[FFmpeg-devel] [PATCH] avcodec/mpegpicture: guard "stride changed" failures against unknown uvlinesize

[Aman Gupta]

On Tue, Apr 10, 2018 at 3:15 PM, Michael Niedermayer <michael at niedermayer.cc
> wrote:

> On Tue, Apr 10, 2018 at 10:20:07AM -0700, Aman Gupta wrote:
> > From: Aman Gupta <aman at tmm1.net>
> >
> > Before adding uvlinesize check, I was seeing failures decoding
> > some samples (shown with extra logging added):
> >
> > [mpeg2video @ 0x7fa193818c00] get_buffer() failed (stride changed:
> linesize=1280/1280 uvlinesize=0/640)
> > [mpeg2video @ 0x7fa193818c00] get_buffer() failed (stride changed:
> linesize=1280/1280 uvlinesize=0/640)
> > ---
> >  libavcodec/mpegpicture.c | 9 ++++++---
> >  1 file changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/libavcodec/mpegpicture.c b/libavcodec/mpegpicture.c
> > index 2be670cdbc..80898c161c 100644
> > --- a/libavcodec/mpegpicture.c
> > +++ b/libavcodec/mpegpicture.c
> > @@ -148,10 +148,13 @@ static int alloc_frame_buffer(AVCodecContext
> *avctx,  Picture *pic,
> >          }
> >      }
> >
> > -    if (linesize && (linesize   != pic->f->linesize[0] ||
> > -                     uvlinesize != pic->f->linesize[1])) {
> > +    if (linesize && uvlinesize &&
> > +        (linesize   != pic->f->linesize[0] ||
> > +         uvlinesize != pic->f->linesize[1])) {
>
> without checking if this can be reached with grayscale.
> If it can then uvlinesize may be 0 in which case this change would be
> wrong
>

I forgot to mention that I was only seeing this bug with gray mode
(--enable-gray + AV_CODEC_FLAG_GRAY), and that after this patch it works as
expected.

Aman


>
> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> Observe your enemies, for they first find out your faults. -- Antisthenes
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
>