Skip to contentSkip to conversation

drm: msm: a6xx: fix gpu failure after system resume

PATCH In review Akhil P Oommen 4 participants 5 messages Jul 17, 2020

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 drm: msm: a6xx: fix gpu failure after system resume

On targets where GMU is available, GMU takes over the ownership of GX GDSC during its initialization. So, move the refcount-get on GX PD before we initialize the GMU. This ensures that nobody can collapse the GX GDSC once GMU owns the GX GDSC. This patch fixes some GMU OOB errors seen during GPU wake up during a system resume.

---
Changes from v1:
- Reworded the commit text
- Added Reported-by & Tested-by tags

 drivers/gpu/drm/msm/adreno/a6xx_gmu.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/gpu/drm/msm/adreno/a6xx_gmu.c +10 -8
index 21e77d6..1d33020 100644
@@ -854,10 +854,19 @@ int a6xx_gmu_resume(struct a6xx_gpu *a6xx_gpu)
 	/* Turn on the resources */
 	pm_runtime_get_sync(gmu->dev);
 
+	/*
+	 * "enable" the GX power domain which won't actually do anything but it
+	 * will make sure that the refcounting is correct in case we need to
+	 * bring down the GX after a GMU failure
+	 */
+	if (!IS_ERR_OR_NULL(gmu->gxpd))
+		pm_runtime_get_sync(gmu->gxpd);
+
 	/* Use a known rate to bring up the GMU */
 	clk_set_rate(gmu->core_clk, 200000000);
 	ret = clk_bulk_prepare_enable(gmu->nr_clocks, gmu->clocks);
 	if (ret) {
+		pm_runtime_put(gmu->gxpd);
 		pm_runtime_put(gmu->dev);
 		return ret;
 	}
@@ -903,19 +912,12 @@ int a6xx_gmu_resume(struct a6xx_gpu *a6xx_gpu)
 	else
 		a6xx_hfi_set_freq(gmu, gmu->current_perf_index);
 
-	/*
-	 * "enable" the GX power domain which won't actually do anything but it
-	 * will make sure that the refcounting is correct in case we need to
-	 * bring down the GX after a GMU failure
-	 */
-	if (!IS_ERR_OR_NULL(gmu->gxpd))
-		pm_runtime_get(gmu->gxpd);
-
 out:
 	/* On failure, shut down the GMU to leave it in a good state */
 	if (ret) {
 		disable_irq(gmu->gmu_irq);
 		a6xx_rpmh_stop(gmu);
+		pm_runtime_put(gmu->gxpd);
 		pm_runtime_put(gmu->dev);
 	}
 
Signature · 1 line
2.7.4
Reviewed-by on v2 1/1
#1 lore raw
Quoted 9 lines · Akhil P Oommen

On targets where GMU is available, GMU takes over the ownership of GX GDSC during its initialization. So, move the refcount-get on GX PD before we initialize the GMU. This ensures that nobody can collapse the GX GDSC once GMU owns the GX GDSC. This patch fixes some GMU OOB errors seen during GPU wake up during a system resume.

The Signed-off-by needs to be at the end but I think Rob can do that for you.

Quoted 56 lines

---
Changes from v1:
- Reworded the commit text
- Added Reported-by & Tested-by tags

 drivers/gpu/drm/msm/adreno/a6xx_gmu.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/gpu/drm/msm/adreno/a6xx_gmu.c +10 -8
index 21e77d6..1d33020 100644
@@ -854,10 +854,19 @@ int a6xx_gmu_resume(struct a6xx_gpu *a6xx_gpu)
 	/* Turn on the resources */
 	pm_runtime_get_sync(gmu->dev);
 
+	/*
+	 * "enable" the GX power domain which won't actually do anything but it
+	 * will make sure that the refcounting is correct in case we need to
+	 * bring down the GX after a GMU failure
+	 */
+	if (!IS_ERR_OR_NULL(gmu->gxpd))
+		pm_runtime_get_sync(gmu->gxpd);
+
 	/* Use a known rate to bring up the GMU */
 	clk_set_rate(gmu->core_clk, 200000000);
 	ret = clk_bulk_prepare_enable(gmu->nr_clocks, gmu->clocks);
 	if (ret) {
+		pm_runtime_put(gmu->gxpd);
 		pm_runtime_put(gmu->dev);
 		return ret;
 	}
@@ -903,19 +912,12 @@ int a6xx_gmu_resume(struct a6xx_gpu *a6xx_gpu)
 	else
 		a6xx_hfi_set_freq(gmu, gmu->current_perf_index);
 
