Chris@axboe Looks like this bug (but not the fix) has made it to the Ubuntu 24.04.4 kernel (6.17.0-20-generic): https://github.com/axboe/liburing/issues/1509 (it was not in 24.04.3), which is unfortunate as that's what GitHub has rolled out as "ubuntu-latest" on their GHA runners 🫠do you know of a work-around or detection we can apply?
Jens Axboe@chrisvest You'd really need to ping Ubuntu to sort out their kernel. I always have a hard time with distro kernels that aren't based on a stable release, as they end up kind of wonky because of it. I can ask around and see if anyone has an ubuntu kernel contact so we can get it sorted. But again, distro non-stable kernels, it's such a pita and not a business I want to deal with really.
@axboe ok, I’ll try to bother the MS people since they seem to do their own custom kernels on top of Ubuntu.
@chrisvest Since I'm a nice guy, I did a stable kind of backport for the defunct 6.17-stable. It has any io_uring commit you'd want, and the backport of the af_unix fix as well because of commit:
commit df30285b3670bf52e1e5512e4d4482bec5e93c16
Author: Kuniyuki Iwashima <[email protected]>
Date: Wed Jul 2 22:35:18 2025 +0000
af_unix: Introduce SO_INQ.
which is what introduced this bug.
Find it here: https://web.git.kernel.org/pub/scm/linux/kernel/git/axboe/stable.git/
@chrisvest I sent an email to Manuel Diewald as well, who _seems_ to be the person at Canonical spinning these kernels. Offered to keep doing io_uring related backports too, should be in both our interests to ensure that a) those are done when appropriate, and b) that they are done by someone who knows the code.
We'll see what comes of this... It was bad enough with their 6.8 based kernel, would be nice to not repeat that with a newer one. But for God's sake, just use a stable kernel, Ubuntu...
@chrisvest private reply with your email and I'll fwd it to you.