,18 @@ deleted: /* * We leak table_toast_map here (among other things), but since we're - * going away soon, it's not a problem. + * going away soon, it's not a problem normally. But when using Valgrind, + * release some stuff to reduce complaints about leaked storage. */ +#ifdef USE_VALGRIND + hash_destroy(table_toast_map); + FreeTupleDesc(pg_class_desc); + if (bstrategy) + pfree(bstrategy); +#endif + + /* Run the rest in xact context, mainly to avoid Valgrind leak warnings */ + MemoryContextSwitchTo(TopTransactionContext); /* * Update pg_database.datfrozenxid, and truncate pg_xact if possible. We diff --git a/src/backend/postmaster/bgworker.c b/src/backend/postmaster/bgworker.c index 116ddf7b835..1ad65c237c3 100644 --- a/src/backend/postmaster/bgworker.c +++ b/src/backend/postmaster/bgworker.c @@ -613,6 +613,7 @@ ResetBackgroundWorkerCrashTimes(void) * resetting. */ rw->rw_crashed_at = 0; + rw->rw_pid = 0; /* * If there was anyone waiting for it, they're history. diff --git a/src/backend/postmaster/checkpointer.c b/src/backend/postmaster/checkpointer.c index 2809e298a44..8490148a47d 100644 --- a/src/backend/postmaster/checkpointer.c +++ b/src/backend/postmaster/checkpointer.c @@ -130,6 +130,13 @@ typedef struct int num_requests; /* current # of requests */ int max_requests; /* allocated array size */ + + int head; /* Index of the first request in the ring + * buffer */ + int tail; /* Index of the last request in the ring + * buffer */ + + /* The ring buffer of pending checkpointer requests */ CheckpointerRequest requests[FLEXIBLE_ARRAY_MEMBER]; } CheckpointerShmemStruct; @@ -138,6 +145,12 @@ static CheckpointerShmemStruct *CheckpointerShmem; /* interval for calling AbsorbSyncRequests in CheckpointWriteDelay */ #define WRITES_PER_ABSORB 1000 +/* Maximum number of checkpointer requests to process in one batch */ +#define CKPT_REQ_BATCH_SIZE 10000 + +/* Max number of requests the checkpointer request queue can hold */ +#define MAX_CHECKPOINT_REQUESTS 10000000 + /* * GUC parameters */ @@ -973,7 +986,8 @@ CheckpointerShmemInit(void) */ MemSet(CheckpointerShmem, 0, size); SpinLockInit(&CheckpointerShmem->ckpt_lck); - CheckpointerShmem->max_requests = NBuffers; + CheckpointerShmem->max_requests = Min(NBuffers, MAX_CHECKPOINT_REQUESTS); + CheckpointerShmem->head = CheckpointerShmem->tail = 0; ConditionVariableInit(&CheckpointerShmem->start_cv); ConditionVariableInit(&CheckpointerShmem->done_cv); } @@ -1201,6 +1215,7 @@ ForwardSyncRequest(const FileTag *ftag, SyncRequestType type) { CheckpointerRequest *request; bool too_full; + int insert_pos; if (!IsUnderPostmaster) return false; /* probably shouldn't even get here */ @@ -1224,10 +1239,14 @@ ForwardSyncRequest(const FileTag *ftag, SyncRequestType type) } /* OK, insert request */ - request = &CheckpointerShmem->requests[CheckpointerShmem->num_requests++]; + insert_pos = CheckpointerShmem->tail; + request = &CheckpointerShmem->requests[insert_pos]; request->ftag = *ftag; request->type = type; + CheckpointerShmem->tail = (CheckpointerShmem->tail + 1) % CheckpointerShmem->max_requests; + CheckpointerShmem->num_requests++; + /* If queue is more than half full, nudge the checkpointer to empty it */ too_full = (CheckpointerShmem->num_requests >= CheckpointerShmem->max_requests / 2); @@ -1269,12 +1288,16 @@ CompactCheckpointerRequestQueue(void) struct CheckpointerSlotMapping { CheckpointerRequest request; - int slot; + int ring_idx; }; - int n, - preserve_count; + int n; int num_skipped = 0; + int head; + int max_requests; + int num_requests; + int read_idx, + write_idx; HASHCTL ctl; HTAB *htab; bool *skip_slot; @@ -1286,8 +1309,13 @@ CompactCheckpointerRequestQueue(void) if (CritSectionCount > 0) return false; + max_requests = CheckpointerShmem->max_requests; + num_requests = CheckpointerShmem->num_requests; + /* Initialize skip_slot array */ - skip_slot = palloc0(sizeof(bool) * CheckpointerShmem->num_requests); + skip_slot = palloc0(sizeof(bool) * max_requests); + + head = CheckpointerShmem->head; /* Initialize temporary hash table */ ctl.keysize = sizeof(CheckpointerRequest); @@ -1311,7 +1339,8 @@ CompactCheckpointerRequestQueue(void) * away preceding entries that would end up being canceled anyhow), but * it's not clear that the extra complexity would buy us anything. */ - for (n = 0; n < CheckpointerShmem->num_requests; n++) + read_idx = head; + for (n = 0; n < num_requests; n++) { CheckpointerRequest *request; struct CheckpointerSlotMapping *slotmap; @@ -1324,16 +1353,19 @@ CompactCheckpointerRequestQueue(void) * CheckpointerShmemInit. Note also that RelFileLocator had better * contain no pad bytes. */ - request = &CheckpointerShmem->requests[n]; + request = &CheckpointerShmem->requests[read_idx]; slotmap = hash_search(htab, request, HASH_ENTER, &found); if (found) { /* Duplicate, so mark the previous occurrence as skippable */ - skip_slot[slotmap->slot] = true; + skip_slot[slotmap->ring_idx] = true; num_skipped++; } /* Remember slot containing latest occurrence of this request value */ - slotmap->slot = n; + slotmap->ring_idx = read_idx; + + /* Move to the next request in the ring buffer */ + read_idx = (read_idx + 1) % max_requests; } /* Done with the hash table. */ @@ -1347,17 +1379,34 @@ CompactCheckpointerRequestQueue(void) } /* We found some duplicates; remove them. */ - preserve_count = 0; - for (n = 0; n < CheckpointerShmem->num_requests; n++) + read_idx = write_idx = head; + for (n = 0; n < num_requests; n++) { - if (skip_slot[n]) - continue; - CheckpointerShmem->requests[preserve_count++] = CheckpointerShmem->requests[n]; + /* If this slot is NOT skipped, keep it */ + if (!skip_slot[read_idx]) + { + /* If the read and write positions are different, copy the request */ + if (write_idx != read_idx) + CheckpointerShmem->requests[write_idx] = + CheckpointerShmem->requests[read_idx]; + + /* Advance the write position */ + write_idx = (write_idx + 1) % max_requests; + } + + read_idx = (read_idx + 1) % max_requests; } + + /* + * Update ring buffer state: head remains the same, tail moves, count + * decreases + */ + CheckpointerShmem->tail = write_idx; + CheckpointerShmem->num_requests -= num_skipped; + ereport(DEBUG1, (errmsg_internal("compacted fsync request queue from %d entries to %d entries", - CheckpointerShmem->num_requests, preserve_count))); - CheckpointerShmem->num_requests = preserve_count; + num_requests, CheckpointerShmem->num_requests))); /* Cleanup. */ pfree(skip_slot); @@ -1378,40 +1427,64 @@ AbsorbSyncRequests(void) { CheckpointerRequest *requests = NULL; CheckpointerRequest *request; - int n; + int n, + i; + bool loop; if (!AmCheckpointerProcess()) return; - LWLockAcquire(CheckpointerCommLock, LW_EXCLUSIVE); - - /* - * We try to avoid holding the lock for a long time by copying the request - * array, and processing the requests after releasing the lock. - * - * Once we have cleared the requests from shared memory, we have to PANIC - * if we then fail to absorb them (eg, because our hashtable runs out of - * memory). This is because the system cannot run safely if we are unable - * to fsync what we have been told to fsync. Fortunately, the hashtable - * is so small that the problem is quite unlikely to arise in practice. - */ - n = CheckpointerShmem->num_requests; - if (n > 0) + do { - requests = (CheckpointerRequest *) palloc(n * sizeof(CheckpointerRequest)); - memcpy(requests, CheckpointerShmem->requests, n * sizeof(CheckpointerRequest)); - } + LWLockAcquire(CheckpointerCommLock, LW_EXCLUSIVE); + + /*--- + * We try to avoid holding the lock for a long time by: + * 1. Copying the request array and processing the requests after + * releasing the lock; + * 2. Processing not the whole queue, but only batches of + * CKPT_REQ_BATCH_SIZE at once. + * + * Once we have cleared the requests from shared memory, we must + * PANIC if we then fail to absorb them (e.g., because our hashtable + * runs out of memory). This is because the system cannot run safely + * if we are unable to fsync what we have been told to fsync. + * Fortunately, the hashtable is so small that the problem is quite + * unlikely to arise in practice. + * + * Note: The maximum possible size of a ring buffer is + * MAX_CHECKPOINT_REQUESTS entries, which fit into a maximum palloc + * allocation size of 1Gb. Our maximum batch size, + * CKPT_REQ_BATCH_SIZE, is even smaller. + */ + n = Min(CheckpointerShmem->num_requests, CKPT_REQ_BATCH_SIZE); + if (n > 0) + { + if (!requests) + requests = (CheckpointerRequest *) palloc(n * sizeof(CheckpointerRequest)); - START_CRIT_SECTION(); + for (i = 0; i < n; i++) + { + requests[i] = CheckpointerShmem->requests[CheckpointerShmem->head]; + CheckpointerShmem->head = (CheckpointerShmem->head + 1) % CheckpointerShmem->max_requests; + } - CheckpointerShmem->num_requests = 0; + CheckpointerShmem->num_requests -= n; - LWLockRelease(CheckpointerCommLock); + } + + START_CRIT_SECTION(); + + /* Are there any requests in the queue? If so, keep going. */ + loop = CheckpointerShmem->num_requests != 0; + + LWLockRelease(CheckpointerCommLock); - for (request = requests; n > 0; request++, n--) - RememberSyncRequest(&request->ftag, request->type); + for (request = requests; n > 0; request++, n--) + RememberSyncRequest(&request->ftag, request->type); - END_CRIT_SECTION(); + END_CRIT_SECTION(); + } while (loop); if (requests) pfree(requests); diff --git a/src/backend/postmaster/pmchild.c b/src/backend/postmaster/pmchild.c index cde1d23a4ca..584bb58c8ab 100644 --- a/src/backend/postmaster/pmchild.c +++ b/src/backend/postmaster/pmchild.c @@ -60,6 +60,17 @@ NON_EXEC_STATIC int num_pmchild_slots = 0; dlist_head ActiveChildList; /* + * Dummy pointer to persuade Valgrind that we've not leaked the array of + * PMChild structs. Make it global to ensure the compiler doesn't + * optimize it away. + */ +#ifdef USE_VALGRIND +extern PMChild *pmchild_array; +PMChild *pmchild_array; +#endif + + +/* * MaxLivePostmasterChildren * * This reports the number of postmaster child processes that can be active. @@ -125,8 +136,13 @@ InitPostmasterChildSlots(void) for (int i = 0; i < BACKEND_NUM_TYPES; i++) num_pmchild_slots += pmchild_pools[i].size; - /* Initialize them */ + /* Allocate enough slots, and make sure Valgrind doesn't complain */ slots = palloc(num_pmchild_slots * sizeof(PMChild)); +#ifdef USE_VALGRIND + pmchild_array = slots; +#endif + + /* Initialize them */ slotno = 0; for (int btype = 0; btype < BACKEND_NUM_TYPES; btype++) { diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c index cca9b946e53..e01d9f0cfe8 100644 --- a/src/backend/postmaster/postmaster.c +++ b/src/backend/postmaster/postmaster.c @@ -2630,6 +2630,13 @@ CleanupBackend(PMChild *bp, } bp = NULL; + /* + * In a crash case, exit immediately without resetting background worker + * state. However, if restart_after_crash is enabled, the background + * worker state (e.g., rw_pid) still needs be reset so the worker can + * restart after crash recovery. This reset is handled in + * ResetBackgroundWorkerCrashTimes(), not here. + */ if (crashed) { HandleChildCrash(bp_pid, exitstatus, procname); diff --git a/src/backend/replication/libpqwalreceiver/libpqwalreceiver.c b/src/backend/replication/libpqwalreceiver/libpqwalreceiver.c index f7b5d093681..239641bfbb6 100644 --- a/src/backend/replication/libpqwalreceiver/libpqwalreceiver.c +++ b/src/backend/replication/libpqwalreceiver/libpqwalreceiver.c @@ -232,6 +232,9 @@ libpqrcv_connect(const char *conninfo, bool replication, bool logical, errhint("Target server's authentication method must be changed, or set password_required=false in the subscription parameters."))); } + PQsetNoticeReceiver(conn->streamConn, libpqsrv_notice_receiver, + "received message via replication"); + /* * Set always-secure search path for the cases where the connection is * used to run SQL queries, so malicious users can't get control. @@ -418,31 +421,22 @@ libpqrcv_identify_system(WalReceiverConn *conn, TimeLineID *primary_tli) "IDENTIFY_SYSTEM", WAIT_EVENT_LIBPQWALRECEIVER_RECEIVE); if (PQresultStatus(res) != PGRES_TUPLES_OK) - { - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("could not receive database system identifier and timeline ID from " "the primary server: %s", pchomp(PQerrorMessage(conn->streamConn))))); - } /* * IDENTIFY_SYSTEM returns 3 columns in 9.3 and earlier, and 4 columns in * 9.4 and onwards. */ if (PQnfields(res) < 3 || PQntuples(res) != 1) - { - int ntuples = PQntuples(res); - int nfields = PQnfields(res); - - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("invalid response from primary server"), errdetail("Could not identify system: got %d rows and %d fields, expected %d rows and %d or more fields.", - ntuples, nfields, 1, 3))); - } + PQntuples(res), PQnfields(res), 1, 3))); primary_sysid = pstrdup(PQgetvalue(res, 0, 0)); *primary_tli = pg_strtoint32(PQgetvalue(res, 0, 1)); PQclear(res); @@ -604,13 +598,10 @@ libpqrcv_startstreaming(WalReceiverConn *conn, return false; } else if (PQresultStatus(res) != PGRES_COPY_BOTH) - { - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("could not start WAL streaming: %s", pchomp(PQerrorMessage(conn->streamConn))))); - } PQclear(res); return true; } @@ -718,26 +709,17 @@ libpqrcv_readtimelinehistoryfile(WalReceiverConn *conn, cmd, WAIT_EVENT_LIBPQWALRECEIVER_RECEIVE); if (PQresultStatus(res) != PGRES_TUPLES_OK) - { - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("could not receive timeline history file from " "the primary server: %s", pchomp(PQerrorMessage(conn->streamConn))))); - } if (PQnfields(res) != 2 || PQntuples(res) != 1) - { - int ntuples = PQntuples(res); - int nfields = PQnfields(res); - - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("invalid response from primary server"), errdetail("Expected 1 tuple with 2 fields, got %d tuples with %d fields.", - ntuples, nfields))); - } + PQntuples(res), PQnfields(res)))); *filename = pstrdup(PQgetvalue(res, 0, 0)); *len = PQgetlength(res, 0, 1); @@ -841,13 +823,10 @@ libpqrcv_receive(WalReceiverConn *conn, char **buffer, return -1; } else - { - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("could not receive data from WAL stream: %s", pchomp(PQerrorMessage(conn->streamConn))))); - } } if (rawlen < -1) ereport(ERROR, @@ -971,13 +950,10 @@ libpqrcv_create_slot(WalReceiverConn *conn, const char *slotname, pfree(cmd.data); if (PQresultStatus(res) != PGRES_TUPLES_OK) - { - PQclear(res); ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("could not create replication slot \"%s\": %s", slotname, pchomp(PQerrorMessage(conn->streamConn))))); - } if (lsn) *lsn = DatumGetLSN(DirectFunctionCall1Coll(pg_lsn_in, InvalidOid, diff --git a/src/backend/replication/logical/applyparallelworker.c b/src/backend/replication/logical/applyparallelworker.c index d25085d3515..1fa931a7422 100644 --- a/src/backend/replication/logical/applyparallelworker.c +++ b/src/backend/replication/logical/applyparallelworker.c @@ -441,7 +441,8 @@ pa_launch_parallel_worker(void) MySubscription->name, MyLogicalRepWorker->userid, InvalidOid, - dsm_segment_handle(winfo->dsm_seg)); + dsm_segment_handle(winfo->dsm_seg), + false); if (launched) { diff --git a/src/backend/replication/logical/launcher.c b/src/backend/replication/logical/launcher.c index 4aed0dfcebb..37377f7eb63 100644 --- a/src/backend/replication/logical/launcher.c +++ b/src/backend/replication/logical/launcher.c @@ -32,6 +32,7 @@ #include "postmaster/interrupt.h" #include "replication/logicallauncher.h" #include "replication/origin.h" +#include "replication/slot.h" #include "replication/walreceiver.h" #include "replication/worker_internal.h" #include "storage/ipc.h" @@ -91,7 +92,6 @@ static dshash_table *last_start_times = NULL; static bool on_commit_launcher_wakeup = false; -static void ApplyLauncherWakeup(void); static void logicalrep_launcher_onexit(int code, Datum arg); static void logicalrep_worker_onexit(int code, Datum arg); static void logicalrep_worker_detach(void); @@ -100,6 +100,9 @@ static int logicalrep_pa_worker_count(Oid subid); static void logicalrep_launcher_attach_dshmem(void); static void ApplyLauncherSetWorkerStartTime(Oid subid, TimestampTz start_time); static TimestampTz ApplyLauncherGetWorkerStartTime(Oid subid); +static void compute_min_nonremovable_xid(LogicalRepWorker *worker, TransactionId *xmin); +static bool acquire_conflict_slot_if_exists(void); +static void advance_conflict_slot_xmin(TransactionId new_xmin); /* @@ -148,6 +151,7 @@ get_subscription_list(void) sub->owner = subform->subowner; sub->enabled = subform->subenabled; sub->name = pstrdup(NameStr(subform->subname)); + sub->retaindeadtuples = subform->subretaindeadtuples; /* We don't fill fields we are not interested in. */ res = lappend(res, sub); @@ -309,7 +313,8 @@ logicalrep_workers_find(Oid subid, bool only_running, bool acquire_lock) bool logicalrep_worker_launch(LogicalRepWorkerType wtype, Oid dbid, Oid subid, const char *subname, Oid userid, - Oid relid, dsm_handle subworker_dsm) + Oid relid, dsm_handle subworker_dsm, + bool retain_dead_tuples) { BackgroundWorker bgw; BackgroundWorkerHandle *bgw_handle; @@ -328,10 +333,13 @@ logicalrep_worker_launch(LogicalRepWorkerType wtype, * - must be valid worker type * - tablesync workers are only ones to have relid * - parallel apply worker is the only kind of subworker + * - The replication slot used in conflict detection is created when + * retain_dead_tuples is enabled */ Assert(wtype != WORKERTYPE_UNKNOWN); Assert(is_tablesync_worker == OidIsValid(relid)); Assert(is_parallel_apply_worker == (subworker_dsm != DSM_HANDLE_INVALID)); + Assert(!retain_dead_tuples || MyReplicationSlot); ereport(DEBUG1, (errmsg_internal("starting logical replication worker for subscription \"%s\"", @@ -454,6 +462,9 @@ retry: worker->stream_fileset = NULL; worker->leader_pid = is_parallel_apply_worker ? MyProcPid : InvalidPid; worker->parallel_apply = is_parallel_apply_worker; + worker->oldest_nonremovable_xid = retain_dead_tuples + ? MyReplicationSlot->data.xmin + : InvalidTransactionId; worker->last_lsn = InvalidXLogRecPtr; TIMESTAMP_NOBEGIN(worker->last_send_time); TIMESTAMP_NOBEGIN(worker->last_recv_time); @@ -779,6 +790,8 @@ logicalrep_worker_detach(void) } LWLockRelease(LogicalRepWorkerLock); + + list_free(workers); } /* Block concurrent access. */ @@ -1118,7 +1131,10 @@ ApplyLauncherWakeupAtCommit(void) on_commit_launcher_wakeup = true; } -static void +/* + * Wakeup the launcher immediately. + */ +void ApplyLauncherWakeup(void) { if (LogicalRepCtx->launcher_pid != 0) @@ -1150,6 +1166,12 @@ ApplyLauncherMain(Datum main_arg) */ BackgroundWorkerInitializeConnection(NULL, NULL, 0); + /* + * Acquire the conflict detection slot at startup to ensure it can be + * dropped if no longer needed after a restart. + */ + acquire_conflict_slot_if_exists(); + /* Enter main loop */ for (;;) { @@ -1159,6 +1181,9 @@ ApplyLauncherMain(Datum main_arg) MemoryContext subctx; MemoryContext oldctx; long wait_time = DEFAULT_NAPTIME_PER_CYCLE; + bool can_advance_xmin = true; + bool retain_dead_tuples = false; + TransactionId xmin = InvalidTransactionId; CHECK_FOR_INTERRUPTS(); @@ -1168,7 +1193,14 @@ ApplyLauncherMain(Datum main_arg) ALLOCSET_DEFAULT_SIZES); oldctx = MemoryContextSwitchTo(subctx); - /* Start any missing workers for enabled subscriptions. */ + /* + * Start any missing workers for enabled subscriptions. + * + * Also, during the iteration through all subscriptions, we compute + * the minimum XID required to protect deleted tuples for conflict + * detection if one of the subscription enables retain_dead_tuples + * option. + */ sublist = get_subscription_list(); foreach(lc, sublist) { @@ -1178,6 +1210,38 @@ ApplyLauncherMain(Datum main_arg) TimestampTz now; long elapsed; + if (sub->retaindeadtuples) + { + retain_dead_tuples = true; + + /* + * Can't advance xmin of the slot unless all the subscriptions + * with retain_dead_tuples are enabled. This is required to + * ensure that we don't advance the xmin of + * CONFLICT_DETECTION_SLOT if one of the subscriptions is not + * enabled. Otherwise, we won't be able to detect conflicts + * reliably for such a subscription even though it has set the + * retain_dead_tuples option. + */ + can_advance_xmin &= sub->enabled; + + /* + * Create a replication slot to retain information necessary + * for conflict detection such as dead tuples, commit + * timestamps, and origins. + * + * The slot is created before starting the apply worker to + * prevent it from unnecessarily maintaining its + * oldest_nonremovable_xid. + * + * The slot is created even for a disabled subscription to + * ensure that conflict-related information is available when + * applying remote changes that occurred before the + * subscription was enabled. + */ + CreateConflictDetectionSlot(); + } + if (!sub->enabled) continue; @@ -1186,7 +1250,27 @@ ApplyLauncherMain(Datum main_arg) LWLockRelease(LogicalRepWorkerLock); if (w != NULL) - continue; /* worker is running already */ + { + /* + * Compute the minimum xmin required to protect dead tuples + * required for conflict detection among all running apply + * workers that enables retain_dead_tuples. + */ + if (sub->retaindeadtuples && can_advance_xmin) + compute_min_nonremovable_xid(w, &xmin); + + /* worker is running already */ + continue; + } + + /* + * Can't advance xmin of the slot unless all the workers + * corresponding to subscriptions with retain_dead_tuples are + * running, disabling the further computation of the minimum + * nonremovable xid. + */ + if (sub->retaindeadtuples) + can_advance_xmin = false; /* * If the worker is eligible to start now, launch it. Otherwise, @@ -1210,7 +1294,8 @@ ApplyLauncherMain(Datum main_arg) if (!logicalrep_worker_launch(WORKERTYPE_APPLY, sub->dbid, sub->oid, sub->name, sub->owner, InvalidOid, - DSM_HANDLE_INVALID)) + DSM_HANDLE_INVALID, + sub->retaindeadtuples)) { /* * We get here either if we failed to launch a worker @@ -1230,6 +1315,20 @@ ApplyLauncherMain(Datum main_arg) } } + /* + * Drop the CONFLICT_DETECTION_SLOT slot if there is no subscription + * that requires us to retain dead tuples. Otherwise, if required, + * advance the slot's xmin to protect dead tuples required for the + * conflict detection. + */ + if (MyReplicationSlot) + { + if (!retain_dead_tuples) + ReplicationSlotDropAcquired(); + else if (can_advance_xmin) + advance_conflict_slot_xmin(xmin); + } + /* Switch back to original memory context. */ MemoryContextSwitchTo(oldctx); /* Clean the temporary memory. */ @@ -1258,6 +1357,125 @@ ApplyLauncherMain(Datum main_arg) } /* + * Determine the minimum non-removable transaction ID across all apply workers + * for subscriptions that have retain_dead_tuples enabled. Store the result + * in *xmin. + */ +static void +compute_min_nonremovable_xid(LogicalRepWorker *worker, TransactionId *xmin) +{ + TransactionId nonremovable_xid; + + Assert(worker != NULL); + + /* + * The replication slot for conflict detection must be created before the + * worker starts. + */ + Assert(MyReplicationSlot); + + SpinLockAcquire(&worker->relmutex); + nonremovable_xid = worker->oldest_nonremovable_xid; + SpinLockRelease(&worker->relmutex); + + Assert(TransactionIdIsValid(nonremovable_xid)); + + if (!TransactionIdIsValid(*xmin) || + TransactionIdPrecedes(nonremovable_xid, *xmin)) + *xmin = nonremovable_xid; +} + +/* + * Acquire the replication slot used to retain information for conflict + * detection, if it exists. + * + * Return true if successfully acquired, otherwise return false. + */ +static bool +acquire_conflict_slot_if_exists(void) +{ + if (!SearchNamedReplicationSlot(CONFLICT_DETECTION_SLOT, true)) + return false; + + ReplicationSlotAcquire(CONFLICT_DETECTION_SLOT, true, false); + return true; +} + +/* + * Advance the xmin the replication slot used to retain information required + * for conflict detection. + */ +static void +advance_conflict_slot_xmin(TransactionId new_xmin) +{ + Assert(MyReplicationSlot); + Assert(TransactionIdIsValid(new_xmin)); + Assert(TransactionIdPrecedesOrEquals(MyReplicationSlot->data.xmin, new_xmin)); + + /* Return if the xmin value of the slot cannot be advanced */ + if (TransactionIdEquals(MyReplicationSlot->data.xmin, new_xmin)) + return; + + SpinLockAcquire(&MyReplicationSlot->mutex); + MyReplicationSlot->effective_xmin = new_xmin; + MyReplicationSlot->data.xmin = new_xmin; + SpinLockRelease(&MyReplicationSlot->mutex); + + elog(DEBUG1, "updated xmin: %u", MyReplicationSlot->data.xmin); + + ReplicationSlotMarkDirty(); + ReplicationSlotsComputeRequiredXmin(false); + + /* + * Like PhysicalConfirmReceivedLocation(), do not save slot information + * each time. This is acceptable because all concurrent transactions on + * the publisher that require the data preceding the slot's xmin should + * have already been applied and flushed on the subscriber before the xmin + * is advanced. So, even if the slot's xmin regresses after a restart, it + * will be advanced again in the next cycle. Therefore, no data required + * for conflict detection will be prematurely removed. + */ + return; +} + +/* + * Create and acquire the replication slot used to retain information for + * conflict detection, if not yet. + */ +void +CreateConflictDetectionSlot(void) +{ + TransactionId xmin_horizon; + + /* Exit early, if the replication slot is already created and acquired */ + if (MyReplicationSlot) + return; + + ereport(LOG, + errmsg("creating replication conflict detection slot")); + + ReplicationSlotCreate(CONFLICT_DETECTION_SLOT, false, RS_PERSISTENT, false, + false, false); + + LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE); + + xmin_horizon = GetOldestSafeDecodingTransactionId(false); + + SpinLockAcquire(&MyReplicationSlot->mutex); + MyReplicationSlot->effective_xmin = xmin_horizon; + MyReplicationSlot->data.xmin = xmin_horizon; + SpinLockRelease(&MyReplicationSlot->mutex); + + ReplicationSlotsComputeRequiredXmin(true); + + LWLockRelease(ProcArrayLock); + + /* Write this slot to disk */ + ReplicationSlotMarkDirty(); + ReplicationSlotSave(); +} + +/* * Is current process the logical replication launcher? */ bool diff --git a/src/backend/replication/logical/reorderbuffer.c b/src/backend/replication/logical/reorderbuffer.c index 7b4e8629553..34cf05668ae 100644 --- a/src/backend/replication/logical/reorderbuffer.c +++ b/src/backend/replication/logical/reorderbuffer.c @@ -2599,7 +2599,7 @@ ReorderBufferProcessTXN(ReorderBuffer *rb, ReorderBufferTXN *txn, if (++changes_count >= CHANGES_THRESHOLD) { - rb->update_progress_txn(rb, txn, change->lsn); + rb->update_progress_txn(rb, txn, prev_lsn); changes_count = 0; } } @@ -4917,7 +4917,7 @@ StartupReorderBuffer(void) continue; /* if it cannot be a slot, skip the directory */ - if (!ReplicationSlotValidateName(logical_de->d_name, DEBUG2)) + if (!ReplicationSlotValidateName(logical_de->d_name, true, DEBUG2)) continue; /* diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c index e4fd6347fd1..d3356bc84ee 100644 --- a/src/backend/replication/logical/tablesync.c +++ b/src/backend/replication/logical/tablesync.c @@ -316,7 +316,8 @@ process_syncing_tables_for_sync(XLogRecPtr current_lsn) UpdateSubscriptionRelState(MyLogicalRepWorker->subid, MyLogicalRepWorker->relid, MyLogicalRepWorker->relstate, - MyLogicalRepWorker->relstate_lsn); + MyLogicalRepWorker->relstate_lsn, + false); /* * End streaming so that LogRepWorkerWalRcvConn can be used to drop @@ -425,6 +426,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn) ListCell *lc; bool started_tx = false; bool should_exit = false; + Relation rel = NULL; Assert(!IsTransactionState()); @@ -492,7 +494,17 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn) * worker to remove the origin tracking as if there is any * error while dropping we won't restart it to drop the * origin. So passing missing_ok = true. + * + * Lock the subscription and origin in the same order as we + * are doing during DDL commands to avoid deadlocks. See + * AlterSubscription_refresh. */ + LockSharedObject(SubscriptionRelationId, MyLogicalRepWorker->subid, + 0, AccessShareLock); + + if (!rel) + rel = table_open(SubscriptionRelRelationId, RowExclusiveLock); + ReplicationOriginNameForLogicalRep(MyLogicalRepWorker->subid, rstate->relid, originname, @@ -504,7 +516,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn) */ UpdateSubscriptionRelState(MyLogicalRepWorker->subid, rstate->relid, rstate->state, - rstate->lsn); + rstate->lsn, true); } } else @@ -555,7 +567,14 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn) * This is required to avoid any undetected deadlocks * due to any existing lock as deadlock detector won't * be able to detect the waits on the latch. + * + * Also close any tables prior to the commit. */ + if (rel) + { + table_close(rel, NoLock); + rel = NULL; + } CommitTransactionCommand(); pgstat_report_stat(false); } @@ -615,13 +634,19 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn) MySubscription->name, MyLogicalRepWorker->userid, rstate->relid, - DSM_HANDLE_INVALID); + DSM_HANDLE_INVALID, + false); } } } } } + /* Close table if opened */ + if (rel) + table_close(rel, NoLock); + + if (started_tx) { /* @@ -1413,7 +1438,8 @@ LogicalRepSyncTableStart(XLogRecPtr *origin_startpos) UpdateSubscriptionRelState(MyLogicalRepWorker->subid, MyLogicalRepWorker->relid, MyLogicalRepWorker->relstate, - MyLogicalRepWorker->relstate_lsn); + MyLogicalRepWorker->relstate_lsn, + false); CommitTransactionCommand(); pgstat_report_stat(true); @@ -1546,7 +1572,8 @@ LogicalRepSyncTableStart(XLogRecPtr *origin_startpos) UpdateSubscriptionRelState(MyLogicalRepWorker->subid, MyLogicalRepWorker->relid, SUBREL_STATE_FINISHEDCOPY, - MyLogicalRepWorker->relstate_lsn); + MyLogicalRepWorker->relstate_lsn, + false); CommitTransactionCommand(); diff --git a/src/backend/replication/logical/worker.c b/src/backend/replication/logical/worker.c index c5fb627aa56..b59221c4d06 100644 --- a/src/backend/replication/logical/worker.c +++ b/src/backend/replication/logical/worker.c @@ -132,6 +132,96 @@ * failover = true when creating the subscription. Enabling failover allows us * to smoothly transition to the promoted standby, ensuring that we can * subscribe to the new primary without losing any data. + * + * RETAIN DEAD TUPLES + * ---------------------- + * Each apply worker that enabled retain_dead_tuples option maintains a + * non-removable transaction ID (oldest_nonremovable_xid) in shared memory to + * prevent dead rows from being removed prematurely when the apply worker still + * needs them to detect conflicts reliably. This helps to retain the required + * commit_ts module information, which further helps to detect + * update_origin_differs and delete_origin_differs conflicts reliably, as + * otherwise, vacuum freeze could remove the required information. + * + * The logical replication launcher manages an internal replication slot named + * "pg_conflict_detection". It asynchronously aggregates the non-removable + * transaction ID from all apply workers to determine the appropriate xmin for + * the slot, thereby retaining necessary tuples. + * + * The non-removable transaction ID in the apply worker is advanced to the + * oldest running transaction ID once all concurrent transactions on the + * publisher have been applied and flushed locally. The process involves: + * + * - RDT_GET_CANDIDATE_XID: + * Call GetOldestActiveTransactionId() to take oldestRunningXid as the + * candidate xid. + * + * - RDT_REQUEST_PUBLISHER_STATUS: + * Send a message to the walsender requesting the publisher status, which + * includes the latest WAL write position and information about transactions + * that are in the commit phase. + * + * - RDT_WAIT_FOR_PUBLISHER_STATUS: + * Wait for the status from the walsender. After receiving the first status, + * do not proceed if there are concurrent remote transactions that are still + * in the commit phase. These transactions might have been assigned an + * earlier commit timestamp but have not yet written the commit WAL record. + * Continue to request the publisher status (RDT_REQUEST_PUBLISHER_STATUS) + * until all these transactions have completed. + * + * - RDT_WAIT_FOR_LOCAL_FLUSH: + * Advance the non-removable transaction ID if the current flush location has + * reached or surpassed the last received WAL position. + * + * The overall state progression is: GET_CANDIDATE_XID -> + * REQUEST_PUBLISHER_STATUS -> WAIT_FOR_PUBLISHER_STATUS -> (loop to + * REQUEST_PUBLISHER_STATUS till concurrent remote transactions end) -> + * WAIT_FOR_LOCAL_FLUSH -> loop back to GET_CANDIDATE_XID. + * + * Retaining the dead tuples for this period is sufficient for ensuring + * eventual consistency using last-update-wins strategy, as dead tuples are + * useful for detecting conflicts only during the application of concurrent + * transactions from remote nodes. After applying and flushing all remote + * transactions that occurred concurrently with the tuple DELETE, any + * subsequent UPDATE from a remote node should have a later timestamp. In such + * cases, it is acceptable to detect an update_missing scenario and convert the + * UPDATE to an INSERT when applying it. But, detecting concurrent remote + * transactions with earlier timestamps than the DELETE is necessary, as the + * UPDATEs in remote transactions should be ignored if their timestamp is + * earlier than that of the dead tuples. + * + * Note that advancing the non-removable transaction ID is not supported if the + * publisher is also a physical standby. This is because the logical walsender + * on the standby can only get the WAL replay position but there may be more + * WALs that are being replicated from the primary and those WALs could have + * earlier commit timestamp. + * + * Similarly, when the publisher has subscribed to another publisher, + * information necessary for conflict detection cannot be retained for + * changes from origins other than the publisher. This is because publisher + * lacks the information on concurrent transactions of other publishers to + * which it subscribes. As the information on concurrent transactions is + * unavailable beyond subscriber's immediate publishers, the non-removable + * transaction ID might be advanced prematurely before changes from other + * origins have been fully applied. + * + * XXX Retaining information for changes from other origins might be possible + * by requesting the subscription on that origin to enable retain_dead_tuples + * and fetching the conflict detection slot.xmin along with the publisher's + * status. In the RDT_WAIT_FOR_PUBLISHER_STATUS phase, the apply worker could + * wait for the remote slot's xmin to reach the oldest active transaction ID, + * ensuring that all transactions from other origins have been applied on the + * publisher, thereby getting the latest WAL position that includes all + * concurrent changes. However, this approach may impact performance, so it + * might not worth the effort. + * + * XXX It seems feasible to get the latest commit's WAL location from the + * publisher and wait till that is applied. However, we can't do that + * because commit timestamps can regress as a commit with a later LSN is not + * guaranteed to have a later timestamp than those with earlier LSNs. Having + * said that, even if that is possible, it won't improve performance much as + * the apply always lag and moves slowly as compared with the transactions + * on the publisher. *------------------------------------------------------------------------- */ @@ -140,6 +230,7 @@ #include <sys/stat.h> #include <unistd.h> +#include "access/commit_ts.h" #include "access/table.h" #include "access/tableam.h" #include "access/twophase.h" @@ -148,6 +239,7 @@ #include "catalog/pg_inherits.h" #include "catalog/pg_subscription.h" #include "catalog/pg_subscription_rel.h" +#include "commands/subscriptioncmds.h" #include "commands/tablecmds.h" #include "commands/trigger.h" #include "executor/executor.h" @@ -166,12 +258,14 @@ #include "replication/logicalrelation.h" #include "replication/logicalworker.h" #include "replication/origin.h" +#include "replication/slot.h" #include "replication/walreceiver.h" #include "replication/worker_internal.h" #include "rewrite/rewriteHandler.h" #include "storage/buffile.h" #include "storage/ipc.h" #include "storage/lmgr.h" +#include "storage/procarray.h" #include "tcop/tcopprot.h" #include "utils/acl.h" #include "utils/dynahash.h" @@ -268,6 +362,78 @@ typedef enum TRANS_PARALLEL_APPLY, } TransApplyAction; +/* + * The phases involved in advancing the non-removable transaction ID. + * + * See comments atop worker.c for details of the transition between these + * phases. + */ +typedef enum +{ + RDT_GET_CANDIDATE_XID, + RDT_REQUEST_PUBLISHER_STATUS, + RDT_WAIT_FOR_PUBLISHER_STATUS, + RDT_WAIT_FOR_LOCAL_FLUSH +} RetainDeadTuplesPhase; + +/* + * Critical information for managing phase transitions within the + * RetainDeadTuplesPhase. + */ +typedef struct RetainDeadTuplesData +{ + RetainDeadTuplesPhase phase; /* current phase */ + XLogRecPtr remote_lsn; /* WAL write position on the publisher */ + + /* + * Oldest transaction ID that was in the commit phase on the publisher. + * Use FullTransactionId to prevent issues with transaction ID wraparound, + * where a new remote_oldestxid could falsely appear to originate from the + * past and block advancement. + */ + FullTransactionId remote_oldestxid; + + /* + * Next transaction ID to be assigned on the publisher. Use + * FullTransactionId for consistency and to allow straightforward + * comparisons with remote_oldestxid. + */ + FullTransactionId remote_nextxid; + + TimestampTz reply_time; /* when the publisher responds with status */ + + /* + * Publisher transaction ID that must be awaited to complete before + * entering the final phase (RDT_WAIT_FOR_LOCAL_FLUSH). Use + * FullTransactionId for the same reason as remote_nextxid. + */ + FullTransactionId remote_wait_for; + + TransactionId candidate_xid; /* candidate for the non-removable + * transaction ID */ + TimestampTz flushpos_update_time; /* when the remote flush position was + * updated in final phase + * (RDT_WAIT_FOR_LOCAL_FLUSH) */ + + /* + * The following fields are used to determine the timing for the next + * round of transaction ID advancement. + */ + TimestampTz last_recv_time; /* when the last message was received */ + TimestampTz candidate_xid_time; /* when the candidate_xid is decided */ + int xid_advance_interval; /* how much time (ms) to wait before + * attempting to advance the + * non-removable transaction ID */ +} RetainDeadTuplesData; + +/* + * The minimum (100ms) and maximum (3 minutes) intervals for advancing + * non-removable transaction IDs. The maximum interval is a bit arbitrary but + * is sufficient to not cause any undue network traffic. + */ +#define MIN_XID_ADVANCE_INTERVAL 100 +#define MAX_XID_ADVANCE_INTERVAL 180000 + /* errcontext tracker */ static ApplyErrorCallbackArg apply_error_callback_arg = { @@ -332,6 +498,13 @@ static XLogRecPtr skip_xact_finish_lsn = InvalidXLogRecPtr; /* BufFile handle of the current streaming file */ static BufFile *stream_fd = NULL; +/* + * The remote WAL position that has been applied and flushed locally. We record + * and use this information both while sending feedback to the server and + * advancing oldest_nonremovable_xid. + */ +static XLogRecPtr last_flushpos = InvalidXLogRecPtr; + typedef struct SubXactInfo { TransactionId xid; /* XID of the subxact */ @@ -372,6 +545,19 @@ static void stream_close_file(void); static void send_feedback(XLogRecPtr recvpos, bool force, bool requestReply); +static void maybe_advance_nonremovable_xid(RetainDeadTuplesData *rdt_data, + bool status_received); +static bool can_advance_nonremovable_xid(RetainDeadTuplesData *rdt_data); +static void process_rdt_phase_transition(RetainDeadTuplesData *rdt_data, + bool status_received); +static void get_candidate_xid(RetainDeadTuplesData *rdt_data); +static void request_publisher_status(RetainDeadTuplesData *rdt_data); +static void wait_for_publisher_status(RetainDeadTuplesData *rdt_data, + bool status_received); +static void wait_for_local_flush(RetainDeadTuplesData *rdt_data); +static void adjust_xid_advance_interval(RetainDeadTuplesData *rdt_data, + bool new_xid_found); + static void apply_handle_commit_internal(LogicalRepCommitData *commit_data); static void apply_handle_insert_internal(ApplyExecutionData *edata, ResultRelInfo *relinfo, @@ -3577,6 +3763,7 @@ LogicalRepApplyLoop(XLogRecPtr last_received) bool ping_sent = false; TimeLineID tli; ErrorContextCallback errcallback; + RetainDeadTuplesData rdt_data = {0}; /* * Init the ApplyMessageContext which we clean up after each replication @@ -3655,6 +3842,8 @@ LogicalRepApplyLoop(XLogRecPtr last_received) last_recv_timestamp = GetCurrentTimestamp(); ping_sent = false; + rdt_data.last_recv_time = last_recv_timestamp; + /* Ensure we are reading the data into our memory context. */ MemoryContextSwitchTo(ApplyMessageContext); @@ -3681,6 +3870,8 @@ LogicalRepApplyLoop(XLogRecPtr last_received) UpdateWorkerStats(last_received, send_time, false); apply_dispatch(&s); + + maybe_advance_nonremovable_xid(&rdt_data, false); } else if (c == 'k') { @@ -3696,8 +3887,31 @@ LogicalRepApplyLoop(XLogRecPtr last_received) last_received = end_lsn; send_feedback(last_received, reply_requested, false); + + maybe_advance_nonremovable_xid(&rdt_data, false); + UpdateWorkerStats(last_received, timestamp, true); } + else if (c == 's') /* Primary status update */ + { + rdt_data.remote_lsn = pq_getmsgint64(&s); + rdt_data.remote_oldestxid = FullTransactionIdFromU64((uint64) pq_getmsgint64(&s)); + rdt_data.remote_nextxid = FullTransactionIdFromU64((uint64) pq_getmsgint64(&s)); + rdt_data.reply_time = pq_getmsgint64(&s); + + /* + * This should never happen, see + * ProcessStandbyPSRequestMessage. But if it happens + * due to a bug, we don't want to proceed as it can + * incorrectly advance oldest_nonremovable_xid. + */ + if (XLogRecPtrIsInvalid(rdt_data.remote_lsn)) + elog(ERROR, "cannot get the latest WAL position from the publisher"); + + maybe_advance_nonremovable_xid(&rdt_data, true); + + UpdateWorkerStats(last_received, rdt_data.reply_time, false); + } /* other message types are purposefully ignored */ MemoryContextReset(ApplyMessageContext); @@ -3710,6 +3924,11 @@ LogicalRepApplyLoop(XLogRecPtr last_received) /* confirm all writes so far */ send_feedback(last_received, false, false); + /* Reset the timestamp if no message was received */ + rdt_data.last_recv_time = 0; + + maybe_advance_nonremovable_xid(&rdt_data, false); + if (!in_remote_transaction && !in_streamed_transaction) { /* @@ -3744,6 +3963,14 @@ LogicalRepApplyLoop(XLogRecPtr last_received) else wait_time = NAPTIME_PER_CYCLE; + /* + * Ensure to wake up when it's possible to advance the non-removable + * transaction ID. + */ + if (rdt_data.phase == RDT_GET_CANDIDATE_XID && + rdt_data.xid_advance_interval) + wait_time = Min(wait_time, rdt_data.xid_advance_interval); + rc = WaitLatchOrSocket(MyLatch, WL_SOCKET_READABLE | WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH, @@ -3807,6 +4034,8 @@ LogicalRepApplyLoop(XLogRecPtr last_received) send_feedback(last_received, requestReply, requestReply); + maybe_advance_nonremovable_xid(&rdt_data, false); + /* * Force reporting to ensure long idle periods don't lead to * arbitrarily delayed stats. Stats can only be reported outside @@ -3842,7 +4071,6 @@ send_feedback(XLogRecPtr recvpos, bool force, bool requestReply) static XLogRecPtr last_recvpos = InvalidXLogRecPtr; static XLogRecPtr last_writepos = InvalidXLogRecPtr; - static XLogRecPtr last_flushpos = InvalidXLogRecPtr; XLogRecPtr writepos; XLogRecPtr flushpos; @@ -3921,6 +4149,367 @@ send_feedback(XLogRecPtr recvpos, bool force, bool requestReply) } /* + * Attempt to advance the non-removable transaction ID. + * + * See comments atop worker.c for details. + */ +static void +maybe_advance_nonremovable_xid(RetainDeadTuplesData *rdt_data, + bool status_received) +{ + if (!can_advance_nonremovable_xid(rdt_data)) + return; + + process_rdt_phase_transition(rdt_data, status_received); +} + +/* + * Preliminary check to determine if advancing the non-removable transaction ID + * is allowed. + */ +static bool +can_advance_nonremovable_xid(RetainDeadTuplesData *rdt_data) +{ + /* + * It is sufficient to manage non-removable transaction ID for a + * subscription by the main apply worker to detect conflicts reliably even + * for table sync or parallel apply workers. + */ + if (!am_leader_apply_worker()) + return false; + + /* No need to advance if retaining dead tuples is not required */ + if (!MySubscription->retaindeadtuples) + return false; + + return true; +} + +/* + * Process phase transitions during the non-removable transaction ID + * advancement. See comments atop worker.c for details of the transition. + */ +static void +process_rdt_phase_transition(RetainDeadTuplesData *rdt_data, + bool status_received) +{ + switch (rdt_data->phase) + { + case RDT_GET_CANDIDATE_XID: + get_candidate_xid(rdt_data); + break; + case RDT_REQUEST_PUBLISHER_STATUS: + request_publisher_status(rdt_data); + break; + case RDT_WAIT_FOR_PUBLISHER_STATUS: + wait_for_publisher_status(rdt_data, status_received); + break; + case RDT_WAIT_FOR_LOCAL_FLUSH: + wait_for_local_flush(rdt_data); + break; + } +} + +/* + * Workhorse for the RDT_GET_CANDIDATE_XID phase. + */ +static void +get_candidate_xid(RetainDeadTuplesData *rdt_data) +{ + TransactionId oldest_running_xid; + TimestampTz now; + + /* + * Use last_recv_time when applying changes in the loop to avoid + * unnecessary system time retrieval. If last_recv_time is not available, + * obtain the current timestamp. + */ + now = rdt_data->last_recv_time ? rdt_data->last_recv_time : GetCurrentTimestamp(); + + /* + * Compute the candidate_xid and request the publisher status at most once + * per xid_advance_interval. Refer to adjust_xid_advance_interval() for + * details on how this value is dynamically adjusted. This is to avoid + * using CPU and network resources without making much progress. + */ + if (!TimestampDifferenceExceeds(rdt_data->candidate_xid_time, now, + rdt_data->xid_advance_interval)) + return; + + /* + * Immediately update the timer, even if the function returns later + * without setting candidate_xid due to inactivity on the subscriber. This + * avoids frequent calls to GetOldestActiveTransactionId. + */ + rdt_data->candidate_xid_time = now; + + /* + * Consider transactions in the current database, as only dead tuples from + * this database are required for conflict detection. + */ + oldest_running_xid = GetOldestActiveTransactionId(false, false); + + /* + * Oldest active transaction ID (oldest_running_xid) can't be behind any + * of its previously computed value. + */ + Assert(TransactionIdPrecedesOrEquals(MyLogicalRepWorker->oldest_nonremovable_xid, + oldest_running_xid)); + + /* Return if the oldest_nonremovable_xid cannot be advanced */ + if (TransactionIdEquals(MyLogicalRepWorker->oldest_nonremovable_xid, + oldest_running_xid)) + { + adjust_xid_advance_interval(rdt_data, false); + return; + } + + adjust_xid_advance_interval(rdt_data, true); + + rdt_data->candidate_xid = oldest_running_xid; + rdt_data->phase = RDT_REQUEST_PUBLISHER_STATUS; + + /* process the next phase */ + process_rdt_phase_transition(rdt_data, false); +} + +/* + * Workhorse for the RDT_REQUEST_PUBLISHER_STATUS phase. + */ +static void +request_publisher_status(RetainDeadTuplesData *rdt_data) +{ + static StringInfo request_message = NULL; + + if (!request_message) + { + MemoryContext oldctx = MemoryContextSwitchTo(ApplyContext); + + request_message = makeStringInfo(); + MemoryContextSwitchTo(oldctx); + } + else + resetStringInfo(request_message); + + /* + * Send the current time to update the remote walsender's latest reply + * message received time. + */ + pq_sendbyte(request_message, 'p'); + pq_sendint64(request_message, GetCurrentTimestamp()); + + elog(DEBUG2, "sending publisher status request message"); + + /* Send a request for the publisher status */ + walrcv_send(LogRepWorkerWalRcvConn, + request_message->data, request_message->len); + + rdt_data->phase = RDT_WAIT_FOR_PUBLISHER_STATUS; + + /* + * Skip calling maybe_advance_nonremovable_xid() since further transition + * is possible only once we receive the publisher status message. + */ +} + +/* + * Workhorse for the RDT_WAIT_FOR_PUBLISHER_STATUS phase. + */ +static void +wait_for_publisher_status(RetainDeadTuplesData *rdt_data, + bool status_received) +{ + /* + * Return if we have requested but not yet received the publisher status. + */ + if (!status_received) + return; + + if (!FullTransactionIdIsValid(rdt_data->remote_wait_for)) + rdt_data->remote_wait_for = rdt_data->remote_nextxid; + + /* + * Check if all remote concurrent transactions that were active at the + * first status request have now completed. If completed, proceed to the + * next phase; otherwise, continue checking the publisher status until + * these transactions finish. + * + * It's possible that transactions in the commit phase during the last + * cycle have now finished committing, but remote_oldestxid remains older + * than remote_wait_for. This can happen if some old transaction came in + * the commit phase when we requested status in this cycle. We do not + * handle this case explicitly as it's rare and the benefit doesn't + * justify the required complexity. Tracking would require either caching + * all xids at the publisher or sending them to subscribers. The condition + * will resolve naturally once the remaining transactions are finished. + * + * Directly advancing the non-removable transaction ID is possible if + * there are no activities on the publisher since the last advancement + * cycle. However, it requires maintaining two fields, last_remote_nextxid + * and last_remote_lsn, within the structure for comparison with the + * current cycle's values. Considering the minimal cost of continuing in + * RDT_WAIT_FOR_LOCAL_FLUSH without awaiting changes, we opted not to + * advance the transaction ID here. + */ + if (FullTransactionIdPrecedesOrEquals(rdt_data->remote_wait_for, + rdt_data->remote_oldestxid)) + rdt_data->phase = RDT_WAIT_FOR_LOCAL_FLUSH; + else + rdt_data->phase = RDT_REQUEST_PUBLISHER_STATUS; + + /* process the next phase */ + process_rdt_phase_transition(rdt_data, false); +} + +/* + * Workhorse for the RDT_WAIT_FOR_LOCAL_FLUSH phase. + */ +static void +wait_for_local_flush(RetainDeadTuplesData *rdt_data) +{ + Assert(!XLogRecPtrIsInvalid(rdt_data->remote_lsn) && + TransactionIdIsValid(rdt_data->candidate_xid)); + + /* + * We expect the publisher and subscriber clocks to be in sync using time + * sync service like NTP. Otherwise, we will advance this worker's + * oldest_nonremovable_xid prematurely, leading to the removal of rows + * required to detect conflicts reliably. This check primarily addresses + * scenarios where the publisher's clock falls behind; if the publisher's + * clock is ahead, subsequent transactions will naturally bear later + * commit timestamps, conforming to the design outlined atop worker.c. + * + * XXX Consider waiting for the publisher's clock to catch up with the + * subscriber's before proceeding to the next phase. + */ + if (TimestampDifferenceExceeds(rdt_data->reply_time, + rdt_data->candidate_xid_time, 0)) + ereport(ERROR, + errmsg_internal("oldest_nonremovable_xid transaction ID could be advanced prematurely"), + errdetail_internal("The clock on the publisher is behind that of the subscriber.")); + + /* + * Do not attempt to advance the non-removable transaction ID when table + * sync is in progress. During this time, changes from a single + * transaction may be applied by multiple table sync workers corresponding + * to the target tables. So, it's necessary for all table sync workers to + * apply and flush the corresponding changes before advancing the + * transaction ID, otherwise, dead tuples that are still needed for + * conflict detection in table sync workers could be removed prematurely. + * However, confirming the apply and flush progress across all table sync + * workers is complex and not worth the effort, so we simply return if not + * all tables are in the READY state. + * + * It is safe to add new tables with initial states to the subscription + * after this check because any changes applied to these tables should + * have a WAL position greater than the rdt_data->remote_lsn. + */ + if (!AllTablesyncsReady()) + return; + + /* + * Update and check the remote flush position if we are applying changes + * in a loop. This is done at most once per WalWriterDelay to avoid + * performing costly operations in get_flush_position() too frequently + * during change application. + */ + if (last_flushpos < rdt_data->remote_lsn && rdt_data->last_recv_time && + TimestampDifferenceExceeds(rdt_data->flushpos_update_time, + rdt_data->last_recv_time, WalWriterDelay)) + { + XLogRecPtr writepos; + XLogRecPtr flushpos; + bool have_pending_txes; + + /* Fetch the latest remote flush position */ + get_flush_position(&writepos, &flushpos, &have_pending_txes); + + if (flushpos > last_flushpos) + last_flushpos = flushpos; + + rdt_data->flushpos_update_time = rdt_data->last_recv_time; + } + + /* Return to wait for the changes to be applied */ + if (last_flushpos < rdt_data->remote_lsn) + return; + + /* + * Reaching here means the remote WAL position has been received, and all + * transactions up to that position on the publisher have been applied and + * flushed locally. So, we can advance the non-removable transaction ID. + */ + SpinLockAcquire(&MyLogicalRepWorker->relmutex); + MyLogicalRepWorker->oldest_nonremovable_xid = rdt_data->candidate_xid; + SpinLockRelease(&MyLogicalRepWorker->relmutex); + + elog(DEBUG2, "confirmed flush up to remote lsn %X/%X: new oldest_nonremovable_xid %u", + LSN_FORMAT_ARGS(rdt_data->remote_lsn), + rdt_data->candidate_xid); + + /* Notify launcher to update the xmin of the conflict slot */ + ApplyLauncherWakeup(); + + /* + * Reset all data fields except those used to determine the timing for the + * next round of transaction ID advancement. We can even use + * flushpos_update_time in the next round to decide whether to get the + * latest flush position. + */ + rdt_data->phase = RDT_GET_CANDIDATE_XID; + rdt_data->remote_lsn = InvalidXLogRecPtr; + rdt_data->remote_oldestxid = InvalidFullTransactionId; + rdt_data->remote_nextxid = InvalidFullTransactionId; + rdt_data->reply_time = 0; + rdt_data->remote_wait_for = InvalidFullTransactionId; + rdt_data->candidate_xid = InvalidTransactionId; + + /* process the next phase */ + process_rdt_phase_transition(rdt_data, false); +} + +/* + * Adjust the interval for advancing non-removable transaction IDs. + * + * We double the interval to try advancing the non-removable transaction IDs + * if there is no activity on the node. The maximum value of the interval is + * capped by wal_receiver_status_interval if it is not zero, otherwise to a + * 3 minutes which should be sufficient to avoid using CPU or network + * resources without much benefit. + * + * The interval is reset to a minimum value of 100ms once there is some + * activity on the node. + * + * XXX The use of wal_receiver_status_interval is a bit arbitrary so we can + * consider the other interval or a separate GUC if the need arises. + */ +static void +adjust_xid_advance_interval(RetainDeadTuplesData *rdt_data, bool new_xid_found) +{ + if (!new_xid_found && rdt_data->xid_advance_interval) + { + int max_interval = wal_receiver_status_interval + ? wal_receiver_status_interval * 1000 + : MAX_XID_ADVANCE_INTERVAL; + + /* + * No new transaction ID has been assigned since the last check, so + * double the interval, but not beyond the maximum allowable value. + */ + rdt_data->xid_advance_interval = Min(rdt_data->xid_advance_interval * 2, + max_interval); + } + else + { + /* + * A new transaction ID was found or the interval is not yet + * initialized, so set the interval to the minimum value. + */ + rdt_data->xid_advance_interval = MIN_XID_ADVANCE_INTERVAL; + } +} + +/* * Exit routine for apply workers due to subscription parameter changes. */ static void @@ -4708,6 +5297,30 @@ InitializeLogRepWorker(void) apply_worker_exit(); } + /* + * Restart the worker if retain_dead_tuples was enabled during startup. + * + * At this point, the replication slot used for conflict detection might + * not exist yet, or could be dropped soon if the launcher perceives + * retain_dead_tuples as disabled. To avoid unnecessary tracking of + * oldest_nonremovable_xid when the slot is absent or at risk of being + * dropped, a restart is initiated. + * + * The oldest_nonremovable_xid should be initialized only when the + * retain_dead_tuples is enabled before launching the worker. See + * logicalrep_worker_launch. + */ + if (am_leader_apply_worker() && + MySubscription->retaindeadtuples && + !TransactionIdIsValid(MyLogicalRepWorker->oldest_nonremovable_xid)) + { + ereport(LOG, + errmsg("logical