Brian D Chase wrote:
> 
> On Tue, 14 Sep 1999, Dann Lunsford wrote:
> 
> > My *tentative* conclusion is that the 8MB I removed were bad.  Anybody
> > got a memory test?  Another possibility is that there's some sort of
> > 32MB boundary problem in VS3100m76's.  My money is on bad memory (I
> > mean that literally, I'm putting in an order for 8MB to replace this
> > from Keyways).
> 
> Don't rush on that just yet.  I'd suggest checking things further.
> Rotate the suspected bad chip back in and pull one of the ones you think
> is good. Try crashing the system again.  I don't know of any conclusive
> memory boundary issues, but I definitely wouldn't rule it out.

Good advice, and I've taken the last few days to test.  Two suspect 4MB
modules;
with either of them in (i.e., a 28MB machine) I experience none of the 
strange crashes -- /etc/security runs fine, telnetting into the machine 
works, etc.  Still get inetd.core, sh.core in / on shutdown, but that 
appears to be a different issue, as does the gzip.core generated on
some *.tgz files.  If I put both modules in, *BLOOEY*.  Does not
appear to generate a crash dump, dangit.  I was almost looking forward
to
digging into one.    Almost.

I have also tried putting both my suspects in and taking out one of the
newer ones.  Machine seems fine at 28MB in this config, also.  So, 
there appears to be a problem with this VS3100m76 at 32MB.  It only
shows up under NetBSD, not VMS.  I am thinking we have a memory boundary
problem here, people.  Anyone else out there seeing this?  Or am I the
only one crazy enough to put 32MB in a VS3100m76? :-)  Wouldn't 
really surprise me...

Thanks for the help, Brian.

Dann Lunsford.