[Intel-gfx] [PATCH 02/10] drm/i915: Repeat flush of idle work during suspend

[Chris Wilson]

The idle work handler is self-arming - if it detects that it needs to
run again it will queue itself from its work handler. Take greater care
when trying to drain the idle work, and double check that it is flushed.

The free worker has a similar issue where it is armed by an RCU task
which may be running concurrently with us.

This should hopefully help with the sporadic WARN_ON(dev_priv->gt.awake)
from i915_gem_suspend.

v2: Reuse drain_freed_objects.
v3: Don't try to flush the freed objects from the shrinker, as it may be
underneath the struct_mutex already.
v4: do while and comment upon the excess rcu_barrier in drain_freed_objects

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c |  2 +-
 drivers/gpu/drm/i915/i915_drv.c     |  3 +--
 drivers/gpu/drm/i915/i915_drv.h     | 13 +++++++++++++
 drivers/gpu/drm/i915/i915_gem.c     | 10 ++++++++--
 4 files changed, 23 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index a5552a1ca01f..655b6719c3da 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -4124,7 +4124,7 @@ i915_drop_caches_set(void *data, u64 val)
 
 	if (val & DROP_FREED) {
 		synchronize_rcu();
-		flush_work(&dev_priv->mm.free_work);
+		i915_gem_drain_freed_objects(dev_priv);
 	}
 
 	return ret;
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 6428588518aa..2c020eafada6 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -545,8 +545,7 @@ static void i915_gem_fini(struct drm_i915_private *dev_priv)
 	i915_gem_context_fini(dev_priv);
 	mutex_unlock(&dev_priv->drm.struct_mutex);
 
-	rcu_barrier();
-	flush_work(&dev_priv->mm.free_work);
+	i915_gem_drain_freed_objects(dev_priv);
 
 	WARN_ON(!list_empty(&dev_priv->context_list));
 }
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 1a914095efdd..3794b0933531 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3208,6 +3208,19 @@ i915_gem_object_create_from_data(struct drm_i915_private *dev_priv,
 void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file);
 void i915_gem_free_object(struct drm_gem_object *obj);
 
+static inline void i915_gem_drain_freed_objects(struct drm_i915_private *i915)
+{
+	/* A single pass should suffice to release all the freed objects (along
+	 * most call paths) , but be a little more paranoid in that freeing
+	 * the objects does take a little amount of time, during which the rcu
+	 * callbacks could have added new objects into the freed list, and
+	 * armed the work again.
+	 */
+	do {
+		rcu_barrier();
+	} while (flush_work(&i915->mm.free_work));
+}
+
 struct i915_vma * __must_check
 i915_gem_object_ggtt_pin(struct drm_i915_gem_object *obj,
 			 const struct i915_ggtt_view *view,
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 5a08e8c187e8..bdb37bd3813b 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4263,8 +4263,14 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv)
 
 	cancel_delayed_work_sync(&dev_priv->gpu_error.hangcheck_work);
 	cancel_delayed_work_sync(&dev_priv->gt.retire_work);
-	flush_delayed_work(&dev_priv->gt.idle_work);
-	flush_work(&dev_priv->mm.free_work);
+
+	/* As the idle_work is rearming if it detects a race, play safe and
+	 * repeat the flush until it is definitely idle.
+	 */
+	while (flush_delayed_work(&dev_priv->gt.idle_work))
+		;
+
+	i915_gem_drain_freed_objects(dev_priv);
 
 	/* Assert that we sucessfully flushed all the work and
 	 * reset the GPU back to its idle, low power state.
-- 
2.11.0