replication worker for subscription \"%s\" will restart because the option %s was enabled during startup", + MySubscription->name, "retain_dead_tuples")); + + apply_worker_exit(); + } + /* Setup synchronous commit according to the user's wishes */ SetConfigOption("synchronous_commit", MySubscription->synccommit, PGC_BACKEND, PGC_S_OVERRIDE); @@ -4864,6 +5477,14 @@ DisableSubscriptionAndExit(void) errmsg("subscription \"%s\" has been disabled because of an error", MySubscription->name)); + /* + * Skip the track_commit_timestamp check when disabling the worker due to + * an error, as verifying commit timestamps is unnecessary in this + * context. + */ + if (MySubscription->retaindeadtuples) + CheckSubDeadTupleRetention(false, true, WARNING); + proc_exit(0); } diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c index 082b4d9d327..f4c977262c5 100644 --- a/src/backend/replication/pgoutput/pgoutput.c +++ b/src/backend/replication/pgoutput/pgoutput.c @@ -297,10 +297,12 @@ parse_output_parameters(List *options, PGOutputData *data) bool two_phase_option_given = false; bool origin_option_given = false; + /* Initialize optional parameters to defaults */ data->binary = false; data->streaming = LOGICALREP_STREAM_OFF; data->messages = false; data->two_phase = false; + data->publish_no_origin = false; foreach(lc, options) { diff --git a/src/backend/replication/slot.c b/src/backend/replication/slot.c index e44ad576bc7..8605776ad86 100644 --- a/src/backend/replication/slot.c +++ b/src/backend/replication/slot.c @@ -47,6 +47,7 @@ #include "miscadmin.h" #include "pgstat.h" #include "postmaster/interrupt.h" +#include "replication/logicallauncher.h" #include "replication/slotsync.h" #include "replication/slot.h" #include "replication/walsender_private.h" @@ -172,6 +173,7 @@ static SyncStandbySlotsConfigData *synchronized_standby_slots_config; static XLogRecPtr ss_oldest_flush_lsn = InvalidXLogRecPtr; static void ReplicationSlotShmemExit(int code, Datum arg); +static bool IsSlotForConflictCheck(const char *name); static void ReplicationSlotDropPtr(ReplicationSlot *slot); /* internal persistency functions */ @@ -258,13 +260,17 @@ ReplicationSlotShmemExit(int code, Datum arg) /* * Check whether the passed slot name is valid and report errors at elevel. * + * An error will be reported for a reserved replication slot name if + * allow_reserved_name is set to false. + * * Slot names may consist out of [a-z0-9_]{1,NAMEDATALEN-1} which should allow * the name to be used as a directory name on every supported OS. * * Returns whether the directory name is valid or not if elevel < ERROR. */ bool -ReplicationSlotValidateName(const char *name, int elevel) +ReplicationSlotValidateName(const char *name, bool allow_reserved_name, + int elevel) { const char *cp; @@ -300,10 +306,32 @@ ReplicationSlotValidateName(const char *name, int elevel) return false; } } + + if (!allow_reserved_name && IsSlotForConflictCheck(name)) + { + ereport(elevel, + errcode(ERRCODE_RESERVED_NAME), + errmsg("replication slot name \"%s\" is reserved", + name), + errdetail("The name \"%s\" is reserved for the conflict detection slot.", + CONFLICT_DETECTION_SLOT)); + + return false; + } + return true; } /* + * Return true if the replication slot name is "pg_conflict_detection". + */ +static bool +IsSlotForConflictCheck(const char *name) +{ + return (strcmp(name, CONFLICT_DETECTION_SLOT) == 0); +} + +/* * Create a new replication slot and mark it as used by this backend. * * name: Name of the slot @@ -330,7 +358,12 @@ ReplicationSlotCreate(const char *name, bool db_specific, Assert(MyReplicationSlot == NULL); - ReplicationSlotValidateName(name, ERROR); + /* + * The logical launcher or pg_upgrade may create or migrate an internal + * slot, so using a reserved name is allowed in these cases. + */ + ReplicationSlotValidateName(name, IsBinaryUpgrade || IsLogicalLauncher(), + ERROR); if (failover) { @@ -582,6 +615,17 @@ retry: } /* + * Do not allow users to acquire the reserved slot. This scenario may + * occur if the launcher that owns the slot has terminated unexpectedly + * due to an error, and a backend process attempts to reuse the slot. + */ + if (!IsLogicalLauncher() && IsSlotForConflictCheck(name)) + ereport(ERROR, + errcode(ERRCODE_UNDEFINED_OBJECT), + errmsg("cannot acquire replication slot \"%s\"", name), + errdetail("The slot is reserved for conflict detection and can only be acquired by logical replication launcher.")); + + /* * This is the slot we want; check if it's active under some other * process. In single user mode, we don't need this check. */ diff --git a/src/backend/replication/syncrep_scanner.l b/src/backend/replication/syncrep_scanner.l index 7dec1f869c7..02004d621e7 100644 --- a/src/backend/replication/syncrep_scanner.l +++ b/src/backend/replication/syncrep_scanner.l @@ -157,17 +157,16 @@ syncrep_yyerror(SyncRepConfigData **syncrep_parse_result_p, char **syncrep_parse { struct yyguts_t *yyg = (struct yyguts_t *) yyscanner; /* needed for yytext * macro */ - char *syncrep_parse_error_msg = *syncrep_parse_error_msg_p; /* report only the first error in a parse operation */ - if (syncrep_parse_error_msg) + if (*syncrep_parse_error_msg_p) return; if (yytext[0]) - syncrep_parse_error_msg = psprintf("%s at or near \"%s\"", - message, yytext); + *syncrep_parse_error_msg_p = psprintf("%s at or near \"%s\"", + message, yytext); else - syncrep_parse_error_msg = psprintf("%s at end of input", - message); + *syncrep_parse_error_msg_p = psprintf("%s at end of input", + message); } void diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c index 28b8591efa5..ee911394a23 100644 --- a/src/backend/replication/walsender.c +++ b/src/backend/replication/walsender.c @@ -65,6 +65,7 @@ #include "funcapi.h" #include "libpq/libpq.h" #include "libpq/pqformat.h" +#include "libpq/protocol.h" #include "miscadmin.h" #include "nodes/replnodes.h" #include "pgstat.h" @@ -84,6 +85,7 @@ #include "storage/ipc.h" #include "storage/pmsignal.h" #include "storage/proc.h" +#include "storage/procarray.h" #include "tcop/dest.h" #include "tcop/tcopprot.h" #include "utils/acl.h" @@ -258,6 +260,7 @@ static void StartLogicalReplication(StartReplicationCmd *cmd); static void ProcessStandbyMessage(void); static void ProcessStandbyReplyMessage(void); static void ProcessStandbyHSFeedbackMessage(void); +static void ProcessStandbyPSRequestMessage(void); static void ProcessRepliesIfAny(void); static void ProcessPendingWrites(void); static void WalSndKeepalive(bool requestReply, XLogRecPtr writePtr); @@ -733,13 +736,13 @@ HandleUploadManifestPacket(StringInfo buf, off_t *offset, switch (mtype) { - case 'd': /* CopyData */ + case PqMsg_CopyData: maxmsglen = PQ_LARGE_MESSAGE_LIMIT; break; - case 'c': /* CopyDone */ - case 'f': /* CopyFail */ - case 'H': /* Flush */ - case 'S': /* Sync */ + case PqMsg_CopyDone: + case PqMsg_CopyFail: + case PqMsg_Flush: + case PqMsg_Sync: maxmsglen = PQ_SMALL_MESSAGE_LIMIT; break; default: @@ -761,19 +764,19 @@ HandleUploadManifestPacket(StringInfo buf, off_t *offset, /* Process the message */ switch (mtype) { - case 'd': /* CopyData */ + case PqMsg_CopyData: AppendIncrementalManifestData(ib, buf->data, buf->len); return true; - case 'c': /* CopyDone */ + case PqMsg_CopyDone: return false; - case 'H': /* Sync */ - case 'S': /* Flush */ + case PqMsg_Sync: + case PqMsg_Flush: /* Ignore these while in CopyOut mode as we do elsewhere. */ return true; - case 'f': + case PqMsg_CopyFail: ereport(ERROR, (errcode(ERRCODE_QUERY_CANCELED), errmsg("COPY from stdin failed: %s", @@ -1567,7 +1570,7 @@ WalSndWriteData(LogicalDecodingContext *ctx, XLogRecPtr lsn, TransactionId xid, tmpbuf.data, sizeof(int64)); /* output previously gathered data in a CopyData packet */ - pq_putmessage_noblock('d', ctx->out->data, ctx->out->len); + pq_putmessage_noblock(PqMsg_CopyData, ctx->out->data, ctx->out->len); CHECK_FOR_INTERRUPTS(); @@ -2303,7 +2306,7 @@ ProcessRepliesIfAny(void) case PqMsg_CopyDone: if (!streamingDoneSending) { - pq_putmessage_noblock('c', NULL, 0); + pq_putmessage_noblock(PqMsg_CopyDone, NULL, 0); streamingDoneSending = true; } @@ -2355,6 +2358,10 @@ ProcessStandbyMessage(void) ProcessStandbyHSFeedbackMessage(); break; + case 'p': + ProcessStandbyPSRequestMessage(); + break; + default: ereport(COMMERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), @@ -2702,6 +2709,60 @@ ProcessStandbyHSFeedbackMessage(void) } /* + * Process the request for a primary status update message. + */ +static void +ProcessStandbyPSRequestMessage(void) +{ + XLogRecPtr lsn = InvalidXLogRecPtr; + TransactionId oldestXidInCommit; + FullTransactionId nextFullXid; + FullTransactionId fullOldestXidInCommit; + WalSnd *walsnd = MyWalSnd; + TimestampTz replyTime; + + /* + * This shouldn't happen because we don't support getting primary status + * message from standby. + */ + if (RecoveryInProgress()) + elog(ERROR, "the primary status is unavailable during recovery"); + + replyTime = pq_getmsgint64(&reply_message); + + /* + * Update shared state for this WalSender process based on reply data from + * standby. + */ + SpinLockAcquire(&walsnd->mutex); + walsnd->replyTime = replyTime; + SpinLockRelease(&walsnd->mutex); + + /* + * Consider transactions in the current database, as only these are the + * ones replicated. + */ + oldestXidInCommit = GetOldestActiveTransactionId(true, false); + nextFullXid = ReadNextFullTransactionId(); + fullOldestXidInCommit = FullTransactionIdFromAllowableAt(nextFullXid, + oldestXidInCommit); + lsn = GetXLogWriteRecPtr(); + + elog(DEBUG2, "sending primary status"); + + /* construct the message... */ + resetStringInfo(&output_message); + pq_sendbyte(&output_message, 's'); + pq_sendint64(&output_message, lsn); + pq_sendint64(&output_message, (int64) U64FromFullTransactionId(fullOldestXidInCommit)); + pq_sendint64(&output_message, (int64) U64FromFullTransactionId(nextFullXid)); + pq_sendint64(&output_message, GetCurrentTimestamp()); + + /* ... and send it wrapped in CopyData */ + pq_putmessage_noblock(PqMsg_CopyData, output_message.data, output_message.len); +} + +/* * Compute how long send/receive loops should sleep. * * If wal_sender_timeout is enabled we want to wake up in time to send @@ -3246,7 +3307,7 @@ XLogSendPhysical(void) wal_segment_close(xlogreader); /* Send CopyDone */ - pq_putmessage_noblock('c', NULL, 0); + pq_putmessage_noblock(PqMsg_CopyDone, NULL, 0); streamingDoneSending = true; WalSndCaughtUp = true; @@ -3374,7 +3435,7 @@ retry: memcpy(&output_message.data[1 + sizeof(int64) + sizeof(int64)], tmpbuf.data, sizeof(int64)); - pq_putmessage_noblock('d', output_message.data, output_message.len); + pq_putmessage_noblock(PqMsg_CopyData, output_message.data, output_message.len); sentPtr = endptr; @@ -4080,7 +4141,7 @@ WalSndKeepalive(bool requestReply, XLogRecPtr writePtr) pq_sendbyte(&output_message, requestReply ? 1 : 0); /* ... and send it wrapped in CopyData */ - pq_putmessage_noblock('d', output_message.data, output_message.len); + pq_putmessage_noblock(PqMsg_CopyData, output_message.data, output_message.len); /* Set local flag */ if (requestReply) diff --git a/src/backend/storage/aio/README.md b/src/backend/storage/aio/README.md index f10b5c7e31e..72ae3b3737d 100644 --- a/src/backend/storage/aio/README.md +++ b/src/backend/storage/aio/README.md @@ -94,7 +94,7 @@ pgaio_io_register_callbacks(ioh, PGAIO_HCB_SHARED_BUFFER_READV, 0); * * In this example we're reading only a single buffer, hence the 1. */ -pgaio_io_set_handle_data_32(ioh, (uint32 *) buffer, 1); +pgaio_io_set_handle_data_32(ioh, (uint32 *) &buffer, 1); /* * Pass the AIO handle to lower-level function. When operating on the level of @@ -119,8 +119,9 @@ pgaio_io_set_handle_data_32(ioh, (uint32 *) buffer, 1); * e.g. due to reaching a limit on the number of unsubmitted IOs, and even * complete before smgrstartreadv() returns. */ +void *page = BufferGetBlock(buffer); smgrstartreadv(ioh, operation->smgr, forknum, blkno, - BufferGetBlock(buffer), 1); + &page, 1); /* * To benefit from AIO, it is beneficial to perform other work, including diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index 6afdd28dba6..67431208e7f 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -2743,12 +2743,10 @@ ExtendBufferedRelShared(BufferManagerRelation bmr, * because mdread doesn't complain about reads beyond EOF (when * zero_damaged_pages is ON) and so a previous attempt to read a block * beyond EOF could have left a "valid" zero-filled buffer. - * Unfortunately, we have also seen this case occurring because of - * buggy Linux kernels that sometimes return an lseek(SEEK_END) result - * that doesn't account for a recent write. In that situation, the - * pre-existing buffer would contain valid data that we don't want to - * overwrite. Since the legitimate cases should always have left a - * zero-filled buffer, complain if not PageIsNew. + * + * This has also been observed when relation was overwritten by + * external process. Since the legitimate cases should always have + * left a zero-filled buffer, complain if not PageIsNew. */ if (existing_id >= 0) { @@ -2778,8 +2776,7 @@ ExtendBufferedRelShared(BufferManagerRelation bmr, ereport(ERROR, (errmsg("unexpected data beyond EOF in block %u of relation %s", existing_hdr->tag.blockNum, - relpath(bmr.smgr->smgr_rlocator, fork).str), - errhint("This has been seen to occur with buggy kernels; consider updating your system."))); + relpath(bmr.smgr->smgr_rlocator, fork).str))); /* * We *must* do smgr[zero]extend before succeeding, else the page diff --git a/src/backend/storage/buffer/localbuf.c b/src/backend/storage/buffer/localbuf.c index 3da9c41ee1d..3c0d20f4659 100644 --- a/src/backend/storage/buffer/localbuf.c +++ b/src/backend/storage/buffer/localbuf.c @@ -932,10 +932,11 @@ GetLocalBufferStorage(void) num_bufs = Min(num_bufs, MaxAllocSize / BLCKSZ); /* Buffers should be I/O aligned. */ - cur_block = (char *) - TYPEALIGN(PG_IO_ALIGN_SIZE, - MemoryContextAlloc(LocalBufferContext, - num_bufs * BLCKSZ + PG_IO_ALIGN_SIZE)); + cur_block = MemoryContextAllocAligned(LocalBufferContext, + num_bufs * BLCKSZ, + PG_IO_ALIGN_SIZE, + 0); + next_buf_in_block = 0; num_bufs_in_block = num_bufs; } diff --git a/src/backend/storage/ipc/latch.c b/src/backend/storage/ipc/latch.c index c6aefd2f688..beadeb5e46a 100644 --- a/src/backend/storage/ipc/latch.c +++ b/src/backend/storage/ipc/latch.c @@ -187,9 +187,11 @@ WaitLatch(Latch *latch, int wakeEvents, long timeout, if (!(wakeEvents & WL_LATCH_SET)) latch = NULL; ModifyWaitEvent(LatchWaitSet, LatchWaitSetLatchPos, WL_LATCH_SET, latch); - ModifyWaitEvent(LatchWaitSet, LatchWaitSetPostmasterDeathPos, - (wakeEvents & (WL_EXIT_ON_PM_DEATH | WL_POSTMASTER_DEATH)), - NULL); + + if (IsUnderPostmaster) + ModifyWaitEvent(LatchWaitSet, LatchWaitSetPostmasterDeathPos, + (wakeEvents & (WL_EXIT_ON_PM_DEATH | WL_POSTMASTER_DEATH)), + NULL); if (WaitEventSetWait(LatchWaitSet, (wakeEvents & WL_TIMEOUT) ? timeout : -1, diff --git a/src/backend/storage/ipc/procarray.c b/src/backend/storage/ipc/procarray.c index 2418967def6..bf987aed8d3 100644 --- a/src/backend/storage/ipc/procarray.c +++ b/src/backend/storage/ipc/procarray.c @@ -2814,8 +2814,10 @@ GetRunningTransactionData(void) * * Similar to GetSnapshotData but returns just oldestActiveXid. We include * all PGPROCs with an assigned TransactionId, even VACUUM processes. - * We look at all databases, though there is no need to include WALSender - * since this has no effect on hot standby conflicts. + * + * If allDbs is true, we look at all databases, though there is no need to + * include WALSender since this has no effect on hot standby conflicts. If + * allDbs is false, skip processes attached to other databases. * * This is never executed during recovery so there is no need to look at * KnownAssignedXids. @@ -2823,9 +2825,12 @@ GetRunningTransactionData(void) * We don't worry about updating other counters, we want to keep this as * simple as possible and leave GetSnapshotData() as the primary code for * that bookkeeping. + * + * inCommitOnly indicates getting the oldestActiveXid among the transactions + * in the commit critical section. */ TransactionId -GetOldestActiveTransactionId(void) +GetOldestActiveTransactionId(bool inCommitOnly, bool allDbs) { ProcArrayStruct *arrayP = procArray; TransactionId *other_xids = ProcGlobal->xids; @@ -2852,6 +2857,8 @@ GetOldestActiveTransactionId(void) for (index = 0; index < arrayP->numProcs; index++) { TransactionId xid; + int pgprocno = arrayP->pgprocnos[index]; + PGPROC *proc = &allProcs[pgprocno]; /* Fetch xid just once - see GetNewTransactionId */ xid = UINT32_ACCESS_ONCE(other_xids[index]); @@ -2859,6 +2866,13 @@ GetOldestActiveTransactionId(void) if (!TransactionIdIsNormal(xid)) continue; + if (inCommitOnly && + (proc->delayChkptFlags & DELAY_CHKPT_IN_COMMIT) == 0) + continue; + + if (!allDbs && proc->databaseId != MyDatabaseId) + continue; + if (TransactionIdPrecedes(xid, oldestRunningXid)) oldestRunningXid = xid; diff --git a/src/backend/storage/ipc/procsignal.c b/src/backend/storage/ipc/procsignal.c index a9bb540b55a..087821311cc 100644 --- a/src/backend/storage/ipc/procsignal.c +++ b/src/backend/storage/ipc/procsignal.c @@ -728,7 +728,11 @@ procsignal_sigusr1_handler(SIGNAL_ARGS) void SendCancelRequest(int backendPID, const uint8 *cancel_key, int cancel_key_len) { - Assert(backendPID != 0); + if (backendPID == 0) + { + ereport(LOG, (errmsg("invalid cancel request with PID 0"))); + return; + } /* * See if we have a matching backend. Reading the pss_pid and diff --git a/src/backend/storage/lmgr/generate-lwlocknames.pl b/src/backend/storage/lmgr/generate-lwlocknames.pl index 4441b7cba0c..cd3e43c448a 100644 --- a/src/backend/storage/lmgr/generate-lwlocknames.pl +++ b/src/backend/storage/lmgr/generate-lwlocknames.pl @@ -10,7 +10,6 @@ use Getopt::Long; my $output_path = '.'; my $lastlockidx = -1; -my $continue = "\n"; GetOptions('outdir:s' => \$output_path); @@ -28,18 +27,24 @@ print $h "/* there is deliberately not an #ifndef LWLOCKNAMES_H here */\n\n"; # -# First, record the predefined LWLocks listed in wait_event_names.txt. We'll -# cross-check those with the ones in lwlocklist.h. +# First, record the predefined LWLocks and built-in tranches listed in +# wait_event_names.txt. We'll cross-check those with the ones in lwlocklist.h. # +my @wait_event_tranches; my @wait_event_lwlocks; my $record_lwlocks = 0; +my $in_tranches = 0; while (<$wait_event_names>) { chomp; # Check for end marker. - last if /^# END OF PREDEFINED LWLOCKS/; + if (/^# END OF PREDEFINED LWLOCKS/) + { + $in_tranches = 1; + next; + } # Skip comments and empty lines. next if /^#/; @@ -55,13 +60,29 @@ while (<$wait_event_names>) # Go to the next line if we are not yet recording LWLocks. next if not $record_lwlocks; + # Stop recording if we reach another section. + last if /^Section:/; + # Record the LWLock. (my $waiteventname, my $waitevendocsentence) = split(/\t/, $_); - push(@wait_event_lwlocks, $waiteventname); + + if ($in_tranches) + { + push(@wait_event_tranches, $waiteventname); + } + else + { + push(@wait_event_lwlocks, $waiteventname); + } } +# +# While gathering the list of predefined LWLocks, cross-check the lists in +# lwlocklist.h with the wait events we just recorded. +# my $in_comment = 0; -my $i = 0; +my $lwlock_count = 0; +my $tranche_count = 0; while (<$lwlocklist>) { chomp; @@ -82,40 +103,72 @@ while (<$lwlocklist>) next; } - die "unable to parse lwlocklist.h line \"$_\"" - unless /^PG_LWLOCK\((\d+),\s+(\w+)\)$/; + # + # Gather list of predefined LWLocks and cross-check with the wait events. + # + if (/^PG_LWLOCK\((\d+),\s+(\w+)\)$/) + { + my ($lockidx, $lockname) = ($1, $2); - (my $lockidx, my $lockname) = ($1, $2); + die "lwlocklist.h not in order" if $lockidx < $lastlockidx; + die "lwlocklist.h has duplicates" if $lockidx == $lastlockidx; - die "lwlocklist.h not in order" if $lockidx < $lastlockidx; - die "lwlocklist.h has duplicates" if $lockidx == $lastlockidx; + die "$lockname defined in lwlocklist.h but missing from " + . "wait_event_names.txt" + if $lwlock_count >= scalar @wait_event_lwlocks; + die "lists of predefined LWLocks do not match (first mismatch at " + . "$wait_event_lwlocks[$lwlock_count] in wait_event_names.txt and " + . "$lockname in lwlocklist.h)" + if $wait_event_lwlocks[$lwlock_count] ne $lockname; - die "$lockname defined in lwlocklist.h but missing from " - . "wait_event_names.txt" - if $i >= scalar @wait_event_lwlocks; - die "lists of predefined LWLocks do not match (first mismatch at " - . "$wait_event_lwlocks[$i] in wait_event_names.txt and $lockname in " - . "lwlocklist.h)" - if $wait_event_lwlocks[$i] ne $lockname; - $i++; + $lwlock_count++; - while ($lastlockidx < $lockidx - 1) + while ($lastlockidx < $lockidx - 1) + { + ++$lastlockidx; + } + $lastlockidx = $lockidx; + + # Add a "Lock" suffix to each lock name, as the C code depends on that. + printf $h "#define %-32s (&MainLWLockArray[$lockidx].lock)\n", + $lockname . "Lock"; + + next; + } + + # + # Cross-check the built-in LWLock tranches with the wait events. + # + if (/^PG_LWLOCKTRANCHE\((\w+),\s+(\w+)\)$/) { - ++$lastlockidx; - $continue = ",\n"; + my ($tranche_id, $tranche_name) = ($1, $2); + + die "$tranche_name defined in lwlocklist.h but missing from " + . "wait_event_names.txt" + if $tranche_count >= scalar @wait_event_tranches; + die + "lists of built-in LWLock tranches do not match (first mismatch at " + . "$wait_event_tranches[$tranche_count] in wait_event_names.txt and " + . "$tranche_name in lwlocklist.h)" + if $wait_event_tranches[$tranche_count] ne $tranche_name; + + $tranche_count++; + + next; } - $lastlockidx = $lockidx; - $continue = ",\n"; - # Add a "Lock" suffix to each lock name, as the C code depends on that - printf $h "#define %-32s (&MainLWLockArray[$lockidx].lock)\n", - $lockname . "Lock"; + die "unable to parse lwlocklist.h line \"$_\""; } die - "$wait_event_lwlocks[$i] defined in wait_event_names.txt but missing from " - . "lwlocklist.h" - if $i < scalar @wait_event_lwlocks; + "$wait_event_lwlocks[$lwlock_count] defined in wait_event_names.txt but " + . " missing from lwlocklist.h" + if $lwlock_count < scalar @wait_event_lwlocks; + +die + "$wait_event_tranches[$tranche_count] defined in wait_event_names.txt but " + . "missing from lwlocklist.h" + if $tranche_count < scalar @wait_event_tranches; print $h "\n"; printf $h "#define NUM_INDIVIDUAL_LWLOCKS %s\n", $lastlockidx + 1; diff --git a/src/backend/storage/lmgr/lwlock.c b/src/backend/storage/lmgr/lwlock.c index 46f44bc4511..ec9c345ffdf 100644 --- a/src/backend/storage/lmgr/lwlock.c +++ b/src/backend/storage/lmgr/lwlock.c @@ -122,9 +122,8 @@ StaticAssertDecl((LW_VAL_EXCLUSIVE & LW_FLAG_MASK) == 0, * own tranche. We absorb the names of these tranches from there into * BuiltinTrancheNames here. * - * 2. There are some predefined tranches for built-in groups of locks. - * These are listed in enum BuiltinTrancheIds in lwlock.h, and their names - * appear in BuiltinTrancheNames[] below. + * 2. There are some predefined tranches for built-in groups of locks defined + * in lwlocklist.h. We absorb the names of these tranches, too. * * 3. Extensions can create new tranches, via either RequestNamedLWLockTranche * or LWLockRegisterTranche. The names of these that are known in the current @@ -135,49 +134,10 @@ StaticAssertDecl((LW_VAL_EXCLUSIVE & LW_FLAG_MASK) == 0, */ static const char *const BuiltinTrancheNames[] = { #define PG_LWLOCK(id, lockname) [id] = CppAsString(lockname), +#define PG_LWLOCKTRANCHE(id, lockname) [LWTRANCHE_##id] = CppAsString(lockname), #include "storage/lwlocklist.h" #undef PG_LWLOCK - [LWTRANCHE_XACT_BUFFER] = "XactBuffer", - [LWTRANCHE_COMMITTS_BUFFER] = "CommitTsBuffer", - [LWTRANCHE_SUBTRANS_BUFFER] = "SubtransBuffer", - [LWTRANCHE_MULTIXACTOFFSET_BUFFER] = "MultiXactOffsetBuffer", - [LWTRANCHE_MULTIXACTMEMBER_BUFFER] = "MultiXactMemberBuffer", - [LWTRANCHE_NOTIFY_BUFFER] = "NotifyBuffer", - [LWTRANCHE_SERIAL_BUFFER] = "SerialBuffer", - [LWTRANCHE_WAL_INSERT] = "WALInsert", - [LWTRANCHE_BUFFER_CONTENT] = "BufferContent", - [LWTRANCHE_REPLICATION_ORIGIN_STATE] = "ReplicationOriginState", - [LWTRANCHE_REPLICATION_SLOT_IO] = "ReplicationSlotIO", - [LWTRANCHE_LOCK_FASTPATH] = "LockFastPath", - [LWTRANCHE_BUFFER_MAPPING] = "BufferMapping", - [LWTRANCHE_LOCK_MANAGER] = "LockManager", - [LWTRANCHE_PREDICATE_LOCK_MANAGER] = "PredicateLockManager", - [LWTRANCHE_PARALLEL_HASH_JOIN] = "ParallelHashJoin", - [LWTRANCHE_PARALLEL_BTREE_SCAN] = "ParallelBtreeScan", - [LWTRANCHE_PARALLEL_QUERY_DSA] = "ParallelQueryDSA", - [LWTRANCHE_PER_SESSION_DSA] = "PerSessionDSA", - [LWTRANCHE_PER_SESSION_RECORD_TYPE] = "PerSessionRecordType", - [LWTRANCHE_PER_SESSION_RECORD_TYPMOD] = "PerSessionRecordTypmod", - [LWTRANCHE_SHARED_TUPLESTORE] = "SharedTupleStore", - [LWTRANCHE_SHARED_TIDBITMAP] = "SharedTidBitmap", - [LWTRANCHE_PARALLEL_APPEND] = "ParallelAppend", - [LWTRANCHE_PER_XACT_PREDICATE_LIST] = "PerXactPredicateList", - [LWTRANCHE_PGSTATS_DSA] = "PgStatsDSA", - [LWTRANCHE_PGSTATS_HASH] = "PgStatsHash", - [LWTRANCHE_PGSTATS_DATA] = "PgStatsData", - [LWTRANCHE_LAUNCHER_DSA] = "LogicalRepLauncherDSA", - [LWTRANCHE_LAUNCHER_HASH] = "LogicalRepLauncherHash", - [LWTRANCHE_DSM_REGISTRY_DSA] = "DSMRegistryDSA", - [LWTRANCHE_DSM_REGISTRY_HASH] = "DSMRegistryHash", - [LWTRANCHE_COMMITTS_SLRU] = "CommitTsSLRU", - [LWTRANCHE_MULTIXACTOFFSET_SLRU] = "MultixactOffsetSLRU", - [LWTRANCHE_MULTIXACTMEMBER_SLRU] = "MultixactMemberSLRU", - [LWTRANCHE_NOTIFY_SLRU] = "NotifySLRU", - [LWTRANCHE_SERIAL_SLRU] = "SerialSLRU", - [LWTRANCHE_SUBTRANS_SLRU] = "SubtransSLRU", - [LWTRANCHE_XACT_SLRU] = "XactSLRU", - [LWTRANCHE_PARALLEL_VACUUM_DSA] = "ParallelVacuumDSA", - [LWTRANCHE_AIO_URING_COMPLETION] = "AioUringCompletion", +#undef PG_LWLOCKTRANCHE }; StaticAssertDecl(lengthof(BuiltinTrancheNames) == diff --git a/src/backend/tcop/backend_startup.c b/src/backend/tcop/backend_startup.c index ad0af5edc1f..14d5fc0b196 100644 --- a/src/backend/tcop/backend_startup.c +++ b/src/backend/tcop/backend_startup.c @@ -492,7 +492,7 @@ static int ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) { int32 len; - char *buf; + char *buf = NULL; ProtocolVersion proto; MemoryContext oldcontext; @@ -516,7 +516,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) * scanners, which may be less benign, but it's not really our job to * notice those.) */ - return STATUS_ERROR; + goto fail; } if (pq_getbytes(((char *) &len) + 1, 3) == EOF) @@ -526,7 +526,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) ereport(COMMERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("incomplete startup packet"))); - return STATUS_ERROR; + goto fail; } len = pg_ntoh32(len); @@ -538,7 +538,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) ereport(COMMERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("invalid length of startup packet"))); - return STATUS_ERROR; + goto fail; } /* @@ -554,7 +554,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) ereport(COMMERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION), errmsg("incomplete startup packet"))); - return STATUS_ERROR; + goto fail; } pq_endmsgread(); @@ -568,7 +568,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) { ProcessCancelRequestPacket(port, buf, len); /* Not really an error, but we don't want to proceed further */ - return STATUS_ERROR; + goto fail; } if (proto == NEGOTIATE_SSL_CODE && !ssl_done) @@ -607,14 +607,16 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) ereport(COMMERROR, (errcode_for_socket_access(), errmsg("failed to send SSL negotiation response: %m"))); - return STATUS_ERROR; /* close the connection */ + goto fail; /* close the connection */ } #ifdef USE_SSL if (SSLok == 'S' && secure_open_server(port) == -1) - return STATUS_ERROR; + goto fail; #endif + pfree(buf); + /* * At this point we should have no data already buffered. If we do, * it was received before we performed the SSL handshake, so it wasn't @@ -661,14 +663,16 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) ereport(COMMERROR, (errcode_for_socket_access(), errmsg("failed to send GSSAPI negotiation response: %m"))); - return STATUS_ERROR; /* close the connection */ + goto fail; /* close the connection */ } #ifdef ENABLE_GSS if (GSSok == 'G' && secure_open_gssapi(port) == -1) - return STATUS_ERROR; + goto fail; #endif + pfree(buf); + /* * At this point we should have no data already buffered. If we do, * it was received before we performed the GSS handshake, so it wasn't @@ -863,7 +867,16 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done) */ MemoryContextSwitchTo(oldcontext); + pfree(buf); + return STATUS_OK; + +fail: + /* be tidy, just to avoid Valgrind complaints */ + if (buf) + pfree(buf); + + return STATUS_ERROR; } /* diff --git a/src/backend/tcop/postgres.c b/src/backend/tcop/postgres.c index 2f8c3d5f918..0cecd464902 100644 --- a/src/backend/tcop/postgres.c +++ b/src/backend/tcop/postgres.c @@ -988,6 +988,7 @@ pg_plan_queries(List *querytrees, const char *query_string, int cursorOptions, stmt->stmt_location = query->stmt_location; stmt->stmt_len = query->stmt_len; stmt->queryId = query->queryId; + stmt->planOrigin = PLAN_STMT_INTERNAL; } else { diff --git a/src/backend/tcop/pquery.c b/src/backend/tcop/pquery.c index d1593f38b35..08791b8f75e 100644 --- a/src/backend/tcop/pquery.c +++ b/src/backend/tcop/pquery.c @@ -1350,24 +1350,15 @@ PortalRunMulti(Portal portal, PopActiveSnapshot(); /* - * If a query completion data was supplied, use it. Otherwise use the - * portal's query completion data. - * - * Exception: Clients expect INSERT/UPDATE/DELETE tags to have counts, so - * fake them with zeros. This can happen with DO INSTEAD rules if there - * is no replacement query of the same type as the original. We print "0 - * 0" here because technically there is no query of the matching tag type, - * and printing a non-zero count for a different query type seems wrong, - * e.g. an INSERT that does an UPDATE instead should not print "0 1" if - * one row was updated. See QueryRewrite(), step 3, for details. + * If a command tag was requested and we did not fill in a run-time- + * determined tag above, copy the parse-time tag from the Portal. (There + * might not be any tag there either, in edge cases such as empty prepared + * statements. That's OK.) */ - if (qc && qc->commandTag == CMDTAG_UNKNOWN) - { - if (portal->qc.commandTag != CMDTAG_UNKNOWN) - CopyQueryCompletion(qc, &portal->qc); - /* If the caller supplied a qc, we should have set it by now. */ - Assert(qc->commandTag != CMDTAG_UNKNOWN); - } + if (qc && + qc->commandTag == CMDTAG_UNKNOWN && + portal->qc.commandTag != CMDTAG_UNKNOWN) + CopyQueryCompletion(qc, &portal->qc); } /* diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c index 4c1faf5575c..4f4191b0ea6 100644 --- a/src/backend/tcop/utility.c +++ b/src/backend/tcop/utility.c @@ -1234,6 +1234,7 @@ ProcessUtilitySlow(ParseState *pstate, wrapper->utilityStmt = stmt; wrapper->stmt_location = pstmt->stmt_location; wrapper->stmt_len = pstmt->stmt_len; + wrapper->planOrigin = PLAN_STMT_INTERNAL; ProcessUtility(wrapper, queryString, @@ -1964,6 +1965,7 @@ ProcessUtilityForAlterTable(Node *stmt, AlterTableUtilityContext *context) wrapper->utilityStmt = stmt; wrapper->stmt_location = context->pstmt->stmt_location; wrapper->stmt_len = context->pstmt->stmt_len; + wrapper->planOrigin = PLAN_STMT_INTERNAL; ProcessUtility(wrapper, context->queryString, diff --git a/src/backend/tsearch/dict_ispell.c b/src/backend/tsearch/dict_ispell.c index 63bd193a78a..debfbf956cc 100644 --- a/src/backend/tsearch/dict_ispell.c +++ b/src/backend/tsearch/dict_ispell.c @@ -47,24 +47,30 @@ dispell_init(PG_FUNCTION_ARGS) if (strcmp(defel->defname, "dictfile") == 0) { + char *filename; + if (dictloaded) ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE), errmsg("multiple DictFile parameters"))); - NIImportDictionary(&(d->obj), - get_tsearch_config_filename(defGetString(defel), - "dict")); + filename = get_tsearch_config_filename(defGetString(defel), + "dict"); + NIImportDictionary(&(d->obj), filename); + pfree(filename); dictloaded = true; } else if (strcmp(defel->defname, "afffile") == 0) { + char *filename; + if (affloaded) ereport(ERROR, (errcode(ERRCODE_INVALID_PARAMETER_VALUE), errmsg("multiple AffFile parameters"))); - NIImportAffixes(&(d->obj), - get_tsearch_config_filename(defGetString(defel), - "affix")); + filename = get_tsearch_config_filename(defGetString(defel), + "affix"); + NIImportAffixes(&(d->obj), filename); + pfree(filename); affloaded = true; } else if (strcmp(defel->defname, "stopwords") == 0) diff --git a/src/backend/tsearch/dict_synonym.c b/src/backend/tsearch/dict_synonym.c index 0da5a9d6868..c2773eb01ad 100644 --- a/src/backend/tsearch/dict_synonym.c +++ b/src/backend/tsearch/dict_synonym.c @@ -199,6 +199,7 @@ skipline: } tsearch_readline_end(&trst); + pfree(filename); d->len = cur; qsort(d->syn, d->len, sizeof(Syn), compareSyn); diff --git a/src/backend/tsearch/dict_thesaurus.c b/src/backend/tsearch/dict_thesaurus.c index 1bebe36a691..1e6bbde1ca7 100644 --- a/src/backend/tsearch/dict_thesaurus.c +++ b/src/backend/tsearch/dict_thesaurus.c @@ -167,17 +167,17 @@ addWrd(DictThesaurus *d, char *b, char *e, uint32 idsubst, uint16 nwrd, uint16 p static void thesaurusRead(const char *filename, DictThesaurus *d) { + char *real_filename = get_tsearch_config_filename(filename, "ths"); tsearch_readline_state trst; uint32 idsubst = 0; bool useasis = false; char *line; - filename = get_tsearch_config_filename(filename, "ths"); - if (!tsearch_readline_begin(&trst, filename)) + if (!tsearch_readline_begin(&trst, real_filename)) ereport(ERROR, (errcode(ERRCODE_CONFIG_FILE_ERROR), errmsg("could not open thesaurus file \"%s\": %m", - filename))); + real_filename))); while ((line = tsearch_readline(&trst)) != NULL) { @@ -297,6 +297,7 @@ thesaurusRead(const char *filename, DictThesaurus *d) d->nsubst = idsubst; tsearch_readline_end(&trst); + pfree(real_filename); } static TheLexeme * diff --git a/src/backend/utils/activity/pgstat.c b/src/backend/utils/activity/pgstat.c index 8b57845e870..6bc91ce0dad 100644 --- a/src/backend/utils/activity/pgstat.c +++ b/src/backend/utils/activity/pgstat.c @@ -212,6 +212,11 @@ int pgstat_fetch_consistency = PGSTAT_FETCH_CONSISTENCY_CACHE; PgStat_LocalState pgStatLocal; +/* + * Track pending reports for fixed-numbered stats, used by + * pgstat_report_stat(). + */ +bool pgstat_report_fixed = false; /* ---------- * Local data @@ -370,7 +375,6 @@ static const PgStat_KindInfo pgstat_kind_builtin_infos[PGSTAT_KIND_BUILTIN_SIZE] .shared_data_off = offsetof(PgStatShared_Backend, stats), .shared_data_len = sizeof(((PgStatShared_Backend *) 0)->stats), - .have_static_pending_cb = pgstat_backend_have_pending_cb, .flush_static_cb = pgstat_backend_flush_cb, .reset_timestamp_cb = pgstat_backend_reset_timestamp_cb, }, @@ -437,7 +441,6 @@ static const PgStat_KindInfo pgstat_kind_builtin_infos[PGSTAT_KIND_BUILTIN_SIZE] .shared_data_len = sizeof(((PgStatShared_IO *) 0)->stats), .flush_static_cb = pgstat_io_flush_cb, - .have_static_pending_cb = pgstat_io_have_pending_cb, .init_shmem_cb = pgstat_io_init_shmem_cb, .reset_all_cb = pgstat_io_reset_all_cb, .snapshot_cb = pgstat_io_snapshot_cb, @@ -455,7 +458,6 @@ static const PgStat_KindInfo pgstat_kind_builtin_infos[PGSTAT_KIND_BUILTIN_SIZE] .shared_data_len = sizeof(((PgStatShared_SLRU *) 0)->stats), .flush_static_cb = pgstat_slru_flush_cb, - .have_static_pending_cb = pgstat_slru_have_pending_cb, .init_shmem_cb = pgstat_slru_init_shmem_cb, .reset_all_cb = pgstat_slru_reset_all_cb, .snapshot_cb = pgstat_slru_snapshot_cb, @@ -474,7 +476,6 @@ static const PgStat_KindInfo pgstat_kind_builtin_infos[PGSTAT_KIND_BUILTIN_SIZE] .init_backend_cb = pgstat_wal_init_backend_cb, .flush_static_cb = pgstat_wal_flush_cb, - .have_static_pending_cb = pgstat_wal_have_pending_cb, .init_shmem_cb = pgstat_wal_init_shmem_cb, .reset_all_cb = pgstat_wal_reset_all_cb, .snapshot_cb = pgstat_wal_snapshot_cb, @@ -708,29 +709,10 @@ pgstat_report_stat(bool force) } /* Don't expend a clock check if nothing to do */ - if (dlist_is_empty(&pgStatPending)) + if (dlist_is_empty(&pgStatPending) && + !pgstat_report_fixed) { - bool do_flush = false; - - /* Check for pending stats */ - for (PgStat_Kind kind = PGSTAT_KIND_MIN; kind <= PGSTAT_KIND_MAX; kind++) - { - const PgStat_KindInfo *kind_info = pgstat_get_kind_info(kind); - - if (!kind_info) - continue; - if (!kind_info->have_static_pending_cb) - continue; - - if (kind_info->have_static_pending_cb()) - { - do_flush = true; - break; - } - } - - if (!do_flush) - return 0; + return 0; } /* @@ -784,16 +766,19 @@ pgstat_report_stat(bool force) partial_flush |= pgstat_flush_pending_entries(nowait); /* flush of other stats kinds */ - for (PgStat_Kind kind = PGSTAT_KIND_MIN; kind <= PGSTAT_KIND_MAX; kind++) + if (pgstat_report_fixed) { - const PgStat_KindInfo *kind_info = pgstat_get_kind_info(kind); + for (PgStat_Kind kind = PGSTAT_KIND_MIN; kind <= PGSTAT_KIND_MAX; kind++) + { + const PgStat_KindInfo *kind_info = pgstat_get_kind_info(kind); - if (!kind_info) - continue; - if (!kind_info->flush_static_cb) - continue; + if (!kind_info) + continue; + if (!kind_info->flush_static_cb) + continue; - partial_flush |= kind_info->flush_static_cb(nowait); + partial_flush |= kind_info->flush_static_cb(nowait); + } } last_flush = now; @@ -815,6 +800,7 @@ pgstat_report_stat(bool force) } pending_since = 0; + pgstat_report_fixed = false; return 0; } diff --git a/src/backend/utils/activity/pgstat_backend.c b/src/backend/utils/activity/pgstat_backend.c index 51256277e8d..8714a85e2d9 100644 --- a/src/backend/utils/activity/pgstat_backend.c +++ b/src/backend/utils/activity/pgstat_backend.c @@ -66,6 +66,7 @@ pgstat_count_backend_io_op_time(IOObject io_object, IOContext io_context, io_time); backend_has_iostats = true; + pgstat_report_fixed = true; } void @@ -81,6 +82,7 @@ pgstat_count_backend_io_op(IOObject io_object, IOContext io_context, PendingBackendStats.pending_io.bytes[io_object][io_context][io_op] += bytes; backend_has_iostats = true; + pgstat_report_fixed = true; } /* @@ -302,18 +304,6 @@ pgstat_flush_backend(bool nowait, bits32 flags) } /* - * Check if there are any backend stats waiting for flush. - */ -bool -pgstat_backend_have_pending_cb(void) -{ - if (!pgstat_tracks_backend_bktype(MyBackendType)) - return false; - - return (backend_has_iostats || pgstat_backend_wal_have_pending()); -} - -/* * Callback to flush out locally pending backend statistics. * * If some stats could not be flushed due to lock contention, return true. diff --git a/src/backend/utils/activity/pgstat_io.c b/src/backend/utils/activity/pgstat_io.c index d8d26379a57..13ae57ed649 100644 --- a/src/backend/utils/activity/pgstat_io.c +++ b/src/backend/utils/activity/pgstat_io.c @@ -80,6 +80,7 @@ pgstat_count_io_op(IOObject io_object, IOContext io_context, IOOp io_op, pgstat_count_backend_io_op(io_object, io_context, io_op, cnt, bytes); have_iostats = true; + pgstat_report_fixed = true; } /* @@ -168,15 +169,6 @@ pgstat_fetch_stat_io(void) } /* - * Check if there any IO stats waiting for flush. - */ -bool -pgstat_io_have_pending_cb(void) -{ - return have_iostats; -} - -/* * Simpler wrapper of pgstat_io_flush_cb() */ void diff --git a/src/backend/utils/activity/pgstat_slru.c b/src/backend/utils/activity/pgstat_slru.c index b9e940dde45..7bd8744accb 100644 --- a/src/backend/utils/activity/pgstat_slru.c +++ b/src/backend/utils/activity/pgstat_slru.c @@ -144,15 +144,6 @@ pgstat_get_slru_index(const char *name) } /* - * Check if there are any SLRU stats entries waiting for flush. - */ -bool -pgstat_slru_have_pending_cb(void) -{ - return have_slrustats; -} - -/* * Flush out locally pending SLRU stats entries * * If nowait is true, this function returns false on lock failure. Otherwise @@ -247,6 +238,7 @@ get_slru_entry(int slru_idx) Assert((slru_idx >= 0) && (slru_idx < SLRU_NUM_ELEMENTS)); have_slrustats = true; + pgstat_report_fixed = true; return &pending_SLRUStats[slru_idx]; } diff --git a/src/backend/utils/activity/pgstat_wal.c b/src/backend/utils/activity/pgstat_wal.c index 16a1ecb4d90..0d04480d2f6 100644 --- a/src/backend/utils/activity/pgstat_wal.c +++ b/src/backend/utils/activity/pgstat_wal.c @@ -72,6 +72,15 @@ pgstat_fetch_stat_wal(void) } /* + * To determine whether WAL usage happened. + */ +static inline bool +pgstat_wal_have_pending(void) +{ + return pgWalUsage.wal_records != prevWalUsage.wal_records; +} + +/* * Calculate how much WAL usage counters have increased by subtracting the * previous counters from the current ones. * @@ -92,7 +101,7 @@ pgstat_wal_flush_cb(bool nowait) * This function can be called even if nothing at all has happened. Avoid * taking lock for nothing in that case. */ - if (!pgstat_wal_have_pending_cb()) + if (!pgstat_wal_have_pending()) return false; /* @@ -136,15 +145,6 @@ pgstat_wal_init_backend_cb(void) prevWalUsage = pgWalUsage; } -/* - * To determine whether WAL usage happened. - */ -bool -pgstat_wal_have_pending_cb(void) -{ - return pgWalUsage.wal_records != prevWalUsage.wal_records; -} - void pgstat_wal_init_shmem_cb(void *stats) { diff --git a/src/backend/utils/activity/wait_event_names.txt b/src/backend/utils/activity/wait_event_names.txt index 4da68312b5f..0be307d2ca0 100644 --- a/src/backend/utils/activity/wait_event_names.txt +++ b/src/backend/utils/activity/wait_event_names.txt @@ -356,9 +356,13 @@ AioWorkerSubmissionQueue "Waiting to access AIO worker submission queue." # # END OF PREDEFINED LWLOCKS (DO NOT CHANGE THIS LINE) # -# Predefined LWLocks (i.e., those declared in lwlocknames.h) must be listed -# in the section above and must be listed in the same order as in -# lwlocknames.h. Other LWLocks must be listed in the section below. +# Predefined LWLocks (i.e., those declared at the top of lwlocknames.h) must be +# listed in the section above and must be listed in the same order as in +# lwlocknames.h. +# +# Likewise, the built-in LWLock tranches (i.e., those declared at the bottom of +# lwlocknames.h) must be listed in the section below and must be listed in the +# same order as in lwlocknames.h. # XactBuffer "Waiting for I/O on a transaction status SLRU buffer." diff --git a/src/backend/utils/adt/bytea.c b/src/backend/utils/adt/bytea.c index 2e539c2504e..6e7b914c563 100644 --- a/src/backend/utils/adt/bytea.c +++ b/src/backend/utils/adt/bytea.c @@ -182,27 +182,21 @@ bytea_overlay(bytea *t1, bytea *t2, int sp, int sl) * * Non-printable characters must be passed as '\nnn' (octal) and are * converted to internal form. '\' must be passed as '\\'. - * ereport(ERROR, ...) if bad form. - * - * BUGS: - * The input is scanned twice. - * The error checking of input is minimal. */ Datum byteain(PG_FUNCTION_ARGS) { char *inputText = PG_GETARG_CSTRING(0); Node *escontext = fcinfo->context; + size_t len = strlen(inputText); + size_t bc; char *tp; char *rp; - int bc; bytea *result; /* Recognize hex input */ if (inputText[0] == '\\' && inputText[1] == 'x') { - size_t len = strlen(inputText); - bc = (len - 2) / 2 + VARHDRSZ; /* maximum possible length */ result = palloc(bc); bc = hex_decode_safe(inputText + 2, len - 2, VARDATA(result), @@ -213,33 +207,7 @@ byteain(PG_FUNCTION_ARGS) } /* Else, it's the traditional escaped style */ - for (bc = 0, tp = inputText; *tp != '\0'; bc++) - { - if (tp[0] != '\\') - tp++; - else if ((tp[0] == '\\') && - (tp[1] >= '0' && tp[1] <= '3') && - (tp[2] >= '0' && tp[2] <= '7') && - (tp[3] >= '0' && tp[3] <= '7')) - tp += 4; - else if ((tp[0] == '\\') && - (tp[1] == '\\')) - tp += 2; - else - { - /* - * one backslash, not followed by another or ### valid octal - */ - ereturn(escontext, (Datum) 0, - (errcode(ERRCODE_INVALID_TEXT_REPRESENTATION), - errmsg("invalid input syntax for type %s", "bytea"))); - } - } - - bc += VARHDRSZ; - - result = (bytea *) palloc(bc); - SET_VARSIZE(result, bc); + result = (bytea *) palloc(len + VARHDRSZ); /* maximum possible length */ tp = inputText; rp = VARDATA(result); @@ -247,21 +215,21 @@ byteain(PG_FUNCTION_ARGS) { if (tp[0] != '\\') *rp++ = *tp++; - else if ((tp[0] == '\\') && - (tp[1] >= '0' && tp[1] <= '3') && + else if ((tp[1] >= '0' && tp[1] <= '3') && (tp[2] >= '0' && tp[2] <= '7') && (tp[3] >= '0' && tp[3] <= '7')) { - bc = VAL(tp[1]); - bc <<= 3; - bc += VAL(tp[2]); - bc <<= 3; - *rp++ = bc + VAL(tp[3]); + int v; + + v = VAL(tp[1]); + v <<= 3; + v += VAL(tp[2]); + v <<= 3; + *rp++ = v + VAL(tp[3]); tp += 4; } - else if ((tp[0] == '\\') && - (tp[1] == '\\')) + else if (tp[1] == '\\') { *rp++ = '\\'; tp += 2; @@ -269,7 +237,7 @@ byteain(PG_FUNCTION_ARGS) else { /* - * We should never get here. The first pass should not allow it. + * one backslash, not followed by another or ### valid octal */ ereturn(escontext, (Datum) 0, (errcode(ERRCODE_INVALID_TEXT_REPRESENTATION), @@ -277,6 +245,9 @@ byteain(PG_FUNCTION_ARGS) } } + bc = rp - VARDATA(result); /* actual length */ + SET_VARSIZE(result, bc + VARHDRSZ); + PG_RETURN_BYTEA_P(result); } diff --git a/src/backend/utils/adt/jsonb_util.c b/src/backend/utils/adt/jsonb_util.c index c8b6c15e059..82b807d067a 100644 --- a/src/backend/utils/adt/jsonb_util.c +++ b/src/backend/utils/adt/jsonb_util.c @@ -277,22 +277,16 @@ compareJsonbContainers(JsonbContainer *a, JsonbContainer *b) else { /* - * It's safe to assume that the types differed, and that the va - * and vb values passed were set. - * - * If the two values were of the same container type, then there'd - * have been a chance to observe the variation in the number of - * elements/pairs (when processing WJB_BEGIN_OBJECT, say). They're - * either two heterogeneously-typed containers, or a container and - * some scalar type. - * - * We don't have to consider the WJB_END_ARRAY and WJB_END_OBJECT - * cases here, because we would have seen the corresponding - * WJB_BEGIN_ARRAY and WJB_BEGIN_OBJECT tokens first, and - * concluded that they don't match. + * It's not possible for one iterator to report end of array or + * object while the other one reports something else, because we + * would have detected a length mismatch when we processed the + * container-start tokens above. Likewise we can't see WJB_DONE + * from one but not the other. So we have two different-type + * containers, or a container and some scalar type, or two + * different scalar types. Sort on the basis of the type code. */ - Assert(ra != WJB_END_ARRAY && ra != WJB_END_OBJECT); - Assert(rb != WJB_END_ARRAY && rb != WJB_END_OBJECT); + Assert(ra != WJB_DONE && ra != WJB_END_ARRAY && ra != WJB_END_OBJECT); + Assert(rb != WJB_DONE && rb != WJB_END_ARRAY && rb != WJB_END_OBJECT); Assert(va.type != vb.type); Assert(va.type != jbvBinary); @@ -852,15 +846,20 @@ JsonbIteratorInit(JsonbContainer *container) * It is our job to expand the jbvBinary representation without bothering them * with it. However, clients should not take it upon themselves to touch array * or Object element/pair buffers, since their element/pair pointers are - * garbage. Also, *val will not be set when returning WJB_END_ARRAY or - * WJB_END_OBJECT, on the assumption that it's only useful to access values - * when recursing in. + * garbage. + * + * *val is not meaningful when the result is WJB_DONE, WJB_END_ARRAY or + * WJB_END_OBJECT. However, we set val->type = jbvNull in those cases, + * so that callers may assume that val->type is always well-defined. */ JsonbIteratorToken JsonbIteratorNext(JsonbIterator **it, JsonbValue *val, bool skipNested) { if (*it == NULL) + { + val->type = jbvNull; return WJB_DONE; + } /* * When stepping into a nested container, we jump back here to start @@ -898,6 +897,7 @@ recurse: * nesting). */ *it = freeAndGetParent(*it); + val->type = jbvNull; return WJB_END_ARRAY; } @@ -951,6 +951,7 @@ recurse: * of nesting). */ *it = freeAndGetParent(*it); + val->type = jbvNull; return WJB_END_OBJECT; } else @@ -995,8 +996,10 @@ recurse: return WJB_VALUE; } - elog(ERROR, "invalid iterator state"); - return -1; + elog(ERROR, "invalid jsonb iterator state"); + /* satisfy compilers that don't know that elog(ERROR) doesn't return */ + val->type = jbvNull; + return WJB_DONE; } /* diff --git a/src/backend/utils/adt/pg_upgrade_support.c b/src/backend/utils/adt/pg_upgrade_support.c index d44f8c262ba..a4f8b4faa90 100644 --- a/src/backend/utils/adt/pg_upgrade_support.c +++ b/src/backend/utils/adt/pg_upgrade_support.c @@ -21,6 +21,7 @@ #include "commands/extension.h" #include "miscadmin.h" #include "replication/logical.h" +#include "replication/logicallauncher.h" #include "replication/origin.h" #include "replication/worker_internal.h" #include "storage/lmgr.h" @@ -410,3 +411,21 @@ binary_upgrade_replorigin_advance(PG_FUNCTION_ARGS) PG_RETURN_VOID(); } + +/* + * binary_upgrade_create_conflict_detection_slot + * + * Create a replication slot to retain information necessary for conflict + * detection such as dead tuples, commit timestamps, and origins. + */ +Datum +binary_upgrade_create_conflict_detection_slot(PG_FUNCTION_ARGS) +{ + CHECK_IS_BINARY_UPGRADE; + + CreateConflictDetectionSlot(); + + ReplicationSlotRelease(); + + PG_RETURN_VOID(); +} diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c index ce6a626eba2..17fbfa9b410 100644 --- a/src/backend/utils/adt/selfuncs.c +++ b/src/backend/utils/adt/selfuncs.c @@ -3798,18 +3798,25 @@ estimate_multivariate_bucketsize(PlannerInfo *root, RelOptInfo *inner, List *hashclauses, Selectivity *innerbucketsize) { - List *clauses = list_copy(hashclauses); - List *otherclauses = NIL; - double ndistinct = 1.0; + List *clauses; + List *otherclauses; + double ndistinct; if (list_length(hashclauses) <= 1) - + { /* * Nothing to do for a single clause. Could we employ univariate * extended stat here? */ return hashclauses; + } + /* "clauses" is the list of hashclauses we've not dealt with yet */ + clauses = list_copy(hashclauses); + /* "otherclauses" holds clauses we are going to return to caller */ + otherclauses = NIL; + /* current estimate of ndistinct */ + ndistinct = 1.0; while (clauses != NIL) { ListCell *lc; @@ -3874,12 +3881,13 @@ estimate_multivariate_bucketsize(PlannerInfo *root, RelOptInfo *inner, group_rel = root->simple_rel_array[relid]; } else if (group_relid != relid) - + { /* * Being in the group forming state we don't need other * clauses. */ continue; + } /* * We're going to add the new clause to the varinfos list. We diff --git a/src/backend/utils/adt/tid.c b/src/backend/utils/adt/tid.c index 1b0df111717..39dab3e42df 100644 --- a/src/backend/utils/adt/tid.c +++ b/src/backend/utils/adt/tid.c @@ -84,7 +84,7 @@ tidin(PG_FUNCTION_ARGS) /* * Cope with possibility that unsigned long is wider than BlockNumber, in * which case strtoul will not raise an error for some values that are out - * of the range of BlockNumber. (See similar code in oidin().) + * of the range of BlockNumber. (See similar code in uint32in_subr().) */ #if SIZEOF_LONG > 4 if (cvt != (unsigned long) blockNumber && diff --git a/src/backend/utils/adt/xml.c b/src/backend/utils/adt/xml.c index f7b731825fc..182e8f75db7 100644 --- a/src/backend/utils/adt/xml.c +++ b/src/backend/utils/adt/xml.c @@ -1769,7 +1769,7 @@ xml_doctype_in_content(const xmlChar *str) * xmloption_arg, but a DOCTYPE node in the input can force DOCUMENT mode). * * If parsed_nodes isn't NULL and we parse in CONTENT mode, the list - * of parsed nodes from the xmlParseInNodeContext call will be returned + * of parsed nodes from the xmlParseBalancedChunkMemory call will be returned * to *parsed_nodes. (It is caller's responsibility to free that.) * * Errors normally result in ereport(ERROR), but if escontext is an @@ -1795,6 +1795,7 @@ xml_parse(text *data, XmlOptionType xmloption_arg, PgXmlErrorContext *xmlerrcxt; volatile xmlParserCtxtPtr ctxt = NULL; volatile xmlDocPtr doc = NULL; + volatile int save_keep_blanks = -1; /* * This step looks annoyingly redundant, but we must do it to have a @@ -1822,7 +1823,6 @@ xml_parse(text *data, XmlOptionType xmloption_arg, PG_TRY(); { bool parse_as_document = false; - int options; int res_code; size_t count = 0; xmlChar *version = NULL; @@ -1853,18 +1853,6 @@ xml_parse(text *data, XmlOptionType xmloption_arg, parse_as_document = true; } - /* - * Select parse options. - * - * Note that here we try to apply DTD defaults (XML_PARSE_DTDATTR) - * according to SQL/XML:2008 GR 10.16.7.d: 'Default values defined by - * internal DTD are applied'. As for external DTDs, we try to support - * them too (see SQL/XML:2008 GR 10.16.7.e), but that doesn't really - * happen because xmlPgEntityLoader prevents it. - */ - options = XML_PARSE_NOENT | XML_PARSE_DTDATTR - | (preserve_whitespace ? 0 : XML_PARSE_NOBLANKS); - /* initialize output parameters */ if (parsed_xmloptiontype != NULL) *parsed_xmloptiontype = parse_as_document ? XMLOPTION_DOCUMENT : @@ -1874,11 +1862,26 @@ xml_parse(text *data, XmlOptionType xmloption_arg, if (parse_as_document) { + int options; + + /* set up parser context used by xmlCtxtReadDoc */ ctxt = xmlNewParserCtxt(); if (ctxt == NULL || xmlerrcxt->err_occurred) xml_ereport(xmlerrcxt, ERROR, ERRCODE_OUT_OF_MEMORY, "could not allocate parser context"); + /* + * Select parse options. + * + * Note that here we try to apply DTD defaults (XML_PARSE_DTDATTR) + * according to SQL/XML:2008 GR 10.16.7.d: 'Default values defined + * by internal DTD are applied'. As for external DTDs, we try to + * support them too (see SQL/XML:2008 GR 10.16.7.e), but that + * doesn't really happen because xmlPgEntityLoader prevents it. + */ + options = XML_PARSE_NOENT | XML_PARSE_DTDATTR + | (preserve_whitespace ? 0 : XML_PARSE_NOBLANKS); + doc = xmlCtxtReadDoc(ctxt, utf8string, NULL, /* no URL */ "UTF-8", @@ -1900,10 +1903,7 @@ xml_parse(text *data, XmlOptionType xmloption_arg, } else { - xmlNodePtr root; - xmlNodePtr oldroot PG_USED_FOR_ASSERTS_ONLY; - - /* set up document with empty root node to be the context node */ + /* set up document that xmlParseBalancedChunkMemory will add to */ doc = xmlNewDoc(version); if (doc == NULL || xmlerrcxt->err_occurred) xml_ereport(xmlerrcxt, ERROR, ERRCODE_OUT_OF_MEMORY, @@ -1916,43 +1916,22 @@ xml_parse(text *data, XmlOptionType xmloption_arg, "could not allocate XML document"); doc->standalone = standalone; - root = xmlNewNode(NULL, (const xmlChar *) "content-root"); - if (root == NULL || xmlerrcxt->err_occurred) - xml_ereport(xmlerrcxt, ERROR, ERRCODE_OUT_OF_MEMORY, - "could not allocate xml node"); - - /* - * This attaches root to doc, so we need not free it separately; - * and there can't yet be any old root to free. - */ - oldroot = xmlDocSetRootElement(doc, root); - Assert(oldroot == NULL); + /* set parse options --- have to do this the ugly way */ + save_keep_blanks = xmlKeepBlanksDefault(preserve_whitespace ? 