[openchrome-devel] [PATCH/RFT v1 0/6] drm/via: drop use of deprecated headers drmP.h and drm_os_linux.h

[Daniel Vetter]

On Fri, Jul 19, 2019 at 11:05:44AM +0200, Michel Dänzer wrote:
> On 2019-07-19 8:07 a.m., Sam Ravnborg wrote:
> > On Thu, Jul 18, 2019 at 05:37:31PM +0200, Sam Ravnborg wrote:
> >> This is some janitorial updates to the via driver
> >> that is required to get rid of deprecated headers
> >> in the drm subsystem.
> >>
> >> The first three patches are trivial, where
> >> the dependencies on drmP.h and drm_os_linux are dropped.
> >>
> >> The remaining three patches drop use of DRM_WAIT_ON().
> >> They are replaced by wait_event_interruptible_timeout().
> >> These patches could use a more critical review.
> > 
> > The differences between DRM_WAIT_ON() and
> > wait_event_interruptible_timeout() are bigger than anticipated.
> > 
> > The conversion I did for drm_vblank.c is bogus thus I expect
> > the conversion done for via is also bogus.
> 
> What exactly is the problem though? Can you share information about the
> failures you're seeing?
> 
> There was some discussion about DRM_WAIT_ON() "polling" on IRC. I assume
> that refers to it only sleeping for up to 0.01s before checking the
> condition again. In contrast, wait_event_interruptible_timeout() checks
> the condition once, then sleeps up to the full timeout before checking
> it again.
> 
> If that makes a difference for drm_wait_vblank_ioctl, it indicates that
> some other code which updates the vblank count or clears vblank->enabled
> doesn't wake up the vblank->queue.

The return values differ, latest patch from Sam passes CI:

https://patchwork.freedesktop.org/patch/318909/

I have vague memories from exactly this fun when I ripped out DRM_WAIT_ON
from i915.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch