[sr-dev] [kamailio/kamailio] core kamailio 5.4 (#2740)

[descartin]

Dear all

we have seen a core in a running production kamailio instance today, i attach you the bt in dropbox

version: kamailio 5.4.4 (x86_64/linux) 0ddb01-dirty
centos7 3.10.0-693.17.1.el7.x86_64 #1 SMP Thu Jan 25 20:13:58 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

https://www.dropbox.com/s/3dg72cbuz7nvx5b/core_kamailio_21mayo?dl=0
https://www.dropbox.com/s/iyko32fwfxj8aam/core_kamailio_21mayo_2?dl=0

seems there might be a relation async module when delaying a request after communicating with the rtpengine ?¿
could you please take a look?

thanks a lot and regards
david



--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/kamailio/kamailio/issues/2740

[Henning Westerholt]

Program terminated with signal 11, Segmentation fault.
#0 0x00007f4ee113ac1d in async_ms_sleep (msg=0x7f4ee5cf11b0, milliseconds=250, act=0x7f4ee5e37ca8, cbname=0x0) at async_sleep.c:381
381 if(_async_ms_list->len >= MAX_MS_SLEEP_QUEUE) {

[Henning Westerholt]

This is from the second backtrace, the first one is looking more like a crash during the kamailio shutdown. Could it be possible that you have a loop in your cfg? Just wondered from the large number of run_action, do_action calls in the trace. Kamailio should not crash from it, sure.

[descartin]

hello Henning

mmm maybe could be but i wouldn't have expected that. Seems the call is a call in where we receive a BYE from callee before que first ACK in sent from the caller (dialog in state 3). So i was using async_ms_sleep to delay it 250ms and try to avoid 200ok (of the first invite) retransmissions. In some other configurations we also use that async delay.
The difference here is we used async_route (in secs) before, and now used async_ms_sleep. Anycase the actions done after are the same.

Could it be any issue with the usage of the delay of ms?

thanks a lot and best regards
david

[descartin]

Hello Henning

i noticed that i forgot to set ms_timer param. MAybe being with the 0 default value, causes a loop when using async_ms_sleep?

best regards
david

[Daniel-Constantin Mierla]

@descartin: can you get from core file with gdb the output for:

```
frame 0
info locals
p _async_ms_list
p *_async_ms_list
```

[descartin]

hello Daniel
sorry for the delay answering

```
(gdb) frame 0
#0 0x00007f4ee113ac1d in async_ms_sleep (msg=0x7f4ee5cf11b0, milliseconds=250, act=0x7f4ee5e37ca8, cbname=0x0) at async_sleep.c:381
381 if(_async_ms_list->len >= MAX_MS_SLEEP_QUEUE) {
(gdb) info locals
ai = 0x7fff943fcc30
dsize = 0
t = 0x0
tindex = 0
tlabel = 15
atp = 0x0
at = 0x7e
__FUNCTION__ = "async_ms_sleep"
now = {tv_sec = 139976839739840, tv_usec = 112}
upause = {tv_sec = 140735680595712, tv_usec = 5221289}
(gdb) p _async_ms_list
$1 = (struct async_ms_list *) 0x0
(gdb) p *_async_ms_list
Cannot access memory at address 0x0
```

best regards
david escartin

[Daniel-Constantin Mierla]

Have you set:

* https://www.kamailio.org/docs/modules/stable/modules/async.html#idm56

[descartin]

hello Daniel

yes i set this parameter after we had the crash

regards
david

[Daniel-Constantin Mierla]

If it crashed after you set it, then reopen. Otherwise, there was a safety check pushed to repo to avoid crashes when it is not set.

[Daniel-Constantin Mierla]

Closed #2740.