[asterisk-bugs] [Asterisk 0012513]: RTP is not reinvited back to Asterisk when a native bridge is broken by AMI redirect on both channels

[noreply at bugs.digium.com]

The following issue has been SUBMITTED. 
====================================================================== 
http://bugs.digium.com/view.php?id=12513 
====================================================================== 
Reported By:                mneuhauser
Assigned To:                
====================================================================== 
Project:                    Asterisk
Issue ID:                   12513
Category:                   Channels/chan_sip/General
Reproducibility:            always
Severity:                   major
Priority:                   normal
Status:                     new
Asterisk Version:           SVN 
SVN Branch (only for SVN checkouts, not tarball releases):  1.4  
SVN Revision (number only!): 114602 
Disclaimer on File?:        N/A 
Request Review:              
====================================================================== 
Date Submitted:             04-24-2008 07:47 CDT
Last Modified:              04-24-2008 07:47 CDT
====================================================================== 
Summary:                    RTP is not reinvited back to Asterisk when a native
bridge is broken by AMI redirect on both channels
Description: 
Two SIP channels are natively bridged, i.e., RTP flows directly between
them and not through Asterisk. A Redirect command is received on AMI that
redirects both channels of the bridge to some extension (to add both to a
conference, but it does not matter). The problem is. that the RTP of only
one of the channels is reinvited back to Asterisk but not the other's.

This happens:
* manager.c:action_redirect() locks both channels and calls
ast_async_goto() on both of them
* pbx.c:ast_async_goto() does a
ast_explicit_goto()/ast_softhangup_nolock() on one channel (the one with
chan->pbx set) and a masquerade on the other
* the redirected channels are both handled by one thread in
rtp.c:bridge_native_loop()
* rtp.c:bridge_native_loop() terminates because a softhangup was performed
on the non-masqueraded channel and prX->set_rtp_peer(cX, NULL, NULL, 0, 0)
is called on both channels (to reinvited RTP back to Asterisk), but one of
the channels has already been masqueraded and is a zombie, i.e.,
set_rtp_peer() is NOT called for this channel and RTP is not reinvited back
to Asterisk.

I think the easiest way to fix this is to perform the
re-invite-back-to-Asterisk in the chan_sip.c:sip_fixup() function that is
called during all masquerade operations performed on SIP channels (see
attached patch). I think this is save to do there because a channel can not
be masqueraded *into* a native bridge, only *out* of one. So there is no
danger that this breaks a native bridge that should stay up.

====================================================================== 

Issue History 
Date Modified   Username       Field                    Change               
====================================================================== 
04-24-08 07:47  mneuhauser     Asterisk Version          => SVN             
04-24-08 07:47  mneuhauser     SVN Branch (only for SVN checkou =>  1.4         
  
04-24-08 07:47  mneuhauser     SVN Revision (number only!) => 114602          
======================================================================