[asterisk-bugs] [JIRA] (ASTERISK-28643) Deadlock, possibly in Parking, maybe in combination with AMI status messages.

[George Joseph (JIRA)]

     [ https://issues.asterisk.org/jira/browse/ASTERISK-28643?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

George Joseph updated ASTERISK-28643:
-------------------------------------

    Assignee: Steve Sether  (was: Unassigned)
      Status: Waiting for Feedback  (was: Triage)

I got it but without the binaries I can't do much with it.
Can you run the following on the machine that produced the coredump?
{noformat}
$ sudo /var/lib/asterisk/scripts/ast_coredumper --tarball-coredumps --no-default-search /tmp/core.usi-vf13-mtka.ravon.net-2019-11-14T15_20_29-0600
{noformat}

This will tarball up the coredump, the text files and the asterisk binaries in use.

Now I'll have to see if we've still got a CentOS 6 system hanging around. :)




> Deadlock, possibly in Parking, maybe in combination with AMI status messages.
> -----------------------------------------------------------------------------
>
>                 Key: ASTERISK-28643
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-28643
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: PBX/General
>    Affects Versions: 16.6.1
>         Environment: Centos 6.10
>            Reporter: Steve Sether
>            Assignee: Steve Sether
>            Severity: Critical
>         Attachments: core_show_locks-2019-11-14-crash1.txt, full-asterisk-crash.gz
>
>
> Asterisk creates a deadlock, and stops responding to any SIP messages when I park a call. No new incoming calls are accepted, no outgoing calls are accepted, etc.  Some initial investigation using gdb and the core file show this might indicate it's related to generating STATUS messages from the AMI.  This happened when only one or maybe two calls were on the system.  Looking through the lock file it looks like it's a simple deadlock between two threads, each of which holing locks the other needs.   One of which is in parking code.
> I was able to reproduce it initially a couple times, (and got some debug information out of it, including from a version where core show locks was compiled in).  For whatever reason, I can't seem to reproduce it again, even under the same conditions.



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