[PATCH] dma-buf: add more reservation object locking wrappers

Koenig, Christian Christian.Koenig at amd.com
Mon Aug 5 07:31:22 UTC 2019

Am 31.07.19 um 15:58 schrieb Chris Wilson:
> Quoting Christian König (2019-07-31 14:34:28)
>> Am 31.07.19 um 14:33 schrieb Chris Wilson:
>>> Quoting Christian König (2019-07-31 12:38:53)
>>>> Complete the abstraction of the ww_mutex inside the reservation object.
>>>> This allows us to add more handling and debugging to the reservation
>>>> object in the future.
>>>> Signed-off-by: Christian König <christian.koenig at amd.com>
>>> Looks entirely mechanical,
>>> Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
>>> Quietly opines for s/reservation_object/dma_reservation/
>> I was thinking about that as well because "reservation_object" is just a
>> rather long name and not very descriptive.
>> But I'm not sure if dma_reservation fits either. How about something
>> like dma_cntrl?
> Reservation kind of works because of historical usage, but is itself
> rather long. Control doesn't mean much to me. dma_sequence? Maybe just
> dma_syncpt, as the snapshot of fences is itself a fence /
> synchronisation point. Though that is at odds with other usage of
> syncpt, we have an unordered collection of fences across multiple
> timelines, as opposed to a single point along a timeline.
> Fwiw, we use i915_active for the similar purpose of tracking the active
> collection of fences, so maybe dma_active?

In amdgpu we have amdgpu_sync for that, but that isn't much descriptive 

Ok, let's stick with dma_reservation for now unless somebody has a 
better idea.


> -Chris

More information about the dri-devel mailing list