Bug 32023

Summary:	Kernel 6.3 is in big trouble when using the nouveau driver
Product:	Mageia	Reporter:	Mike Burgener <mburgener>
Component:	RPM Packages	Assignee:	Kernel and Drivers maintainers <kernel>
Status:	RESOLVED FIXED	QA Contact:
Severity:	critical
Priority:	release_blocker	CC:	davidwhodgins, fri, stephengermany
Version:	Cauldron
Target Milestone:	---
Hardware:	All
OS:	Linux
Whiteboard:
Source RPM:		CVE:
Status comment:
Attachments:	Relevant system specs

Description Mike Burgener 2023-06-16 07:39:45 CEST

Hi everybody,
looks like Kernel 6.3 is in big trouble when using the nouveau driver, see https://www.phoronix.com/news/Avoid-Nouveau-Linux-6.3

Taken from https://blog.mageia.org/en/2023/05/24/the-release-of-beta2-brings-mageia-9-stable-closer-to-reality/ mageia uses 6.3.3 on Mga9, thus this is release-critical, as it may break everything even already in early install stage when using the nouveau driver. please have a look at it and decide if it is critical after review.

BTW i could not find the mga9 Kernel 6.3 package anywhere on https://madb.mageia.org/

Kind regards

Mike Burgener 2023-06-16 07:44:12 CEST

Priority: Normal => release_blocker

Comment 1 Mike Burgener 2023-06-16 08:48:48 CEST

Also see https://gitlab.freedesktop.org/drm/nouveau/-/issues/213

Comment 2 Mike Burgener 2023-06-16 08:53:04 CEST

A patch seems available now in the gitlab freedesktop issue

Comment 3 Morgan Leijström 2023-06-16 09:43:38 CEST

Cauldron/mga9 is currently at 6.3.8

I guess our kernel wizard is into it already

CC: (none) => fri
Assignee: bugsquad => kernel

Comment 4 Mike Burgener 2023-06-16 10:23:38 CEST

(In reply to Morgan Leijström from comment #3)
> Cauldron/mga9 is currently at 6.3.8
> 
> I guess our kernel wizard is into it already

Perfect, is it tmb? If so, he is not a wizard, he is a reincarnation of Merlin in terms of Kernels :D

Comment 5 Dave Hodgins 2023-06-16 20:40:59 CEST

From the gitlab bug report, there's a patch being tested
https://gitlab.freedesktop.org/drm/nouveau/uploads/150dc8a040dc18aee72fa12d7c506bc3/0001-nouveau-fix-client-work-fence-deletion-race.patch

CC: (none) => davidwhodgins

Comment 6 Thomas Backlund 2023-06-16 20:53:10 CEST

(In reply to Morgan Leijström from comment #3)
> Cauldron/mga9 is currently at 6.3.8
> 
> I guess our kernel wizard is into it already

Yes, 
I'm aware of the issue and have been monitoring it while upstream folks was figuring out how to root-cause and fix it...

I'm currently travelling so it will be a few days before a new kernel will land ...

Comment 7 Stephen Germany 2023-06-17 14:14:06 CEST

I have not had that issue.  Maybe hit-n-miss or only certain Nvidia chips?  Just guessing..

Created an attachment with specs.

CC: (none) => stephengermany

Comment 8 Stephen Germany 2023-06-17 14:17:04 CEST

Created attachment 13878 [details]
Relevant system specs

Comment 9 Dave Hodgins 2023-06-17 21:45:45 CEST

From the report, it only causes major problems some cards. The problem
with memory corruption, is that it may or may not cause noticeable issues
depending on what gets corrupted. If it triggers file-system corruption
the damage is permanent, but depending on which files get corrupted, the
corruption may or may not be noticeable.

Comment 10 Thomas Backlund 2023-06-19 18:46:42 CEST

Upstream fix merged in kernel-6.3.8-2.mga9 (currently building) and will be on public mirrors in some hours...

Resolution: (none) => FIXED
Status: NEW => RESOLVED