Remote hybrid echo not being canceled by hardware ec

Jessie,

I understand that the EC is only working on active calls. What I'm trying to say is that if I place a call, or receive a call, and it happens to be on channel 1, the EC works. If it ends up on any other channel (admittedly I've only tested 2-6), including if channel 1 is currently on a call, the EC doesn't work on that channel. During the day, there are 2-8 active channels at any given time. Again, if I happen to get channel 1, everything works. Otherwise, it doesn't. What concerns me is that /proc/dadh/1 shows the echo canceller for every call, even though it doesn't work (unless the call I'm testing with happens to be on channel 1).

-CJO-

I have yet more info. After unloading all modules and reinserting them with the following in /etc/modprobe.conf:
options wctdm24xxp debug=2
options wct4xxp debug=16
options opvxd115 debug=16


Echo cancelling is working on all channels. Admittedly, I have only tested channels 1, 2, 4, 11, 15, 16, 22 and 23. I will play with the modules while you're on holiday to see if I can reliably reproduce the results I was seeing earlier with only one channel having echo cancellation. dmesg does not complain about anything after the debug and I can see the echo-can being turned on and off for each channel.

Happy New Year!

-CJO-

After 48 hours and 1148 calls, we are now getting reports of echo again. I have confirmed that the echo canceler is not working on most channels--I was unable to test them all. dmesg shows the ec turning on and off but it is no longer working. I have also confirmed that echo cancelling DOES work on channel 1, just not the other channels, so this can be reproduced.

Please tell me what to set on the modules to get the full debug. I will also be happy to give your developers access to my system to diagnose and correct this problem. We are also willing to pay a premium to make this a priority. I will plan on reloading the modules every evening until you return from holiday.

So to confirm my suspicions, the echo canceler works after a module reload and stops working some time later. I see nothing in dmesg indicating an error. Here is the dmesg during the test calls. I was using channel 14 and 23:

echocan: Card is 0, Channel is 14, Span is 0, offset is 72 length 128
echocan: Card is 0, Channel is 5, Span is 0, offset is 36 length 128
echocan: Card is 0, Channel is 4, Span is 0, offset is 32 length 0
Unassigning channel 0/4!
Unassigning channel 0/4!
echocan: Card is 0, Channel is 14, Span is 0, offset is 72 length 0
Unassigning channel 0/14!
Unassigning channel 0/14!
echocan: Card is 0, Channel is 23, Span is 0, offset is 108 length 128
echocan: Card is 0, Channel is 2, Span is 0, offset is 24 length 0
Unassigning channel 0/2!
Unassigning channel 0/2!
echocan: Card is 0, Channel is 23, Span is 0, offset is 108 length 0
Unassigning channel 0/23!
Unassigning channel 0/23!


This is the dmesg from a working echocan call:

echocan: Card is 0, Channel is 1, Span is 0, offset is 20 length 128
echocan: Card is 0, Channel is 1, Span is 0, offset is 20 length 0
Unassigning channel 0/1!
Unassigning channel 0/1!

Hello,
Card was replaced last thursday. No echo on friday. By monday morning, echo was back on all channels except channel 1. Modules reloaded monday evening, no echo today.

I found something about a race condition causing hardware echo to not be enabled correctly in every 2.2 up to 2.2.1.1, and fixed in 2.2.1.2. I'm going to take a stab at 2.4 and see what happens unless you guys have any idea.

Also, can a mod change the title of this thread to something more descriptive. Originally, I had thought that the echo canceler just didn't work, but that's not the case.

Update:
Upgraded to 2.4 kernel and asterisk dahdi modules and tools Sunday evening. No echo was present after upgrade or the following day. Today, the echo reappeared roughly around 10AM our time. I know know that because I happened to test a random channel this morning--which had no echo. Some time later, tested again, call had echo. Tried 5 more channles, all had echo except channel 1.

What is our next course of action? Do you guys want to get into the box and take a look?

I should clarify my last post, not "2.4 kernel", but dahdi 2.4 asterisk and kernel modules.
Also, I see this in the release notes for the DAHDI drivers. Could this problem still be present in OpenVOX's code?

2009-11-12 23:45 +0000 [r7572-7584] Shaun Ruffell <This email address is being protected from spambots. You need JavaScript enabled to view it.> * voicebus: Fix race when enabling/disabling hardware echocan. This closes a race condition where it was possible for the driver to believe it has enabled the VPMADT032 when in fact, it really has not. This fixes a regression introduced in dahdi-linux 2.2.0. (issue #15724)