The DECpc XL server series

I inherited two of these older servers, which were top-of-the-line when originally purchased. By today's standards they are underpowered in the CPU and RAM department, but still have impressive disk IO with a fairly decent onboard SCSI controller. They have the largest physical capacity for mounting and powering hard drives of any system I've had the pleasure of owning so far, so they are great picks if all you want is a simple file server. Mine came loaded with barracuda drives, so much the better. Here I keep notes on the issues I had to deal with getting a recent (08/2004 testing) installation of Debian Linux working on these systems.

These systems have three PCI slots and a bunch of EISA slots. Having very few EISA parts worth using I stuck to the PCI -- three slots being plenty when you don't have to use one for SCSI. As EISA systems they predate PnP, and so all the hardware must be configured. Especially if you alter the EISA configuration by adding or more likely removing cards, you'll be wanting to run an EISA configuration utility to allow the machines to boot without a keyboard acknowlegement. Hewlette Packard now owns DEC, and they have done a fairly good job at preserving support files online for DEC products. However their indexing system misses one of the critical files, though it is still available by typing in the URL.

The two disks you need are an update the BIOS of the DECpc XL servers and a "SCU" disk, which includes the "ECU". Don't get confused by this, because in earlier rleases the "SCU" and "ECU" were on separate disks for a total of three. The following links are the only disks that are needed to configure the system: BIOS 2.0.8 disk SCU 1.12 disk (includes ECU).

There are some scary warnings on the HP site that state some of the disks there only work on "DECpc XL" and not on "ALL Prioris XL Servers". As far as I can tell, at least as far as the 590 is concerned the two are the same animal. One of mine says DECpc on the case, the other Prioris, and the above disks work fine on both of them. The BIOS should be flashed first, then the SCU/ECU disk should be used.

These files are self-extracting programs that will want to write a floppy disk out, so you'll need a system with a floppy and either MSDOS (6.22 is what is on the disks and probably your best bet) or dosemu with freedos. A note about using native DOS -- you'll need an installation that is on a hard drive because the utility unconditionally uses drive a: and gets unhappy if that is the same drive you booted DOS from. A note on using freedos/dosemu -- for some reason the utility seemed to fail on my installation unless I ussued a "dir a:" before running the extraction program. Doing so seems to cache some of the floppy contents into RAM and works around this problem.

The SCU is pretty self-explanatory and will automatically delete missing cards. You'll also want to assign IRQs and PCI bus master status to any used PCI slots.

Being EISA systems these boards lack the advanced BIOS calls that Linux needs to properly determine the memory size. Linux will only detect 64M of RAM. I was using kernel 2.6, so I simply added the following line to my LILO setup to get my full 128M recognized:

append="memmap=exactmap memmap=640K@0 memmap=127M@1M"

Do note that the 2.4 kernel parameters may be different ("mem" versus "memmap") and that simply using "mem=128M" will not work past a certain kernel revision, because past that point the simple "mem=" parameter can only be used to reduce the total memory. I have not been able to ascertain for sure if the system does actually implement caching for this much RAM, but have observed no dramatic slowdown so I assume that the motherboard is not one of the broken variety which don't work well with more than 64MB.

I only have one CPU for each of these systems so I have not played with SMP support.

This document is Copyright (C) 2004 Brian S. Julin and may be
reproduced and/or used under the terms of the GFDL version 1.2
with no invariant sections.