Horton

Re: Issue in GIN fast-insert: XLogBeginInsert + Read/LockBuffer ordering

Alvaro Herrera <alvherre@alvh.no-ip.org>

View thread View on postgr.es
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Michael Paquier <michael@paquier.xyz>, Zhang Mingli <zmlpostgres@gmail.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>, Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: 2022-10-25T07:37:08Z
Lists: pgsql-hackers 
On 2022-Oct-25, Tom Lane wrote:

> Michael Paquier <michael@paquier.xyz> writes:
> > On Mon, Oct 24, 2022 at 02:22:16PM +0200, Alvaro Herrera wrote:
> >> I confess I don't understand why is it important that XLogBeginInsert is
> >> called inside the critical section.  It seems to me that that part is
> >> only a side-effect of having to acquire the buffer locks in the critical
> >> section.  Right?
> 
> > Yeah, you are right that it would not matter for XLogBeginInsert(),
> > though I'd like to think that this is a good practice on consistency
> > grounds with anywhere else, and we respect what's documented in the
> > README.
> 
> Yeah --- it's documented that way, and there doesn't seem to be
> a good reason not to honor that here.

Okay, so if we follow this argument, then the logical conclusion is that
this *should* be backpatched, after all.

-- 
Álvaro Herrera        Breisgau, Deutschland  —  https://www.EnterpriseDB.com/
Maybe there's lots of data loss but the records of data loss are also lost.
(Lincoln Yeoh)

Commits

  1. Fix ordering issue with WAL operations in GIN fast insert path