Horton

Re: [PATCH] OAuth: fix performance bug with stuck multiplexer events

Jacob Champion <jacob.champion@enterprisedb.com>

View thread
From: Jacob Champion <jacob.champion@enterprisedb.com>
To: Thomas Munro <thomas.munro@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Daniel Gustafsson <daniel@yesql.se>, Peter Eisentraut <peter@eisentraut.org>
Date: 2025-08-07T18:11:02Z
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 →

  1. oauth: Always link with -lm for floor()

  2. oauth: Add unit tests for multiplexer handling

  3. oauth: Ensure unused socket registrations are removed

  4. oauth: Remove expired timers from the multiplexer

  5. oauth: Remove stale events from the kqueue multiplexer

  6. oauth: Track total call count during a client flow

On Wed, Aug 6, 2025 at 6:46 PM Thomas Munro <thomas.munro@gmail.com> wrote:
> "Unlike epoll descriptors, kqueue descriptors only transition from
> readable to unreadable when kevent() is called and finds nothing,
> after removing level-triggered conditions that have gone away.  We
> therefore need a dummy kevent() call after operations might have been
> performed on the monitored sockets or timer_fd.  Any event returned is
> ignored here, but it also remains queued (being level-triggered) and
> leaves the descriptor readable.  This is a no-op for epoll
> descriptors."

I really like this; I'm working it into the doc comment.

> FWIW I re-read the kqueue paper's discussion of the goals of making
> kqueue descriptors themselves monitorable/pollable, and it seems it
> was mainly intended for hierarchies of kqueues, like your timer_fd,
> with the specific aim of expressing priorities.  It doesn't talk about
> giving them to code that doesn't know it has a kqueue fd (the client)
> and never calls kevent() and infers the events instead (libcurl).

Interesting! It would be nice if they papered over this for us, but I
guess that's water under the bridge.

> s/signalled/signaled/ (= US spelling) in a couple of places.

Ah. Will fix(?) or else lobby the dictionary companies.

Thank you so much for the reviews!

--Jacob