Summary:ASTERISK-12687: Crash in ast_cdr_start() when Local channel is involved
Reporter:mdu113 (mdu113)Labels:
Date Opened:2008-09-05 11:40:59Date Closed:2011-06-07 14:00:35
Versions:Frequency of
Environment:Attachments:( 0) bt_full.txt
( 1) cdr_bug.txt
Description:Scenario: a SIP call arrives and get split into several branches using Local channel. A callee in one of the branch answers the call and then hangs up. Asterisk crashes at point when callee hangs up and gdb suggests (AFAIU) that crash happened while in the ast_cdr_start().
This happens only if hangup initiated by callee. If hangup initiated by caller then everything's fine.



exten => 17189282005,1,Dial(Local/111@fork-xyz&Local/650@fork-xyz&Local/101@fork-xyz,10)

exten => 111,1,Dial(SIP/xyz011101,10)
exten => 111,n,Hangup
exten => 650,1,Dial(SIP/poly_650_01,10)
exten => 650,n,Hangup
exten => 101,1,Dial(SIP/xyz010101,10)
exten => 101,n,Hangup

Console output and backtrace is in attached cdr_bug.txt
Comments:By: mdu113 (mdu113) 2008-09-05 11:44:32

Forgot to say, as well as in issue 13364, whatever causing it was done after r118858. Revision 118858 doesn't have this issue

By: Jeff Peeler (jpeeler) 2008-09-05 22:52:34

This can be duplicated without dialing multiple local channels. It looks like the scenario goes like this:

exten => 17189282005,1,Dial(Local/111@fork-xyz)

SIP/poly_650_01 calls extension 17189282005.

(chan, peer  chan, peer)
SIP/poly_650_01 <-> Local/111@fork-xyz-94a6,1 <-> Local/111@fork-xyz-94a6,2 <-> SIP/xyz011101

The two local channels are hung up leaving:
SIP/poly_650_01 <-> SIP/xyz011101

I haven't looked much closer, but the orig_chan_cdr in ast_bridge_call looks very suspect. One of the peers is definitely not valid since it was one of the local channels.

By: mdu113 (mdu113) 2008-09-08 13:13:14

jpeeler, you're right. I can confirm that it happens even with single local channel.

By: Alan Graham (zerohalo) 2008-09-08 13:28:29

Can confirm the same here with r141565 - single channel causes crash - BT attached.

By: Steve Murphy (murf) 2008-09-11 08:42:34

Sorry for the wait!

I'm closing this bug 'no change required', because it was fixed when
the side issue crash in 13409 was fixed by Russell.

I labbed it up and hung up on both sides, and no crash.