[FFmpeg-devel] [PATCH 3/4] avformat/mxfdec: Check for avio_read() failure in mxf_read_strong_ref_array()

[Michael Niedermayer]

On Mon, Mar 14, 2022 at 06:08:32PM +0100, Marton Balint wrote:
> 
> 
> On Mon, 14 Mar 2022, Michael Niedermayer wrote:
> 
> > On Sun, Mar 13, 2022 at 04:52:25PM +0100, Marton Balint wrote:
> > > 
> > > 
> > > On Sun, 13 Mar 2022, Michael Niedermayer wrote:
> > > 
> > > > Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> > > > ---
> > > > libavformat/mxfdec.c | 8 +++++++-
> > > > 1 file changed, 7 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/libavformat/mxfdec.c b/libavformat/mxfdec.c
> > > > index d7cdd22c8a..828fc0f9f1 100644
> > > > --- a/libavformat/mxfdec.c
> > > > +++ b/libavformat/mxfdec.c
> > > > @@ -932,6 +932,7 @@ static int mxf_read_cryptographic_context(void *arg, AVIOContext *pb, int tag, i
> > > > 
> > > > static int mxf_read_strong_ref_array(AVIOContext *pb, UID **refs, int *count)
> > > > {
> > > > +    int64_t ret;
> > > >     unsigned c = avio_rb32(pb);
> > > > 
> > > >     //avio_read() used int
> > > > @@ -946,7 +947,12 @@ static int mxf_read_strong_ref_array(AVIOContext *pb, UID **refs, int *count)
> > > >         return AVERROR(ENOMEM);
> > > >     }
> > > >     avio_skip(pb, 4); /* useless size of objects, always 16 according to specs */
> > > > -    avio_read(pb, (uint8_t *)*refs, *count * sizeof(UID));
> > > > +    ret = avio_read(pb, (uint8_t *)*refs, *count * sizeof(UID));
> > > > +    if (ret != *count * sizeof(UID)) {
> > > > +        *count = ret < 0 ? 0   : ret / sizeof(UID);
> > > 
> > 
> > > I suggest you hard fail if the read count is not the expected, do not
> > > silently ignore corrupt file.
> > > 
> > > Regards,
> > > Marton
> > > 
> > > > +        return   ret < 0 ? ret : AVERROR_INVALIDDATA;
> > 
> > Does it not hard fail here ?
> 
> Yeah, it does, sorry. This extra count calculation confused me... I'd just
> probably set it to 0 in case of a partial read, same as in case of an error,
> but fine either way I guess.

i wanted to set the count to be most correct value. So if someone was
debuging this and tried to just not fail it would already have a useable
value in count.

thx
[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

it is not once nor twice but times without number that the same ideas make
their appearance in the world. -- Aristotle
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20220315/d2f6db17/attachment.sig>