On Fri, 5 Jun 2009, Iñaki Baz Castillo wrote:
2009/6/5 Aymeric Moizard jack@atosc.org:
Hi Again, (sorry for wrong subject previous post),
I'm sure this is not enough to help, but I just got a crash using kamailio 1.5.
Jun 5 10:24:43 ns26829 /usr/sbin/kamailio[29050]: CRITICAL:core:anchor_lump: offset exceeds message size (682 > 675) aborting... Jun 5 10:24:43 ns26829 /usr/sbin/kamailio[29127]: CRITICAL:core:receive_fd: EOF on 21 Jun 5 10:24:43 ns26829 /usr/sbin/kamailio[29036]: INFO:core:handle_sigs: child process 29050 exited by a signal 6 Jun 5 10:24:43 ns26829 /usr/sbin/kamailio[29036]: INFO:core:handle_sigs: core was not generated Jun 5 10:24:43 ns26829 /usr/sbin/kamailio[29036]: INFO:core:handle_sigs: terminating due to SIGCHLD
What is that "offset exceeds message size"? What kind of trace should I take if this happen again?
Please, configure your system and Kamailio to generate verbose coredump.
Such a message occurred some time ago in ALL OpenSer/Kamailio versions due when the "Content-Length" header value was greater than the real body size, and a body modification was performed (as force_rtpproxy). The bug was that the code used the "Content-Length" header value instead of the real body size, so a segmentfault occurred due to wrong memory access.
This could be a similar issue, please generate a coredump. Regards.
My kamailio -V shows "kamailio 1.5.0-tls"
I'm pretty sure this is right: the last INVITE that was made comes from a user that was using a ALG box. I received a call from this same guy earlier in the morning with traces and have seen this ALG was modifying some INVITE contents... I guess this ALG hasn't changed properly the Content-Length and I'm now pretty sure this is it!
I'll check where to grab the fix. Tks for the valuable help!
Aymeric
-- Iñaki Baz Castillo ibc@aliax.net