public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Hans Verkuil <hverkuil+cisco@kernel.org>
To: Pavan Bobba <opensource206@gmail.com>,
	mchehab@kernel.org, hverkuil@kernel.org, ribalda@chromium.org,
	laurent.pinchart@ideasonboard.com, yunkec@google.com,
	sakari.ailus@linux.intel.com, james.cowgill@blaize.com,
	Nicolas Dufresne <nicolas.dufresne@collabora.com>
Cc: linux-media@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3] media: v4l2-ctrls: add full AV1 profile validation in validate_av1_sequence()
Date: Wed, 22 Oct 2025 09:14:03 +0200	[thread overview]
Message-ID: <8199bec4-b9e1-4d6e-98da-a4d7eb667437@kernel.org> (raw)
In-Reply-To: <20250913105252.26886-1-opensource206@gmail.com>

Hi Pavan,

On 13/09/2025 12:52, Pavan Bobba wrote:
> Complete the "TODO: PROFILES" by enforcing profile-specific and
> monochrome constraints as defined by the AV1 specification
> (Section 5.5.2, "Color config syntax").
> 
> The validator now checks:
> 
>  - Flags: reject any unknown bits set in sequence->flags
>  - Profile range: only profiles 0..2 are valid
>  - Profile 0: 8/10-bit only, subsampling must be 4:2:0 (sx=1, sy=1),
>    monochrome allowed
>  - Profile 1: 8/10-bit only, subsampling must be 4:4:4 (sx=0, sy=0),
>    monochrome forbidden
>  - Profile 2:
>     * 8/10-bit: only 4:2:2 allowed (sx=1, sy=0)
>     * 12-bit: 4:4:4 (sx=0, sy=0), 4:2:2 (sx=1, sy=0), or 4:2:0 (sx=1, sy=1)
>       allowed
>  - Monochrome path (all profiles except 1): forces subsampling_x=1,
>    subsampling_y=1, separate_uv_delta_q=0
> 
> These checks prevent userspace from providing invalid AV1 sequence
> headers that would otherwise be accepted, leading to undefined driver
> or hardware behavior.

This patch was merged in our media-committers next branch, but I noticed that
it now fails the v4l2-compliance test for the visl driver.

The cause is that the new validation now fails with the default values for
this control as set in std_init_compound().

You can test this yourself by loading the visl driver and then running
v4l2-compliance -d /dev/videoX -E --verbose
(-E stops at the first error)

Can you provide a patch to initialize this control with sane values?

Apologies for not noticing this before: there are some issues with the automatic
regression tests in our CI, so the tests weren't run.

Regards,

	Hans

> 
> Signed-off-by: Pavan Bobba <opensource206@gmail.com>
> ---
> v1 -> v2 : Added more checks for subsampling combinations per profile.
>          : Added a TODO note in the function header for checks to be implemented later.
> 
> v2 -> v3 : Patch generated properly with all the changes
> 
>  drivers/media/v4l2-core/v4l2-ctrls-core.c | 125 +++++++++++++++++-----
>  1 file changed, 100 insertions(+), 25 deletions(-)
> 
> diff --git a/drivers/media/v4l2-core/v4l2-ctrls-core.c b/drivers/media/v4l2-core/v4l2-ctrls-core.c
> index 98b960775e87..fa03341588e4 100644
> --- a/drivers/media/v4l2-core/v4l2-ctrls-core.c
> +++ b/drivers/media/v4l2-core/v4l2-ctrls-core.c
> @@ -827,39 +827,114 @@ static int validate_av1_frame(struct v4l2_ctrl_av1_frame *f)
>  	return 0;
>  }
>  
> +/**
> + * validate_av1_sequence - validate AV1 sequence header fields
> + * @s: control struct from userspace
> + *
> + * Implements AV1 spec §5.5.2 color_config() checks that are
> + * possible with the current v4l2_ctrl_av1_sequence definition.
> + *
> + * TODO: extend validation once additional fields such as
> + *       color_primaries, transfer_characteristics,
> + *       matrix_coefficients, and chroma_sample_position
> + *       are added to the uAPI.
> + *
> + * Returns 0 if valid, -EINVAL otherwise.
> + */
>  static int validate_av1_sequence(struct v4l2_ctrl_av1_sequence *s)
>  {
> -	if (s->flags &
> -	~(V4L2_AV1_SEQUENCE_FLAG_STILL_PICTURE |
> -	 V4L2_AV1_SEQUENCE_FLAG_USE_128X128_SUPERBLOCK |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_FILTER_INTRA |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_INTRA_EDGE_FILTER |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_INTERINTRA_COMPOUND |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_MASKED_COMPOUND |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_WARPED_MOTION |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_DUAL_FILTER |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_ORDER_HINT |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_JNT_COMP |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_REF_FRAME_MVS |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_SUPERRES |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_CDEF |
> -	 V4L2_AV1_SEQUENCE_FLAG_ENABLE_RESTORATION |
> -	 V4L2_AV1_SEQUENCE_FLAG_MONO_CHROME |
> -	 V4L2_AV1_SEQUENCE_FLAG_COLOR_RANGE |
> -	 V4L2_AV1_SEQUENCE_FLAG_SUBSAMPLING_X |
> -	 V4L2_AV1_SEQUENCE_FLAG_SUBSAMPLING_Y |
> -	 V4L2_AV1_SEQUENCE_FLAG_FILM_GRAIN_PARAMS_PRESENT |
> -	 V4L2_AV1_SEQUENCE_FLAG_SEPARATE_UV_DELTA_Q))
> -		return -EINVAL;
> +	const bool mono  = s->flags & V4L2_AV1_SEQUENCE_FLAG_MONO_CHROME;
> +	const bool sx    = s->flags & V4L2_AV1_SEQUENCE_FLAG_SUBSAMPLING_X;
> +	const bool sy    = s->flags & V4L2_AV1_SEQUENCE_FLAG_SUBSAMPLING_Y;
> +	const bool uv_dq = s->flags & V4L2_AV1_SEQUENCE_FLAG_SEPARATE_UV_DELTA_Q;
>  
> -	if (s->seq_profile == 1 && s->flags & V4L2_AV1_SEQUENCE_FLAG_MONO_CHROME)
> +	/* 1. Reject unknown flags */
> +	if (s->flags &
> +	    ~(V4L2_AV1_SEQUENCE_FLAG_STILL_PICTURE |
> +	      V4L2_AV1_SEQUENCE_FLAG_USE_128X128_SUPERBLOCK |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_FILTER_INTRA |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_INTRA_EDGE_FILTER |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_INTERINTRA_COMPOUND |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_MASKED_COMPOUND |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_WARPED_MOTION |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_DUAL_FILTER |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_ORDER_HINT |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_JNT_COMP |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_REF_FRAME_MVS |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_SUPERRES |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_CDEF |
> +	      V4L2_AV1_SEQUENCE_FLAG_ENABLE_RESTORATION |
> +	      V4L2_AV1_SEQUENCE_FLAG_MONO_CHROME |
> +	      V4L2_AV1_SEQUENCE_FLAG_COLOR_RANGE |
> +	      V4L2_AV1_SEQUENCE_FLAG_SUBSAMPLING_X |
> +	      V4L2_AV1_SEQUENCE_FLAG_SUBSAMPLING_Y |
> +	      V4L2_AV1_SEQUENCE_FLAG_FILM_GRAIN_PARAMS_PRESENT |
> +	      V4L2_AV1_SEQUENCE_FLAG_SEPARATE_UV_DELTA_Q))
>  		return -EINVAL;
>  
> -	/* reserved */
> +	/* 2. Profile range */
>  	if (s->seq_profile > 2)
>  		return -EINVAL;
>  
> -	/* TODO: PROFILES */
> +	/* 3. Monochrome shortcut */
> +	if (mono) {
> +		/* Profile 1 forbids monochrome */
> +		if (s->seq_profile == 1)
> +			return -EINVAL;
> +
> +		/* Mono → subsampling must look like 4:0:0: sx=1, sy=1 */
> +		if (!sx || !sy)
> +			return -EINVAL;
> +
> +		/* separate_uv_delta_q must be 0 */
> +		if (uv_dq)
> +			return -EINVAL;
> +
> +		return 0;
> +	}
> +
> +	/* 4. Profile-specific rules */
> +	switch (s->seq_profile) {
> +	case 0:
> +		/* Profile 0: only 8/10-bit, subsampling=4:2:0 (sx=1, sy=1) */
> +		if (s->bit_depth != 8 && s->bit_depth != 10)
> +			return -EINVAL;
> +		if (!(sx && sy))
> +			return -EINVAL;
> +		break;
> +
> +	case 1:
> +		/* Profile 1: only 8/10-bit, subsampling=4:4:4 (sx=0, sy=0) */
> +		if (s->bit_depth != 8 && s->bit_depth != 10)
> +			return -EINVAL;
> +		if (sx || sy)
> +			return -EINVAL;
> +		break;
> +
> +	case 2:
> +		/* Profile 2: 8/10/12-bit allowed */
> +		if (s->bit_depth != 8 && s->bit_depth != 10 &&
> +		    s->bit_depth != 12)
> +			return -EINVAL;
> +
> +		if (s->bit_depth == 12) {
> +			if (!sx) {
> +				/* 4:4:4 → sy must be 0 */
> +				if (sy)
> +					return -EINVAL;
> +			} else {
> +				/* sx=1 → sy=0 (4:2:2) or sy=1 (4:2:0) */
> +				if (sy != 0 && sy != 1)
> +					return -EINVAL;
> +			}
> +		} else {
> +			/* 8/10-bit → only 4:2:2 allowed (sx=1, sy=0) */
> +			if (!(sx && !sy))
> +				return -EINVAL;
> +		}
> +		break;
> +	}
> +
>  	return 0;
>  }
>  


  parent reply	other threads:[~2025-10-22  7:14 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-09-13 10:52 [PATCH v3] media: v4l2-ctrls: add full AV1 profile validation in validate_av1_sequence() Pavan Bobba
2025-09-27  4:51 ` opensource india
2025-09-27  8:56   ` Daniel Almeida
2025-10-07 10:50     ` opensource india
2025-10-07 13:49       ` Daniel Almeida
2025-10-18  9:21         ` opensource india
2025-10-22  7:14 ` Hans Verkuil [this message]
2025-10-23 10:32   ` opensource india
2025-10-23 10:43     ` Hans Verkuil
2025-10-23 13:03       ` John Cox
2025-10-23 13:14         ` Hans Verkuil
     [not found]           ` <CAKPKb88Tov27+c227p8k0KAuZtm_LNNxDkf=5YBfDYw94afFPw@mail.gmail.com>
2025-10-27 15:54             ` Hans Verkuil
2025-10-28  6:35               ` opensource india

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8199bec4-b9e1-4d6e-98da-a4d7eb667437@kernel.org \
    --to=hverkuil+cisco@kernel.org \
    --cc=hverkuil@kernel.org \
    --cc=james.cowgill@blaize.com \
    --cc=laurent.pinchart@ideasonboard.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-media@vger.kernel.org \
    --cc=mchehab@kernel.org \
    --cc=nicolas.dufresne@collabora.com \
    --cc=opensource206@gmail.com \
    --cc=ribalda@chromium.org \
    --cc=sakari.ailus@linux.intel.com \
    --cc=yunkec@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox