Subject: Re: Kernel doesn't probe memory size correctly...
To: Bill Harris , Joel Reicher <joel@panacea.null.org>
From: Michael L. Hitch <mhitch@lightning.oscs.montana.edu>
List: port-pmax
Date: 05/06/1998 08:26:16
On May  6,  9:03am, Bill Harris wrote:
> On my 240,  I just bought 12  8meg cards.  I currently have two 32's in the
> machine.  Can I mix and match?  I'd like to have at least 1 32 meg card and
> the other 12 8meg cards if the system will recognize it.

  That will depend upon how the 5000/240 configures the memory space.  A
sophisticated and smart design (and probably more expensive) might be able
to map each memory slot separately to make the memory appear contiguous.
I would rather doubt the DECstations would be able to do this.

  My guess would be that what ever memory module is in the first slot
would set the memory size for the the rest of the slots.  If this is the
case, then a 32M card in the first slot would result in each of the slots
being configured on a 32M boundary.  If you put the 32M cards in the first
two slots, and the 8M cards in the following slots, the first 8M board
would then be contiguous with the two 32M boards (giving 72M), but the
following 8M boards would have non-existant space between them.  If the
8M boards appear multiple times within the 32M space, you will lose the
way the current memory probe is done (since it will appear that each 8M
board fills the 32M space for that slot).  If the memory design is such
that the 8M board only maps on the first 8M of the 32M slot space, then
NetBSD will work, but only use one of the 8M boards.

  The memory probde could be made smarter by checking for multiple
non-contiguous memory (since the -current kernel should now be able
to handle this), and by checking for memory modules that appear
multiple times in the address space.

Michael

-- 
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA