[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#597828: linux-image-2.6.32-5-amd64: [xen] debug options + sysrq-t => sending NMI to all CPUs: BUG: unable to handle kernel paging request



On Fri, 2010-09-24 at 14:02 -0700, Jeremy Fitzhardinge wrote:
> On 09/24/2010 01:28 PM, Ian Campbell wrote:
> >> I think the Intel MCE people have made NMI work in pvops, but I didn't
> >> look closely.
> >>
> >> But from a pvops perspective, I think the tricky part is sending an NMI
> >> rather than receiving.
> > It's not just "HYPERVISOR_vcpu_op(VCPUOP_send_nmi, cpu, NULL)"?
> 
> That part's simple, but how to expose it in pvops?  A "send NMI" would
> be the lowest-level version, but not very generic.

It seems like generally useful functionality to me, wouldn't other
sub-archs want it too? e.g. UV appears to use a non-x86 APIC mechanism
for sending IPIs, although its not clear if the non-x86-ness of their
APIC extends to sending NMIs differently or not.

> We could do something like "dump all CPU backtraces" as a high-level operation
> without needing any NMIs/IPIs.

I think the only hypercalls which return VCPU state are domctl's and
hence not much use to us in the kernel.

Ian.
-- 
Ian Campbell
Current Noise: Judas Priest - Leather Rebel

Have you noticed that all you need to grow healthy, vigorous grass is a
crack in your sidewalk?




Reply to: