Subject: RE: multisegment memory problems
To: Steve Woodford <swoodfor@bluews.com>
From: Todd Vierling <tv@pobox.com>
List: tech-kern
Date: 07/14/1998 10:08:15
On Tue, 14 Jul 1998, Steve Woodford wrote:

: 	>Can we collect some data of the machines with problems?
: 	>Are you all running UVM? Anything else these machines have in
: common?
: 
: Shark (233MHz StrongARM, with 32Mb RAM) running -current as of just over
: a week ago. ISTR Jason saying that UVM is now mandatory on the Shark.
: Todd's machine is also a Shark, and should be pretty much identical to
: mine (with or without local disk).

Reproduced on a Shark both with and without local disk.  Haven't tried
backing down to a "Mach VM with MNN" setup, which is at this point rather
painful to do.  So yes, it's with UVM.

This happens with two DIMMs installed (IIRC, that's actually four memory
segments, 8MB each), and I haven't been able to reproduce it with only one
16MB DIMM installed (funky problems with high load conditions and NFS swap 
prevent me from getting that far through the compile).

Mark Brinicombe was not, however, able to reproduce it on another arm32
machine, the CATS board, which has a _single_ 32MB memory segment.  I'm more
and more wondering if there's some special magic in MNN that's just being
slightly misused.... :>

-- 
-- Todd Vierling (Personal tv@pobox.com; Bus. todd_vierling@xn.xerox.com)