udev and number of devices 

Newsgroups: gmane.linux.distributions.grml.user
Date: 2006-12-04

http://thread.gmane.org/gmane.linux.distributions.grml.user/844/focus=847

> > > My HD used to have 15 partitions, now I created another one, the 16th, but
> > > it is not recognized by udev [...]
[...]
> > Such a limit does not exist for udev:
> > root <at> grml ~ # ls -la /dev/hda*
> Thanks for the respond mika. glad to know that. The difference is mine is
> sda and yours is hda.
> I think the answer I get from gmane.linux.debian.user by Greg Folkert is
> right, my SATA is recognized as SCSI device, which limits to 15 partitions
> by SCSI Standard.

Ah right.

> However, both responds from gmane.linux.debian.user implied that the sda16
> should be recognized as sdb (under both linux & windows). any idea?

It's a limit of the kernel.

A possible solution might be the patch from:

http://lwn.net/Articles/110426/

But we have to review and test that one. It's on the todolist: http://bts.grml.org/grml/issue42 - but it's definitely queued for post 0.9, we had the full freeze for grml 0.9 and grml-small 0.3 right in those minutes. And 2.6.19 will have some further changes due to Libata PATA merge as well….

two distros on one hd? 

Newsgroups:  comp.os.linux.misc
Date:        Sun, 26 Oct 2003 01:55:06 GMT
> > Anyone have pointers (links) to info on installing a second distro and
> > booting between the two?  Right now I have rh9, about 60G free space and
> > wanting to take a look at gentoo.
>
> simple. I've had 4 booting off the same hard drive.
> The way I did it was, install the first, let it configure lilo so it's
> bootable.
>
> Then install the second, third and fourth, selecting the same /home and swap
> partitions. (But tell them NOT to write lilo to the MBR)
>
> Then, boot into the first, mount the other 3, copy their kernels to the
> bootable linux's /boot directory, edit lilo and add those 3 kernels with
> image=vmlinuz-mandrake
>    label=mandrake
>    root=/dev/hda<whatever>
>
> image=vmlinuz-suse
>    label=suse
>    root=/dev/hda<whateverelse>
>
> etc
> run lilo, and you're away.
> the system map files will probably need copying too, but they should have
> names based on their kernel, like System.map-2.4.20-1, so they won't clash
> between distros.

You really don't need the System.map files, unless you're doing kernel debugging, (although some kernel messages to syslogd to actually use some of the symbols picked up from the map files if available).

Likewise, you don't need to copy the kernel images from the other partitions if you mount those other partitions somewhere and create symlinks in /boot pointing to them—during the time you are writing lilo to the MBR (at that time, lilo will write the disk block addresses, so that the symlinks are not needed later).

Or you can use grub, and not bother with the above, since grub can find the kernel images via its understanding of filesystems in which the kernel images are located. In grub, you specify which partition have the other kernels via the 'root' parameter, and the pathname to them via the 'kernel' parameter.

Dave Brown Austin, TX

Inconsistent Partition Table on Different HDD 

Newsgroups: comp.os.linux.setup
Date: Thu, 12 Dec 2002 15:16:03 -0400
> The old HDD (Conner CT204 4.3GB ATA-66) which I originally intended to
> install RedHat 8 happened to crash last week and I replaced it with a
> new Western Digital one (10GB ATA-100). But when I try to install
> RedHat 8 after reinstalling Win98 onto the new HDD, the same problem
> that I used to have with the old HDD resurface again.
> RedHat installation complains that the partition table is inconsistent
> and the message given is not very informative on what is actually
> wrong, hence hard to troubleshoot and fix whatever problem it's
> having.
> It reports:
> The partition table on /dev/hda is inconsistent. There are many
> reasons why this might be the case. Often, the reason is that Linux
> detected the BIOS geometry incorrectly. However, this does not appear
> to be the case here. It is safe to ignore, but ignoring may cause
> (fixable) problems with some boot loaders, and may cause problems with
> FAT file systems. Using LBA is recommended.
> And it reports the same problem for /tmp/hda.
> What I suspect now are either:
> (1) my motherboard (Intel AL440LX) BIOS is mis-reporting or Linux is
> unable to obtain the correct geometry information from the BIOS.
> (2) The FDISK.EXE that comes with Win98 are doing partitioning that
> are not compliant with standards.
> If it is case 1, I believe I would have to try to do a BIOS update.
> For case 2, I may have to give up on Win98, clean out the HDD totally
> and remove all partitions made by FDISK.EXE and use RedHat Linux as
> the sole OS.
> So far, running a Google search on Linux newsgroup has not been
> helpful in understanding the problem and how to resolve it. Hope
> someone can help out in this issue. Thanks in advance.
> To aid in understanding the partitioning on my HDD, I've included the
> information as reported by Disk Druid. Currently, I've a primary
> partition to hold my Win98 (2GB) and an extended partition with 1
> logical partition to hold some user's data (2GB). I was planning to
> use the remaining free space (5509M) to create another primary
> partition and a Linux swap for RedHat 8.

Hi,

Based on the information you gave above, I suggest that you

Partition Magic would be quite straight forward. If you don't have it, use RedHat's Disk Druid, or anything but M$ fdisk. :-) The point is "one single tool". Otherwise, there will be endless trouble.

FYI, I now personally think that the sfdisk, which comes with every distro, is the best tool I ever used. I was first scared off by its man pages, and spent lots of lots time in using other tools (like cfdisk, Partition Magic, Disk Druid), and trying to solve the problem arisen from using them. Only after having so many troubles already, did I sit down and read its man pages thoroughly, and play with it on the new disk that I just had. Had I done that early, I won't be dealing the troubles I used to have…

The point: it is good to have new disk to play with. :-) I played with it while my working Win98 and RH8 are all on it! :-)

Good luck!

T

How to rename partition? 

Newsgroups: comp.os.linux.misc
Date: 2002-10-12 08:00:38 PST
> > I've modified /etc/fstab, renamed /data to /misc (after umounting
> > /data), and rebooted.
>
> Did you also run "e2label" to update the partition's label from "/data" to
> "/misc"?

BTW, the use of labels to identify partitions was not popular until RH 7.x time. I personally feel it to be a danger. It has the disadvantage that there might be other partitions on the system which have same labels, but the system may be ignoring those labels, until you use one in an fstab. In my case, I have several Linux installations on the same disk, and some shared partitions, so managing labels is no easier than keeping track of device names.

First, a little more precise terminology: a partition is a device with a name, /dev/hdb1; a filesystem has a name in reference to its location in a / directory struction, /data. True, it can be labeled, but isn't done so automatically (except, perhaps by the RH et al. installer).

So you could have accomplished the above by: /dev/hdb1 /data ext3 defaults 0 3 and not encountered the problem.

(The only advantage I can see to using labels is that device names can be rearranged if you delete a partition, requiring you to go change /etc/fstab. But that's a situation rarely encountered.)

Dave Brown

Mounting file-systems by label rather than device name 

http://www.debian-administration.org/articles/522#comment_15

by Steve on 1 May 2007

*Tags*: filesystems, grub, blkid obtian list disk by uuid,

When you're dealing with multiple drives, both fixed and removable, it can get hard to remember which is which. Remembering to mount /dev/sda1 in one place and /dev/sdc5 in another. The solution to this problem is to use labels instead of partition names when referring to them, and here we'll show how that can be done.

There are two things you need to do to start using labels:

To add a label you must use the tune2fs command, specifying the partition to modify and the label to add.

Upon my home system I have only a single disk /dev/sda which has three partitions:

   Name        Flags           Size (MB)
-------------------------------------------
   sda1        Linux ext3       20003.89
   sda2        swap              3002.23
   sda3        Linux LVM       297064.22

To add the label "root" to my root partition, /dev/sda1, I'd run:

root@mine:~# tune2fs -Lroot /dev/sda1

Once I've got a label in place it becomes visible in my partition list when I run "cfdisk /dev/sda", so I know it has worked.

Once the label has been put into place there are now two places which I need to update the system in order to use it:

In both cases the change is the same, instead of specifying /dev/sda1 I specify LABEL=root.

My /etc/fstab file now looks like this:

#
# /etc/fstab - static file system information.
#
LABEL=root      /        ext3    defaults,errors=remount-ro 0       1
/dev/sda2       none     swap    sw                         0       0
...
...

The grub configuration file was updated to change:

# kopt=root=/dev/sda1 ro acpi=off noapic

into this:

# kopt=root=LABEL=root ro acpi=off noapic

Once this was done the kernel(s) can be updated by running update-grub. This updated my Xen kernels and my Debian kernels to include the new root - for example:

title           Xen 3.0.3-1-amd64 / Debian GNU/Linux, kernel 2.6.18-4-xen-amd64
root            (hd0,0)
kernel          /boot/xen-3.0.3-1-amd64.gz
module          /boot/vmlinuz-2.6.18-4-xen-amd64 root=LABEL=root ro acpi=off noapic console=tty0
module          /boot/initrd.img-2.6.18-4-xen-amd64
savedefault

The advantage of doing this is that I can now use identical configuration files upon each of my hosts, and rely upon the labels to make sure the correct partition is used as my root filesystem. This is very useful when you're configuring your systems via CFEngine, or similar.

Mounting file-systems by label rather than device name 

By mounting using labels, does it mean that if I were to move the partitions around, for e.g. move my root partition from /dev/hda6 to /dev/hda7 without making changes to GRUB or /etc/fstab, that my Linux will start up without any problems?

— ramsamba

Label disk/partitions are nice, but it's kinda dangerous when you move your disk from one machine to another, specially with common names for partitions such as 'root', 'usr', 'var', etc…

Mounting file-systems by label rather than device name 

> Label disk/partitions are nice, but it's kinda dangerous...

Yes.

In addition to that, it can be a security problem, i.e. people can intentionally label their partitions in a way that would create name clashes with vital partitions.

I don't know what bad could be done with that, but I suppose that, for example, a USB disk labelled "root", "var", etc., could be used to gain access to a system: even if booting from a USB disk is disabled, having the device plugged at boot time may allow the partition to be mounted instead of the expected one (root partition, /var, etc.).

— davux

You could use UUIDs instead (which can be gathered by running /sbin/blkid).

The only problem I've found is with Linux RAID1 devices - you have to specify /dev/md0 or whatever, otherwise the boot process may try to mount one of the mirrors first (/dev/sda1 for example) and then when the raid tries to assemble, it fails, halting the boot process.

by Anonymous (72.66.xx.xx) on Mon 7 May 2007

documented on: 2007.07.29

advantage of disk label in ext3 

Newsgroups: comp.os.linux.misc
Date: Sat, 07 Dec 2002 18:54:02 GMT
> Can anyone explain what the advantage of using disk labels in ext3
> instead of using plain /dev/hd? ?

If you're talking about /etc/fstab entries, one potential advantage that springs to mind is that if the partition numbers change (say, because you repartition or add support for BSD partitions or the like), the /etc/fstab entries with labels will still work. Another advantage is that it might be easier to understand the purpose of partitions when examining /etc/fstab — but that is, IMHO, counteracted by the fact that it obscures the true layout of those partitions.

> The reason I'm asking is that the other day, when I put my two disk on
> one computer, my RH8 get very confused because there were two disks
> labeled /export. It refused to boot, even to single mode, even after I
> have set one of the disk to 'none' in BIOS!

Ouch!

> More over, what is the tools that I can use to handle the disk label?
> none of the disk tools that I know of can do it, fdisk, cfdisk, sfdisk,
> parted...
tune2fs -L {volume-label} /dev/{ID}

device filename for the partition.

Rod Smith

advantage of disk label in ext3 

> More over, what is the tools that I can use to handle the disk label?
$ apropos label
e2label (8)          - Change the label on an ext2 filesystem

Bill Marcum

advantage of disk label in ext3 

$ e2label /dev/hda
e2label: Bad magic number in super-block while trying to open /dev/hda
Couldn't find valid filesystem superblock.
$ e2label /dev/hda2
/

/etc/fstab LABEL=/ stuff 

Newsgroups: comp.os.linux.misc
Date: 2002-11-12 12:46:23 PST
> What exactly is the purpose of the LABEL= phrase in /etc/fstab ?
> Surely mount tells one what is mounted where if one wants to know ?
>
> I think the LABEL= idea was a really bad one.
> It just causes confusion without adding to knowledge.

I don't think it was a bad idea - and after all you don't *have* to use it. If you type "man fstab" then you will find:

Instead  of giving the device explicitly, one may indicate the (ext2 or
XFS) filesystem that is to be mounted by its UUID or volume label (cf.
        e2label(8) or xfs_admin(8)), writing LABEL=<label> or
UUID=<uuid>, e.g., `LABEL=Boot'  or
`UUID=3e6be9de-8139-11d1-9106-a43f08d823a6'.   This
        will make the system more robust: adding or removing a SCSI disk
changes the disk device name but not the filesystem volume label.

I think that says it all. You can view and change labels with e2label, mount obviously enumerates all labels present in all ext2/3 filesystems and mounts the appropriate one.

