linux-kernel - Re: [PATCH 1/5] media: videodev2: add Compressed Framebuffer pixel formats

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

Message-ID: <02aa06fd8397b77c9a75d3a8399cb55d3b4d39c1.camel@ndufresne.ca>
Date:   Fri, 05 Jun 2020 11:35:50 -0400
From:   Nicolas Dufresne <nicolas@...fresne.ca>
To:     Neil Armstrong <narmstrong@...libre.com>, hverkuil-cisco@...all.nl
Cc:     linux-media@...r.kernel.org, linux-amlogic@...ts.infradead.org,
        linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        Maxime Jourdan <mjourdan@...libre.com>
Subject: Re: [PATCH 1/5] media: videodev2: add Compressed Framebuffer pixel
 formats

Le jeudi 04 juin 2020 à 15:53 +0200, Neil Armstrong a écrit :
> From: Maxime Jourdan <mjourdan@...libre.com>
> 
> Add two generic Compressed Framebuffer pixel formats to be used
> with a modifier when imported back in another subsystem like DRM/KMS.
> 
> These pixel formats represents generic 8bits and 10bits compressed buffers
> with a vendor specific layout.
> 
> These are aligned with the DRM_FORMAT_YUV420_8BIT and DRM_FORMAT_YUV420_10BIT
> used to describe the underlying compressed buffers used for ARM Framebuffer
> Compression. In the Amlogic case, the compression is different but the
> underlying buffer components is the same.
> 
> Signed-off-by: Maxime Jourdan <mjourdan@...libre.com>
> Signed-off-by: Neil Armstrong <narmstrong@...libre.com>
> ---
>  drivers/media/v4l2-core/v4l2-ioctl.c | 2 ++
>  include/uapi/linux/videodev2.h       | 9 +++++++++
>  2 files changed, 11 insertions(+)
> 
> diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c
> index 2322f08a98be..8f14adfd5bc5 100644
> --- a/drivers/media/v4l2-core/v4l2-ioctl.c
> +++ b/drivers/media/v4l2-core/v4l2-ioctl.c
> @@ -1447,6 +1447,8 @@ static void v4l_fill_fmtdesc(struct v4l2_fmtdesc *fmt)
>  		case V4L2_PIX_FMT_S5C_UYVY_JPG:	descr = "S5C73MX interleaved UYVY/JPEG"; break;
>  		case V4L2_PIX_FMT_MT21C:	descr = "Mediatek Compressed Format"; break;
>  		case V4L2_PIX_FMT_SUNXI_TILED_NV12: descr = "Sunxi Tiled NV12 Format"; break;
> +		case V4L2_PIX_FMT_YUV420_8BIT:	descr = "Compressed YUV 4:2:0 8-bit Format"; break;
> +		case V4L2_PIX_FMT_YUV420_10BIT:	descr = "Compressed YUV 4:2:0 10-bit Format"; break;

When I read the DRM documentation [0], I'm reading that YUV420_8BIT
definition matches V4L2_PIX_FMT_YVU420 and V4L2_PIX_FMT_YVU420M fully.
In fact, on DRM side, to represent that format you want to expose here,
they will strictly combine this generic format (documented un-
compressed) with a modifier generated with the macro
DRM_FORMAT_MOD_ARM_AFBC(*). And only the combination represent a unique
and share-able format.

In absence of modifier in V4L2 API, this compressed format should be
named accordingly to the compressed algorithm used (something like
FMT_YUV420_8BIT_AML_FBC). So I believe these format name cannot be
copied as-is like this, as they can only introduce more ambiguity in
the already quite hard to follow naming of pixel formats. In fact, it
is very common to see multiple different vendor compressions on the
same SoC, so I don't really believe a "generic" compressed format make
sense. To site one, the IMX8M, which got Verrisillicon/Vivante DEC400
on the GPU, and the Hantro G2 compression format. Both will apply to
NV12 class of format so in DRM they would be NV12 + modifier, and the
combination forms the unique format. Now, in term of sharing, they must
be differiented by userspace, as support for compression/decompression
is heterogeneous (in that case the GPU does not support Hantro G2
decompression or compression, and the VPU does not support DEC400).

I'll remind that the modifier implementation has great value and is
much more scalable then the current V4L2 approach. There has been some
early proposal for this, maybe it's time to prioritize because this
list will starts growing with hundred or even thousands or format,
which is clearly indicated by the increase of modifier generator macro
on the DRM side.

>  		default:
>  			if (fmt->description[0])
>  				return;
> diff --git a/include/uapi/linux/videodev2.h b/include/uapi/linux/videodev2.h
> index c3a1cf1c507f..90b9949acb8a 100644
> --- a/include/uapi/linux/videodev2.h
> +++ b/include/uapi/linux/videodev2.h
> @@ -705,6 +705,15 @@ struct v4l2_pix_format {
>  #define V4L2_PIX_FMT_FWHT     v4l2_fourcc('F', 'W', 'H', 'T') /* Fast Walsh Hadamard Transform (vicodec) */
>  #define V4L2_PIX_FMT_FWHT_STATELESS     v4l2_fourcc('S', 'F', 'W', 'H') /* Stateless FWHT (vicodec) */
>  
> +/*
> + * Compressed Luminance+Chrominance meta-formats
> + * In these formats, the component ordering is specified (Y, followed by U
> + * then V), but the exact Linear layout is undefined.
> + * These formats can only be used with a non-Linear modifier.
> + */
> +#define V4L2_PIX_FMT_YUV420_8BIT	v4l2_fourcc('Y', 'U', '0', '8') /* 1-plane YUV 4:2:0 8-bit */
> +#define V4L2_PIX_FMT_YUV420_10BIT	v4l2_fourcc('Y', 'U', '1', '0') /* 1-plane YUV 4:2:0 10-bit */
> +
>  /*  Vendor-specific formats   */
>  #define V4L2_PIX_FMT_CPIA1    v4l2_fourcc('C', 'P', 'I', 'A') /* cpia1 YUV */
>  #define V4L2_PIX_FMT_WNVA     v4l2_fourcc('W', 'N', 'V', 'A') /* Winnov hw compress */