On Tue, 11 Jan 2000, Chris Tribo wrote:

>  I seem to remember rreading from DEC documentation that the only
> situation in which you were (supposed to) be able to have mixed memory was
> when you had 14 32MB modules and then you could put an 8MB module in the
> very last slot. 

  I suspect that might have been because Ultrix was not able to deal with
non-contiguous memory.  NetBSD can do that easily - but the pmax doesn't
currently do that (yet).

>  My 5000/240 came with 448 MB in 14 32MB modules and the 1MB Prestoserve
> module in the last slot. I don't remember reading any scenarios about
> mixing 32MB, 8MB, and the Prestoserve card; which leads me to believe that
> it wasn't possible (at least under Ultrix). Is this a hardware limitation
> or a software limitation? Good info for the DEC HW FAQ :-)

  I'm fairly sure it's a software limitation.  I've got a Prestoserve in
one of my machines.  My memory sizing modifications looks for the 1MB
memory segment and excludes it from the available memory.  It's been a
while, but I think I added a 32MB memory module the slot before the
Prestoserve in addition to the 8MB boards in the other slots, and it
worked fine.

>  As a side note, is there a way to test memory from NetBSD so that I could
> target specific areas with a psuedo-random and/or sequential read/write
> test? On one of my 5000/240 motherboards, the PROM RAM test checked out
> all 360 some odd MB of RAM, I would boot and go into X, start a kernel
> make and the system would panic with ECC errors. I did this a couple dozen
> times and each time it was in a different address. I pulled the MB and
> switched to another one which seems to be happily working. Is there such a
> testing utilitiy available? Even if I couldn't test the first chip since
> the OS is residing there, a Read/Write/Verify memory testing program would
> be very nice to have.

  I don't know of any supported way to do this.  I suppose one could hack
something into the kernel to allow a user program to allocate explict
physical memory.  I've modified mem.c a bit to let me access uncached
memory space via a '/dev/mem'-type device.  I was using that to access
the Prestoserve memory so that I could read/write that memory.

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