Re: Non-reproducible AIO failure

Andres Freund <andres@anarazel.de>

From: Andres Freund <andres@anarazel.de>

To: Konstantin Knizhnik <knizhnik@garret.ru>

Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>

Date: 2025-06-16T15:11:29Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

aio: Stop using enum bitfields due to bad code generation
- ce161b194e84 18.0 landed
- 5865150b6d53 19 (unreleased) landed
amcheck: Fix posting tree checks in gin_index_check()
- 0cf205e122ae 18.0 cited
aio: Add missing memory barrier when waiting for IO handle
- e9a3615a5224 18.0 landed

Hi,

On 2025-06-16 14:11:39 +0300, Konstantin Knizhnik wrote:
> One more update: with the proposed patch (memory barrier before
> `ConditionVariableBroadcast` in `pgaio_io_process_completion`

I don't see how that barrier could be required for correctness -
ConditionVariableBroadcast() is a barrier itself (as the comment above the
call notes, too).

On 2025-06-15 14:48:43 +0300, Konstantin Knizhnik wrote:
> Also I think that replacing bitfields with `uint8` and may be even with
> `int`, is good idea at least to avoids false sharing.

I don't think there's false sharing here. And even if there were, the
granularity at which false sharing occurs is a cache line size, so either 64
or 128 byte.

I unfortunately can't repro this issue so far.  I don't think it's the same
problem as my patch fixes, so I'll push my patch.  How exactly did you
reproduce the probelm?

Greetings,

Andres Freund