drm/i915: Drop no-semaphore boosting

Now that we have fast timeslicing on semaphores, we no longer need to
prioritise none-semaphore work as we will yield any work blocked on a
semaphore to the next in the queue. Previously with no timeslicing,
blocking on the semaphore caused extremely bad scheduling with multiple
clients utilising multiple rings. Now, there is no impact and we can
remove the complication.

Signed-off-by: Chris Wilson <[e-mail]>
Reviewed-by: Mika Kuoppala <[e-mail]>
Link: https://patchwork.freedesktop.org/patch/msgid/[e-mail]

drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c

···
 	/* Check that the context wasn't destroyed before submission */
 	if (likely(!intel_context_is_closed(eb->context))) {
 		attr = eb->gem_context->sched;
-
-		/*
-		 * Boost actual workloads past semaphores!
-		 *
-		 * With semaphores we spin on one engine waiting for another,
-		 * simply to reduce the latency of starting our work when
-		 * the signaler completes. However, if there is any other
-		 * work that we could be doing on this engine instead, that
-		 * is better utilisation and will reduce the overall duration
-		 * of the current work. To avoid PI boosting a semaphore
-		 * far in the distance past over useful work, we keep a history
-		 * of any semaphore use along our dependency chain.
-		 */
-		if (!(rq->sched.flags & I915_SCHED_HAS_SEMAPHORE_CHAIN))
-			attr.priority |= I915_PRIORITY_NOSEMAPHORE;
 	} else {
 		/* Serialise with context_close via the add_to_timeline */
 		i915_request_set_error_once(rq, -ENOENT);

drivers/gpu/drm/i915/gt/intel_lrc.c

···
 	if (i915_request_has_nopreempt(rq))
 		prio = I915_PRIORITY_UNPREEMPTABLE;
 
-	/*
-	 * On unwinding the active request, we give it a priority bump
-	 * if it has completed waiting on any semaphore. If we know that
-	 * the request has already started, we can prevent an unwanted
-	 * preempt-to-idle cycle by taking that into account now.
-	 */
-	if (__i915_request_has_started(rq))
-		prio |= I915_PRIORITY_NOSEMAPHORE;
-
 	return prio;
 }
 

drivers/gpu/drm/i915/gt/selftest_context.c

···
 
 	/* Opencode i915_request_add() so we can keep the timeline locked. */
 	__i915_request_commit(rq);
+	rq->sched.attr.priority = I915_PRIORITY_BARRIER;
 	__i915_request_queue(rq, NULL);
 
 	timeout = i915_request_wait(rq, 0, HZ / 10);

drivers/gpu/drm/i915/i915_priolist_types.h

···
 	I915_PRIORITY_DISPLAY,
 };
 
-#define I915_USER_PRIORITY_SHIFT 1
+#define I915_USER_PRIORITY_SHIFT 0
 #define I915_USER_PRIORITY(x) ((x) << I915_USER_PRIORITY_SHIFT)
 
 #define I915_PRIORITY_COUNT BIT(I915_USER_PRIORITY_SHIFT)
 #define I915_PRIORITY_MASK (I915_PRIORITY_COUNT - 1)
-
-#define I915_PRIORITY_NOSEMAPHORE	((u8)BIT(0))
 
 /* Smallest priority value that cannot be bumped. */
 #define I915_PRIORITY_INVALID (INT_MIN | (u8)I915_PRIORITY_MASK)

drivers/gpu/drm/i915/i915_request.c

···
 	}
 	spin_unlock_irq(&signal->lock);
 
-	/* Copy across semaphore status as we need the same behaviour */
-	rq->sched.flags |= signal->sched.flags;
 	return 0;
 }
 
···
 	spin_unlock(&request->lock);
 
 	/* We've already spun, don't charge on resubmitting. */
-	if (request->sched.semaphores && i915_request_started(request)) {
-		request->sched.attr.priority |= I915_PRIORITY_NOSEMAPHORE;
+	if (request->sched.semaphores && i915_request_started(request))
 		request->sched.semaphores = 0;
-	}
 
 	/*
 	 * We don't need to wake_up any waiters on request->execute, they
···
 	return NOTIFY_DONE;
 }
 
-static void irq_semaphore_cb(struct irq_work *wrk)
-{
-	struct i915_request *rq =
-		container_of(wrk, typeof(*rq), semaphore_work);
-
-	i915_schedule_bump_priority(rq, I915_PRIORITY_NOSEMAPHORE);
-	i915_request_put(rq);
-}
-
 static int __i915_sw_fence_call
 semaphore_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
 {
···
 
 	switch (state) {
 	case FENCE_COMPLETE:
-		if (!(READ_ONCE(rq->sched.attr.priority) & I915_PRIORITY_NOSEMAPHORE)) {
-			i915_request_get(rq);
-			init_irq_work(&rq->semaphore_work, irq_semaphore_cb);
-			irq_work_queue(&rq->semaphore_work);
-		}
 		break;
 
 	case FENCE_FREE:
···
 		    gfp_t gfp)
 {
 	const intel_engine_mask_t mask = READ_ONCE(from->engine)->mask;
+	struct i915_sw_fence *wait = &to->submit;
 
 	if (!intel_context_use_semaphores(to->context))
 		goto await_fence;
···
 		goto await_fence;
 
 	to->sched.semaphores |= mask;
-	to->sched.flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
-	return 0;
+	wait = &to->semaphore;
 
 await_fence:
-	return i915_sw_fence_await_dma_fence(&to->submit,
+	return i915_sw_fence_await_dma_fence(wait,
 					     &from->fence, 0,
 					     I915_FENCE_GFP);
 }
···
 		ret = emit_semaphore_wait(to, from, I915_FENCE_GFP);
 	if (ret < 0)
 		return ret;
-
-	if (to->sched.flags & I915_SCHED_HAS_SEMAPHORE_CHAIN) {
-		ret = i915_sw_fence_await_dma_fence(&to->semaphore,
-						    &from->fence, 0,
-						    I915_FENCE_GFP);
-		if (ret < 0)
-			return ret;
-	}
-
-	if (from->sched.flags & I915_SCHED_HAS_EXTERNAL_CHAIN)
-		to->sched.flags |= I915_SCHED_HAS_EXTERNAL_CHAIN;
 
 	return 0;
 }
···
 	if (ctx)
 		attr = ctx->sched;
 	rcu_read_unlock();
-
-	if (!(rq->sched.flags & I915_SCHED_HAS_SEMAPHORE_CHAIN))
-		attr.priority |= I915_PRIORITY_NOSEMAPHORE;
 
 	__i915_request_queue(rq, &attr);
 

drivers/gpu/drm/i915/i915_request.h

···
 	};
 	struct list_head execute_cb;
 	struct i915_sw_fence semaphore;
-	struct irq_work semaphore_work;
 
 	/*
 	 * A list of everyone we wait upon, and everyone who waits upon us.

drivers/gpu/drm/i915/i915_scheduler.c

···
 	GEM_BUG_ON(rb_first_cached(&execlists->queue) !=
 		   rb_first(&execlists->queue.rb_root));
 
-	last_prio = (INT_MAX >> I915_USER_PRIORITY_SHIFT) + 1;
+	last_prio = INT_MAX;
 	for (rb = rb_first_cached(&execlists->queue); rb; rb = rb_next(rb)) {
 		const struct i915_priolist *p = to_priolist(rb);
 
-		GEM_BUG_ON(p->priority >= last_prio);
+		GEM_BUG_ON(p->priority > last_prio);
 		last_prio = p->priority;
 
 		GEM_BUG_ON(!p->used);
···
 		dep->waiter = node;
 		dep->flags = flags;
 
-		/* Keep track of whether anyone on this chain has a semaphore */
-		if (signal->flags & I915_SCHED_HAS_SEMAPHORE_CHAIN &&
-		    !node_started(signal))
-			node->flags |= I915_SCHED_HAS_SEMAPHORE_CHAIN;
-
 		/* All set, now publish. Beware the lockless walkers. */
 		list_add_rcu(&dep->signal_link, &node->signalers_list);
 		list_add_rcu(&dep->wait_link, &signal->waiters_list);
 
+		/* Propagate the chains */
+		node->flags |= signal->flags;
 		ret = true;
 	}
 

drivers/gpu/drm/i915/i915_scheduler_types.h

···
 	struct list_head link;
 	struct i915_sched_attr attr;
 	unsigned int flags;
-#define I915_SCHED_HAS_SEMAPHORE_CHAIN	BIT(0)
-#define I915_SCHED_HAS_EXTERNAL_CHAIN	BIT(1)
+#define I915_SCHED_HAS_EXTERNAL_CHAIN	BIT(0)
 	intel_engine_mask_t semaphores;
 };