Greg KH wrote:
> I got the following oops on 2.5.41 + latest bk tree as of 3 pm PST under
> heavy ide load:
> 
> It's a UP box running with SMP and preempt enabled.
> 
> Unable to handle kernel NULL pointer dereference at virtual address 00000004
> *pde = 00000000
> Oops: 0002
> hid uhci-hcd usbcore  
> CPU:    0
> EIP:    0060:[<c0125292>]    Not tainted
> EFLAGS: 00010016
> EIP is at run_timer_tasklet+0x102/0x220
 > <snip>
> Call Trace:
>  [<c01215c0>] tasklet_hi_action+0x80/0xd0
>  [<c01212bb>] do_softirq+0x5b/0xc0
>  [<c01114d1>] smp_apic_timer_interrupt+0x111/0x140
>  [<c0109b0f>] do_IRQ+0x18f/0x230
>  [<c0108256>] apic_timer_interrupt+0x1a/0x20

Notice that eip is actually in run_timer_tasklet and __run_timers is 
inlined there.  You're actually crashing in __run_timers, in the same 
place as me, before Ingo's fix.  Try the attached patch that I got 
from Ingo this morning.  There is still some kind of problem for me, 
but it might be a completely different one.

My oops is while running Specweb, so high interrupt load for me too.
-- 
Dave Hansen
haveblue@us.ibm.com