[PATCH v3 3/7] drm/admgpu: Serialize RAS recovery work directly into reset domain queue.

[Luben Tuikov]

On 2022-05-30 03:49, Christian König wrote:
> Am 25.05.22 um 21:04 schrieb Andrey Grodzovsky:
>> Save the extra usless work schedule.
>>
>> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky at amd.com>
> 
> Acked-by: Christian König <christian.koenig at amd.com>
> 
> Maybe Luben want to take a look as well, he has done some RAS review in 
> the past.

Looks good to me. I understand this is a change to sync up with the other changes
Andrey has been doing to GPU recovery.

We'll need to test this "in the wild", and thoroughly.

Acked-by: Luben Tuikov <luben.tuikov at amd.com>

Regards,
Luben

> 
> Thanks,
> Christian.
> 
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 6 ++++--
>>   1 file changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
>> index 31207f7eec02..a439c04223b5 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
>> @@ -35,6 +35,8 @@
>>   #include "amdgpu_xgmi.h"
>>   #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h"
>>   #include "atom.h"
>> +#include "amdgpu_reset.h"
>> +
>>   #ifdef CONFIG_X86_MCE_AMD
>>   #include <asm/mce.h>
>>   
>> @@ -1920,7 +1922,7 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
>>   	}
>>   
>>   	if (amdgpu_device_should_recover_gpu(ras->adev))
>> -		amdgpu_device_gpu_recover(ras->adev, NULL);
>> +		amdgpu_device_gpu_recover_imp(ras->adev, NULL);
>>   	atomic_set(&ras->in_recovery, 0);
>>   }
>>   
>> @@ -2928,7 +2930,7 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev)
>>   	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
>>   
>>   	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0)
>> -		schedule_work(&ras->recovery_work);
>> +		amdgpu_reset_domain_schedule(ras->adev->reset_domain, &ras->recovery_work);
>>   	return 0;
>>   }
>>   
>