[Intel-gfx] [PATCH 12/12] drm/i915: Cache the reset_counter for the request

[Chris Wilson]

On Wed, Nov 25, 2015 at 10:12:53AM +0100, Daniel Vetter wrote:
> On Tue, Nov 24, 2015 at 09:43:32PM +0000, Chris Wilson wrote:
> > The wait-request interface still has the wait-seqno legacy of having to
> > acquire the reset_counter under the mutex before calling. That is quite
> > hairy and causes a major complication later where we want to amalgamate
> > waiters.
> 
> Yeah, that's a sensible change. But since I don't yet understand where
> exactly the current logic results in a wedge gpu I can't convince myself
> that the new one is ok.

Just to show where I am going with the interface change and why I think
it so important (other than it removes a hairy dance that we've have had
to copy all around the code):

http://cgit.freedesktop.org/~ickle/linux-2.6/commit/?h=nightly&id=75e73616f72f14ab82037adb1b2b054e2e500c21
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre