[FFmpeg-devel] [PATCH] avcodec/mpegaudiodec_template: Check CRCs for layer1 and layer2

[Lynne]

Aug 3, 2020, 18:31 by michael at niedermayer.cc:

> This differs from the MPEG specification as the actual real world
> files do compute their CRC over variable areas and not the fixed
> ones listed in the specification. This is also the reason for
> the complexity of this code and the need to perform the CRC
> check for layer2 in the middle of layer2 decoding.
>
> Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> ---
>  libavcodec/mpegaudiodec_template.c | 61 +++++++++++++++++++++---------
>  1 file changed, 44 insertions(+), 17 deletions(-)
>
> diff --git a/libavcodec/mpegaudiodec_template.c b/libavcodec/mpegaudiodec_template.c
> index 3d7e3ba4f2..d84e4f1cb6 100644
> --- a/libavcodec/mpegaudiodec_template.c
> +++ b/libavcodec/mpegaudiodec_template.c
> @@ -89,6 +89,7 @@ typedef struct MPADecodeContext {
>  MPADSPContext mpadsp;
>  AVFloatDSPContext *fdsp;
>  AVFrame *frame;
> +    int crc;
>

Make this a uint32_t, its what we always store CRCs as even if they're 8 bits.
Thought it was a flag on first read.



>  } MPADecodeContext;
>  
>  #define HEADER_SIZE 4
> @@ -500,12 +501,43 @@ static void imdct12(INTFLOAT *out, SUINTFLOAT *in)
>  out[11] = in0 + in5;
>  }
>  
> +static int handle_crc(MPADecodeContext *s, int sec_len)
> +{
> +    if (s->error_protection && (s->err_recognition & AV_EF_CRCCHECK)) {
> +        const uint8_t *buf = s->gb.buffer - HEADER_SIZE;
> +        int sec_byte_len  = sec_len >> 3;
> +        int sec_rem_bits  = sec_len & 7;
> +        const AVCRC *crc_tab = av_crc_get_table(AV_CRC_16_ANSI);
> +        uint8_t tmp_buf[4];
> +        uint32_t crc_val = av_crc(crc_tab, UINT16_MAX, &buf[2], 2);
> +        crc_val = av_crc(crc_tab, crc_val, &buf[6], sec_byte_len);
> +
> +        AV_WB32(tmp_buf,
> +            ((buf[6 + sec_byte_len] & (0xFF00>>sec_rem_bits))<<24) +
> +            ((unsigned)(s->crc<<16) >> sec_rem_bits));
>

If s->crc is a uint32_t then you don't need to cast it here.
Also argument alignment seems wrong, we align each new line of a function
call/macro with the starting bracket rather than just 4 spaces.
And some spaces in between the shift operators would be nice.


> +        crc_val = av_crc(crc_tab, crc_val, tmp_buf, 3);
> +
> +        if (crc_val) {
> +            av_log(s->avctx, AV_LOG_ERROR, "CRC mismatch %X!\n", crc_val);
>

Good idea, printing the difference was very useful when I was debugging,
we should do that everywhere.


How this works was a little confusing at first but makes sense for me,
so apart from those comments patch LGTM, thanks.