]> granicus.if.org Git - postgresql/commit
Fix RBM_ZERO_AND_LOCK mode to not acquire lock on local buffers.
authorHeikki Linnakangas <heikki.linnakangas@iki.fi>
Wed, 13 May 2015 06:44:43 +0000 (09:44 +0300)
committerHeikki Linnakangas <heikki.linnakangas@iki.fi>
Wed, 13 May 2015 07:06:38 +0000 (10:06 +0300)
commitbd1cfde70e0f6165cd794488e88a57df9c233b43
tree427ed180435f01dec96b5b161fa5eac0564f6ce3
parentb93c8eaf8c35485175358c552cd5adce0e197ec9
Fix RBM_ZERO_AND_LOCK mode to not acquire lock on local buffers.

Commit 81c45081 introduced a new RBM_ZERO_AND_LOCK mode to ReadBuffer, which
takes a lock on the buffer before zeroing it. However, you cannot take a
lock on a local buffer, and you got a segfault instead. The version of that
patch committed to master included a check for !isLocalBuf, and therefore
didn't crash, but oddly I missed that in the back-patched versions. This
patch adds that check to the back-branches too.

RBM_ZERO_AND_LOCK mode is only used during WAL replay, and in hash indexes.
WAL replay only deals with shared buffers, so the only way to trigger the
bug is with a temporary hash index.

Reported by Artem Ignatyev, analysis by Tom Lane.
src/backend/storage/buffer/bufmgr.c