[Asterisk-Users] Segfaults with NAT clients

[Mark Spencer]

If you run asterisk with -vvvgc you can force Asterisk to dump core when
it crashes.  then you can run:

gdb ./asterisk core.<foo>
.
.
.
(gdb) bt

And that will give you a backtrace of where the crash occured.  Even just
the backtrace will give some useful information, but if you have even a
vague idea of how to use gdb, you can use it to get more detailed
information by changing the frame and looking at specific data elements.

Remember, the core file is useless except on the system where you compiled
it, so DON'T SEND ME A CORE FILE, but *do* send the backtrace results.

Mark

On 19 Mar 2003, Matthew Farley wrote:

> Another puzzler from my asterisk system:
>
> For reference, here is my basic setup:
> -Asterisk (latest CVS as of this afternoon (3/19/2003)
> -All clients are Cisco ATA 186 units
> -Public IP for Asterisk server
> -All clients behind NAT firewalls
>
> Everything was working fine with my Asterisk system until I added a
> client behind a firewall that has three numbers in the final octet of
> its public IP number.
>
> For instance:
>
> A client behind a firewall with public IP of 63.77.102.4 would be fine
> A client behind a firewall with public IP of 63.77.102.221 would fail
>
> The failure does not present itself until an INVITE is issued
> (registration works fine). When the INVITE happens to or from such an
> IP, though, Asterisk segfaults. After studying the output of sip debug,
> I noticed that somehow the Contact: field is getting messed up.
>
> The field should look like:
> Contact: <sip:3117000 at 63.91.19.135:5060;user=phone;transport=udp>
>
> Instead, it looks like:
> Contact: <sip:3117000 at 63.91.19.0:5060;user=phone;transport=udp>
>
> I checked the settings in the ATA, and they are all correct, with no
> reference whatsoever to '63.91.19.0'. The ATA is on the same firmware as
> all of my working units.
>
> I can only assume that Asterisk is somehow losing that last octet when
> it deals with a NAT client behind a firewall which has a three-digit
> number at the end. Does this seem accurate, or am I missing something
> here?
>
> I have attached an excerpt of sip debug output showing this problem
> (line 22). Any suggestions that anyone could offer for
> troubleshooting/fixing this problem would be greatly appreciated.
>
> Thanks!
> -Matt
>
>
> --
> Matthew Farley <asterisk at wheatstate.net>
>