Summary:	ASTERISK-21374: [patch] One-way Audio With auto_* NAT Settings When SIP Calls Initiated By PBX
Reporter:	Michael L. Young (elguero)	Labels:
Date Opened:	2013-04-02 14:53:00	Date Closed:	2013-05-13 16:15:38
Priority:	Major	Regression?	Yes
Status:	Closed/Complete	Components:	Channels/chan_sip/General
Versions:	11.3.0	Frequency of Occurrence
Related Issues:	cannot be resolved before completion of ASTERISK-21225 [patch] Setting nat=force_rport in [general] sip.conf will never work causes ASTERISK-22236 REGISTER reply send to bad port with nat=yes(or force_rport,comedia) in 11.5.0 causes ASTERISK-26457 [patch] force_rport,auto_comedia: No NAT detection triggered. is related to ASTERISK-20904 RFC1918 NAT Issue On Prune
Environment:		Attachments:	( 0) asterisk-21374-auto-nat-outgoing-fix.diff ( 1) asterisk-21374-fix-crash-and-rt-peers.diff ( 2) gdb.txt
Description:	I found another case where the force_rport and comedia flags are not being set automatically when using the new auto_* settings.  This time it involves calls initiated by the PBX. When we reload asterisk the default flags turned on and off by auto_force_rport (force_rport) and auto_comedia (comedia) go back to the default setting of off.  These flags are turned on, as needed, when a peer re-registers or initiates a call.  This would apply to even just having the default global setting "nat=auto_force_rport". Everything is good except in the following scenario: We reload Asterisk and the peer's registration has not expired.  We load in the default settings for the peer which turns force_rport and comedia back to off.  Since the peer has not re-registered or placed a call yet, they remain off.  We then initiate a call to the peer from the PBX.  The force_rport and comedia flags stay off.  If NAT is involved, we end up with one-way audio since we never checked to see if the peer is behind NAT or not.
Comments:	By: Michael L. Young (elguero) 2013-04-02 14:54:32.264-0500 The patch from ASTERISK-21225 needs to be applied as well as this patch. By: David Brillert (aragon) 2013-04-10 13:13:23.579-0500 Ping I'm waiting for this patch to get committed but I don't see any progress on https://reviewboard.asterisk.org/r/2421/ By: Michael L. Young (elguero) 2013-04-10 13:15:14.210-0500 Yep, me too.  Have you been able to test the patch?  If so, I take it you have had good results? By: David Brillert (aragon) 2013-04-10 13:48:08.226-0500 I haven't tested the patch.  Only my hosted production servers seem effected and I have no way to reproduce on any development systems.  I'd prefer to wait until the patch passes review and then try it on the production system... By: Michael L. Young (elguero) 2013-04-12 12:16:27.227-0500 @David, The patch has been committed and should show up in Asterisk 11.5. By: jmls (jmls) 2013-04-25 15:46:00.251-0500 well, this kinda broke asterisk for me ;) using an ip address instead of a peer name causes a crash. backtrace attached By: Michael L. Young (elguero) 2013-04-25 16:14:12.754-0500 Doh!  Sorry about that... it even made it through code review... I'll work on fixing that up. By: David Brillert (aragon) 2013-04-27 22:43:25.822-0500 Another ping ;) By: Michael L. Young (elguero) 2013-04-30 21:04:29.546-0500 I have been testing this patch, [^asterisk-21374-fix-crash-and-rt-peers.diff], with success.  It is based off of Asterisk 11 branch.  Tomorrow will be a good test on a production server. If anyone else could test this, that would be great to get more confirmation that this patch works as intended. Thanks By: David Brillert (aragon) 2013-05-01 13:59:47.917-0500 I held off upgrading to 11.4-rc1 due to the crash report. So I never experienced the crash. Hopefully jmls can test since he reported the issue. I am glad to see you wrote a patch though since that means that soon I can finally upgrade to 11.4.0 or 11.5, or SVN (whenever this gets committed).  I have been holding off while waiting for this ticket to get closed. By: David Brillert (aragon) 2013-05-08 08:32:25.895-0500 Michael, I really wish I could test this patch but I can't. I'm still waiting for a commit and jmls hasn't replied. How about throwing your patch asterisk-21374-fix-crash-and-rt-peers.diff onto the review board so we can get it shipped? By: Michael L. Young (elguero) 2013-05-08 23:19:20.454-0500 David, This crash would happen right now in current 11 branch, not 11.4.0-rc1. I was kind of hoping for some additional confirmation besides my own testing plus I have been busy as well. Will work on getting this on review board. By: David Brillert (aragon) 2013-05-09 11:18:55.160-0500 Hooray, I see a ship it from Mark Michelson! I'm really looking forward to 11.5 By: David Brillert (aragon) 2013-05-13 12:18:52.515-0500 Ping The patch on reviewboard got a ship it. Will you be committing soon so it makes it way into 11.5 or are you waiting for more ship its? I can't test an uncommitted patch on a production server... When you commit I will probably a do an SVN checkout for production.