Re: [PATCH 2/3] drm/i915: Do not call retire_requests from wait_for_rendering

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, Mar 28, 2014 at 03:58:25PM -0700, Volkin, Bradley D wrote:
> On Mon, Mar 17, 2014 at 05:21:55AM -0700, Chris Wilson wrote:
> > @@ -1949,58 +1956,58 @@ static unsigned long
> >  __i915_gem_shrink(struct drm_i915_private *dev_priv, long target,
> >  		  bool purgeable_only)
> >  {
> > -	struct list_head still_bound_list;
> > -	struct drm_i915_gem_object *obj, *next;
> > +	struct list_head still_in_list;
> > +	struct drm_i915_gem_object *obj;
> >  	unsigned long count = 0;
> >  
> > -	list_for_each_entry_safe(obj, next,
> > -				 &dev_priv->mm.unbound_list,
> > -				 global_list) {
> > -		if ((i915_gem_object_is_purgeable(obj) || !purgeable_only) &&
> > -		    i915_gem_object_put_pages(obj) == 0) {
> > -			count += obj->base.size >> PAGE_SHIFT;
> > -			if (count >= target)
> > -				return count;
> > -		}
> > -	}
> > -
> >  	/*
> > -	 * As we may completely rewrite the bound list whilst unbinding
> > +	 * As we may completely rewrite the (un)bound list whilst unbinding
> >  	 * (due to retiring requests) we have to strictly process only
> >  	 * one element of the list at the time, and recheck the list
> >  	 * on every iteration.
> 
> Is it still true that we could retire requests on this path? I see that
> currently we will retire requests via:
> i915_vma_unbind -> i915_gem_object_finish_gpu -> i915_gem_object_wait_rendering.
> 
> But we've taken the explicit request retirement out of the wait_rendering path.
> Have I missed somewhere that it could still happen?

Yes, as wait_rendering doesn't retire all the requests, we may still have
a request associated with the bo. This will then cause us to call
i915_gem_object_retire() during i915_gem_object_put_pages() (through
i915_gem_object_set_to_cpu_domain) thereby discard the last active
reference and destroying the object unless we take care.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx




[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux