Drop slot's LWLock before returning from SaveSlotToPath()
authorPeter Eisentraut <peter@eisentraut.org>
Thu, 26 Mar 2020 10:51:39 +0000 (11:51 +0100)
committerPeter Eisentraut <peter@eisentraut.org>
Thu, 26 Mar 2020 13:08:15 +0000 (14:08 +0100)
commitce28a43ffa89b49584e75d6bb9f8ae03a8e13151
tree0ad2ce62fdd83f70be5c2c7162053a89053502ec
parent2af8894c409d0464d4f29449ac39d7d229ccc6ef
Drop slot's LWLock before returning from SaveSlotToPath()

When SaveSlotToPath() is called with elevel=LOG, the early exits didn't
release the slot's io_in_progress_lock.

This could result in a walsender being stuck on the lock forever.  A
possible way to get into this situation is if the offending code paths
are triggered in a low disk space situation.

Author: Pavan Deolasee <pavan.deolasee@2ndquadrant.com>
Reported-by: Craig Ringer <craig@2ndquadrant.com>
Discussion: http://www.postgresql.org/message-id/flat/56a138c5-de61-f553-7e8f-6789296de785%402ndquadrant.com
src/backend/replication/slot.c