drm-next-misc merge breaks vmwgfx

[Daniel Vetter]

On Thu, Apr 6, 2017 at 4:56 PM, Thomas Hellstrom <thellstrom at vmware.com> wrote:
> On 04/06/2017 04:47 PM, Daniel Vetter wrote:
>> On Thu, Apr 6, 2017 at 4:10 PM, Thomas Hellstrom <thellstrom at vmware.com> wrote:
>>> Also, when testing the tip of drm-misc-next (with the non-atomic vmwgfx)
>>> there appeared to be warnings about a non-NULL
>>> dev->mode_config.acquire_ctx. I'll see if I can reproduce those, but
>>> perhaps removing the line
>>>
>>> dev->mode_config.acquire_ctx = &ctx
>>>
>>> in drm_mode_setcrtc()
>>>
>>> is part of the problem.
>> Hm, where do you hit that? And by tip of drm-misc-next, do you mean
>> the very latest state, which includes atomic vmwgfx, or is this with
>> the non-atomic vmwgfx? Please paste the backtraces (and for which tree
>> they are).
>> -Daniel
>
> Actually this must have been from a confused rebase-bisect state
> somewhere. I can't reproduce this.

Have you changed CONFIG_DEBUG_WW_MUTEX_SLOWPATH perhaps? Without this
you shouldn't be able to hit any retry path (so minus warnings the
patch you bisected won't have an affected), with you will hit the
deadlock retry paths pseudo-randomly. It might take a few trials to
hit it again, depending upon timing.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch