From ff4e078773b48260a36587b54537ca54d929898f Mon Sep 17 00:00:00 2001 From: Heikki Linnakangas Date: Wed, 15 Jun 2011 11:43:05 +0300 Subject: [PATCH] Make non-MVCC snapshots exempt from predicate locking. Scans with non-MVCC snapshots, like in REINDEX, are basically non-transactional operations. The DDL operation itself might participate in SSI, but there's separate functions for that. Kevin Grittner and Dan Ports, with some changes by me. --- src/backend/access/heap/heapam.c | 23 ++-- src/backend/access/index/indexam.c | 15 +-- src/backend/access/nbtree/nbtsearch.c | 33 +++-- src/backend/executor/nodeSeqscan.c | 3 +- src/backend/storage/lmgr/predicate.c | 168 ++++++++++++++++---------- src/include/storage/predicate.h | 9 +- 6 files changed, 151 insertions(+), 100 deletions(-) diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c index 01a492e496..b947c11f7d 100644 --- a/src/backend/access/heap/heapam.c +++ b/src/backend/access/heap/heapam.c @@ -274,7 +274,8 @@ heapgetpage(HeapScanDesc scan, BlockNumber page) else valid = HeapTupleSatisfiesVisibility(&loctup, snapshot, buffer); - CheckForSerializableConflictOut(valid, scan->rs_rd, &loctup, buffer); + CheckForSerializableConflictOut(valid, scan->rs_rd, &loctup, + buffer, snapshot); if (valid) scan->rs_vistuples[ntup++] = lineoff; @@ -469,7 +470,8 @@ heapgettup(HeapScanDesc scan, snapshot, scan->rs_cbuf); - CheckForSerializableConflictOut(valid, scan->rs_rd, tuple, scan->rs_cbuf); + CheckForSerializableConflictOut(valid, scan->rs_rd, tuple, + scan->rs_cbuf, snapshot); if (valid && key != NULL) HeapKeyTest(tuple, RelationGetDescr(scan->rs_rd), @@ -478,7 +480,7 @@ heapgettup(HeapScanDesc scan, if (valid) { if (!scan->rs_relpredicatelocked) - PredicateLockTuple(scan->rs_rd, tuple); + PredicateLockTuple(scan->rs_rd, tuple, snapshot); LockBuffer(scan->rs_cbuf, BUFFER_LOCK_UNLOCK); return; } @@ -747,7 +749,7 @@ heapgettup_pagemode(HeapScanDesc scan, if (valid) { if (!scan->rs_relpredicatelocked) - PredicateLockTuple(scan->rs_rd, tuple); + PredicateLockTuple(scan->rs_rd, tuple, scan->rs_snapshot); scan->rs_cindex = lineindex; return; } @@ -755,7 +757,7 @@ heapgettup_pagemode(HeapScanDesc scan, else { if (!scan->rs_relpredicatelocked) - PredicateLockTuple(scan->rs_rd, tuple); + PredicateLockTuple(scan->rs_rd, tuple, scan->rs_snapshot); scan->rs_cindex = lineindex; return; } @@ -1470,9 +1472,9 @@ heap_fetch(Relation relation, valid = HeapTupleSatisfiesVisibility(tuple, snapshot, buffer); if (valid) - PredicateLockTuple(relation, tuple); + PredicateLockTuple(relation, tuple, snapshot); - CheckForSerializableConflictOut(valid, relation, tuple, buffer); + CheckForSerializableConflictOut(valid, relation, tuple, buffer, snapshot); LockBuffer(buffer, BUFFER_LOCK_UNLOCK); @@ -1588,11 +1590,12 @@ heap_hot_search_buffer(ItemPointer tid, Relation relation, Buffer buffer, /* If it's visible per the snapshot, we must return it */ valid = HeapTupleSatisfiesVisibility(&heapTuple, snapshot, buffer); - CheckForSerializableConflictOut(valid, relation, &heapTuple, buffer); + CheckForSerializableConflictOut(valid, relation, &heapTuple, buffer, + snapshot); if (valid) { ItemPointerSetOffsetNumber(tid, offnum); - PredicateLockTuple(relation, &heapTuple); + PredicateLockTuple(relation, &heapTuple, snapshot); if (all_dead) *all_dead = false; return true; @@ -1750,7 +1753,7 @@ heap_get_latest_tid(Relation relation, * result candidate. */ valid = HeapTupleSatisfiesVisibility(&tp, snapshot, buffer); - CheckForSerializableConflictOut(valid, relation, &tp, buffer); + CheckForSerializableConflictOut(valid, relation, &tp, buffer, snapshot); if (valid) *tid = ctid; diff --git a/src/backend/access/index/indexam.c b/src/backend/access/index/indexam.c index 0208765964..31d705c056 100644 --- a/src/backend/access/index/indexam.c +++ b/src/backend/access/index/indexam.c @@ -126,7 +126,7 @@ do { \ } while(0) static IndexScanDesc index_beginscan_internal(Relation indexRelation, - int nkeys, int norderbys); + int nkeys, int norderbys, Snapshot snapshot); /* ---------------------------------------------------------------- @@ -234,7 +234,7 @@ index_beginscan(Relation heapRelation, { IndexScanDesc scan; - scan = index_beginscan_internal(indexRelation, nkeys, norderbys); + scan = index_beginscan_internal(indexRelation, nkeys, norderbys, snapshot); /* * Save additional parameters into the scandesc. Everything else was set @@ -259,7 +259,7 @@ index_beginscan_bitmap(Relation indexRelation, { IndexScanDesc scan; - scan = index_beginscan_internal(indexRelation, nkeys, 0); + scan = index_beginscan_internal(indexRelation, nkeys, 0, snapshot); /* * Save additional parameters into the scandesc. Everything else was set @@ -275,7 +275,7 @@ index_beginscan_bitmap(Relation indexRelation, */ static IndexScanDesc index_beginscan_internal(Relation indexRelation, - int nkeys, int norderbys) + int nkeys, int norderbys, Snapshot snapshot) { IndexScanDesc scan; FmgrInfo *procedure; @@ -284,7 +284,7 @@ index_beginscan_internal(Relation indexRelation, GET_REL_PROCEDURE(ambeginscan); if (!(indexRelation->rd_am->ampredlocks)) - PredicateLockRelation(indexRelation); + PredicateLockRelation(indexRelation, snapshot); /* * We hold a reference count to the relcache entry throughout the scan. @@ -602,7 +602,8 @@ index_getnext(IndexScanDesc scan, ScanDirection direction) scan->xs_cbuf); CheckForSerializableConflictOut(valid, scan->heapRelation, - heapTuple, scan->xs_cbuf); + heapTuple, scan->xs_cbuf, + scan->xs_snapshot); if (valid) { @@ -624,7 +625,7 @@ index_getnext(IndexScanDesc scan, ScanDirection direction) else scan->xs_next_hot = InvalidOffsetNumber; - PredicateLockTuple(scan->heapRelation, heapTuple); + PredicateLockTuple(scan->heapRelation, heapTuple, scan->xs_snapshot); LockBuffer(scan->xs_cbuf, BUFFER_LOCK_UNLOCK); diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c index 2ce2bc2f00..0771d19463 100644 --- a/src/backend/access/nbtree/nbtsearch.c +++ b/src/backend/access/nbtree/nbtsearch.c @@ -64,10 +64,7 @@ _bt_search(Relation rel, int keysz, ScanKey scankey, bool nextkey, /* If index is empty and access = BT_READ, no root page is created. */ if (!BufferIsValid(*bufP)) - { - PredicateLockRelation(rel); /* Nothing finer to lock exists. */ return (BTStack) NULL; - } /* Loop iterates once per level descended in the tree */ for (;;) @@ -92,11 +89,7 @@ _bt_search(Relation rel, int keysz, ScanKey scankey, bool nextkey, page = BufferGetPage(*bufP); opaque = (BTPageOpaque) PageGetSpecialPointer(page); if (P_ISLEAF(opaque)) - { - if (access == BT_READ) - PredicateLockPage(rel, BufferGetBlockNumber(*bufP)); break; - } /* * Find the appropriate item on the internal page, and get the child @@ -855,9 +848,16 @@ _bt_first(IndexScanDesc scan, ScanDirection dir) if (!BufferIsValid(buf)) { - /* Only get here if index is completely empty */ + /* + * We only get here if the index is completely empty. + * Lock relation because nothing finer to lock exists. + */ + PredicateLockRelation(rel, scan->xs_snapshot); return false; } + else + PredicateLockPage(rel, BufferGetBlockNumber(buf), + scan->xs_snapshot); /* initialize moreLeft/moreRight appropriately for scan direction */ if (ScanDirectionIsForward(dir)) @@ -1153,7 +1153,7 @@ _bt_steppage(IndexScanDesc scan, ScanDirection dir) opaque = (BTPageOpaque) PageGetSpecialPointer(page); if (!P_IGNORE(opaque)) { - PredicateLockPage(rel, blkno); + PredicateLockPage(rel, blkno, scan->xs_snapshot); /* see if there are any matches on this page */ /* note that this will clear moreRight if we can stop */ if (_bt_readpage(scan, dir, P_FIRSTDATAKEY(opaque))) @@ -1201,7 +1201,7 @@ _bt_steppage(IndexScanDesc scan, ScanDirection dir) opaque = (BTPageOpaque) PageGetSpecialPointer(page); if (!P_IGNORE(opaque)) { - PredicateLockPage(rel, BufferGetBlockNumber(so->currPos.buf)); + PredicateLockPage(rel, BufferGetBlockNumber(so->currPos.buf), scan->xs_snapshot); /* see if there are any matches on this page */ /* note that this will clear moreLeft if we can stop */ if (_bt_readpage(scan, dir, PageGetMaxOffsetNumber(page))) @@ -1363,11 +1363,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost) buf = _bt_gettrueroot(rel); if (!BufferIsValid(buf)) - { - /* empty index... */ - PredicateLockRelation(rel); /* Nothing finer to lock exists. */ return InvalidBuffer; - } page = BufferGetPage(buf); opaque = (BTPageOpaque) PageGetSpecialPointer(page); @@ -1444,13 +1440,16 @@ _bt_endpoint(IndexScanDesc scan, ScanDirection dir) if (!BufferIsValid(buf)) { - /* empty index... */ - PredicateLockRelation(rel); /* Nothing finer to lock exists. */ + /* + * Empty index. Lock the whole relation, as nothing finer to lock + * exists. + */ + PredicateLockRelation(rel, scan->xs_snapshot); so->currPos.buf = InvalidBuffer; return false; } - PredicateLockPage(rel, BufferGetBlockNumber(buf)); + PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot); page = BufferGetPage(buf); opaque = (BTPageOpaque) PageGetSpecialPointer(page); Assert(P_ISLEAF(opaque)); diff --git a/src/backend/executor/nodeSeqscan.c b/src/backend/executor/nodeSeqscan.c index 1e566b2d50..f356874b44 100644 --- a/src/backend/executor/nodeSeqscan.c +++ b/src/backend/executor/nodeSeqscan.c @@ -113,7 +113,8 @@ SeqRecheck(SeqScanState *node, TupleTableSlot *slot) TupleTableSlot * ExecSeqScan(SeqScanState *node) { - PredicateLockRelation(node->ss_currentRelation); + PredicateLockRelation(node->ss_currentRelation, + node->ss_currentScanDesc->rs_snapshot); node->ss_currentScanDesc->rs_relpredicatelocked = true; return ExecScan((ScanState *) node, (ExecScanAccessMtd) SeqNext, diff --git a/src/backend/storage/lmgr/predicate.c b/src/backend/storage/lmgr/predicate.c index 02f5d033b2..141d7ff20a 100644 --- a/src/backend/storage/lmgr/predicate.c +++ b/src/backend/storage/lmgr/predicate.c @@ -148,9 +148,11 @@ * predicate lock maintenance * RegisterSerializableTransaction(Snapshot snapshot) * RegisterPredicateLockingXid(void) - * PredicateLockRelation(Relation relation) - * PredicateLockPage(Relation relation, BlockNumber blkno) - * PredicateLockTuple(Relation relation, HeapTuple tuple) + * PredicateLockRelation(Relation relation, Snapshot snapshot) + * PredicateLockPage(Relation relation, BlockNumber blkno, + * Snapshot snapshot) + * PredicateLockTuple(Relation relation, HeapTuple tuple, + * Snapshot snapshot) * PredicateLockPageSplit(Relation relation, BlockNumber oldblkno, * BlockNumber newblkno); * PredicateLockPageCombine(Relation relation, BlockNumber oldblkno, @@ -160,7 +162,8 @@ * * conflict detection (may also trigger rollback) * CheckForSerializableConflictOut(bool visible, Relation relation, - * HeapTupleData *tup, Buffer buffer) + * HeapTupleData *tup, Buffer buffer, + * Snapshot snapshot) * CheckForSerializableConflictIn(Relation relation, HeapTupleData *tup, * Buffer buffer) * CheckTableForSerializableConflictIn(Relation relation) @@ -258,26 +261,6 @@ #define SxactIsROUnsafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_UNSAFE) != 0) #define SxactIsMarkedForDeath(sxact) (((sxact)->flags & SXACT_FLAG_MARKED_FOR_DEATH) != 0) -/* - * Is this relation exempt from predicate locking? We don't do predicate - * locking on system or temporary relations. - */ -#define SkipPredicateLocksForRelation(relation) \ - (((relation)->rd_id < FirstBootstrapObjectId) \ - || RelationUsesLocalBuffers(relation)) - -/* - * When a public interface method is called for serializing a relation within - * the current transaction, this is the test to see if we should do a quick - * return. - */ -#define SkipSerialization(relation) \ - ((!IsolationIsSerializable()) \ - || ((MySerializableXact == InvalidSerializableXact)) \ - || ReleasePredicateLocksIfROSafe() \ - || SkipPredicateLocksForRelation(relation)) - - /* * Compute the hash code associated with a PREDICATELOCKTARGETTAG. * @@ -444,7 +427,6 @@ static void PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag); static void DropAllPredicateLocksFromTable(const Relation relation, bool transfer); static void SetNewSxactGlobalXmin(void); -static bool ReleasePredicateLocksIfROSafe(void); static void ClearOldPredicateLocks(void); static void ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial, bool summarize); @@ -454,6 +436,91 @@ static void FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer); static void OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer); + +/*------------------------------------------------------------------------*/ + +/* + * Does this relation participate in predicate locking? Temporary and system + * relations are exempt. + */ +static inline bool +PredicateLockingNeededForRelation(Relation relation) +{ + return !(relation->rd_id < FirstBootstrapObjectId || + RelationUsesLocalBuffers(relation)); +} + +/* + * When a public interface method is called for a read, this is the test to + * see if we should do a quick return. + * + * Note: this function has side-effects! If this transaction has been flagged + * as RO-safe since the last call, we release all predicate locks and reset + * MySerializableXact. That makes subsequent calls to return quickly. + * + * This is marked as 'inline' to make to eliminate the function call overhead + * in the common case that serialization is not needed. + */ +static inline bool +SerializationNeededForRead(Relation relation, Snapshot snapshot) +{ + /* Nothing to do if this is not a serializable transaction */ + if (MySerializableXact == InvalidSerializableXact) + return false; + + /* + * Don't acquire locks or conflict when scanning with a special snapshot. + * This excludes things like CLUSTER and REINDEX. They use the wholesale + * functions TransferPredicateLocksToHeapRelation() and + * CheckTableForSerializableConflictIn() to participate serialization, but + * the scans involved don't need serialization. + */ + if (!IsMVCCSnapshot(snapshot)) + return false; + + /* + * Check if we have just become "RO-safe". If we have, immediately release + * all locks as they're not needed anymore. This also resets + * MySerializableXact, so that subsequent calls to this function can exit + * quickly. + * + * A transaction is flagged as RO_SAFE if all concurrent R/W + * transactions commit without having conflicts out to an earlier + * snapshot, thus ensuring that no conflicts are possible for this + * transaction. + */ + if (SxactIsROSafe(MySerializableXact)) + { + ReleasePredicateLocks(false); + return false; + } + + /* Check if the relation doesn't participate in predicate locking */ + if (!PredicateLockingNeededForRelation(relation)) + return false; + + return true; /* no excuse to skip predicate locking */ +} + +/* + * Like SerializationNeededForRead(), but called on writes. + * The logic is the same, but there is no snapshot and we can't be RO-safe. + */ +static inline bool +SerializationNeededForWrite(Relation relation) +{ + /* Nothing to do if this is not a serializable transaction */ + if (MySerializableXact == InvalidSerializableXact) + return false; + + /* Check if the relation doesn't participate in predicate locking */ + if (!PredicateLockingNeededForRelation(relation)) + return false; + + return true; /* no excuse to skip predicate locking */ +} + + /*------------------------------------------------------------------------*/ /* @@ -2244,11 +2311,11 @@ PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag) * Clear any finer-grained predicate locks this session has on the relation. */ void -PredicateLockRelation(const Relation relation) +PredicateLockRelation(const Relation relation, const Snapshot snapshot) { PREDICATELOCKTARGETTAG tag; - if (SkipSerialization(relation)) + if (!SerializationNeededForRead(relation, snapshot)) return; SET_PREDICATELOCKTARGETTAG_RELATION(tag, @@ -2267,11 +2334,12 @@ PredicateLockRelation(const Relation relation) * Clear any finer-grained predicate locks this session has on the relation. */ void -PredicateLockPage(const Relation relation, const BlockNumber blkno) +PredicateLockPage(const Relation relation, const BlockNumber blkno, + const Snapshot snapshot) { PREDICATELOCKTARGETTAG tag; - if (SkipSerialization(relation)) + if (!SerializationNeededForRead(relation, snapshot)) return; SET_PREDICATELOCKTARGETTAG_PAGE(tag, @@ -2289,13 +2357,14 @@ PredicateLockPage(const Relation relation, const BlockNumber blkno) * Skip if this is a temporary table. */ void -PredicateLockTuple(const Relation relation, const HeapTuple tuple) +PredicateLockTuple(const Relation relation, const HeapTuple tuple, + const Snapshot snapshot) { PREDICATELOCKTARGETTAG tag; ItemPointer tid; TransactionId targetxmin; - if (SkipSerialization(relation)) + if (!SerializationNeededForRead(relation, snapshot)) return; /* @@ -2666,7 +2735,7 @@ DropAllPredicateLocksFromTable(const Relation relation, bool transfer) if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)) return; - if (SkipPredicateLocksForRelation(relation)) + if (!PredicateLockingNeededForRelation(relation)) return; dbId = relation->rd_node.dbNode; @@ -2881,7 +2950,7 @@ PredicateLockPageSplit(const Relation relation, const BlockNumber oldblkno, if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)) return; - if (SkipPredicateLocksForRelation(relation)) + if (!PredicateLockingNeededForRelation(relation)) return; Assert(oldblkno != newblkno); @@ -3310,30 +3379,6 @@ ReleasePredicateLocks(const bool isCommit) } } -/* - * ReleasePredicateLocksIfROSafe - * Check if the current transaction is read only and operating on - * a safe snapshot. If so, release predicate locks and return - * true. - * - * A transaction is flagged as RO_SAFE if all concurrent R/W - * transactions commit without having conflicts out