diff options
Diffstat (limited to 'src')
57 files changed, 1242 insertions, 355 deletions
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c index b0891998b24..5553c20fee8 100644 --- a/src/backend/access/transam/xlog.c +++ b/src/backend/access/transam/xlog.c @@ -703,7 +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); @@ -5371,11 +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)); @@ -5618,6 +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 @@ -5646,25 +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; @@ -5672,7 +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; @@ -5680,7 +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: @@ -6325,6 +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. * diff --git a/src/backend/access/transam/xlogrecovery.c b/src/backend/access/transam/xlogrecovery.c index e8f3ba00caa..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); diff --git a/src/backend/catalog/system_views.sql b/src/backend/catalog/system_views.sql index f6eca09ee15..77c693f630e 100644 --- a/src/backend/catalog/system_views.sql +++ b/src/backend/catalog/system_views.sql @@ -1399,6 +1399,7 @@ CREATE VIEW pg_stat_subscription_stats AS ss.confl_insert_exists, ss.confl_update_origin_differs, ss.confl_update_exists, + ss.confl_update_deleted, ss.confl_update_missing, ss.confl_delete_origin_differs, ss.confl_delete_missing, diff --git a/src/backend/commands/analyze.c b/src/backend/commands/analyze.c index 7111d5d5334..40d66537ad7 100644 --- a/src/backend/commands/analyze.c +++ b/src/backend/commands/analyze.c @@ -690,8 +690,8 @@ do_analyze_rel(Relation onerel, const VacuumParams params, * only do it for inherited stats. (We're never called for not-inherited * stats on partitioned tables anyway.) * - * Reset the changes_since_analyze counter only if we analyzed all - * columns; otherwise, there is still work for auto-analyze to do. + * Reset the mod_since_analyze counter only if we analyzed all columns; + * otherwise, there is still work for auto-analyze to do. */ if (!inh) pgstat_report_analyze(onerel, totalrows, totaldeadrows, diff --git a/src/backend/executor/execReplication.c b/src/backend/executor/execReplication.c index f262e7a66f7..68184f5d671 100644 --- a/src/backend/executor/execReplication.c +++ b/src/backend/executor/execReplication.c @@ -14,12 +14,14 @@ #include "postgres.h" +#include "access/commit_ts.h" #include "access/genam.h" #include "access/gist.h" #include "access/relscan.h" #include "access/tableam.h" #include "access/transam.h" #include "access/xact.h" +#include "access/heapam.h" #include "catalog/pg_am_d.h" #include "commands/trigger.h" #include "executor/executor.h" @@ -36,7 +38,7 @@ static bool tuples_equal(TupleTableSlot *slot1, TupleTableSlot *slot2, - TypeCacheEntry **eq); + TypeCacheEntry **eq, Bitmapset *columns); /* * Setup a ScanKey for a search in the relation 'rel' for a tuple 'key' that @@ -221,7 +223,7 @@ retry: if (eq == NULL) eq = palloc0(sizeof(*eq) * outslot->tts_tupleDescriptor->natts); - if (!tuples_equal(outslot, searchslot, eq)) + if (!tuples_equal(outslot, searchslot, eq, NULL)) continue; } @@ -277,10 +279,13 @@ retry: /* * Compare the tuples in the slots by checking if they have equal values. + * + * If 'columns' is not null, only the columns specified within it will be + * considered for the equality check, ignoring all other columns. */ static bool tuples_equal(TupleTableSlot *slot1, TupleTableSlot *slot2, - TypeCacheEntry **eq) + TypeCacheEntry **eq, Bitmapset *columns) { int attrnum; @@ -306,6 +311,14 @@ tuples_equal(TupleTableSlot *slot1, TupleTableSlot *slot2, continue; /* + * Ignore columns that are not listed for checking. + */ + if (columns && + !bms_is_member(att->attnum - FirstLowInvalidHeapAttributeNumber, + columns)) + continue; + + /* * If one value is NULL and other is not, then they are certainly not * equal */ @@ -380,7 +393,7 @@ retry: /* Try to find the tuple */ while (table_scan_getnextslot(scan, ForwardScanDirection, scanslot)) { - if (!tuples_equal(scanslot, searchslot, eq)) + if (!tuples_equal(scanslot, searchslot, eq, NULL)) continue; found = true; @@ -456,6 +469,236 @@ BuildConflictIndexInfo(ResultRelInfo *resultRelInfo, Oid conflictindex) } /* + * If the tuple is recently dead and was deleted by a transaction with a newer + * commit timestamp than previously recorded, update the associated transaction + * ID, commit time, and origin. This helps ensure that conflict detection uses + * the most recent and relevant deletion metadata. + */ +static void +update_most_recent_deletion_info(TupleTableSlot *scanslot, + TransactionId oldestxmin, + TransactionId *delete_xid, + TimestampTz *delete_time, + RepOriginId *delete_origin) +{ + BufferHeapTupleTableSlot *hslot; + HeapTuple tuple; + Buffer buf; + bool recently_dead = false; + TransactionId xmax; + TimestampTz localts; + RepOriginId localorigin; + + hslot = (BufferHeapTupleTableSlot *) scanslot; + + tuple = ExecFetchSlotHeapTuple(scanslot, false, NULL); + buf = hslot->buffer; + + LockBuffer(buf, BUFFER_LOCK_SHARE); + + /* + * We do not consider HEAPTUPLE_DEAD status because it indicates either + * tuples whose inserting transaction was aborted (meaning there is no + * commit timestamp or origin), or tuples deleted by a transaction older + * than oldestxmin, making it safe to ignore them during conflict + * detection (See comments atop worker.c for details). + */ + if (HeapTupleSatisfiesVacuum(tuple, oldestxmin, buf) == HEAPTUPLE_RECENTLY_DEAD) + recently_dead = true; + + LockBuffer(buf, BUFFER_LOCK_UNLOCK); + + if (!recently_dead) + return; + + xmax = HeapTupleHeaderGetUpdateXid(tuple->t_data); + if (!TransactionIdIsValid(xmax)) + return; + + /* Select the dead tuple with the most recent commit timestamp */ + if (TransactionIdGetCommitTsData(xmax, &localts, &localorigin) && + TimestampDifferenceExceeds(*delete_time, localts, 0)) + { + *delete_xid = xmax; + *delete_time = localts; + *delete_origin = localorigin; + } +} + +/* + * Searches the relation 'rel' for the most recently deleted tuple that matches + * the values in 'searchslot' and is not yet removable by VACUUM. The function + * returns the transaction ID, origin, and commit timestamp of the transaction + * that deleted this tuple. + * + * 'oldestxmin' acts as a cutoff transaction ID. Tuples deleted by transactions + * with IDs >= 'oldestxmin' are considered recently dead and are eligible for + * conflict detection. + * + * Instead of stopping at the first match, we scan all matching dead tuples to + * identify most recent deletion. This is crucial because only the latest + * deletion is relevant for resolving conflicts. + * + * For example, consider a scenario on the subscriber where a row is deleted, + * re-inserted, and then deleted again only on the subscriber: + * + * - (pk, 1) - deleted at 9:00, + * - (pk, 1) - deleted at 9:02, + * + * Now, a remote update arrives: (pk, 1) -> (pk, 2), timestamped at 9:01. + * + * If we mistakenly return the older deletion (9:00), the system may wrongly + * apply the remote update using a last-update-wins strategy. Instead, we must + * recognize the more recent deletion at 9:02 and skip the update. See + * comments atop worker.c for details. Note, as of now, conflict resolution + * is not implemented. Consequently, the system may incorrectly report the + * older tuple as the conflicted one, leading to misleading results. + * + * The commit timestamp of the deleting transaction is used to determine which + * tuple was deleted most recently. + */ +bool +RelationFindDeletedTupleInfoSeq(Relation rel, TupleTableSlot *searchslot, + TransactionId oldestxmin, + TransactionId *delete_xid, + RepOriginId *delete_origin, + TimestampTz *delete_time) +{ + TupleTableSlot *scanslot; + TableScanDesc scan; + TypeCacheEntry **eq; + Bitmapset *indexbitmap; + TupleDesc desc PG_USED_FOR_ASSERTS_ONLY = RelationGetDescr(rel); + + Assert(equalTupleDescs(desc, searchslot->tts_tupleDescriptor)); + + *delete_xid = InvalidTransactionId; + *delete_origin = InvalidRepOriginId; + *delete_time = 0; + + /* + * If the relation has a replica identity key or a primary key that is + * unusable for locating deleted tuples (see + * IsIndexUsableForFindingDeletedTuple), a full table scan becomes + * necessary. In such cases, comparing the entire tuple is not required, + * since the remote tuple might not include all column values. Instead, + * the indexed columns alone are suffcient to identify the target tuple + * (see logicalrep_rel_mark_updatable). + */ + indexbitmap = RelationGetIndexAttrBitmap(rel, + INDEX_ATTR_BITMAP_IDENTITY_KEY); + + /* fallback to PK if no replica identity */ + if (!indexbitmap) + indexbitmap = RelationGetIndexAttrBitmap(rel, + INDEX_ATTR_BITMAP_PRIMARY_KEY); + + eq = palloc0(sizeof(*eq) * searchslot->tts_tupleDescriptor->natts); + + /* + * Start a heap scan using SnapshotAny to identify dead tuples that are + * not visible under a standard MVCC snapshot. Tuples from transactions + * not yet committed or those just committed prior to the scan are + * excluded in update_most_recent_deletion_info(). + */ + scan = table_beginscan(rel, SnapshotAny, 0, NULL); + scanslot = table_slot_create(rel, NULL); + + table_rescan(scan, NULL); + + /* Try to find the tuple */ + while (table_scan_getnextslot(scan, ForwardScanDirection, scanslot)) + { + if (!tuples_equal(scanslot, searchslot, eq, indexbitmap)) + continue; + + update_most_recent_deletion_info(scanslot, oldestxmin, delete_xid, + delete_time, delete_origin); + } + + table_endscan(scan); + ExecDropSingleTupleTableSlot(scanslot); + + return *delete_time != 0; +} + +/* + * Similar to RelationFindDeletedTupleInfoSeq() but using index scan to locate + * the deleted tuple. + */ +bool +RelationFindDeletedTupleInfoByIndex(Relation rel, Oid idxoid, + TupleTableSlot *searchslot, + TransactionId oldestxmin, + TransactionId *delete_xid, + RepOriginId *delete_origin, + TimestampTz *delete_time) +{ + Relation idxrel; + ScanKeyData skey[INDEX_MAX_KEYS]; + int skey_attoff; + IndexScanDesc scan; + TupleTableSlot *scanslot; + TypeCacheEntry **eq = NULL; + bool isIdxSafeToSkipDuplicates; + TupleDesc desc PG_USED_FOR_ASSERTS_ONLY = RelationGetDescr(rel); + + Assert(equalTupleDescs(desc, searchslot->tts_tupleDescriptor)); + Assert(OidIsValid(idxoid)); + + *delete_xid = InvalidTransactionId; + *delete_time = 0; + *delete_origin = InvalidRepOriginId; + + isIdxSafeToSkipDuplicates = (GetRelationIdentityOrPK(rel) == idxoid); + + scanslot = table_slot_create(rel, NULL); + + idxrel = index_open(idxoid, RowExclusiveLock); + + /* Build scan key. */ + skey_attoff = build_replindex_scan_key(skey, rel, idxrel, searchslot); + + /* + * Start an index scan using SnapshotAny to identify dead tuples that are + * not visible under a standard MVCC snapshot. Tuples from transactions + * not yet committed or those just committed prior to the scan are + * excluded in update_most_recent_deletion_info(). + */ + scan = index_beginscan(rel, idxrel, SnapshotAny, NULL, skey_attoff, 0); + + index_rescan(scan, skey, skey_attoff, NULL, 0); + + /* Try to find the tuple */ + while (index_getnext_slot(scan, ForwardScanDirection, scanslot)) + { + /* + * Avoid expensive equality check if the index is primary key or + * replica identity index. + */ + if (!isIdxSafeToSkipDuplicates) + { + if (eq == NULL) + eq = palloc0(sizeof(*eq) * scanslot->tts_tupleDescriptor->natts); + + if (!tuples_equal(scanslot, searchslot, eq, NULL)) + continue; + } + + update_most_recent_deletion_info(scanslot, oldestxmin, delete_xid, + delete_time, delete_origin); + } + + index_endscan(scan); + + index_close(idxrel, NoLock); + + ExecDropSingleTupleTableSlot(scanslot); + + return *delete_time != 0; +} + +/* * Find the tuple that violates the passed unique index (conflictindex). * * If the conflicting tuple is found return true, otherwise false. 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/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 8908603464c..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; @@ -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/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/replication/logical/conflict.c b/src/backend/replication/logical/conflict.c index 97c4e26b586..2fd3e8bbda5 100644 --- a/src/backend/replication/logical/conflict.c +++ b/src/backend/replication/logical/conflict.c @@ -29,6 +29,7 @@ static const char *const ConflictTypeNames[] = { [CT_UPDATE_EXISTS] = "update_exists", [CT_UPDATE_MISSING] = "update_missing", [CT_DELETE_ORIGIN_DIFFERS] = "delete_origin_differs", + [CT_UPDATE_DELETED] = "update_deleted", [CT_DELETE_MISSING] = "delete_missing", [CT_MULTIPLE_UNIQUE_CONFLICTS] = "multiple_unique_conflicts" }; @@ -176,6 +177,7 @@ errcode_apply_conflict(ConflictType type) case CT_UPDATE_ORIGIN_DIFFERS: case CT_UPDATE_MISSING: case CT_DELETE_ORIGIN_DIFFERS: + case CT_UPDATE_DELETED: case CT_DELETE_MISSING: return errcode(ERRCODE_T_R_SERIALIZATION_FAILURE); } @@ -261,6 +263,26 @@ errdetail_apply_conflict(EState *estate, ResultRelInfo *relinfo, break; + case CT_UPDATE_DELETED: + if (localts) + { + if (localorigin == InvalidRepOriginId) + appendStringInfo(&err_detail, _("The row to be updated was deleted locally in transaction %u at %s."), + localxmin, timestamptz_to_str(localts)); + else if (replorigin_by_oid(localorigin, true, &origin_name)) + appendStringInfo(&err_detail, _("The row to be updated was deleted by a different origin \"%s\" in transaction %u at %s."), + origin_name, localxmin, timestamptz_to_str(localts)); + + /* The origin that modified this row has been removed. */ + else + appendStringInfo(&err_detail, _("The row to be updated was deleted by a non-existent origin in transaction %u at %s."), + localxmin, timestamptz_to_str(localts)); + } + else + appendStringInfo(&err_detail, _("The row to be updated was deleted.")); + + break; + case CT_UPDATE_MISSING: appendStringInfoString(&err_detail, _("Could not find the row to be updated.")); break; diff --git a/src/backend/replication/logical/launcher.c b/src/backend/replication/logical/launcher.c index 742d9ba68e9..37377f7eb63 100644 --- a/src/backend/replication/logical/launcher.c +++ b/src/backend/replication/logical/launcher.c @@ -790,6 +790,8 @@ logicalrep_worker_detach(void) } LWLockRelease(LogicalRepWorkerLock); + + list_free(workers); } /* Block concurrent access. */ diff --git a/src/backend/replication/logical/slotsync.c b/src/backend/replication/logical/slotsync.c index 2f0c08b8fbd..37738440113 100644 --- a/src/backend/replication/logical/slotsync.c +++ b/src/backend/replication/logical/slotsync.c @@ -1059,14 +1059,14 @@ ValidateSlotSyncParams(int elevel) { /* * Logical slot sync/creation requires wal_level >= logical. - * - * Since altering the wal_level requires a server restart, so error out in - * this case regardless of elevel provided by caller. */ if (wal_level < WAL_LEVEL_LOGICAL) - ereport(ERROR, + { + ereport(elevel, errcode(ERRCODE_INVALID_PARAMETER_VALUE), errmsg("replication slot synchronization requires \"wal_level\" >= \"logical\"")); + return false; + } /* * A physical replication slot(primary_slot_name) is required on the diff --git a/src/backend/replication/logical/worker.c b/src/backend/replication/logical/worker.c index b59221c4d06..89e241c8392 100644 --- a/src/backend/replication/logical/worker.c +++ b/src/backend/replication/logical/worker.c @@ -138,9 +138,9 @@ * 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 + * needs them to detect update_deleted conflicts. Additionally, 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 @@ -185,10 +185,10 @@ * 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. + * UPDATE to an INSERT when applying it. But, for concurrent remote + * transactions with earlier timestamps than the DELETE, detecting + * update_deleted 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 @@ -576,6 +576,12 @@ static bool FindReplTupleInLocalRel(ApplyExecutionData *edata, Relation localrel Oid localidxoid, TupleTableSlot *remoteslot, TupleTableSlot **localslot); +static bool FindDeletedTupleInLocalRel(Relation localrel, + Oid localidxoid, + TupleTableSlot *remoteslot, + TransactionId *delete_xid, + RepOriginId *delete_origin, + TimestampTz *delete_time); static void apply_handle_tuple_routing(ApplyExecutionData *edata, TupleTableSlot *remoteslot, LogicalRepTupleData *newtup, @@ -2912,17 +2918,31 @@ apply_handle_update_internal(ApplyExecutionData *edata, } else { + ConflictType type; TupleTableSlot *newslot = localslot; + /* + * Detecting whether the tuple was recently deleted or never existed + * is crucial to avoid misleading the user during confict handling. + */ + if (FindDeletedTupleInLocalRel(localrel, localindexoid, remoteslot, + &conflicttuple.xmin, + &conflicttuple.origin, + &conflicttuple.ts) && + conflicttuple.origin != replorigin_session_origin) + type = CT_UPDATE_DELETED; + else + type = CT_UPDATE_MISSING; + /* Store the new tuple for conflict reporting */ slot_store_data(newslot, relmapentry, newtup); /* - * The tuple to be updated could not be found. Do nothing except for - * emitting a log message. + * The tuple to be updated could not be found or was deleted. Do + * nothing except for emitting a log message. */ - ReportApplyConflict(estate, relinfo, LOG, CT_UPDATE_MISSING, - remoteslot, newslot, list_make1(&conflicttuple)); + ReportApplyConflict(estate, relinfo, LOG, type, remoteslot, newslot, + list_make1(&conflicttuple)); } /* Cleanup. */ @@ -3143,6 +3163,112 @@ FindReplTupleInLocalRel(ApplyExecutionData *edata, Relation localrel, } /* + * Determine whether the index can reliably locate the deleted tuple in the + * local relation. + * + * An index may exclude deleted tuples if it was re-indexed or re-created during + * change application. Therefore, an index is considered usable only if the + * conflict detection slot.xmin (conflict_detection_xmin) is greater than the + * index tuple's xmin. This ensures that any tuples deleted prior to the index + * creation or re-indexing are not relevant for conflict detection in the + * current apply worker. + * + * Note that indexes may also be excluded if they were modified by other DDL + * operations, such as ALTER INDEX. However, this is acceptable, as the + * likelihood of such DDL changes coinciding with the need to scan dead + * tuples for the update_deleted is low. + */ +static bool +IsIndexUsableForFindingDeletedTuple(Oid localindexoid, + TransactionId conflict_detection_xmin) +{ + HeapTuple index_tuple; + TransactionId index_xmin; + + index_tuple = SearchSysCache1(INDEXRELID, ObjectIdGetDatum(localindexoid)); + + if (!HeapTupleIsValid(index_tuple)) /* should not happen */ + elog(ERROR, "cache lookup failed for index %u", localindexoid); + + /* + * No need to check for a frozen