[PATCH] drm/amdgpu: Fix one list corruption when create queue fails

[Felix Kuehling]

Am 2022-07-07 um 09:39 schrieb philip yang:
>
>
> On 2022-07-07 06:28, xinhui pan wrote:
>> Queue would be freed when create_queue_cpsch fails
>> So lets do queue cleanup otherwise various list and memory issues
>> happen.
>
> This bug was introduced when adding MES support, as we used to ignore 
> execute_queues_cpsch return value. Cleanup and return error to user 
> space looks good to me.
>
This is similar to the queue destroy failure you looked at. A failure in 
execute_queues_cpsch doesn't really indicate that the queue creation 
failed. There is nothing specifically wrong with this queue. It just 
means that HWS is probably hanging. So this problem will be handled with 
a GPU reset anyway.

Regards,
   Felix


> Reviewed-by: Philip Yang <Philip.Yang at amd.com>
>
>> Signed-off-by: xinhui pan<xinhui.pan at amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 11 +++++------
>>   1 file changed, 5 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>> index 93a0b6995430..e83725a28106 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
>> @@ -1674,14 +1674,13 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
>>   	if (q->properties.is_active) {
>>   		increment_queue_count(dqm, qpd, q);
>>   
>> -		if (!dqm->dev->shared_resources.enable_mes) {
>> +		if (!dqm->dev->shared_resources.enable_mes)
>>   			retval = execute_queues_cpsch(dqm,
>> -					     KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
>> -		} else {
>> +					KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
>> +		else
>>   			retval = add_queue_mes(dqm, q, qpd);
>> -			if (retval)
>> -				goto cleanup_queue;
>> -		}
>> +		if (retval)
>> +			goto cleanup_queue;
>>   	}
>>   
>>   	/*