Olivier Chapuis About the author: In real life, the author does mathematics, in particular logic, algebra and theory of computation. He recently discovered Linux and that was for him a true revelation. He loves Muriel and his current hobby is scuba diving. He also enjoys Brittany, the sea, boating, skiing and good food. Content: |
Abstract:
In this article, I explain how to remove windows from a machine which contains both Linux and windows, and moreover, how to use the available space for Linux. This article is intended for naïve users (I am one of them myself).
It occurred to me one day, about two months after having installed Linux, that I wanted to remove windows. The reason was simply that I didn't use it anymore! I imagine that a number of new Linux users are in the same situation, and I wrote this document to help them (the gurus can also read it and suggest improvements). If you are in this situation, two strategies are available.
The first is to reinstall Linux. The advantage of this strategy is that with a little Linux experience, you will make better choices on installation, in particular those to do with partition sizes. The main disadvantage is that you will loose everything. You can avoid this in part by backing up everything that isn't on your installation CDROM. But reinstalling this backup will cost you time and sweat.
The second strategy is to remove windows (which is easy), then use the space available for Linux. That is to say, <<repartition>>, an operation a bit tricky and little documented. The advantage of this method is if you succeed, your system will be <<as before>> but with more disk space. The second advantage is that if you fail, you will always be able to use the first strategy. In this document, I explain how I did it. Of course, the method will depend on your configuration. In particular, if you have an enormous virgin disk, you can probably first backup your files to this disk and then install from scratch. I imagine that this is a safer strategy which gives a better result. Me, I back up onto 1.44 Mb floppies. In any case, I hope the strategy described below will help a new user.
I would like to thank Olivier Tharan for his detailed response to a message posted in fr.col.moderated which was version -1 of this document. I would also like to thank Finn Bo Jorgensen and Chmouel Boudjnah for their response.
- Preparation: backup, examine and take notes on certain important files before proceeding. Read the document to understand what you are going to do. (section 3 below).
- Remove windows (section 4 below).
- Repartition, the tricky bit (section 5 below).
Above all, back up the essential files. In particular, the configuration files and your work. I won't say anything more below.
A preliminary step is to understand what you are going to do. There is some explanation of the commands and files in what follows, but it is very incomplete. Read the man pages! In this case, fdisk, mk2ext, cp (a not so stupid command), lilo, lilo.conf and fstab (there are also man pages for certain configuration files). Very detailed docs on lilo are in the directory /usr/doc/lilo-0.20/doc/ (if your system is correctly installed). In particular there is a <<user guide>> of 46 pages! For fdisk see the file /usr/doc/util-linux-2.7/README.fdisk, also very complete. In writing this document I learned of a very useful mini-Howto which is relevant: Hard Disk Upgrade Mini How-To.
First, here is my partition table:
[olivier@snoopy olivier]# /sbin/fdisk -l (as su) Disk /dev/hda: 128 heads, 63 sectors, 788 cylinders Units = cylinders of 8064 * 512 bytes Device Boot Begin Start End Blocks Id System /dev/hda1 * 1 1 397 1600672+ c Unknown /dev/hda2 398 398 601 822528 83 Linux native /dev/hda3 602 602 627 104832 82 Linux swap /dev/hda4 628 628 788 649152 83 Linux native |
We see that windows is on hda1. It is not unusual for fdisk to say <<Unknown>> for a FAT32 partition. My kernel (currently 2.0.34) knows how to access this type of filesystem, but that's not the case for my fdisk. To demonstrate, here is the output of df (with the dos partition mounted).
[olivier@snoopy olivier]# df Filesystem 1024-blocks Used Available Capacity Mounted on /dev/hda2 796065 506992 247947 67% / /dev/hda4 628351 226366 369528 38% /home /dev/hda1 1594456 850420 744036 53% /mnt/dos |
You will have to edit two configuration files /etc/fstab and /etc/lilo.conf. They will look something like this:
[olivier@snoopy olivier] less /etc/fstab /dev/hda2 / ext2 defaults 1 1 /dev/hda4 /home ext2 defaults 1 2 /dev/hda3 swap swap defaults 0 0 /dev/fd0 /mnt/floppy ext2 user,noauto 0 0 /dev/cdrom /mnt/cdrom iso9660 user,noauto,ro 0 0 none proc proc defaults 0 0 /dev/hda1 /mnt/dos msdos user,noauto 0 0 |
[olivier@snoopy olivier] less /etc/lilo.conf boot=/dev/hda map=/boot/map install=/boot/boot.b prompt timeout=500 image=/boot/vmlinuz-2.0.34-1 label=linux root=/dev/hda2 read-only other=/dev/hda1 label=dos table=/dev/hda |
Now you must do everything as root. For psychological reasons, I advise you to log in as root (don't su from your account). If you use vi to edit, you may have to use w! to save. In any case, I advise you to verify that your modifications were indeed saved (leave your editor and look at the file with less).
