On Fri, Mar 24, 2017 at 04:16:28AM +0100, Mario Kleiner wrote: > Looks good to me. As a further optimization, i think we could move > the vblank_disable_fn() call outside/below the > spin_unlock_irqrestore for event_lock, as vblank_disable_fn() > doesn't need any locks held at call time, so slightly reduce > event_lock hold time. Don't know if it is worth it. Hmm, I was cautious since I don't trust myself around the vbl spinlocks. But yes, the vblank->refcount is controlled by vbl_lock and that is checked inside the disable callback, so we can move it out from the vblank_event_lock. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
