AS4100/533 Machine checks still on 5.4-RELEASE

[Wilko Bulte]

On Fri, Nov 04, 2005 at 08:56:01AM -0600, Bill Harris wrote..
> I replaced one, then the other CPU in this machine over the last 3
> months.   It works great for a few weeks, then dumps again.   Now, it's
> every couple of days.
> 
> I've cvsupped the kernel source from the release tree to make sure
> I had the latest patches just in case.   I'm at a loss to what's causing
> this.   I've got a similarly equipped AS4100 at home, that's running
> fine with the same code.   This is a web and border mail server, but
> it's not very heavily loaded.   I'm guessing that the perl script is my 
> MRTG
> scripts that run every 5 mins or so.
> 
> It crashed again last night..
> 
> 
> unexpected machine check:
> 
>     mces    = 0x1
>     vector  = 0x670
>     param   = 0xfffffc0000004838
>     pc      = 0x1601fdd10
>     ra      = 0x1601fda8c
>     curproc = 0xfffffc00301cf490
>         pid = 29125, comm = perl5.8.6
> 
> panic: machine check
> cpuid = 0
> Uptime: 10h40m39s
> Cannot dump. No dump device defined.
> Automatic reboot in 15 seconds - press a key on the console to abort
> Rebooting...
> 
> Since I've replaced both processors, what should I do next?

A machinecheck 670 is a CPU detected error.  The manuals say that typically
the CPU that detected the error is broken ;-)  But FreeBSD does not contain
the smarts of Tru64 or VMS that allows further diagnosis than this.

-- 
Wilko Bulte				wilko at FreeBSD.org