pgsql: Fix RI fast-path permission checks

From: Amit Langote <amitlan(at)postgresql(dot)org>
To: pgsql-committers(at)lists(dot)postgresql(dot)org
Subject: pgsql: Fix RI fast-path permission checks
Date: 2026-09-19 02:05:59
Message-ID: E1x7kSl-00000000Kzz-3T27@gemulon.postgresql.org
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-committers

Fix RI fast-path permission checks

The fast path required table-level SELECT on the referenced table,
rejecting checks that the SPI path allows with column-level grants.
It also omitted the UPDATE privilege required by FOR KEY SHARE.

When table privileges do not suffice, use ExecCheckOneRelPerms() with
the referenced key columns as selectedCols and an empty updatedCols.
This accepts SELECT on all referenced columns and UPDATE on any column,
the same privileges the executor would require for the SELECT ... FOR
KEY SHARE the SPI path runs. Keep the table-privilege check as a
shortcut that avoids constructing a column bitmap in the usual case.

Add missing regression test coverage for the fixed cases.

Reported-by: Nikolay Samokhvalov <nik(at)postgres(dot)ai>
Author: Nikolay Samokhvalov <nik(at)postgres(dot)ai>
Co-authored-by: Amit Langote <amitlangote09(at)gmail(dot)com>
Discussion: https://www.postgr.es/m/CAM527d9BgPjeOOYmbCBTd57R145qHCk-dzw9qNq%2BnOrDq1j__A%40mail.gmail.com
Backpatch-through: 19

Branch
------
REL_19_STABLE

Details
-------
https://git.postgresql.org/pg/commitdiff/2fc654e1133e987319d54661f0f7ee2babb33488

Modified Files
--------------
src/backend/utils/adt/ri_triggers.c | 45 +++++++++++++++++------
src/test/regress/expected/foreign_key.out | 55 ++++++++++++++++++++++++++--
src/test/regress/sql/foreign_key.sql | 59 +++++++++++++++++++++++++++++--
3 files changed, 145 insertions(+), 14 deletions(-)

Browse pgsql-committers by date

  From Date Subject
Next Message Amit Langote 2026-09-19 02:06:13 pgsql: Fix RI fast-path permission checks
Previous Message Heikki Linnakangas 2026-09-18 21:32:54 pgsql: Check that oldestXID and oldestMulti are consistent at pg_upgrad