[Chilli] Chilli 1.2.9 sudently crash

[David Bird]

Thanks for the gdb output. The "this" pointer being NULL in
dhcp_data_req() is certainly a problem. The "dhcp connection" must have
been cleared out without the "app connection" being reset -- certainly
an edge case which "should" never happen. I've committed a check in that
function to prevent it going further and segfaulting in this case. If
could give it a test to see if the crash continues. 

Cheers,
David


On Fri, 2012-09-07 at 10:25 +0430, Mohsen Saeedi wrote:
> Hi David
> I had this problem too. chilli crash in layer 2 with some mac address
> that sends a dhcp request. i installed gdb and abrtd service for core
> dump.
> i have many dump for several crashes. if it is helpful i can send it
> to you. we had this problem with 1.2.9 and 1.3-dev too. 
> it happened when chilli receive a dhcp request for some mac address
> not for all. 
> I attached our dump. i think it is a important chilli bug and it is
> better to fix. 
> and i checked the abrtd dump and each backtrace dump was same as other
> dump. for this reason, i attached just one the dumps. 
> Thanks
> 
> 
> David Bird <david at coova.com> wrote on Thu, 06 Sep 2012 21:25:27 -0700:
> > You can google for how to setup your system to save core dumps on
> > crashing... if you compile chilli with -g option, it would be helpful
> > too in combination. You can then run gdb on your coredump and binary for
> > a glimpse at where it crashed. What version is this? if not a recent
> > one, you should try upgrading. Also, and features you are using (compile
> > and runtime) might be relevant. 
> > 
> > On Wed, 2012-09-05 at 21:39 +0200, John Paul wrote:
> > > Hi,
> > > I deployed a coova chilli solution based on freeradius coupled with postgres.
> > > The customer related several crash, followed each time by a reboot.
> > > In that case, it's not really obvious to obtain the crash's reasons...
> > > 
> > > Recently, once I was with the customer, a crash arose, chilli was dead without 
> > > any particular reason. Nothing in dmesh, either in /var/log/messages.
> > > 
> > > The only way I had on that time, was to open a wired ssh connection and 
> > > restart chilli. Curiously, I lost the ssh connection, never met until that 
> > > case. Ping or any other ssh try was responding.
> > > 
> > > I didn't have enough time to diagnose what happened, I rebooted the server.
> > > 
> > > As I read, I'm not alone in that case.
> > > 
> > > Is it possible to have some trace when chili crash without to start it in a 
> > > full debug mode ?
> > > 
> > > Thanks
> > > J. Paul
> > > _______________________________________________
> > > Chilli mailing list
> > > Chilli at coova.org
> > > http://lists.coova.org/cgi-bin/mailman/listinfo/chilli
> > 
> > _______________________________________________
> > Chilli mailing list
> > Chilli at coova.org
> > http://lists.coova.org/cgi-bin/mailman/listinfo/chilli