[Intel-gfx] [PATCH] drm/i915: Do an optimistic is-busy? check first

[Daniel Vetter]

On Thu, Sep 5, 2013 at 6:46 PM, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> That's incorrect, the client will be told the object is busy.

Oh right, I've completely misread your patch and somehow thought
you're comparing the obj->last_rendering_seqno against the current
seqno ... Mea culpa.

We need at least a ACCESS_RING around the obj->ring deref for
otherwise gcc is allowed to be sneaky and reload a NULL pointer after
the NULL check and let us oops. The other thing that now crossed my
mind is what we're going to do with fbc and friends for legacy
userspace, in case the frontbuffer tracking ever happens. But I guess
we could also do a racy check for that here and avoid any lock
grabbing.

Also the handrolled olr check needs a coment for dense me to not trip
over it ;-)

Cheers, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch