[Intel-gfx] [PATCH] drm/i915: fix hpd work vs. flush_work in the pageflip code deadlock

[Chris Wilson]

On Mon, Sep 02, 2013 at 08:49:01AM +0200, Daniel Vetter wrote:
> Historically we've run our own driver hotplug handling in our own
> work-queue, which then launched the drm core hotplug handling in the
> system workqueue. This is important since we flush our own driver
> workqueue in the pageflip code while hodling modeset locks, and only
> the drm hotplug code grabbed these locks. But with
> 
> commit 69787f7da6b2adc4054357a661aaa1701a9ca76f
> Author: Daniel Vetter <daniel.vetter at ffwll.ch>
> Date:   Tue Oct 23 18:23:34 2012 +0000
> 
>     drm: run the hpd irq event code directly
> 
> this was changed and now we could deadlock in our flip handler if
> there's a hotplug work blocking the progress of the crucial unpin
> works. So this broke the careful deadlock avoidance implemented in
> 
> commit b4a98e57fc27854b5938fc8b08b68e5e68b91e1f
> Author: Chris Wilson <chris at chris-wilson.co.uk>
> Date:   Thu Nov 1 09:26:26 2012 +0000
> 
>     drm/i915: Flush outstanding unpin tasks before pageflipping
> 
> Since the rule thus far has been that work items on our own workqueue
> may never grab modeset locks simply restore that rule again.
> 
> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Stuart Abercrombie <sabercrombie at chromium.org>
> Reported-by: Stuart Abercrombie <sabercrombie at chromium.org>
> References: http://permalink.gmane.org/gmane.comp.freedesktop.xorg.drivers.intel/26239
> Cc: stable at vger.kernel.org
> Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>

That wins for simplicity, and it is indeed the only caller that requires
mode_config.lock, so

Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>

Bonus would a reminder in i915_drv.h to say that we cannot put items
that require mode_config.lock onto the wq, and that they should go onto
the global workqueue instead. 
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre