Which part of a Linux would object the removal of an optional hard disk?

[saikee]

I have strange behaviour of Suse 9.1 Prof. It doesn't like me removing the slave drive from the PC while nearly all the other distros don't seem to mind.

The slave was added later for additional installation of the Linux.

For some reason whenever I remove the slave drive Suse would refuse to boot. I have checked several time in Suse's

/boot/grub/device.map
/boot/grub/menu.lst
/etc/fstab
/etc/mtab

to ensure the slave drive /dev/hdc has no entry in any of the above.

Suse was installed before the slave was added. The funny thing is that I can insert any slave drive just to satisfy Suse to boot.

Is there anywhere I could check to see why Suse depends on the presence of a slave drive?

[kevinalm]

Most likely by removing the drive you effectively alter the grub hd(n) assisnments such that grub stage 1 can't find something, ie the /boot/grub dir. Plugging in a drive restores the ordering. My guess anyway. If you post specifics we can probably figure out exactly why it fails. Your partitioning scheme, hd positions, the location of /boot/grub and so on.

[saikee]

I shall do another check tonight but it is unlikely related to Grub because

(a) The Suse was installed with the slave disk presence
(b) There are 16 distros installed and many of them also have Grub and they don't have a problem with the slave drive being removed.
(c) Unlike Lilo Grub doesn't check the validity of the partitions it has to boot. I have a Grub menu with partitions assigned for booting but they were empty and unformatted. Only after I install the distros then these partitions become bootable. Grub is quite happy to have them around.

[Sepero]

Originally posted by saikee
I have strange behaviour of Suse 9.1 Prof. It doesn't like me removing the slave drive...

For some reason whenever I remove the slave drive Suse would refuse to boot.Do you get an error message? At what point during boot does it stop?

[saikee]

The graphic screen of Suse just stops with the blue bar progess to a staic position. Pressing F2 as suggested by Suse will see the screen scrolling mad non-stop. Never can read the screen. Only manage to press ctrl+alt+del to reboot.

[Sepero]

Sorry to sound dumb, but have you tried the "scroll lock" key?

[saikee]

As dumb as you expect, never cross my mind to do that! Will investigate tonight.

[kevinalm]

What I had in mind is illustrated as follows:
You have four hds, hda,b,c d. Then
hd(0) =hda
hd(1) =hdb
hd(2) =hdc
hd(3) =hdd

But if you remove hdc then

hd(0) =hda
hd(1) =hdb
hd(2) =hdd

If your active menu.lst and company are on hdd then grub will no longer boot. This may not be your problem, but it would be what I would first suspect.

Regards,
Kevin

[saikee]

kevinalm,

Good try but I don't think that is the case as I am not using the last two disks as they are my personal data with which I only mount if access is needed.

Grub also doesn't check the disk order until the partition is booted but as my Suse was installed before the slave drive was introduced hdc was not involved at all in its set up initially anyway.

Here is what I have found so far.

(a) After the slave has been removed I could boot Fedora C2, Kanotix and Mandrake 9.2, all of which use Grub. That proves Grub is not to blame.

(b) Suse failed as usual. I couldn't stop the screen scrolling either as the scroll key didn't respond.

Here is the device.map
(hd0) /dev/hde
(fd0) /dev/fd0

Truncated Grub menu (boots 16 Linux), Suse in hde6 known to other distros as hda6
# /boot/grub/menu.lst of Suse 9.1 professional

color white/blue black/light-gray
default 1
timeout 100
gfxmenu (hd0,5)/boot/message

title Suse 9.1
kernel (hd0,5)/boot/vmlinuz root=/dev/hde6 vga=0x31a splash=silent desktop resume=/dev/hde5 showopts
initrd (hd0,5)/boot/initrd

title 1st Windows XP in Maxtor IDE hd0
hide (hd2,0)
hide (hd1,0)
unhide (hd0,0)
root (hd0,0)
chainloader +1

title Fedora Core 2
root (hd0,1)
chainloader +1

title Debian Sarge
root (hd0,2)
chainloader +1

title Floppy
root (fd0)
chainloader +1

title Failsafe
kernel (hd0,5)/boot/vmlinuz root=/dev/hde6 showopts ide=nodma apm=off acpi=off vga=normal noresume nosmp noapic maxcpus=0 3
initrd (hd0,5)/boot/initrd

Here is the fstab

/dev/hde6 / ext3 acl,user_xattr 1 1
/dev/hde5 swap swap pri=42 0 0
devpts /dev/pts devpts mode=0620,gid=5 0 0
proc /proc proc defaults 0 0
usbfs /proc/bus/usb usbfs noauto 0 0
sysfs /sys sysfs noauto 0 0
/dev/cdrecorder /media/cdrecorder subfs fs=cdfss,ro,procuid,nosuid,nodev,exec,iocharset=ut f8 0 0
/dev/fd0 /media/floppy subfs fs=floppyfss,procuid,nodev,nosuid,sync 0 0

and the mtab

/dev/hde6 / ext3 rw,acl,user_xattr 0 0
/dev/hde6 / ext3 rw,acl,user_xattr 0 0

[retsaw]

It's a bit weird that SUSE refers to your hda as hde. Does it still refer to your hdc as hdc? I do think that the fact it refers to hda as hde may mess up the drive order when you remove your slave drive as far as SUSE is concerned.

Also have you tried the failsafe option to boot SUSE? Another thought I had was it may be having a problem looking for the resume partition, so the "noresume" kernel option that the failsafe menu entry has would test that, but I'm not too sure about this possiblilty as the partition is specified as being on the same partition as SUSE.

[saikee]

retsaw,

Yes Suse call my hda as hde and the hdc as hdf. I have reported previously in the forum too.

While most the other distros call my Sata as hde Suse calls it hda!

Luckily Mandrake 10 does exactly the same so it may not be Suse's fault. I have since concentrate my effort on another box with older components. All the Linux, including Suse, call the first IDE hda, reserves the hdb for the CD writer and use hdc for the second IDE disk which is alone on it own cable. Suse in that box call the Sata as hde while some distros use sda instead.

Basically I don't have a choice but just follow whatever the name chosen by the distros.

On the failsafe booting choice of Suse, I have tried it. The display is in command line mode but it stopped after

checking files system
fsck 1.34 (25-Jul-2003), with the hard drive LED permanently on, indicating busy activities. I think it is gone because the condition has remain static for the last 20 minutes.

Like I said it is a strange behaviour

[kevinalm]

Originally posted by retsaw
It's a bit weird that SUSE refers to your hda as hde. Does it still refer to your hdc as hdc? I do think that the fact it refers to hda as hde may mess up the drive order when you remove your slave drive as far as SUSE is concerned.


I agree. To me this suggests that it is a drive ordering problem. I think an investigation as to why suse sees the drives differently than the other distros would be worthwhile. Remarkable what you can do with grub by the way. 16 linux partitions and 1 xp. Quite an accomplishment. Congrats. :D

[saikee]

kevinalm,

Just another twist in the tale.

The Suse 9.1 was the first distro I installed and it named the first IDE disk as hde while all the others use hda except Mandraake 10. I have also Mandrake 9.2 installed and it names the IDE disk hda. It is possible that my hard disks set up, with a mixture of IDE and Sata, is a bit weird as I use caddies and mobile racks for all my hard disks. Thus shifting around may have upset the ordering sequence initially. The disks are all on cable select and so I can swap the master and slave in seconds. Now I know Linux doesn't like it and am much more careful.

I actually took a backup disk of the 16 distros + XP and forced fed into a second machine and managed to fired up all the 16 Linux by amending just the display drivers. The Windows XP was the only one totally dead and I have to replace it with a different license.

To make it more interesting, after obtaining some idea from this forum, I even managed to amend the hde reference of Suse back to hda in the second machine which I now use most. The same hde-to-hda amendments fail to work with the first machine. The amendments were carried out to the files I listed earlier.

I come to the conclusion that may be the system wasn't entirely "clean" when I first installed Suse. It could be something like creating a partition and putting data inside immediately without rebooting the computer first (using only as a suggestive example as I don't really know).

I have been expanding the 16 distros, first to 28 and then now there are 32 in the box booting them the Suse' Grub. I did it by reducing the partition size which has been standardised to 5Gb/Linux. I have also learned to use Grub to boot 2 additional versions of Windows and a DOS.

In the 5 months into the operating system I have been continuously amazed by the sheer power and flexibility of Linux and at the same time realising how bad Windows has been.

[Sepero]

On the failsafe booting choice of Suse, I have tried it. The display is in command line mode but it stopped after

checking files system
fsck 1.34 (25-Jul-2003), with the hard drive LED permanently on, indicating busy activities. I think it is gone because the condition has remain static for the last 20 minutes.Originally posted by kevinalm
I agree. To me this suggests that it is a drive ordering problem.With only this information to go on, I would have to agree with kevinalm. Although I'm still not convinced it's a problem with grub. It sounds more like a problem with your initial root(initrd) trying to guess about something that's not there.

I'm sorry I can't help more, but without the original error message, it's like trying to debug proprietary software.

[retsaw]

I know that whether a SATA drive is refered to as sdx or hdx is dependant on the kernel configuration as I was changing some settings in my Slackware 2.6.7 kernel to enable DMA on my DVD drive and my SATA drive switched from being sda to hde. Weirdly though when I copied my Slackware kernel config and did a make oldconfig to build my 2.6.8 gentoo kernel it switched back to being sda.

It really would be helpful if you could manage to read the error message when booting SUSE. I think a problem with GRUB can be ruled out as it starts booting the kernel and it doesn't give problems with other distros also your fstab looks fine. To get to the bottom of this you may have to examine SUSE's boot scripts perhaps even starting from the initrd (though I doubt there would be anything in the initrd causing the problem).