[Intel-gfx] [PATCH 3/3] drm/i915: kicking rings considered harmful

[Daniel Vetter]

On Sat, Oct 01, 2011 at 11:09:33PM -0700, Ben Widawsky wrote:
> On Sat, 01 Oct 2011 20:59:28 -0700
> Eric Anholt <eric at anholt.net> wrote:
> 
> > On Sat,  1 Oct 2011 19:15:19 -0700, Ben Widawsky <ben at bwidawsk.net> wrote:
> > > From: Daniel Vetter <daniel.vetter at ffwll.ch>
> > > 
> > > Only do it in the hope of resurrecting the gpu. Disable when reset is
> > > disabled because it seems to tremendously increases our changes to
> > > actually capture an error_state before the system goes all belly-up.
> > 
> > This patch needs a better subject, like "Disable stuck semaphore kick when
> > GPU reset is disabled", instead of something suggesting that kicking
> > rings is bad in general and going to be removed by the patch.
> 
> I'm in favor of removing the ring kick code. I've never seen it actually work,
> and Chris mentioned something on IRC about how it's no longer quite as
> necessary. I'll let Daniel handle the authoritative response though.

Yeah, just kill the stuck semaphore ring kick code. If we deadlock on
semaphores or something like that, we likely have a massive problem and a
full reset sounds like the best course of action. Otoh I think we need to
retain the kick ring stuck in MI_WAIT code, at least to keep on supporting
old (broken) userspace.

Cheers, Daniel
-- 
Daniel Vetter
Mail: daniel at ffwll.ch
Mobile: +41 (0)79 365 57 48