I think this is quite useful, especially if you are cloning partitions since it won't matter what partition order the destination computer has.

But I admit, it can be confusing if you don't know what it does …

Ingmar

/etc/fstab LABEL=/ stuff 

> Could someone explain to me how this LABEL=/ stuff in newer
> (Redhat) /etc/fstab works?

Well the label shouldn't really have anything to do with your problems … it's just a way of avoiding to hardcode the partition numbers into an fstab file, instead you use "labels" that can be created / viewed with "e2label"

> I had a root dump of a Redhat 8.0 ext3 FS which I restored to a
> partition with an ext2 FS and I'm getting a kernel panic because
> the kernel loads an ext3 module while the FS now is a
> ext2 FS.

I don't think that I understand this correctly, but something on the side. ext3 is really just ext2 + journal. E.g. to convert an ext2 filesystem to ext3 you just run one command ('tune2fs -j /dev/hdaX') to create the journal.

You can also always force mount to a certain filesystem type in fstab

Ingmar

question: deleting/undeleting disk partitions 

Newsgroups: alt.solaris.x86,comp.os.misc,comp.os.linux.misc,comp.unix.solaris

>Can anyone recommend/ point to some source where I can get info on
>reading/editing disk partition tables?

I guess I'm answering my own question here, but what the hell — thanks to all who replied anyway. Andries Brower at http://www.win.tue.nl/~aeb/partitions/partition_tables.html has lots of useful info. Also I found a shareware program DiskEdit from Vinoj Software that lets you edit partition tables in hex. I used it to zero out everything from x1BE through x1FD, in effect deleting the master partition table, then created a solaris overlay (without overwriting existing partitions, and then restored the appropriate entries in the partition table by hand. Easy enough!

question: deleting/undeleting disk partitions 

>partitions.  Rather than backing up and restoring very large
>partitions, I've saw a post (via deja) about a workaround which
>involves effectively hiding surrounding partitions by temporarily
>deleting them.  Later one can undelete them by restoring some entries
>in the partition table.

I've done this using Linux fdisk, but only on systems which had Windows and Linux, so I can't speak to what Solaris might do.

The Linux fdisk program (as of a couple of years ago anyway) was non-destructive, as least as far as Linux goes. (I had a disk that had 20 partitions on it —don't ask— and RedHat and SuSe installers would crash when they saw that many. I had to delete them, do the install (on a partition =< 16), then put them back with Linux fdisk. The filesystems on those "high" partitions were unaffected.

I've also "re-typed" partitions to hide them from Windows and Linux installers. Windows looks for the signature byte on booting and applies letters to the drives. RedHat looks for Linux signatures when you do an update (although now, I believe it lists the ones it finds and asks which is the partition to update). Linux fdisk has a number of signatures you can put on a partition which are ignored by Windows, etc.

Dave Brown Austin, TX

question: deleting/undeleting disk partitions 

>on pc type systems with the dos partition scheme the mbr can only
>define 4 partitions and each extended partition (dos format anyways)
>was restricted to 4 additional.  this would allow up to 16 usable
>partiotions and 4 extended ("real") partitions. this in concept allows
>20 partitions with the first 4 allways used as extended.  it would be
>concievably possible to have extended within extended to get more,
>however I don't know if this is possible.
>
>comment?

Uh-uh. There can only be one extended partition, but said extended partition can contain any number of logical partitions[0]. The first extended partition in the MBR points to a certain sector on the drive, which contains another 446-byte boot record and a 64-byte partition table.

The partition table in this sector contains an entry describing a logical partition, and a pointer to the next 512-byte sector that contains the next logical partition.[1] This is the standard in the x86 world; other architectures have very different ways of doing things, and certain x86 OSes like *BSD can use a "whole disk" way of doing things that is incompatible with a standard PC partition table. The standard is there so that the x86 BIOS can find a valid partition and start a bootloader without straining its little mind.

[0] Sort of. In Linux, an IDE disk with more than 63 partitions will have all partitions after the 63rd unusable. The limit is 15 for SCSI disks. Most people find they can live with these limits. [1] The details may be wrong here; it's late and I can't find my technical reference on this crud.

Matt G

Partition overlapped 

Newsgroups: comp.os.linux.misc,linux.redhat.install,linux.redhat.misc

> I just don't get it how people keep creating logicals like this.

I disagree with you. Here is my partition table for one drive. I have windows and 4 distros of linux going just fine. The order is reversed and I did it with linux fdisk, much like he did with gpart. You will see that hda7 and hda8 are out of order.

Partition overlapped 

> > delete partitions hda5 and hda6 , and recreate them in the right order.
> > print the EXACT CHS values of these partitions on to a piece of paper.  If
> > you recreate them with these numbers (with hda5 and hda6 switched!) you
> > won't even lose data. A typo is disastrous though. Therefor backup what
> > you can before you begin this operation.
>
> yeah, I was wondering how can change the order back. which tool do you
> recommend that can do this trick? I know using M$ tools will definately
> wipe everything...

But linux fdisk variants will not run a mkfs, so use one of those So fdisk/cfdisk/sfdisk : whichever one you like the best, but don't forget to read manpages first if you don't know the programs yet.

Partition overlapped 

> In the first example, the partitions 3 and 4 are overlapped.
> In the second example, you have a native Linux logical in a
> W95 extended- is this possible?
>

It is possible to have any partitiontype inside any extended. (don't put another extended in though) Whether it's usable or not depends on the OS that reads the table. Putting a windows ID inside a linux extended is also possible. Linux won't care, windows wouldn't be able to find it though.

The overlap should not be possible, but poor partitioning programs exist. Normally you wouldn't be able to create such a table though.

documented on: 2001.01.08 Mon 13:36:59

partition table changes with lilo 

http://groups.google.com/groups?selm=39dfa0e0.4940561%40news.inet.tele.dk

Newsgroups: comp.os.linux.setup
Date: 2000-10-03 02:03:59 PST
>I want to boot DOS, Win98, and Linux, and I have a problem with the
>partition table changes necessary to do this.  The only partition types
>defined in lilo according to some text file in the docs directory are
>DOS_12, DOS_16_small, and DOS_16_big.  So what do I do if I want to
>hide/unhide a FAT32 partition?  Ideally I would hide the Win98 partition
>from DOS when DOS boots, and then unhide it when Win98 boots.
>
>In case it's not clear what I'm asking for, it should eventually look like
>this:
>
>other=/dev/hda1                                        #this is DOS
>the usual...
>change
>                    partition=/dev/hda1
>                    activate
>                    set=DOS_16_big_normal
>                    partition=/dev/dha2                #the Win98 partition
>                    deactivate
>                    set=xxxxx_hidden                  #here's the problem
>                                                                 #how do I
>refer to the FAT32 system?
>
>Note that this is just from my head, not from lilo.conf so some syntax is
>probably off, but I think it's good enough to show what I'm looking for.

From the Lilo readme file it seems as it is possible to define your own change rules. Look for "change-rules".

Primary FAT32 ending cylinder 1024 or later: 0x0C, hidden 0x1C

Other FAT32: 0x0B, hidden 0x1B

Svend Olaf

win2k and linux 

http://groups.google.com/groups?selm=b%251h6.129816%24Z2.1670246%40nnrp1.uunet.ca

Newsgroups: comp.os.linux.hardware
Date: 2001-02-09 18:49:54 PST
> i am going to install rh6.1  and win2k .
> Which one should i install first i have 20 gb.
>
> I can mange to install win98 and linux and it works
> perfectly fine.

You'll get this around Saturday morning. If you have time during the week about half a day, preferably the afternoon half ) then proceed. Otherwise wait till the weekend. The switching from one os to another can get a little confusing.

  1. Leave the MBR alone from lilo's point of view , and keep it in the linux boot partition. leave the dos mbr in place.

  2. Make a separate boot partition for each of the microsoft OS'es ( that way they will have less opportunity to muck up other installations ).

  3. Make sure the boot partitions of the OSes are within the 1024 cylinder limit.

  4. When installing 98 or nt4 or win2k , make sure that the appropriate boot partition is marked bootable AND active.

  5. Win98's C: drive must not be in the same partition as win2k's C: drive , or you'll have hell to pay for down the road. Best way is for the first MS reable partition to be NTFS ( win98 can't natively read it ) , and the second MS readable to be fat32 with win98 on it. Win2k will be able to read win98's boot partition, but it will not be "C:" it will be something else . ( F in my case ) .

  6. Have a battle plan :) .

For example :

I have Linix, Win2k and win98 on the same hdd ( 27GB maxtor ) , and they behave ;) .

