[FFmpeg-devel] [PATCH 2/2] avformat/mpc8: fix hang with fuzzed file

[wm4]

On Tue, 3 Feb 2015 22:00:11 +0100
Reimar Döffinger <Reimar.Doeffinger at gmx.de> wrote:

> On Tue, Feb 03, 2015 at 09:54:51PM +0100, wm4 wrote:
> > On Tue, 3 Feb 2015 21:47:57 +0100
> > Reimar Döffinger <Reimar.Doeffinger at gmx.de> wrote:
> > 
> > > On Tue, Feb 03, 2015 at 07:04:12PM +0100, wm4 wrote:
> > > > This can lead to an endless loop by seeking back a few bytes after each
> > > > attempted chunk read. Assuming negative sizes are always invalid, this
> > > > is easy to fix. Other code in this demuxer treats negative sizes as
> > > > invalid as well.
> > > > 
> > > > Fixes ticket #4262.
> > > > ---
> > > >  libavformat/mpc8.c | 4 ++++
> > > >  1 file changed, 4 insertions(+)
> > > > 
> > > > diff --git a/libavformat/mpc8.c b/libavformat/mpc8.c
> > > > index d6ca338..6524c7e 100644
> > > > --- a/libavformat/mpc8.c
> > > > +++ b/libavformat/mpc8.c
> > > > @@ -223,6 +223,10 @@ static int mpc8_read_header(AVFormatContext *s)
> > > >      while(!avio_feof(pb)){
> > > >          pos = avio_tell(pb);
> > > >          mpc8_get_chunk_header(pb, &tag, &size);
> > > > +        if (size < 0) {
> > > 
> > > Isn't the only way for this to become negative for a too
> > > large uint64_t to be assigned to a int64_t?
> > > I.e. undefined behaviour.
> > > In that case this isn't quite the right way in the strictest sense,
> > > though it is likely to work "normally".
> > 
> > This is what mpc8_get_chunk_header() does:
> > 
> >     pos = avio_tell(pb);
> >     *tag = avio_rl16(pb);
> >     *size = ffio_read_varlen(pb);
> >     *size -= avio_tell(pb) - pos;
> > 
> > ffio_read_varlen() returns uint64_t, so I don't think it's supposed to
> > read negative values. But the last line could make it negative anyway
> > if the chunk size field read here is inconsistent.
> 
> The problem is in this line:
> *size = ffio_read_varlen(pb);
> *size is int64_t.
> If ffio_read_varlen returns a value > 0x7f..ffull then this
> assignment is undefined, which a fairly risky thing to allow
> for a security-relevant check.

AFAIK it's implementation-defined, rather than undefined. Which means
it works on all CPUs with complement-of-2 integers anyway.

If you prefer, I can change them all to unsigned...

> That is ignoring the fact that I am not exactly happy that
> ffio_read_varlen will happily read until EOF of a multi-GB file,
> which doesn't seem like reasonable behaviour at least in this
> use-case.

OK, that's a separate problem though.