Re: [PATCH] lockless MCE i386 port

From: Andi Kleen
Date: Sun Nov 14 2004 - 03:54:50 EST


On Sat, Nov 13, 2004 at 11:36:42AM -0700, Zwane Mwaikambo wrote:
> On Fri, 12 Nov 2004, Zwane Mwaikambo wrote:
>
> > Andi fixed the locking issues with respect to printk and MCEs on x86_64,
> > this is a port of said code with a few small changes due to i386 currently
> > supporting the extended MCE MSRs on intel processors (also present on
> > intel x86_64). The addition of hooks allows for the Pentium and P4 MCE
> > drivers to print additional/extended information. I've also converted the
> > P4 thermal monitor driver to avoid printing from interrupt context.
> >
> > Tested on P4 and K7
> >
> > Signed-off-by: Zwane Mwaikambo <zwane@xxxxxxxxxxx>
> >
> > arch/i386/kernel/cpu/mcheck/k7.c | 68 ++++++++++-----
> > arch/i386/kernel/cpu/mcheck/mce.c | 109 ++++++++++++++++++++++++-
> > arch/i386/kernel/cpu/mcheck/mce.h | 53 ++++++++++++
> > arch/i386/kernel/cpu/mcheck/p4.c | 165 ++++++++++++++++++++++++++------------
> > arch/i386/kernel/cpu/mcheck/p5.c | 27 ++++--
> > arch/i386/kernel/cpu/mcheck/p6.c | 66 +++++++++------
> > 6 files changed, 382 insertions(+), 106 deletions(-)
>
> I flipped the printk output for the m.addrl/m.addrh, here is the patch
> rediffed.

Looks good from a first look.

One issue is that you will need people to run the mcelog cron job
and create /dev/mcelog, otherwise they won't see any non fatal
warnings anymore.

I'm actually considering to add a tasklet/bit on x86-64 to printk
a one line warning when any events are in the log. Perhaps that
would be a good idea here too to make the migration smoother.

-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/