projects / users / heikki / postgres.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 3582b22) | patch
Plug race in dsa_attach.
authorRobert Haas <rhaas@postgresql.org>
Wed, 29 Mar 2017 13:44:29 +0000 (09:44 -0400)
committerRobert Haas <rhaas@postgresql.org>
Wed, 29 Mar 2017 13:48:39 +0000 (09:48 -0400)
commitfddf45b38097d14301d249fbeebca32e40233bd2
tree9f8fa462f059b3a18ef02380570084e6b1c18717tree
parent3582b223d494cd505c6a22da98bcb49f99196645commit | diff
Plug race in dsa_attach.

With sufficiently bad luck, it was possible for a parallel worker to
attempt attach to a DSA area after all other backends have detached
from it, which is not legal.  If the worker had waited a little longer
to get started, the DSM itself would have been destroyed, which is why
this wasn't noticed before.

Thomas Munro, per a report from Andreas Seltenreich

Discussion: http://postgr.es/m/87h92g83t3.fsf@credativ.de
src/backend/utils/mmgr/dsa.c diff | blob | blame | history
Heikki's random PostgreSQL patches