hd install went haywire

[Joseph Schiller]

Hi,

I've been trying to install Knoppix 3.2 5/3/03 release on the hard drive to no avail. The 'knx-hdinstall' script fails while cfdisk is trying to repartition the hard drive. I've started with 6 GB drive divided into 3 partitions, primary DOS 2GB each. I've tried to create 500 MB swap and 3.5 GB root, however, mkfs generated an error "attempt to read block from filesystem resulted in short read while creating root directory." Now I can't boot from CD, I'm getting pdcraid error reading superblock and htpraid error reading superblock. At one point even BIOS could not detect the hard drive. After booting back into Windows95, I tried fdisk to delete the non-DOS partition. Booted again Knoppix from CD, but now launching the script, cfdisk cannot read the hard drive at all, I get 0 MB and not able to detect the swap partition, the script fails. Now, I realized that I should have created 256 MB swap and 2 ext2 partitions for the install. What should I use to repartition the drive? Any suggestions? I don't have a copy of Partition Magic.

Regards,

Joseph Schiller

[fingers99]

You could try Ranish Partition Manager. It's not as pretty pretty as PM, but reputed to be better in many respects. A Google should find it. It's free (as in in beer) and while I don't think it's GPL is at least written by someone who respects Linux.

But I'd go for swap (Linux swap) and ext3 for the root partition.

[Stephen]

If you get the lastest version of Knoppix (2003-05-20) you could use Qtparted which is included and is similar to Partition Magic.

[Joseph Schiller]

Hi,

I've tried Knoppix 3.2 version 5/21/03. Thank you. The QTParted program has been enhanced with new features. It does resize partition really great, however, it didn't work on bare drive. I've wiped it clean, dumped Windoze, and experimented with reiserFS and ext3. I've discovered running MicroScope that the hard drive has an error in the "system sector." Whatever that may be. The drive is temperamental, it sounds really bad, like an old LP record when the needle finished playing all the tracks, some a rocker arm is hitting an aperture while its trying to do some kind of write operation beyond where the sector ends. I thought this might be a virus. After low level format, I used fdisk to create win-32 partition, because when I launched QTparted, it just crashed not sure what I wanted it to do, I guess. I also bought a new 40 gig drive. Then using CFSK I changed the single partition to Linux swap 500 MB and remainder /home using reiser. I could not boot, I got 'kernel panic' message. I'm planning to try out also copying iso image onto 1.2 gig drive and using it as if it were a CD-ROM. Will let you know how it turns out. Unfortunately, "ars longa vita brevis" and there is so many options yet to try.

Regards,

Joseph Schiller

[Stephen]

Hi,

I've tried Knoppix 3.2 version 5/21/03. Then using CFSK I changed the single partition to Linux swap 500 MB and remainder /home using reiser. I could not boot, I got 'kernel panic' message.

Did you make the file systems on the drive. eg sudo mkswap /dev/hd?? for the /swap or sudo mkreiserfs /dev/hd?? for the reiser filesystem for the /home before you re-booted?

[Joseph Schiller]

Hi,

I've tried low level format, but it crashed at some point using MicroScope utilities. I then tried fdisk to set primary and format c: booting from a recovery diskette. I got following dubious message:

Formatting 6,149.85M
Trying to recover allocationunit 1,564,562
Not ready
Format terminated

So I ran extensive test of the hard drive. MicroScope gives interesting parameters:

Win9x-32 partition type
status bootable
starting head 1
starting sector 1
starting cyl. 0
ending head 254
ending sector 63
ending cylinder 783
start abs sec 63
number of sectors 12594897
boot signature 55AA

Also, there are two types of params BIOS and LBA which are off by 1 (BIOS counts 785 cylinders, LBA counts 784 cylinders). I duno whether 783 cylinders is better than 784 or 785? Running fixed disk LBA read/write test I found

drive 0 cylinder 780
head 181 sector 31
pass 0 error code 10
"uncorrectable crc/ecc errror"

and the error log was full of msgs of this type on cylinder 780. I gather that the drive is defective from cylinder 780. There are options available to map the drive to skip the bad section, but I think it's better to abandon the drive at this point than pull out hairs when it goes belly up later.

At this point, I went ahead with hd install to test the behavior of Knoppix on a defective drive with mixed results.

