Skip to contentSkip to conversation

drm/amdgpu: stop cp resume when compute ring test failed

PATCH In review Hawking Zhang 4 participants 7 messages Apr 23, 2020

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

PATCH 1/2 drm/amdgpu: stop cp resume when compute ring test failed

driver should stop cp resume once compute
ring test failed

Change-Id I4cd3328f38e0755d0c877484936132d204c9fe50
---
 drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 4 +++-
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 4 +++-
 drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++-
 3 files changed, 9 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +3 -1
index b2f10e3..fcee758 100644
@@ -3132,7 +3132,9 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
 		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
 	}
 
 	return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +3 -1
index 6c56ced..8dc8e90 100644
@@ -4781,7 +4781,9 @@ static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
 		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
 	}
 
 	return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +3 -1
index 09aa5f5..20937059 100644
@@ -3846,7 +3846,9 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
 		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
 	}
 
 	gfx_v9_0_enable_gui_idle_interrupt(adev, true);
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#1 lore raw

PATCH 2/2 drm/amdgpu: drop the unused local variable

local variable kiq won't be used in function
gfx_v8_0_ring_emit_rreg

Change-Id I6229987c8ce43ff0d55e1fae15ede9cb0827f76d
---
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 1 -
 1 file changed, 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +0 -1
index 8dc8e90..9644614 100644
@@ -6395,7 +6395,6 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4));
 	amdgpu_ring_write(ring, 0 |	/* src: register*/
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#2 lore raw

RE: [PATCH 2/2] drm/amdgpu: drop the unused local variable

Hi Hawking

Can you help also remove the same local variable kiq for gfx_v10_0_ring_emit_rreg? Thanks in advance. After that , Reviewed-by: Yintian Tao <[email protected]>

Best Regards
Yintian Tao
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Hawking Zhang
Sent: 2020年4月23日 17:02
To: [email protected]
Cc: Zhang, Hawking <[email protected]>
Subject: [PATCH 2/2] drm/amdgpu: drop the unused local variable

local variable kiq won't be used in function gfx_v8_0_ring_emit_rreg

Change-Id I6229987c8ce43ff0d55e1fae15ede9cb0827f76d
---
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 1 -
 1 file changed, 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +0 -1
index 8dc8e90..9644614 100644
@@ -6395,7 +6395,6 @@ static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4));
 	amdgpu_ring_write(ring, 0 |	/* src: register*/

--
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cyintian.tao%40amd.com%7C3e351ebdd7ff45259e6108d7e76502d0%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637232293332601376&amp;sdata=dH2o%2Bl9%2FjqzP2IIN43qIDrbpmmQpXjPwgGvyaAc%2B1L8%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw
In reply to Hawking Zhang

Am 23.04.20 um 11:01 schrieb Hawking Zhang:

Quoted 2 lines

driver should stop cp resume once compute
ring test failed

Mhm intentionally ignored those errors because the compute rings sometimes doesn't come up again after a GPU reset.

We even have the necessary logic in the SW scheduler to redirect the jobs to another compute ring when one fails to come up again.

Christian.

Quoted 44 lines
Change-Id I4cd3328f38e0755d0c877484936132d204c9fe50
---
  drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 4 +++-
  drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 4 +++-
  drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++-
  3 files changed, 9 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +3 -1
index b2f10e3..fcee758 100644
@@ -3132,7 +3132,9 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
 
  	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
  		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
  	}
 
  	return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +3 -1
index 6c56ced..8dc8e90 100644
@@ -4781,7 +4781,9 @@ static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
 
  	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
  		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
  	}
 
  	return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +0 -0
index 09aa5f5..20937059 100644

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) · line 3854

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +3 -1
@@ -3846,7 +3846,9 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
 
  	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
  		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
  	}
 
  	gfx_v9_0_enable_gui_idle_interrupt(adev, true);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#4 lore raw

[AMD Official Use Only - Internal Distribution Only]

Would you mind to enable this and try it again? The recent gpu reset testing on vega20 looks very positive.

Regards,
Hawking
-----Original Message-----
From: Christian König <[email protected]>
Sent: Thursday, April 23, 2020 20:31
To: Zhang, Hawking <[email protected]>; [email protected]
Subject: Re: [PATCH 1/2] drm/amdgpu: stop cp resume when compute ring test failed

Am 23.04.20 um 11:01 schrieb Hawking Zhang:

Quoted 1 line

driver should stop cp resume once compute ring test failed

Mhm intentionally ignored those errors because the compute rings sometimes doesn't come up again after a GPU reset.

We even have the necessary logic in the SW scheduler to redirect the jobs to another compute ring when one fails to come up again.

Christian.

Quoted 59 lines
Change-Id I4cd3328f38e0755d0c877484936132d204c9fe50
---
  drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 4 +++-
  drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 4 +++-
  drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++-
  3 files changed, 9 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +3 -1
index b2f10e3..fcee758 100644
@@ -3132,7 +3132,9 @@ static int gfx_v7_0_cp_compute_resume(struct 
 amdgpu_device *adev)
 
  	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
  		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
  	}
 

return 0;

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +3 -1
index 6c56ced..8dc8e90 100644
@@ -4781,7 +4781,9 @@ static int gfx_v8_0_cp_test_all_rings(struct 
 amdgpu_device *adev)
 
  	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
  		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
  	}
 

return 0;

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +3 -1
index 09aa5f5..20937059 100644
@@ -3846,7 +3846,9 @@ static int gfx_v9_0_cp_resume(struct 
 amdgpu_device *adev)
 
  	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
  		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
  	}
 
Quoted 1 line
  	gfx_v9_0_enable_gui_idle_interrupt(adev, true);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#5 lore raw

Yeah, we certainly could try this again. But maybe split that up into individual patches for gfx7/8/9.

In other words make it easy to revert if this still doesn't work well on gfx7 or some other generation.

Christian.

Am 23.04.20 um 15:43 schrieb Zhang, Hawking:

Quoted 78 lines

[AMD Official Use Only - Internal Distribution Only]

Would you mind to enable this and try it again? The recent gpu reset testing on vega20 looks very positive.

Regards,
Hawking
-----Original Message-----
From: Christian König <[email protected]>
Sent: Thursday, April 23, 2020 20:31
To: Zhang, Hawking <[email protected]>; [email protected]
Subject: Re: [PATCH 1/2] drm/amdgpu: stop cp resume when compute ring test failed

Am 23.04.20 um 11:01 schrieb Hawking Zhang:

Quoted 1 line

driver should stop cp resume once compute ring test failed

Mhm intentionally ignored those errors because the compute rings sometimes doesn't come up again after a GPU reset.

We even have the necessary logic in the SW scheduler to redirect the jobs to another compute ring when one fails to come up again.

Christian.

Quoted 58 lines
Change-Id I4cd3328f38e0755d0c877484936132d204c9fe50
---
   drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 4 +++-
   drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 4 +++-
   drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++-
   3 files changed, 9 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +3 -1
index b2f10e3..fcee758 100644
@@ -3132,7 +3132,9 @@ static int gfx_v7_0_cp_compute_resume(struct
 amdgpu_device *adev)
 
   	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
   		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
   	}
 
   	return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +3 -1
index 6c56ced..8dc8e90 100644
@@ -4781,7 +4781,9 @@ static int gfx_v8_0_cp_test_all_rings(struct
 amdgpu_device *adev)
 
   	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
   		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
   	}
 
   	return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +3 -1
index 09aa5f5..20937059 100644
@@ -3846,7 +3846,9 @@ static int gfx_v9_0_cp_resume(struct
 amdgpu_device *adev)
 
   	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
   		ring = &adev->gfx.compute_ring[i];
-		amdgpu_ring_test_helper(ring);
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
   	}
 
Quoted 5 lines
Quoted 1 line
   	gfx_v9_0_enable_gui_idle_interrupt(adev, true);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#6 lore raw
Quoted 6 lines · Christian König

Yeah, we certainly could try this again. But maybe split that up into individual patches for gfx7/8/9.

In other words make it easy to revert if this still doesn't work well on gfx7 or some other generation.

Yeah, unless there is a good reason, I don't think we should do this. IIRC, compute rings randomly fail to recover on a lot of hw generations.

Alex

Quoted 88 lines

Christian.

Am 23.04.20 um 15:43 schrieb Zhang, Hawking:

Quoted 78 lines

[AMD Official Use Only - Internal Distribution Only]

Would you mind to enable this and try it again? The recent gpu reset testing on vega20 looks very positive.

Regards,
Hawking
-----Original Message-----
From: Christian König <[email protected]>
Sent: Thursday, April 23, 2020 20:31
To: Zhang, Hawking <[email protected]>; [email protected]
Subject: Re: [PATCH 1/2] drm/amdgpu: stop cp resume when compute ring test failed

Am 23.04.20 um 11:01 schrieb Hawking Zhang:

Quoted 1 line

driver should stop cp resume once compute ring test failed

Mhm intentionally ignored those errors because the compute rings sometimes doesn't come up again after a GPU reset.

We even have the necessary logic in the SW scheduler to redirect the jobs to another compute ring when one fails to come up again.

Christian.

Quoted 58 lines
Change-Id I4cd3328f38e0755d0c877484936132d204c9fe50
---
   drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 4 +++-
   drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 4 +++-
   drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++-
   3 files changed, 9 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +3 -1
index b2f10e3..fcee758 100644
@@ -3132,7 +3132,9 @@ static int gfx_v7_0_cp_compute_resume(struct
 amdgpu_device *adev)
 
     for (i = 0; i < adev->gfx.num_compute_rings; i++) {
             ring = &adev->gfx.compute_ring[i];
-            amdgpu_ring_test_helper(ring);
+            r = amdgpu_ring_test_helper(ring);
+            if (r)
+                    return r;
     }
 
     return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +3 -1
index 6c56ced..8dc8e90 100644
@@ -4781,7 +4781,9 @@ static int gfx_v8_0_cp_test_all_rings(struct
 amdgpu_device *adev)
 
     for (i = 0; i < adev->gfx.num_compute_rings; i++) {
             ring = &adev->gfx.compute_ring[i];
-            amdgpu_ring_test_helper(ring);
+            r = amdgpu_ring_test_helper(ring);
+            if (r)
+                    return r;
     }
 
     return 0;
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +3 -1
index 09aa5f5..20937059 100644
@@ -3846,7 +3846,9 @@ static int gfx_v9_0_cp_resume(struct
 amdgpu_device *adev)
 
     for (i = 0; i < adev->gfx.num_compute_rings; i++) {
             ring = &adev->gfx.compute_ring[i];
-            amdgpu_ring_test_helper(ring);
+            r = amdgpu_ring_test_helper(ring);
+            if (r)
+                    return r;
     }
 
Quoted 5 lines
Quoted 1 line
     gfx_v9_0_enable_gui_idle_interrupt(adev, true);

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Quoted 4 lines

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx