Re: new FFMpeg stuff

[Michael Niedermayer <michaelni () gmx at>]

Hi

On Tue, Jul 09, 2013 at 06:49:34AM +0200, Moritz Muehlenhoff wrote:
> Kurt Seifried wrote:
>
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > https://bugs.gentoo.org/show_bug.cgi?id=476218
> >
> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=38229362529ed1619d8ebcc81ecde85b23b45895

This should have been fixed by b21ba20cc83c80fe56192fee3626a8087f37d806 in ffmpeg (Apr 22 2012)


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=e30b068ef79f604ff439418da07f7e2efd01d4ea

This should have been fixed by 780d45473c32fa356c8ce385c3ea4692567c3228 in ffmpeg (Sep 24 2011)


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=6765ee7b9cba46818a45b051438b2552f0a1b70a

This seems listed as buffer overflow but as far as i can tell it fixes
just a null pointer dereference. If you want to assign CVEs to all
null pointer dereferences and out of array reads that got fixed then
quiete a few more CVEs are needed.

Also see: a9456c7c5ca883b5a3947e59a9fba5587e18e119


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=b36e1893ef3430f039c1eaddeedcbb378f9c4444

This was fixed in 4b35ee0b7c0c4cbac3541a25a5e8c00b657c8f95 in ffmpeg (Dec 28 2011)


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=7388c0c58601477db076e2e74e8b11f8a644384a


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=95a57d26d8653d21f0dab1aff3558ee944853dbf

This was fixed in c49d94487c6135325930cbc4a8cd96d38ef6653e in ffmpeg (Jun 6 2013)
Note, this issue shouldnt affect any ffmpeg releases as the code was
added more recently


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=b564784a207b1395d2b5a41e580539df04651096

Same as above jpeg2000dec.c wasnt in any releases yet as of today,
what was in the releases was j2kdec.c but that was marked as
experimental


> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=78962d3df49afe5011b572656ecfe940bd5fbf2e
> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=cf04af2086be105ff86088357b83d672d38417d9
> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=eae63e3c156f784ee0612422f0c95131ea913c14
> > http://git.videolan.org/?p=ffmpeg.git;a=commit;h=fd54dd028bc9f7bfb80ebf823a533dc84b73f936

Same as above



> > Correct me if I'm wrong but most of these seem to deserve CVEs and
> > none have been assigned, correct?
> >
> > http://ffmpeg.org/security.html
>
> These appear to be new, but I'm not sure how previous CVE IDs were assigned for ffmpeg/libav.
> E.g. CVE-2013-0878 seems to be from a Google CNA, right? (At least CVE-2013-0879 is for Chrome)
>
> All these issues (and all the ones in previous rounds) were found through fuzzing done
> at Google by Mateusz "j00ru" Jurczyk and Gynvael Coldwind.

I dont know about the libav side, for the ffmpeg side CVEs where
provided by "google" for all serious issues that where found. Which
issues where serious could in general only be assesed after the issues
where fixed so values where available only after the fixes where
commited.


> It would be very, very welcome if CVE assignments from either ffmpeg or libav for any
> such issues would have a reference to the filename of the fuzzed file triggering the problem.

> With the diverging code bases between ffmpeg and libav [1] it becomes very complicated
> to properly track down if one of the two is affected.

yes, its a big headache for us as well. Especialy for me as iam always
merging all improvments and fixes from libav into ffmpeg ...

[...]

Thanks

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Observe your enemies, for they first find out your faults. -- Antisthenes

Attachment: signature.asc
Description: Digital signature