[FFmpeg-devel] [PATCH] libavformat/mov.c : Avoid seek request with invalid seek offset ( 9223372036854775799 ) after EOS

[Malviya, Janpriya]

Hello zhilizhao , 

Thank you for help and support. 

When are all three patches going to merge in master?

Regards 
Janpriya.

-----Original Message-----
From: ffmpeg-devel <ffmpeg-devel-bounces at ffmpeg.org> On Behalf Of "zhilizhao(???)"
Sent: 26 April 2022 01:58 PM
To: FFmpeg development discussions and patches <ffmpeg-devel at ffmpeg.org>
Cc: Banerjee, Debasmit <Debasmit_Banerjee at bose.com>; Patel, Pratik <Pratik_Patel at bose.com>; Van Iderstine, David <David_Van_Iderstine at bose.com>
Subject: Re: [FFmpeg-devel] [PATCH] libavformat/mov.c : Avoid seek request with invalid seek offset ( 9223372036854775799 ) after EOS



> On Apr 22, 2022, at 4:37 PM, Malviya, Janpriya <Janpriya_Malviya at bose.com> wrote:
> 
> Hello ffmpeg Dev team,
> 
> While integrating FFmpeg for M4A , MP4 stream & facing issue that mov demuxer trying to request seek for invalid seek offset.
> 
> Issue details:-
> 	- Modified doc/examples/demuxing_decoding.c example to support Custom IO call-back because we have our  own mechanism to read data from source
> 	-  In custom IO seek call back , we do not support “whence” values AVSEEK_SIZE & SEEK_END , used to find stream size ( returning -1 in both the cases ).
> 	- In attached screen shot , where we are observing seek with invalid seek offset and the seek requested after End of stream is detected
> 	- Seek offset value 9223372036854775799 always the same ( I think INT64_MAX – 8 ,  8 is length Atom type & size field )
> 	- If we allow to return proper stream size in seek callback then this problem is not observed.

The real issue isn’t about invalid seek offset, but use invalid box size/type due to EOF.

Add another check just after the eof check in while condition isn’t efficient. I tried another method:

[1/3] avformat/mov: remove an always true condition https://urldefense.com/v3/__http://ffmpeg.org/pipermail/ffmpeg-devel/2022-April/295712.html__;!!I6pijIezNA!y_KUjRpt0ie8shWRhndmGEosSUA0XTPqZLDL6wo1oiEcGS9zlqkjvgeW02KIZRriFN_4zSOSF55eUMBwYFpEi8BJ9_E$ 

[2/3] reindent

[3/3] avformat/mov: fix use invalid box size/type due to eof https://urldefense.com/v3/__http://ffmpeg.org/pipermail/ffmpeg-devel/2022-April/295713.html__;!!I6pijIezNA!y_KUjRpt0ie8shWRhndmGEosSUA0XTPqZLDL6wo1oiEcGS9zlqkjvgeW02KIZRriFN_4zSOSF55eUMBwYFpE8kolcu0$ 

> 
> Fix :
> In attached patch to verify EOS condition before going to parse atom data.
> 
> Signed-off-by: Janpriya Malviya <Janpriya_Malviya at bose.com>
> ---
> libavformat/mov.c | 4 ++++
> 1 file changed, 4 insertions(+)
> 
> diff --git a/libavformat/mov.c b/libavformat/mov.c index 
> 6c847de..9704550 100644
> --- a/libavformat/mov.c
> +++ b/libavformat/mov.c
> @@ -7556,6 +7556,10 @@ static int mov_read_default(MOVContext *c, AVIOContext *pb, MOVAtom atom)
>         if (atom.size >= 8) {
>             a.size = avio_rb32(pb);
>             a.type = avio_rl32(pb);
> +
> +            if(avio_feof(pb))
> +                continue;
> +
>             if (((a.type == MKTAG('f','r','e','e') && c->moov_retry) ||
>                   a.type == MKTAG('h','o','o','v')) &&
>                 a.size >= 8 &&
> --
> 2.7.4
> 
> <issueImage.png><reproduce_issue.c>___________________________________
> ____________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://urldefense.com/v3/__https://ffmpeg.org/mailman/listinfo/ffmpeg
> -devel__;!!I6pijIezNA!y_KUjRpt0ie8shWRhndmGEosSUA0XTPqZLDL6wo1oiEcGS9z
> lqkjvgeW02KIZRriFN_4zSOSF55eUMBwYFpE04pB2pw$
> 
> To unsubscribe, visit link above, or email 
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
> 

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel at ffmpeg.org
https://urldefense.com/v3/__https://ffmpeg.org/mailman/listinfo/ffmpeg-devel__;!!I6pijIezNA!y_KUjRpt0ie8shWRhndmGEosSUA0XTPqZLDL6wo1oiEcGS9zlqkjvgeW02KIZRriFN_4zSOSF55eUMBwYFpE04pB2pw$ 

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