Subject: Getting a Multia to like NetBSD 1.4
To: None <>
From: Todd Whitesel <>
List: port-alpha
Date: 05/25/1999 20:24:07
I have a Multia here with a 3.2 gig scsi drive on the internal 53c810 card,
internal floppy, and 32 MB, but otherwise straight from CPU Micromart.

Back in January I managed to get the october snapshot to install, and recently
I brought it up to 1.4.

After checking the FAQ et al., I still have some unresolved mysteries.

    1. I can create the two install floppies just fine, but when I actually
	boot from them I get about seven soft error diagnostic banners on
	disk1of2, and another six or so on disk2of2. This does not prevent
	it from starting the kernel, turning the screen into an imitation Sun
	console, and running sysinst, but the errors are somewhat disturbing.

    2. If I use 'dd if=/dev/fd0a of=/tmp/disk1of2' on the first install
	floppy I get all kinds of diagnostics like these:

fd0a: soft error (corrected) reading fsbn 0 of 0-3
fd0a: soft error (corrected) reading fsbn 20 of 20-23
fd0a: soft error (corrected) reading fsbn 24 of 24-27
fd0a: soft error (corrected) reading fsbn 40 of 40-43

	Writing back to the floppy gives even more, some of which say
	'writing' instead of 'reading'.

    3. 'portmap -l' spewed a lot because of broadcasts from certain PCs on
	the net at work where the multia is currently connected. I got rid
	of the -l in rc.conf and that made portmap shut up.

    4. Sometimes when the multia is idle, it will panic with a vm fault of
	some kind, but the frequency of this does not appear to be heat
	related (yes I have it upright in the plastic stand with plenty of
	open space around it). A sample:

fatal kernel trap:

    trap entry = 0x2 (memory management fault)
    a0         = 0x200
    a1         = 0x1
    a2         = 0x0
    pc         = 0xfffffc0000423088
    ra         = 0xfffffc0000422c90
    curproc    = 0x0

panic: trap

fatal kernel trap:

    trap entry = 0x2 (memory management fault)
    a0         = 0x80
    a1         = 0x1
    a2         = 0x0
    pc         = 0xfffffc0000370bb0
    ra         = 0xfffffc000036facc
    curproc    = 0xfffffc00008c6988
        pid = 216, comm = portmap

panic: trap

    5. Other times (also while the system is idle) ncr0 will report a script
	error, and the next program run after that will hang solid, but the
	console terminal is fine -- signals sent from the terminal have
	no effect and just print as ^C, ^/, etc. Unfortunately this output
	never makes it into /var/log/messages.

    6. I have some plain-VGA monitors lying around, and found that getting the
	multia to use them is problematic. In particular, the documentation
	for jumpers W3-W6 to set the SRM video mode is wrong. I found no
	correspondence whatsoever between what the multia service manual says
	and what my multia actually does. By trial and error I found a jumper
	setting that gives me ancient VGA -- but then I discovered that the
	imitation Sun console is only about 53x20 characters in that mode, so
	sysinst blows chunks. Is there any way to force VGA text mode on a

    7. Just now (while I was typing all this) it spontaneously halted to the
	SRM console when I wasn't looking. Attempting to boot it delivered
	a stream of "processor correctable error through vector 00000063"
	messages (which I can pause with scroll lock, how nice) and it appears
	to be very unhappy now so I turned it off. This is beginning to look
	more like the infamous multia heat death, but as I have no prior
	experience with the multia I figured I should appeal to port-alpha for

Todd Whitesel
toddpw @