Ok, win98 may not be one's choice of "stable" os, but gaming under win2k is pretty bad unless one plays one of the mainstream games like quake , unreal etc. and I'd much rather fly an F-15 simulator or something.With that out of the way, here's how I set up my disk : ( I used partition magic 5 to lay out the partitions ) Hint : use the 2 "rescue floppies" .That's the same thing, only, works in dos .

hda1 to hda4 are the primary partitions. Any one of these can be an extended partition , ie, one that can contains sub-partitions I think one can have about 64 partitions on a single ide hard disk ( 1-63 ) . The PC architecture ( more precisely the BIOS ) is limited to booting one of 4 partitions in the partition table ( those 4 are the primary ( or first ) partitions. It will boot the partition that is marked "active" .

Here's the layout.

active        hda1    10 MB    ext2 /boot   ( contains the kernel image )
bootable   hda2    2gig       ntfs5  win2k partition
bootable  hda3     2gig       fat32    win98 partitoin
extended  hda4     4010GB to 27 gig  - my extended partition

Now inside the extended partition:

hda5      2GB       fat32    (for windows )
hda6      200mb  ext2 / (root )
hda7      128mb  swap
hda8      128mb  swap
hda9      100MB  ext2 /var
hda10    2GB      ext2 /usr
hda11    500MB ext2 /root ( root user's home )
hda12    500MB  ext2 /tmp
hda13    6GB        fat32  ( for the games )
hda14    4GB        fat32   ( data exchange MS to linux and vice versa )
hda15    2GB        ext2     /home
hda16    4gig      ext2      /usr/local/datastore

Since hda1 is marked as active, the MBR code will boot it always, and guess what sits in the first 512k of that partition…LILO. So I see lilo first now, which allows me to boot one of several linux kernels, or win98 or NT5 ( win2k) .

If I wanted to put ,say, WinME ( thank you, but I'd rather not ) , I would load up fdisk or partition magic , and set the win98 or the win2k partition as active , reboot , and proceed with the install install. As far as the OS in question is concerned, it sees a primary active partition, and happily parks itself there .

Once the nauseating cycle of reboots is completed, in goes the partition magic disk, and I set the linux boot partition to active , taking care to "unhide" the other two primaries. This will avoid those "NTOSKRNL.exe not found" messages that sometimes prevents win2k from booting .( Why ? )

  1. Use partition magic to make the partitions , but with one exception : the win2k partition should be fat32 for installation. ( if your win2k cd is not bootable, use the win98 floppy to boot and access the cdrom. )

  2. mark the win2k as active, reboot and install win2k

    If you don't install win98 ,skip to step 6.
  3. hide the win2k partition ( don;t convert it yet ) .

  4. mark the second 2 gig as active, reboot and install win98.

  5. mark the win2k as active,and "unhide" the win98 partitoin.

  6. boot into win2k, and convert to ntfs. ( you will have to reboot for it to convert the file system.)

  7. Reboot and come up with the Partiton Magic disks.

  8. mark the first partition ( ext2 /boot ) as active, and make sure to "unhide" win2k partition.

  9. If you had put in win98 , unhide the win98 partition.

  10. Reboot and start the linux install

  11. When it asks you where to put lilo, put into the first sector or /boot .

  12. Provide appropriate names for the other boot options ( the installation will id ntfs as HPFS and win98 as dos ).

You're set.

From the above you may be thinking "that's complicated". But it is not, when you understand the underlying processes and ideas , like the co-relation between dos volume labels ( C: , D: etc.. ) and the partitions and their attributes .

A warning about win2k: Make sure your bios is compatable with win2k . I had a "minor" problem which had to do with the fact that my hdd geomertry had more than 1024 cylinders ( hdd > 8034MB ) , even in LBA.

Symptoms:

  1. NT4 will fail to completely boot the multiprocessor kernel for the installation .

  2. Win2k will complete the install, but will generate "STOP" errors ( among other IRQ_LEVEL_ errors in between random reboots) after the final reboot .

  3. But win98 and linux work fine.

In my case, a bios _downgrade_ fixed it.

Another warning about win2k : Once installed , the OS takes up 1 gig of space ( well 996MB is what 's used on my win2k boot partition). So atleast 2 gig is necessary . Also , it is best _not_ to grab a huge 5 or 10 gig partition as the win2k boot partition. In the event of an unexpected reboot , you will have to check all the "dirty" partitions… Usually after a boot, the only partition that is "dirty" is just the boot partition, so the file system check should take much lesser time if it is small. Also , when win2k decides to go south, it will take only itself down, and keep your data safe ( which is preferably on another partition ) .

hope I have not scared/bored you off.

joseph lascaux.ca