to an earlier - * snapshot, thus ensuring that no conflicts are possible for this - * transaction. Thus, we call this function as part of the - * SkipSerialization check on all public interface methods. - */ -static bool -ReleasePredicateLocksIfROSafe(void) -{ - if (SxactIsROSafe(MySerializableXact)) - { - ReleasePredicateLocks(false); - return true; - } - else - return false; -} - /* * Clear old predicate locks, belonging to committed transactions that are no * longer interesting to any in-progress transaction. @@ -3679,7 +3724,8 @@ XidIsConcurrent(TransactionId xid) */ void CheckForSerializableConflictOut(const bool visible, const Relation relation, - const HeapTuple tuple, const Buffer buffer) + const HeapTuple tuple, const Buffer buffer, + const Snapshot snapshot) { TransactionId xid; SERIALIZABLEXIDTAG sxidtag; @@ -3687,7 +3733,7 @@ CheckForSerializableConflictOut(const bool visible, const Relation relation, SERIALIZABLEXACT *sxact; HTSV_Result htsvResult; - if (SkipSerialization(relation)) + if (!SerializationNeededForRead(relation, snapshot)) return; if (SxactIsMarkedForDeath(MySerializableXact)) @@ -4064,7 +4110,7 @@ CheckForSerializableConflictIn(const Relation relation, const HeapTuple tuple, { PREDICATELOCKTARGETTAG targettag; - if (SkipSerialization(relation)) + if (!SerializationNeededForWrite(relation)) return; if (SxactIsMarkedForDeath(MySerializableXact)) @@ -4160,7 +4206,7 @@ CheckTableForSerializableConflictIn(const Relation relation) if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)) return; - if (SkipSerialization(relation)) + if (!SerializationNeededForWrite(relation)) return; /* diff --git a/src/include/storage/predicate.h b/src/include/storage/predicate.h index 760c76cff0..aae5326fd8 100644 --- a/src/include/storage/predicate.h +++ b/src/include/storage/predicate.h @@ -44,16 +44,17 @@ extern bool PageIsPredicateLocked(const Relation relation, const BlockNumber blk /* predicate lock maintenance */ extern Snapshot RegisterSerializableTransaction(Snapshot snapshot); extern void RegisterPredicateLockingXid(const TransactionId xid); -extern void PredicateLockRelation(const Relation relation); -extern void PredicateLockPage(const Relation relation, const BlockNumber blkno); -extern void PredicateLockTuple(const Relation relation, const HeapTuple tuple); +extern void PredicateLockRelation(const Relation relation, const Snapshot snapshot); +extern void PredicateLockPage(const Relation relation, const BlockNumber blkno, const Snapshot snapshot); +extern void PredicateLockTuple(const Relation relation, const HeapTuple tuple, const Snapshot snapshot); extern void PredicateLockPageSplit(const Relation relation, const BlockNumber oldblkno, const BlockNumber newblkno); extern void PredicateLockPageCombine(const Relation relation, const BlockNumber oldblkno, const BlockNumber newblkno); extern void TransferPredicateLocksToHeapRelation(const Relation relation); extern void ReleasePredicateLocks(const bool isCommit); /* conflict detection (may also trigger rollback) */ -extern void CheckForSerializableConflictOut(const bool valid, const Relation relation, const HeapTuple tuple, const Buffer buffer); +extern void CheckForSerializableConflictOut(const bool valid, const Relation relation, const HeapTuple tuple, + const Buffer buffer, const Snapshot snapshot); extern void CheckForSerializableConflictIn(const Relation relation, const HeapTuple tuple, const Buffer buffer); extern void CheckTableForSerializableConflictIn(const Relation relation); -- 2.40.0