[RFC PATCH v2 09/17] gpu: host1x: DMA fences and userspace fence creation

[Mikko Perttunen]

On 9/11/20 1:00 AM, Dmitry Osipenko wrote:
> 05.09.2020 13:34, Mikko Perttunen пишет:
> ...
>>   
>> +static void action_signal_fence(struct host1x_waitlist *waiter)
>> +{
>> +	struct host1x_syncpt_fence *f = waiter->data;
>> +
>> +	host1x_fence_signal(f);
>> +}
>> +
>>   typedef void (*action_handler)(struct host1x_waitlist *waiter);
>>   
>>   static const action_handler action_handlers[HOST1X_INTR_ACTION_COUNT] = {
>>   	action_submit_complete,
>>   	action_wakeup,
>>   	action_wakeup_interruptible,
>> +	action_signal_fence,
>>   };
> 
> My expectation is that we should remove the host1x-waiter entirely. It
> comes from 2011/2012 era of the host1x driver and now duplicates
> functionality provided by the dma-fence and drm-scheduler. Perhaps it
> could be okay to re-use existing code for the starter, but this is
> something to keep in mind that it may be better not to put much effort
> into the older code.
> 

Agreed, it should be cleaned up and probably replaced. I made only 
minimal changes here to get all my tests working, as I didn't want to do 
a full refactoring in this patch series.

Mikko