You are going to use the commands fdisk and mke2fs. These commands work with hard disk partitions and all other storage systems. It is better to use these commands on partitions that are not mounted. It is all done in three steps.
- If your partition is mounted, unmount it: umount /mnt/dos.
- Start fdisk. Change the identity of hda1: we will tell linux that this partition will now be a Linux native partition. Do this with the command t (t, 1, 83). Look at: p , the only thing which changes is the Unkown which becomes linux native and the c which becomes 83. Save and quit: w. Look again at: fdisk -l. In quitting fdisk I received the following error message:
The partition table has been altered! Calling ioctl() to re-read partition table. Syncing disks. Re-read table failed with error 16: Device or resource busy. Reboot your system to ensure the partition table is updated. |
As far as I know, this message is normal and harmless: I didn't reboot.
- Format: mke2fs /dev/hda1. Plenty of numbers appear then the system works for a bit and poof! windows 98 is no longer! The prudent will add the -c option which does a rapid check for bad blocks.
There you have it. Windows is gone. I wasn't especially thrilled by this (I don't have anything personal against win98, as opposed to Microsoft ...). But I was happy for Linux ...
Good, now I have a 1.5 Gb partition with nothing on it, a 600 Mb partition for /home and an 800 Mb for / (and a swap partition of 100 Mb). There is 100 Mb in /home/myaccount which should be in /usr/local/ (I will leave it on /dev/hda2) and there is 10 Mb in /usr/local. I will move / to hda1 and put /usr/local on hda2. This will give:
Filesystem 1024-blocks Used Available Capacity Mounted on /dev/hda1 1600672 465000 1130672 29% / /dev/hda2 796065 110000 696065 13% /usr/local /dev/hda4 628351 40739 587612 6% /home |
Enough space for a dozen years ....
Now to make it so. To access hda1 you must mount it somewhere. Generally you use /mnt.
- Create a new directory in /mnt, lets say /mnt/tmp, to provide a proper mount point for hda1 (mount doesn't create directories): mkdir /mnt/tmp.
- Mount hda1 on /mnt/tmp : mount -t ext2 /dev/hda1 /mnt/tmp (clearly -t ext2 gives the filesystem type).
- Copy / to /mnt/tmp/ while preserving its structure: cp -ax / /mnt/tmp/. For the -a option, the man page says: <<Preserve as much possible of the structure and attributes of the original file in the copy. The same as -dpR>>. See the cp man page for the significance of d, p and R, which will better explain the a. The -x option asks to skip the subdirectories which are on different filesystems from the one the copy started on. In my case, this avoids copying /home onto /dev/hda1. The copy operation will take some time (about 15 min with my IDE disk and my AMD K6 at 266 MHz), now for the usual phrase: go have a coffee or a double whisky while waiting.
- There are still some things missing from hda1. Because of the -x option, we are missing the two directories (empty) /home and /proc. So mkdir /mnt/tmp/home and mkdir /mnt/tmp/proc. These directories are the mount points that must be there so that /home (which is on hda3) and /proc can be mounted on the root.
Now things get a little complicated. We have / (=/-/home) on hda2 but also on /mnt/tmp/. We are going to get ready to restart. We must edit fstab and lilo.conf. Linux must be told that we are going to restart with the / which is on /mnt/tmp/. Thus, for fstab the file which we must modify is in fact /mnt/tmp/etc/fstab. For lilo.conf it is different. We must edit /etc/lilo.conf, run lilo and also copy the new /etc/lilo.conf to /mnt/tmp/etc/lilo.conf (to have a consistent system). Or we could edit only /mnt/tmp/etc/lilo.conf and run lilo with the -C option telling it that it must use a different configuration file. I chose the second method.
