Skip to contentSkip to conversation

Revert "drm/amdgpu: use GMC v9 KIQ workaround only for the GFXHUB"

PATCH In review Chengming Gui 4 participants 4 messages Nov 16, 2018 – Jun 6, 2019 2 messages not in the archive

Subsystems (2)
DRM drivers Radeon and amdgpu DRM drivers
lore
Share & tools
lore

Chengming Gui

author
#0 lore raw

PATCH 1/1 Revert "drm/amdgpu: use GMC v9 KIQ workaround only for the GFXHUB"

With GFXOFF enabled, this patch will cause PCO amdgpu_test failed, but GFXOFF is necessary for PCO, so revert the patch.

This reverts commit b83761bb0b09ec11c924afe9d88e458cb16a0372.

Signed-off-by Jack Gui <[email protected]>
---
 drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +3 -3
index 811231e..14ca4d8 100644
@@ -338,9 +338,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev,
 		struct amdgpu_vmhub *hub = &adev->vmhub[i];
 		u32 tmp = gmc_v9_0_get_invalidate_req(vmid, flush_type);
 
-		if (i == AMDGPU_GFXHUB && !adev->in_gpu_reset &&
-		    adev->gfx.kiq.ring.sched.ready &&
-		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		if (adev->gfx.kiq.ring.sched.ready &&
+		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
+		    !adev->in_gpu_reset) {
 			uint32_t req = hub->vm_inv_eng0_req + eng;
 			uint32_t ack = hub->vm_inv_eng0_ack + eng;
Signature · 6 lines
2.7.4

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

A message that is not in the archive (sent off-list or not yet received): 1542356446-29149-1-git-send-email-Jack.Gui-5C7GfCeVMHo@public.gmane.org

Reviewed-by Huang Rui
#2 lore raw
Quoted 9 lines

-----Original Message-----
From: amd-gfx [mailto:[email protected]] On Behalf Of Chengming Gui
Sent: Friday, November 16, 2018 4:21 PM
To: [email protected]
Cc: Gui, Jack <[email protected]>
Subject: [PATCH revert] Revert "drm/amdgpu: use GMC v9 KIQ workaround only for the GFXHUB"

Quoted 5 lines

With GFXOFF enabled, this patch will cause PCO amdgpu_test failed, but GFXOFF is necessary for PCO, so revert the patch.

This reverts commit b83761bb0b09ec11c924afe9d88e458cb16a0372.

Signed-off-by Jack Gui <[email protected]>
Reviewed-by Huang Rui <[email protected]>
Quoted 28 lines
---
 drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +2 -3
index 811231e..14ca4d8 100644
@@ -338,9 +338,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct
 amdgpu_device *adev,
 		struct amdgpu_vmhub *hub = &adev->vmhub[i];
 		u32 tmp = gmc_v9_0_get_invalidate_req(vmid, flush_type);
 
-		if (i == AMDGPU_GFXHUB && !adev->in_gpu_reset &&
-		    adev->gfx.kiq.ring.sched.ready &&
-		    (amdgpu_sriov_runtime(adev)
 || !amdgpu_sriov_vf(adev))) {
+		if (adev->gfx.kiq.ring.sched.ready &&
+		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))

&&

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +1 -1
+		    !adev->in_gpu_reset) {
 			uint32_t req = hub->vm_inv_eng0_req + eng;
 			uint32_t ack = hub->vm_inv_eng0_ack + eng;
 
--

2.7.4

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

A message that is not in the archive (sent off-list or not yet received): BY2PR12MB0040C512097EA4B6EA662658ECDD0-K//h7OWB4q7Zvl48JdS6+wdYzm3356FpvxpqHgZTriW3zl9H0oFU5g@public.gmane.org

#4 lore raw

Am 16.11.18 um 09:23 schrieb Huang, Ray:

Quoted 9 lines
Quoted 9 lines

-----Original Message-----
From: amd-gfx [mailto:[email protected]] On Behalf Of Chengming Gui
Sent: Friday, November 16, 2018 4:21 PM
To: [email protected]
Cc: Gui, Jack <[email protected]>
Subject: [PATCH revert] Revert "drm/amdgpu: use GMC v9 KIQ workaround only for the GFXHUB"

Quoted 6 lines
Quoted 5 lines

With GFXOFF enabled, this patch will cause PCO amdgpu_test failed, but GFXOFF is necessary for PCO, so revert the patch.

This reverts commit b83761bb0b09ec11c924afe9d88e458cb16a0372.

Signed-off-by Jack Gui <[email protected]>
Reviewed-by Huang Rui <[email protected]>

Please add a comment like "/* This is necessary for a HW workaround under SRIOV as well as GFXOFF under bare metal */".

