]> granicus.if.org Git - postgresql/commit
shm_mq: Third attempt at fixing nowait behavior in shm_mq_receive.
authorRobert Haas <rhaas@postgresql.org>
Tue, 3 Nov 2015 14:12:52 +0000 (09:12 -0500)
committerRobert Haas <rhaas@postgresql.org>
Tue, 3 Nov 2015 14:12:52 +0000 (09:12 -0500)
commit4efe26cbd3ef0d85656bf00ac9e5bd41cc8a2a36
tree85aa24973256684c420378fe4f3264a3c9bfc01d
parent0279f62fdca7ff19ef34a836f5201935dc7f627c
shm_mq: Third attempt at fixing nowait behavior in shm_mq_receive.

Commit a1480ec1d3bacb9acb08ec09f22bc25bc033115b purported to fix the
problems with commit b2ccb5f4e6c81305386edb34daf7d1d1e1ee112a, but it
didn't completely fix them.  The problem is that the checks were
performed in the wrong order, leading to a race condition.  If the
sender attached, sent a message, and detached after the receiver
called shm_mq_get_sender and before the receiver called
shm_mq_counterparty_gone, we'd incorrectly return SHM_MQ_DETACHED
before all messages were read.  Repair by reversing the order of
operations, and add a long comment explaining why this new logic is
(hopefully) correct.
src/backend/storage/ipc/shm_mq.c