public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/msm/a6xx: Fix recovery vs runpm race
@ 2023-12-18 15:59 Rob Clark
  2023-12-22 19:57 ` Akhil P Oommen
  0 siblings, 1 reply; 3+ messages in thread
From: Rob Clark @ 2023-12-18 15:59 UTC (permalink / raw)
  To: dri-devel
  Cc: freedreno, linux-arm-msm, Rob Clark, David Heidelberg, Rob Clark,
	Abhinav Kumar, Dmitry Baryshkov, Sean Paul, Marijn Suijten,
	David Airlie, Daniel Vetter, Konrad Dybcio, Akhil P Oommen,
	Danylo Piliaiev, Bjorn Andersson, open list

From: Rob Clark <robdclark@chromium.org>

a6xx_recover() is relying on the gpu lock to serialize against incoming
submits doing a runpm get, as it tries to temporarily balance out the
runpm gets with puts in order to power off the GPU.  Unfortunately this
gets worse when we (in a later patch) will move the runpm get out of the
scheduler thread/work to move it out of the fence signaling path.

Instead we can just simplify the whole thing by using force_suspend() /
force_resume() instead of trying to be clever.

Reported-by: David Heidelberg <david.heidelberg@collabora.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10272
Fixes: abe2023b4cea ("drm/msm/gpu: Push gpu lock down past runpm")
Signed-off-by: Rob Clark <robdclark@chromium.org>
---
 drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 12 ++----------
 1 file changed, 2 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
index 268737e59131..a5660d63535b 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
@@ -1244,12 +1244,7 @@ static void a6xx_recover(struct msm_gpu *gpu)
 	dev_pm_genpd_add_notifier(gmu->cxpd, &gmu->pd_nb);
 	dev_pm_genpd_synced_poweroff(gmu->cxpd);
 
-	/* Drop the rpm refcount from active submits */
-	if (active_submits)
-		pm_runtime_put(&gpu->pdev->dev);
-
-	/* And the final one from recover worker */
-	pm_runtime_put_sync(&gpu->pdev->dev);
+	pm_runtime_force_suspend(&gpu->pdev->dev);
 
 	if (!wait_for_completion_timeout(&gmu->pd_gate, msecs_to_jiffies(1000)))
 		DRM_DEV_ERROR(&gpu->pdev->dev, "cx gdsc didn't collapse\n");
@@ -1258,10 +1253,7 @@ static void a6xx_recover(struct msm_gpu *gpu)
 
 	pm_runtime_use_autosuspend(&gpu->pdev->dev);
 
-	if (active_submits)
-		pm_runtime_get(&gpu->pdev->dev);
-
-	pm_runtime_get_sync(&gpu->pdev->dev);
+	pm_runtime_force_resume(&gpu->pdev->dev);
 
 	gpu->active_submits = active_submits;
 	mutex_unlock(&gpu->active_lock);
-- 
2.43.0


^ permalink raw reply related	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2023-12-22 20:15 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-12-18 15:59 [PATCH] drm/msm/a6xx: Fix recovery vs runpm race Rob Clark
2023-12-22 19:57 ` Akhil P Oommen
2023-12-22 20:15   ` Rob Clark

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox