From: Tom Lane <tgl@sss.pgh.pa.us>
To: Masahiko Sawada <masahiko.sawada@2ndquadrant.com>
cc: Kyotaro Horiguchi <horikyota.ntt@gmail.com>,
        Fujii Masao <masao.fujii@oss.nttdata.com>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Race condition in SyncRepGetSyncStandbysPriority
In-reply-to: 
 <CA+fd4k4oJZYmi8+U=_8_CH3Lwk2ZQoeV2rmGkYtvoAnHXsQsTQ@mail.gmail.com>
References: <23215.1587051546@sss.pgh.pa.us>
 <20200417.145834.1847060913969327935.horikyota.ntt@gmail.com>
 <3f58063d-2544-aea8-07aa-7e0c35484187@oss.nttdata.com>
 <20200417.170706.1152908438290368702.horikyota.ntt@gmail.com>
 <3020.1587137510@sss.pgh.pa.us>
 <CA+fd4k4oJZYmi8+U=_8_CH3Lwk2ZQoeV2rmGkYtvoAnHXsQsTQ@mail.gmail.com>
Comments: In-reply-to Masahiko Sawada <masahiko.sawada@2ndquadrant.com>
	message dated "Sat, 18 Apr 2020 12:38:54 +0900"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <22663.1587225619.1@sss.pgh.pa.us>
Date: Sat, 18 Apr 2020 12:00:19 -0400
Message-ID: <22664.1587225619@sss.pgh.pa.us>
Precedence: bulk

Masahiko Sawada <masahiko.sawada@2ndquadrant.com> writes:
> On Sat, 18 Apr 2020 at 00:31, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> +   /* Quick out if not even configured to be synchronous */
>> +   if (SyncRepConfig == NULL)
>> +       return false;

> I felt strange a bit that we do the above check in
> SyncRepGetSyncRecPtr() because SyncRepReleaseWaiters() which is the
> only caller says the following before calling it:

Notice there was such a test in SyncRepGetSyncRecPtr already --- I just
moved it to be before doing some work instead of after.

> Can we either change it to an assertion, move it to before acquiring
> SyncRepLock in SyncRepReleaseWaiters or just remove it?

I have no objection to that in principle, but it seems like it's a
change in SyncRepGetSyncRecPtr's API that is not necessary to fix
this bug.  So I'd rather leave it to happen along with the larger
API changes (getting rid of am_sync) that are proposed for v14.

			regards, tom lane