Re: [PATCH] media: v4l2-subdev: Fix stream handling for crop API

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

On Wed, Apr 03, 2024 at 09:51:01AM +0300, Tomi Valkeinen wrote:
> On 02/04/2024 23:11, Laurent Pinchart wrote:
> > On Tue, Apr 02, 2024 at 03:23:03PM +0000, Sakari Ailus wrote:
> >> On Tue, Apr 02, 2024 at 12:11:30PM +0300, Tomi Valkeinen wrote:
> >>> On 02/04/2024 11:46, Sakari Ailus wrote:
> >>>> On Tue, Apr 02, 2024 at 11:44:07AM +0300, Laurent Pinchart wrote:
> >>>>> On Tue, Apr 02, 2024 at 08:20:22AM +0000, Sakari Ailus wrote:
> >>>>>> Moi,
> >>>>>>
> >>>>>> On Tue, Apr 02, 2024 at 02:37:25AM +0300, Laurent Pinchart wrote:
> >>>>>>> When support for streams was added to the V4L2 subdev API, the
> >>>>>>> v4l2_subdev_crop structure was extended with a stream field, but the
> >>>>>>> field was not handled in the core code that translates the
> >>>>>>> VIDIOC_SUBDEV_[GS]_CROP ioctls to the selection API. Fix it.
> >>>>>>
> >>>>>> The field is indeed in the UAPI headers. But do we want to support the CROP
> >>>>>> IOCTL for streams? Shouldn't the callers be using the [GS]_SELECTION
> >>>>>> instead?
> >>>>>
> >>>>> They should, but if the field is there, we should support it :-) The
> >>>>> alternative is to remove it. It will cause failures in v4l2-compliance
> >>>>> that we'll need to handle though.
> >>>>
> >>>> I'd prefer to stick to selections here, this is new functionality so
> >>>> [GS]_CROP support isn't required. I don't have a strong opinion on the
> >>>> matter though.
> >>>
> >>> Maybe it's easier to just support the stream field, instead of making
> >>> [GS]_CROP the odd case which looks like it should support streams, but then
> >>> doesn't...
> >>
> >> It's an old IOCTL already replaced by the [GS]_SELECTION. I mainly write
> >> kernel space software but overall I think it's better if we can provide a
> >> single API for controlling cropping instead of two with similar
> >> functionality, of which the user then should choose from.
> >>
> >> It should be also documented in this context if we choose support
> >> [GS]_CROP.
> >>
> >> So I believe we have less work to do and have a better result if we just
> >> drop the stream field there. :-)
> > 
> > I tend to agree, even if that's only a slight preference. Tomi, if
> > you're fine with this, I'll update the patch.
> 
> I'm fine with it. So we now should move the 'stream' field back to 
> reserved, and add documentation saying that [GS]_CROP works differently 
> than the other ioctls?

That's the idea. I'll send patches.

-- 
Regards,

Laurent Pinchart