[RFC PATCH] Bug during kexec...not all cpus are stopped
vgoyal at redhat.com
Tue Oct 12 18:17:17 EDT 2010
On Mon, Oct 11, 2010 at 03:10:11PM -0700, Eric W. Biederman wrote:
> Vivek Goyal <vgoyal at redhat.com> writes:
> > On Mon, Oct 11, 2010 at 12:41:23PM -0700, Alok Kataria wrote:
> > I don't think that kdump path uses smp_send_stop().
> It doesn't.
> > IIUC, on x86, we directly send NMI to other cpus.
> > native_machine_crash_shutdown()
> > kdump_nmi_shootdown_cpus()
> > nmi_shootdown_cpus()
> > smp_send_nmi_allbutself
> > apic->send_IPI_allbutself(NMI_VECTOR);
> > So above description should be limited to only panic() path.
> Is it actually confusing? With respect to documenting the line
> of thinking it seems reasonable.
No, just wanted to point out that let us modify the changelog to remove
keyword "kdump" from it.
> > On a side note, I am wondering why panic() and kdump path can't share the
> > shutdown routine.
> Hysterical raisins. Andi's change to smp_send_stop says that NMIs not
> working on some boxes. When someone wants to weed through all of the
> insanity it would probably be good to get the panic and the kdump paths
> sharing code. For now simply separating panic and reboot should be
> enough, and it lets the code evolve where it needs to.
Ok. Agreed that atleast conceptually kdump and panic() path should share
the code. But that's a different problem altogether and this patch can go in.
More information about the kexec