What is MCE?

What is MCE?

[Norman Diamond]

Sometimes:
MCE: The hardware reports a non fatal, correctable incident occurred on CPU
0.
Bank 3: e20000000002010a

Sometimes:
MCE: The hardware reports a non fatal, correctable incident occurred on CPU
0.
Bank 3: a200000000080a01

Obviously this isn't a Linux error, Linux is being kind enough to report a
hardware error to me, but I don't know how to interpret these error flags.
I don't even know what MCE is but whoever developed it for Linux surely must
know.  Please, how can I find out what this is?

Re: What is MCE?

[Randy.Dunlap]

On Sun, 28 Dec 2003 11:33:24 +0900 "Norman Diamond" <ndiamond@wta.att.ne.jp> wrote:

| Sometimes:
| MCE: The hardware reports a non fatal, correctable incident occurred on CPU
| 0.
| Bank 3: e20000000002010a
| 
| Sometimes:
| MCE: The hardware reports a non fatal, correctable incident occurred on CPU
| 0.
| Bank 3: a200000000080a01
| 
| Obviously this isn't a Linux error, Linux is being kind enough to report a
| hardware error to me, but I don't know how to interpret these error flags.
| I don't even know what MCE is but whoever developed it for Linux surely must
| know.  Please, how can I find out what this is?

It's an x86 Machine Check Exception, documented in the Intel
IA-32 docs.  And probably in some other x86-like docs.

There's a 'parsemce' program around that attempts to decode some
of those bits.  It might or might not help you out.
It's in this directory:
  http://www.kernel.org/pub/linux/kernel/people/davej/tools/

--
~Randy

Re: What is MCE?

[Joshua Kwan]

[-- Attachment #1: Type: text/plain, Size: 692 bytes --]

On Sun, Dec 28, 2003 at 11:33:24AM +0900, Norman Diamond wrote:
> Sometimes:
> MCE: The hardware reports a non fatal, correctable incident occurred on CPU
> 0.
> Bank 3: a200000000080a01
> 
> Obviously this isn't a Linux error, Linux is being kind enough to report a
> hardware error to me, but I don't know how to interpret these error flags.
> I don't even know what MCE is but whoever developed it for Linux surely must
> know.  Please, how can I find out what this is?

MCE stands for Machine Check Exception.

This mailing list reply provides a pretty good explanation of it:

http://lists.debian.org/debian-user/2000/debian-user-200009/msg01677.html

-- 
Joshua Kwan

[-- Attachment #2: Type: application/pgp-signature, Size: 827 bytes --]