Re: Status on CPU hotplug issues

[Bryce Harrington]

On Mon, Oct 09, 2006 at 02:40:24PM -0700, Randy Dunlap wrote:
> On Sat, 7 Oct 2006 21:57:49 +0000 Pavel Machek wrote:
> 
> > Hi!
> > 
> > > 1.  Oops offlining cpu twice on AMD64 (but not on EM64t)
> > >     with the 2.6.18-git22 kernel
> > > 
> > >     Reported to hotplug lists 10/05:
> > >       http://lists.osdl.org/pipermail/hotplug_sig/2006-October/000680.html
> > > 
> > >     To recreate: offline, online, and then offline a CPU, then oopses
> > >       http://crucible.osdl.org/runs/2397/sysinfo/amd01.console
> > >       http://crucible.osdl.org/runs/2397/sysinfo/amd01.2/proc/config
> > > 
> > >     Here's a snippet of the oops:
> > > 
> > > # echo 0 > /sys/devices/system/cpu/cpu1/online
> > > 
> > >  Unable to handle kernel NULL pointer dereference at 0000000000000000 RIP:
> > >  [<ffffffff80255287>] __drain_pages+0x29/0x5f
> > > PGD 7e56d067 PUD 7ee80067 PMD 0
> > > Oops: 0000 [1] PREEMPT SMP
> > > CPU 0
> > > Modules linked in:
> > > Pid: 7203, comm: bash Tainted: G   M  2.6.18-git22 #1
> >                                  ~~~~~
> > kernel is unhappy here. Forced module unload?
> 
> Machine check exception.  'G' is Good, same place where 'P'
> for proprietary would be.  But yes, kernel or machine is unhappy.

To followup on this issue...

I found a BIOS update for the motherboard of this machine indicating it
includes a fix for MCE during hibernate operations; my guess is that
cpu hotplug may be triggering this bug.

Meanwhile, we checked against a couple other different AMD64 systems;
these are behaving correctly.

Anyway, thanks for the pointers, it sounds like this is probably just a
hardware issue.  I'll report back if I find differently.

Bryce

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/