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

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

 



On Thu, Sep 05, 2013 at 04:48:54PM +0200, Daniel Vetter wrote:
> On Thu, Sep 05, 2013 at 03:29:01PM +0100, Chris Wilson wrote:
> > The busy-ioctl is frequently called as it is used by clients treating
> > buffer objects like fences and polling for completion. These frequent
> > calls are especially subject to contention with multiple clients leading
> > to a lot of busy-waiting due to mutex_spin_on_owner()). We can, in most
> > cases, report whether the bo is idle without touching struct_mutex, and
> > only resort to hitting the lock if we need to queue flushes for the bo.
> > 
> > Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> > ---
> >  drivers/gpu/drm/i915/i915_gem.c | 27 +++++++++++++++++++++------
> >  1 file changed, 21 insertions(+), 6 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> > index 3b942ce..fed0b39 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.c
> > +++ b/drivers/gpu/drm/i915/i915_gem.c
> > @@ -4068,18 +4068,34 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
> >  {
> >  	struct drm_i915_gem_busy *args = data;
> >  	struct drm_i915_gem_object *obj;
> > +	struct intel_ring_buffer *ring;
> >  	int ret;
> >  
> > -	ret = i915_mutex_lock_interruptible(dev);
> > -	if (ret)
> > -		return ret;
> > -
> >  	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
> >  	if (&obj->base == NULL) {
> >  		ret = -ENOENT;
> > -		goto unlock;
> > +		goto out_unlocked;
> >  	}
> >  
> > +	/* Do an optimistic check for activity - we don't care about userspace
> > +	 * racing with itself, that is always problematic.
> > +	 */
> > +	ring = obj->ring;
> > +	if (ring && obj->last_read_seqno == ring->outstanding_lazy_seqno)
> > +		goto lock_and_flush;
> 
> Feels a bit too tricky ... How useful is just an
> 
> 	if (ACCESS_ONCE(obj->active))
> 		goto lock_and_flush;

That's actually racy as hell since active is in a bitfield. So maybe just
ACCESS_ONCE(obj->ring)?
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
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