[Home]

Summary:DAHLIN-00258: weird sound with a native bridged isdn-bri connection
Reporter:Daniel (danielp)Labels:
Date Opened:2011-10-07 11:03:36Date Closed:2011-11-04 09:54:25
Priority:MajorRegression?
Status:Closed/CompleteComponents:wcb4xxp
Versions:2.5.0 Frequency of
Occurrence
Related
Issues:
Environment:Asterisk 1.8.7.0 with DAHDI-Linux 2.5.0.1 and echo canceler OSLEC on Kernel 2.6.33.5Attachments:( 0) debug.txt
( 1) faxsound.wav
Description:We have connected ISDN-PSTN (BRI, PtP TE-mode) and a Tenovis-ISDN-telephone-system (PtP NT-mode) to a Junghanns.NET HFC-8S (wcb4xxp). (And some SIP-phones.)

The connection will go over native bridging between DAHDI/pstn and DAHDI/isdn-telsystem.

With DAHDI-Linux 2.4.1.2 it works fine.

Now with DAHDI-Linux 2.5.0 very often the sound is very weird. Caller and callee couldn't understand any word.

What can i do, to give you more debugging information?
Comments:By: Shaun Ruffell (sruffell) 2011-10-07 11:05:59.140-0500

Daniel, could you try with the current tip of the 2.5 branch.  I fear [r10207|http://svnview.digium.com/svn/dahdi?view=rev&rev=10207] might be related here.
By: Daniel (danielp) 2011-10-07 11:34:42.533-0500

Sorry, same problem with actual SVN-branch-2.5-r10219M.
By: Shaun Ruffell (sruffell) 2011-10-07 11:42:51.168-0500

Would it be possible to grab a short dahdi_monitor recording that exhibits the strange audio?
By: Daniel (danielp) 2011-10-07 11:48:30.759-0500

weird faxsound
By: Shaun Ruffell (sruffell) 2011-10-10 12:31:22.939-0500

Thanks for the attachment, I should be able to try to simulate this here today and hopefully come up with some more questions depending on the results of that test.  So...it's in my court right now.
By: Daniel (danielp) 2011-10-10 15:14:05.106-0500

echo 127 >/sys/module/wct4xxp/parameters/debug
By: Daniel (danielp) 2011-10-10 15:14:30.816-0500

I have made some tests. Recompiling dahdi-linux, dahdi-tools and asterisk. Disable echo canceller. Reboot server. All unsuccessful. dahdi-linux-2.4.1.2 is the last version which work without problems for me.

I activated debugging with "echo 127 >/sys/module/wct4xxp/parameters/debug". It looks normal. See attached file "debug.txt".

Thank you!
By: Daniel (danielp) 2011-11-01 06:34:57.584-0500

For information: upgrade to kernel 2.6.35.14 does not help.
By: Shaun Ruffell (sruffell) 2011-11-02 11:14:00.268-0500

Just to make sure:

Do you see (MASTER) when you run the following command:

{noformat}
for span in $(ls /proc/dahdi); do head -n1 /proc/dahdi/$span; done
{noformat}

The audio is strange in that you have 2 milliseconds and then 8 milliseconds of silence (running with CONFIG_HZ set to 100?)....so it does seem by all accounts very similar to [r10207|http://svnview.digium.com/svn/dahdi?view=rev&rev=10207].  Also, when you're in this state are any of the spans showing any other alarms?

One thing you can do to check is in include/dahdi/kernel.h (in the dahdi-linux sources) change DAHDI_CB_SIZE from 2 to something like 16.  This will use more memory but allow more samples of channels in conference to be buffered up.  If this resolves the problem for you, then it must be that sometimes the BRI card span is not the master when it should be.
By: Shaun Ruffell (sruffell) 2011-11-02 14:03:19.172-0500

Ok...this is the same thing as r10207, but the wcb4xxp driver does something funny by setting the "RUNNING" flag when the spans are first started.  I'll commit a change to master / 2.5 here soon for you to try out. Sorry about the delay on this.
By: Shaun Ruffell (sruffell) 2011-11-02 14:48:51.035-0500

dahdi-linux 2.6.0-rc1 (contains [r10304|http://svnview.digium.com/svn/dahdi?view=rev&rev=10304]) which I believe will resolve this issue for you.

the rc1 should be tagged momentarily, but if not, you could try dahdi-linux trunk which has that commit in it.
By: Shaun Ruffell (sruffell) 2011-11-02 15:04:50.062-0500

:( and I just realized I did not put this issue number in the commit or add your name as the reporter.  I'll be sure to add a note to the release announcement. Sorry about that.

I've also moved that change down onto the 2.5 branch too if you want to try the tip of the 2.5 branch instead of 2.6.0-rc1 / trunk.
By: Daniel (danielp) 2011-11-02 15:56:09.585-0500

Hey, i'm running SVN-trunk-r10304 and it works fine! Tomorrow i will extensive test some scenarios, to be sure that the error no longer occurs. Thank you very much!
By: Shaun Ruffell (sruffell) 2011-11-04 09:54:25.593-0500

Daniel, I'll go ahead and close this right now as fixed in 2.6.0. Feel free to reopen if this is a hasty close.  Thanks for reporting and testing.