Re: [FFmpeg-devel] avformat/mxfenc: SMPTE RDD 48:2018 Amd 1:2022 (FFV1 in MXF) support

[Michael Niedermayer <michael@niedermayer.cc>]

[-- Attachment #1.1: Type: text/plain, Size: 6358 bytes --]

On Mon, Apr 03, 2023 at 12:07:06AM +0200, Jerome Martinez wrote:
> On 02/04/2023 22:07, Michael Niedermayer wrote:
> > On Sat, Apr 01, 2023 at 07:11:45PM +0200, Jerome Martinez wrote:
> > [...]
> > >   mxfenc.c |   10 ++++++----
> > >   1 file changed, 6 insertions(+), 4 deletions(-)
> > > 2b3a2cc619fcaf2dd1da7aa4b566b390eced62a3  0001-avformat-mxfenc-replace-ffv1-version-related-assert-.patch
> > >  From 65f6fba9434fa9ab4b5d2b91c3f930c1e3cb4bef Mon Sep 17 00:00:00 2001
> > > From: Jerome Martinez <jerome@mediaarea.net>
> > > Date: Sat, 1 Apr 2023 19:04:25 +0200
> > > Subject: [PATCH] avformat/mxfenc: replace ffv1 version related assert by error
> > >   message
> > > 
> > > ---
> > >   libavformat/mxfenc.c | 10 ++++++----
> > >   1 file changed, 6 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/libavformat/mxfenc.c b/libavformat/mxfenc.c
> > > index 9eba208ffb..0a1ae5353c 100644
> > > --- a/libavformat/mxfenc.c
> > > +++ b/libavformat/mxfenc.c
> > > @@ -2489,8 +2489,8 @@ static int mxf_parse_ffv1_frame(AVFormatContext *s, AVStream *st, AVPacket *pkt)
> > >           ff_init_range_decoder(&c, st->codecpar->extradata, st->codecpar->extradata_size);
> > >           ff_build_rac_states(&c, 0.05 * (1LL << 32), 256 - 8);
> > >           v = get_ffv1_unsigned_symbol(&c, state);
> > > -        av_assert0(v >= 2);
> > > -        if (v > 4) {
> > > +        if (v < 2) {
> > > +            av_log(s, AV_LOG_ERROR, "Invalid version in ffv1 global header\n");
> > >               return 0;
> > >           }
> > should this not also fail with v > 4 as before ?
> 
> I preferred during the change to copy the code in ffv1dec.c for more
> coherency, and ffv1dec.c considers version 5 in ver01 header as invalid and
> version 5 with extra metadata as legit.
> Actually, a bigger issue is that ffv1dec.c accepts version 5 as if it can
> decode it, e.g. with hacked ffv1 stream having version 5:
> 
> $ ./ffmpeg -hide_banner -loglevel error -i level5_like_level1.mkv -f null
> /dev/null 2> >(grep -i  "version")
> [ffv1 @ 0x7fffc04c81c0] invalid version 5 in ver01 header
> $ mediainfo --Details=1 level5_like_level3.mkv | grep "version"
> 1B1        version:                         5 (0x00000005) -
> Error=FFV1-HEADER-version-LATERVERSION:1
> $ ./ffmpeg -hide_banner -loglevel error -i level5_like_level3.mkv -f null
> /dev/null && echo $?
> 0
> 
> It works only because my hacked file is a version 4 in reality but it should
> be rejected for when we have a version 5 not compatible with version 4.
> I suggest to keep the previous patch as is for coherency with ffv1dec.c then
> add the new attached patch that add a check on ffv1 version > 4 for both
> ffv1dec.c and mxfenc.c; for ffv1dec.c I also check micro_version for version
> 4 because FFV1 spec hints that micro_version before the one flagged as
> stable may be incompatible with previous micro_versions.
> With the additional patch:
> $ ./ffmpeg -hide_banner -loglevel error -i level5_like_level3.mkv -f null
> /dev/null
> [ffv1 @ 0x7fffc290e640] unsupported version 5
> Error while opening decoder for input stream #0:0 : Not yet implemented in
> FFmpeg, patches welcome
> $ ./ffmpeg -hide_banner -loglevel error -i level4.3.mkv -f null /dev/null
> [ffv1 @ 0x7fffc5b27640] unsupported version 4 micro_version 3
> Error while opening decoder for input stream #0:0 : Not yet implemented in
> FFmpeg, patches welcome
> $ ./ffmpeg -hide_banner -loglevel error -i level5_like_level3.mkv -r 25 -c:v
> copy -f mxf /dev/null
> [ffv1 @ 0x7fffe422bf00] unsupported version 5
> [mxf @ 0x7fffda8b4080] unsupported ffv1 version 5
> [out#0/mxf @ 0x7fffd34470c0] Error muxing a packet
> 
> Jérôme
> 

>  libavcodec/ffv1dec.c |   10 ++++++++++
>  libavformat/mxfenc.c |    4 ++++
>  2 files changed, 14 insertions(+)
> 14eb2ae57e1918a903d379640e466bdcb5a73a41  0001-avcodec-ffv1dec-reject-unsupported-ffv1-versions.patch
> From 4434ebddb53524f52346b4fbb1126489d8864810 Mon Sep 17 00:00:00 2001
> From: Jerome Martinez <jerome@mediaarea.net>
> Date: Mon, 3 Apr 2023 00:04:53 +0200
> Subject: [PATCH] avcodec/ffv1dec: reject unsupported ffv1 versions
> 
> And add similar check in libavformat/mxfenc
> ---
>  libavcodec/ffv1dec.c | 10 ++++++++++
>  libavformat/mxfenc.c |  4 ++++
>  2 files changed, 14 insertions(+)
> 
> diff --git a/libavcodec/ffv1dec.c b/libavcodec/ffv1dec.c
> index 180d24e695..7aaee7ec5e 100644
> --- a/libavcodec/ffv1dec.c
> +++ b/libavcodec/ffv1dec.c
> @@ -439,11 +439,21 @@ static int read_extra_header(FFV1Context *f)
>          av_log(f->avctx, AV_LOG_ERROR, "Invalid version in global header\n");
>          return AVERROR_INVALIDDATA;
>      }
> +    if (f->version > 4) {
> +        av_log(f->avctx, AV_LOG_ERROR, "unsupported version %d\n",
> +            f->version);
> +        return AVERROR_PATCHWELCOME;
> +    }
>      if (f->version > 2) {
>          c->bytestream_end -= 4;
>          f->micro_version = get_symbol(c, state, 0);
>          if (f->micro_version < 0)
>              return AVERROR_INVALIDDATA;
> +        if (f->version == 4 && f->micro_version > 2) {
> +            av_log(f->avctx, AV_LOG_ERROR, "unsupported version 4 micro_version %d\n",
> +                f->micro_version);
> +            return AVERROR_PATCHWELCOME;
> +        }
>      }
>      f->ac = get_symbol(c, state, 0);

you do not know if the decoder will have any problem with these files



>  
> diff --git a/libavformat/mxfenc.c b/libavformat/mxfenc.c
> index 0a1ae5353c..df5bb55ad8 100644
> --- a/libavformat/mxfenc.c
> +++ b/libavformat/mxfenc.c
> @@ -2493,6 +2493,10 @@ static int mxf_parse_ffv1_frame(AVFormatContext *s, AVStream *st, AVPacket *pkt)
>              av_log(s, AV_LOG_ERROR, "Invalid version in ffv1 global header\n");
>              return 0;
>          }
> +        if (v > 4) {
> +            av_log(s, AV_LOG_ERROR, "unsupported ffv1 version %d\n", v);
> +            return 0;
> +        }
>          sc->micro_version = get_ffv1_unsigned_symbol(&c, state);
>      } else {
>          uint8_t keystate = 128;

mxfenc change is ok

thx

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The misfortune of the wise is better than the prosperity of the fool.
-- Epicurus

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 251 bytes --]

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".