[FFmpeg-devel] [PATCH] Vorbis-in-Ogg: Do not set timebase to invalid values

[Jason Garrett-Glaser]

On Fri, Jan 28, 2011 at 10:22 PM, Reimar D?ffinger
<Reimar.Doeffinger at gmx.de> wrote:
> On 29 Jan 2011, at 05:42, M?ns Rullg?rd <mans at mansr.com> wrote:
>> Janne Grunau <janne-ffmpeg at jannau.net> writes:
>>
>>> From: Reimar D?ffinger <Reimar.Doeffinger at gmx.de>
>>>
>>> cherry picked from git.videolan.org repo
>>>
>>> Janne
>>> ---8<---
>>> Avoids an assert when the sample rate is invalid and the timebase
>>> is thus set to e.g. 1/0.
>>> Sample file is http://samples.mplayerhq.hu/ogg/fuzzed-srate-crash.ogg
>>> ---
>>> libavformat/oggparsevorbis.c | ? 10 +++++++---
>>> 1 files changed, 7 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/libavformat/oggparsevorbis.c b/libavformat/oggparsevorbis.c
>>> index cdb0266..d743d25 100644
>>> --- a/libavformat/oggparsevorbis.c
>>> +++ b/libavformat/oggparsevorbis.c
>>> @@ -221,6 +221,7 @@ vorbis_header (AVFormatContext * s, int idx)
>>> ? ? if (os->buf[os->pstart] == 1) {
>>> ? ? ? ? const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
>>> ? ? ? ? unsigned blocksize, bs0, bs1;
>>> + ? ? ? ?int srate;
>>>
>>> ? ? ? ? if (os->psize != 30)
>>> ? ? ? ? ? ? return -1;
>>> @@ -229,7 +230,7 @@ vorbis_header (AVFormatContext * s, int idx)
>>> ? ? ? ? ? ? return -1;
>>>
>>> ? ? ? ? st->codec->channels = bytestream_get_byte(&p);
>>> - ? ? ? ?st->codec->sample_rate = bytestream_get_le32(&p);
>>> + ? ? ? ?srate = bytestream_get_le32(&p);
>>> ? ? ? ? p += 4; // skip maximum bitrate
>>> ? ? ? ? st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
>>> ? ? ? ? p += 4; // skip minimum bitrate
>>> @@ -249,8 +250,11 @@ vorbis_header (AVFormatContext * s, int idx)
>>> ? ? ? ? st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
>>> ? ? ? ? st->codec->codec_id = CODEC_ID_VORBIS;
>>>
>>> - ? ? ? ?st->time_base.num = 1;
>>> - ? ? ? ?st->time_base.den = st->codec->sample_rate;
>>> + ? ? ? ?if (srate > 0) {
>>> + ? ? ? ? ? ?st->codec->sample_rate = srate;
>>> + ? ? ? ? ? ?st->time_base.num = 1;
>>> + ? ? ? ? ? ?st->time_base.den = srate;
>>> + ? ? ? ?}
>>> ? ? } else if (os->buf[os->pstart] == 3) {
>>> ? ? ? ? if (os->psize > 8)
>>> ? ? ? ? ? ? ff_vorbis_comment (s, &st->metadata, os->buf + os->pstart + 7, os->psize - 8);
>>> --
>>> 1.7.4.rc2
>>
>> I still want to know why common code doesn't catch this. ?Replicating
>> this check in each and every demuxer is insane.
>
> I already explained it and nobody cared.
> I also gave a reason why, even if the common code would catch it this is of advantage.
> In addition relying on common code is risky, streams can be added at any time, demuxers might parse headers even if they are in the middle of the stream, unless data is validated immediately (i.e. in the demuxer) there is a significant risk that it will end up outside FFmpeg, in the application.
> However common code could set the timebase properly/better if the decoder figures out the sample rate, but I don't feel like investigating where to do this.

Maybe it could be done in the wrapper functions, e.g. the utility
functions, not per-demuxer?  So it's still done in libav*, but not in
each codec repeatedly.

Jason