Subject: why not REALEXTMEM
To: None <port-i386@netbsd.org>
From: Henry Nelson <henry@irm.nara.kindai.ac.jp>
List: port-i386
Date: 02/17/2000 23:13:23
Finally got around to testing my suspicious 64MB edo using "testmem.exe"
(Thanks Craig Hart!).  Anyway, only the 61Mb and 63Mb locations fail.

I recompiled the kernel with "options REALEXTMEM=60416", and was able
then to install the modules and compile a kernel, which I was never able
to do before (always stopped with a fatal "page fault trap, code=0").

In GENERIC it says:
# The following options override the memory sizes passed in from the boot
# block.  Use them *only* if the boot block is unable to determine the correct
# values.
The emphasis on "only" has me worried.  By setting REALEXTMEM to 60416,
aren't simply the last 4MB not used, so that in effect the bad memory
locations are isolated?  Okay to assume I'm _fairly_ safe?

henry nelson