MCE handler in 2.6.37 kernel

MCE handler in 2.6.37 kernel

[Mukesh Rathor]

Hi Jeremy,

I am playing around with MCE stuff, and looking to test MCE handler in
the guest kernel next. Just wondering why we are not installing
MCE handler in the hypervisor in 2.6.37 kernel.


arch/x86/xen/enlighten.c:

cvt_gate_to_trap():

#ifdef CONFIG_X86_MCE
        } else if (addr == (unsigned long)machine_check) {
                return 0;
#endif


thanks,
Mukesh

Re: MCE handler in 2.6.37 kernel

[Ian Campbell]

On Wed, 2011-02-02 at 03:39 +0000, Mukesh Rathor wrote: 
> Hi Jeremy,
> 
> I am playing around with MCE stuff, and looking to test MCE handler in
> the guest kernel next. Just wondering why we are not installing
> MCE handler in the hypervisor in 2.6.37 kernel.

In these cases the most likely reason is that nobody with an interest in
MCE has yet stepped up to port the feature to the pvops kernel. Patches
always welcome though ;-)

Ian.

Re: MCE handler in 2.6.37 kernel

[Jeremy Fitzhardinge]

On 02/01/2011 07:39 PM, Mukesh Rathor wrote:
> Hi Jeremy,
>
> I am playing around with MCE stuff, and looking to test MCE handler in
> the guest kernel next. Just wondering why we are not installing
> MCE handler in the hypervisor in 2.6.37 kernel.
>
>
> arch/x86/xen/enlighten.c:
>
> cvt_gate_to_trap():
>
> #ifdef CONFIG_X86_MCE
>         } else if (addr == (unsigned long)machine_check) {
>                 return 0;
> #endif

There's a patch from the Intel folks to implement MCE, but I hadn't
added it to my upstream todo list yet.  I think it's pretty small, so no
problem to add it - esp if you can actually test it.

    J

Re: MCE handler in 2.6.37 kernel

[Mukesh Rathor]

On Wed, 02 Feb 2011 09:48:14 -0800
Jeremy Fitzhardinge <jeremy@goop.org> wrote:

> On 02/01/2011 07:39 PM, Mukesh Rathor wrote:
> > Hi Jeremy,
> >
> > I am playing around with MCE stuff, and looking to test MCE handler
> > in the guest kernel next. Just wondering why we are not installing
> > MCE handler in the hypervisor in 2.6.37 kernel.
> >
> >
> > arch/x86/xen/enlighten.c:
> >
> > cvt_gate_to_trap():
> >
> > #ifdef CONFIG_X86_MCE
> >         } else if (addr == (unsigned long)machine_check) {
> >                 return 0;
> > #endif
> 
> There's a patch from the Intel folks to implement MCE, but I hadn't
> added it to my upstream todo list yet.  I think it's pretty small, so
> no problem to add it - esp if you can actually test it.
> 
>     J

Yeah, I can do basic testing if you send or point me to the Intel patch
you mention.

thanks
Mukesh