On Tue, Nov 20, 2018 at 10:41:42AM +0100, Hans Verkuil wrote:
> On 11/20/2018 10:27 AM, Sakari Ailus wrote:
> > Hi Hans,
> > 
> > On Tue, Nov 20, 2018 at 09:58:43AM +0100, Hans Verkuil wrote:
> >> Add new buffer capability flags to indicate if the VIDIOC_PREPARE_BUF or
> >> VIDIOC_CREATE_BUFS ioctls are supported.
> > 
> > Are there practical benefits from the change for the user space?
> 
> The more important ioctl to know about is PREPARE_BUF. I noticed this when 
> working
> on v4l2-compliance: the only way to know for an application if PREPARE_BUF 
> exists
> is by trying it, but then you already have prepared a buffer. That's not what 
> you
> want in the application, you need a way to know up front if prepare_buf is 
> present
> or not without having to actually execute it.
> 
> CREATE_BUFS was added because not all drivers support it. It can be dropped 
> since
> it is possible to test for the existence of CREATE_BUFS without actually 
> allocating
> anything, but if I'm adding V4L2_BUF_CAP_SUPPORTS_PREPARE_BUF anyway, then it 
> is
> trivial to add V4L2_BUF_CAP_SUPPORTS_CREATE_BUFS as well to avoid an 
> additional
> ioctl call.
> 
> Hmm, I should have explained this in the commit log.

Please add:

Acked-by: Sakari Ailus <[email protected]>

-- 
Sakari Ailus
[email protected]

Reply via email to