linux-kernel - Re: [PATCH 1/3] media: rkisp1: Set format defaults based on requested color space

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20250301010252.GI7342@pendragon.ideasonboard.com>
Date: Sat, 1 Mar 2025 03:02:52 +0200
From: Laurent Pinchart <laurent.pinchart@...asonboard.com>
To: Stefan Klug <stefan.klug@...asonboard.com>
Cc: linux-media@...r.kernel.org, Dafna Hirschfeld <dafna@...tmail.com>,
	Mauro Carvalho Chehab <mchehab@...nel.org>,
	Heiko Stuebner <heiko@...ech.de>,
	linux-rockchip@...ts.infradead.org,
	linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/3] media: rkisp1: Set format defaults based on
 requested color space

Hi Stefan,

Thank you for the patch.

On Thu, Feb 27, 2025 at 12:44:59PM +0100, Stefan Klug wrote:
> When color space JPEG is requested, the ISP sets the quantization
> incorrectly to limited range. To fix that, set the xfer_func, ycbcr_enc
> and quantization to the defaults for the requested color space if they
> are not specified explicitly.

The commit message fails to explain why you're addressing xfer_func and
ycbcr_enc to fix the quantization issue.

> Do this only in case we are converting
> from RAW to YUV.

And this should explain why.

> Signed-off-by: Stefan Klug <stefan.klug@...asonboard.com>
> ---
>  .../media/platform/rockchip/rkisp1/rkisp1-isp.c   | 15 ++++++++++++++-
>  1 file changed, 14 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c b/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c
> index d94917211828..468f5a7d03c7 100644
> --- a/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c
> +++ b/drivers/media/platform/rockchip/rkisp1/rkisp1-isp.c
> @@ -680,10 +680,23 @@ static void rkisp1_isp_set_src_fmt(struct rkisp1_isp *isp,

Adding a bit more context:

	set_csc = format->flags & V4L2_MBUS_FRAMEFMT_SET_CSC;

	if (set_csc && src_info->pixel_enc == V4L2_PIXEL_ENC_YUV) {

If V4L2_MBUS_FRAMEFMT_SET_CSC isn't set, the colorspace fields on the
source pad will be copied from the sink pad, which doesn't seem right.
It's a separate issue, but fixing both together may lead to better code.

>  		if (sink_info->pixel_enc == V4L2_PIXEL_ENC_BAYER) {
>  			if (format->colorspace != V4L2_COLORSPACE_DEFAULT)
>  				src_fmt->colorspace = format->colorspace;
> -			if (format->xfer_func != V4L2_XFER_FUNC_DEFAULT)
> +
> +			if (format->xfer_func == V4L2_XFER_FUNC_DEFAULT)

Are you sure the condition should be inverted ?

>  				src_fmt->xfer_func = format->xfer_func;
> +			else
> +				src_fmt->xfer_func =
> +					V4L2_MAP_XFER_FUNC_DEFAULT(format->colorspace);
> +
>  			if (format->ycbcr_enc != V4L2_YCBCR_ENC_DEFAULT)
>  				src_fmt->ycbcr_enc = format->ycbcr_enc;
> +			else
> +				src_fmt->ycbcr_enc =
> +					V4L2_MAP_YCBCR_ENC_DEFAULT(format->colorspace);
> +
> +			if (format->quantization == V4L2_QUANTIZATION_DEFAULT)
> +				src_fmt->quantization =
> +					V4L2_MAP_QUANTIZATION_DEFAULT(false,
> +						format->colorspace, format->ycbcr_enc);

Shouldn't this use src_fmt instead of format ?

I think quantization handling could be moved below.

>  		}
>  
>  		if (format->quantization != V4L2_QUANTIZATION_DEFAULT)

-- 
Regards,

Laurent Pinchart