[Intel-gfx] [PATCH 1/2] drm/i915: signal first fence from irq handler if complete

[Chris Wilson]

On Wed, Feb 22, 2017 at 04:53:35PM +0000, Tvrtko Ursulin wrote:
> 
> On 17/02/2017 15:51, Chris Wilson wrote:
> > 	if (engine->irq_seqno_barrier &&
> >-	    rcu_access_pointer(engine->breadcrumbs.irq_seqno_bh) == current &&
> > 	    test_and_clear_bit(ENGINE_IRQ_BREADCRUMB, &engine->irq_posted)) {
> >-		struct task_struct *tsk;
> >+		unsigned long flags;
> >
> > 		/* The ordering of irq_posted versus applying the barrier
> > 		 * is crucial. The clearing of the current irq_posted must
> >@@ -4058,17 +4061,17 @@ __i915_request_irq_complete(const struct drm_i915_gem_request *req)
> > 		 * the seqno before we believe it coherent since they see
> > 		 * irq_posted == false but we are still running).
> > 		 */
> >-		rcu_read_lock();
> >-		tsk = rcu_dereference(engine->breadcrumbs.irq_seqno_bh);
> >-		if (tsk && tsk != current)
> >+		spin_lock_irqsave(&engine->breadcrumbs.lock, flags);
> >+		if (engine->breadcrumbs.first_wait &&
> >+		    engine->breadcrumbs.first_wait->tsk != current)
> > 			/* Note that if the bottom-half is changed as we
> > 			 * are sending the wake-up, the new bottom-half will
> > 			 * be woken by whomever made the change. We only have
> > 			 * to worry about when we steal the irq-posted for
> > 			 * ourself.
> > 			 */
> >-			wake_up_process(tsk);
> >-		rcu_read_unlock();
> >+			wake_up_process(engine->breadcrumbs.first_wait->tsk);
> >+		spin_unlock_irqrestore(&engine->breadcrumbs.lock, flags);
> 
> Worth caching &engine->breadcrumbs maybe?

Makes no difference to object code, but makes it more pleasant to read,
so yes.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre