Fix ReinitializeParallelDSM to tolerate finding no error queues.
authorRobert Haas <rhaas@postgresql.org>
Tue, 28 Nov 2017 17:05:52 +0000 (12:05 -0500)
committerRobert Haas <rhaas@postgresql.org>
Tue, 28 Nov 2017 17:19:19 +0000 (12:19 -0500)
commitdba6e75c1113dbfc7b10d93cc6c0babd8b7d852b
treea7ffa4b9cbe774b91f85e953c3a608e5220cb67b
parentec7629dfb98a34cf4e459cf6b27febfd33ae9ad5
Fix ReinitializeParallelDSM to tolerate finding no error queues.

Commit d4663350646ca0c069a36d906155a0f7e3372eb7 changed things so
that shm_toc_lookup would fail with an error rather than silently
returning NULL in the hope that such failures would be reported
in a useful way rather than via a system crash.  However, it
overlooked the fact that the lookup of PARALLEL_KEY_ERROR_QUEUE
in ReinitializeParallelDSM is expected to fail when no DSM segment
was created in the first place; in that case, we end up with a
backend-private memory segment that still contains an entry for
PARALLEL_KEY_FIXED but no others.  Consequently a benign failure
to initialize parallelism can escalate into an elog(ERROR);
repair.

Discussion: http://postgr.es/m/CA+Tgmob8LFw55DzH1QEREpBEA9RJ_W_amhBFCVZ6WMwUhVpOqg@mail.gmail.com
src/backend/access/transam/parallel.c