I have two pentium boxes running NetBSD-1.0.  Both are Micron systems
with Micronics motherboards; one is a p90, the other a P100.  Their
peripherals are identical: a BT-946C scsi interface and a Diamond 864
Dram video card.

Both machines will run successfully (with an uptime of over a month,
when they are rebooted for maintenance) with 32 or 64 MB of memory.
However, If I try running 96MB in either system, "something
happens".

Symptoms:
1] The dmesg kernel buffer is corrupted shortly after boot.

2] When the kernel is logging messages quickly (for example,
a slew of mbi not found) Processes often have their address
space corrupted with the text of the kernel messages.


I browsed the kernel, and didn't notice (which may be the problem)
any overtly fixed size resources which would overflow with the
addition of the additional memory.  Well, ... the only thing which is
slightly "suspicious" is the default limit of the kernel malloc arena
to 6MB, which I haven't traced the impact of.

Has anyone else seen a problem like this, know how to fix it, or can
point me in the right direction?

Thanks for any help!
| Joe Burger	U of WI-Madison Computer Sciences | "No matter where you go,
| "bolo"	bolo@cs.wisc.edu   uwvax!bolo     |  There you are."
|               http://www.cs.wisc.edu/~bolo/     |     -- Buckaroo Banzai