[PATCH 3/3] drm: do not sleep on vblank while holding a mutex
daniel at ffwll.ch
Fri Oct 28 07:51:49 PDT 2011
On Fri, Oct 28, 2011 at 07:10:51AM -0500, Ilija Hadzic wrote:
> I'll keep it then and figure out the best mutex/spinlock to use. It
> can be anything that exists on one-per-CRTC basis (vblank waits on
> different CTCs are not contending). The critical section is from
> that switch in which vblwait->request.sequence is incremented until
> it is assigned to dev->last_vblank_wait[crtc] (and we are only
> protecting that section against itself, executed in contexts of
> different PIDs).
> I guess we settled now on this patch (other comments will be
> addressed in a different set of patches).
If you settle on keeping ->last_vblank_wait I think the easiest is to wrap
it with the dev->vbl_lock spinlock everywhere.
Mail: daniel at ffwll.ch
Mobile: +41 (0)79 365 57 48
More information about the dri-devel