drm/vkms: deadlock between dev->event_lock and timer

[Daniel Stone]

On Mon, 18 Sept 2023 at 23:02, Helen Koike <helen.koike at collabora.com> wrote:
> On 14/09/2023 05:12, Daniel Vetter wrote:
> > Also yes how that landed without anyone running lockdep is ... not good. I
> > guess we need a lockdep enabled drm ci target that runs vkms tests asap
> > :-)
>
> btw, I just executed a draft version of vkms targed on the ci, we do get
> the warning:
>
> https://gitlab.freedesktop.org/helen.fornazier/linux/-/jobs/49156305#L623
>
> I'm just not sure if tests would fail (since it is a warning) and it has
> a chance to be ignored if people don't look at the logs (unless if igt
> already handles that).

Hmm, dmesg-warn is already a separate igt test status. I guess it just
needs to be handled explicitly.

> I still need to do some adjustments (it seems the tests is hanging
> somewhere and we got a timeout) but we should have vkms in drm ci soon.

Might be due to the locking explosion? The kernels should probably
have soft-lockup detection enabled as well as lockdep.

Cheers,
Daniel