Warning appeared after c8b5a95 ("drm/amdgpu: Fix desktop freezed after gpu-reset")

[Alex Deucher]

On Mon, Jun 19, 2023 at 9:05 AM Christian Kastner <ckk at debian.org> wrote:
>
> Hi,
>
> On a Debian 12 ("bookworm") system, I observed a new warning when I
> upgraded from kernel 6.1.25 to 6.1.27. This is on a system with an RX
> 6800 XT GPU and 3500X processor.
>
> I've traced it down to commit c8b5a95 ("drm/amdgpu: Fix desktop freezed
> after gpu-reset"). Rebuilding the 6.1.27 kernel without this change
> makes the warning disappear.
>
> I can reliably trigger this (and another) warning with
>
>   $ sudo cat /sys/kernel/debug/dri/0/amdgpu_test_ib
>   run ib test:
>   ib ring tests passed.
>
> 5 or 6 seconds after this, two warnings are printed. I see these same
> two warnings on system shutdown (or, at least, they looked similar
> enough to the above that I didn't check for identity).
>
> I've attached
>   (1) the dmesg output after modprobe'ing amdgpu
>   (2) the dmesg output after triggering amdgpu_test_ib
>
> The system in question is only used for ROCm development. I haven't
> observed any other side effects there, other than the warning. There's
> no monitor attached. So I can't speak to the effect of a desktop freeze.

The warnings are harmless, but they have been fixed[1] and the fixes
are making their way back to stable kernels.

[1] - https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=08c677cb0b436a96a836792bb35a8ec5de4999c2

Alex