On Saturday, June 21, 2003, at 00:22 Europe/Helsinki, Manuel Bouyer 
wrote:
> On Fri, Jun 20, 2003 at 10:27:05AM +0300, fulcrum@mbnet.fi wrote:
>> I've got this old Compaq Presario 7220 from a friend of mine, and I 
>> was
>> supposed to install NetBSD on it. The problem is that it won't 
>> install.
>> The installation stops in the following message, when newfs is about 
>> to
>> create the partitions;
>>
>> === clip ===
>> wd0: no disk, Failed: command failed
>> Command: /sbin/newfs /dev/rwd0a
>>
>> newfs: /dev/rwd0a: open for read: Device not configured
>> === clap ===
>
> Hum, this could be that disklabel failed to create the partitions. 
> What does
> fdisk and disklabel says ?

I tried the installation process (without sysinst this time) following 
the instructions in documentation 
(http://www.mclink.it/personal/MG2508/nbsdeng/ap-inst.html).

#dd if=/dev/zero of=/dev/rwd0 bs=1m count=1
1+0 records in
1+0 records out
1048576 bytes transferred in 0.233 secs (4500326 bytes/sec)
# disklabel -I -i wd0
partition>
[... editing of the partitions ...]
partition> W
Label disk [n] y
Label written
partition> Q
# newfs -m 5 -b 4096 -f 512 /dev/rwd0a
wd0: no disklabel
newfs: /dev/rwd0a: open for read: Device not configured
#

>> I have been trying different methods, but none of them seem to work.
>> E.g. I installed and tested NetBSD on another computer successfully,
>> then put the working harddrive in the Compaq with no success (kernel
>> was loaded, but the boot process stopped saying it couldn't read
>> root-filesystem).
>
> This is, however, strange. What is the exact error message ?

It was silly of me that I didn't write it down, and I currently dont 
have the opportunity to test it. If I recall correctly it was "root 
filesystem unreadable". There were no more info.

>> My theory is that Compaq protects some part of the harddrive (MBR?)
>> where it stores its information (this computer is one of those with no
>> real BIOS). Windows/DOS/Linux are however installed correctly.
>
> What's your disk layout ? It's possible that the BIOS reserve some 
> space
> (probably at end of disk) for deep sleep.

The machine has an old 1Gb hardrive completely dedicated to NetBSD. I 
actually tried to leave a small empty DOS partition to the 
beginning/end/both of the disk, but it didn't work out. It generates a 
message like this:

=== message ===
Command: /sbin/newfs /dev/rwd0a

/dev/rwd0a: [...]
super-block backups (for fsck -b #) at:
	32,pciide0:0:0: lost interrupt
	type: ata tc_bcount: 4096 tc_skip: 4096
wd0a: device timeout writing fsbn 89208 of 89200-89215 (wd0 bn 192024; 
cn19 tn 8 sn9), retrying
wd0: soft errror (corrected)
	88800, 177568, 26636
uid 0 connn newfs on /: filesystem full
/: write failed, filesystem is full
=== message ===

=== a note found from netbsd.org 
(http://www.netbsd.org/Ports/i386/faq.html#determining_geometry) ===
If a disk has a valid DOS MBR partition, (created by DOS, Windows,
pfdisk, or similar) it will contain the "correct" BIOS geometry. NetBSD
will use this as the BIOS geometry during sysinst. Unfortunately on an
empty disk, or one without the MBR set up properly, NetBSD will get the
BIOS geometry wrong. Ideally this information should be passed in from
the boot program, but there is no certain method of determining the
mapping of bios disk to netbsd disk, though it should be possible to
improve on the current 0% hitrate.
=== a note found from netbsd.org ===

Anyways the number of heads, cylinders and sectors found by NetBSD 
matches the true numbers.