[0/2] media: v4l2: map UVC_CT_ROLL_ABSOLUTE_CONTROL

Hi,

Some new UVC cameras can report whether they are mounted in 'portrait 
mode'.

In order to userspace getting to know this, this series maps 
UVC_CT_ROLL_ABSOLUTE_CONTROL to V4L2_CID_ROLL_ABSOLUTE.

Tested with Logitech StreamCam and OBSBOT Tiny 2.

Gergo Koteles (2):
  media: v4l2: ctrls: Add ROLL_ABSOLUTE control
  media: v4l2: map UVC_CT_ROLL_ABSOLUTE_CONTROL

 .../userspace-api/media/v4l/ext-ctrls-camera.rst         | 5 +++++
 drivers/media/usb/uvc/uvc_ctrl.c                         | 9 +++++++++
 drivers/media/v4l2-core/v4l2-ctrls-defs.c                | 1 +
 include/uapi/linux/v4l2-controls.h                       | 2 ++
 4 files changed, 17 insertions(+)

On Wed, Jun 12, 2024 at 08:31:14PM +0200, Gergo Koteles wrote:
> Some new UVC cameras can report whether they are mounted in 'portrait
> mode'.
> 
> Current roll degrees (-90, 0, 90, 180) are reported with
> UVC_CT_ROLL_ABSOLUTE_CONTROL.

UVC_CT_ROLL_ABSOLUTE_CONTROL is about controlling the motion of the
camera along the roll axis, while this patch series sounds like you want
to support reporting the mounting orientation of the device, not cause
the device to rotate. Is that right ?

If that's the case, the right V4L2 control to use would be
V4L2_CID_CAMERA_SENSOR_ROTATION. Mapping it to
UVC_CT_ROLL_ABSOLUTE_CONTROL is problematic though, as
UVC_CT_ROLL_ABSOLUTE_CONTROL is not meant for this in the UVC spec. We
would likely need a quirk to control how it gets used.

> Map UVC_CT_ROLL_ABSOLUTE_CONTROL to V4L2_CID_ROLL_ABSOLUTE to
> make it available to userspace.
> 
> Signed-off-by: Gergo Koteles <soyer@irl.hu>
> ---
>  drivers/media/usb/uvc/uvc_ctrl.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/media/usb/uvc/uvc_ctrl.c b/drivers/media/usb/uvc/uvc_ctrl.c
> index 4b685f883e4d..bc3272b6ceb1 100644
> --- a/drivers/media/usb/uvc/uvc_ctrl.c
> +++ b/drivers/media/usb/uvc/uvc_ctrl.c
> @@ -748,6 +748,15 @@ static const struct uvc_control_mapping uvc_ctrl_mappings[] = {
>  		.v4l2_type	= V4L2_CTRL_TYPE_BOOLEAN,
>  		.data_type	= UVC_CTRL_DATA_TYPE_BOOLEAN,
>  	},
> +	{
> +		.id		= V4L2_CID_ROLL_ABSOLUTE,
> +		.entity		= UVC_GUID_UVC_CAMERA,
> +		.selector	= UVC_CT_ROLL_ABSOLUTE_CONTROL,
> +		.size		= 16,
> +		.offset		= 0,
> +		.v4l2_type	= V4L2_CTRL_TYPE_INTEGER,
> +		.data_type	= UVC_CTRL_DATA_TYPE_SIGNED,
> +	},
>  };
>  
>  const struct uvc_control_mapping uvc_ctrl_power_line_mapping_limited = {

Hi Laurent,

On Wed, 2024-06-12 at 23:31 +0300, Laurent Pinchart wrote:
> On Wed, Jun 12, 2024 at 08:31:14PM +0200, Gergo Koteles wrote:
> > Some new UVC cameras can report whether they are mounted in 'portrait
> > mode'.
> > 
> > Current roll degrees (-90, 0, 90, 180) are reported with
> > UVC_CT_ROLL_ABSOLUTE_CONTROL.
> 
> UVC_CT_ROLL_ABSOLUTE_CONTROL is about controlling the motion of the
> camera along the roll axis, while this patch series sounds like you want
> to support reporting the mounting orientation of the device, not cause
> the device to rotate. Is that right ?

The SET_CUR of UVC_CT_ROLL_ABSOLUTE_CONTROL is optional, so I think
it's for reporting also.

The cameras I've tested before can't roll the sensor, they only report
the angle/orientation.

Yes, I was thinking to support the mounting orientation report, however
I tried a friend's DJI Osmo Pocket 3, which can roll the sensor in +-
30˚ with ROLL_ABSOLUTE UVC control. It can also be
Panned/Tilted/Zoomed, very cool.

I think mapping UVC_CT_ROLL_ABSOLUTE_CONTROL to V4L2_CID_ROLL_ABSOLUTE
would be useful. I can reword the commit message a bit if you'd like.

> 
> If that's the case, the right V4L2 control to use would be
> V4L2_CID_CAMERA_SENSOR_ROTATION. Mapping it to
> UVC_CT_ROLL_ABSOLUTE_CONTROL is problematic though, as
> UVC_CT_ROLL_ABSOLUTE_CONTROL is not meant for this in the UVC spec. We
> would likely need a quirk to control how it gets used.
> 

I can also create a quirk for the other two cameras to map
UVC_CT_ROLL_ABSOLUTE_CONTROL to V4L2_CID_CAMERA_SENSOR_ROTATION.

Maybe it can be detected if CT_ROLL_ABSOLUTE_CONTROL doesn't have
SET_CUR, default 0, step 90.

> 
Best regards,
Gergo