Re: Use standard die() handler for SIGTERM in bgworkers

From: Heikki Linnakangas <hlinnaka(at)iki(dot)fi>
To: Kirill Reshke <reshkekirill(at)gmail(dot)com>
Cc: Andres Freund <andres(at)anarazel(dot)de>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Use standard die() handler for SIGTERM in bgworkers
Date: 2026-02-18 18:07:43
Message-ID: d7eb0f7d-c6e1-40d3-9373-5564ff20c27c@iki.fi
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 18/02/2026 10:45, Kirill Reshke wrote:
> On Wed, 18 Feb 2026 at 02:18, Heikki Linnakangas <hlinnaka(at)iki(dot)fi> wrote:
>> + <para>
>> + The default signal handlers merely set interrupt flags
>> + that are processed later by <function>CHECK_FOR_INTERRUPTS()</function>.
>> + <function>CHECK_FOR_INTERRUPTS()</function> should be called in any
>> + long-running loop to ensure that the background worker doesn't prevent the
>> + system from shutting down in a timely fashion.
>> + </para>
>
> Unless you are holding LWLock (or, more generally, inside the CRIT
> section). Should we update wording for this case, or is it maybe
> already assumed in doc?

It's OK to call CHECK_FOR_INTERRUPTS() while holding locks or in a
critical section. It's a no-op in that case.

I wouldn't go into the details here on how to use
CHECK_FOR_INTERRUPTS(). It would be good to document all that somewhere,
but it should go to some more general section on how to write server C
functions as it's applicable to any C code. For background workers, I
think the above is good.

Pushed, thanks for the review!

- Heikki

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2026-02-18 18:39:05 Re: meson: Allow disabling static libraries
Previous Message Nathan Bossart 2026-02-18 18:03:19 Re: convert SpinLock* macros to static inline functions