SIP TCP Bug

Steven Howes · ‎12-01-2010

Hi,

Think I may have found a bug in the SIP TCP...

We've got a 504 and a 508 using SIP TCP. If they lose their internet connection (but not network connection, i.e. still has LAN access) for enough time, they fail to re-register (but think they are trying...)

I've managed to syslog from one of them, it thinks it is sending an ICMP packet and it is failing, however a tcpdump on the gateway shows the handset isn't even sending the packet.

Output on syslog is as follows:

Dec 1 11:01:13 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:13 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:01:13 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:13 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:01:13 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:13 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:01:14 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:14 172.16.0.47 [0]SIP:RegFailed;Retry in 120s

Dec 1 11:01:29 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:29 172.16.0.47 [0]SIP:RegFailed;Retry in 120s

Dec 1 11:01:43 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:43 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:01:43 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:43 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:01:43 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:43 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:01:44 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:44 172.16.0.47 [0]SIP:RegFailed;Retry in 120s

Dec 1 11:01:59 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:01:59 172.16.0.47 [0]SIP:RegFailed;Retry in 120s

Dec 1 11:02:13 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:02:13 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:02:13 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:02:13 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:02:13 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:02:13 172.16.0.47 SUBS Failed 19 30 -1

Dec 1 11:02:14 172.16.0.47 [0]SIP:ICMP Error -1 (c36ea214:5060, 16777226)

Dec 1 11:02:14 172.16.0.47 [0]SIP:RegFailed;Retry in 120s

Which to me looks like it is sending a packet and thinks it gets an ICMP failure. But they gateway sees nothing. Network config on the phone is intact, and a reboot fixes it so its not a network/gateway problem. We've seen this in multiple locations (one with ISA, one with an 877, one with an SRP527W).

Have attached (censored) spa config. All removed information replaced with xxxxxx.

Steve

Steven Howes · ‎12-01-2010

Update: It also cannot retrieve XML from the same host. It's like all outbound TCP connections are failing (inbound works because I can't get web admin)

Dec 1 11:52:21 172.16.0.47 Starting XML service @ http://p1.voice.xxxxxx.net.uk/xml/svcs/?mac=$MA

Dec 1 11:52:21 172.16.0.47 ********setSoftKeys for 94a02180 to type 0 with 0 items

Dec 1 11:52:21 172.16.0.47 CMX_new(), URL as http://p1.voice.xxxxxx.net.uk/xml/svcs/?mac=002584d8ceaa after macro expand

Dec 1 11:52:21 172.16.0.47 cme services url=http://p1.voice.xxxxxx.net.uk/xml/svcs/?mac=002584d8ceaa

Dec 1 11:52:21 172.16.0.47 create CMX_new @ 94a02180, init cbData 0

Steven Howes · ‎12-06-2010

Any thoughts anyone?..

Andrew Hickman · ‎12-09-2010

Hi Steve,

This is a known issue and should be fixed in the 7.4.7 release that is due to be posted before the end of the month.

Cheers

Andy

Steven Howes · ‎12-10-2010

Hi Andy,

Thanks. Do you know if this bug kills all TCP or just that to a specific host? We've had one or two phones that reach our failover server (using SRV records) and failing to jump back to the primary afterwards.

Steve

Steven Howes · ‎05-25-2011

The initial problem was fixed as expected, but the failure to jump back to primary is still present (even in 7.4.8). Are Cisco aware of anything that may prevent it retrying the primary server once it returns? The proxy fallback interval is set correctly. I'll attach syslogs shortly of two phones on the same LAN, one fails back to the primary, one doesn't.