1 : 0); /* allow empty content */ if (*(utf8string + count)) { - xmlNodePtr node_list = NULL; - xmlParserErrors res; - - res = xmlParseInNodeContext(root, - (char *) utf8string + count, - strlen((char *) utf8string + count), - options, - &node_list); - - if (res != XML_ERR_OK || xmlerrcxt->err_occurred) + res_code = xmlParseBalancedChunkMemory(doc, NULL, NULL, 0, + utf8string + count, + parsed_nodes); + if (res_code != 0 || xmlerrcxt->err_occurred) { - xmlFreeNodeList(node_list); xml_errsave(escontext, xmlerrcxt, ERRCODE_INVALID_XML_CONTENT, "invalid XML content"); goto fail; } - - if (parsed_nodes != NULL) - *parsed_nodes = node_list; - else - xmlFreeNodeList(node_list); } } @@ -1961,6 +1940,8 @@ fail: } PG_CATCH(); { + if (save_keep_blanks != -1) + xmlKeepBlanksDefault(save_keep_blanks); if (doc != NULL) xmlFreeDoc(doc); if (ctxt != NULL) @@ -1972,6 +1953,9 @@ fail: } PG_END_TRY(); + if (save_keep_blanks != -1) + xmlKeepBlanksDefault(save_keep_blanks); + if (ctxt != NULL) xmlFreeParserCtxt(ctxt); diff --git a/src/backend/utils/cache/evtcache.c b/src/backend/utils/cache/evtcache.c index ce596bf5638..b9d5a5998be 100644 --- a/src/backend/utils/cache/evtcache.c +++ b/src/backend/utils/cache/evtcache.c @@ -78,7 +78,6 @@ BuildEventTriggerCache(void) { HASHCTL ctl; HTAB *cache; - MemoryContext oldcontext; Relation rel; Relation irel; SysScanDesc scan; @@ -110,9 +109,6 @@ BuildEventTriggerCache(void) (Datum) 0); } - /* Switch to correct memory context. */ - oldcontext = MemoryContextSwitchTo(EventTriggerCacheContext); - /* Prevent the memory context from being nuked while we're rebuilding. */ EventTriggerCacheState = ETCS_REBUILD_STARTED; @@ -145,6 +141,7 @@ BuildEventTriggerCache(void) bool evttags_isnull; EventTriggerCacheEntry *entry; bool found; + MemoryContext oldcontext; /* Get next tuple. */ tup = systable_getnext_ordered(scan, ForwardScanDirection); @@ -171,6 +168,9 @@ BuildEventTriggerCache(void) else continue; + /* Switch to correct memory context. */ + oldcontext = MemoryContextSwitchTo(EventTriggerCacheContext); + /* Allocate new cache item. */ item = palloc0(sizeof(EventTriggerCacheItem)); item->fnoid = form->evtfoid; @@ -188,6 +188,9 @@ BuildEventTriggerCache(void) entry->triggerlist = lappend(entry->triggerlist, item); else entry->triggerlist = list_make1(item); + + /* Restore previous memory context. */ + MemoryContextSwitchTo(oldcontext); } /* Done with pg_event_trigger scan. */ @@ -195,9 +198,6 @@ BuildEventTriggerCache(void) index_close(irel, AccessShareLock); relation_close(rel, AccessShareLock); - /* Restore previous memory context. */ - MemoryContextSwitchTo(oldcontext); - /* Install new cache. */ EventTriggerCache = cache; @@ -240,6 +240,8 @@ DecodeTextArrayToBitmapset(Datum array) } pfree(elems); + if ((Pointer) arr != DatumGetPointer(array)) + pfree(arr); return bms; } diff --git a/src/backend/utils/cache/plancache.c b/src/backend/utils/cache/plancache.c index 89a1c79e984..6661d2c6b73 100644 --- a/src/backend/utils/cache/plancache.c +++ b/src/backend/utils/cache/plancache.c @@ -463,8 +463,7 @@ CompleteCachedPlan(CachedPlanSource *plansource, /* * Save the final parameter types (or other parameter specification data) - * into the source_context, as well as our other parameters. Also save - * the result tuple descriptor. + * into the source_context, as well as our other parameters. */ MemoryContextSwitchTo(source_context); @@ -480,9 +479,25 @@ CompleteCachedPlan(CachedPlanSource *plansource, plansource->parserSetupArg = parserSetupArg; plansource->cursor_options = cursor_options; plansource->fixed_result = fixed_result; - plansource->resultDesc = PlanCacheComputeResultDesc(querytree_list); + /* + * Also save the result tuple descriptor. PlanCacheComputeResultDesc may + * leak some cruft; normally we just accept that to save a copy step, but + * in USE_VALGRIND mode be tidy by running it in the caller's context. + */ +#ifdef USE_VALGRIND + MemoryContextSwitchTo(oldcxt); + plansource->resultDesc = PlanCacheComputeResultDesc(querytree_list); + if (plansource->resultDesc) + { + MemoryContextSwitchTo(source_context); + plansource->resultDesc = CreateTupleDescCopy(plansource->resultDesc); + MemoryContextSwitchTo(oldcxt); + } +#else + plansource->resultDesc = PlanCacheComputeResultDesc(querytree_list); MemoryContextSwitchTo(oldcxt); +#endif plansource->is_complete = true; plansource->is_valid = true; @@ -1283,6 +1298,7 @@ GetCachedPlan(CachedPlanSource *plansource, ParamListInfo boundParams, CachedPlan *plan = NULL; List *qlist; bool customplan; + ListCell *lc; /* Assert caller is doing things in a sane order */ Assert(plansource->magic == CACHEDPLANSOURCE_MAGIC); @@ -1385,6 +1401,13 @@ GetCachedPlan(CachedPlanSource *plansource, ParamListInfo boundParams, plan->is_saved = true; } + foreach(lc, plan->stmt_list) + { + PlannedStmt *pstmt = (PlannedStmt *) lfirst(lc); + + pstmt->planOrigin = customplan ? PLAN_STMT_CACHE_CUSTOM : PLAN_STMT_CACHE_GENERIC; + } + return plan; } diff --git a/src/backend/utils/cache/ts_cache.c b/src/backend/utils/cache/ts_cache.c index 18cccd778fd..e8ae53238d0 100644 --- a/src/backend/utils/cache/ts_cache.c +++ b/src/backend/utils/cache/ts_cache.c @@ -321,7 +321,9 @@ lookup_ts_dictionary_cache(Oid dictId) /* * Init method runs in dictionary's private memory context, and we - * make sure the options are stored there too + * make sure the options are stored there too. This typically + * results in a small amount of memory leakage, but it's not worth + * complicating the API for tmplinit functions to avoid it. */ oldcontext = MemoryContextSwitchTo(entry->dictCtx); diff --git a/src/backend/utils/cache/typcache.c b/src/backend/utils/cache/typcache.c index f9aec38a11f..6a347698edf 100644 --- a/src/backend/utils/cache/typcache.c +++ b/src/backend/utils/cache/typcache.c @@ -1171,9 +1171,6 @@ load_domaintype_info(TypeCacheEntry *typentry) elog(ERROR, "domain \"%s\" constraint \"%s\" has NULL conbin", NameStr(typTup->typname), NameStr(c->conname)); - /* Convert conbin to C string in caller context */ - constring = TextDatumGetCString(val); - /* Create the DomainConstraintCache object and context if needed */ if (dcc == NULL) { @@ -1189,9 +1186,8 @@ load_domaintype_info(TypeCacheEntry *typentry) dcc->dccRefCount = 0; } - /* Create node trees in DomainConstraintCache's context */ - oldcxt = MemoryContextSwitchTo(dcc->dccContext); - + /* Convert conbin to a node tree, still in caller's context */ + constring = TextDatumGetCString(val); check_expr = (Expr *) stringToNode(constring); /* @@ -1206,10 +1202,13 @@ load_domaintype_info(TypeCacheEntry *typentry) */ check_expr = expression_planner(check_expr); + /* Create only the minimally needed stuff in dccContext */ + oldcxt = MemoryContextSwitchTo(dcc->dccContext); + r = makeNode(DomainConstraintState); r->constrainttype = DOM_CONSTRAINT_CHECK; r->name = pstrdup(NameStr(c->conname)); - r->check_expr = check_expr; + r->check_expr = copyObject(check_expr); r->check_exprstate = NULL; MemoryContextSwitchTo(oldcxt); diff --git a/src/backend/utils/hash/dynahash.c b/src/backend/utils/hash/dynahash.c index 1ad155d446e..81da03629f0 100644 --- a/src/backend/utils/hash/dynahash.c +++ b/src/backend/utils/hash/dynahash.c @@ -22,10 +22,11 @@ * lookup key's hash value as a partition number --- this will work because * of the way calc_bucket() maps hash values to bucket numbers. * - * For hash tables in shared memory, the memory allocator function should - * match malloc's semantics of returning NULL on failure. For hash tables - * in local memory, we typically use palloc() which will throw error on - * failure. The code in this file has to cope with both cases. + * The memory allocator function should match malloc's semantics of returning + * NULL on failure. (This is essential for hash tables in shared memory. + * For hash tables in local memory, we used to use palloc() which will throw + * error on failure; but we no longer do, so it's untested whether this + * module will still cope with that behavior.) * * dynahash.c provides support for these types of lookup keys: * @@ -98,6 +99,7 @@ #include "access/xact.h" #include "common/hashfn.h" +#include "lib/ilist.h" #include "port/pg_bitutils.h" #include "storage/shmem.h" #include "storage/spin.h" @@ -195,6 +197,7 @@ struct HASHHDR long ssize; /* segment size --- must be power of 2 */ int sshift; /* segment shift = log2(ssize) */ int nelem_alloc; /* number of entries to allocate at once */ + bool isfixed; /* if true, don't enlarge */ #ifdef HASH_STATISTICS @@ -227,7 +230,6 @@ struct HTAB MemoryContext hcxt; /* memory context if default allocator used */ char *tabname; /* table name (for error messages) */ bool isshared; /* true if table is in shared memory */ - bool isfixed; /* if true, don't enlarge */ /* freezing a shared table isn't allowed, so we can keep state here */ bool frozen; /* true = no more inserts allowed */ @@ -236,6 +238,16 @@ struct HTAB Size keysize; /* hash key length in bytes */ long ssize; /* segment size --- must be power of 2 */ int sshift; /* segment shift = log2(ssize) */ + + /* + * In a USE_VALGRIND build, non-shared hashtables keep an slist chain of + * all the element blocks they have allocated. This pacifies Valgrind, + * which would otherwise often claim that the element blocks are "possibly + * lost" for lack of any non-interior pointers to their starts. + */ +#ifdef USE_VALGRIND + slist_head element_blocks; +#endif }; /* @@ -618,8 +630,10 @@ hash_create(const char *tabname, long nelem, const HASHCTL *info, int flags) } } + /* Set isfixed if requested, but not till after we build initial entries */ if (flags & HASH_FIXED_SIZE) - hashp->isfixed = true; + hctl->isfixed = true; + return hashp; } @@ -644,6 +658,8 @@ hdefault(HTAB *hashp) hctl->ssize = DEF_SEGSIZE; hctl->sshift = DEF_SEGSIZE_SHIFT; + hctl->isfixed = false; /* can be enlarged */ + #ifdef HASH_STATISTICS hctl->accesses = hctl->collisions = 0; #endif @@ -1708,23 +1724,51 @@ element_alloc(HTAB *hashp, int nelem, int freelist_idx) { HASHHDR *hctl = hashp->hctl; Size elementSize; + Size requestSize; + char *allocedBlock; HASHELEMENT *firstElement; HASHELEMENT *tmpElement; HASHELEMENT *prevElement; int i; - if (hashp->isfixed) + if (hctl->isfixed) return false; /* Each element has a HASHELEMENT header plus user data. */ elementSize = MAXALIGN(sizeof(HASHELEMENT)) + MAXALIGN(hctl->entrysize); + requestSize = nelem * elementSize; + + /* Add space for slist_node list link if we need one. */ +#ifdef USE_VALGRIND + if (!hashp->isshared) + requestSize += MAXALIGN(sizeof(slist_node)); +#endif + + /* Allocate the memory. */ CurrentDynaHashCxt = hashp->hcxt; - firstElement = (HASHELEMENT *) hashp->alloc(nelem * elementSize); + allocedBlock = hashp->alloc(requestSize); - if (!firstElement) + if (!allocedBlock) return false; + /* + * If USE_VALGRIND, each allocated block of elements of a non-shared + * hashtable is chained into a list, so that Valgrind won't think it's + * been leaked. + */ +#ifdef USE_VALGRIND + if (hashp->isshared) + firstElement = (HASHELEMENT *) allocedBlock; + else + { + slist_push_head(&hashp->element_blocks, (slist_node *) allocedBlock); + firstElement = (HASHELEMENT *) (allocedBlock + MAXALIGN(sizeof(slist_node))); + } +#else + firstElement = (HASHELEMENT *) allocedBlock; +#endif + /* prepare to link all the new entries into the freelist */ prevElement = NULL; tmpElement = firstElement; diff --git a/src/backend/utils/init/miscinit.c b/src/backend/utils/init/miscinit.c index 43b4dbccc3d..65d8cbfaed5 100644 --- a/src/backend/utils/init/miscinit.c +++ b/src/backend/utils/init/miscinit.c @@ -1183,7 +1183,6 @@ UnlinkLockFiles(int status, Datum arg) /* Should we complain if the unlink fails? */ } /* Since we're about to exit, no need to reclaim storage */ - lock_files = NIL; /* * Lock file removal should always be the last externally visible action diff --git a/src/backend/utils/init/postinit.c b/src/backend/utils/init/postinit.c index c86ceefda94..641e535a73c 100644 --- a/src/backend/utils/init/postinit.c +++ b/src/backend/utils/init/postinit.c @@ -417,12 +417,11 @@ CheckMyDatabase(const char *name, bool am_superuser, bool override_allow_connect datum = SysCacheGetAttrNotNull(DATABASEOID, tup, Anum_pg_database_datctype); ctype = TextDatumGetCString(datum); - if (pg_perm_setlocale(LC_COLLATE, collate) == NULL) - ereport(FATAL, - (errmsg("database locale is incompatible with operating system"), - errdetail("The database was initialized with LC_COLLATE \"%s\", " - " which is not recognized by setlocale().", collate), - errhint("Recreate the database with another locale or install the missing locale."))); + /* + * Historcally, we set LC_COLLATE from datcollate, as well. That's no + * longer necessary because all collation behavior is handled through + * pg_locale_t. + */ if (pg_perm_setlocale(LC_CTYPE, ctype) == NULL) ereport(FATAL, diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c index 667df448732..e404c345e6e 100644 --- a/src/backend/utils/misc/guc.c +++ b/src/backend/utils/misc/guc.c @@ -249,6 +249,7 @@ static void reapply_stacked_values(struct config_generic *variable, const char *curvalue, GucContext curscontext, GucSource cursource, Oid cursrole); +static void free_placeholder(struct config_string *pHolder); static bool validate_option_array_item(const char *name, const char *value, bool skipIfNoPermissions); static void write_auto_conf_file(int fd, const char *filename, ConfigVariable *head); @@ -4722,8 +4723,13 @@ AlterSystemSetConfigFile(AlterSystemStmt *altersysstmt) * the config file cannot cause postmaster start to fail, so we * don't have to be too tense about possibly installing a bad * value.) + * + * As an exception, we skip this check if this is a RESET command + * for an unknown custom GUC, else there'd be no way for users to + * remove such settings with reserved prefixes. */ - (void) assignable_custom_variable_name(name, false, ERROR); + if (value || !valid_custom_variable_name(name)) + (void) assignable_custom_variable_name(name, false, ERROR); } /* @@ -5018,16 +5024,8 @@ define_custom_variable(struct config_generic *variable) set_config_sourcefile(name, pHolder->gen.sourcefile, pHolder->gen.sourceline); - /* - * Free up as much as we conveniently can of the placeholder structure. - * (This neglects any stack items, so it's possible for some memory to be - * leaked. Since this can only happen once per session per variable, it - * doesn't seem worth spending much code on.) - */ - set_string_field(pHolder, pHolder->variable, NULL); - set_string_field(pHolder, &pHolder->reset_val, NULL); - - guc_free(pHolder); + /* Now we can free the no-longer-referenced placeholder variable */ + free_placeholder(pHolder); } /* @@ -5127,6 +5125,25 @@ reapply_stacked_values(struct config_generic *variable, } /* + * Free up a no-longer-referenced placeholder GUC variable. + * + * This neglects any stack items, so it's possible for some memory to be + * leaked. Since this can only happen once per session per variable, it + * doesn't seem worth spending much code on. + */ +static void +free_placeholder(struct config_string *pHolder) +{ + /* Placeholders are always STRING type, so free their values */ + Assert(pHolder->gen.vartype == PGC_STRING); + set_string_field(pHolder, pHolder->variable, NULL); + set_string_field(pHolder, &pHolder->reset_val, NULL); + + guc_free(unconstify(char *, pHolder->gen.name)); + guc_free(pHolder); +} + +/* * Functions for extensions to call to define their custom GUC variables. */ void @@ -5286,9 +5303,7 @@ MarkGUCPrefixReserved(const char *className) /* * Check for existing placeholders. We must actually remove invalid - * placeholders, else future parallel worker startups will fail. (We - * don't bother trying to free associated memory, since this shouldn't - * happen often.) + * placeholders, else future parallel worker startups will fail. */ hash_seq_init(&status, guc_hashtab); while ((hentry = (GUCHashEntry *) hash_seq_search(&status)) != NULL) @@ -5312,6 +5327,8 @@ MarkGUCPrefixReserved(const char *className) NULL); /* Remove it from any lists it's in, too */ RemoveGUCFromLists(var); + /* And free it */ + free_placeholder((struct config_string *) var); } } @@ -6711,6 +6728,7 @@ validate_option_array_item(const char *name, const char *value, { struct config_generic *gconf; + bool reset_custom; /* * There are three cases to consider: @@ -6729,16 +6747,21 @@ validate_option_array_item(const char *name, const char *value, * it's assumed to be fully validated.) * * name is not known and can't be created as a placeholder. Throw error, - * unless skipIfNoPermissions is true, in which case return false. + * unless skipIfNoPermissions or reset_custom is true. If reset_custom is + * true, this is a RESET or RESET ALL operation for an unknown custom GUC + * with a reserved prefix, in which case we want to fall through to the + * placeholder case described in the preceding paragraph (else there'd be + * no way for users to remove them). Otherwise, return false. */ - gconf = find_option(name, true, skipIfNoPermissions, ERROR); - if (!gconf) + reset_custom = (!value && valid_custom_variable_name(name)); + gconf = find_option(name, true, skipIfNoPermissions || reset_custom, ERROR); + if (!gconf && !reset_custom) { /* not known, failed to make a placeholder */ return false; } - if (gconf->flags & GUC_CUSTOM_PLACEHOLDER) + if (!gconf || gconf->flags & GUC_CUSTOM_PLACEHOLDER) { /* * We cannot do any meaningful check on the value, so only permissions diff --git a/src/backend/utils/misc/ps_status.c b/src/backend/utils/misc/ps_status.c index e08b26e8c14..4df25944deb 100644 --- a/src/backend/utils/misc/ps_status.c +++ b/src/backend/utils/misc/ps_status.c @@ -100,6 +100,17 @@ static void flush_ps_display(void); static int save_argc; static char **save_argv; +/* + * Valgrind seems not to consider the global "environ" variable as a valid + * root pointer; so when we allocate a new environment array, it claims that + * data is leaked. To fix that, keep our own statically-allocated copy of the + * pointer. (Oddly, this doesn't seem to be a problem for "argv".) + */ +#if defined(PS_USE_CLOBBER_ARGV) && defined(USE_VALGRIND) +extern char **ps_status_new_environ; +char **ps_status_new_environ; +#endif + /* * Call this early in startup to save the original argc/argv values. @@ -206,6 +217,11 @@ save_ps_display_args(int argc, char **argv) } new_environ[i] = NULL; environ = new_environ; + + /* See notes about Valgrind above. */ +#ifdef USE_VALGRIND + ps_status_new_environ = new_environ; +#endif } /* diff --git a/src/backend/utils/mmgr/alignedalloc.c b/src/backend/utils/mmgr/alignedalloc.c index 7eea695de62..b1be7426914 100644 --- a/src/backend/utils/mmgr/alignedalloc.c +++ b/src/backend/utils/mmgr/alignedalloc.c @@ -45,6 +45,15 @@ AlignedAllocFree(void *pointer) GetMemoryChunkContext(unaligned)->name, chunk); #endif + /* + * Create a dummy vchunk covering the start of the unaligned chunk, but + * not overlapping the aligned chunk. This will be freed while pfree'ing + * the unaligned chunk, keeping Valgrind happy. Then when we return to + * the outer pfree, that will clean up the vchunk for the aligned chunk. + */ + VALGRIND_MEMPOOL_ALLOC(GetMemoryChunkContext(unaligned), unaligned, + (char *) pointer - (char *) unaligned); + /* Recursively pfree the unaligned chunk */ pfree(unaligned); } @@ -123,6 +132,15 @@ AlignedAllocRealloc(void *pointer, Size size, int flags) VALGRIND_MAKE_MEM_DEFINED(pointer, old_size); memcpy(newptr, pointer, Min(size, old_size)); + /* + * Create a dummy vchunk covering the start of the old unaligned chunk, + * but not overlapping the aligned chunk. This will be freed while + * pfree'ing the old unaligned chunk, keeping Valgrind happy. Then when + * we return to repalloc, it will move the vchunk for the aligned chunk. + */ + VALGRIND_MEMPOOL_ALLOC(ctx, unaligned, + (char *) pointer - (char *) unaligned); + pfree(unaligned); return newptr; diff --git a/src/backend/utils/mmgr/aset.c b/src/backend/utils/mmgr/aset.c index 666ecd8f78d..9ef109ca586 100644 --- a/src/backend/utils/mmgr/aset.c +++ b/src/backend/utils/mmgr/aset.c @@ -103,6 +103,8 @@ #define ALLOC_BLOCKHDRSZ MAXALIGN(sizeof(AllocBlockData)) #define ALLOC_CHUNKHDRSZ sizeof(MemoryChunk) +#define FIRST_BLOCKHDRSZ (MAXALIGN(sizeof(AllocSetContext)) + \ + ALLOC_BLOCKHDRSZ) typedef struct AllocBlockData *AllocBlock; /* forward reference */ @@ -458,6 +460,21 @@ AllocSetContextCreateInternal(MemoryContext parent, * we'd leak the header/initial block if we ereport in this stretch. */ + /* Create a vpool associated with the context */ + VALGRIND_CREATE_MEMPOOL(set, 0, false); + + /* + * Create a vchunk covering both the AllocSetContext struct and the keeper + * block's header. (Perhaps it would be more sensible for these to be two + * separate vchunks, but doing that seems to tickle bugs in some versions + * of Valgrind.) We must have these vchunks, and also a vchunk for each + * subsequently-added block header, so that Valgrind considers the + * pointers within them while checking for leaked memory. Note that + * Valgrind doesn't distinguish between these vchunks and those created by + * mcxt.c for the user-accessible-data chunks