[FFmpeg-devel] [PATCH] avformat/options_table: Set the default maximum number of streams to 100

wm4 nfxjfg at googlemail.com
Sun Dec 11 17:13:42 EET 2016


On Sun, 11 Dec 2016 15:29:32 +0100
Hendrik Leppkes <h.leppkes at gmail.com> wrote:

> On Sat, Dec 10, 2016 at 11:36 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
> > Hi,
> >
> > On Sat, Dec 10, 2016 at 5:24 PM, Carl Eugen Hoyos <ceffmpeg at gmail.com>
> > wrote:
> >  
> >> 2016-12-10 14:07 GMT+01:00 Ronald S. Bultje <rsbultje at gmail.com>:  
> >> > Hi,
> >> >
> >> > On Sat, Dec 10, 2016 at 7:11 AM, Carl Eugen Hoyos <ceffmpeg at gmail.com>
> >> > wrote:
> >> >  
> >> >> 2016-12-09 12:56 GMT+01:00 Ronald S. Bultje <rsbultje at gmail.com>:
> >> >>  
> >> >> > On IRC, we discussed at what values OOM start occurring, which
> >> >> > seems to be around 30k-60k  
> >> >>
> >> >> This is not true, why do you think so?  
> >> >
> >> > http://lists.ffmpeg.org/pipermail/ffmpeg-devel-irc/  
> >> 2016-December/003980.html
> >>
> >> http://www.openwall.com/lists/oss-security/2016/12/08/1
> >> Iiuc (which isn't sure at all) this is about 26000 streams.  
> >
> >
> > And how does that fundamentally change the discussion? The question was
> > about orders of magnitude (think powers of 10), so you just changes the
> > answer from log10(30k)=4.48 to log10(26k)=4.41. In both cases, a sensible
> > limit is something like exp10(3) or exp10(4), but exp10(2) is not necessary
> > IMHO,
> >  
> 
> If we must have a pre-defined limit (which I still find a questionable
> choice to limit resource usage in general, OOM itself is not an
> inherently "evil" thing, perfectly valid files could cause OOM just
> due to sheer size and if OOM is a crash then something else needs
> fixing to check mallocs),

+1

> then 1000 (ie. exp10(3)) (or 1024 because we
> like power of 2 numbers) is probably going to be enough for all sorts
> of valid files. Even odd mpegts files with loads of empty streams in
> empty programs would never reach that.

Subtitle streams, font attachments, attached images, and other things
could also inflate the number of streams, so 1000 seems a bit low.


More information about the ffmpeg-devel mailing list