On Thu, Mar 24, 2011 at 8:01 PM, Christoph Lameter <cl@xxxxxxxxx> wrote:
> On Thu, 24 Mar 2011, Pekka Enberg wrote:
>
>> On Thu, Mar 24, 2011 at 7:43 PM, Christoph Lameter <cl@xxxxxxxxx> wrote:
>> > The bug should only trigger on old AMD64 boxes that do not support
>> > cmpxchg16b.
>>
>> Yup. Ingo is it possible to see /proc/cpuinfo of one of the affected
>> boxes? I'll try your config but I'm pretty sure the problem doesn't
>> trigger here. Like I said, I think the problem is that alternative
>> instructions are not patched early enough for cmpxchg16b emulation to
>> work for kmem_cache_init(). I tried my check_bugs() patch but it hangs
>> during boot. I'll see if I can cook up a patch that does
>> alternative_instructions() before kmem_cache_init() because I think
>> those *should* be available during boot too.
>
> I forced the fallback to the _emu function to occur but could not trigger
> the bug in kvm.
That's not the problem. I'm sure the fallback is just fine. What I'm
saying is that the fallback is *not patched* to kernel text on Ingo's
machines because alternative_instructions() happens late in the boot!
So the problem is that on Ingo's boxes (that presumably have old AMD
CPUs) we execute cmpxchg16b, not the fallback code.
Pekka
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>