From ca392df8dd183f93575dd21f1b76390656ca9550 Mon Sep 17 00:00:00 2001 From: Tom Lane Date: Tue, 2 Apr 2024 14:59:04 -0400 Subject: [PATCH] Avoid deadlock during orphan temp table removal. If temp tables have dependencies (such as sequences) then it's possible for autovacuum's cleanup of orphan temp tables to deadlock against an incoming backend that's trying to clean out the temp namespace for its own use. That can happen because RemoveTempRelations' performDeletion call can visit objects within the namespace in an order different from the order in which a per-table deletion will visit them. To fix, observe that performDeletion will begin by taking an exclusive lock on the temp namespace (even though it won't actually delete it). So, if we can get a shared lock on the namespace, we can be sure we're not running concurrently with RemoveTempRelations, while also not conflicting with ordinary use of the namespace. This requires introducing a conditional version of LockDatabaseObject, but that's no big deal. (It's surprising we've got along without that this long.) Report and patch by Mikhail Zhilin. Back-patch to all supported branches. Discussion: https://postgr.es/m/c43ce028-2bc2-4865-9b89-3f706246eed5@postgrespro.ru --- src/backend/postmaster/autovacuum.c | 21 +++++++++++++++- src/backend/storage/lmgr/lmgr.c | 38 +++++++++++++++++++++++++++++ src/include/storage/lmgr.h | 2 ++ 3 files changed, 60 insertions(+), 1 deletion(-) diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c index 18cb8a8e8d6..192837bfa36 100644 --- a/src/backend/postmaster/autovacuum.c +++ b/src/backend/postmaster/autovacuum.c @@ -75,6 +75,7 @@ #include "catalog/dependency.h" #include "catalog/namespace.h" #include "catalog/pg_database.h" +#include "catalog/pg_namespace.h" #include "commands/dbcommands.h" #include "commands/vacuum.h" #include "lib/ilist.h" @@ -2293,6 +2294,24 @@ do_autovacuum(void) continue; } + /* + * Try to lock the temp namespace, too. Even though we have lock on + * the table itself, there's a risk of deadlock against an incoming + * backend trying to clean out the temp namespace, in case this table + * has dependencies (such as sequences) that the backend's + * performDeletion call might visit in a different order. If we can + * get AccessShareLock on the namespace, that's sufficient to ensure + * we're not running concurrently with RemoveTempRelations. If we + * can't, back off and let RemoveTempRelations do its thing. + */ + if (!ConditionalLockDatabaseObject(NamespaceRelationId, + classForm->relnamespace, 0, + AccessShareLock)) + { + UnlockRelationOid(relid, AccessExclusiveLock); + continue; + } + /* OK, let's delete it */ ereport(LOG, (errmsg("autovacuum: dropping orphan temp table \"%s.%s.%s\"", @@ -2310,7 +2329,7 @@ do_autovacuum(void) /* * To commit the deletion, end current transaction and start a new - * one. Note this also releases the lock we took. + * one. Note this also releases the locks we took. */ CommitTransactionCommand(); StartTransactionCommand(); diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c index 2db0424ad94..b717ba2a4b9 100644 --- a/src/backend/storage/lmgr/lmgr.c +++ b/src/backend/storage/lmgr/lmgr.c @@ -991,6 +991,44 @@ LockDatabaseObject(Oid classid, Oid objid, uint16 objsubid, AcceptInvalidationMessages(); } +/* + * ConditionalLockDatabaseObject + * + * As above, but only lock if we can get the lock without blocking. + * Returns true iff the lock was acquired. + */ +bool +ConditionalLockDatabaseObject(Oid classid, Oid objid, uint16 objsubid, + LOCKMODE lockmode) +{ + LOCKTAG tag; + LOCALLOCK *locallock; + LockAcquireResult res; + + SET_LOCKTAG_OBJECT(tag, + MyDatabaseId, + classid, + objid, + objsubid); + + res = LockAcquireExtended(&tag, lockmode, false, true, true, &locallock); + + if (res == LOCKACQUIRE_NOT_AVAIL) + return false; + + /* + * Now that we have the lock, check for invalidation messages; see notes + * in LockRelationOid. + */ + if (res != LOCKACQUIRE_ALREADY_CLEAR) + { + AcceptInvalidationMessages(); + MarkLockClear(locallock); + } + + return true; +} + /* * UnlockDatabaseObject */ diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h index b009559229b..ddac2d2c841 100644 --- a/src/include/storage/lmgr.h +++ b/src/include/storage/lmgr.h @@ -92,6 +92,8 @@ extern void SpeculativeInsertionWait(TransactionId xid, uint32 token); /* Lock a general object (other than a relation) of the current database */ extern void LockDatabaseObject(Oid classid, Oid objid, uint16 objsubid, LOCKMODE lockmode); +extern bool ConditionalLockDatabaseObject(Oid classid, Oid objid, + uint16 objsubid, LOCKMODE lockmode); extern void UnlockDatabaseObject(Oid classid, Oid objid, uint16 objsubid, LOCKMODE lockmode); -- 2.39.5