-	/*
-	 * "enable" the GX power domain which won't actually do anything but it
-	 * will make sure that the refcounting is correct in case we need to
-	 * bring down the GX after a GMU failure
-	 */
-	if (!IS_ERR_OR_NULL(gmu->gxpd))
-		pm_runtime_get(gmu->gxpd);
-
 out:
 	/* On failure, shut down the GMU to leave it in a good state */
 	if (ret) {
 		disable_irq(gmu->gmu_irq);
 		a6xx_rpmh_stop(gmu);
+		pm_runtime_put(gmu->gxpd);
 		pm_runtime_put(gmu->dev);
 	}
 
Signature · 1 line
2.7.4
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
#2 lore raw

Hi,

Quoted 8 lines · Jordan Crouse
Quoted 5 lines · Akhil P Oommen

On targets where GMU is available, GMU takes over the ownership of GX GDSC during its initialization. So, move the refcount-get on GX PD before we initialize the GMU. This ensures that nobody can collapse the GX GDSC once GMU owns the GX GDSC. This patch fixes some GMU OOB errors seen during GPU wake up during a system resume.

Quoted 5 lines
Quoted 3 lines

The Signed-off-by needs to be at the end but I think Rob can do that for you.

It does? I've always been told that this is supposed to be roughly a log of what happens. In that sense you added your SoB before the review/test happened so it should come before. I know some maintainers seem to do things differently but that seems to be the most common. In that sense, I think the order that Akhil has is correct. ...but, obviously, it's up to the maintainer. ;-)
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#3 lore raw
Quoted 16 lines · Doug Anderson

Hi,

Quoted 12 lines · Jordan Crouse
Quoted 9 lines · Akhil P Oommen

On targets where GMU is available, GMU takes over the ownership of GX GDSC during its initialization. So, move the refcount-get on GX PD before we initialize the GMU. This ensures that nobody can collapse the GX GDSC once GMU owns the GX GDSC. This patch fixes some GMU OOB errors seen during GPU wake up during a system resume.

Quoted 8 lines
Quoted 1 line

The Signed-off-by needs to be at the end but I think Rob can do that for you.

It does? I've always been told that this is supposed to be roughly a log of what happens. In that sense you added your SoB before the review/test happened so it should come before. I know some maintainers seem to do things differently but that seems to be the most common. In that sense, I think the order that Akhil has is correct. ...but, obviously, it's up to the maintainer. ;-)

yeah, I chronological order was my understanding too.. but presumably the Reported-by happened before the Signed-of-by (which is how I reordered things when applying the patch)

BR,
-R
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#4 lore raw

Hi,

Quoted 27 lines · Rob Clark
Quoted 16 lines · Doug Anderson

Hi,

Quoted 12 lines · Jordan Crouse
Quoted 9 lines · Akhil P Oommen

On targets where GMU is available, GMU takes over the ownership of GX GDSC during its initialization. So, move the refcount-get on GX PD before we initialize the GMU. This ensures that nobody can collapse the GX GDSC once GMU owns the GX GDSC. This patch fixes some GMU OOB errors seen during GPU wake up during a system resume.

Quoted 8 lines
Quoted 1 line

The Signed-off-by needs to be at the end but I think Rob can do that for you.

It does? I've always been told that this is supposed to be roughly a log of what happens. In that sense you added your SoB before the review/test happened so it should come before. I know some maintainers seem to do things differently but that seems to be the most common. In that sense, I think the order that Akhil has is correct. ...but, obviously, it's up to the maintainer. ;-)

Quoted 3 lines

yeah, I chronological order was my understanding too.. but presumably the Reported-by happened before the Signed-of-by (which is how I reordered things when applying the patch)

Doh! Yeah, I somehow read that as Reviewed-by. Thanks! :-)

-Doug
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel