machine check on 4100 5.4-RELEASE

[Wilko Bulte]

On Fri, Aug 19, 2005 at 01:09:17AM +0200, Bernd Walter wrote..
> On Thu, Aug 18, 2005 at 10:24:42AM -0500, harrisb at rcisd.org wrote:
> > Here's the panic info:
> > 
> > --------------------------------------------
> > Mounting root from ufs:/dev/da0a
> > 
> > unexpected machine check:
> > 
> >     mces    = 0x1
> >     vector  = 0x670
> >     param   = 0xfffffc0000004e10
> >     pc      = 0xfffffc000072faa8
> >     ra      = 0xfffffc000072bba4
> >     curproc = 0xfffffc002e7cc000
> >         pid = 693, comm = perl5.8.6
> > 
> 
> I could be completly wrong, but vector 0x670 (on AS4100) should be
> CPU detected errors, such as cache failure - either B-Cache or
> internal, which points to a CPU module.

Correct.  Indicates hardware failure.  Just checked the fault analysis
docs at work.

> Are they always the same, or will they happen at random times?
> 
> In any case you should check the SRM error log - might be something
> interesting in there.
> If they are logged they should appear as MCHK 670 events.
> 
> > On Wed, Aug 17, 2005 at 04:53:28PM -0500, harrisb at rcisd.org wrote:
> > > All of a sudden, I'm getting regular crashes with machine check's.
> > > 
> > > I've pulled one of the 533 CPU's, which didn't help, and now am 
> > > wondering if it's possible that my instance of Mysql with all it's 
> > > unaligned errors
> > > could possibly cause it crash?  I've stopped the mysql daemon for a
> > > while just to see if it stabilizes.   Anyone have any ideas? 
> > > 
> > > It will crash after it's been up for days, and then immediately after 
> > > reboot.
> > 
> > Details about the machine checks would be interesting.
> > 
> > Unaligned errors in userland are corrected or the appplication is
> > terminated, depending on configuration.
> > Only unaligned faults inside the kernel are fatal.
> > 
> > > I keep thinking hardware, but all the srm test fine.
> > 
> > Hard- and software is possible, but without further details this is
> > hard to say.
> 
> -- 
> B.Walter                   BWCT                http://www.bwct.de
> bernd at bwct.de                                  info at bwct.de
> 
> _______________________________________________
> freebsd-alpha at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-alpha
> To unsubscribe, send any mail to "freebsd-alpha-unsubscribe at freebsd.org"
--- end of quoted text ---

-- 
Wilko Bulte				wilko at FreeBSD.org