Re: [RFC 0/5] rework fences on struct sync_file

[Gustavo Padovan]

2016-06-24 Christian König <deathsimple@xxxxxxxxxxx>:

> Am 24.06.2016 um 15:17 schrieb Gustavo Padovan:
> > Hi Christian,
> > 
> > 2016-06-24 Christian König <christian.koenig@xxxxxxx>:
> > 
> > > Am 23.06.2016 um 17:29 schrieb Gustavo Padovan:
> > > > From: Gustavo Padovan <gustavo.padovan@xxxxxxxxxxxxxxx>
> > > > 
> > > > Hi all,
> > > > 
> > > > This is an attempt to improve fence support on Sync File. The basic idea
> > > > is to have only sync_file->fence and store all fences there, either as
> > > > normal fences or fence_arrays. That way we can remove some potential
> > > > duplication when using fence_array with sync_file: the duplication of the array
> > > > of fences and the duplication of fence_add_callback() for all fences.
> > > > 
> > > > Now when creating a new sync_file during the merge process sync_file_set_fence()
> > > > will set sync_file->fence based on the number of fences for that sync_file. If
> > > > there is more than one fence a fence_array is created. One important advantage
> > > > approach is that we only add one fence callback now, no matter how many fences
> > > > there are in a sync_file - the individual callbacks are added by fence_array.
> > > > 
> > > > Two fence ops had to be created to help abstract the difference between handling
> > > > fences and fences_arrays: .teardown() and .get_fences(). The former run needed
> > > > on fence_array, and the latter just return a copy of all fences in the fence.
> > > > I'm not so sure about adding those two, speacially .get_fences(). What do you
> > > > think?
> > > Clearly not a good idea to add this a fence ops, cause those are specialized
> > > functions for only a certain fence implementation (the fence_array).
> > Are you refering only to .get_fences()?
> 
> That comment was only for the get_fences() operation, but the teardown()
> callback looks very suspicious to me as well.
> 
> Can you explain once more why that should be necessary?

When the sync_file owner exits we need to clean up it and that means releasing
the fence too, however with fence_array we can't just call fence_put()
as a extra reference to array->base for each fence is held when enabling
signalling. Thus we need a prior step, that I called teardown(), to
remove the callback for not signaled fences and put the extra
references.

Another way to do this would be:

	if (fence_is_array(sync_file->fence))
		fence_array_destroy(to_fence_array(sync_file->fence));
	else
		fence_put(sync_file_fence);

This would avoid the extra ops, maybe we should go this way.

	Gustavo