transaction ID, as + * TransactionIdPrecedes() manages it internally, treating it as falling + * behind the conflict_detection_xmin. + */ + index_xmin = HeapTupleHeaderGetXmin(index_tuple->t_data); + + ReleaseSysCache(index_tuple); + + return TransactionIdPrecedes(index_xmin, conflict_detection_xmin); +} + +/* + * Attempts to locate a deleted tuple in the local relation that matches the + * values of the tuple received from the publication side (in 'remoteslot'). + * The search is performed using either the replica identity index, primary + * key, other available index, or a sequential scan if necessary. + * + * Returns true if the deleted tuple is found. If found, the transaction ID, + * origin, and commit timestamp of the deletion are stored in '*delete_xid', + * '*delete_origin', and '*delete_time' respectively. + */ +static bool +FindDeletedTupleInLocalRel(Relation localrel, Oid localidxoid, + TupleTableSlot *remoteslot, + TransactionId *delete_xid, RepOriginId *delete_origin, + TimestampTz *delete_time) +{ + TransactionId oldestxmin; + ReplicationSlot *slot; + + /* + * Return false if either dead tuples are not retained or commit timestamp + * data is not available. + */ + if (!MySubscription->retaindeadtuples || !track_commit_timestamp) + return false; + + /* + * For conflict detection, we use the conflict slot's xmin value instead + * of invoking GetOldestNonRemovableTransactionId(). The slot.xmin acts as + * a threshold to identify tuples that were recently deleted. These tuples + * are not visible to concurrent transactions, but we log an + * update_deleted conflict if such a tuple matches the remote update being + * applied. + * + * Although GetOldestNonRemovableTransactionId() can return a value older + * than the slot's xmin, for our current purpose it is acceptable to treat + * tuples deleted by transactions prior to slot.xmin as update_missing + * conflicts. + * + * Ideally, we would use oldest_nonremovable_xid, which is directly + * maintained by the leader apply worker. However, this value is not + * available to table synchronization or parallel apply workers, making + * slot.xmin a practical alternative in those contexts. + */ + slot = SearchNamedReplicationSlot(CONFLICT_DETECTION_SLOT, true); + + Assert(slot); + + SpinLockAcquire(&slot->mutex); + oldestxmin = slot->data.xmin; + SpinLockRelease(&slot->mutex); + + Assert(TransactionIdIsValid(oldestxmin)); + + if (OidIsValid(localidxoid) && + IsIndexUsableForFindingDeletedTuple(localidxoid, oldestxmin)) + return RelationFindDeletedTupleInfoByIndex(localrel, localidxoid, + remoteslot, oldestxmin, + delete_xid, delete_origin, + delete_time); + else + return RelationFindDeletedTupleInfoSeq(localrel, remoteslot, + oldestxmin, delete_xid, + delete_origin, delete_time); +} + +/* * This handles insert, update, delete on a partitioned table. */ static void @@ -3260,18 +3386,35 @@ apply_handle_tuple_routing(ApplyExecutionData *edata, remoteslot_part, &localslot); if (!found) { + ConflictType type; TupleTableSlot *newslot = localslot; + /* + * Detecting whether the tuple was recently deleted or + * never existed is crucial to avoid misleading the user + * during confict handling. + */ + if (FindDeletedTupleInLocalRel(partrel, + part_entry->localindexoid, + remoteslot_part, + &conflicttuple.xmin, + &conflicttuple.origin, + &conflicttuple.ts) && + conflicttuple.origin != replorigin_session_origin) + type = CT_UPDATE_DELETED; + else + type = CT_UPDATE_MISSING; + /* Store the new tuple for conflict reporting */ slot_store_data(newslot, part_entry, newtup); /* - * The tuple to be updated could not be found. Do nothing - * except for emitting a log message. + * The tuple to be updated could not be found or was + * deleted. Do nothing except for emitting a log message. */ ReportApplyConflict(estate, partrelinfo, LOG, - CT_UPDATE_MISSING, remoteslot_part, - newslot, list_make1(&conflicttuple)); + type, remoteslot_part, newslot, + list_make1(&conflicttuple)); return; } @@ -4172,8 +4315,8 @@ 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. + * subscription by the main apply worker to detect update_deleted reliably + * even for table sync or parallel apply workers. */ if (!am_leader_apply_worker()) return false; @@ -4374,10 +4517,11 @@ wait_for_local_flush(RetainDeadTuplesData *rdt_data) * 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. + * required to detect update_deleted 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. 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/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/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/adt/pgstatfuncs.c b/src/backend/utils/adt/pgstatfuncs.c index 1c12ddbae49..c756c2bebaa 100644 --- a/src/backend/utils/adt/pgstatfuncs.c +++ b/src/backend/utils/adt/pgstatfuncs.c @@ -2171,7 +2171,7 @@ pg_stat_get_replication_slot(PG_FUNCTION_ARGS) Datum pg_stat_get_subscription_stats(PG_FUNCTION_ARGS) { -#define PG_STAT_GET_SUBSCRIPTION_STATS_COLS 11 +#define PG_STAT_GET_SUBSCRIPTION_STATS_COLS 12 Oid subid = PG_GETARG_OID(0); TupleDesc tupdesc; Datum values[PG_STAT_GET_SUBSCRIPTION_STATS_COLS] = {0}; @@ -2197,15 +2197,17 @@ pg_stat_get_subscription_stats(PG_FUNCTION_ARGS) INT8OID, -1, 0); TupleDescInitEntry(tupdesc, (AttrNumber) 6, "confl_update_exists", INT8OID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 7, "confl_update_missing", + TupleDescInitEntry(tupdesc, (AttrNumber) 7, "confl_update_deleted", INT8OID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 8, "confl_delete_origin_differs", + TupleDescInitEntry(tupdesc, (AttrNumber) 8, "confl_update_missing", INT8OID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 9, "confl_delete_missing", + TupleDescInitEntry(tupdesc, (AttrNumber) 9, "confl_delete_origin_differs", INT8OID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 10, "confl_multiple_unique_conflicts", + TupleDescInitEntry(tupdesc, (AttrNumber) 10, "confl_delete_missing", INT8OID, -1, 0); - TupleDescInitEntry(tupdesc, (AttrNumber) 11, "stats_reset", + TupleDescInitEntry(tupdesc, (AttrNumber) 11, "confl_multiple_unique_conflicts", + INT8OID, -1, 0); + TupleDescInitEntry(tupdesc, (AttrNumber) 12, "stats_reset", TIMESTAMPTZOID, -1, 0); BlessTupleDesc(tupdesc); 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 0c506d320b1..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; 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/error/elog.c b/src/backend/utils/error/elog.c index 47af743990f..afce1a8e1f0 100644 --- a/src/backend/utils/error/elog.c +++ b/src/backend/utils/error/elog.c @@ -1128,12 +1128,15 @@ set_backtrace(ErrorData *edata, int num_skip) nframes = backtrace(buf, lengthof(buf)); strfrms = backtrace_symbols(buf, nframes); - if (strfrms == NULL) - return; - - for (int i = num_skip; i < nframes; i++) - appendStringInfo(&errtrace, "\n%s", strfrms[i]); - free(strfrms); + if (strfrms != NULL) + { + for (int i = num_skip; i < nframes; i++) + appendStringInfo(&errtrace, "\n%s", strfrms[i]); + free(strfrms); + } + else + appendStringInfoString(&errtrace, + "insufficient memory for backtrace generation"); } #else appendStringInfoString(&errtrace, diff --git a/src/backend/utils/hash/dynahash.c b/src/backend/utils/hash/dynahash.c index 42e9be274fc..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" @@ -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 }; /* @@ -1712,6 +1724,8 @@ 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; @@ -1723,12 +1737,38 @@ element_alloc(HTAB *hashp, int nelem, int freelist_idx) /* 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/misc/guc.c b/src/backend/utils/misc/guc.c index ce5449f2878..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); @@ -5023,16 +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); } /* @@ -5132,6 +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 @@ -5291,9 +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) @@ -5317,6 +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); } } 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 ce01dce9861..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); } /* @@ -1170,8 +1183,6 @@ MemoryContextCreate(MemoryContext node, node->nextchild = NULL; node->allowInCritSection = false; } - - VALGRIND_CREATE_MEMPOOL(node, 0, false); } /* @@ -1454,7 +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 */ @@ -1491,10 +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)); @@ -1522,12 +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; } @@ -1561,16 +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); } /* @@ -1580,9 +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 @@ -1605,10 +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; } diff --git a/src/bin/pg_basebackup/pg_recvlogical.c b/src/bin/pg_basebackup/pg_recvlogical.c index 8a5dd24e6c9..0e9d2e23947 100644 --- a/src/bin/pg_basebackup/pg_recvlogical.c +++ b/src/bin/pg_basebackup/pg_recvlogical.c @@ -517,7 +517,7 @@ StreamLogicalLog(void) } /* - * Read the header of the XLogData message, enclosed in the CopyData + * Read the header of the WALData message, enclosed in the CopyData * message. We only need the WAL location field (dataStart), the rest * of the header is ignored. */ @@ -605,7 +605,7 @@ StreamLogicalLog(void) /* * We're doing a client-initiated clean exit and have sent CopyDone to * the server. Drain any messages, so we don't miss a last-minute - * ErrorResponse. The walsender stops generating XLogData records once + * ErrorResponse. The walsender stops generating WALData records once * it sees CopyDone, so expect this to finish quickly. After CopyDone, * it's too late for sendFeedback(), even if this were to take a long * time. Hence, use synchronous-mode PQgetCopyData(). diff --git a/src/bin/pg_basebackup/receivelog.c b/src/bin/pg_basebackup/receivelog.c index d6b7f117fa3..f2b54d3c501 100644 --- a/src/bin/pg_basebackup/receivelog.c +++ b/src/bin/pg_basebackup/receivelog.c @@ -38,8 +38,8 @@ static int CopyStreamReceive(PGconn *conn, long timeout, pgsocket stop_socket, char **buffer); static bool ProcessKeepaliveMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, XLogRecPtr blockpos, TimestampTz *last_status); -static bool ProcessXLogDataMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, - XLogRecPtr *blockpos); +static bool ProcessWALDataMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, + XLogRecPtr *blockpos); static PGresult *HandleEndOfCopyStream(PGconn *conn, StreamCtl *stream, char *copybuf, XLogRecPtr blockpos, XLogRecPtr *stoppos); static bool CheckCopyStreamStop(PGconn *conn, StreamCtl *stream, XLogRecPtr blockpos); @@ -831,7 +831,7 @@ HandleCopyStream(PGconn *conn, StreamCtl *stream, } else if (copybuf[0] == 'w') { - if (!ProcessXLogDataMsg(conn, stream, copybuf, r, &blockpos)) + if (!ProcessWALDataMsg(conn, stream, copybuf, r, &blockpos)) goto error; /* @@ -1041,11 +1041,11 @@ ProcessKeepaliveMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, } /* - * Process XLogData message. + * Process WALData message. */ static bool -ProcessXLogDataMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, - XLogRecPtr *blockpos) +ProcessWALDataMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, + XLogRecPtr *blockpos) { int xlogoff; int bytes_left; @@ -1054,13 +1054,13 @@ ProcessXLogDataMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, /* * Once we've decided we don't want to receive any more, just ignore any - * subsequent XLogData messages. + * subsequent WALData messages. */ if (!(still_sending)) return true; /* - * Read the header of the XLogData message, enclosed in the CopyData + * Read the header of the WALData message, enclosed in the CopyData * message. We only need the WAL location field (dataStart), the rest of * the header is ignored. */ @@ -1162,7 +1162,7 @@ ProcessXLogDataMsg(PGconn *conn, StreamCtl *stream, char *copybuf, int len, return false; } still_sending = false; - return true; /* ignore the rest of this XLogData packet */ + return true; /* ignore the rest of this WALData packet */ } } } diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c index b1ac8d7b509..f3a353a61a5 100644 --- a/src/bin/pg_dump/pg_dump.c +++ b/src/bin/pg_dump/pg_dump.c @@ -449,8 +449,6 @@ main(int argc, char **argv) bool data_only = false; bool schema_only = false; bool statistics_only = false; - bool with_data = false; - bool with_schema = false; bool with_statistics = false; bool no_data = false; bool no_schema = false; @@ -514,6 +512,7 @@ main(int argc, char **argv) {"section", required_argument, NULL, 5}, {"serializable-deferrable", no_argument, &dopt.serializable_deferrable, 1}, {"snapshot", required_argument, NULL, 6}, + {"statistics", no_argument, NULL, 22}, {"statistics-only", no_argument, NULL, 18}, {"strict-names", no_argument, &strict_names, 1}, {"use-set-session-authorization", no_argument, &dopt.use_setsessauth, 1}, @@ -528,9 +527,6 @@ main(int argc, char **argv) {"no-toast-compression", no_argument, &dopt.no_toast_compression, 1}, {"no-unlogged-table-data", no_argument, &dopt.no_unlogged_table_data, 1}, {"no-sync", no_argument, NULL, 7}, - {"with-data", no_argument, NULL, 22}, - {"with-schema", no_argument, NULL, 23}, - {"with-statistics", no_argument, NULL, 24}, {"on-conflict-do-nothing", no_argument, &dopt.do_nothing, 1}, {"rows-per-insert", required_argument, NULL, 10}, {"include-foreign-data", required_argument, NULL, 11}, @@ -798,14 +794,6 @@ main(int argc, char **argv) break; case 22: - with_data = true; - break; - - case 23: - with_schema = true; - break; - - case 24: with_statistics = true; break; @@ -852,24 +840,17 @@ main(int argc, char **argv) if (statistics_only && no_statistics) pg_fatal("options --statistics-only and --no-statistics cannot be used together"); - /* reject conflicting "with-" and "no-" options */ - if (with_data && no_data) - pg_fatal("options --with-data and --no-data cannot be used together"); - if (with_schema && no_schema) - pg_fatal("options --with-schema and --no-schema cannot be used together"); + /* reject conflicting "no-" options */ if (with_statistics && no_statistics) - pg_fatal("options --with-statistics and --no-statistics cannot be used together"); + pg_fatal("options --statistics and --no-statistics cannot be used together"); - /* reject conflicting "-only" and "with-" options */ - if (data_only && (with_schema || with_statistics)) - pg_fatal("options %s and %s cannot be used together", - "-a/--data-only", with_schema ? "--with-schema" : "--with-statistics"); - if (schema_only && (with_data || with_statistics)) + /* reject conflicting "-only" options */ + if (data_only && with_statistics) pg_fatal("options %s and %s cannot be used together", - "-s/--schema-only", with_data ? "--with-data" : "--with-statistics"); - if (statistics_only && (with_data || with_schema)) + "-a/--data-only", "--statistics"); + if (schema_only && with_statistics) pg_fatal("options %s and %s cannot be used together", - "--statistics-only", with_data ? "--with-data" : "--with-schema"); + "-s/--schema-only", "--statistics"); if (schema_only && foreign_servers_include_patterns.head != NULL) pg_fatal("options -s/--schema-only and --include-foreign-data cannot be used together"); @@ -889,9 +870,9 @@ main(int argc, char **argv) * of the checks above. */ dopt.dumpData = ((dopt.dumpData && !schema_only && !statistics_only) || - (data_only || with_data)) && !no_data; + data_only) && !no_data; dopt.dumpSchema = ((dopt.dumpSchema && !data_only && !statistics_only) || - (schema_only || with_schema)) && !no_schema; + schema_only) && !no_schema; dopt.dumpStatistics = ((dopt.dumpStatistics && !schema_only && !data_only) || (statistics_only || with_statistics)) && !no_statistics; @@ -1364,6 +1345,7 @@ help(const char *progname) printf(_(" --sequence-data include sequence data in dump\n")); printf(_(" --serializable-deferrable wait until the dump can run without anomalies\n")); printf(_(" --snapshot=SNAPSHOT use given snapshot for the dump\n")); + printf(_(" --statistics dump the statistics\n")); printf(_(" --statistics-only dump only the statistics, not schema or data\n")); printf(_(" --strict-names require table and/or schema include patterns to\n" " match at least one entity each\n")); @@ -1372,9 +1354,6 @@ help(const char *progname) printf(_(" --use-set-session-authorization\n" " use SET SESSION AUTHORIZATION commands instead of\n" " ALTER OWNER commands to set ownership\n")); - printf(_(" --with-data dump the data\n")); - printf(_(" --with-schema dump the schema\n")); - printf(_(" --with-statistics dump the statistics\n")); printf(_("\nConnection options:\n")); printf(_(" -d, --dbname=DBNAME database to dump\n")); diff --git a/src/bin/pg_dump/pg_dumpall.c b/src/bin/pg_dump/pg_dumpall.c index 87d10df07c4..27aa1b65698 100644 --- a/src/bin/pg_dump/pg_dumpall.c +++ b/src/bin/pg_dump/pg_dumpall.c @@ -105,8 +105,6 @@ static int no_subscriptions = 0; static int no_toast_compression = 0; static int no_unlogged_table_data = 0; static int no_role_passwords = 0; -static int with_data = 0; -static int with_schema = 0; static int with_statistics = 0; static int server_version; static int load_via_partition_root = 0; @@ -180,11 +178,9 @@ main(int argc, char *argv[]) {"no-sync", no_argument, NULL, 4}, {"no-toast-compression", no_argument, &no_toast_compression, 1}, {"no-unlogged-table-data", no_argument, &no_unlogged_table_data, 1}, - {"with-data", no_argument, &with_data, 1}, - {"with-schema", no_argument, &with_schema, 1}, - {"with-statistics", no_argument, &with_statistics, 1}, {"on-conflict-do-nothing", no_argument, &on_conflict_do_nothing, 1}, {"rows-per-insert", required_argument, NULL, 7}, + {"statistics", no_argument, &with_statistics, 1}, {"statistics-only", no_argument, &statistics_only, 1}, {"filter", required_argument, NULL, 8}, {"sequence-data", no_argument, &sequence_data, 1}, @@ -475,12 +471,8 @@ main(int argc, char *argv[]) appendPQExpBufferStr(pgdumpopts, " --no-toast-compression"); if (no_unlogged_table_data) appendPQExpBufferStr(pgdumpopts, " --no-unlogged-table-data"); - if (with_data) - appendPQExpBufferStr(pgdumpopts, " --with-data"); - if (with_schema) - appendPQExpBufferStr(pgdumpopts, " --with-schema"); if (with_statistics) - appendPQExpBufferStr(pgdumpopts, " --with-statistics"); + appendPQExpBufferStr(pgdumpopts, " --statistics"); if (on_conflict_do_nothing) appendPQExpBufferStr(pgdumpopts, " --on-conflict-do-nothing"); if (statistics_only) @@ -712,13 +704,11 @@ help(void) printf(_(" --quote-all-identifiers quote all identifiers, even if not key words\n")); printf(_(" --rows-per-insert=NROWS number of rows per INSERT; implies --inserts\n")); printf(_(" --sequence-data include sequence data in dump\n")); + printf(_(" --statistics dump the statistics\n")); printf(_(" --statistics-only dump only the statistics, not schema or data\n")); printf(_(" --use-set-session-authorization\n" " use SET SESSION AUTHORIZATION commands instead of\n" " ALTER OWNER commands to set ownership\n")); - printf(_(" --with-data dump the data\n")); - printf(_(" --with-schema dump the schema\n")); - printf(_(" --with-statistics dump the statistics\n")); printf(_("\nConnection options:\n")); printf(_(" -d, --dbname=CONNSTR connect using connection string\n")); diff --git a/src/bin/pg_dump/pg_restore.c b/src/bin/pg_dump/pg_restore.c index 2c727b9f156..6c129278bc5 100644 --- a/src/bin/pg_dump/pg_restore.c +++ b/src/bin/pg_dump/pg_restore.c @@ -82,8 +82,6 @@ main(int argc, char **argv) static int no_subscriptions = 0; static int strict_names = 0; static int statistics_only = 0; - static int with_data = 0; - static int with_schema = 0; static int with_statistics = 0; struct option cmdopts[] = { @@ -139,9 +137,7 @@ main(int argc, char **argv) {"no-security-labels", no_argument, &no_security_labels, 1}, {"no-subscriptions", no_argument, &no_subscriptions, 1}, {"no-statistics", no_argument, &no_statistics, 1}, - {"with-data", no_argument, &with_data, 1}, - {"with-schema", no_argument, &with_schema, 1}, - {"with-statistics", no_argument, &with_statistics, 1}, + {"statistics", no_argument, &with_statistics, 1}, {"statistics-only", no_argument, &statistics_only, 1}, {"filter", required_argument, NULL, 4}, @@ -373,24 +369,17 @@ main(int argc, char **argv) if (statistics_only && no_statistics) pg_fatal("options --statistics-only and --no-statistics cannot be used together"); - /* reject conflicting "with-" and "no-" options */ - if (with_data && no_data) - pg_fatal("options --with-data and --no-data cannot be used together"); - if (with_schema && no_schema) - pg_fatal("options --with-schema and --no-schema cannot be used together"); + /* reject conflicting "no-" options */ if (with_statistics && no_statistics) - pg_fatal("options --with-statistics and --no-statistics cannot be used together"); + pg_fatal("options --statistics and --no-statistics cannot be used together"); - /* reject conflicting "only-" and "with-" options */ - if (data_only && (with_schema || with_statistics)) + /* reject conflicting "only-" options */ + if (data_only && with_statistics) pg_fatal("options %s and %s cannot be used together", - "-a/--data-only", with_schema ? "--with-schema" : "--with-statistics"); - if (schema_only && (with_data || with_statistics)) + "-a/--data-only", "--statistics"); + if (schema_only && with_statistics) pg_fatal("options %s and %s cannot be used together", - "-s/--schema-only", with_data ? "--with-data" : "--with-statistics"); - if (statistics_only && (with_data || with_schema)) - pg_fatal("options %s and %s cannot be used together", - "--statistics-only", with_data ? "--with-data" : "--with-schema"); + "-s/--schema-only", "--statistics"); if (data_only && opts->dropSchema) pg_fatal("options -c/--clean and -a/--data-only cannot be used together"); @@ -415,9 +404,9 @@ main(int argc, char **argv) * of the checks above. */ opts->dumpData = ((opts->dumpData && !schema_only && !statistics_only) || - (data_only || with_data)) && !no_data; + data_only) && !no_data; opts->dumpSchema = ((opts->dumpSchema && !data_only && !statistics_only) || - (schema_only || with_schema)) && !no_schema; + schema_only) && !no_schema; opts->dumpStatistics = ((opts->dumpStatistics && !schema_only && !data_only) || (statistics_only || with_statistics)) && !no_statistics; @@ -558,6 +547,7 @@ usage(const char *progname) printf(_(" --no-table-access-method do not restore table access methods\n")); printf(_(" --no-tablespaces do not restore tablespace assignments\n")); printf(_(" --section=SECTION restore named section (pre-data, data, or post-data)\n")); + printf(_(" --statistics restore the statistics\n")); printf(_(" --statistics-only restore only the statistics, not schema or data\n")); printf(_(" --strict-names require table and/or schema include patterns to\n" " match at least one entity each\n")); @@ -565,9 +555,6 @@ usage(const char *progname) printf(_(" --use-set-session-authorization\n" " use SET SESSION AUTHORIZATION commands instead of\n" " ALTER OWNER commands to set ownership\n")); - printf(_(" --with-data restore the data\n")); - printf(_(" --with-schema restore the schema\n")); - printf(_(" --with-statistics restore the statistics\n")); printf(_("\nConnection options:\n")); printf(_(" -h, --host=HOSTNAME database server host or socket directory\n")); diff --git a/src/bin/pg_dump/t/002_pg_dump.pl b/src/bin/pg_dump/t/002_pg_dump.pl index d597842908e..a86b38466de 100644 --- a/src/bin/pg_dump/t/002_pg_dump.pl +++ b/src/bin/pg_dump/t/002_pg_dump.pl @@ -68,7 +68,7 @@ my %pgdump_runs = ( '--no-data', '--sequence-data', '--binary-upgrade', - '--with-statistics', + '--statistics', '--dbname' => 'postgres', # alternative way to specify database ], restore_cmd => [ @@ -76,7 +76,7 @@ my %pgdump_runs = ( '--format' => 'custom', '--verbose', '--file' => "$tempdir/binary_upgrade.sql", - '--with-statistics', + '--statistics', "$tempdir/binary_upgrade.dump", ], }, @@ -90,13 +90,13 @@ my %pgdump_runs = ( '--format' => 'custom', '--compress' => '1', '--file' => "$tempdir/compression_gzip_custom.dump", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--file' => "$tempdir/compression_gzip_custom.sql", - '--with-statistics', + '--statistics', "$tempdir/compression_gzip_custom.dump", ], command_like => { @@ -119,7 +119,7 @@ my %pgdump_runs = ( '--format' => 'directory', '--compress' => 'gzip:1', '--file' => "$tempdir/compression_gzip_dir", - '--with-statistics', + '--statistics', 'postgres', ], # Give coverage for manually compressed blobs.toc files during @@ -137,7 +137,7 @@ my %pgdump_runs = ( 'pg_restore', '--jobs' => '2', '--file' => "$tempdir/compression_gzip_dir.sql", - '--with-statistics', + '--statistics', "$tempdir/compression_gzip_dir", ], }, @@ -150,7 +150,7 @@ my %pgdump_runs = ( '--format' => 'plain', '--compress' => '1', '--file' => "$tempdir/compression_gzip_plain.sql.gz", - '--with-statistics', + '--statistics', 'postgres', ], # Decompress the generated file to run through the tests. @@ -169,13 +169,13 @@ my %pgdump_runs = ( '--format' => 'custom', '--compress' => 'lz4', '--file' => "$tempdir/compression_lz4_custom.dump", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--file' => "$tempdir/compression_lz4_custom.sql", - '--with-statistics', + '--statistics', "$tempdir/compression_lz4_custom.dump", ], command_like => { @@ -198,7 +198,7 @@ my %pgdump_runs = ( '--format' => 'directory', '--compress' => 'lz4:1', '--file' => "$tempdir/compression_lz4_dir", - '--with-statistics', + '--statistics', 'postgres', ], # Verify that data files were compressed @@ -210,7 +210,7 @@ my %pgdump_runs = ( 'pg_restore', '--jobs' => '2', '--file' => "$tempdir/compression_lz4_dir.sql", - '--with-statistics', + '--statistics', "$tempdir/compression_lz4_dir", ], }, @@ -223,7 +223,7 @@ my %pgdump_runs = ( '--format' => 'plain', '--compress' => 'lz4', '--file' => "$tempdir/compression_lz4_plain.sql.lz4", - '--with-statistics', + '--statistics', 'postgres', ], # Decompress the generated file to run through the tests. @@ -245,13 +245,13 @@ my %pgdump_runs = ( '--format' => 'custom', '--compress' => 'zstd', '--file' => "$tempdir/compression_zstd_custom.dump", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--file' => "$tempdir/compression_zstd_custom.sql", - '--with-statistics', + '--statistics', "$tempdir/compression_zstd_custom.dump", ], command_like => { @@ -273,7 +273,7 @@ my %pgdump_runs = ( '--format' => 'directory', '--compress' => 'zstd:1', '--file' => "$tempdir/compression_zstd_dir", - '--with-statistics', + '--statistics', 'postgres', ], # Give coverage for manually compressed blobs.toc files during @@ -294,7 +294,7 @@ my %pgdump_runs = ( 'pg_restore', '--jobs' => '2', '--file' => "$tempdir/compression_zstd_dir.sql", - '--with-statistics', + '--statistics', "$tempdir/compression_zstd_dir", ], }, @@ -308,7 +308,7 @@ my %pgdump_runs = ( '--format' => 'plain', '--compress' => 'zstd:long', '--file' => "$tempdir/compression_zstd_plain.sql.zst", - '--with-statistics', + '--statistics', 'postgres', ], # Decompress the generated file to run through the tests. @@ -327,7 +327,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/clean.sql", '--clean', - '--with-statistics', + '--statistics', '--dbname' => 'postgres', # alternative way to specify database ], }, @@ -338,7 +338,7 @@ my %pgdump_runs = ( '--clean', '--if-exists', '--encoding' => 'UTF8', # no-op, just for testing - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -357,7 +357,7 @@ my %pgdump_runs = ( '--create', '--no-reconnect', # no-op, just for testing '--verbose', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -376,7 +376,7 @@ my %pgdump_runs = ( dump_cmd => [ 'pg_dump', '--no-sync', '--file' => "$tempdir/defaults.sql", - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -385,7 +385,7 @@ my %pgdump_runs = ( dump_cmd => [ 'pg_dump', '--no-sync', '--file' => "$tempdir/defaults_no_public.sql", - '--with-statistics', + '--statistics', 'regress_pg_dump_test', ], }, @@ -395,7 +395,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--clean', '--file' => "$tempdir/defaults_no_public_clean.sql", - '--with-statistics', + '--statistics', 'regress_pg_dump_test', ], }, @@ -404,7 +404,7 @@ my %pgdump_runs = ( dump_cmd => [ 'pg_dump', '--no-sync', '--file' => "$tempdir/defaults_public_owner.sql", - '--with-statistics', + '--statistics', 'regress_public_owner', ], }, @@ -419,14 +419,14 @@ my %pgdump_runs = ( 'pg_dump', '--format' => 'custom', '--file' => "$tempdir/defaults_custom_format.dump", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--format' => 'custom', '--file' => "$tempdir/defaults_custom_format.sql", - '--with-statistics', + '--statistics', "$tempdir/defaults_custom_format.dump", ], command_like => { @@ -451,14 +451,14 @@ my %pgdump_runs = ( 'pg_dump', '--format' => 'directory', '--file' => "$tempdir/defaults_dir_format", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--format' => 'directory', '--file' => "$tempdir/defaults_dir_format.sql", - '--with-statistics', + '--statistics', "$tempdir/defaults_dir_format", ], command_like => { @@ -484,13 +484,13 @@ my %pgdump_runs = ( '--format' => 'directory', '--jobs' => 2, '--file' => "$tempdir/defaults_parallel", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--file' => "$tempdir/defaults_parallel.sql", - '--with-statistics', + '--statistics', "$tempdir/defaults_parallel", ], }, @@ -502,14 +502,14 @@ my %pgdump_runs = ( 'pg_dump', '--format' => 'tar', '--file' => "$tempdir/defaults_tar_format.tar", - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--format' => 'tar', '--file' => "$tempdir/defaults_tar_format.sql", - '--with-statistics', + '--statistics', "$tempdir/defaults_tar_format.tar", ], }, @@ -518,7 +518,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/exclude_dump_test_schema.sql", '--exclude-schema' => 'dump_test', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -527,7 +527,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/exclude_test_table.sql", '--exclude-table' => 'dump_test.test_table', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -536,7 +536,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/exclude_measurement.sql", '--exclude-table-and-children' => 'dump_test.measurement', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -546,7 +546,7 @@ my %pgdump_runs = ( '--file' => "$tempdir/exclude_measurement_data.sql", '--exclude-table-data-and-children' => 'dump_test.measurement', '--no-unlogged-table-data', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -556,7 +556,7 @@ my %pgdump_runs = ( '--file' => "$tempdir/exclude_test_table_data.sql", '--exclude-table-data' => 'dump_test.test_table', '--no-unlogged-table-data', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -575,7 +575,7 @@ my %pgdump_runs = ( '--file' => "$tempdir/pg_dumpall_globals.sql", '--globals-only', '--no-sync', - '--with-statistics', + '--statistics', ], }, pg_dumpall_globals_clean => { @@ -585,14 +585,14 @@ my %pgdump_runs = ( '--globals-only', '--clean', '--no-sync', - '--with-statistics', + '--statistics', ], }, pg_dumpall_dbprivs => { dump_cmd => [ 'pg_dumpall', '--no-sync', '--file' => "$tempdir/pg_dumpall_dbprivs.sql", - '--with-statistics', + '--statistics', ], }, pg_dumpall_exclude => { @@ -602,7 +602,7 @@ my %pgdump_runs = ( '--file' => "$tempdir/pg_dumpall_exclude.sql", '--exclude-database' => '*dump_test*', '--no-sync', - '--with-statistics', + '--statistics', ], }, no_toast_compression => { @@ -610,7 +610,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/no_toast_compression.sql", '--no-toast-compression', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -619,7 +619,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/no_large_objects.sql", '--no-large-objects', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -628,7 +628,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/no_policies.sql", '--no-policies', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -637,7 +637,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/no_privs.sql", '--no-privileges', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -646,7 +646,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/no_owner.sql", '--no-owner', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -655,7 +655,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/no_table_access_method.sql", '--no-table-access-method', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -664,7 +664,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/only_dump_test_schema.sql", '--schema' => 'dump_test', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -675,7 +675,7 @@ my %pgdump_runs = ( '--table' => 'dump_test.test_table', '--lock-wait-timeout' => (1000 * $PostgreSQL::Test::Utils::timeout_default), - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -686,7 +686,7 @@ my %pgdump_runs = ( '--table-and-children' => 'dump_test.measurement', '--lock-wait-timeout' => (1000 * $PostgreSQL::Test::Utils::timeout_default), - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -696,7 +696,7 @@ my %pgdump_runs = ( '--file' => "$tempdir/role.sql", '--role' => 'regress_dump_test_role', '--schema' => 'dump_test_second_schema', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -709,13 +709,13 @@ my %pgdump_runs = ( '--file' => "$tempdir/role_parallel", '--role' => 'regress_dump_test_role', '--schema' => 'dump_test_second_schema', - '--with-statistics', + '--statistics', 'postgres', ], restore_cmd => [ 'pg_restore', '--file' => "$tempdir/role_parallel.sql", - '--with-statistics', + '--statistics', "$tempdir/role_parallel", ], }, @@ -744,7 +744,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/section_pre_data.sql", '--section' => 'pre-data', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -753,7 +753,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/section_data.sql", '--section' => 'data', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -762,7 +762,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', '--file' => "$tempdir/section_post_data.sql", '--section' => 'post-data', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -773,7 +773,7 @@ my %pgdump_runs = ( '--schema' => 'dump_test', '--large-objects', '--no-large-objects', - '--with-statistics', + '--statistics', 'postgres', ], }, @@ -789,7 +789,7 @@ my %pgdump_runs = ( 'pg_dump', '--no-sync', "--file=$tempdir/no_data_no_schema.sql", '--no-data', '--no-schema', 'postgres', - '--with-statistics', + '--statistics', ], }, statistics_only => { @@ -803,7 +803,7 @@ my %pgdump_runs = ( dump_cmd => [ 'pg_dump', '--no-sync', "--file=$tempdir/no_schema.sql", '--no-schema', - '--with-statistics', 'postgres', + '--statistics', 'postgres', ], },); @@ -5206,10 +5206,10 @@ command_fails_like( '--port' => $port, '--strict-names', '--schema-only', - '--with-statistics', + '--statistics', ], - qr/\Qpg_dump: error: options -s\/--schema-only and --with-statistics cannot be used together\E/, - 'cannot use --schema-only and --with-statistics together'); + qr/\Qpg_dump: error: options -s\/--schema-only and --statistics cannot be used together\E/, + 'cannot use --schema-only and --statistics together'); command_fails_like( [ diff --git a/src/bin/pg_upgrade/dump.c b/src/bin/pg_upgrade/dump.c index 183f08ce1e8..55f6e7b4d9c 100644 --- a/src/bin/pg_upgrade/dump.c +++ b/src/bin/pg_upgrade/dump.c @@ -58,7 +58,7 @@ generate_old_dump(void) (user_opts.transfer_mode == TRANSFER_MODE_SWAP) ? "" : "--sequence-data", log_opts.verbose ? "--verbose" : "", - user_opts.do_statistics ? "--with-statistics" : "--no-statistics", + user_opts.do_statistics ? "--statistics" : "--no-statistics", log_opts.dumpdir, sql_file_name, escaped_connstr.data); diff --git a/src/bin/pgbench/pgbench.c b/src/bin/pgbench/pgbench.c index 497a936c141..125f3c7bbbe 100644 --- a/src/bin/pgbench/pgbench.c +++ b/src/bin/pgbench/pgbench.c @@ -3495,6 +3495,8 @@ doRetry(CState *st, pg_time_usec_t *now) static int discardUntilSync(CState *st) { + bool received_sync = false; + /* send a sync */ if (!PQpipelineSync(st->con)) { @@ -3509,10 +3511,21 @@ discardUntilSync(CState *st) PGresult *res = PQgetResult(st->con); if (PQresultStatus(res) == PGRES_PIPELINE_SYNC) + received_sync = true; + else if (received_sync) { - PQclear(res); - res = PQgetResult(st->con); + /* + * PGRES_PIPELINE_SYNC must be followed by another + * PGRES_PIPELINE_SYNC or NULL; otherwise, assert failure. + */ Assert(res == NULL); + + /* + * Reset ongoing sync count to 0 since all PGRES_PIPELINE_SYNC + * results have been discarded. + */ + st->num_syncs = 0; + PQclear(res); break; } PQclear(res); diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h index 5173d422d46..750a9d8a09b 100644 --- a/src/include/catalog/catversion.h +++ b/src/include/catalog/catversion.h @@ -57,6 +57,6 @@ */ /* yyyymmddN */ -#define CATALOG_VERSION_NO 202507231 +#define CATALOG_VERSION_NO 202508041 #endif diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat index 3ee8fed7e53..118d6da1ace 100644 --- a/src/include/catalog/pg_proc.dat +++ b/src/include/catalog/pg_proc.dat @@ -5688,9 +5688,9 @@ { oid => '6231', descr => 'statistics: information about subscription stats', proname => 'pg_stat_get_subscription_stats', provolatile => 's', proparallel => 'r', prorettype => 'record', proargtypes => 'oid', - proallargtypes => '{oid,oid,int8,int8,int8,int8,int8,int8,int8,int8,int8,timestamptz}', - proargmodes => '{i,o,o,o,o,o,o,o,o,o,o,o}', - proargnames => '{subid,subid,apply_error_count,sync_error_count,confl_insert_exists,confl_update_origin_differs,confl_update_exists,confl_update_missing,confl_delete_origin_differs,confl_delete_missing,confl_multiple_unique_conflicts,stats_reset}', + proallargtypes => '{oid,oid,int8,int8,int8,int8,int8,int8,int8,int8,int8,int8,timestamptz}', + proargmodes => '{i,o,o,o,o,o,o,o,o,o,o,o,o}', + proargnames => '{subid,subid,apply_error_count,sync_error_count,confl_insert_exists,confl_update_origin_differs,confl_update_exists,confl_update_deleted,confl_update_missing,confl_delete_origin_differs,confl_delete_missing,confl_multiple_unique_conflicts,stats_reset}', prosrc => 'pg_stat_get_subscription_stats' }, { oid => '6118', descr => 'statistics: information about subscription', proname => 'pg_stat_get_subscription', prorows => '10', proisstrict => 'f', diff --git a/src/include/executor/executor.h b/src/include/executor/executor.h index 104b059544d..a71502efeed 100644 --- a/src/include/executor/executor.h +++ b/src/include/executor/executor.h @@ -14,6 +14,7 @@ #ifndef EXECUTOR_H #define EXECUTOR_H +#include "datatype/timestamp.h" #include "executor/execdesc.h" #include "fmgr.h" #include "nodes/lockoptions.h" @@ -759,7 +760,18 @@ extern bool RelationFindReplTupleByIndex(Relation rel, Oid idxoid, TupleTableSlot *outslot); extern bool RelationFindReplTupleSeq(Relation rel, LockTupleMode lockmode, TupleTableSlot *searchslot, TupleTableSlot *outslot); - +extern bool RelationFindDeletedTupleInfoSeq(Relation rel, + TupleTableSlot *searchslot, + TransactionId oldestxmin, + TransactionId *delete_xid, + RepOriginId *delete_origin, + TimestampTz *delete_time); +extern bool RelationFindDeletedTupleInfoByIndex(Relation rel, Oid idxoid, + TupleTableSlot *searchslot, + TransactionId oldestxmin, + TransactionId *delete_xid, + RepOriginId *delete_origin, + TimestampTz *delete_time); extern void ExecSimpleRelationInsert(ResultRelInfo *resultRelInfo, EState *estate, TupleTableSlot *slot); extern void ExecSimpleRelationUpdate(ResultRelInfo *resultRelInfo, diff --git a/src/include/replication/conflict.h b/src/include/replication/conflict.h index 6c59125f256..ff3cb8416ec 100644 --- a/src/include/replication/conflict.h +++ b/src/include/replication/conflict.h @@ -32,6 +32,9 @@ typedef enum /* The updated row value violates unique constraint */ CT_UPDATE_EXISTS, + /* The row to be updated was concurrently deleted by a different origin */ + CT_UPDATE_DELETED, + /* The row to be updated is missing */ CT_UPDATE_MISSING, diff --git a/src/include/replication/worker_internal.h b/src/include/replication/worker_internal.h index 0c7b8440a61..7c0204dd6f4 100644 --- a/src/include/replication/worker_internal.h +++ b/src/include/replication/worker_internal.h @@ -87,8 +87,9 @@ typedef struct LogicalRepWorker bool parallel_apply; /* - * The changes made by this and later transactions must be retained to - * ensure reliable conflict detection during the apply phase. + * Changes made by this transaction and subsequent ones must be preserved. + * This ensures that update_deleted conflicts can be accurately detected + * during the apply phase of logical replication by this worker. * * The logical replication launcher manages an internal replication slot * named "pg_conflict_detection". It asynchronously collects this ID to diff --git a/src/include/utils/catcache.h b/src/include/utils/catcache.h index 277ec33c00b..00808e23f49 100644 --- a/src/include/utils/catcache.h +++ b/src/include/utils/catcache.h @@ -87,6 +87,14 @@ typedef struct catcache typedef struct catctup { + /* + * Each tuple in a cache is a member of a dlist that stores the elements + * of its hash bucket. We keep each dlist in LRU order to speed repeated + * lookups. Keep the dlist_node field first so that Valgrind understands + * the struct is reachable. + */ + dlist_node cache_elem; /* list member of per-bucket list */ + int ct_magic; /* for identifying CatCTup entries */ #define CT_MAGIC 0x57261502 @@ -99,13 +107,6 @@ typedef struct catctup Datum keys[CATCACHE_MAXKEYS]; /* - * Each tuple in a cache is a member of a dlist that stores the elements - * of its hash bucket. We keep each dlist in LRU order to speed repeated - * lookups. - */ - dlist_node cache_elem; /* list member of per-bucket list */ - - /* * A tuple marked "dead" must not be returned by subsequent searches. * However, it won't be physically deleted from the cache until its * refcount goes to zero. (If it's a member of a CatCList, the list's @@ -158,13 +159,17 @@ typedef struct catctup */ typedef struct catclist { + /* + * Keep the dlist_node field first so that Valgrind understands the struct + * is reachable. + */ + dlist_node cache_elem; /* list member of per-catcache list */ + int cl_magic; /* for identifying CatCList entries */ #define CL_MAGIC 0x52765103 uint32 hash_value; /* hash value for lookup keys */ - dlist_node cache_elem; /* list member of per-catcache list */ - /* * Lookup keys for the entry, with the first nkeys elements being valid. * All by-reference are separately allocated. diff --git a/src/include/utils/memdebug.h b/src/include/utils/memdebug.h index 7309271834b..80692dcef93 100644 --- a/src/include/utils/memdebug.h +++ b/src/include/utils/memdebug.h @@ -29,6 +29,7 @@ #define VALGRIND_MEMPOOL_ALLOC(context, addr, size) do {} while (0) #define VALGRIND_MEMPOOL_FREE(context, addr) do {} while (0) #define VALGRIND_MEMPOOL_CHANGE(context, optr, nptr, size) do {} while (0) +#define VALGRIND_MEMPOOL_TRIM(context, addr, size) do {} while (0) #endif diff --git a/src/pl/plpgsql/src/pl_comp.c b/src/pl/plpgsql/src/pl_comp.c index ee961425a5b..f6976689a69 100644 --- a/src/pl/plpgsql/src/pl_comp.c +++ b/src/pl/plpgsql/src/pl_comp.c @@ -177,6 +177,7 @@ plpgsql_compile_callback(FunctionCallInfo fcinfo, yyscan_t scanner; Datum prosrcdatum; char *proc_source; + char *proc_signature; HeapTuple typeTup; Form_pg_type typeStruct; PLpgSQL_variable *var; @@ -223,6 +224,9 @@ plpgsql_compile_callback(FunctionCallInfo fcinfo, plpgsql_check_syntax = forValidator; plpgsql_curr_compile = function; + /* format_procedure leaks memory, so run it in temp context */ + proc_signature = format_procedure(fcinfo->flinfo->fn_oid); + /* * All the permanent output of compilation (e.g. parse tree) is kept in a * per-function memory context, so it can be reclaimed easily. @@ -237,7 +241,7 @@ plpgsql_compile_callback(FunctionCallInfo fcinfo, ALLOCSET_DEFAULT_SIZES); plpgsql_compile_tmp_cxt = MemoryContextSwitchTo(func_cxt); - function->fn_signature = format_procedure(fcinfo->flinfo->fn_oid); + function->fn_signature = pstrdup(proc_signature); MemoryContextSetIdentifier(func_cxt, function->fn_signature); function->fn_oid = fcinfo->flinfo->fn_oid; function->fn_input_collation = fcinfo->fncollation; @@ -1673,6 +1677,11 @@ plpgsql_parse_wordrowtype(char *ident) { Oid classOid; Oid typOid; + TypeName *typName; + MemoryContext oldCxt; + + /* Avoid memory leaks in long-term function context */ + oldCxt = MemoryContextSwitchTo(plpgsql_compile_tmp_cxt); /* * Look up the relation. Note that because relation rowtypes have the @@ -1695,9 +1704,12 @@ plpgsql_parse_wordrowtype(char *ident) errmsg("relation \"%s\" does not have a composite type", ident))); + typName = makeTypeName(ident); + + MemoryContextSwitchTo(oldCxt); + /* Build and return the row type struct */ - return plpgsql_build_datatype(typOid, -1, InvalidOid, - makeTypeName(ident)); + return plpgsql_build_datatype(typOid, -1, InvalidOid, typName); } /* ---------- @@ -1711,6 +1723,7 @@ plpgsql_parse_cwordrowtype(List *idents) Oid classOid; Oid typOid; RangeVar *relvar; + TypeName *typName; MemoryContext oldCxt; /* @@ -1733,11 +1746,12 @@ plpgsql_parse_cwordrowtype(List *idents) errmsg("relation \"%s\" does not have a composite type", relvar->relname))); + typName = makeTypeNameFromNameList(idents); + MemoryContextSwitchTo(oldCxt); /* Build and return the row type struct */ - return plpgsql_build_datatype(typOid, -1, InvalidOid, - makeTypeNameFromNameList(idents)); + return plpgsql_build_datatype(typOid, -1, InvalidOid, typName); } /* @@ -1952,6 +1966,8 @@ plpgsql_build_recfield(PLpgSQL_rec *rec, const char *fldname) * origtypname is the parsed form of what the user wrote as the type name. * It can be NULL if the type could not be a composite type, or if it was * identified by OID to begin with (e.g., it's a function argument type). + * origtypname is in short-lived storage and must be copied if we choose + * to incorporate it into the function's parse tree. */ PLpgSQL_type * plpgsql_build_datatype(Oid typeOid, int32 typmod, @@ -2070,7 +2086,7 @@ build_datatype(HeapTuple typeTup, int32 typmod, errmsg("type %s is not composite", format_type_be(typ->typoid)))); - typ->origtypname = origtypname; + typ->origtypname = copyObject(origtypname); typ->tcache = typentry; typ->tupdesc_id = typentry->tupDesc_identifier; } diff --git a/src/pl/plpgsql/src/pl_gram.y b/src/pl/plpgsql/src/pl_gram.y index 7b672ea5179..17568d82554 100644 --- a/src/pl/plpgsql/src/pl_gram.y +++ b/src/pl/plpgsql/src/pl_gram.y @@ -3853,6 +3853,7 @@ parse_datatype(const char *string, int location, yyscan_t yyscanner) int32 typmod; sql_error_callback_arg cbarg; ErrorContextCallback syntax_errcontext; + MemoryContext oldCxt; cbarg.location = location; cbarg.yyscanner = yyscanner; @@ -3862,9 +3863,14 @@ parse_datatype(const char *string, int location, yyscan_t yyscanner) syntax_errcontext.previous = error_context_stack; error_context_stack = &syntax_errcontext; - /* Let the main parser try to parse it under standard SQL rules */ + /* + * Let the main