Skip to contentSkip to conversation

drm/amdgpu: default noretry=0 for navi1x and newer (v2)

PATCH In review Alex Deucher 3 participants 3 messages Nov 30, 2020

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

PATCH 1/1 drm/amdgpu: default noretry=0 for navi1x and newer (v2)

There are no performance advantages to setting it to 1 and it causes stability issues in some cases.

v2: simplify the code

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +0 -5
index fede84509dbc..b07c47cacd28 100644
@@ -422,11 +422,6 @@ void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
 
 	switch (adev->asic_type) {
 	case CHIP_VEGA20:
-	case CHIP_NAVI10:
-	case CHIP_NAVI14:
-	case CHIP_SIENNA_CICHLID:
-	case CHIP_NAVY_FLOUNDER:
-	case CHIP_DIMGREY_CAVEFISH:
 		/*
 		 * noretry = 0 will cause kfd page fault tests fail
 		 * for some ASICs, so set default to 1 for these ASICs.
Signature · 6 lines
2.25.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Reviewed-by on v1 1/1
#1 lore raw

Am 2020-11-30 um 2:25 p.m. schrieb Alex Deucher:

Quoted 2 lines

There are no performance advantages to setting it to 1 and it causes stability issues in some cases.

See my later email on the other thread. There was another consideration besides performance. It had to do with retry fault handling, which depended on being able to reroute the interrupts to the IH1 ring, which the security policy doesn't allow on many boards. With Christian's recent changes to reroute retry interrupts to a software ring, this dependency is gone and it should now be safe to enable retry on most boards.

Quoted 5 lines

v2: simplify the code

Quoted 10 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +0 -0
index fede84509dbc..b07c47cacd28 100644
@@ -422,11 +422,6 @@ void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
 

drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c void amdgpu_gmc_noretry_set(struct amdgpu_device *adev) · line 427

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +0 -5
 	switch (adev->asic_type) {
 	case CHIP_VEGA20:
-	case CHIP_NAVI10:
-	case CHIP_NAVI14:
-	case CHIP_SIENNA_CICHLID:
-	case CHIP_NAVY_FLOUNDER:
-	case CHIP_DIMGREY_CAVEFISH:
 		/*
 		 * noretry = 0 will cause kfd page fault tests fail
 		 * for some ASICs, so set default to 1 for these ASICs.

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

Reviewed-by
#2 lore raw

Am 30.11.20 um 20:52 schrieb Felix Kuehling:

Quoted 9 lines

Am 2020-11-30 um 2:25 p.m. schrieb Alex Deucher:

Quoted 2 lines

There are no performance advantages to setting it to 1 and it causes stability issues in some cases.

See my later email on the other thread. There was another consideration besides performance. It had to do with retry fault handling, which depended on being able to reroute the interrupts to the IH1 ring, which the security policy doesn't allow on many boards. With Christian's recent changes to reroute retry interrupts to a software ring, this dependency is gone and it should now be safe to enable retry on most boards.

It's not the best solution from a performance point of view, but it indeed seems to get things working as expected.

Quoted 7 lines
Quoted 6 lines

v2: simplify the code

Quoted 12 lines
Quoted 12 lines

---
  drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 5 -----
  1 file changed, 5 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +0 -0
index fede84509dbc..b07c47cacd28 100644
@@ -422,11 +422,6 @@ void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
 
Quoted 14 lines

drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c void amdgpu_gmc_noretry_set(struct amdgpu_device *adev) · line 427

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +0 -5
  	switch (adev->asic_type) {
  	case CHIP_VEGA20:
-	case CHIP_NAVI10:
-	case CHIP_NAVI14:
-	case CHIP_SIENNA_CICHLID:
-	case CHIP_NAVY_FLOUNDER:
-	case CHIP_DIMGREY_CAVEFISH:
  		/*
  		 * noretry = 0 will cause kfd page fault tests fail
  		 * for some ASICs, so set default to 1 for these ASICs.

_______________________________________________
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