From 87985cc9252296f11db3cacc155e8d710d2e9b9c Mon Sep 17 00:00:00 2001 From: Alexander Korotkov Date: Tue, 26 Mar 2024 01:27:56 +0200 Subject: [PATCH] Allow locking updated tuples in tuple_update() and tuple_delete() Currently, in read committed transaction isolation mode (default), we have the following sequence of actions when tuple_update()/tuple_delete() finds the tuple updated by the concurrent transaction. 1. Attempt to update/delete tuple with tuple_update()/tuple_delete(), which returns TM_Updated. 2. Lock tuple with tuple_lock(). 3. Re-evaluate plan qual (recheck if we still need to update/delete and calculate the new tuple for update). 4. Second attempt to update/delete tuple with tuple_update()/tuple_delete(). This attempt should be successful, since the tuple was previously locked. This commit eliminates step 2 by taking the lock during the first tuple_update()/tuple_delete() call. The heap table access method saves some effort by checking the updated tuple once instead of twice. Future undo-based table access methods, which will start from the latest row version, can immediately place a lock there. Also, this commit makes tuple_update()/tuple_delete() optionally save the old tuple into the dedicated slot. That saves efforts on re-fetching tuples in certain cases. The code in nodeModifyTable.c is simplified by removing the nested switch/case. Discussion: https://postgr.es/m/CAPpHfdua-YFw3XTprfutzGp28xXLigFtzNbuFY8yPhqeq6X5kg%40mail.gmail.com Reviewed-by: Aleksander Alekseev, Pavel Borisov, Vignesh C, Mason Sharp Reviewed-by: Andres Freund, Chris Travers --- src/backend/access/heap/heapam.c | 205 +++++++++---- src/backend/access/heap/heapam_handler.c | 94 ++++-- src/backend/access/table/tableam.c | 26 +- src/backend/commands/trigger.c | 55 +--- src/backend/executor/execReplication.c | 19 +- src/backend/executor/nodeModifyTable.c | 353 ++++++++++------------- src/include/access/heapam.h | 19 +- src/include/access/tableam.h | 73 +++-- src/include/commands/trigger.h | 4 +- 9 files changed, 502 insertions(+), 346 deletions(-) diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c index cc67dd813d..9a8c8e3348 100644 --- a/src/backend/access/heap/heapam.c +++ b/src/backend/access/heap/heapam.c @@ -2496,10 +2496,11 @@ xmax_infomask_changed(uint16 new_infomask, uint16 old_infomask) } /* - * heap_delete - delete a tuple + * heap_delete - delete a tuple, optionally fetching it into a slot * * See table_tuple_delete() for an explanation of the parameters, except that - * this routine directly takes a tuple rather than a slot. + * this routine directly takes a tuple rather than a slot. Also, we don't + * place a lock on the tuple in this function, just fetch the existing version. * * In the failure cases, the routine fills *tmfd with the tuple's t_ctid, * t_xmax (resolving a possible MultiXact, if necessary), and t_cmax (the last @@ -2508,8 +2509,9 @@ xmax_infomask_changed(uint16 new_infomask, uint16 old_infomask) */ TM_Result heap_delete(Relation relation, ItemPointer tid, - CommandId cid, Snapshot crosscheck, bool wait, - TM_FailureData *tmfd, bool changingPart) + CommandId cid, Snapshot crosscheck, int options, + TM_FailureData *tmfd, bool changingPart, + TupleTableSlot *oldSlot) { TM_Result result; TransactionId xid = GetCurrentTransactionId(); @@ -2587,7 +2589,7 @@ l1: (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), errmsg("attempted to delete invisible tuple"))); } - else if (result == TM_BeingModified && wait) + else if (result == TM_BeingModified && (options & TABLE_MODIFY_WAIT)) { TransactionId xwait; uint16 infomask; @@ -2728,7 +2730,30 @@ l1: tmfd->cmax = HeapTupleHeaderGetCmax(tp.t_data); else tmfd->cmax = InvalidCommandId; - UnlockReleaseBuffer(buffer); + + /* + * If we're asked to lock the updated tuple, we just fetch the + * existing tuple. That let's the caller save some resources on + * placing the lock. + */ + if (result == TM_Updated && + (options & TABLE_MODIFY_LOCK_UPDATED)) + { + BufferHeapTupleTableSlot *bslot; + + Assert(TTS_IS_BUFFERTUPLE(oldSlot)); + bslot = (BufferHeapTupleTableSlot *) oldSlot; + + LockBuffer(buffer, BUFFER_LOCK_UNLOCK); + bslot->base.tupdata = tp; + ExecStorePinnedBufferHeapTuple(&bslot->base.tupdata, + oldSlot, + buffer); + } + else + { + UnlockReleaseBuffer(buffer); + } if (have_tuple_lock) UnlockTupleTuplock(relation, &(tp.t_self), LockTupleExclusive); if (vmbuffer != InvalidBuffer) @@ -2902,8 +2927,24 @@ l1: */ CacheInvalidateHeapTuple(relation, &tp, NULL); - /* Now we can release the buffer */ - ReleaseBuffer(buffer); + /* Fetch the old tuple version if we're asked for that. */ + if (options & TABLE_MODIFY_FETCH_OLD_TUPLE) + { + BufferHeapTupleTableSlot *bslot; + + Assert(TTS_IS_BUFFERTUPLE(oldSlot)); + bslot = (BufferHeapTupleTableSlot *) oldSlot; + + bslot->base.tupdata = tp; + ExecStorePinnedBufferHeapTuple(&bslot->base.tupdata, + oldSlot, + buffer); + } + else + { + /* Now we can release the buffer */ + ReleaseBuffer(buffer); + } /* * Release the lmgr tuple lock, if we had it. @@ -2935,8 +2976,8 @@ simple_heap_delete(Relation relation, ItemPointer tid) result = heap_delete(relation, tid, GetCurrentCommandId(true), InvalidSnapshot, - true /* wait for commit */ , - &tmfd, false /* changingPart */ ); + TABLE_MODIFY_WAIT /* wait for commit */ , + &tmfd, false /* changingPart */ , NULL); switch (result) { case TM_SelfModified: @@ -2963,10 +3004,11 @@ simple_heap_delete(Relation relation, ItemPointer tid) } /* - * heap_update - replace a tuple + * heap_update - replace a tuple, optionally fetching it into a slot * * See table_tuple_update() for an explanation of the parameters, except that - * this routine directly takes a tuple rather than a slot. + * this routine directly takes a tuple rather than a slot. Also, we don't + * place a lock on the tuple in this function, just fetch the existing version. * * In the failure cases, the routine fills *tmfd with the tuple's t_ctid, * t_xmax (resolving a possible MultiXact, if necessary), and t_cmax (the last @@ -2975,9 +3017,9 @@ simple_heap_delete(Relation relation, ItemPointer tid) */ TM_Result heap_update(Relation relation, ItemPointer otid, HeapTuple newtup, - CommandId cid, Snapshot crosscheck, bool wait, + CommandId cid, Snapshot crosscheck, int options, TM_FailureData *tmfd, LockTupleMode *lockmode, - TU_UpdateIndexes *update_indexes) + TU_UpdateIndexes *update_indexes, TupleTableSlot *oldSlot) { TM_Result result; TransactionId xid = GetCurrentTransactionId(); @@ -3154,7 +3196,7 @@ l2: result = HeapTupleSatisfiesUpdate(&oldtup, cid, buffer); /* see below about the "no wait" case */ - Assert(result != TM_BeingModified || wait); + Assert(result != TM_BeingModified || (options & TABLE_MODIFY_WAIT)); if (result == TM_Invisible) { @@ -3163,7 +3205,7 @@ l2: (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE), errmsg("attempted to update invisible tuple"))); } - else if (result == TM_BeingModified && wait) + else if (result == TM_BeingModified && (options & TABLE_MODIFY_WAIT)) { TransactionId xwait; uint16 infomask; @@ -3367,7 +3409,30 @@ l2: tmfd->cmax = HeapTupleHeaderGetCmax(oldtup.t_data); else tmfd->cmax = InvalidCommandId; - UnlockReleaseBuffer(buffer); + + /* + * If we're asked to lock the updated tuple, we just fetch the + * existing tuple. That let's the caller save some resouces on + * placing the lock. + */ + if (result == TM_Updated && + (options & TABLE_MODIFY_LOCK_UPDATED)) + { + BufferHeapTupleTableSlot *bslot; + + Assert(TTS_IS_BUFFERTUPLE(oldSlot)); + bslot = (BufferHeapTupleTableSlot *) oldSlot; + + LockBuffer(buffer, BUFFER_LOCK_UNLOCK); + bslot->base.tupdata = oldtup; + ExecStorePinnedBufferHeapTuple(&bslot->base.tupdata, + oldSlot, + buffer); + } + else + { + UnlockReleaseBuffer(buffer); + } if (have_tuple_lock) UnlockTupleTuplock(relation, &(oldtup.t_self), *lockmode); if (vmbuffer != InvalidBuffer) @@ -3846,7 +3911,26 @@ l2: /* Now we can release the buffer(s) */ if (newbuf != buffer) ReleaseBuffer(newbuf); - ReleaseBuffer(buffer); + + /* Fetch the old tuple version if we're asked for that. */ + if (options & TABLE_MODIFY_FETCH_OLD_TUPLE) + { + BufferHeapTupleTableSlot *bslot; + + Assert(TTS_IS_BUFFERTUPLE(oldSlot)); + bslot = (BufferHeapTupleTableSlot *) oldSlot; + + bslot->base.tupdata = oldtup; + ExecStorePinnedBufferHeapTuple(&bslot->base.tupdata, + oldSlot, + buffer); + } + else + { + /* Now we can release the buffer */ + ReleaseBuffer(buffer); + } + if (BufferIsValid(vmbuffer_new)) ReleaseBuffer(vmbuffer_new); if (BufferIsValid(vmbuffer)) @@ -4054,8 +4138,8 @@ simple_heap_update(Relation relation, ItemPointer otid, HeapTuple tup, result = heap_update(relation, otid, tup, GetCurrentCommandId(true), InvalidSnapshot, - true /* wait for commit */ , - &tmfd, &lockmode, update_indexes); + TABLE_MODIFY_WAIT /* wait for commit */ , + &tmfd, &lockmode, update_indexes, NULL); switch (result) { case TM_SelfModified: @@ -4118,12 +4202,14 @@ get_mxact_status_for_lock(LockTupleMode mode, bool is_update) * tuples. * * Output parameters: - * *tuple: all fields filled in - * *buffer: set to buffer holding tuple (pinned but not locked at exit) + * *slot: BufferHeapTupleTableSlot filled with tuple * *tmfd: filled in failure cases (see below) * * Function results are the same as the ones for table_tuple_lock(). * + * If *slot already contains the target tuple, it takes advantage on that by + * skipping the ReadBuffer() call. + * * In the failure cases other than TM_Invisible, the routine fills * *tmfd with the tuple's t_ctid, t_xmax (resolving a possible MultiXact, * if necessary), and t_cmax (the last only for TM_SelfModified, @@ -4134,15 +4220,14 @@ get_mxact_status_for_lock(LockTupleMode mode, bool is_update) * See README.tuplock for a thorough explanation of this mechanism. */ TM_Result -heap_lock_tuple(Relation relation, HeapTuple tuple, +heap_lock_tuple(Relation relation, ItemPointer tid, TupleTableSlot *slot, CommandId cid, LockTupleMode mode, LockWaitPolicy wait_policy, - bool follow_updates, - Buffer *buffer, TM_FailureData *tmfd) + bool follow_updates, TM_FailureData *tmfd) { TM_Result result; - ItemPointer tid = &(tuple->t_self); ItemId lp; Page page; + Buffer buffer; Buffer vmbuffer = InvalidBuffer; BlockNumber block; TransactionId xid, @@ -4154,8 +4239,24 @@ heap_lock_tuple(Relation relation, HeapTuple tuple, bool skip_tuple_lock = false; bool have_tuple_lock = false; bool cleared_all_frozen = false; + BufferHeapTupleTableSlot *bslot = (BufferHeapTupleTableSlot *) slot; + HeapTuple tuple = &bslot->base.tupdata; - *buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid)); + Assert(TTS_IS_BUFFERTUPLE(slot)); + + /* Take advantage if slot already contains the relevant tuple */ + if (!TTS_EMPTY(slot) && + slot->tts_tableOid == relation->rd_id && + ItemPointerCompare(&slot->tts_tid, tid) == 0 && + BufferIsValid(bslot->buffer)) + { + buffer = bslot->buffer; + IncrBufferRefCount(buffer); + } + else + { + buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid)); + } block = ItemPointerGetBlockNumber(tid); /* @@ -4164,21 +4265,22 @@ heap_lock_tuple(Relation relation, HeapTuple tuple, * in the middle of changing this, so we'll need to recheck after we have * the lock. */ - if (PageIsAllVisible(BufferGetPage(*buffer))) + if (PageIsAllVisible(BufferGetPage(buffer))) visibilitymap_pin(relation, block, &vmbuffer); - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); - page = BufferGetPage(*buffer); + page = BufferGetPage(buffer); lp = PageGetItemId(page, ItemPointerGetOffsetNumber(tid)); Assert(ItemIdIsNormal(lp)); + tuple->t_self = *tid; tuple->t_data = (HeapTupleHeader) PageGetItem(page, lp); tuple->t_len = ItemIdGetLength(lp); tuple->t_tableOid = RelationGetRelid(relation); l3: - result = HeapTupleSatisfiesUpdate(tuple, cid, *buffer); + result = HeapTupleSatisfiesUpdate(tuple, cid, buffer); if (result == TM_Invisible) { @@ -4207,7 +4309,7 @@ l3: infomask2 = tuple->t_data->t_infomask2; ItemPointerCopy(&tuple->t_data->t_ctid, &t_ctid); - LockBuffer(*buffer, BUFFER_LOCK_UNLOCK); + LockBuffer(buffer, BUFFER_LOCK_UNLOCK); /* * If any subtransaction of the current top transaction already holds @@ -4359,12 +4461,12 @@ l3: { result = res; /* recovery code expects to have buffer lock held */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto failed; } } - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); /* * Make sure it's still an appropriate lock, else start over. @@ -4399,7 +4501,7 @@ l3: if (HEAP_XMAX_IS_LOCKED_ONLY(infomask) && !HEAP_XMAX_IS_EXCL_LOCKED(infomask)) { - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); /* * Make sure it's still an appropriate lock, else start over. @@ -4427,7 +4529,7 @@ l3: * No conflict, but if the xmax changed under us in the * meantime, start over. */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) || !TransactionIdEquals(HeapTupleHeaderGetRawXmax(tuple->t_data), xwait)) @@ -4439,7 +4541,7 @@ l3: } else if (HEAP_XMAX_IS_KEYSHR_LOCKED(infomask)) { - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); /* if the xmax changed in the meantime, start over */ if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) || @@ -4467,7 +4569,7 @@ l3: TransactionIdIsCurrentTransactionId(xwait)) { /* ... but if the xmax changed in the meantime, start over */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) || !TransactionIdEquals(HeapTupleHeaderGetRawXmax(tuple->t_data), xwait)) @@ -4489,7 +4591,7 @@ l3: */ if (require_sleep && (result == TM_Updated || result == TM_Deleted)) { - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto failed; } else if (require_sleep) @@ -4514,7 +4616,7 @@ l3: */ result = TM_WouldBlock; /* recovery code expects to have buffer lock held */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto failed; } @@ -4540,7 +4642,7 @@ l3: { result = TM_WouldBlock; /* recovery code expects to have buffer lock held */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto failed; } break; @@ -4580,7 +4682,7 @@ l3: { result = TM_WouldBlock; /* recovery code expects to have buffer lock