* [PATCH 1/1] drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode
@ 2020-05-20 22:51 Felix Kuehling
2020-05-21 13:50 ` Christian König
0 siblings, 1 reply; 4+ messages in thread
From: Felix Kuehling @ 2020-05-20 22:51 UTC (permalink / raw)
To: amd-gfx; +Cc: jay.cornwall, christian.koenig
This fixes an intermittent bug where a root PD clear operation still in
progress could overwrite a PDE update done by the CPU, resulting in a
VM fault.
Fixes: 108b4d928c03 ("drm/amd/amdgpu: Update VM function pointer")
Reported-by: Jay Cornwall <Jay.Cornwall@amd.com>
Tested-by: Jay Cornwall <Jay.Cornwall@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 11 +++++++++--
1 file changed, 9 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 414a0b1c2e5a..7417754e9141 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -3000,10 +3000,17 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm,
!amdgpu_gmc_vram_full_visible(&adev->gmc)),
"CPU update of VM recommended only for large BAR system\n");
- if (vm->use_cpu_for_update)
+ if (vm->use_cpu_for_update) {
+ /* Sync with last SDMA update/clear before switching to CPU */
+ r = amdgpu_bo_sync_wait(vm->root.base.bo,
+ AMDGPU_FENCE_OWNER_UNDEFINED, true);
+ if (r)
+ goto free_idr;
+
vm->update_funcs = &amdgpu_vm_cpu_funcs;
- else
+ } else {
vm->update_funcs = &amdgpu_vm_sdma_funcs;
+ }
dma_fence_put(vm->last_update);
vm->last_update = NULL;
vm->is_compute_context = true;
--
2.17.1
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH 1/1] drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode
2020-05-20 22:51 [PATCH 1/1] drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode Felix Kuehling
@ 2020-05-21 13:50 ` Christian König
2020-05-21 17:06 ` Felix Kuehling
0 siblings, 1 reply; 4+ messages in thread
From: Christian König @ 2020-05-21 13:50 UTC (permalink / raw)
To: Felix Kuehling, amd-gfx; +Cc: jay.cornwall, christian.koenig
Am 21.05.20 um 00:51 schrieb Felix Kuehling:
> This fixes an intermittent bug where a root PD clear operation still in
> progress could overwrite a PDE update done by the CPU, resulting in a
> VM fault.
Mhm, maybe better add this to amdgpu_vm_cpu_prepare().
This way we could (in theory) switch between CPU and SDMA based updates
on the fly elsewhere as well.
Christian.
>
> Fixes: 108b4d928c03 ("drm/amd/amdgpu: Update VM function pointer")
> Reported-by: Jay Cornwall <Jay.Cornwall@amd.com>
> Tested-by: Jay Cornwall <Jay.Cornwall@amd.com>
> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 11 +++++++++--
> 1 file changed, 9 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index 414a0b1c2e5a..7417754e9141 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -3000,10 +3000,17 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm,
> !amdgpu_gmc_vram_full_visible(&adev->gmc)),
> "CPU update of VM recommended only for large BAR system\n");
>
> - if (vm->use_cpu_for_update)
> + if (vm->use_cpu_for_update) {
> + /* Sync with last SDMA update/clear before switching to CPU */
> + r = amdgpu_bo_sync_wait(vm->root.base.bo,
> + AMDGPU_FENCE_OWNER_UNDEFINED, true);
> + if (r)
> + goto free_idr;
> +
> vm->update_funcs = &amdgpu_vm_cpu_funcs;
> - else
> + } else {
> vm->update_funcs = &amdgpu_vm_sdma_funcs;
> + }
> dma_fence_put(vm->last_update);
> vm->last_update = NULL;
> vm->is_compute_context = true;
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH 1/1] drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode
2020-05-21 13:50 ` Christian König
@ 2020-05-21 17:06 ` Felix Kuehling
2020-05-22 8:59 ` Christian König
0 siblings, 1 reply; 4+ messages in thread
From: Felix Kuehling @ 2020-05-21 17:06 UTC (permalink / raw)
To: christian.koenig, amd-gfx; +Cc: jay.cornwall
Am 2020-05-21 um 9:50 a.m. schrieb Christian König:
> Am 21.05.20 um 00:51 schrieb Felix Kuehling:
>> This fixes an intermittent bug where a root PD clear operation still in
>> progress could overwrite a PDE update done by the CPU, resulting in a
>> VM fault.
>
> Mhm, maybe better add this to amdgpu_vm_cpu_prepare().
>
> This way we could (in theory) switch between CPU and SDMA based
> updates on the fly elsewhere as well.
That won't work. I want to wait for FENCE_OWNER_VM fences, so I need to
use FENCE_OWNER_UNDEFINED. But then I would also end up waiting for
FENCE_OWNER_KFD eviction fences, which would trigger unwanted evictions.
