[asterisk-bugs] [Asterisk 0018036]: Incoming SIP UDP packets are ignored.

[Asterisk Bug Tracker]

A NOTE has been added to this issue. 
====================================================================== 
https://issues.asterisk.org/view.php?id=18036 
====================================================================== 
Reported By:                dtryba
Assigned To:                
====================================================================== 
Project:                    Asterisk
Issue ID:                   18036
Category:                   Channels/chan_sip/General
Reproducibility:            random
Severity:                   minor
Priority:                   normal
Status:                     feedback
Asterisk Version:           1.6.2.13 
JIRA:                        
Regression:                 No 
Reviewboard Link:            
SVN Branch (only for SVN checkouts, not tarball releases): N/A 
SVN Revision (number only!):  
Request Review:              
====================================================================== 
Date Submitted:             2010-09-23 07:53 CDT
Last Modified:              2010-11-11 06:17 CST
====================================================================== 
Summary:                    Incoming SIP UDP packets are ignored.
Description: 
UDP SIP datagrams are not processed by Asterisk. tcpdump shows them
arriving on the network interface but they don't show up in Asterisk. This
lasts for about 10m (RTP and SIP/TCP or IAX functions just fine in this
period), Asterisk then starts to "Really destroying SIP dialog" (100+ in
1s) and suddenly all is well again. Outbound UDP packets are being send and
the other side replies (but obviously gets ignored).

Last UDP datagram received:
[2010-09-22 18:49:20] VERBOSE[17369] chan_sip.c:
<--- SIP read from UDP:213.247.xxx.xxx:1024 --->

next datagram arrives 9m and 11s later:
[2010-09-22 18:58:31] VERBOSE[17369] chan_sip.c:
<--- SIP read from UDP:84.246.xxx.xxx:1042 --->

Versions affected are 1.6.2.9 (backported from Debian/unstable to
Debian/stable) and 1.6.2.13 (Debian 1.6.2.9-2 with patches applied to get
to version 1.6.2.13). This suddenly started happening 1 weeks ago, before
this machine had been running in the same configuration since 1.6.2.9 was
released by the Debian team (uptime was about 115 days before rebooting
this machine to fix a kernel issue) without any problems.

The only thing changed that week was the installation of fail2ban to block
bruteforce attempts. But removing this (and rebooting the machine) didn't
solve the problem. 

The problem happends at random times, during the day with active calls or
during the night whitout any usage.

====================================================================== 

---------------------------------------------------------------------- 
 (0128777) schmidts (manager) - 2010-11-11 06:17
 https://issues.asterisk.org/view.php?id=18036#c128777 
---------------------------------------------------------------------- 
this deadlock occurs in the scheduler part of the do_monitor section and
this cause asterisk to not receive any further packets.

what i can futher see is that there is a problem with the refcounter for
sip objects like this:

name: simonthuis2
refcount: 1345

name: simonthuis1
refcount: 1286

this should never be on such a high value cause this objects wont be
destroyed clearly.
IMO there is a refcounter bug with TCP/TLS calls in 1.6.2 which cause this
deadlock. 

Issue History 
Date Modified    Username       Field                    Change               
====================================================================== 
2010-11-11 06:17 schmidts       Note Added: 0128777                          
======================================================================