linux-kernel - Re: [PATCH RFT 5/7] media: hantro: jpeg: Add COM segment to JPEG header to align image scan

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 for Android: free password hash cracker in your pocket

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

Message-ID: <YciZZ2hA4uMveN2l@eze-laptop>
Date:   Sun, 26 Dec 2021 13:33:43 -0300
From:   Ezequiel Garcia <ezequiel@...guardiasur.com.ar>
To:     Chen-Yu Tsai <wenst@...omium.org>
Cc:     Philipp Zabel <p.zabel@...gutronix.de>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Hans Verkuil <hverkuil-cisco@...all.nl>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Tomasz Figa <tfiga@...omium.org>, linux-media@...r.kernel.org,
        linux-rockchip@...ts.infradead.org, linux-staging@...ts.linux.dev,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH RFT 5/7] media: hantro: jpeg: Add COM segment to JPEG
 header to align image scan

Hi,

On Fri, Dec 24, 2021 at 04:42:46PM +0800, Chen-Yu Tsai wrote:
> The JPEG header size is not 64-bit aligned. This makes the driver
> require a bounce buffer for the encoded JPEG image scan output.
> 
> Add a COM (comment) segment to the JPEG header so that the header size
> is a multiple of 64 bits. This will then allow dropping the use of the
> bounce buffer, and instead have the hardware write out to the capture
> buffer directly.
> 
> Signed-off-by: Chen-Yu Tsai <wenst@...omium.org>
> ---
>  drivers/staging/media/hantro/hantro_jpeg.c | 3 +++
>  drivers/staging/media/hantro/hantro_jpeg.h | 2 +-
>  2 files changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/media/hantro/hantro_jpeg.c b/drivers/staging/media/hantro/hantro_jpeg.c
> index 7d4018bd6876..51e67e5cf86f 100644
> --- a/drivers/staging/media/hantro/hantro_jpeg.c
> +++ b/drivers/staging/media/hantro/hantro_jpeg.c
> @@ -247,6 +247,9 @@ static const unsigned char hantro_jpeg_header[JPEG_HEADER_SIZE] = {
>  	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
>  	0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
>  
> +	/* COM */
> +	0xff, 0xfe, 0x00, 0x03, 0x00,
> +
>  	/* SOS */
>  	0xff, 0xda, 0x00, 0x0c, 0x03, 0x01, 0x00, 0x02,
>  	0x11, 0x03, 0x11, 0x00, 0x3f, 0x00,
> diff --git a/drivers/staging/media/hantro/hantro_jpeg.h b/drivers/staging/media/hantro/hantro_jpeg.h
> index f33c492134e4..0b49d0b82caa 100644
> --- a/drivers/staging/media/hantro/hantro_jpeg.h
> +++ b/drivers/staging/media/hantro/hantro_jpeg.h
> @@ -1,6 +1,6 @@
>  /* SPDX-License-Identifier: GPL-2.0+ */
>  
> -#define JPEG_HEADER_SIZE	619
> +#define JPEG_HEADER_SIZE	624

Can we add some compile-time check for the 8-byte alignment,
so this is always enforced?

Perhaps getting rid of the JPEG_HEADER_SIZE macro,
something like this....


@@ -140,7 +140,7 @@ static const unsigned char chroma_ac_table[] = {
  * and we'll use fixed offsets to change the width, height
  * quantization tables, etc.
  */
-static const unsigned char hantro_jpeg_header[JPEG_HEADER_SIZE] = {
+static const unsigned char hantro_jpeg_header[] = {
        /* SOI */
        0xff, 0xd8,

@@ -304,8 +304,13 @@ void hantro_jpeg_header_assemble(struct hantro_jpeg_ctx *ctx)
 {
        char *buf = ctx->buffer;

-       memcpy(buf, hantro_jpeg_header,
-              sizeof(hantro_jpeg_header));
+       /*
+        * THE JPEG buffer is prepended with the JPEG header,
+        * so 64-bit alignment is needed for DMA.
+        */
+       BUILD_BUG_ON(!IS_ALIGNED(sizeof(hantro_jpeg_header), 8));
+
+       memcpy(buf, hantro_jpeg_header, sizeof(hantro_jpeg_header));

Thanks,
Ezequiel