parser try to parse it under standard SQL rules. The + * parser leaks memory, so run it in temp context. + */ + oldCxt = MemoryContextSwitchTo(plpgsql_compile_tmp_cxt); typeName = typeStringToTypeName(string, NULL); typenameTypeIdAndMod(NULL, typeName, &type_id, &typmod); + MemoryContextSwitchTo(oldCxt); /* Restore former ereport callback */ error_context_stack = syntax_errcontext.previous; diff --git a/src/test/regress/expected/foreign_key.out b/src/test/regress/expected/foreign_key.out index f9bd252444f..dc541d61adf 100644 --- a/src/test/regress/expected/foreign_key.out +++ b/src/test/regress/expected/foreign_key.out @@ -1750,7 +1750,7 @@ Indexes: Referenced by: TABLE "fk_partitioned_fk" CONSTRAINT "fk_partitioned_fk_a_b_fkey" FOREIGN KEY (a, b) REFERENCES fk_notpartitioned_pk(a, b) --- Check the exsting FK trigger +-- Check the existing FK trigger SELECT conname, tgrelid::regclass as tgrel, regexp_replace(tgname, '[0-9]+', 'N') as tgname, tgtype FROM pg_trigger t JOIN pg_constraint c ON (t.tgconstraint = c.oid) WHERE tgrelid IN (SELECT relid FROM pg_partition_tree('fk_partitioned_fk'::regclass) diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out index dce8c672b40..6509fda77a9 100644 --- a/src/test/regress/expected/rules.out +++ b/src/test/regress/expected/rules.out @@ -2179,13 +2179,14 @@ pg_stat_subscription_stats| SELECT ss.subid, ss.confl_insert_exists, ss.confl_update_origin_differs, ss.confl_update_exists, + ss.confl_update_deleted, ss.confl_update_missing, ss.confl_delete_origin_differs, ss.confl_delete_missing, ss.confl_multiple_unique_conflicts, ss.stats_reset FROM pg_subscription s, - LATERAL pg_stat_get_subscription_stats(s.oid) ss(subid, apply_error_count, sync_error_count, confl_insert_exists, confl_update_origin_differs, confl_update_exists, confl_update_missing, confl_delete_origin_differs, confl_delete_missing, confl_multiple_unique_conflicts, stats_reset); + LATERAL pg_stat_get_subscription_stats(s.oid) ss(subid, apply_error_count, sync_error_count, confl_insert_exists, confl_update_origin_differs, confl_update_exists, confl_update_deleted, confl_update_missing, confl_delete_origin_differs, confl_delete_missing, confl_multiple_unique_conflicts, stats_reset); pg_stat_sys_indexes| SELECT relid, indexrelid, schemaname, diff --git a/src/test/regress/sql/foreign_key.sql b/src/test/regress/sql/foreign_key.sql index cfcecb4e911..39174ad1eb9 100644 --- a/src/test/regress/sql/foreign_key.sql +++ b/src/test/regress/sql/foreign_key.sql @@ -1296,7 +1296,7 @@ UPDATE fk_notpartitioned_pk SET b = 2504 WHERE a = 2500; -- check psql behavior \d fk_notpartitioned_pk --- Check the exsting FK trigger +-- Check the existing FK trigger SELECT conname, tgrelid::regclass as tgrel, regexp_replace(tgname, '[0-9]+', 'N') as tgname, tgtype FROM pg_trigger t JOIN pg_constraint c ON (t.tgconstraint = c.oid) WHERE tgrelid IN (SELECT relid FROM pg_partition_tree('fk_partitioned_fk'::regclass) diff --git a/src/test/subscription/t/035_conflicts.pl b/src/test/subscription/t/035_conflicts.pl index 976d53a870e..36aeb14c563 100644 --- a/src/test/subscription/t/035_conflicts.pl +++ b/src/test/subscription/t/035_conflicts.pl @@ -150,7 +150,9 @@ pass('multiple_unique_conflicts detected on a leaf partition during insert'); # Setup a bidirectional logical replication between node_A & node_B ############################################################################### -# Initialize nodes. +# Initialize nodes. Enable the track_commit_timestamp on both nodes to detect +# the conflict when attempting to update a row that was previously modified by +# a different origin. # node_A. Increase the log_min_messages setting to DEBUG2 to debug test # failures. Disable autovacuum to avoid generating xid that could affect the @@ -158,7 +160,8 @@ pass('multiple_unique_conflicts detected on a leaf partition during insert'); my $node_A = $node_publisher; $node_A->append_conf( 'postgresql.conf', - qq{autovacuum = off + qq{track_commit_timestamp = on + autovacuum = off log_min_messages = 'debug2'}); $node_A->restart; @@ -270,6 +273,8 @@ $node_A->psql('postgres', ############################################################################### # Check that dead tuples on node A cannot be cleaned by VACUUM until the # concurrent transactions on Node B have been applied and flushed on Node A. +# Also, check that an update_deleted conflict is detected when updating a row +# that was deleted by a different origin. ############################################################################### # Insert a record @@ -288,6 +293,8 @@ $node_A->poll_query_until('postgres', "SELECT count(*) = 0 FROM pg_stat_activity WHERE backend_type = 'logical replication apply worker'" ); +my $log_location = -s $node_B->logfile; + $node_B->safe_psql('postgres', "UPDATE tab SET b = 3 WHERE a = 1;"); $node_A->safe_psql('postgres', "DELETE FROM tab WHERE a = 1;"); @@ -299,10 +306,30 @@ ok( $stderr =~ qr/1 are dead but not yet removable/, 'the deleted column is non-removable'); +# Ensure the DELETE is replayed on Node B +$node_A->wait_for_catchup($subname_BA); + +# Check the conflict detected on Node B +my $logfile = slurp_file($node_B->logfile(), $log_location); +ok( $logfile =~ + qr/conflict detected on relation "public.tab": conflict=delete_origin_differs.* +.*DETAIL:.* Deleting the row that was modified locally in transaction [0-9]+ at .* +.*Existing local tuple \(1, 3\); replica identity \(a\)=\(1\)/, + 'delete target row was modified in tab'); + +$log_location = -s $node_A->logfile; + $node_A->safe_psql( 'postgres', "ALTER SUBSCRIPTION $subname_AB ENABLE;"); $node_B->wait_for_catchup($subname_AB); +$logfile = slurp_file($node_A->logfile(), $log_location); +ok( $logfile =~ + qr/conflict detected on relation "public.tab": conflict=update_deleted.* +.*DETAIL:.* The row to be updated was deleted locally in transaction [0-9]+ at .* +.*Remote tuple \(1, 3\); replica identity \(a\)=\(1\)/, + 'update target row was deleted in tab'); + # Remember the next transaction ID to be assigned my $next_xid = $node_A->safe_psql('postgres', "SELECT txid_current() + 1;"); @@ -325,6 +352,41 @@ ok( $stderr =~ 'the deleted column is removed'); ############################################################################### +# Ensure that the deleted tuple needed to detect an update_deleted conflict is +# accessible via a sequential table scan. +############################################################################### + +# Drop the primary key from tab on node A and set REPLICA IDENTITY to FULL to +# enforce sequential scanning of the table. +$node_A->safe_psql('postgres', "ALTER TABLE tab REPLICA IDENTITY FULL"); +$node_B->safe_psql('postgres', "ALTER TABLE tab REPLICA IDENTITY FULL"); +$node_A->safe_psql('postgres', "ALTER TABLE tab DROP CONSTRAINT tab_pkey;"); + +# Disable the logical replication from node B to node A +$node_A->safe_psql('postgres', "ALTER SUBSCRIPTION $subname_AB DISABLE"); + +# Wait for the apply worker to stop +$node_A->poll_query_until('postgres', + "SELECT count(*) = 0 FROM pg_stat_activity WHERE backend_type = 'logical replication apply worker'" +); + +$node_B->safe_psql('postgres', "UPDATE tab SET b = 4 WHERE a = 2;"); +$node_A->safe_psql('postgres', "DELETE FROM tab WHERE a = 2;"); + +$log_location = -s $node_A->logfile; + +$node_A->safe_psql( + 'postgres', "ALTER SUBSCRIPTION $subname_AB ENABLE;"); +$node_B->wait_for_catchup($subname_AB); + +$logfile = slurp_file($node_A->logfile(), $log_location); +ok( $logfile =~ + qr/conflict detected on relation "public.tab": conflict=update_deleted.* +.*DETAIL:.* The row to be updated was deleted locally in transaction [0-9]+ at .* +.*Remote tuple \(2, 4\); replica identity full \(2, 2\)/, + 'update target row was deleted in tab'); + +############################################################################### # Check that the replication slot pg_conflict_detection is dropped after # removing all the subscriptions. ############################################################################### diff --git a/src/tools/valgrind.supp b/src/tools/valgrind.supp index 2ad5b81526d..3880007dfb3 100644 --- a/src/tools/valgrind.supp +++ b/src/tools/valgrind.supp @@ -194,3 +194,36 @@ Memcheck:Addr8 fun:pg_numa_touch_mem_if_required } + + +# Memory-leak suppressions +# Note that a suppression rule will silence complaints about memory blocks +# allocated in matching places, but it won't prevent "indirectly lost" +# complaints about blocks that are only reachable via the suppressed blocks. + +# Suppress complaints about stuff leaked during function cache loading. +# Both the PL/pgSQL and SQL-function parsing processes generate some cruft +# within the function's cache context, which doesn't seem worth the trouble +# to get rid of. Moreover, there are cases where CachedFunction structs +# are intentionally leaked because we're unsure if any fn_extra pointers +# remain. +{ + hide_function_cache_leaks + Memcheck:Leak + match-leak-kinds: definite,possible,indirect + + ... + fun:cached_function_compile +} + +# Suppress complaints about stuff leaked during TS dictionary loading. +# Not very much is typically lost there, and preventing it would +# require a risky API change for TS tmplinit functions. +{ + hide_ts_dictionary_leaks + Memcheck:Leak + match-leak-kinds: definite,possible,indirect + + ... + fun:lookup_ts_dictionary_cache +} |