[FFmpeg-devel] [PATCH 4/7] avformat/nistspheredec: Check bits_per_coded_sample and channels

[Anton Khirnov]

Quoting Michael Niedermayer (2021-01-19 19:00:47)
> On Mon, Jan 18, 2021 at 09:06:10PM +0100, Anton Khirnov wrote:
> > Quoting Michael Niedermayer (2021-01-17 00:07:26)
> > > Fixes: signed integer overflow: 80 * 92233009 cannot be represented in type 'int'
> > > Fixes: 26910/clusterfuzz-testcase-minimized-ffmpeg_dem_NISTSPHERE_fuzzer-6669100654919680
> > > 
> > > Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
> > > Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> > > ---
> > >  libavformat/nistspheredec.c | 4 +++-
> > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/libavformat/nistspheredec.c b/libavformat/nistspheredec.c
> > > index 079369929f..fec5c88892 100644
> > > --- a/libavformat/nistspheredec.c
> > > +++ b/libavformat/nistspheredec.c
> > > @@ -80,7 +80,9 @@ static int nist_read_header(AVFormatContext *s)
> > >  
> > >              avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
> > >  
> > > -            st->codecpar->block_align = st->codecpar->bits_per_coded_sample * st->codecpar->channels / 8;
> > > +            if (st->codecpar->bits_per_coded_sample * (uint64_t)st->codecpar->channels / 8 > INT_MAX)
> > > +                return AVERROR_INVALIDDATA;
> > > +            st->codecpar->block_align = st->codecpar->bits_per_coded_sample * (uint64_t)st->codecpar->channels / 8;
> > >  
> > >              if (avio_tell(s->pb) > header_size)
> > >                  return AVERROR_INVALIDDATA;
> > > -- 
> > > 2.17.1
> > 
> > Both bits_per_coded_sample and channels seem to be arbitrary ints, so
> > you are assuming sizeof(int) < sizeof(int64)
> 
> I wonder if we shouldnt limit these 2 fields in a platform independant way
> But if we dont then the change below should do more correct checks

I would be in favor of limiting them both to 16 bits, that should be way
beyond anything needed by any reasonable use case.

> 
> @@ -80,7 +80,11 @@ static int nist_read_header(AVFormatContext *s)
>  
>              avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
>  
> -            st->codecpar->block_align = st->codecpar->bits_per_coded_sample * st->codecpar->channels / 8;
> +            if (st->codecpar->bits_per_coded_sample < 0 || st->codecpar->channels <= 0 ||
> +                st->codecpar->bits_per_coded_sample > UINT64_MAX / st->codecpar->channels ||
> +                st->codecpar->bits_per_coded_sample * (uint64_t)st->codecpar->channels / 8 > INT_MAX)
> +                return AVERROR_INVALIDDATA;
> +            st->codecpar->block_align = st->codecpar->bits_per_coded_sample * (uint64_t)st->codecpar->channels / 8;
>  
>              if (avio_tell(s->pb) > header_size)
>                  return AVERROR_INVALIDDATA;

This looks ok.

Though it would be even better to sanity-check each value for
zero/negative right after the sscanf() that reads it (not insisting you
do it).

-- 
Anton Khirnov