Digium (Asterisk) PCI card crashing system

[afried]

Last night I just did a bare metal install of Ubuntu 9.04 (amd64) then brought up Asterisk 1.4.24.1. The configs for Asterisk had been used on the same box prior to the upgrade. The box also contains a Digium TDM card.

A short time later, I discovered that the box suddenly was running at 80% utilization and that Asterisk had ceased working. Looking at the process list, I found two programs consuming almost all available CPU resourses - DD and klog. My syslog, messages and kern.log files were megs in size and contained essentially garbage. Just prior to the race condition starting, however, were references to DAHDI:

Apr 25 17:17:57 ASTERISK kernel: [ 8020.982920] TDM PCI Master abort

Subsequent logs from kern.log looked like the following:

Apr 25 17:17:57 ASTERISK kernel: 221.21221.2121.2121.2121.2121.2121.2121.2121.21.21.21212121.212121212121.21.2121.21.21.02121.021.21.2121.21.21.21.212121.21.21.2122121.21.21.2121.2121.21.21.2121.21.2121
.21.21.21.2121.21.21.21.2121.21.21.212121.21.2121.21.2121.21.21.2121.21.2121.2121.21.21.21.2121.212121.21.21.21.21.21.21.212121.21.2121.21.212121.2121.21.2121.21.2121.21.21.21.21.2121.2121.21.21.21.21.2
1.21.21.21.2121.21.21.2121.21.21.212121.21.21.2121.21.21.2121.21.21.21.21.2121.2121.21.21.212121.21.21.21.2121.21.21.21.21.2121.21.21.2121.2121.2121.2121.21.2121.21.212121.2121.2121.21.2121.21.21.21.21.
2121.212121.21.212121.21.21.21.21.21.21.2121.21.21.212121.21.21.21.21.2121.21.021.21.21.2122121212121.21.21.2121.2121.21.2121212121.21212121212121221.2121212121212212122121221212122121212121212122121221
212121.2212121.2122121221.21212221212212122121221.2121212122121.2212121221212212121221212122121221212122121221.2122121212121221212212122121221212122121212121212212122121.22212121212122121221.038212121.2
122121.22121212212

This server had been running Asterisk with the same PCI cards without issues under 8.10 and 8.04.

I rebuilt and reconfigured both DAHDI and Asterisk, but the issue remains. It appears that Ubuntu's handling of the PCI bus might be the issue.

The issue occurs intermittently - sometimes minutes after being rebooted, sometimes hours. But the only solution to the problem once it surfaces is a complete reboot.

I deactivated the DAHDI drivers and the system ran without any further problems.

[Lionel Porcheron (lionel.porcheron)]

Hi,

Thanks for your bug report. Do you think you would have a change to test Ubuntu 9.10 on this system with this card?

[Ryan Finnie (fo0bar)]

FYI, I have a TDM800P (similar to the TDM400P but with two module daughterboards, but I only have one populated) running under 9.10 without problem. Unfortunately I cannot test this back to 9.04, but you should definitely try 9.10 to see if that will work. You'll have to migrate your dialplan from 1.4 to 1.6.2, but most of it just involves changing pipes to commas (Directory, VoiceMail, MeetMe, Record).

[afried (andrew-fried)]

I would love to test it, however.... there seems to be a problem
compiling Asterisk under 9.10. It appears that part of the net library
that Asterisk uses is not resolving from Ubuntu 9.10, making it quite
impossible to test further. I don't believe this is an Asterisk bug,
rather, possibly some new upgraded library with a differently named module.

Andrew

Andrew Fried
<email address hidden>

On 1/4/10 4:20 AM, Ryan Finnie wrote:
> FYI, I have a TDM800P (similar to the TDM400P but with two module
> daughterboards, but I only have one populated) running under 9.10
> without problem. Unfortunately I cannot test this back to 9.04, but you
> should definitely try 9.10 to see if that will work. You'll have to
> migrate your dialplan from 1.4 to 1.6.2, but most of it just involves
> changing pipes to commas (Directory, VoiceMail, MeetMe, Record).
>

[Matthias Urlichs (smurf)]

> I would love to test it, however.... there seems to be a problem
> compiling Asterisk under 9.10

Why do you want to do that? Just install the Asterisk 1.6 .debs from 9.10 on the machine.

[Paul Belanger (pabelanger)]

Thank you for taking the time to report this bug and helping to make Ubuntu better. The issue that you reported is one that should be reproducible with the live environment of the Desktop CD of the development release - Lucid Lynx. It would help us greatly if you could test with it so we can work on getting it fixed in the next release of Ubuntu. You can find out more about the development release at http://www.ubuntu.com/testing/ . Thanks again and we appreciate your help.

[PKirkham (pete-kirkham)]

I have a TDM400P card from Digium. I have installed Lucid. I'm using version 2.6.32-22. The system won't boot with the card in a PCI slot whether I've attached the power cable or not. Previously I was using 6.04LTS which worked fine on the same hardware.

[afried (andrew-fried)]

Good luck. I tried upgrading an existing system from 9.04 to 10.4 and
the system crashed bigtime. Apparently the problem was that Asterisk
was unable to read and parse some of the config files like
extensions.conf, particularly if there was an include statement in them.
 Stick with 9.04.

Andrew

Andrew Fried
<email address hidden>

On 6/17/10 4:59 PM, PKirkham wrote:
> I have a TDM400P card from Digium. I have installed Lucid. I'm using
> version 2.6.32-22. The system won't boot with the card in a PCI slot
> whether I've attached the power cable or not. Previously I was using
> 6.04LTS which worked fine on the same hardware.
>

[PKirkham (pete-kirkham)]

This is a bigger problem, the system won't even boot with the card in place.
The problem is reproduceable with the 10.04 Live CD. It boots without the card but hangs with the card. I changed the slot for the card too to see if I could have any influence on IRQs but the perceived behaviour is the same.

With the Live CD the boot gets to the Ubuntu screen with the dots under it and stops with five red dots on the 5th iteration I believe. I don't know if the dots mean anything or if they're just elapsed time. The processor is a Celeron so it doesn't boot very quickly. There's only 512MB of memory but I don't see why that makes any difference.

With th installed system the screen displays a soft hang message which repeats after a minute or two.

[PKirkham (pete-kirkham)]

I've now tried with 9.10 too which boots and runs correctly with the TDM card in place. So what's wrong with 10.4?

[Launchpad Janitor (janitor)]

[Expired for asterisk (Ubuntu) because there has been no activity for 60 days.]