From 989069eaca788cfd61010e5f88c0feb359b40180 Mon Sep 17 00:00:00 2001 From: Heikki Linnakangas Date: Mon, 22 Jul 2024 01:21:20 +0300 Subject: [PATCH 1/8] Remove LOCK_PRINT() call that could point to garbage If a deadlock is detected in ProcSleep, it removes the process from the wait queue and the shared LOCK and PROCLOCK entries. As soon as it does that, the other process holding the lock might release it, and remove the LOCK entry altogether. So on return from ProcSleep(), it's possible that the LOCK entry is no longer valid. --- src/backend/storage/lmgr/lock.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c index 0400a507779..05b3a243b5c 100644 --- a/src/backend/storage/lmgr/lock.c +++ b/src/backend/storage/lmgr/lock.c @@ -1861,8 +1861,6 @@ WaitOnLock(LOCALLOCK *locallock, ResourceOwner owner, bool dontWait) * now. */ awaitedLock = NULL; - LOCK_PRINT("WaitOnLock: aborting on lock", - locallock->lock, locallock->tag.mode); LWLockRelease(LockHashPartitionLock(locallock->hashcode)); /* -- 2.39.2