[Intel-gfx] [PATCH 16/16] drm/i915: Mark the GT as busy before idling the previous request

[Chris Wilson]

In a synchronous setup, we may retire the last request before we
complete allocating the next request. As the last request is retired, we
queue a timer to mark the device as idle, and promptly have to execute
ad cancel that timer once we complete allocating the request and need to
keep the device awake. If we rearrange the mark_busy() to occur before
we retire the previous request, we can skip this ping-pong.

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_request.c | 61 ++++++++++++++++-----------------
 1 file changed, 30 insertions(+), 31 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 9eedd33eb524..87b34153945c 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -265,6 +265,13 @@ static int reserve_seqno(struct intel_engine_cs *engine)
 
 static void unreserve_seqno(struct intel_engine_cs *engine)
 {
+	if (!--engine->i915->gt.active_requests) {
+		GEM_BUG_ON(!engine->i915->gt.awake);
+		mod_delayed_work(engine->i915->wq,
+				 &engine->i915->gt.idle_work,
+				 msecs_to_jiffies(100));
+	}
+
 	GEM_BUG_ON(!engine->timeline->inflight_seqnos);
 	engine->timeline->inflight_seqnos--;
 }
@@ -333,12 +340,6 @@ static void i915_gem_request_retire(struct drm_i915_gem_request *request)
 	list_del_init(&request->link);
 	spin_unlock_irq(&engine->timeline->lock);
 
-	if (!--request->i915->gt.active_requests) {
-		GEM_BUG_ON(!request->i915->gt.awake);
-		mod_delayed_work(request->i915->wq,
-				 &request->i915->gt.idle_work,
-				 msecs_to_jiffies(100));
-	}
 	unreserve_seqno(request->engine);
 	advance_ring(request);
 
@@ -540,6 +541,26 @@ submit_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
 	return NOTIFY_DONE;
 }
 
+static void mark_busy(struct drm_i915_private *i915)
+{
+	if (i915->gt.awake)
+		return;
+
+	GEM_BUG_ON(!i915->gt.active_requests);
+
+	intel_runtime_pm_get_noresume(i915);
+	i915->gt.awake = true;
+
+	intel_enable_gt_powersave(i915);
+	i915_update_gfx_val(i915);
+	if (INTEL_GEN(i915) >= 6)
+		gen6_rps_busy(i915);
+
+	queue_delayed_work(i915->wq,
+			   &i915->gt.retire_work,
+			   round_jiffies_up_relative(HZ));
+}
+
 /**
  * i915_gem_request_alloc - allocate a request structure
  *
@@ -579,6 +600,9 @@ i915_gem_request_alloc(struct intel_engine_cs *engine,
 	if (ret)
 		goto err_unpin;
 
+	if (!dev_priv->gt.active_requests++)
+		mark_busy(dev_priv);
+
 	/* Move the oldest request to the slab-cache (if not in use!) */
 	req = list_first_entry_or_null(&engine->timeline->requests,
 				       typeof(*req), link);
@@ -863,28 +887,6 @@ i915_gem_request_await_object(struct drm_i915_gem_request *to,
 	return ret;
 }
 
-static void i915_gem_mark_busy(const struct intel_engine_cs *engine)
-{
-	struct drm_i915_private *dev_priv = engine->i915;
-
-	if (dev_priv->gt.awake)
-		return;
-
-	GEM_BUG_ON(!dev_priv->gt.active_requests);
-
-	intel_runtime_pm_get_noresume(dev_priv);
-	dev_priv->gt.awake = true;
-
-	intel_enable_gt_powersave(dev_priv);
-	i915_update_gfx_val(dev_priv);
-	if (INTEL_GEN(dev_priv) >= 6)
-		gen6_rps_busy(dev_priv);
-
-	queue_delayed_work(dev_priv->wq,
-			   &dev_priv->gt.retire_work,
-			   round_jiffies_up_relative(HZ));
-}
-
 /*
  * NB: This function is not allowed to fail. Doing so would mean the the
  * request is not being tracked for completion but the work itself is
@@ -966,9 +968,6 @@ void __i915_add_request(struct drm_i915_gem_request *request, bool flush_caches)
 	list_add_tail(&request->ring_link, &ring->request_list);
 	request->emitted_jiffies = jiffies;
 
-	if (!request->i915->gt.active_requests++)
-		i915_gem_mark_busy(engine);
-
 	/* Let the backend know a new request has arrived that may need
 	 * to adjust the existing execution schedule due to a high priority
 	 * request - i.e. we may want to preempt the current request in order
-- 
2.13.3