Subject: Re: my se/30 just died - help!
To: John Ostrowick <jon@macenroe.cs.wits.ac.za>
From: Hauke Fath <hauke@Espresso.Rhein-Neckar.DE>
List: port-mac68k
Date: 01/17/1997 19:24:55
At 12:25 Uhr +0100 17.01.1997, John Ostrowick wrote:
>hi. this is a pretty complex story, so i'll keep it short.

[...]

>i installed from scratch on my new 1gb (conner) and can boot up, however,
>fsck freaks out and says it can't do this disk and i must do it manually.
>so i get a # prompt and type fsck /dev/sd1a and it says, forget it, there
>are some bummed sectors here. i believe it names the last few sectors of
>the disk (huge numbers in the region of 180 000). fsck then either core
>dumps or, more usually, just gets an exception and exits.

What machine? Does your kernel have Allen's ncrscsi or Scott's sbc SCSI
driver? The former is a nice driver from a nice guy, but it is very picky
about hard disks. ;) It has told me nonsense about damaged media on a
Quantum LPS105 before.

>when i run mkfs, it sees all three and i try format the untitled partition,
>and it does it and says 'i don't remmer any errors' - BUT before it starts
>it *does* say, "warning, 690 unused sectors". I presume that is what is
>causing fsck to freak out. So how do i force it to fully format that disk?
>it seems to format most of it and just leave out the last few sectors. the
>result is then that fsck refuses to give a thumbs up to the disk..

DON'T PANIC.

The disk has a varying number of sectors per cylinder. A lot of hitech OSes
like CP/M, MSDOS or Un*xes can't cope with such a revolutionary scheme, so
it tries to map its sectors to a fictitious set of fixed numbers of sectors
per track per head, which leaves a reminder.

Nothing to worry about. Really.

>to summarise:
>
>- my 250mb won't boot and sticks in some crap about nubus JUST AFTER
>'preserving xyz bytes'. I presume a reinstall will fix this but i couldn't
>be bothered since it's the other disk i want to use as my boot disk from
>now on.
>
>- my 1gb will boot but soon dies because fsck/the kernel itself doesn't
>like the disk - reports bad/unused sectors and names about 5 sectors before
>exiting with an exception. As i say, i *think* this is because the mkfs
>program didn't fully format it. I really do not want to format from unix
>CLI because that means reinstalling *again* (done it twice). A weird point
>too: despite having a separate 'untitled' partition, netbsd mounts
>'untitled' and 'root&usr' as if they were the same partition - so i cannot
>run formatting tools from root&usr to hit 'untitled' because as far as
>netbsd is concerned, there is no separate partition. weird weird weird!

1) I am not sure the netbsd kernel can make much sense of an 'untitled'
partition. Make that a 'A/UX Usr Slice 2' type partition; no problem with
having more than one on a disk.

2) If you allocate more (e.g. MacOS) partitions on the disk, make sure the
netbsd partitions come first; the kernel cannot see more than eight of
them. If you fiddle with a netbsd partition afterwards, it is appended to
the END of the partition table and may disappear from the kernel's sight.

3) Give the netbsd newfs(8) a try (e.g. from a miniroot); it is a hell of a
lot faster than Mkfs and I personally trust it more. And don't forget '-O'
if you want to access the partition from the MacOS Installer.

	hauke


	hauke

---
"It's never straight up and down"     (DEVO)