Date: Thu, 14 Nov 2019 23:22:32 +0100
From: Tomas Vondra <tomas.vondra@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Mark Dilger <hornschnorter@gmail.com>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: Missing dependency tracking for TableFunc nodes
Message-ID: <20191114222232.o76u6jdvf53ozjjk@development>
References: <18427.1573508501@sss.pgh.pa.us>
 <997cbcae-7739-f871-84a0-1192380abb54@gmail.com>
 <15228.1573695479@sss.pgh.pa.us>
 <20191114213124.i6xwx24y26iklzds@development>
 <14155.1573767414@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
In-Reply-To: <14155.1573767414@sss.pgh.pa.us>
User-Agent: NeoMutt/20180716
Precedence: bulk

On Thu, Nov 14, 2019 at 04:36:54PM -0500, Tom Lane wrote:
>Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
>> On Wed, Nov 13, 2019 at 08:37:59PM -0500, Tom Lane wrote:
>>> I concur with Tomas' suspicion that this must be a race condition
>>> during DROP STATISTICS.  If we're going to allow people to do that
>>> separately from dropping the table(s), there has to be some kind of
>>> locking around it, and it sounds like there's not :-(
>
>> I think the right thing to do is simply acquire AE lock on the relation
>> in RemoveStatisticsById, per the attached patch. It's possible we'll
>> need to do something more complicated once join stats are added, but
>> for now this should be enough (and backpatchable).
>
>Hm.  No, it's not enough, unless you add more logic to deal with the
>possibility that the stats object is gone by the time you have the
>table lock.  Consider e.g. two concurrent DROP STATISTICS commands,
>or a statistics drop that's cascading from something like a drop
>of a relevant function and so has no earlier table lock.
>

Isn't that solved by RemoveObjects() doing this?

    /* Get an ObjectAddress for the object. */
    address = get_object_address(stmt->removeType,
                                 object,
                                 &relation,
                                 AccessExclusiveLock,
                                 stmt->missing_ok);

I've actually done some debugging before sending the patch, and I think
this prevent the issue you describe.


regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services