> From: Jason Thorpe <thorpej@nas.nasa.gov>
> Date: Tue, 03 Mar 1998 09:49:34 -0800
> 
> On Tue, 3 Mar 1998 09:46:14 -0600 
>  Chuck Cranor <chuck@dworkin.wustl.edu> wrote:
> 
>  > this is a bug.  it isn't supposed to be cleared.   i think certain
>  > proms are touching the area where the message buffer is stored, thus
>  > causing it to be cleared.   i never had the chance to fully track it
>  > down.
> 
> The message buffer should probably be moved to the end of core so that
> it can be resized, too... it's somewhat annoying that the SPARC port
> can't do this, currently...

The end of core is generally owned by the PROM, so you might need
to put it just before the end.  But then, of course you need to
make sure reboot does not clear out that memory...

It might help to know that SunOS uses low memory for the message
buffer.  (I peeked when deciding where to put it on the sun3.)
Also, note that boot programs and such occasionally trash a few
bytes at address zero, and my work-around for that was to just
skip some space at address zero, i.e.:

#define MSGBUFOFF	0x200
#define MSGBUFSIZE	(NBPG - MSGBUFOFF)

So it uses all but the first 256 bytes of page zero.
I suspect this would work on the sparc also.

Gordon