[asterisk-dev] [Code Review]: Fix local channel chains optimizing themselves out of a call.

Thu May 3 19:06:05 CDT 2012


> On May 2, 2012, 3:18 a.m., Alec Davis wrote:
> > Referring to https://issues.asterisk.org/jira/browse/18028
> > 
> > Tested by dialling 10127 (thus 127 looped calls)
> >    the exceptionally long queue, wasn't reported before or after applying this patch.
> >    After patch successfully (very fast) optimizes out all required calls when using the dialplan.
> > 
> > Any more calls (ie 128) I did have hung channels, but was able to hang them up with 'channel request hangup'
> 
> rmudgett wrote:
>     What stuck channels?  If you are referring to the last local channel in the chain running the Echo application, that channel cannot be optimized out.
>     If you called another phone, all local channels can be optimized out.
> 
> Alec Davis wrote:
>     I'm running the same version as last night (when I managed to get stuck local channel twice), and after 20 attempts this morning I cannot repeat the stuck channels.
>     
>     I understand about the last local chan not being able to be optimized out.
>     
>     I had hungup, but luckily the history buffer allows me to see the command I used, 'channel request hangup Local/10000 at phones-5e0c;1'
>     
>     I'll look through debug log to see if I can find more.
> 
> Alec Davis wrote:
>     The hung channels wern't realy hung at all, I'd hookflashed, not hungup, which ends up with 2 echo apps talking together :(
>     
>        -- Executing [h at phones:1] NoOp("Local/10127 at phones-c2f0;2", "") in new stack
>       == Spawn extension (phones, 10127, 2) exited non-zero on 'Local/10127 at phones-c2f0;2'
>         -- Executing [s at echo-test:4] Echo("Local/10000 at phones-4dc3;2", "") in new stack
>         -- Started three way call on channel 33                                        <<<<< Here's the issue, I finished up hook flashing, not hanging up.
>         -- Started music on hold, class 'default', on Local/10000 at phones-4dc3;1
>         -- Starting simple switch on 'DAHDI/33-2'
>         -- Executing [10128 at phones:1] Set("DAHDI/33-2", "i=10127") in new stack
>         -- Executing [10128 at phones:2] Dial("DAHDI/33-2", "Local/10127 at phones") in new stack
>         -- Called Local/10127 at phones
>     
>     
>       == Spawn extension (phones, 10058, 2) exited non-zero on 'Local/10058 at phones-4606;2'
>         -- Executing [h at phones:1] NoOp("Local/10059 at phones-f677;2", "") in new stack
>       == Spawn extension (phones, 10059, 2) exited non-zero on 'Local/10059 at phones-f677;2'
>         -- Executing [h at phones:1] NoOp("Local/10060 at phones-eeeb;2", "") in new stack
>         -- Building conference call with DAHDI/33-1 and DAHDI/33-2
>         -- Stopped music on hold on Local/10000 at phones-4dc3;1
>       == Spawn extension (phones, 10060, 2) exited non-zero on 'Local/10060 at phones-eeeb;2'
>         -- Executing [h at phones:1] NoOp("Local/10061 at phones-dbe3;2", "") in new stack
>       == Spawn extension (phones, 10061, 2) exited non-zero on 'Local/10061 at phones-dbe3;2'
>     ...
>       == Spawn extension (phones, 10127, 2) exited non-zero on 'Local/10127 at phones-7d55;2'
>         -- Executing [s at echo-test:4] Echo("Local/10000 at phones-9e2e;2", "") in new stack
>         -- TRANSFERRING DAHDI/33-2 to DAHDI/33-1
>     [2012-05-03 19:17:49.158442] WARNING[8429]: chan_dahdi.c:3722 _dahdi_get_index: Unable to get index for 'DAHDI/33-1' on channel 33 (dahdi_indicate(), line 9395)
>         -- Hanging up on 'Local/10000 at phones-9e2e;1<MASQ>'
>         -- Hungup 'Local/10000 at phones-9e2e;1<MASQ>'
>         -- Executing [h at phones:1] NoOp("DAHDI/33-2", "") in new stack
>       == Spawn extension (phones, 10128, 2) exited non-zero on 'DAHDI/33-2'
>         -- Hanging up on 'DAHDI/33-2'
>         -- Hungup 'DAHDI/33-2'
>         -- Starting simple switch on 'DAHDI/33-1'
>     
>     asterix*CLI> core show channels
>     Channel              Location             State   Application(Data)
>     Local/10000 at phones-9 s at echo-test:4        Up      Echo()
>     Local/10000 at phones-9 10128 at phones:2       Up      Dial(Local/10127 at phones)
>     Local/10000 at phones-4 s at echo-test:4        Up      Echo()
>     Local/10000 at phones-4 (None)               Up      AppDial((Outgoing Line))
>     DAHDI/33-1           s at phones:1           Rsrvd   (None)
>     
>     asterix*CLI> channel request hangup
>     Local/10000 at phones-4dc3;2  Local/10000 at phones-9e2e;1  Local/10000 at phones-9e2e;2  Local/10000 at phones-4dc3;1
>
> 
> rmudgett wrote:
>     I'm surprised you didn't hit the deadlock in sig_analog that I did.
> 
> rmudgett wrote:
>     I think I see what you did.  You waited for the channels to be optimized out before you attempted to transfer them together.  I did not wait.  While the local channels were trying to optimize themselves out, I tried to transfer the call with a flash hook.  Deadlock ensues.

>You waited for the channels to be optimized out 
As it turned out, yes I had waited for all channels to be optimised out.

>before you attempted to transfer them together
I wasn't trying to transfer, I was just too quick to try again, which resulted in a transfer.


- Alec


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/1894/#review6126
-----------------------------------------------------------


On May 3, 2012, 5:01 p.m., rmudgett wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/1894/
> -----------------------------------------------------------
> 
> (Updated May 3, 2012, 5:01 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> The masquerade torture test (ASTERISK-19807) is a long chain of local channels to be optimized out of a call when answered.
> This patch fixes local channel so it can successfully optimize itself out when there are chains of channels.
> 
> The principle problem was check_bridge() failed to check the ast_channel_masquerade() return value to see if the masquerade was actually setup.  In long chains of local channels, the masquerade occasionally fails to get setup because there is another masquerade already setup on an adjacent local channel in the chain.
> 
> * Checks the return value of ast_channel_masquerade().
> 
> * Made sure that the outgoing local channel (the ;2 channel) also does not have any frames in its queue before the masquerade.
> 
> * The outgoing channel will flush one voice or video frame per optimization attempt.
> 
> * Do the masquerade immediately to ensure that the outgoing channel queue does not get any new frames added and thus unconditionally flushed.
> 
> * Blocks indication -1 (Stop tones event) when the local channel is going to optimize itself out.  When the call is answered, a chain of local channels pass down a -1 indication for each bridge.  This blizzard of -1 events really slows down the optimization process.
> 
> 
> This addresses bug ASTERISK-16711.
>     https://issues.asterisk.org/jira/browse/ASTERISK-16711
> 
> 
> Diffs
> -----
> 
>   /branches/1.8/channels/chan_local.c 365297 
> 
> Diff: https://reviewboard.asterisk.org/r/1894/diff
> 
> 
> Testing
> -------
> 
> Chains of 100 local channels now are able to optimize themselves out.
> 
> 
> Thanks,
> 
> rmudgett
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20120504/0bf921ec/attachment.htm>