Opteron ECC

[James Van Artsdalen]

Several weeks ago I reported an instability in my system WRT ECC.

It turns out that AMD has published its Opteron errata sheet and
errate item 101 appears to be the issue: a bug in the Opteron means
you can't have both "node interleave" and ECC scrubbing on at the same
time.

I've turned off node interleave.  I have no idea what are reasonable
values for ECC scrubbing in ROM setup but it seems more useful to
leave scrubbing on than to enable node interleave, even without a
NUMA-aware kernel.

(this only applies to multiprocessor systems)

> From: James Van Artsdalen <james-freebsd-amd64 at jrv.org>
> Subject: Re: shaky support or my system?
> 
> > Date: Wed, 4 Feb 2004 06:07:19 -0600 (CST)
> > From: James Van Artsdalen <james at jrv.org>
> > 
> > > From: Adriaan de Groot <adridg at cs.kun.nl>
> > > Date: Wed, 4 Feb 2004 12:51:45 +0100
> > > 
> > > > in either fbsd5.2 nor the recent 5.2.1RC. One of the things that makes
> > > > it crash for sure is the /usr/ports/benchmarks/ubench program. It does
> > > 
> > > Don't run that, it'll just tell you that the system is only marginally faster 
> > > than the athlon XP it replaces. 
> > 
> > I've reproduced Janne's panic with ubench twice.  Both show similar
> > outputs from the panic.  Panic #3 is proving more elusive: I'm running
> > ubench for the fourth time in a row now waiting for crash #3.
> 
> Unfortunately it appears my case was a false alarm.
> My Tyan K8W S2885 is not stable unless ECC is turned off in ROM setup.