Re: [PATCH 3/4] drm/sched: always keep selecetd ring sched list in ctx entity

["Christian König" <ckoenig.leichtzumerken@gmail.com>]

Hi James,

please use to_amdgpu_ring(entity->rq->sched) for this.

That's the scheduler which was actually picked for this entity.

Regards,
Christian.

Am 08.09.22 um 15:19 schrieb James Zhu:
> Hi Christian
>
> I need use entity->sched_list to track ring (ring = 
> container_of(sched, struct amdgpu_ring, sched))
>
> during amdgpu_ctx_fini_entity.
>
> I think change here to keep selected ring sched list in 
> entity->sched_list won't change the original logic too much.
>
> Best Regards!
>
> James
>
>
> On 2022-09-08 2:15 a.m., Christian König wrote:
>> Am 07.09.22 um 22:57 schrieb James Zhu:
>>> Always keep selecetd ring sched list in ctx entity.
>>
>> I have no idea what you are doing here, but this certainly doesn't 
>> make sense.
>>
>> Please explain a bit more.
>>
>> Thanks,
>> Christian.
>>
>>>
>>> Signed-off-by: James Zhu <James.Zhu@amd.com>
>>> ---
>>>   drivers/gpu/drm/scheduler/sched_entity.c | 7 ++-----
>>>   1 file changed, 2 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c 
>>> b/drivers/gpu/drm/scheduler/sched_entity.c
>>> index f5595607995b..39dca9cb8e0d 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_entity.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
>>> @@ -71,7 +71,7 @@ int drm_sched_entity_init(struct drm_sched_entity 
>>> *entity,
>>>       entity->guilty = guilty;
>>>       entity->num_sched_list = num_sched_list;
>>>       entity->priority = priority;
>>> -    entity->sched_list = num_sched_list > 1 ? sched_list : NULL;
>>> +    entity->sched_list = sched_list;
>>>       entity->last_scheduled = NULL;
>>>         if(num_sched_list)
>>> @@ -453,7 +453,7 @@ void drm_sched_entity_select_rq(struct 
>>> drm_sched_entity *entity)
>>>       struct drm_sched_rq *rq;
>>>         /* single possible engine and already selected */
>>> -    if (!entity->sched_list)
>>> +    if (entity->num_sched_list <= 1)
>>>           return;
>>>         /* queue non-empty, stay on the same engine */
>>> @@ -482,9 +482,6 @@ void drm_sched_entity_select_rq(struct 
>>> drm_sched_entity *entity)
>>>           entity->rq = rq;
>>>       }
>>>       spin_unlock(&entity->rq_lock);
>>> -
>>> -    if (entity->num_sched_list == 1)
>>> -        entity->sched_list = NULL;
>>>   }
>>>     /**
>>