held */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto failed; } break; @@ -4606,12 +4708,12 @@ l3: { result = res; /* recovery code expects to have buffer lock held */ - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto failed; } } - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); /* * xwait is done, but if xwait had just locked the tuple then some @@ -4633,7 +4735,7 @@ l3: * don't check for this in the multixact case, because some * locker transactions might still be running. */ - UpdateXmaxHintBits(tuple->t_data, *buffer, xwait); + UpdateXmaxHintBits(tuple->t_data, buffer, xwait); } } @@ -4692,9 +4794,9 @@ failed: */ if (vmbuffer == InvalidBuffer && PageIsAllVisible(page)) { - LockBuffer(*buffer, BUFFER_LOCK_UNLOCK); + LockBuffer(buffer, BUFFER_LOCK_UNLOCK); visibilitymap_pin(relation, block, &vmbuffer); - LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE); + LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE); goto l3; } @@ -4757,7 +4859,7 @@ failed: cleared_all_frozen = true; - MarkBufferDirty(*buffer); + MarkBufferDirty(buffer); /* * XLOG stuff. You might think that we don't need an XLOG record because @@ -4777,7 +4879,7 @@ failed: XLogRecPtr recptr; XLogBeginInsert(); - XLogRegisterBuffer(0, *buffer, REGBUF_STANDARD); + XLogRegisterBuffer(0, buffer, REGBUF_STANDARD); xlrec.offnum = ItemPointerGetOffsetNumber(&tuple->t_self); xlrec.xmax = xid; @@ -4798,7 +4900,7 @@ failed: result = TM_Ok; out_locked: - LockBuffer(*buffer, BUFFER_LOCK_UNLOCK); + LockBuffer(buffer, BUFFER_LOCK_UNLOCK); out_unlocked: if (BufferIsValid(vmbuffer)) @@ -4816,6 +4918,9 @@ out_unlocked: if (have_tuple_lock) UnlockTupleTuplock(relation, tid, mode); + /* Put the target tuple to the slot */ + ExecStorePinnedBufferHeapTuple(tuple, slot, buffer); + return result; } diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c index 2b7c702642..6abfe36dec 100644 --- a/src/backend/access/heap/heapam_handler.c +++ b/src/backend/access/heap/heapam_handler.c @@ -45,6 +45,12 @@ #include "utils/builtins.h" #include "utils/rel.h" +static TM_Result heapam_tuple_lock(Relation relation, ItemPointer tid, + Snapshot snapshot, TupleTableSlot *slot, + CommandId cid, LockTupleMode mode, + LockWaitPolicy wait_policy, uint8 flags, + TM_FailureData *tmfd); + static void reform_and_rewrite_tuple(HeapTuple tuple, Relation OldHeap, Relation NewHeap, Datum *values, bool *isnull, RewriteState rwstate); @@ -300,23 +306,55 @@ heapam_tuple_complete_speculative(Relation relation, TupleTableSlot *slot, static TM_Result heapam_tuple_delete(Relation relation, ItemPointer tid, CommandId cid, - Snapshot snapshot, Snapshot crosscheck, bool wait, - TM_FailureData *tmfd, bool changingPart) + Snapshot snapshot, Snapshot crosscheck, int options, + TM_FailureData *tmfd, bool changingPart, + TupleTableSlot *oldSlot) { + TM_Result result; + /* * Currently Deleting of index tuples are handled at vacuum, in case if * the storage itself is cleaning the dead tuples by itself, it is the * time to call the index tuple deletion also. */ - return heap_delete(relation, tid, cid, crosscheck, wait, tmfd, changingPart); + result = heap_delete(relation, tid, cid, crosscheck, options, + tmfd, changingPart, oldSlot); + + /* + * If the tuple has been concurrently updated, then get the lock on it. + * (Do only if caller asked for this by setting the + * TABLE_MODIFY_LOCK_UPDATED option) With the lock held retry of the + * delete should succeed even if there are more concurrent update + * attempts. + */ + if (result == TM_Updated && (options & TABLE_MODIFY_LOCK_UPDATED)) + { + /* + * heapam_tuple_lock() will take advantage of tuple loaded into + * oldSlot by heap_delete(). + */ + result = heapam_tuple_lock(relation, tid, snapshot, + oldSlot, cid, LockTupleExclusive, + (options & TABLE_MODIFY_WAIT) ? + LockWaitBlock : + LockWaitSkip, + TUPLE_LOCK_FLAG_FIND_LAST_VERSION, + tmfd); + + if (result == TM_Ok) + return TM_Updated; + } + + return result; } static TM_Result heapam_tuple_update(Relation relation, ItemPointer otid, TupleTableSlot *slot, CommandId cid, Snapshot snapshot, Snapshot crosscheck, - bool wait, TM_FailureData *tmfd, - LockTupleMode *lockmode, TU_UpdateIndexes *update_indexes) + int options, TM_FailureData *tmfd, + LockTupleMode *lockmode, TU_UpdateIndexes *update_indexes, + TupleTableSlot *oldSlot) { bool shouldFree = true; HeapTuple tuple = ExecFetchSlotHeapTuple(slot, true, &shouldFree); @@ -326,8 +364,8 @@ heapam_tuple_update(Relation relation, ItemPointer otid, TupleTableSlot *slot, slot->tts_tableOid = RelationGetRelid(relation); tuple->t_tableOid = slot->tts_tableOid; - result = heap_update(relation, otid, tuple, cid, crosscheck, wait, - tmfd, lockmode, update_indexes); + result = heap_update(relation, otid, tuple, cid, crosscheck, options, + tmfd, lockmode, update_indexes, oldSlot); ItemPointerCopy(&tuple->t_self, &slot->tts_tid); /* @@ -354,6 +392,31 @@ heapam_tuple_update(Relation relation, ItemPointer otid, TupleTableSlot *slot, if (shouldFree) pfree(tuple); + /* + * If the tuple has been concurrently updated, then get the lock on it. + * (Do only if caller asked for this by setting the + * TABLE_MODIFY_LOCK_UPDATED option) With the lock held retry of the + * update should succeed even if there are more concurrent update + * attempts. + */ + if (result == TM_Updated && (options & TABLE_MODIFY_LOCK_UPDATED)) + { + /* + * heapam_tuple_lock() will take advantage of tuple loaded into + * oldSlot by heap_update(). + */ + result = heapam_tuple_lock(relation, otid, snapshot, + oldSlot, cid, *lockmode, + (options & TABLE_MODIFY_WAIT) ? + LockWaitBlock : + LockWaitSkip, + TUPLE_LOCK_FLAG_FIND_LAST_VERSION, + tmfd); + + if (result == TM_Ok) + return TM_Updated; + } + return result; } @@ -365,7 +428,6 @@ heapam_tuple_lock(Relation relation, ItemPointer tid, Snapshot snapshot, { BufferHeapTupleTableSlot *bslot = (BufferHeapTupleTableSlot *) slot; TM_Result result; - Buffer buffer; HeapTuple tuple = &bslot->base.tupdata; bool follow_updates; @@ -375,9 +437,8 @@ heapam_tuple_lock(Relation relation, ItemPointer tid, Snapshot snapshot, Assert(TTS_IS_BUFFERTUPLE(slot)); tuple_lock_retry: - tuple->t_self = *tid; - result = heap_lock_tuple(relation, tuple, cid, mode, wait_policy, - follow_updates, &buffer, tmfd); + result = heap_lock_tuple(relation, tid, slot, cid, mode, wait_policy, + follow_updates, tmfd); if (result == TM_Updated && (flags & TUPLE_LOCK_FLAG_FIND_LAST_VERSION)) @@ -385,8 +446,6 @@ tuple_lock_retry: /* Should not encounter speculative tuple on recheck */ Assert(!HeapTupleHeaderIsSpeculative(tuple->t_data)); - ReleaseBuffer(buffer); - if (!ItemPointerEquals(&tmfd->ctid, &tuple->t_self)) { SnapshotData SnapshotDirty; @@ -408,6 +467,8 @@ tuple_lock_retry: InitDirtySnapshot(SnapshotDirty); for (;;) { + Buffer buffer = InvalidBuffer; + if (ItemPointerIndicatesMovedPartitions(tid)) ereport(ERROR, (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE), @@ -502,7 +563,7 @@ tuple_lock_retry: /* * This is a live tuple, so try