[asterisk-bugs] [JIRA] (ASTERISK-22857) Deadlock: Locked Here: chan_iax2.c line 9756 (socket_read)

[Matt Jordan (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-22857?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=212080#comment-212080 ] 

Matt Jordan commented on ASTERISK-22857:
----------------------------------------

Hm. That doesn't make a lot of sense.

The offending code is this:

{noformat}
		iax2_lock_owner(fr->callno);
		if (iaxs[fr->callno]) {
			if (iaxs[fr->callno]->owner) {
				struct ast_format_cap *orignative = ast_format_cap_dup(ast_channel_nativeformats(iaxs[fr->callno]->owner));
				struct ast_format_cap *native = ast_channel_nativeformats(iaxs[fr->callno]->owner);
				if (orignative) {
					ast_format_cap_set(native, &f.subclass.format);
					if (ast_channel_readformat(iaxs[fr->callno]->owner)->id) {
						ast_set_read_format(iaxs[fr->callno]->owner, ast_channel_readformat(iaxs[fr->callno]->owner));
					}
					ast_format_cap_copy(native, orignative);
					orignative = ast_format_cap_destroy(orignative);
				}
				ast_channel_unlock(iaxs[fr->callno]->owner);

{noformat}

Richard's patch moved the {{ast_channel_unlock}} from inside the {{if (orignative)}} call to outside of it. The only time that would have made any difference is if the channel did not have native formats on it (which is fairly off nominal) - in which case, the original code would have left the channel locked.

I'm not sure how or why reverting this would cause any difference to the scenario you're seeing.
                
> Deadlock: Locked Here: chan_iax2.c line 9756 (socket_read)
> ----------------------------------------------------------
>
>                 Key: ASTERISK-22857
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-22857
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Channels/chan_iax2
>    Affects Versions: SVN, 11.6.0
>         Environment: 64 bit CentOS, 4GB RAM
>            Reporter: David Brillert
>            Assignee: David Brillert
>            Severity: Critical
>         Attachments: backtrace.txt, core show locks_gdb thread apply all bt full.txt, gdbcorebtbtfull.txt, iax2 debug.txt, valgrind.txt
>
>
> Full deadlock no call processing, no SIP traffic.
> GDB output DONT_OPTIMIZE DEBUG_THREADS BETTER_BACKTRACES thread apply all bt full.txt attached.  The deadlock occurred overnight while the system was idle.
> It is happening on multiple production servers.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.asterisk.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira