Re: [tip:x86/debug] x86, reboot: Use NMI instead of REBOOT_VECTOR to stop cpus

[Don Zickus <dzickus@redhat.com>]

On Wed, Dec 21, 2011 at 10:24:53AM -0800, Yinghai Lu wrote:
> On Wed, Dec 21, 2011 at 6:59 AM, Don Zickus <dzickus@redhat.com> wrote:
> > On Tue, Dec 20, 2011 at 02:38:39PM -0800, Yinghai Lu wrote:
> >> > @@ -230,7 +285,7 @@ struct smp_ops smp_ops = {
> >> >        .smp_prepare_cpus       = native_smp_prepare_cpus,
> >> >        .smp_cpus_done          = native_smp_cpus_done,
> >> >
> >> > -       .stop_other_cpus        = native_stop_other_cpus,
> >> > +       .stop_other_cpus        = native_nmi_stop_other_cpus,
> >> >        .smp_send_reschedule    = native_smp_send_reschedule,
> >> >
> >> >        .cpu_up                 = native_cpu_up,
> >>
> >> this broke kexec on our intel nehalem, westmere and sandbridge platforms.
> >> system get reset while try to kexec second kernel.
> >
> >
> > Hmm. Ok.  Does the reboot path work correctly?
> 
> Yes.
> 
> > Vivek showed me that the
> > kexec and reboot paths do the same shutdowns.  Perhaps the second kernel
> > has trouble dealing with cpus spinning in an NMI context and can't
> > properly reset them.
> 
> not sure.
> when use nonmi_ipi in first kernel, it will work well.

Ok.  I tried taking the cpus out of NMI context and putting them into irq
context with irq_work_queue() but that didn't seem to work.  It seems to
be hanging in the new kernel somewhere.  I'll have to wait until I get
back into the office next year to debug with early_printk and a vga
console.

Cheers,
Don