to lock it again. */ - ReleaseBuffer(buffer); + ExecStorePinnedBufferHeapTuple(tuple, slot, buffer); goto tuple_lock_retry; } @@ -513,7 +574,7 @@ tuple_lock_retry: */ if (tuple->t_data == NULL) { - Assert(!BufferIsValid(buffer)); + ReleaseBuffer(buffer); return TM_Deleted; } @@ -566,9 +627,6 @@ tuple_lock_retry: slot->tts_tableOid = RelationGetRelid(relation); tuple->t_tableOid = slot->tts_tableOid; - /* store in slot, transferring existing pin */ - ExecStorePinnedBufferHeapTuple(tuple, slot, buffer); - return result; } diff --git a/src/backend/access/table/tableam.c b/src/backend/access/table/tableam.c index e57a0b7ea3..8d3675be95 100644 --- a/src/backend/access/table/tableam.c +++ b/src/backend/access/table/tableam.c @@ -287,16 +287,23 @@ simple_table_tuple_insert(Relation rel, TupleTableSlot *slot) * via ereport(). */ void -simple_table_tuple_delete(Relation rel, ItemPointer tid, Snapshot snapshot) +simple_table_tuple_delete(Relation rel, ItemPointer tid, Snapshot snapshot, + TupleTableSlot *oldSlot) { TM_Result result; TM_FailureData tmfd; + int options = TABLE_MODIFY_WAIT; /* wait for commit */ + + /* Fetch old tuple if the relevant slot is provided */ + if (oldSlot) + options |= TABLE_MODIFY_FETCH_OLD_TUPLE; result = table_tuple_delete(rel, tid, GetCurrentCommandId(true), snapshot, InvalidSnapshot, - true /* wait for commit */ , - &tmfd, false /* changingPart */ ); + options, + &tmfd, false /* changingPart */ , + oldSlot); switch (result) { @@ -335,17 +342,24 @@ void simple_table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot, Snapshot snapshot, - TU_UpdateIndexes *update_indexes) + TU_UpdateIndexes *update_indexes, + TupleTableSlot *oldSlot) { TM_Result result; TM_FailureData tmfd; LockTupleMode lockmode; + int options = TABLE_MODIFY_WAIT; /* wait for commit */ + + /* Fetch old tuple if the relevant slot is provided */ + if (oldSlot) + options |= TABLE_MODIFY_FETCH_OLD_TUPLE; result = table_tuple_update(rel, otid, slot, GetCurrentCommandId(true), snapshot, InvalidSnapshot, - true /* wait for commit */ , - &tmfd, &lockmode, update_indexes); + options, + &tmfd, &lockmode, update_indexes, + oldSlot); switch (result) { diff --git a/src/backend/commands/trigger.c b/src/backend/commands/trigger.c index 35eb7180f7..3309b4ebd2 100644 --- a/src/backend/commands/trigger.c +++ b/src/backend/commands/trigger.c @@ -2773,8 +2773,8 @@ ExecBRDeleteTriggers(EState *estate, EPQState *epqstate, void ExecARDeleteTriggers(EState *estate, ResultRelInfo *relinfo, - ItemPointer tupleid, HeapTuple fdw_trigtuple, + TupleTableSlot *slot, TransitionCaptureState *transition_capture, bool is_crosspart_update) { @@ -2783,20 +2783,11 @@ ExecARDeleteTriggers(EState *estate, if ((trigdesc && trigdesc->trig_delete_after_row) || (transition_capture && transition_capture->tcs_delete_old_table)) { - TupleTableSlot *slot = ExecGetTriggerOldSlot(estate, relinfo); - - Assert(HeapTupleIsValid(fdw_trigtuple) ^ ItemPointerIsValid(tupleid)); - if (fdw_trigtuple == NULL) - GetTupleForTrigger(estate, - NULL, - relinfo, - tupleid, - LockTupleExclusive, - slot, - NULL, - NULL, - NULL); - else + /* + * Put the FDW old tuple to the slot. Otherwise, caller is expected + * to have old tuple alredy fetched to the slot. + */ + if (fdw_trigtuple != NULL) ExecForceStoreHeapTuple(fdw_trigtuple, slot, false); AfterTriggerSaveEvent(estate, relinfo, NULL, NULL, @@ -3087,18 +3078,17 @@ ExecBRUpdateTriggers(EState *estate, EPQState *epqstate, * Note: 'src_partinfo' and 'dst_partinfo', when non-NULL, refer to the source * and destination partitions, respectively, of a cross-partition update of * the root partitioned table mentioned in the query, given by 'relinfo'. - * 'tupleid' in that case refers to the ctid of the "old" tuple in the source - * partition, and 'newslot' contains the "new" tuple in the destination - * partition. This interface allows to support the requirements of - * ExecCrossPartitionUpdateForeignKey(); is_crosspart_update must be true in - * that case. + * 'oldslot' contains the "old" tuple in the source partition, and 'newslot' + * contains the "new" tuple in the destination partition. This interface + * allows to support the requirements of ExecCrossPartitionUpdateForeignKey(); + * is_crosspart_update must be true in that case. */ void ExecARUpdateTriggers(EState *estate, ResultRelInfo *relinfo, ResultRelInfo *src_partinfo, ResultRelInfo *dst_partinfo, - ItemPointer tupleid, HeapTuple fdw_trigtuple, + TupleTableSlot *oldslot, TupleTableSlot *newslot, List *recheckIndexes, TransitionCaptureState *transition_capture, @@ -3117,29 +3107,14 @@ ExecARUpdateTriggers(EState *estate, ResultRelInfo *relinfo, * separately for DELETE and INSERT to capture transition table rows. * In such case, either old tuple or new tuple can be NULL. */ - TupleTableSlot *oldslot; - ResultRelInfo *tupsrc; - Assert((src_partinfo != NULL && dst_partinfo != NULL) || !is_crosspart_update); - tupsrc = src_partinfo ? src_partinfo : relinfo; - oldslot = ExecGetTriggerOldSlot(estate, tupsrc); - - if (fdw_trigtuple == NULL && ItemPointerIsValid(tupleid)) - GetTupleForTrigger(estate, - NULL, - tupsrc, - tupleid, - LockTupleExclusive, - oldslot, - NULL, - NULL, - NULL); - else if (fdw_trigtuple != NULL) + if (fdw_trigtuple != NULL) + { + Assert(oldslot); ExecForceStoreHeapTuple(fdw_trigtuple, oldslot, false); - else - ExecClearTuple(oldslot); + } AfterTriggerSaveEvent(estate, relinfo, src_partinfo, dst_partinfo, diff --git a/src/backend/executor/execReplication.c b/src/backend/executor/execReplication.c index d0a89cd577..0cad843fb6 100644 --- a/src/backend/executor/execReplication.c +++ b/src/backend/executor/execReplication.c @@ -577,6 +577,7 @@ ExecSimpleRelationUpdate(ResultRelInfo *resultRelInfo, { List *recheckIndexes = NIL; TU_UpdateIndexes update_indexes; + TupleTableSlot *oldSlot = NULL; /* Compute stored generated columns */ if (rel->rd_att->constr && @@ -590,8 +591,12 @@ ExecSimpleRelationUpdate(ResultRelInfo *resultRelInfo, if (rel->rd_rel->relispartition) ExecPartitionCheck(resultRelInfo, slot, estate, true); + if (resultRelInfo->ri_TrigDesc && + resultRelInfo->ri_TrigDesc->trig_update_after_row) + oldSlot = ExecGetTriggerOldSlot(estate, resultRelInfo); + simple_table_tuple_update(rel, tid, slot, estate->es_snapshot, - &update_indexes); + &update_indexes, oldSlot); if (resultRelInfo->ri_NumIndices > 0 && (update_indexes != TU_None)) recheckIndexes = ExecInsertIndexTuples(resultRelInfo, @@ -602,7 +607,7 @@ ExecSimpleRelationUpdate(ResultRelInfo *resultRelInfo, /* AFTER ROW UPDATE Triggers */ ExecARUpdateTriggers(estate, resultRelInfo, NULL, NULL, - tid, NULL, slot, + NULL, oldSlot, slot, recheckIndexes, NULL, false); list_free(recheckIndexes); @@ -636,12 +641,18 @@ ExecSimpleRelationDelete(ResultRelInfo *resultRelInfo, if (!skip_tuple) { + TupleTableSlot *oldSlot = NULL; + + if (resultRelInfo->ri_TrigDesc && + resultRelInfo->ri_TrigDesc->trig_delete_after_row) + oldSlot = ExecGetTriggerOldSlot(estate, resultRelInfo); + /* OK, delete the tuple */ - simple_table_tuple_delete(rel, tid, estate->es_snapshot); + simple_table_tuple_delete(rel, tid, estate->es_snapshot, oldSlot); /* AFTER ROW DELETE Triggers */ ExecARDeleteTriggers(estate, resultRelInfo, - tid, NULL, NULL, false); + NULL, oldSlot, NULL, false); } } diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c index 5568dd7b95..d1917f2fea 100644 --- a/src/backend/executor/nodeModifyTable.c +++ b/src/backend/executor/nodeModifyTable.c @@ -566,6 +566,15 @@ ExecInitInsertProjection(ModifyTableState *mtstate, table_slot_create(resultRelInfo->ri_RelationDesc, &estate->es_tupleTable); + /* + * In the ON CONFLICT UPDATE case, we will also need a slot for the old + * tuple to calculate the updated tuple on its base. + */ + if (node->onConflictAction == ONCONFLICT_UPDATE) + resultRelInfo->ri_oldTupleSlot = + table_slot_create(resultRelInfo->ri_RelationDesc, + &estate->es_tupleTable); + /* Build ProjectionInfo if needed (it probably isn't). */ if (need_projection) { @@ -1154,7 +1163,7 @@ ExecInsert(ModifyTableContext *context, ExecARUpdateTriggers(estate, resultRelInfo, NULL, NULL, NULL, - NULL, + resultRelInfo->ri_oldTupleSlot, slot, NULL, mtstate->mt_transition_capture, @@ -1334,7 +1343,8 @@ ExecDeletePrologue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, */ static TM_Result ExecDeleteAct(ModifyTableContext *context, ResultRelInfo *resultRelInfo, - ItemPointer tupleid, bool changingPart) + ItemPointer tupleid, bool changingPart, int options, + TupleTableSlot *oldSlot) { EState *estate = context->estate; @@ -1342,9 +1352,10 @@ ExecDeleteAct(ModifyTableContext *context, ResultRelInfo *resultRelInfo, estate->es_output_cid, estate->es_snapshot, estate->es_crosscheck_snapshot, - true /* wait for commit */ , + options, &context->tmfd, - changingPart); + changingPart, + oldSlot); } /* @@ -1353,10 +1364,15 @@ ExecDeleteAct(ModifyTableContext *context, ResultRelInfo *resultRelInfo, * Closing steps of tuple deletion; this invokes AFTER FOR EACH ROW triggers, * including the UPDATE triggers if the deletion is being done as part of a * cross-partition tuple move. + * + * The old tuple is already fetched into ‘slot’ for regular tables. For FDW, + * the old tuple is given as 'oldtuple' and is to be stored in 'slot' when + * needed. */ static void ExecDeleteEpilogue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, - ItemPointer tupleid, HeapTuple oldtuple, bool changingPart) + ItemPointer tupleid, HeapTuple oldtuple, + TupleTableSlot *slot, bool changingPart) { ModifyTableState *mtstate = context->mtstate; EState *estate = context->estate; @@ -1374,8 +1390,8 @@ ExecDeleteEpilogue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, { ExecARUpdateTriggers(estate, resultRelInfo, NULL, NULL, - tupleid, oldtuple, - NULL, NULL, mtstate->mt_transition_capture, + oldtuple, + slot, NULL, NULL, mtstate->mt_transition_capture, false); /* @@ -1386,10 +1402,30 @@ ExecDeleteEpilogue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, } /* AFTER ROW DELETE Triggers */ - ExecARDeleteTriggers(estate, resultRelInfo, tupleid, oldtuple, + ExecARDeleteTriggers(estate, resultRelInfo, oldtuple, slot, ar_delete_trig_tcs, changingPart); } +/* + * Initializes the tuple slot in a ResultRelInfo for DELETE action. + * + * We mark 'projectNewInfoValid' even though the projections themselves + * are not initialized here. + */ +static void +ExecInitDeleteTupleSlot(ModifyTableState *mtstate, + ResultRelInfo *resultRelInfo) +{ + EState *estate = mtstate->ps.state; + + Assert(!resultRelInfo->ri_projectNewInfoValid); + + resultRelInfo->ri_oldTupleSlot = + table_slot_create(resultRelInfo->ri_RelationDesc, + &estate->es_tupleTable); + resultRelInfo->ri_projectNewInfoValid = true; +} + /* ---------------------------------------------------------------- * ExecDelete * @@ -1409,7 +1445,8 @@ ExecDeleteEpilogue(ModifyTableContext *context, ResultRelInfo *resultRelInfo, * part of an UPDATE of partition-key, then the slot returned by * EvalPlanQual() is passed back using output parameter epqreturnslot. * - * Returns RETURNING result if any, otherwise NULL. + * Returns RETURNING result if any, otherwise NULL. The deleted tuple + * to be stored into oldslot independently that. * ---------------------------------------------------------------- */ static TupleTableSlot * @@ -1417,6 +1454,7 @@ ExecDelete(ModifyTableContext *context, ResultRelInfo *resultRelInfo, ItemPointer tupleid, HeapTuple oldtuple, + TupleTableSlot *oldslot, bool processReturning, bool changingPart, bool canSetTag, @@ -1480,6 +1518,15 @@ ExecDelete(ModifyTableContext *context, } else { + int options = TABLE_MODIFY_WAIT | TABLE_MODIFY_FETCH_OLD_TUPLE; + + /* + * Specify that we need to lock and fetch the last tuple version for + * EPQ on appropriate transaction isolation levels. + */ + if (!IsolationUsesXactSnapshot()) + options |= TABLE_MODIFY_LOCK_UPDATED; + /* * delete the tuple * @@ -1490,7 +1537,8 @@ ExecDelete(ModifyTableContext *context, * transaction-snapshot mode transactions. */ ldelete: - result = ExecDeleteAct(context, resultRelInfo, tupleid, changingPart); + result = ExecDeleteAct(context, resultRelInfo, tupleid, changingPart, + options, oldslot); if (tmresult) *tmresult = result; @@ -1537,7 +1585,6 @@ ldelete: case TM_Updated: { - TupleTableSlot *inputslot; TupleTableSlot *epqslot; if (IsolationUsesXactSnapshot()) @@ -1546,87 +1593,29 @@ ldelete: errmsg("could not serialize access due to concurrent update"))); /* - * Already know that we're going to need to do EPQ, so - * fetch tuple directly into the right slot. + * We need to do EPQ. The latest tuple is already found + * and locked as a result of TABLE_MODIFY_LOCK_UPDATED. */ - EvalPlanQualBegin(context->epqstate); - inputslot = EvalPlanQualSlot(context->epqstate, resultRelationDesc, - resultRelInfo->ri_RangeTableIndex); + Assert(context->tmfd.traversed); + epqslot = EvalPlanQual(context->epqstate, + resultRelationDesc, + resultRelInfo->ri_RangeTableIndex, + oldslot); + if (TupIsNull(epqslot)) + /* Tuple not passing quals anymore, exiting... */ + return NULL; - result = table_tuple_lock(resultRelationDesc, tupleid, - estate->es_snapshot, - inputslot, estate->es_output_cid, - LockTupleExclusive, LockWaitBlock, - TUPLE_LOCK_FLAG_FIND_LAST_VERSION, - &context->tmfd); - - switch (result) + /* + * If requested, skip delete and pass back the updated + * row. + */ + if (epqreturnslot) { - case TM_Ok: - Assert(context->tmfd.traversed); - epqslot = EvalPlanQual(context->epqstate, - resultRelationDesc, - resultRelInfo->ri_RangeTableIndex, - inputslot); - if (TupIsNull(epqslot)) - /* Tuple not passing quals anymore, exiting... */ - return NULL; - - /* - * If requested, skip delete and pass back the - * updated row. - */ - if (epqreturnslot) - { - *epqreturnslot = epqslot; - return NULL; - } - else - goto ldelete; - - case TM_SelfModified: - - /* - * This can be reached when following an update - * chain from a tuple updated by another session, - * reaching a tuple that was already updated in - * this transaction. If previously updated by this - * command, ignore the delete, otherwise error - * out. - * - * See also TM_SelfModified response to - * table_tuple_delete() above. - */ - if (context->tmfd.cmax != estate->es_output_cid) - ereport(ERROR, - (errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION), - errmsg("tuple to be deleted was already modified by an operation triggered by the current command"), - errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows."))); - return NULL; - - case TM_Deleted: - /* tuple already deleted; nothing to do */ - return NULL; - - default: - - /* - * TM_Invisible should be impossible because we're - * waiting for updated row versions, and would - * already have errored out if the first version - * is invisible. - * - * TM_Updated should be impossible, because we're - * locking the latest version via - * TUPLE_LOCK_FLAG_FIND_LAST_VERSION. - */ - elog(ERROR, "unexpected table_tuple_lock status: %u", - result); - return NULL; + *epqreturnslot = epqslot; + return NULL; } - - Assert(false); - break; + else + goto ldelete; } case TM_Deleted: @@ -1660,7 +1649,8 @@ ldelete: if (tupleDeleted) *tupleDeleted = true; - ExecDeleteEpilogue(context, resultRelInfo, tupleid, oldtuple, changingPart); + ExecDeleteEpilogue(context, resultRelInfo, tupleid, oldtuple, + oldslot, changingPart); /* Process RETURNING if present and if requested */ if (processReturning && resultRelInfo->ri_projectReturning) @@ -1678,17 +1668,13 @@ ldelete: } else { + /* Copy old tuple to the returning slot */ slot = ExecGetReturningSlot(estate, resultRelInfo); if (oldtuple != NULL) - { ExecForceStoreHeapTuple(oldtuple, slot, false); - } else - { - if (!table_tuple_fetch_row_version(resultRelationDesc, tupleid, - SnapshotAny, slot)) - elog(ERROR, "failed to fetch deleted tuple for DELETE RETURNING"); - } + ExecCopySlot(slot, oldslot); + Assert(!TupIsNull(slot)); } rslot = ExecProcessReturning(resultRelInfo, slot, context->planSlot); @@ -1788,12 +1774,19 @@ ExecCrossPartitionUpdate(ModifyTableContext *context, MemoryContextSwitchTo(oldcxt); } + /* + * Make sure ri_oldTupleSlot is initialized. The old tuple is to be saved + * there by ExecDelete() to save effort on further re-fetching. + */ + if (unlikely(!resultRelInfo->ri_projectNewInfoValid)) + ExecInitUpdateProjection(mtstate, resultRelInfo); + /* * Row movement, part 1. Delete the tuple, but skip RETURNING processing. * We want to return rows from INSERT. */ ExecDelete(context, resultRelInfo, - tupleid, oldtuple, + tupleid, oldtuple, resultRelInfo->ri_oldTupleSlot, false, /* processReturning */ true, /* changingPart */ false, /* canSetTag */ @@ -1834,21 +1827,13 @@ ExecCrossPartitionUpdate(ModifyTableContext *context, return true; else { - /* Fetch the most recent version of old tuple. */ - TupleTableSlot *oldSlot; - - /* ... but first, make sure ri_oldTupleSlot is initialized. */ - if (unlikely(!resultRelInfo->ri_projectNewInfoValid)) - ExecInitUpdateProjection(mtstate, resultRelInfo); - oldSlot = resultRelInfo->ri_oldTupleSlot; - if (!table_tuple_fetch_row_version(resultRelInfo->ri_RelationDesc, - tupleid, - SnapshotAny, - oldSlot)) - elog(ERROR, "failed to fetch tuple being updated"); - /* and project the new tuple to retry the UPDATE with */ + /* + * ExecDelete already fetches the most recent version of old tuple + * to resultRelInfo->ri_oldTupleSlot. So, just project the new + * tuple to retry the UPDATE with. + */ *retry_slot = ExecGetUpdateNewTuple(resultRelInfo, epqslot, - oldSlot); + resultRelInfo->ri_oldTupleSlot); return false; } } @@ -1967,7 +1952,8 @@ ExecUpdatePrepareSlot(ResultRelInfo *resultRelInfo, static TM_Result ExecUpdateAct(ModifyTableContext *context, ResultRelInfo *resultRelInfo, ItemPointer tupleid, HeapTuple oldtuple, TupleTableSlot *slot, - bool canSetTag, UpdateContext *updateCxt) + bool canSetTag, int options, TupleTableSlot *oldSlot, + UpdateContext *updateCxt) { EState *estate = context->estate; Relation resultRelationDesc = resultRelInfo->ri_RelationDesc; @@ -2059,7 +2045,8 @@ lreplace: ExecCrossPartitionUpdateForeignKey(context, resultRelInfo, insert_destrel, - tupleid, slot, + tupleid, + resultRelInfo->ri_oldTupleSlot, inserted_tuple); return TM_Ok; @@ -2102,9 +2089,10 @@ lreplace: estate->es_output_cid, estate->es_snapshot, estate->es_crosscheck_snapshot, - true /* wait for commit */ , + options /* wait for commit */ , &context->tmfd, &updateCxt->lockmode, - &updateCxt->updateIndexes); + &updateCxt->updateIndexes, + oldSlot); return result; } @@ -2118,7 +2106,8 @@ lreplace: static void ExecUpdateEpilogue(ModifyTableContext *context, UpdateContext *updateCxt, ResultRelInfo *resultRelInfo, ItemPointer tupleid, - HeapTuple oldtuple, TupleTableSlot *slot) + HeapTuple oldtuple, TupleTableSlot *slot, + TupleTableSlot *oldslot) { ModifyTableState *mtstate = context->mtstate; List *recheckIndexes = NIL; @@ -2134,7 +2123,7 @@ ExecUpdateEpilogue(ModifyTableContext *context, UpdateContext *updateCxt, /* AFTER ROW UPDATE Triggers */ ExecARUpdateTriggers(context->estate, resultRelInfo, NULL, NULL, - tupleid, oldtuple, slot, + oldtuple, oldslot, slot, recheckIndexes, mtstate->operation == CMD_INSERT ? mtstate->mt_oc_transition_capture : @@ -2223,7 +2212,7 @@ ExecCrossPartitionUpdateForeignKey(ModifyTableContext *context, /* Perform the root table's triggers. */ ExecARUpdateTriggers(context->estate, rootRelInfo, sourcePartInfo, destPartInfo, - tupleid, NULL, newslot, NIL, NULL, true); + NULL, oldslot, newslot, NIL, NULL, true); } /* ---------------------------------------------------------------- @@ -2246,6 +2235,7 @@ ExecCrossPartitionUpdateForeignKey(ModifyTableContext *context, * no relevant triggers. * * slot contains the new tuple value to be stored. + * oldslot is the slot to store the old tuple. * planSlot is the output of the ModifyTable's subplan; we use it * to access values from other input tables (for RETURNING), * row-ID junk columns, etc. @@ -2256,7 +2246,7 @@ ExecCrossPartitionUpdateForeignKey(ModifyTableContext *context, static TupleTableSlot * ExecUpdate(ModifyTableContext *context, ResultRelInfo *resultRelInfo, ItemPointer tupleid, HeapTuple oldtuple, TupleTableSlot *slot, - bool canSetTag) + TupleTableSlot *oldslot, bool canSetTag, bool locked) { EState *estate = context->estate; Relation resultRelationDesc = resultRelInfo->ri_RelationDesc; @@ -2309,6 +2299,16 @@ ExecUpdate(ModifyTableContext *context, ResultRelInfo *resultRelInfo, } else { + int options = TABLE_MODIFY_WAIT | TABLE_MODIFY_FETCH_OLD_TUPLE; + + /* + * Specify that we need to lock and fetch the last tuple version for + * EPQ on appropriate transaction isolation levels if the tuple isn't + * locked already. + */ + if (!locked && !IsolationUsesXactSnapshot()) + options |= TABLE_MODIFY_LOCK_UPDATED; + /* * If we generate a new candidate tuple after EvalPlanQual testing, we * must loop back here to try again. (We don't need to redo triggers, @@ -2318,7 +2318,7 @@ ExecUpdate(ModifyTableContext *context, ResultRelInfo *resultRelInfo, */ redo_act: result = ExecUpdateAct(context, resultRelInfo, tupleid, oldtuple, slot, - canSetTag, &updateCxt); + canSetTag, options, oldslot, &updateCxt); /* * If ExecUpdateAct reports that a cross-partition update was done, @@ -2369,88 +2369,32 @@ redo_act: case TM_Updated: { - TupleTableSlot *inputslot; TupleTableSlot *epqslot; - TupleTableSlot *oldSlot; if (IsolationUsesXactSnapshot()) ereport(ERROR, (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE), errmsg("could not serialize access due to concurrent update"))); + /* Shouldn't get there if the tuple was previously locked */ + Assert(!locked); + /* - * Already know that we're going to need to do EPQ, so - * fetch tuple directly into the right slot. + * We need to do EPQ. The latest tuple is already found + * and locked as a result of TABLE_MODIFY_LOCK_UPDATED. */ - inputslot = EvalPlanQualSlot(context->epqstate, resultRelationDesc, - resultRelInfo->ri_RangeTableIndex); - - result = table_tuple_lock(resultRelationDesc, tupleid, - estate->es_snapshot, - inputslot, estate->es_output_cid, - updateCxt.lockmode, LockWaitBlock, - TUPLE_LOCK_FLAG_FIND_LAST_VERSION, - &context->tmfd); - - switch (result) - { - case TM_Ok: - Assert(context->tmfd.traversed); - - epqslot = EvalPlanQual(context->epqstate, - resultRelationDesc, - resultRelInfo->ri_RangeTableIndex, - inputslot); - if (TupIsNull(epqslot)) - /* Tuple not passing quals anymore, exiting... */ - return NULL; - - /* Make sure ri_oldTupleSlot is initialized. */ - if (unlikely(!resultRelInfo->ri_projectNewInfoValid)) - ExecInitUpdateProjection(context->mtstate, - resultRelInfo); - - /* Fetch the most recent version of old tuple. */ - oldSlot = resultRelInfo->ri_oldTupleSlot; - if (!table_tuple_fetch_row_version(resultRelationDesc, - tupleid, - SnapshotAny, - oldSlot)) - elog(ERROR, "failed to fetch tuple being updated"); - slot = ExecGetUpdateNewTuple(resultRelInfo, - epqslot, oldSlot); - goto redo_act; - - case TM_Deleted: - /* tuple already deleted; nothing to do */ - return NULL; - - case TM_SelfModified: - - /* - * This can be reached when following an update - * chain from a tuple updated by another session, - * reaching a tuple that was already updated in - * this transaction. If previously modified by - * this command, ignore the redundant update, - * otherwise error out. - * - * See also TM_SelfModified response to - * table_tuple_update() above. - */ - if (context->tmfd.cmax != estate->es_output_cid) - ereport(ERROR, - (errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION), - errmsg("tuple to be updated was already modified by an operation triggered by the current command"), - errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows."))); - return NULL; - - default: - /* see table_tuple_lock call in ExecDelete() */ - elog(ERROR, "unexpected table_tuple_lock status: %u", - result); - return NULL; - } + Assert(context->tmfd.traversed); + epqslot = EvalPlanQual(context->epqstate, + resultRelationDesc, + resultRelInfo->ri_RangeTableIndex, + oldslot); + if (TupIsNull(epqslot)) + /* Tuple not passing quals anymore, exiting... */ + return NULL; + slot = ExecGetUpdateNewTuple(resultRelInfo, + epqslot, + oldslot); + goto redo_act; } break; @@ -2474,7 +2418,7 @@ redo_act: (estate->es_processed)++; ExecUpdateEpilogue(context, &updateCxt, resultRelInfo, tupleid, oldtuple, - slot); + slot, oldslot); /* Process RETURNING if present */ if (resultRelInfo->ri_projectReturning) @@ -2692,7 +2636,8 @@ ExecOnConflictUpdate(ModifyTableContext *context, *returning = ExecUpdate(context, resultRelInfo, conflictTid, NULL, resultRelInfo->ri_onConflict->oc_ProjSlot, - canSetTag); + existing, + canSetTag, true); /* * Clear out existing tuple, as there might not be another conflict among @@ -2934,6 +2879,7 @@ lmerge_matched: { result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL, newslot, canSetTag, + TABLE_MODIFY_WAIT, NULL, &updateCxt); /* @@ -2956,7 +2902,8 @@ lmerge_matched: if (result == TM_Ok) { ExecUpdateEpilogue(context, &updateCxt, resultRelInfo, - tupleid, NULL, newslot); + tupleid, NULL, newslot, + resultRelInfo->ri_oldTupleSlot); mtstate->mt_merge_updated += 1; } break; @@ -2987,12 +2934,12 @@ lmerge_matched: } else result = ExecDeleteAct(context, resultRelInfo, tupleid, - false); + false, TABLE_MODIFY_WAIT, NULL); if (result == TM_Ok) { ExecDeleteEpilogue(context, resultRelInfo, tupleid, NULL, - false); + resultRelInfo->ri_oldTupleSlot, false); mtstate->mt_merge_deleted += 1; } break; @@ -4006,12 +3953,18 @@ ExecModifyTable(PlanState *pstate) /* Now apply the update. */ slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple, - slot, node->canSetTag); + slot, resultRelInfo->ri_oldTupleSlot, + node->canSetTag, false); break; case CMD_DELETE: + /* Initialize slot for DELETE to fetch the old tuple */ + if (unlikely(!resultRelInfo->ri_projectNewInfoValid)) + ExecInitDeleteTupleSlot(node, resultRelInfo); + slot = ExecDelete(&context, resultRelInfo, tupleid, oldtuple, - true, false, node->canSetTag, NULL, NULL, NULL); + resultRelInfo->ri_oldTupleSlot, true, false, + node->canSetTag, NULL, NULL, NULL); break; case CMD_MERGE: diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h index 368c570a0f..f112245373 100644 --- a/src/include/access/heapam.h +++ b/src/include/access/heapam.h @@ -284,19 +284,22 @@ extern void heap_multi_insert(Relation relation, struct TupleTableSlot **slots, int ntuples, CommandId cid, int options, BulkInsertState bistate); extern TM_Result heap_delete(Relation relation, ItemPointer tid, - CommandId cid, Snapshot crosscheck, bool wait, - struct TM_FailureData *tmfd, bool changingPart); + CommandId cid, Snapshot crosscheck, int options, + struct TM_FailureData *tmfd, bool changingPart, + TupleTableSlot *oldSlot); extern void heap_finish_speculative(Relation relation, ItemPointer tid); extern void heap_abort_speculative(Relation relation, ItemPointer tid); extern TM_Result heap_update(Relation relation, ItemPointer otid, HeapTuple newtup, - CommandId cid, Snapshot crosscheck, bool wait, + CommandId cid, Snapshot crosscheck, int options, struct TM_FailureData *tmfd, LockTupleMode *lockmode, - TU_UpdateIndexes *update_indexes); -extern TM_Result heap_lock_tuple(Relation relation, HeapTuple tuple, - CommandId cid, LockTupleMode mode, LockWaitPolicy wait_policy, - bool follow_updates, - Buffer *buffer, struct TM_FailureData *tmfd); + TU_UpdateIndexes *update_indexes, + TupleTableSlot *oldSlot); +extern TM_Result heap_lock_tuple(Relation relation, ItemPointer tid, + TupleTableSlot *slot, + CommandId cid, LockTupleMode mode, + LockWaitPolicy wait_policy, bool follow_updates, + struct TM_FailureData *tmfd); extern void heap_inplace_update(Relation relation, HeapTuple tuple); extern bool heap_prepare_freeze_tuple(HeapTupleHeader tuple, diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h index 65834caeb1..fc0e702715 100644 --- a/src/include/access/tableam.h +++ b/src/include/access/tableam.h @@ -259,6 +259,15 @@ typedef struct TM_IndexDeleteOp /* Follow update chain and lock latest version of tuple */ #define TUPLE_LOCK_FLAG_FIND_LAST_VERSION (1 << 1) +/* + * "options" flag bits for table_tuple_update and table_tuple_delete, + * Wait for any conflicting update to commit/abort */ +#define TABLE_MODIFY_WAIT 0x0001 +/* Fetch the existing tuple into a dedicated slot */ +#define TABLE_MODIFY_FETCH_OLD_TUPLE 0x0002 +/* On concurrent update, follow the update chain and lock latest version of tuple */ +#define TABLE_MODIFY_LOCK_UPDATED 0x0004 + /* Typedef for callback function for table_index_build_scan */ typedef void (*IndexBuildCallback) (Relation index, @@ -528,9 +537,10 @@ typedef struct TableAmRoutine CommandId cid, Snapshot snapshot, Snapshot crosscheck, - bool wait, + int options, TM_FailureData *tmfd, - bool changingPart); + bool changingPart, + TupleTableSlot *oldSlot); /* see table_tuple_update() for reference about parameters */ TM_Result (*tuple_update) (Relation rel, @@ -539,10 +549,11 @@ typedef struct TableAmRoutine CommandId cid, Snapshot snapshot, Snapshot crosscheck, - bool wait, + int options, TM_FailureData *tmfd, LockTupleMode *lockmode, - TU_UpdateIndexes *update_indexes); + TU_UpdateIndexes *update_indexes, + TupleTableSlot *oldSlot); /* see table_tuple_lock() for reference about parameters */ TM_Result (*tuple_lock) (Relation rel, @@ -1463,7 +1474,7 @@ table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots, } /* - * Delete a tuple. + * Delete a tuple (and optionally lock the last tuple version). * * NB: do not call this directly unless prepared to deal with * concurrent-update conditions. Use simple_table_tuple_delete instead. @@ -1474,11 +1485,21 @@ table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots, * cid - delete command ID (used for visibility test, and stored into * cmax if successful) * crosscheck - if not InvalidSnapshot, also check tuple against this - * wait - true if should wait for any conflicting update to commit/abort + * options: + * If TABLE_MODIFY_WAIT, wait for any conflicting update to commit/abort. + * If TABLE_MODIFY_FETCH_OLD_TUPLE option is given, the existing tuple is + * fetched into oldSlot when the update is successful. + * If TABLE_MODIFY_LOCK_UPDATED option is given and the tuple is + * concurrently updated, then the last tuple version is locked and fetched + * into oldSlot. + * * Output parameters: * tmfd - filled in failure cases (see below) * changingPart - true iff the tuple is being moved to another partition * table due to an update of the partition key. Otherwise, false. + * oldSlot - slot to save the deleted or locked tuple. Can be NULL if none of + * TABLE_MODIFY_FETCH_OLD_TUPLE or TABLE_MODIFY_LOCK_UPDATED options + * is specified. * * Normal, successful return value is TM_Ok, which means we did actually * delete it. Failure return codes are TM_SelfModified, TM_Updated, and @@ -1490,16 +1511,18 @@ table_multi_insert(Relation rel, TupleTableSlot **slots, int nslots, */ static inline TM_Result table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid, - Snapshot snapshot, Snapshot crosscheck, bool wait, - TM_FailureData *tmfd, bool changingPart) + Snapshot snapshot, Snapshot crosscheck, int options, + TM_FailureData *tmfd, bool changingPart, + TupleTableSlot *oldSlot) { return rel->rd_tableam->tuple_delete(rel, tid, cid, snapshot, crosscheck, - wait, tmfd, changingPart); + options, tmfd, changingPart, + oldSlot); } /* - * Update a tuple. + * Update a tuple (and optionally lock the last tuple version). * * NB: do not call this directly unless you are prepared to deal with * concurrent-update conditions. Use simple_table_tuple_update instead. @@ -1511,13 +1534,23 @@ table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid, * cid - update command ID (used for visibility test, and stored into * cmax/cmin if successful) * crosscheck - if not InvalidSnapshot, also check old tuple against this - * wait - true if should wait for any conflicting update to commit/abort + * options: + * If TABLE_MODIFY_WAIT, wait for any conflicting update to commit/abort. + * If TABLE_MODIFY_FETCH_OLD_TUPLE option is given, the existing tuple is + * fetched into oldSlot when the update is successful. + * If TABLE_MODIFY_LOCK_UPDATED option is given and the tuple is + * concurrently updated, then the last tuple version is locked and fetched + * into oldSlot. + * * Output parameters: * tmfd - filled in failure cases (see below) * lockmode - filled with lock mode acquired on tuple * update_indexes - in success cases this is set to true if new index entries * are required for this tuple - * + * oldSlot - slot to save the deleted or locked tuple. Can be NULL if none of + * TABLE_MODIFY_FETCH_OLD_TUPLE or TABLE_MODIFY_LOCK_UPDATED options + * is specified. + * Normal, successful return value is TM_Ok, which means we did actually * update it. Failure return codes are TM_SelfModified, TM_Updated, and * TM_BeingModified (the last only possible if wait == false). @@ -1535,13 +1568,15 @@ table_tuple_delete(Relation rel, ItemPointer tid, CommandId cid, static inline TM_Result table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot, CommandId cid, Snapshot snapshot, Snapshot crosscheck, - bool wait, TM_FailureData *tmfd, LockTupleMode *lockmode, - TU_UpdateIndexes *update_indexes) + int options, TM_FailureData *tmfd, LockTupleMode *lockmode, + TU_UpdateIndexes *update_indexes, + TupleTableSlot *oldSlot) { return rel->rd_tableam->tuple_update(rel, otid, slot, cid, snapshot, crosscheck, - wait, tmfd, - lockmode, update_indexes); + options, tmfd, + lockmode, update_indexes, + oldSlot); } /* @@ -2083,10 +2118,12 @@ table_scan_sample_next_tuple(TableScanDesc scan, extern void simple_table_tuple_insert(Relation rel, TupleTableSlot *slot); extern void simple_table_tuple_delete(Relation rel, ItemPointer tid, - Snapshot snapshot); + Snapshot snapshot, + TupleTableSlot *oldSlot); extern void simple_table_tuple_update(Relation rel, ItemPointer otid, TupleTableSlot *slot, Snapshot snapshot, - TU_UpdateIndexes *update_indexes); + TU_UpdateIndexes *update_indexes, + TupleTableSlot *oldSlot); /* ---------------------------------------------------------------------------- diff --git a/src/include/commands/trigger.h b/src/include/commands/trigger.h index 8a5a9fe642..cb968d03ec 100644 --- a/src/include/commands/trigger.h +++ b/src/include/commands/trigger.h @@ -216,8 +216,8 @@ extern bool ExecBRDeleteTriggers(EState *estate, TM_FailureData *tmfd); extern void ExecARDeleteTriggers(EState *estate, ResultRelInfo *relinfo, - ItemPointer tupleid, HeapTuple fdw_trigtuple, + TupleTableSlot *slot, TransitionCaptureState *transition_capture, bool is_crosspart_update); extern bool ExecIRDeleteTriggers(EState *estate, @@ -240,8 +240,8 @@ extern void ExecARUpdateTriggers(EState *estate, ResultRelInfo *relinfo, ResultRelInfo *src_partinfo, ResultRelInfo *dst_partinfo, - ItemPointer tupleid, HeapTuple fdw_trigtuple, + TupleTableSlot *oldslot, TupleTableSlot *newslot, List *recheckIndexes, TransitionCaptureState *transition_capture,