we allocate. + */ + VALGRIND_MEMPOOL_ALLOC(set, set, FIRST_BLOCKHDRSZ); + /* Fill in the initial block's block header */ block = KeeperBlock(set); block->aset = set; @@ -585,6 +602,14 @@ AllocSetReset(MemoryContext context) #ifdef CLOBBER_FREED_MEMORY wipe_mem(block, block->freeptr - ((char *) block)); #endif + + /* + * We need to free the block header's vchunk explicitly, although + * the user-data vchunks within will go away in the TRIM below. + * Otherwise Valgrind complains about leaked allocations. + */ + VALGRIND_MEMPOOL_FREE(set, block); + free(block); } block = next; @@ -592,6 +617,14 @@ AllocSetReset(MemoryContext context) Assert(context->mem_allocated == keepersize); + /* + * Instruct Valgrind to throw away all the vchunks associated with this + * context, except for the one covering the AllocSetContext and + * keeper-block header. This gets rid of the vchunks for whatever user + * data is getting discarded by the context reset. + */ + VALGRIND_MEMPOOL_TRIM(set, set, FIRST_BLOCKHDRSZ); + /* Reset block size allocation sequence, too */ set->nextBlockSize = set->initBlockSize; } @@ -648,6 +681,9 @@ AllocSetDelete(MemoryContext context) freelist->first_free = (AllocSetContext *) oldset->header.nextchild; freelist->num_free--; + /* Destroy the context's vpool --- see notes below */ + VALGRIND_DESTROY_MEMPOOL(oldset); + /* All that remains is to free the header/initial block */ free(oldset); } @@ -675,13 +711,24 @@ AllocSetDelete(MemoryContext context) #endif if (!IsKeeperBlock(set, block)) + { + /* As in AllocSetReset, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(set, block); free(block); + } block = next; } Assert(context->mem_allocated == keepersize); + /* + * Destroy the vpool. We don't seem to need to explicitly free the + * initial block's header vchunk, nor any user-data vchunks that Valgrind + * still knows about; they'll all go away automatically. + */ + VALGRIND_DESTROY_MEMPOOL(set); + /* Finally, free the context header, including the keeper block */ free(set); } @@ -716,6 +763,9 @@ AllocSetAllocLarge(MemoryContext context, Size size, int flags) if (block == NULL) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(set, block, ALLOC_BLOCKHDRSZ); + context->mem_allocated += blksize; block->aset = set; @@ -922,6 +972,9 @@ AllocSetAllocFromNewBlock(MemoryContext context, Size size, int flags, if (block == NULL) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(set, block, ALLOC_BLOCKHDRSZ); + context->mem_allocated += blksize; block->aset = set; @@ -1104,6 +1157,10 @@ AllocSetFree(void *pointer) #ifdef CLOBBER_FREED_MEMORY wipe_mem(block, block->freeptr - ((char *) block)); #endif + + /* As in AllocSetReset, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(set, block); + free(block); } else @@ -1184,6 +1241,7 @@ AllocSetRealloc(void *pointer, Size size, int flags) * realloc() to make the containing block bigger, or smaller, with * minimum space wastage. */ + AllocBlock newblock; Size chksize; Size blksize; Size oldblksize; @@ -1223,14 +1281,21 @@ AllocSetRealloc(void *pointer, Size size, int flags) blksize = chksize + ALLOC_BLOCKHDRSZ + ALLOC_CHUNKHDRSZ; oldblksize = block->endptr - ((char *) block); - block = (AllocBlock) realloc(block, blksize); - if (block == NULL) + newblock = (AllocBlock) realloc(block, blksize); + if (newblock == NULL) { /* Disallow access to the chunk header. */ VALGRIND_MAKE_MEM_NOACCESS(chunk, ALLOC_CHUNKHDRSZ); return MemoryContextAllocationFailure(&set->header, size, flags); } + /* + * Move the block-header vchunk explicitly. (mcxt.c will take care of + * moving the vchunk for the user data.) + */ + VALGRIND_MEMPOOL_CHANGE(set, block, newblock, ALLOC_BLOCKHDRSZ); + block = newblock; + /* updated separately, not to underflow when (oldblksize > blksize) */ set->header.mem_allocated -= oldblksize; set->header.mem_allocated += blksize; @@ -1294,7 +1359,7 @@ AllocSetRealloc(void *pointer, Size size, int flags) /* Ensure any padding bytes are marked NOACCESS. */ VALGRIND_MAKE_MEM_NOACCESS((char *) pointer + size, chksize - size); - /* Disallow access to the chunk header . */ + /* Disallow access to the chunk header. */ VALGRIND_MAKE_MEM_NOACCESS(chunk, ALLOC_CHUNKHDRSZ); return pointer; diff --git a/src/backend/utils/mmgr/bump.c b/src/backend/utils/mmgr/bump.c index f7a37d1b3e8..2805d55a2ec 100644 --- a/src/backend/utils/mmgr/bump.c +++ b/src/backend/utils/mmgr/bump.c @@ -45,7 +45,9 @@ #include "utils/memutils_memorychunk.h" #include "utils/memutils_internal.h" -#define Bump_BLOCKHDRSZ MAXALIGN(sizeof(BumpBlock)) +#define Bump_BLOCKHDRSZ MAXALIGN(sizeof(BumpBlock)) +#define FIRST_BLOCKHDRSZ (MAXALIGN(sizeof(BumpContext)) + \ + Bump_BLOCKHDRSZ) /* No chunk header unless built with MEMORY_CONTEXT_CHECKING */ #ifdef MEMORY_CONTEXT_CHECKING @@ -189,6 +191,12 @@ BumpContextCreate(MemoryContext parent, const char *name, Size minContextSize, * Avoid writing code that can fail between here and MemoryContextCreate; * we'd leak the header and initial block if we ereport in this stretch. */ + + /* See comments about Valgrind interactions in aset.c */ + VALGRIND_CREATE_MEMPOOL(set, 0, false); + /* This vchunk covers the BumpContext and the keeper block header */ + VALGRIND_MEMPOOL_ALLOC(set, set, FIRST_BLOCKHDRSZ); + dlist_init(&set->blocks); /* Fill in the initial block's block header */ @@ -262,6 +270,14 @@ BumpReset(MemoryContext context) BumpBlockFree(set, block); } + /* + * Instruct Valgrind to throw away all the vchunks associated with this + * context, except for the one covering the BumpContext and keeper-block + * header. This gets rid of the vchunks for whatever user data is getting + * discarded by the context reset. + */ + VALGRIND_MEMPOOL_TRIM(set, set, FIRST_BLOCKHDRSZ); + /* Reset block size allocation sequence, too */ set->nextBlockSize = set->initBlockSize; @@ -279,6 +295,10 @@ BumpDelete(MemoryContext context) { /* Reset to release all releasable BumpBlocks */ BumpReset(context); + + /* Destroy the vpool -- see notes in aset.c */ + VALGRIND_DESTROY_MEMPOOL(context); + /* And free the context header and keeper block */ free(context); } @@ -318,6 +338,9 @@ BumpAllocLarge(MemoryContext context, Size size, int flags) if (block == NULL) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(set, block, Bump_BLOCKHDRSZ); + context->mem_allocated += blksize; /* the block is completely full */ @@ -455,6 +478,9 @@ BumpAllocFromNewBlock(MemoryContext context, Size size, int flags, if (block == NULL) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(set, block, Bump_BLOCKHDRSZ); + context->mem_allocated += blksize; /* initialize the new block */ @@ -606,6 +632,9 @@ BumpBlockFree(BumpContext *set, BumpBlock *block) wipe_mem(block, ((char *) block->endptr - (char *) block)); #endif + /* As in aset.c, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(set, block); + free(block); } diff --git a/src/backend/utils/mmgr/generation.c b/src/backend/utils/mmgr/generation.c index 18679ad4f1e..cfafc9bf082 100644 --- a/src/backend/utils/mmgr/generation.c +++ b/src/backend/utils/mmgr/generation.c @@ -45,6 +45,8 @@ #define Generation_BLOCKHDRSZ MAXALIGN(sizeof(GenerationBlock)) #define Generation_CHUNKHDRSZ sizeof(MemoryChunk) +#define FIRST_BLOCKHDRSZ (MAXALIGN(sizeof(GenerationContext)) + \ + Generation_BLOCKHDRSZ) #define Generation_CHUNK_FRACTION 8 @@ -221,6 +223,12 @@ GenerationContextCreate(MemoryContext parent, * Avoid writing code that can fail between here and MemoryContextCreate; * we'd leak the header if we ereport in this stretch. */ + + /* See comments about Valgrind interactions in aset.c */ + VALGRIND_CREATE_MEMPOOL(set, 0, false); + /* This vchunk covers the GenerationContext and the keeper block header */ + VALGRIND_MEMPOOL_ALLOC(set, set, FIRST_BLOCKHDRSZ); + dlist_init(&set->blocks); /* Fill in the initial block's block header */ @@ -309,6 +317,14 @@ GenerationReset(MemoryContext context) GenerationBlockFree(set, block); } + /* + * Instruct Valgrind to throw away all the vchunks associated with this + * context, except for the one covering the GenerationContext and + * keeper-block header. This gets rid of the vchunks for whatever user + * data is getting discarded by the context reset. + */ + VALGRIND_MEMPOOL_TRIM(set, set, FIRST_BLOCKHDRSZ); + /* set it so new allocations to make use of the keeper block */ set->block = KeeperBlock(set); @@ -329,6 +345,10 @@ GenerationDelete(MemoryContext context) { /* Reset to release all releasable GenerationBlocks */ GenerationReset(context); + + /* Destroy the vpool -- see notes in aset.c */ + VALGRIND_DESTROY_MEMPOOL(context); + /* And free the context header and keeper block */ free(context); } @@ -365,6 +385,9 @@ GenerationAllocLarge(MemoryContext context, Size size, int flags) if (block == NULL) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(set, block, Generation_BLOCKHDRSZ); + context->mem_allocated += blksize; /* block with a single (used) chunk */ @@ -487,6 +510,9 @@ GenerationAllocFromNewBlock(MemoryContext context, Size size, int flags, if (block == NULL) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(set, block, Generation_BLOCKHDRSZ); + context->mem_allocated += blksize; /* initialize the new block */ @@ -677,6 +703,9 @@ GenerationBlockFree(GenerationContext *set, GenerationBlock *block) wipe_mem(block, block->blksize); #endif + /* As in aset.c, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(set, block); + free(block); } diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c index 15fa4d0a55e..47fd774c7d2 100644 --- a/src/backend/utils/mmgr/mcxt.c +++ b/src/backend/utils/mmgr/mcxt.c @@ -8,6 +8,23 @@ * context-type-specific operations via the function pointers in a * context's MemoryContextMethods struct. * + * A note about Valgrind support: when USE_VALGRIND is defined, we provide + * support for memory leak tracking at the allocation-unit level. Valgrind + * does leak detection by tracking allocated "chunks", which can be grouped + * into "pools". The "chunk" terminology is overloaded, since we use that + * word for our allocation units, and it's sometimes important to distinguish + * those from the Valgrind objects that describe them. To reduce confusion, + * let's use the terms "vchunk" and "vpool" for the Valgrind objects. + * + * We use a separate vpool for each memory context. The context-type-specific + * code is responsible for creating and deleting the vpools, and also for + * creating vchunks to cover its management data structures such as block + * headers. (There must be a vchunk that includes every pointer we want + * Valgrind to consider for leak-tracking purposes.) This module creates + * and deletes the vchunks that cover the caller-visible allocated chunks. + * However, the context-type-specific code must handle cleaning up those + * vchunks too during memory context reset operations. + * * * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group * Portions Copyright (c) 1994, Regents of the University of California @@ -418,8 +435,6 @@ MemoryContextResetOnly(MemoryContext context) context->methods->reset(context); context->isReset = true; - VALGRIND_DESTROY_MEMPOOL(context); - VALGRIND_CREATE_MEMPOOL(context, 0, false); } } @@ -526,8 +541,6 @@ MemoryContextDeleteOnly(MemoryContext context) context->ident = NULL; context->methods->delete_context(context); - - VALGRIND_DESTROY_MEMPOOL(context); } /* @@ -560,9 +573,7 @@ MemoryContextDeleteChildren(MemoryContext context) * the specified context, since that means it will automatically be freed * when no longer needed. * - * There is no API for deregistering a callback once registered. If you - * want it to not do anything anymore, adjust the state pointed to by its - * "arg" to indicate that. + * Note that callers can assume this cannot fail. */ void MemoryContextRegisterResetCallback(MemoryContext context, @@ -578,6 +589,41 @@ MemoryContextRegisterResetCallback(MemoryContext context, } /* + * MemoryContextUnregisterResetCallback + * Undo the effects of MemoryContextRegisterResetCallback. + * + * This can be used if a callback's effects are no longer required + * at some point before the context has been reset/deleted. It is the + * caller's responsibility to pfree the callback struct (if needed). + * + * An assertion failure occurs if the callback was not registered. + * We could alternatively define that case as a no-op, but that seems too + * likely to mask programming errors such as passing the wrong context. + */ +void +MemoryContextUnregisterResetCallback(MemoryContext context, + MemoryContextCallback *cb) +{ + MemoryContextCallback *prev, + *cur; + + Assert(MemoryContextIsValid(context)); + + for (prev = NULL, cur = context->reset_cbs; cur != NULL; + prev = cur, cur = cur->next) + { + if (cur != cb) + continue; + if (prev) + prev->next = cur->next; + else + context->reset_cbs = cur->next; + return; + } + Assert(false); +} + +/* * MemoryContextCallResetCallbacks * Internal function to call all registered callbacks for context. */ @@ -1137,8 +1183,6 @@ MemoryContextCreate(MemoryContext node, node->nextchild = NULL; node->allowInCritSection = false; } - - VALGRIND_CREATE_MEMPOOL(node, 0, false); } /* @@ -1421,7 +1465,13 @@ MemoryContextAllocAligned(MemoryContext context, void *unaligned; void *aligned; - /* wouldn't make much sense to waste that much space */ + /* + * Restrict alignto to ensure that it can fit into the "value" field of + * the redirection MemoryChunk, and that the distance back to the start of + * the unaligned chunk will fit into the space available for that. This + * isn't a limitation in practice, since it wouldn't make much sense to + * waste that much space. + */ Assert(alignto < (128 * 1024 * 1024)); /* ensure alignto is a power of 2 */ @@ -1458,10 +1508,15 @@ MemoryContextAllocAligned(MemoryContext context, alloc_size += 1; #endif - /* perform the actual allocation */ - unaligned = MemoryContextAllocExtended(context, alloc_size, flags); + /* + * Perform the actual allocation, but do not pass down MCXT_ALLOC_ZERO. + * This ensures that wasted bytes beyond the aligned chunk do not become + * DEFINED. + */ + unaligned = MemoryContextAllocExtended(context, alloc_size, + flags & ~MCXT_ALLOC_ZERO); - /* set the aligned pointer */ + /* compute the aligned pointer */ aligned = (void *) TYPEALIGN(alignto, (char *) unaligned + sizeof(MemoryChunk)); @@ -1489,12 +1544,23 @@ MemoryContextAllocAligned(MemoryContext context, set_sentinel(aligned, size); #endif - /* Mark the bytes before the redirection header as noaccess */ - VALGRIND_MAKE_MEM_NOACCESS(unaligned, - (char *) alignedchunk - (char *) unaligned); + /* + * MemoryContextAllocExtended marked the whole unaligned chunk as a + * vchunk. Undo that, instead making just the aligned chunk be a vchunk. + * This prevents Valgrind from complaining that the vchunk is possibly + * leaked, since only pointers to the aligned chunk will exist. + * + * After these calls, the aligned chunk will be marked UNDEFINED, and all + * the rest of the unaligned chunk (the redirection chunk header, the + * padding bytes before it, and any wasted trailing bytes) will be marked + * NOACCESS, which is what we want. + */ + VALGRIND_MEMPOOL_FREE(context, unaligned); + VALGRIND_MEMPOOL_ALLOC(context, aligned, size); - /* Disallow access to the redirection chunk header. */ - VALGRIND_MAKE_MEM_NOACCESS(alignedchunk, sizeof(MemoryChunk)); + /* Now zero (and make DEFINED) just the aligned chunk, if requested */ + if ((flags & MCXT_ALLOC_ZERO) != 0) + MemSetAligned(aligned, 0, size); return aligned; } @@ -1528,16 +1594,12 @@ void pfree(void *pointer) { #ifdef USE_VALGRIND - MemoryContextMethodID method = GetMemoryChunkMethodID(pointer); MemoryContext context = GetMemoryChunkContext(pointer); #endif MCXT_METHOD(pointer, free_p) (pointer); -#ifdef USE_VALGRIND - if (method != MCTX_ALIGNED_REDIRECT_ID) - VALGRIND_MEMPOOL_FREE(context, pointer); -#endif + VALGRIND_MEMPOOL_FREE(context, pointer); } /* @@ -1547,9 +1609,6 @@ pfree(void *pointer) void * repalloc(void *pointer, Size size) { -#ifdef USE_VALGRIND - MemoryContextMethodID method = GetMemoryChunkMethodID(pointer); -#endif #if defined(USE_ASSERT_CHECKING) || defined(USE_VALGRIND) MemoryContext context = GetMemoryChunkContext(pointer); #endif @@ -1572,10 +1631,7 @@ repalloc(void *pointer, Size size) */ ret = MCXT_METHOD(pointer, realloc) (pointer, size, 0); -#ifdef USE_VALGRIND - if (method != MCTX_ALIGNED_REDIRECT_ID) - VALGRIND_MEMPOOL_CHANGE(context, pointer, ret, size); -#endif + VALGRIND_MEMPOOL_CHANGE(context, pointer, ret, size); return ret; } diff --git a/src/backend/utils/mmgr/slab.c b/src/backend/utils/mmgr/slab.c index d32c0d318fb..0e35abcf5a0 100644 --- a/src/backend/utils/mmgr/slab.c +++ b/src/backend/utils/mmgr/slab.c @@ -377,6 +377,11 @@ SlabContextCreate(MemoryContext parent, * we'd leak the header if we ereport in this stretch. */ + /* See comments about Valgrind interactions in aset.c */ + VALGRIND_CREATE_MEMPOOL(slab, 0, false); + /* This vchunk covers the SlabContext only */ + VALGRIND_MEMPOOL_ALLOC(slab, slab, sizeof(SlabContext)); + /* Fill in SlabContext-specific header fields */ slab->chunkSize = (uint32) chunkSize; slab->fullChunkSize = (uint32) fullChunkSize; @@ -451,6 +456,10 @@ SlabReset(MemoryContext context) #ifdef CLOBBER_FREED_MEMORY wipe_mem(block, slab->blockSize); #endif + + /* As in aset.c, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(slab, block); + free(block); context->mem_allocated -= slab->blockSize; } @@ -467,11 +476,23 @@ SlabReset(MemoryContext context) #ifdef CLOBBER_FREED_MEMORY wipe_mem(block, slab->blockSize); #endif + + /* As in aset.c, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(slab, block); + free(block); context->mem_allocated -= slab->blockSize; } } + /* + * Instruct Valgrind to throw away all the vchunks associated with this + * context, except for the one covering the SlabContext. This gets rid of + * the vchunks for whatever user data is getting discarded by the context + * reset. + */ + VALGRIND_MEMPOOL_TRIM(slab, slab, sizeof(SlabContext)); + slab->curBlocklistIndex = 0; Assert(context->mem_allocated == 0); @@ -486,6 +507,10 @@ SlabDelete(MemoryContext context) { /* Reset to release all the SlabBlocks */ SlabReset(context); + + /* Destroy the vpool -- see notes in aset.c */ + VALGRIND_DESTROY_MEMPOOL(context); + /* And free the context header */ free(context); } @@ -567,6 +592,9 @@ SlabAllocFromNewBlock(MemoryContext context, Size size, int flags) if (unlikely(block == NULL)) return MemoryContextAllocationFailure(context, size, flags); + /* Make a vchunk covering the new block's header */ + VALGRIND_MEMPOOL_ALLOC(slab, block, Slab_BLOCKHDRSZ); + block->slab = slab; context->mem_allocated += slab->blockSize; @@ -795,6 +823,10 @@ SlabFree(void *pointer) #ifdef CLOBBER_FREED_MEMORY wipe_mem(block, slab->blockSize); #endif + + /* As in aset.c, free block-header vchunks explicitly */ + VALGRIND_MEMPOOL_FREE(slab, block); + free(block); slab->header.mem_allocated -= slab->blockSize; } |