Mailing List archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[vdr] Re: vdr crashes



I was tuned to a channel for several hours on one provider, then tuned to
the "special" channel on the other provider.

The definition of the "special" channel is one of a dozen on a special
transponder that contains 48 hours of EPG data for all information for
this provider.  I do this every so often to populate my EPG--it only takes
a minute or two.  Shortly after doing this, I got a seg fault and vdr
exited with no warnings or info in the log, in fact it didn't even
acknowledge it had died, it just died silently.

I think there also may be some sort of memory corruption thing going on
because I get messages in my log that have unprintable characters in them
for channel names and some printable characters match fragments of my
channel names in my channels.conf.

I have tried to use efence but to no avail--it doesn't whine of any
illegal memory usage or array out of bounds etc.

This is only a suspicion at the moment, I haven't thoroughly studied all
the code in libsi or libvdr, nor have I ensured eit.c is thread-safe.  It
looks like libsi or libvdr might not be thread-safe, note I emphasize the
word might.  Maybe wrapping the calls from vdr in mutexes would do,
locking list operations in libvdr or libsi would be better. (wouldn't want
to forget a mutex one time)

One thing is clear, and that is that I need to study the code more
thoroughly before pronouncing a bug.

I think there is some wierdness in NEWSTRUCT surrounding section filters,
although I wouldn't want to BEGIN trying to dig through the sources for
it.

_J

In the new year, Klaus Schmidinger wrote:
> Jeremy Hall wrote:
> > 
> > I think there's some sort of memory munging going on in the eit code.
> 
> What makes you believe that there is a problem in the EIT code?
> 
> Klaus
> 
> > I'm
> > not sure, but I have seen similar results a few hours ago using
> > 1.0.2.  Maybe it accounts for our wierd activities in 1.1.x.  Have anybody
> > recently reported memory leaks, memory issues, or pointer walk issues
> > lately?
> > 
> > I get a lot of vdr-related mail, so I don't have time to read it all.
> > 
> > _J
> > 
> > In the new year, Klaus Schmidinger wrote:
> > > Rene Caspari wrote:
> > > >
> > > > I start vdr with screen (screen ./vdr) in background to use vdradmin.
> > > > After a while vdr crashes and /var/log/messages says:
> > > >
> > > > Sep 23 15:37:56 beckmann vdr[11630]: panic level: 8
> > > > Sep 23 15:45:04 beckmann vdr[11630]: panic level: 7
> > > > Sep 23 15:45:07 beckmann vdr[11630]: panic level: 8
> > > > Sep 23 15:45:10 beckmann vdr[11630]: panic level: 9
> > > > Sep 23 15:48:16 beckmann vdr[11630]: panic level: 8
> > > > Sep 23 15:48:19 beckmann vdr[11630]: panic level: 9
> > > > Sep 23 15:48:22 beckmann vdr[11630]: panic level: 10
> > > > Sep 23 15:50:31 beckmann vdr[11630]: panic level: 9
> > > > Sep 23 15:50:34 beckmann vdr[11630]: panic level: 10
> > >
> > > Are you sure there's nothing else between these log entries?
> > >
> > > > Sep 23 15:50:37 beckmann vdr[11630]: saved setup to /video/setup.conf
> > > > Sep 23 15:50:37 beckmann vdr[11630]: =====================
> > > > Sep 23 15:50:37 beckmann vdr[11630]: EPG bugfix statistics
> > > > Sep 23 15:50:37 beckmann vdr[11630]: =====================
> > > > Sep 23 15:50:37 beckmann vdr[11630]: IF SOMEBODY WHO IS IN CHARGE OF THE
> > > > EPG DATA FOR ONE OF THE LISTED
> > > > Sep 23 15:50:37 beckmann vdr[11630]: CHANNELS READS THIS: PLEASE TAKE A
> > > > LOOK AT THE FUNCTION cEventInfo::FixEpgBugs()
> > > > Sep 23 15:50:37 beckmann vdr[11630]: IN VDR/eit.c TO LEARN WHAT'S WRONG
> > > > WITH YOUR DATA, AND FIX IT!
> > > > Sep 23 15:50:37 beckmann vdr[11630]: =====================
> > > > Sep 23 15:50:37 beckmann vdr[11630]: Fix^IHits^IChannels
> > > > Sep 23 15:50:37 beckmann vdr[11630]: 0^I671^IVOX, RAI Uno, n-tv
> > > > Sep 23 15:50:37 beckmann vdr[11630]: 2^I3538^ISat.1, N24, Pro-7, Kabel 1
> > > > Sep 23 15:50:37 beckmann vdr[11630]: 3^I38^IPro-7, ARD, EinsMuXx
> > > > Sep 23 15:50:37 beckmann vdr[11630]: 4^I128^IARD, MDR, EinsFestival,
> > > > EinsMuXx, 3sat, N3
> > > > Sep 23 15:50:37 beckmann vdr[11630]: =====================
> > > > Sep 23 15:50:37 beckmann vdr[11632]: EIT processing thread ended
> > > > (pid=11632) - master
> > > > Sep 23 15:50:37 beckmann vdr[11630]: exiting
> > >
> > > >From the "panic level" entries I would say it's an "emergency exit".
> > > But then again the respective log entires for that are missing.
> > > Please check whether you have omitted any lines here.
> > >
> > > Klaus
> -- 
> _______________________________________________________________
> 
> Klaus Schmidinger                       Phone: +49-8635-6989-10
> CadSoft Computer GmbH                   Fax:   +49-8635-6989-40
> Hofmark 2                               Email:   kls@cadsoft.de
> D-84568 Pleiskirchen, Germany           URL:     www.cadsoft.de
> _______________________________________________________________
> 
> 


-- 
I do not get viruses because I do not use MS software.
If you use Outlook then please do not put my email address in your
address-book so that WHEN you get a virus it won't use my address in the
From field.




Home | Main Index | Thread Index