On Sun, 15 Jul 2012 17:16:34 +0200, Daniel Vetter <daniel at ffwll.ch> wrote:
> On Sun, Jul 15, 2012 at 12:34:22PM +0100, Chris Wilson wrote:
> > The issue is that we stale data in the CPU caches, when we come to
> > swap-out the object, the CPU may short-circuit the reads from those
> > cacheline and so corrupt the context object.
> >
> > Secondary, leaving the context object as being marked in the CPU write
> > domain whilst on the GPU active list is a bad idea and will throw
> > warnings later.
> >
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > Reviewed-by: Ben Widawsky <ben at bwidawsk.net>
> > ---
> > drivers/gpu/drm/i915/i915_gem_context.c | 7 +++++++
> > 1 file changed, 7 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> > index 9ae3f2c..fd978bb 100644
> > --- a/drivers/gpu/drm/i915/i915_gem_context.c
> > +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> > @@ -374,6 +374,13 @@ static int do_switch(struct drm_i915_gem_object *from_obj,
> > if (ret)
> > return ret;
> >
> > + /* Clear this page out of any CPU caches for coherent swap-in/out */
> > + ret = i915_gem_object_set_to_gtt_domain(to->obj, false);
> > + if (ret) {
> > + i915_gem_object_unpin(to->obj);
> > + return ret;
> > + }
> > +
>
> Do I understand things correctly that thanks to clever use of
> set_to_gtt_domain with write = false and not setting a write_domain when
> moving the old context object to the active list we won't block? If so,
> I'll add a short note to that effect to the commit message when merging.
Worse, it was accidentally very clever. :(
Definitely needs a comment and in the future a new function to dtrt
intentionally!
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
