[Intel-gfx] [PATCH 1/6] drm/i915: Do request retirement before marking engines as wedged

[Chris Wilson]

As we declare an engine as wedged, we mark all of its active requests as
in error. However, we don't want to mark successfully completed requests
as in error, which requires us to retire those requests first.

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_gem.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 2709be922512..cb61cec26db5 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2998,10 +2998,15 @@ void i915_gem_set_wedged(struct drm_i915_private *dev_priv)
 	lockdep_assert_held(&dev_priv->drm.struct_mutex);
 	set_bit(I915_WEDGED, &dev_priv->gpu_error.flags);
 
+	/* Retire completed requests first so the list of inflight/incomplete
+	 * requests is accurate and we don't try and mark successful requests
+	 * as in error during __i915_gem_set_wedged_BKL().
+	 */
+	i915_gem_retire_requests(dev_priv);
+
 	stop_machine(__i915_gem_set_wedged_BKL, dev_priv, NULL);
 
 	i915_gem_context_lost(dev_priv);
-	i915_gem_retire_requests(dev_priv);
 
 	mod_delayed_work(dev_priv->wq, &dev_priv->gt.idle_work, 0);
 }
-- 
2.11.0