With that done the patch is Reviewed-by: Christian König <[email protected]> as well.

Thanks,
Christian.

Quoted 29 lines
Quoted 29 lines
---
  drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 6 +++---
  1 file changed, 3 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +2 -3
index 811231e..14ca4d8 100644
@@ -338,9 +338,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct
 amdgpu_device *adev,
  		struct amdgpu_vmhub *hub = &adev->vmhub[i];
  		u32 tmp = gmc_v9_0_get_invalidate_req(vmid, flush_type);
 
-		if (i == AMDGPU_GFXHUB && !adev->in_gpu_reset &&
-		    adev->gfx.kiq.ring.sched.ready &&
-		    (amdgpu_sriov_runtime(adev)
 || !amdgpu_sriov_vf(adev))) {
+		if (adev->gfx.kiq.ring.sched.ready &&
+		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))

&&

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +1 -1
+		    !adev->in_gpu_reset) {
  			uint32_t req = hub->vm_inv_eng0_req + eng;
  			uint32_t ack = hub->vm_inv_eng0_ack + eng;
 
--

2.7.4

Quoted 8 lines
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

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

#5 lore raw
In reply to a message not in the archive

Not sure whether it is related to this change, when I boot a system with amdgpu blacklisted, then modprobe amdgpu, I followed issue. Failing path is gmc_v9_0_flush_gpu_tlb calling amdgpu_virt_kiq_req_write_reg_wait. If I boot w/o amdgpu blacklisted, then it is fine (load amdgpu directly during boot).

[   94.778231] amdgpu 0000:08:00.0: ring uvd_enc_0.1 uses VM inv eng 6 on hub 1
[   94.778234] amdgpu 0000:08:00.0: ring vce0 uses VM inv eng 7 on hub 1
[   94.778236] amdgpu 0000:08:00.0: ring vce1 uses VM inv eng 8 on hub 1
[   94.778239] amdgpu 0000:08:00.0: ring vce2 uses VM inv eng 9 on hub 1
[   94.778242] [drm] ECC is not present.
[   94.778245] [drm] SRAM ECC is not present.
[   94.779441] [drm] Initialized amdgpu 3.32.0 20150101 for 0000:08:00.0 on
minor 0
[   96.307042] rfkill: input handler enabled
[   97.101621] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101647] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101653] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101663] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101668] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101673] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013D
[   97.101681] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101686] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101691] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101699] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101704] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101708] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101717] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101722] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101726] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101735] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101740] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101744] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101753] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101757] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101762] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101770] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101775] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101779] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101788] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101793] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101797] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   97.101806] amdgpu 0000:08:00.0: [mmhub] no-retry page fault (src_id:0
ring:158 vmid:0 pasid:0, for process  pid 0 thread  pid 0)
[   97.101810] amdgpu 0000:08:00.0:   in page starting at address
0x0000000000570000 from 18
[   97.101815] amdgpu 0000:08:00.0: VM_L2_PROTECTION_FAULT_STATUS:0x0000013C
[   98.711247] failed to write reg 28b4 wait reg 28c6
[  100.315395] failed to write reg 1a6f4 wait reg 1a706
[  101.943593] failed to write reg 28b4 wait reg 28c6

Regards,
Oak

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Chengming Gui
Sent: Friday, November 16, 2018 3:21 AM
To: [email protected]
Cc: Gui, Jack <[email protected]>
Subject: [PATCH revert] Revert "drm/amdgpu: use GMC v9 KIQ workaround only for the GFXHUB"

With GFXOFF enabled, this patch will cause PCO amdgpu_test failed, but GFXOFF is necessary for PCO, so revert the patch.

This reverts commit b83761bb0b09ec11c924afe9d88e458cb16a0372.

Signed-off-by Jack Gui <[email protected]>
---
 drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +3 -3
index 811231e..14ca4d8 100644
@@ -338,9 +338,9 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev,
 		struct amdgpu_vmhub *hub = &adev->vmhub[i];
 		u32 tmp = gmc_v9_0_get_invalidate_req(vmid, flush_type);
 
-		if (i == AMDGPU_GFXHUB && !adev->in_gpu_reset &&
-		    adev->gfx.kiq.ring.sched.ready &&
-		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		if (adev->gfx.kiq.ring.sched.ready &&
+		    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
+		    !adev->in_gpu_reset) {
 			uint32_t req = hub->vm_inv_eng0_req + eng;
 			uint32_t ack = hub->vm_inv_eng0_ack + eng;

--
2.7.4

_______________________________________________
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