This works OK in amdgpu_vm_make_compute because it runs before the
eviction fence is attached to the VM.
Regards,
Felix
>
> Christian.
>
>>
>> Fixes: 108b4d928c03 ("drm/amd/amdgpu: Update VM function pointer")
>> Reported-by: Jay Cornwall <Jay.Cornwall@amd.com>
>> Tested-by: Jay Cornwall <Jay.Cornwall@amd.com>
>> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
>> ---
>> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 11 +++++++++--
>> 1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> index 414a0b1c2e5a..7417754e9141 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>> @@ -3000,10 +3000,17 @@ int amdgpu_vm_make_compute(struct
>> amdgpu_device *adev, struct amdgpu_vm *vm,
>> !amdgpu_gmc_vram_full_visible(&adev->gmc)),
>> "CPU update of VM recommended only for large BAR system\n");
>> - if (vm->use_cpu_for_update)
>> + if (vm->use_cpu_for_update) {
>> + /* Sync with last SDMA update/clear before switching to CPU */
>> + r = amdgpu_bo_sync_wait(vm->root.base.bo,
>> + AMDGPU_FENCE_OWNER_UNDEFINED, true);
>> + if (r)
>> + goto free_idr;
>> +
>> vm->update_funcs = &amdgpu_vm_cpu_funcs;
>> - else
>> + } else {
>> vm->update_funcs = &amdgpu_vm_sdma_funcs;
>> + }
>> dma_fence_put(vm->last_update);
>> vm->last_update = NULL;
>> vm->is_compute_context = true;
>
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH 1/1] drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode
2020-05-21 17:06 ` Felix Kuehling
@ 2020-05-22 8:59 ` Christian König
0 siblings, 0 replies; 4+ messages in thread
From: Christian König @ 2020-05-22 8:59 UTC (permalink / raw)
To: Felix Kuehling, amd-gfx; +Cc: jay.cornwall
Am 21.05.20 um 19:06 schrieb Felix Kuehling:
> Am 2020-05-21 um 9:50 a.m. schrieb Christian König:
>> Am 21.05.20 um 00:51 schrieb Felix Kuehling:
>>> This fixes an intermittent bug where a root PD clear operation still in
>>> progress could overwrite a PDE update done by the CPU, resulting in a
>>> VM fault.
>> Mhm, maybe better add this to amdgpu_vm_cpu_prepare().
>>
>> This way we could (in theory) switch between CPU and SDMA based
>> updates on the fly elsewhere as well.
> That won't work. I want to wait for FENCE_OWNER_VM fences, so I need to
> use FENCE_OWNER_UNDEFINED. But then I would also end up waiting for
> FENCE_OWNER_KFD eviction fences, which would trigger unwanted evictions.
>
> This works OK in amdgpu_vm_make_compute because it runs before the
> eviction fence is attached to the VM.
Ok, in this case the patch is Reviewed-by: Christian König
<christian.koenig@amd.com>.
>
> Regards,
> Felix
>
>
>> Christian.
>>
>>> Fixes: 108b4d928c03 ("drm/amd/amdgpu: Update VM function pointer")
>>> Reported-by: Jay Cornwall <Jay.Cornwall@amd.com>
>>> Tested-by: Jay Cornwall <Jay.Cornwall@amd.com>
>>> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
>>> ---
>>> drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 11 +++++++++--
>>> 1 file changed, 9 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> index 414a0b1c2e5a..7417754e9141 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
>>> @@ -3000,10 +3000,17 @@ int amdgpu_vm_make_compute(struct
>>> amdgpu_device *adev, struct amdgpu_vm *vm,
>>> !amdgpu_gmc_vram_full_visible(&adev->gmc)),
>>> "CPU update of VM recommended only for large BAR system\n");
>>> - if (vm->use_cpu_for_update)
>>> + if (vm->use_cpu_for_update) {
>>> + /* Sync with last SDMA update/clear before switching to CPU */
>>> + r = amdgpu_bo_sync_wait(vm->root.base.bo,
>>> + AMDGPU_FENCE_OWNER_UNDEFINED, true);
>>> + if (r)
>>> + goto free_idr;
>>> +
>>> vm->update_funcs = &amdgpu_vm_cpu_funcs;
>>> - else
>>> + } else {
>>> vm->update_funcs = &amdgpu_vm_sdma_funcs;
>>> + }
>>> dma_fence_put(vm->last_update);
>>> vm->last_update = NULL;
>>> vm->is_compute_context = true;
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2020-05-22 9:00 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-05-20 22:51 [PATCH 1/1] drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode Felix Kuehling
2020-05-21 13:50 ` Christian König
2020-05-21 17:06 ` Felix Kuehling
2020-05-22 8:59 ` Christian König
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox