Re: [PATCH v5 1/5] media: uapi: Introduce new control for video encoder ROI

[Hans Verkuil]

On 15/08/2026 21:22, Deepa Guthyappa Madivalara wrote:
> Add custom control, V4L2_CID_MPEG_VIDEO_ROI_MB_DELTA_QP for

A small terminology issue: custom -> array

Custom controls are driver-specific controls. But this adds a standard
array control. Please check this series for the use of 'custom' and
replace it with 'array'.

> video encoder Region of Interest to allow applications to specify
> different quality levels for specific regions in video frames. Define
> V4L2_CID_MPEG_VIDEO_ROI_MB_SIZE to retrieve supported MB size based on
> codec from the hardware encoder and set the corresponding ROI MB delta_qp
> map that adjusts quantization relative to the frame QP's base value.
> 
> This enables use cases like prioritizing quality for faces in video
> conferencing or important objects in surveillance footage while reducing
> bitrate for less critical areas.
> 
> Signed-off-by: Deepa Guthyappa Madivalara <deepa.madivalara@xxxxxxxxxxxxxxxx>
> ---
>  .../userspace-api/media/v4l/ext-ctrls-codec.rst       | 19 +++++++++++++++++++
>  .../media/v4l/videodev2.h.rst.exceptions              |  1 +
>  .../userspace-api/media/v4l/vidioc-g-ext-ctrls.rst    |  4 ++++
>  .../userspace-api/media/v4l/vidioc-queryctrl.rst      |  6 ++++++
>  include/media/v4l2-ctrls.h                            |  1 +
>  include/uapi/linux/v4l2-controls.h                    |  2 ++
>  include/uapi/linux/videodev2.h                        |  2 ++
>  7 files changed, 35 insertions(+)
> 
> diff --git a/Documentation/userspace-api/media/v4l/ext-ctrls-codec.rst b/Documentation/userspace-api/media/v4l/ext-ctrls-codec.rst
> index ab865a1a6ba929f1725ce2ab0fd3aeee4a69329a..c16b1c5cd1c02b7f66b5f13433eb039eb615b35e 100644
> --- a/Documentation/userspace-api/media/v4l/ext-ctrls-codec.rst
> +++ b/Documentation/userspace-api/media/v4l/ext-ctrls-codec.rst
> @@ -1674,6 +1674,25 @@ enum v4l2_mpeg_video_h264_hierarchical_coding_type -
>      Codecs need to always use the specified range, rather then a HW custom range.
>      Applicable to encoders
>  
> +``V4L2_CID_MPEG_VIDEO_ROI_MB_DELTA_QP (__s8)``

__s8 -> __s8 array

Otherwise it is not clear that this is a compound control containing an array.

> +    This custom control is used to set ROI MB map delta_Qp for the whole frame.

Here is another use of 'custom' were 'array' is meant.

> +    The frame is divided into a grid of MB Size by MB Size pixels (MB Size is
> +    obtained from querying ``V4L2_CID_MPEG_VIDEO_ROI_MB_SIZE``). Each block
> +    is configured with delta_Qp in raster order. The valid range for delta_Qp
> +    is encoder dependent.
> +    Applicable to encoders.
> +
> +``V4L2_CID_MPEG_VIDEO_ROI_MB_SIZE`` (integer)
> +    A read-only control that reports the MB Size used for the ROI delta_QP map.
> +    The supported MB Size depends on the encoder codec.
> +    Only square MBs are supported. The value returned represents the
> +    width and height of each block in pixels. The ROI delta_QP map provided
> +    through the ROI control must use the MB Size reported by this control.
> +    The number of entries in the ROI delta_QP array is determined by the
> +    frame dimensions and the reported MB Size. For example, for a QCIF frame
> +    (176x144) if reported MB Size is 16, the delta_QP array must contain
> +    99 entries (176 x 144 / 16 x 16).

What happens if the width/height is not nicely dividable by 16? E.g. 1920x1080.

> +
>  .. raw:: latex
>  
>      \normalsize
> diff --git a/Documentation/userspace-api/media/v4l/videodev2.h.rst.exceptions b/Documentation/userspace-api/media/v4l/videodev2.h.rst.exceptions
> index 6182b4e2d2ee002c0d3562e1cc3bc2fd9c4286d6..b036ed3ec2f0744175cad1cbc4e73fe828e0c955 100644
> --- a/Documentation/userspace-api/media/v4l/videodev2.h.rst.exceptions
> +++ b/Documentation/userspace-api/media/v4l/videodev2.h.rst.exceptions
> @@ -141,6 +141,7 @@ replace symbol V4L2_CTRL_TYPE_STRING :c:type:`V4L.v4l2_ctrl_type`
>  replace symbol V4L2_CTRL_TYPE_U16 :c:type:`V4L.v4l2_ctrl_type`
>  replace symbol V4L2_CTRL_TYPE_U32 :c:type:`V4L.v4l2_ctrl_type`
>  replace symbol V4L2_CTRL_TYPE_U8 :c:type:`V4L.v4l2_ctrl_type`
> +replace symbol V4L2_CTRL_TYPE_S8 :c:type:`V4L.v4l2_ctrl_type`
>  replace symbol V4L2_CTRL_TYPE_MPEG2_SEQUENCE :c:type:`V4L.v4l2_ctrl_type`
>  replace symbol V4L2_CTRL_TYPE_MPEG2_PICTURE :c:type:`V4L.v4l2_ctrl_type`
>  replace symbol V4L2_CTRL_TYPE_MPEG2_QUANTISATION :c:type:`V4L.v4l2_ctrl_type`
> diff --git a/Documentation/userspace-api/media/v4l/vidioc-g-ext-ctrls.rst b/Documentation/userspace-api/media/v4l/vidioc-g-ext-ctrls.rst
> index d78328152b7552f8f0d30a9e2f7c104bfd4b70d3..fe5f8992389c32d8f35f432229ef2f734002644e 100644
> --- a/Documentation/userspace-api/media/v4l/vidioc-g-ext-ctrls.rst
> +++ b/Documentation/userspace-api/media/v4l/vidioc-g-ext-ctrls.rst
> @@ -188,6 +188,10 @@ still cause this situation.
>        - ``p_u32``
>        - A pointer to a matrix control of unsigned 32-bit values. Valid if
>  	this control is of type ``V4L2_CTRL_TYPE_U32``.
> +    * - __s8 *
> +      - ``p_s8``
> +      - A pointer to a matrix control of signed 8-bit values. Valid if
> +	this control is of type ``V4L2_CTRL_TYPE_S8``.
>      * - __s32 *
>        - ``p_s32``
>        - A pointer to a matrix control of signed 32-bit values. Valid if
> diff --git a/Documentation/userspace-api/media/v4l/vidioc-queryctrl.rst b/Documentation/userspace-api/media/v4l/vidioc-queryctrl.rst
> index 82c8b52e771ce9301491b3f752e501a8fe7e2434..57e1652ae72e0324742142446acf32a16b94941c 100644
> --- a/Documentation/userspace-api/media/v4l/vidioc-queryctrl.rst
> +++ b/Documentation/userspace-api/media/v4l/vidioc-queryctrl.rst
> @@ -425,6 +425,12 @@ See also the examples in :ref:`control`.
>        - any
>        - An unsigned 32-bit valued control ranging from minimum to maximum
>  	inclusive. The step value indicates the increment between values.
> +    * - ``V4L2_CTRL_TYPE_S8``
> +      - any
> +      - any
> +      - any
> +      - A signed 8-bit valued control ranging from minimum to maximum
> +	inclusive. The step value indicates the increment between values.
>      * - ``V4L2_CTRL_TYPE_MPEG2_QUANTISATION``
>        - n/a
>        - n/a
> diff --git a/include/media/v4l2-ctrls.h b/include/media/v4l2-ctrls.h
> index 327976b14d50b73a631e4d2ed95d9f4d146c45ab..a1806ddbc797efa52e83cd3f685ef70d5b5483d2 100644
> --- a/include/media/v4l2-ctrls.h
> +++ b/include/media/v4l2-ctrls.h
> @@ -64,6 +64,7 @@ union v4l2_ctrl_ptr {
>  	s32 *p_s32;
>  	s64 *p_s64;
>  	u8 *p_u8;
> +	s8 *p_s8;
>  	u16 *p_u16;
>  	u32 *p_u32;
>  	char *p_char;
> diff --git a/include/uapi/linux/v4l2-controls.h b/include/uapi/linux/v4l2-controls.h
> index affec0ab4781507be761cb791ff737460acf9cb7..00a2c48630bd198faa9a2b833a9aa346d8357555 100644
> --- a/include/uapi/linux/v4l2-controls.h
> +++ b/include/uapi/linux/v4l2-controls.h
> @@ -920,6 +920,8 @@ enum v4l2_mpeg_video_av1_level {
>  };
>  
>  #define V4L2_CID_MPEG_VIDEO_AVERAGE_QP  (V4L2_CID_CODEC_BASE + 657)
> +#define V4L2_CID_MPEG_VIDEO_ROI_MB_DELTA_QP  (V4L2_CID_CODEC_BASE + 658)
> +#define V4L2_CID_MPEG_VIDEO_ROI_MB_SIZE  (V4L2_CID_CODEC_BASE + 659)
>  
>  /*  MPEG-class control IDs specific to the CX2341x driver as defined by V4L2 */
>  #define V4L2_CID_CODEC_CX2341X_BASE				(V4L2_CTRL_CLASS_CODEC | 0x1000)
> diff --git a/include/uapi/linux/videodev2.h b/include/uapi/linux/videodev2.h
> index 5373dba640fae3d19247ab1c74ba336e05dc0fe0..95d8369e0d3d1e434978b9e576e0ef122691c65c 100644
> --- a/include/uapi/linux/videodev2.h
> +++ b/include/uapi/linux/videodev2.h
> @@ -1884,6 +1884,7 @@ struct v4l2_ext_control {
>  		__s64 value64;
>  		char __user *string;
>  		__u8 __user *p_u8;
> +		__s8 __user *p_s8;
>  		__u16 __user *p_u16;
>  		__u32 __user *p_u32;
>  		__s32 __user *p_s32;
> @@ -1961,6 +1962,7 @@ enum v4l2_ctrl_type {
>  	V4L2_CTRL_TYPE_U8	     = 0x0100,
>  	V4L2_CTRL_TYPE_U16	     = 0x0101,
>  	V4L2_CTRL_TYPE_U32	     = 0x0102,
> +	V4L2_CTRL_TYPE_S8		 = 0x0103,
>  	V4L2_CTRL_TYPE_AREA          = 0x0106,
>  	V4L2_CTRL_TYPE_RECT	     = 0x0107,
>  
> 

Hmm, I see that patch 2/5 adds the v4l2-core changes for the new controls.

Can you rework patches 1 and 2? It is much easier to review if the first
patch adds V4L2_CTRL_TYPE_S8 support in both headers, documentation and v4l2-core,
and the second patch does the same for the new controls.

Currently the patches mix adding the new type and new controls, and that's actually
hard to review.

Regards,

	Hans