From: Tom Lane <tgl@sss.pgh.pa.us>
To: Stephen Frost <sfrost@snowman.net>
cc: exclusion@gmail.com, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #17062: Assert failed in RemoveRoleFromObjectPolicy() on DROP
 OWNED policy applied to duplicate role
In-reply-to: <1066940.1623964553@sss.pgh.pa.us>
References: <17062-11f471ae3199ca23@postgresql.org>
 <1066940.1623964553@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Thu, 17 Jun 2021 17:15:53 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1067499.1623965562.1@sss.pgh.pa.us>
Date: Thu, 17 Jun 2021 17:32:42 -0400
Message-ID: <1067500.1623965562@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1067500.1623965562%40sss.pgh.pa.us>
Precedence: bulk

I wroteL
> PG Bug reporting form <noreply@postgresql.org> writes:
>> When executing the following query:
>> CREATE USER role1;
>> CREATE TABLE t1(id int);
>> CREATE POLICY p1 ON t1 TO role1,role1 USING (true);
>> DROP OWNED BY role1;

>> The server halts with the failed assertion:

> Nice.  Seems to be that way at least as far back as 9.6, too.

So the proximate problem is RemoveRoleFromObjectPolicy's unfounded
assumption that there are no duplicate OIDs in a pg_policy.polroles
entry.  But that function has got some other serious problems too:

* Locking.  It acquires lock on the policy's relation only after
it looks up the pg_policy entry.  By that point the entry could
be gone or modified.

* Why is it expensively reconstructing the dependencies of the
policy expressions?  Those aren't going to be changed by this
operation.  AFAICS it ought to be sufficient to remove and
rebuild the policy's shared dependencies.

I wonder whether other operations on policies share either
of these issues.

			regards, tom lane