Summary:ASTERISK-02096: X100P driver crashes 4kstacks kernel
Reporter:firefury (firefury)Labels:
Date Opened:2004-07-23 10:43:33Date Closed:2004-09-25 02:37:46
Versions:Frequency of
Environment:Attachments:( 0) minicom.cap
Description:The X100P driver produces an IRQ stack overflow on a kernel compiled in 4k-stacks mode.  This will increasingly cause problems since all distributions are moving towards 4k-stacks kernels for performance reasons.

I estimate about 25% of calls (both incoming and outgoing) cause the crash - the kernel reports a stack overflow and then there appear to be a number of oopses immediately after the overflow and a complete kernel lockup which are probably collateral damage cause by the stack being trashed.

I have attached the log of the kernel output.
Comments:By: Olle Johansson (oej) 2004-07-23 10:51:21

Please confirm version of Asterisk and Zaptel drivers

By: firefury (firefury) 2004-07-23 11:55:28

Asterisk-0.9.1, Zaptel drivers checked out of CVS on 21st July 2004

By: Olle Johansson (oej) 2004-07-23 11:59:15

Please check with latest code and confirm if the problem is still there. Update all libraries (zaptel, zapata) and Asterisk.

By: Mark Spencer (markster) 2004-07-23 12:23:43

Also, have you changed the zaptel Makefile, sources, headers in any way whatsoever, or is it a completely stock cvs checkout?

By: firefury (firefury) 2004-07-23 15:41:51

All completely stock, nothing was modified.

By: Olle Johansson (oej) 2004-07-23 16:28:52

OK. Then update to latest CVS and confirm if you still have the problem.

By: Mark Spencer (markster) 2004-07-24 10:01:14

Also, can you demonstrate it only occurs with 4k stacks and is not tied to another configuration option (e.g. preemptive kernel)?

By: Mark Spencer (markster) 2004-07-24 10:16:32

Also, if you can verify the crash you have is tied to 4k stacks, can you also tell us more about when the crash occurs (e.g. at the start of the call, at the end of the call, randomly during a call, only during a call)

By: Mark Spencer (markster) 2004-07-24 12:18:04

Okay, I did some analysis on the stack using "stackcheck" and i've substantially reduced the stack requirements of zaptel, so go ahead and check it should be good now.

By: Mark Spencer (markster) 2004-07-26 15:48:35

I'm just going to assume this one is fixed.  The bug placer can reopen it if it isn't.