Re: [RFC][PATCH] linux-2.5.59_lost-tick_A0

[john stultz <johnstul@us.ibm.com>]

On Tue, 2003-01-21 at 15:32, Andrew Morton wrote:
> john stultz wrote:
> > 
> > All,
> >         This patch addresses the following problem: Linux cannot properly
> > handle the case where interrupts are disabled for longer then two ticks.
> > 
> 
> Question is: who is holding interrupts off for so long?

Unfortunately in my situation there is a card which can cause 30ms
stalls in an SMI handler. Yuck, I know. :P

> It might be better to implement a detection scheme inside
> the timer interrupt handler: if the time since last interrupt
> exceeds two ticks, whine and drop a backtrace.
> 
> That backtrace will point up into the local_irq_enable() in
> the offending code, and we can go fix it?

Hmm, clever! That would be a very good check for software caused stalls.
I'll try to drop that in. 

thanks for the feedback!
-john