Skip to contentSkip to conversation

drm/amdgpu: skip mes self test for gc 11.0.3 in recover

PATCH In review YuBiao Wang 2 participants 5 messages Oct 19, 2022 – Oct 20, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu: skip mes self test for gc 11.0.3 in recover

Temporary disable mes self teset for gc 11.0.3 during gpu_recovery.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1 -1
index e0445e8cc342..5b8362727226 100644
@@ -5381,7 +5381,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
 		}
 
-		if (adev->enable_mes)
+		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))
 			amdgpu_mes_self_test(tmp_adev);
 
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled) {
Signature · 1 line
2.25.1
#1 lore raw
Quoted 2 lines · YuBiao Wang

Temporary disable mes self teset for gc 11.0.3 during gpu_recovery.

Is this "temporary" as in "we'll revert this commit later", or is it "temporary" as in the code execution itself?

Quoted 10 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
index e0445e8cc342..5b8362727226 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c int amdgpu_device_gpu_recover(struct amdgpu_device *adev, · line 5385

Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1 -1
@@ -5381,7 +5381,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
 		}
 
-		if (adev->enable_mes)
+		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))
 			amdgpu_mes_self_test(tmp_adev);

Is this just for this version of the IP or this and any newer versions?

Regards,
Luben

Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled) {
#2 lore raw

Hi Luben,

As far as I know of this is only for gc 11.0.3. Mes self test is also skipped in mes late init for this version of IP.

Best Regards,
Yubiao Wang

-----Original Message-----
From: Tuikov, Luben <[email protected]>
Sent: Wednesday, October 19, 2022 11:53 AM
To: Wang, YuBiao <[email protected]>; [email protected]
Cc: Grodzovsky, Andrey <[email protected]>; Quan, Evan <[email protected]>; Chen, Horace <[email protected]>; Koenig, Christian <[email protected]>; Deucher, Alexander <[email protected]>; Xiao, Jack <[email protected]>; Zhang, Hawking <[email protected]>; Liu, Monk <[email protected]>; Xu, Feifei <[email protected]>; Wang, Yang(Kevin) <[email protected]>
Subject: Re: [PATCH] drm/amdgpu: skip mes self test for gc 11.0.3 in recover

Quoted 2 lines · YuBiao Wang

Temporary disable mes self teset for gc 11.0.3 during gpu_recovery.

Is this "temporary" as in "we'll revert this commit later", or is it "temporary" as in the code execution itself?

Quoted 11 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
index e0445e8cc342..5b8362727226 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c int amdgpu_device_gpu_recover(struct amdgpu_device *adev, · line 5386

Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +2 -1
@@ -5381,7 +5381,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
 		}
 
-		if (adev->enable_mes)
+		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != 
+IP_VERSION(11, 0, 3))
 			amdgpu_mes_self_test(tmp_adev);

Is this just for this version of the IP or this and any newer versions?

Regards,
Luben

Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && 

!job_signaled) {

#3 lore raw
In reply to Luben Tuikov

Hi Luben,

Quoted 2 lines
Quoted 2 lines

Is this "temporary" as in "we'll revert this commit later", or is it "temporary" as in the code execution itself? Is this just for this version of the IP or this and any newer versions?

I suppose that it is meant to be reverted later. There is a similar patch in commit c25a7a8bf19a98578ad27aaaa78082276ea1557c which also temporarily skip mes self test only for gc_11.0.3 during mes late init, which was reviewed by @Zhang, Hawking. My patch is to also skip mes self test during gpu recover since self test will also cause failure during reset.

Best Regards,
Yubiao Wang

-----Original Message-----
From: Tuikov, Luben <[email protected]>
Sent: Wednesday, October 19, 2022 11:53 AM
To: Wang, YuBiao <[email protected]>; [email protected]
Cc: Grodzovsky, Andrey <[email protected]>; Quan, Evan <[email protected]>; Chen, Horace <[email protected]>; Koenig, Christian <[email protected]>; Deucher, Alexander <[email protected]>; Xiao, Jack <[email protected]>; Zhang, Hawking <[email protected]>; Liu, Monk <[email protected]>; Xu, Feifei <[email protected]>; Wang, Yang(Kevin) <[email protected]>
Subject: Re: [PATCH] drm/amdgpu: skip mes self test for gc 11.0.3 in recover

Quoted 2 lines · YuBiao Wang

Temporary disable mes self teset for gc 11.0.3 during gpu_recovery.

Is this "temporary" as in "we'll revert this commit later", or is it "temporary" as in the code execution itself?

Quoted 11 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
index e0445e8cc342..5b8362727226 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c int amdgpu_device_gpu_recover(struct amdgpu_device *adev, · line 5386

Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +2 -1
@@ -5381,7 +5381,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
 		}
 
-		if (adev->enable_mes)
+		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != 
+IP_VERSION(11, 0, 3))
 			amdgpu_mes_self_test(tmp_adev);

Is this just for this version of the IP or this and any newer versions?

Regards,
Luben

Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && 

!job_signaled) {

Acked-by
#4 lore raw

Hi YuBiao,

Ah, okay, there's a precedent for such a change then.

Regards,
Luben

Quoted 50 lines · YuBiao

Hi Luben,

Quoted 2 lines
Quoted 2 lines

Is this "temporary" as in "we'll revert this commit later", or is it "temporary" as in the code execution itself? Is this just for this version of the IP or this and any newer versions?

I suppose that it is meant to be reverted later. There is a similar patch in commit c25a7a8bf19a98578ad27aaaa78082276ea1557c which also temporarily skip mes self test only for gc_11.0.3 during mes late init, which was reviewed by @Zhang, Hawking. My patch is to also skip mes self test during gpu recover since self test will also cause failure during reset.

Best Regards,
Yubiao Wang

-----Original Message-----
From: Tuikov, Luben <[email protected]>
Sent: Wednesday, October 19, 2022 11:53 AM
To: Wang, YuBiao <[email protected]>; [email protected]
Cc: Grodzovsky, Andrey <[email protected]>; Quan, Evan <[email protected]>; Chen, Horace <[email protected]>; Koenig, Christian <[email protected]>; Deucher, Alexander <[email protected]>; Xiao, Jack <[email protected]>; Zhang, Hawking <[email protected]>; Liu, Monk <[email protected]>; Xu, Feifei <[email protected]>; Wang, Yang(Kevin) <[email protected]>
Subject: Re: [PATCH] drm/amdgpu: skip mes self test for gc 11.0.3 in recover

Quoted 2 lines · YuBiao Wang

Temporary disable mes self teset for gc 11.0.3 during gpu_recovery.

Is this "temporary" as in "we'll revert this commit later", or is it "temporary" as in the code execution itself?

Quoted 11 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
index e0445e8cc342..5b8362727226 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c int amdgpu_device_gpu_recover(struct amdgpu_device *adev, · line 5386

Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +2 -1
@@ -5381,7 +5381,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
 		}
 
-		if (adev->enable_mes)
+		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != 
+IP_VERSION(11, 0, 3))
 			amdgpu_mes_self_test(tmp_adev);

Is this just for this version of the IP or this and any newer versions?

Regards,
Luben

Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +0 -0
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && 

!job_signaled) {