[asterisk-bugs] [JIRA] (ASTERISK-28209) res_pjsip_t38: Passthrough causes crash when re-invite collision

[Joshua C. Colp (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-28209?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=245773#comment-245773 ] 

Joshua C. Colp commented on ASTERISK-28209:
-------------------------------------------

I think what is happening is that the re-invite collision is causing the pending media state to get cleared out, so when the re-invite goes out again the negotiation does not have the proper state to act on that it assumes to exist and causes a crash. The pending state needs to remain on a 491, as the re-invite will indeed go out - albeit just delayed.

> res_pjsip_t38: Passthrough causes crash when re-invite collision
> ----------------------------------------------------------------
>
>                 Key: ASTERISK-28209
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-28209
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Resources/res_pjsip_t38
>    Affects Versions: 16.1.0
>         Environment: Ubuntu 16.04 LTS, E3-1285, 32GB RAM
>            Reporter: Gregory Massel
>            Assignee: Unassigned
>              Labels: fax, pjsip
>         Attachments: core-dump-2018-12-13-08h04-brief.txt, core-dump-2018-12-13-08h04-full.txt, core-dump-2018-12-13-08h04-locks.txt, core-dump-2018-12-13-08h04-thread1.txt, core-dump-2018-12-13-08h42-brief.txt, core-dump-2018-12-13-08h42-full.txt, core-dump-2018-12-13-08h42-locks.txt, core-dump-2018-12-13-08h42-thread1.txt, t38call2.pcap, t38call.pcap
>
>
> After upgrading to 16.1.0, Asterisk crashed six times in the space of an hour. After downgrading to 13.24.0, it has been rock solid for a number of days. Prior to that it was on 13.23.1 and rock solid for weeks.
> I cannot isolate which component is causing the issue, however, the one possible area in which this server is unique (versus another similar server that is running stable on 16.1.0) is that it uses res_odbc and func_odbc extensively. But blaming those components would be speculative.
> I have attached two backtraces from two different crash events.



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