pgsql: Teach DSM registry to retry entry initialization if needed. - Mailing list pgsql-committers

From Nathan Bossart
Subject pgsql: Teach DSM registry to retry entry initialization if needed.
Date
Msg-id E1vOMrr-001YSV-24@gemulon.postgresql.org
Whole thread Raw
List pgsql-committers
Teach DSM registry to retry entry initialization if needed.

If DSM registry entry initialization fails, backends could try to
use an uninitialized DSM segment, DSA, or dshash table (since the
entry is still added to the registry).  To fix, restructure the
code so that the registry retries initialization as needed.  This
commit also modifies pg_get_dsm_registry_allocations() to leave out
partially-initialized entries, as they shouldn't have any allocated
memory.

DSM registry entry initialization shouldn't fail often in practice,
but retrying was deemed better than leaving entries in a
permanently failed state (as was done by commit 1165a933aa, which
has since been reverted).

Suggested-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://postgr.es/m/E1vJHUk-006I7r-37%40gemulon.postgresql.org
Backpatch-through: 17

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/dbdc717ac6743074c3a55fc5c380638c91d24afd

Modified Files
--------------
src/backend/storage/ipc/dsm_registry.c | 121 +++++++++++++++++++++------------
1 file changed, 78 insertions(+), 43 deletions(-)


pgsql-committers by date:

Previous
From: Jeff Davis
Date:
Subject: pgsql: Allow pg_locale_t APIs to work when ctype_is_c.
Next
From: David Rowley
Date:
Subject: pgsql: Have the planner replace COUNT(ANY) with COUNT(*), when possible