Booted from CD-ROM an from root shell I did
mkswap /dev/hda1 and swapon -a to activate. It did use the whole drive at this point. Using cfsk I've created two partitions /dev/hda1 as swap and remainder /dev/hda2 as reiserfs. I think it would be advisable to place a limit on the /root partition as a precaution, and either add another extended /dev/hda4 or add /var for print-spooler and email-spooler. Launching the script knx-hdinstall asked me again to do this part. Then approx 40% into copying files, I've got an i/o error from which the script recovered. Upon completion, I've selected to use LILO in the MBR, but this did not seem to work. I think it's due to user error, the partitions are not kosher. Here's the output from /etc/fstab:

# filesystem mountpoint type options dump pass
/dev/hda2 / reiserfs defaults 0 1
/dev/hda1 none swap sw 0 0
proc /proc proc defaults 0 0
/dev/fd0 /floppy vfat defaults,user,noauto,showexec,umask=022 0 0
/dev/cdrom /cdrom iso9660 defaults,ro,user,noexec,noauto 0 0
# partitions found by Knoppix
#/dev/hda1 /mnt/hda1 auto noauto,users,exec 0 0

That extra #/dev/hda1 is confusing.

Also, here's the /etc/lilo.conf file:

vga=791
lba32
root=/dev/hda2
install=/boot/boot-menu.b
map=/boot/map
delay=20
append="hda=scsi hdb=scsi hdc=scsi hdd=scsi hde=scsi hdf=scsi hdg=scsi
apm=power-off nomce"
default=Linux

image=/boot/vmlinuz-2.4.20=xfs
label=Linux
initrd=/boot/initrd.gz
read-only

image=/vmlinuz.old
label=LinuxOLD
read-only
optional

I've created a boot disk during hd install and can only boot from the floppy. The MBR needs to be ammended. All sugestions will be sincerely appreciated.

Reagards,

Joseph Schiller

[Stephen]

That is just the partitions found by knoppix during the install the # in front of the #/dev/hda1 means it is commented out and will not be used. Once you have booted with the floppy and are up and running then open a console window and su and enter the root password then in the root console /sbin/lilo to write lilo to the master boot record.

How does it work with only the 40% copied to the drive have you checked to see what parts did not get copied?

BTW gotta admire your persistence getting it going.

[Joseph Schiller]

Hi,

Thanks. BTW, I couldn't have done it without you.

I could'nt tell what was missing in the install, at 40% in the script I was getting input/output errors, but the script recovered and completed. I like the auto recovery feature. So if something was corrupted, the install did not live long enough for me to discover what that was.

I reinstalled again, this time I've set up hda1 as 500 MB swap, hda2 4500 MB reiser, and hda3 1500 MB reiser, but running CFDISK, I noticed that hda2 is not reiserFS at all, it went in as ext2. This time I did not have to edit of configure LILO, it did write itself to MBR during script install. But I am dumbfounded that hda2 is not reiserFS. I opted for that specifically because the hd is defective. I'm not sure how that happened, but I used console and the commands that I learned "mkreiserfs /hda2" and "mkreiserfs /hda3", but I must have goofed running cfdisk when seperating out 1500 MB partition. At the moment everything works great, save a minor glitch with /etc/resolv.conf which I don't understand yet. I've never had to edit it manually before, and I've discovered it had a sym link to /etc/dhcpc/resolv.conf, and I followed exactly the same steps.

Regards,

Joseph Schiller

[Stephen]

Hi,

Thanks. BTW, I couldn't have done it without you.

I could'nt tell what was missing in the install, at 40% in the script I was getting input/output errors, but the script recovered and completed. I like the auto recovery feature. So if something was corrupted, the install did not live long enough for me to discover what that was.

I reinstalled again, this time I've set up hda1 as 500 MB swap, hda2 4500 MB reiser, and hda3 1500 MB reiser, but running CFDISK, I noticed that hda2 is not reiserFS at all, it went in as ext2. This time I did not have to edit of configure LILO, it did write itself to MBR during script install. But I am dumbfounded that hda2 is not reiserFS. I opted for that specifically because the hd is defective. I'm not sure how that happened, but I used console and the commands that I learned "mkreiserfs /hda2" and "mkreiserfs /hda3", but I must have goofed running cfdisk when seperating out 1500 MB partition. At the moment everything works great, save a minor glitch with /etc/resolv.conf which I don't understand yet. I've never had to edit it manually before, and I've discovered it had a sym link to /etc/dhcpc/resolv.conf, and I followed exactly the same steps.

Regards,

Joseph Schiller

cfdisk set the partition as a linux partition which IIRC is set by default to ext2 you just had to use the type option to set it as reiser. Any time I have installed I usually use cfdisk and format the partitions before I run the script and just quit from cfdisk when it comes up and continue with the script.

I think your resolv.conf problems have something to do with the groups your user belongs too I have responded to your other post although it could be one of those annoying things that get changed when you install and you have to configure the settings manually.