Andrew Morton wrote:
> Brad Laue <brad@brad-x.com> wrote:
> 
>> Brad Laue wrote:
>> > Hopefully the attached shows some irregularity. If not, I'll have to 
>> > reply back in a few weeks when the problem recurs over the course of time.
>>
>> And without further ado, the attachment. It's been a long day. :)
> 
> 
> It beats me.  Something must be waking up ksoftirqd all the time.
> 
> If you have time, could you please apply the below, then wait for ksoftirqd
> to go bad again and then run:
> 
> 
> 	dmesg -c
> 	echo 1 > /proc/sys/debug/0 ; sleep 1; echo 0 > /proc/sys/debug/0
> 	dmesg -s 1000000 > /tmp/foo
> 
> and then send foo?
> 

Okay, it looks like the 'pppoe' process is definitely causing all the
soft interrupts. Puzzling, because on other systems with similar
hardware this doesn't happen. Ran the machine without iptables running
(eek!!!) just to make sure, and the same result occurred, so it doesn't
seem to be the fact that packets are going through an elaborate ruleset
that's causing it. Kernel mode PPPoE doesn't seem to change things much.

Hard to figure out from a kernel perspective why this many soft 
interrupts are being used, but I'm just getting a grasp on what their 
purpose is to begin with.

Hope this offers some insight.

Brad