Re: ci.rocm.debian.net: gfx1100, gfx1101 broken with (possibly) 6.12.11

[Mario Limonciello <superm1@gmail.com>]

At least amdgpu changes to 6.12.11 nothing really stands out.  Could 
your roll back to 6.12.10 to see if that helps and confirm it was really 
a kernel regression?

On 2/5/25 13:59, Christian Kastner wrote:
> Hi,
>
> I just noticed that gfx1100, gfx1101 started failing/tmpfailing
> recently, eg [1, 2].
>
> In both cases, I looked at some of the logs, and the failures seem to
> coincide with the 2025-01-26 upload of kernel 6.12.11. Tests still
> completed successfully with 6.12.10 before that.
>
> However, annoyingly I also updated the host (gfx1100 and gfx1101 run in
> QEMU VMs) on that day and while the updates were virtually all from the
> 12.9 release, I wouldn't rule out that maybe something broke on the
> host.
>
> I did not investigate this on bare metal I won't get to that anytime
> soon. In case anyone else wants to try to reproduce, maybe even bisect
> this. 6.12.11 did have some driver updates [3].
>
> Best,
> Christian
>
> [1]: https://ci.rocm.debian.net/packages/r/rocrand/unstable/amd64+gfx1100/
> [2]: https://ci.rocm.debian.net/packages/r/rocrand/unstable/amd64+gfx1101/
> [3]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/log/?h=v6.12.11