- /mnt/tmp/etc/fstab
This file indicates how and where to mount everything that is mountable on startup. When we restart, we want the root / to be hda1, so we need the first line of /etc/fstab as below. Of course, we can't mount two roots! Since we have to work on /dev/hda2, I'm mounting it on /mnt/tmp. The rest stays the same.
/dev/hda1 / ext2 defaults 1 1 /dev/hda2 /mnt/tmp ext2 defaults 1 2 /dev/hda4 /home ext2 defaults 1 3 /dev/hda3 swap swap defaults 0 0 /dev/fd0 /mnt/floppy ext2 user,noauto 0 0 /dev/cdrom /mnt/cdrom iso9660 user,noauto,ro 0 0 none /proc proc defaults 0 0 |
- /mnt/tmp/etc/lilo.conf.
lilo.conf is the configuration file for lilo, the little program on the MBR. I won't go into the details. If some people aren't using lilo, I very strongly advise them to switch (lodelin is an absurdity in our case, and why boot off a floppy these days?). Here we tell just lilo that now the kernel is on hda1 (it was on hda2 before) and of course we remove everything concerning windows.
boot=/dev/hda map=/boot/map install=/boot/boot.b prompt timeout=500 image=/boot/vmlinuz-2.0.34-1 label=linux root=/dev/hda1 read-only |
- Run lilo: /sbin/lilo -v -C /mnt/tmp/etc/lilo.conf. The v is so that lilo tells us a little more about what's going on. Here we mustn't see an error message! We can't make mistakes if we want to reboot. We should see something like:
LILO version 20, Copyright 1992-1997 Werner Almesberger Reading boot sector from /dev/hda Merging with /boot/boot.b Boot image: /boot/vmlinuz-2.0.34 Added linux * /boot/boot.0300 exists - no backup copy made. Writing boot sector. |
- Reboot. Some small tests (df, fdisk -l, we see if the filesystem is as we have specified).
- Remove everything from /mnt/tmp/, which we do like this: rm -rf * in the directory /mnt/tmp. Next we copy the directory /usr/local which stays: cp -a * /mnt/tmp after cd /usr/local. Check if everything is okay, then delete /usr/local: rm -fr * in the directory /usr/local (note that we mustn't remove the <<name>> which we will soon give as the mount point for local!).
- Again edit /etc/fstab and change just the second line:
/dev/hda2 /usr/local ext2 defaults 1 2 |
- Reboot. Do some tests.
And the operation is done. I assure you that I was satisfied, especially after I tested a lot of programs and everything worked!
True, I had a little problem. The above isn't exactly as it happened, but an idealization. The advice I can give you is to understand a little about what you are doing which will allow you to understand why it isn't working and to fix the situation. Olivier Tharan advised me to restart in single user mode (I didn't because the only documentatation that I had on single user mode is <<look at the files in single user boot mode>> and I was a little bit lazy). Before the first reboot it is perhaps prudent to have a good reboot disk (which perhaps must be adapted to the situation and I admit to not knowing exactly how to do it...)
Here is the problem I had. Before the first reboot I did nearly everything above. But after rebooting to the stage of mounting the filesystems a bunch of error messages flashed by at high speed. I managed to read: /proc ... error ... error ... /home ... error (without a trace in /var/log/messages). Then (phew) the login prompt appeared. I logged on as root and immediately
[root@snoopy root]$ df Filesystem 1024-blocks Used Available Capacity Mounted on /dev/hda1 1548535 506977 961525 35% / /dev/hda2 796065 506992 247947 67% /mnt/tmp |
Home wasn't mounted. I editted fstab, but all seemed normal. Then I had a brilliant idea (the double whisky?) to mount /home and I was greeted with the message <<mount: mount point /home does not exist>>. Then I understood that I didn't understand what a mount point was and that I had forgotten the option x of cp (I admit to being a little ashamed). I created the mount points /home and /proc and I mounted /home and /proc to see if it was okay. Then the operation was finished as described above.
Webpages maintained by the LinuxFocus Editor team
© Olivier Chapuis LinuxFocus 1999 |