[asterisk-bugs] [JIRA] (ASTERISK-30244) Crash in res_pjsip.so with TCP or TLS transports and Subscriptions

[nappsoft (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-30244?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=260301#comment-260301 ] 

nappsoft commented on ASTERISK-30244:
-------------------------------------

Some additional notes:

- the issue only affects reliable transports (what is clear as this codepath does only exist in conjunction with reliable transports)
- the way I can reproduce this is by not doing a normal reboot of the phone (in which case the phone would shutdown the transport first) but by unplugging and replugging power
- when removing "ast_sip_transport_monitor_unregister" from subscription_persistence_remove in res_pjsip_pubsub.c asterisk doesn't crash (and somehow I didn't find any negative side-effects) => so maybe you can tell me what happens when not unregistering the monitor? will this cause any leaks and shouldn't be done as a workaround? For the moment we are better off with a system with undetectable side-effects than with one with frequent crashes...

As you might ask: we do not use pjsip bundled, but we use the same pjsip version (2.12.1) and use the same config_site.h and patches.

> Crash in res_pjsip.so with TCP or TLS transports and Subscriptions
> ------------------------------------------------------------------
>
>                 Key: ASTERISK-30244
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-30244
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Resources/res_pjsip_pubsub
>    Affects Versions: 18.14.0
>            Reporter: nappsoft
>            Severity: Major
>         Attachments: crash.txt
>
>
> We are observing crashes from time to time on less than one percent of systems. The segfault always happens at the same address in res_pjsip.so
> What they have in common: while we use lots of clients with TCP and TLS connections, only on these systems we have phones (Yealink) connected which subscribe over TLS and/or TCP.
> I can reproduce the crash from time to time when rebooting the phone. The crash happens as soon as it is up again and renewing the subscriptions.
> I was now able to create a backtrace, which you will find attached. 



--
This message was sent by Atlassian JIRA
(v6.2#6252)