Date: Tue, 18 Jun 2019 18:25:47 -0400
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Oleksii Kliukin <alexk@hintbits.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, pgsql-hackers@lists.postgresql.org
Subject: Re: pgsql: Avoid spurious deadlocks when upgrading a tuple lock
Message-ID: <20190618222547.GA18769@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <DB0AA460-7962-4E8D-AC1A-958457DB6563@hintbits.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On 2019-Jun-18, Oleksii Kliukin wrote:

> Sorry, I was confused, as I was looking only at
> https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=de87a084c0a5ac927017cd0834b33a932651cfc9
> 
> without taking your subsequent commit that silences compiler warnings at
> https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=3da73d6839dc47f1f47ca57974bf28e5abd9b572
> into consideration. With that commit, the danger is indeed in resetting the
> skip mechanism on each jump and potentially causing deadlocks.

Yeah, I understand the confusion.

Anyway, as bugs go, this one seems pretty benign.  It would result in a
unexplained deadlock, very rarely, and only for people who use a very
strange locking pattern that includes (row-level) lock upgrades.  I
think it also requires aborted savepoints too, though I don't rule out
the possibility that there might be a way to reproduce this without
that.

I pushed the patch again just now, with the new permutation.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services