Re: cleaning up NMI handling in trap()

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Fri, Mar 18, 2011 at 03:27:00PM -0500, David Young wrote:
> Today I noticed that in trap() on amd64 and i386, the code for handling
> NMI is a rat's nest of #ifdefs.  I've tried to clean it up by adding
> some weak stub implementations for DDB and KGDB and letting the linker
> instead of the preprocessor configure the kernel.  Please have a look at
> the attached patch.  I'd like to commit something like it pretty soon.

The
        printf ("NMI ... going to debugger\n");

is usefull in knowing what's going on. Can we please keep it in, in
one form or another ?

nmi_dispatch() is not available for xen; but it's now called inconditionally
from trap(). Did you forget to add a dummy implementation for XEN ?

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--