[CI,1/2] drm/i915: Switch to kernel context before idling at runtime

Commit Message

Chris Wilson May 25, 2018, 10:24 p.m. UTC

We can reduce our exposure to random neutrinos by resting on the kernel
context having flushed out the user contexts to system memory and
beyond. The corollary is that we then we require two passes through the
idle handler to go to sleep, which on a truly idle system involves an
extra pass through the slow and irregular retire work handler.

Signed-off-by: Chris Wilson <[e-mail]>
Reviewed-by: Mika Kuoppala <[e-mail]>
---
 drivers/gpu/drm/i915/i915_debugfs.c |  8 ++++++--
 drivers/gpu/drm/i915/i915_gem.c     | 29 ++++++++++++++++++++++++-----
 2 files changed, 30 insertions(+), 7 deletions(-)

Patch

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index a8e7761cdc7d..594ee65a6c06 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -4226,8 +4226,12 @@  i915_drop_caches_set(void *data, u64 val)
 		i915_gem_shrink_all(dev_priv);
 	fs_reclaim_release(GFP_KERNEL);
 
-	if (val & DROP_IDLE)
-		drain_delayed_work(&dev_priv->gt.idle_work);
+	if (val & DROP_IDLE) {
+		do {
+			flush_delayed_work(&dev_priv->gt.retire_work);
+			drain_delayed_work(&dev_priv->gt.idle_work);
+		} while (READ_ONCE(dev_priv->gt.awake));
+	}
 
 	if (val & DROP_FREED)
 		i915_gem_drain_freed_objects(dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 05f44ca35a06..c93f5dcb1d82 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -139,6 +139,8 @@  int i915_mutex_lock_interruptible(struct drm_device *dev)
 
 static u32 __i915_gem_park(struct drm_i915_private *i915)
 {
+	GEM_TRACE("\n");
+
 	lockdep_assert_held(&i915->drm.struct_mutex);
 	GEM_BUG_ON(i915->gt.active_requests);
 	GEM_BUG_ON(!list_empty(&i915->gt.active_rings));
@@ -193,6 +195,8 @@  void i915_gem_park(struct drm_i915_private *i915)
 
 void i915_gem_unpark(struct drm_i915_private *i915)
 {
+	GEM_TRACE("\n");
+
 	lockdep_assert_held(&i915->drm.struct_mutex);
 	GEM_BUG_ON(!i915->gt.active_requests);
 
@@ -3504,6 +3508,21 @@  i915_gem_idle_work_handler(struct work_struct *work)
 	if (!READ_ONCE(dev_priv->gt.awake))
 		return;
 
+	/*
+	 * Flush out the last user context, leaving only the pinned
+	 * kernel context resident. When we are idling on the kernel_context,
+	 * no more new requests (with a context switch) are emitted and we
+	 * can finally rest. A consequence is that the idle work handler is
+	 * always called at least twice before idling (and if the system is
+	 * idle that implies a round trip through the retire worker).
+	 */
+	mutex_lock(&dev_priv->drm.struct_mutex);
+	i915_gem_switch_to_kernel_context(dev_priv);
+	mutex_unlock(&dev_priv->drm.struct_mutex);
+
+	GEM_TRACE("active_requests=%d (after switch-to-kernel-context)\n",
+		  READ_ONCE(dev_priv->gt.active_requests));
+
 	/*
 	 * Wait for last execlists context complete, but bail out in case a
 	 * new request is submitted. As we don't trust the hardware, we
@@ -4914,11 +4933,9 @@  static void assert_kernel_context_is_current(struct drm_i915_private *i915)
 
 void i915_gem_sanitize(struct drm_i915_private *i915)
 {
-	if (i915_terminally_wedged(&i915->gpu_error)) {
-		mutex_lock(&i915->drm.struct_mutex);
+	mutex_lock(&i915->drm.struct_mutex);
+	if (i915_terminally_wedged(&i915->gpu_error))
 		i915_gem_unset_wedged(i915);
-		mutex_unlock(&i915->drm.struct_mutex);
-	}
 
 	/*
 	 * If we inherit context state from the BIOS or earlier occupants
@@ -4930,6 +4947,9 @@  void i915_gem_sanitize(struct drm_i915_private *i915)
 	 */
 	if (INTEL_GEN(i915) >= 5 && intel_has_gpu_reset(i915))
 		WARN_ON(intel_gpu_reset(i915, ALL_ENGINES));
+
+	i915_gem_contexts_lost(i915);
+	mutex_unlock(&i915->drm.struct_mutex);
 }
 
 int i915_gem_suspend(struct drm_i915_private *dev_priv)
@@ -4965,7 +4985,6 @@  int i915_gem_suspend(struct drm_i915_private *dev_priv)
 
 		assert_kernel_context_is_current(dev_priv);
 	}
-	i915_gem_contexts_lost(dev_priv);
 	mutex_unlock(&dev->struct_mutex);
 
 	intel_uc_suspend(dev_priv);