On Tue, Aug 06, 2013 at 09:59:46AM +0100, Chris Wilson wrote:
> We can apply the same optimisation tricks as kref_put_mutex() in our
> local equivalent function. However, we have a different locking semantic
> (we unlock ourselves, in kref_put_mutex() the callee unlocks) so that we
> can use the same callbacks for both locked and unlocked kref_put()s and
> so can not simply convert to using kref_put_mutex() directly.
>
> Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
I think longterm we want to move to delayed free callbacks (similar to how
fput works) since the locking with dma-buf and all will simply get too
hairy. But for now this is a neat optimization imo, so
Reviewed-by: Daniel Vetter <daniel.vetter@xxxxxxxx>
> ---
> include/drm/drmP.h | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/include/drm/drmP.h b/include/drm/drmP.h
> index 4b518e0..ee2ef27 100644
> --- a/include/drm/drmP.h
> +++ b/include/drm/drmP.h
> @@ -1626,10 +1626,12 @@ drm_gem_object_unreference(struct drm_gem_object *obj)
> static inline void
> drm_gem_object_unreference_unlocked(struct drm_gem_object *obj)
> {
> - if (obj != NULL) {
> + if (obj && !atomic_add_unless(&obj->refcount.refcount, -1, 1)) {
> struct drm_device *dev = obj->dev;
> +
> mutex_lock(&dev->struct_mutex);
> - kref_put(&obj->refcount, drm_gem_object_free);
> + if (likely(atomic_dec_and_test(&obj->refcount.refcount)))
> + drm_gem_object_free(&obj->refcount);
> mutex_unlock(&dev->struct_mutex);
> }
> }
> --
> 1.8.4.rc1
>
> _______________________________________________
> dri-devel mailing list
> dri-devel@xxxxxxxxxxxxxxxxxxxxx
> http://lists.freedesktop.org/mailman/listinfo/dri-devel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/dri-devel
