[FFmpeg-devel] [PATCH 5/8] avcodec/hevc_ps: check scaling_list_dc_coef

[James Almer]

On 12/18/2020 8:22 PM, Michael Niedermayer wrote:
> Fixes: signed integer overflow: 2147483640 + 8 cannot be represented in type 'int'
> Fixes: 28449/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HEVC_fuzzer-5686013259284480
> 
> Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> ---
>   libavcodec/hevc_ps.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
> index ea6fd536c6..139f3deeda 100644
> --- a/libavcodec/hevc_ps.c
> +++ b/libavcodec/hevc_ps.c
> @@ -816,7 +816,11 @@ static int scaling_list_data(GetBitContext *gb, AVCodecContext *avctx, ScalingLi
>                   next_coef = 8;
>                   coef_num  = FFMIN(64, 1 << (4 + (size_id << 1)));
>                   if (size_id > 1) {
> -                    scaling_list_dc_coef[size_id - 2][matrix_id] = get_se_golomb(gb) + 8;
> +                    int scaling_list_coeff_minus8 = get_se_golomb(gb);
> +                    if (scaling_list_coeff_minus8 < -7 ||
> +                        scaling_list_coeff_minus8 > 247)

LGTM, this check is also done in CBS so i assume it's taken from the spec.

> +                        return AVERROR_INVALIDDATA;
> +                    scaling_list_dc_coef[size_id - 2][matrix_id] = scaling_list_coeff_minus8 + 8;

Not strictly related to this patch, but this array could be made into an 
uint8_t now.

>                       next_coef = scaling_list_dc_coef[size_id - 2][matrix_id];
>                       sl->sl_dc[size_id - 2][matrix_id] = next_coef;
>                   }
>