Now we need to create the filesystems on the partitions. In my case, the
first partition (12MB) is determined to be the root device, the second (12MB)
for swapping, the third (6MB) as /tmp, and the fourth as
/usr.
Normally you create an ext2-filesystem on your partitions. I use
the minix filesystem on my /tmp for easy data transfer between
Linux and AmigaOS. To create the filesystems, you type (as root):
/sbin/mkfs -t <fs-type> <device-node> <size>
or
/sbin/mkfs.<fs-type> <device-node> <size>
In my case, that's
/sbin/mkfs -t ext2 /dev/sdb1 12348
/sbin/mkfs.ext2 /dev/sdb1 12348
for the first partition (root). For ext2 filesystems, you might even leave
out the number determining the size of the partition because Linux can find
that out itself with its ext2 filesystem tools. For minix or other
filesystems, however, you will have to specify the size yourself.
The second partition is used for swapping and does not need a filesystem.
The third is /tmp, with minix filesystem (in my case, but you
can also use ext2):
/sbin/mkfs.minix /dev/sdb3 6223
And the fourth for /usr:
/sbin/mkfs.ext2 /dev/sdb4 71736
Remember: All these device nodes and block sizes are from Ingo's configuration, and are used here only as an example. You definitely will have to use different numbers for your setup.
If you want to make a break here, don't just switch off your Amiga. Type
/sbin/shutdown
and wait until it says that you may turn off the power.
First we need a second mountpoint, so create a directory:
mkdir /mnt2
Mount the partition that is supposed to become your root filesystem (again, all device nodes are examples from Ingo's installation):
mount -t ext2 /dev/sdb1 /mnt
Now mount the CDROM or the AmigaOS partition where the root.tar.gz is
located.
If you want to mount a CD, you will first have to create a device
node for it in /dev (there is already one in the new-filesys
filesystem, but it's wrong). Do this:
cd /dev
rm scd0
mknod scd0 b 11 0
Also, a convenient and usual point to mount a CD-ROM is /cdrom,
so create that directory if it doesn't exist already:
cd /
mkdir cdrom
If you are mounting a harddisk partition, no further steps should be
required to be able to mount it into Linux - unless the device nodes
for your AmigaOS partitions have not yet been created. If that is the
case, you will have to create them yourself using the mknod command.
An example: If you have an AmigaOS partition on the second SCSI disk, 9th
partition (which makes it sdb9 for Linux), you create it by typing:
mknod /dev/sdb9 b 8 25
The formula to determine the numbers is described below in the section "Creating the fstab file".
On Ingo's system, the first harddisk is used for AmigaOS, with 4
partitions, and the Linux archives are in Work:linux (Work: is
the second partition on the first harddisk, so that's /dev/sda2
for Linux):
For CD-ROM:
mount -t iso9660 /dev/scd0 /cdrom
For an AmigaOS partition:
mount -t affs /dev/sda2 /mnt2
Change to the new root:
cd /mnt
And extract the archive:
For CD:
gzip -c -d /cdrom/<path_to_files>/root.tar.gz | tar xvof -
For AmigaOS partition:
gzip -c -d /mnt2/<path_to_files>/root.tar.gz | tar xvof -
This will take some time...
Now for the /usr stuff. There is a slight problem here, since the
filenames in the archive are relative to the parent of the usr
directory: ./usr/*.
Since the ramdisk filesystem already uses /usr, we create a new one
(but first unmount the partition where you installed the root filesystem):
sync ; this will write out all buffer data to disk
cd / ; so we are not "in the way" of the partition
umount /dev/sdb1 ; to unmount the partition
mkdir /mnt/usr ; new directory
Now we mount the usr partition under this directory:
mount -t ext2 /dev/sdb4 /mnt/usr
And extract the archive:
For CD:
gzip -c -d /cdrom/<path_to_files>/usr.tar.gz | tar xvof -
For AmigaOS partition:
gzip -c -d /mnt2/<path_to_files>/usr.tar.gz | tar xvof -
This will take even more time...
After that, unmount everything for a clean shutdown (this also implicitly
performs the above mentioned sync command):
cd /
/sbin/shutdown
Now, hit CTRL-Amiga-Amiga to reboot into AmigaOS.
fstab fileFinally, you have to put together a list of filesystems that are to be mounted automatically when the system is coming up from harddisk.
This file is named /etc/fstab, and here is what it looks like on
Frank's machine (yes, this file has been written by two persons, and
we excuse for jumping from one's configuration to that of the other one of
us all the time - we hope you still keep in mind at any time that all these
are just examples of our personal setups, and will most probably have to
be modified for your machine) :
# /etc/fstab # static file system information, see fstab(5) # # This file is not used by the kernel, but rather by mount(8), umount(8), # swapon(8) and swapoff(8). (Some day fsck(8) will also use this file). # # Comment lines have "#" in the first column. The device field is either # the name of a block special file or a swap file. The device and directory # fields should be absolute pathnames (i.e. start with "/"). Entries that # are to be ignored should have type "ignore". Frequency and pass are numeric # fields for dump(8) and fsck(8) that are not yet used in Linux. You can # leave them empty if you want. # # Since entries in this file are processed sequentially, file systems must # appear after the file systems they are mounted on (/usr before /usr/spool). # # A sample fstab: # # device directory type options # /dev/sda9 / ext2 defaults /proc /proc proc defaults # /dev/sda8 /usr ext2 defaults /dev/sda10 /temp msdos defaults /dev/sda7 /usr2 ext2 defaults #/dev/sda11 none swap swap # # AmigaDOS affs partitions /dev/sdb1 /ados/sys affs ro /dev/sdb2 /ados/dh1 affs ro /dev/sdb3 /ados/dh2 affs ro /dev/sdb4 /ados/temp affs ro /dev/sdb5 /ados/tmp2 affs ro /dev/sda1 /ados/fdh0 affs ro /dev/sda3 /ados/fdh1 affs ro /dev/sda6 /ados/fdh4 affs ro # /dev/sda4 /mnt3 ext2 defaults /dev/sda5 /mnt4 ext2 defaults
The lines starting with a # character are comment lines.
All other lines denote what filesystems are to be mounted and where.
The structure of these lines is quite easy:
The first column is the path to the device node. For SCSI
harddisks, they start with /dev/sd.., for IDE harddisks use
/dev/hd... The second column gives the mount point. In my
case, /dev/sda9 is the root partition (/). The third column
denotes the filesystem type - you should have learnt about that
from the previous paragraphs.
The options column can give a couple of arguments which are
especially interesting for NFS mounted filesystems. All Amiga FFS
partitions are mounted ro which means "read-only". Everything
else (except the swap partition) is set to defaults.
The line with the /proc entry might look a bit strange. The /proc filesystem is a kernel-internal pseudo-filesystem for showing processes
and their environment as files. This makes it easy for some tools
to get ahold of kernel-related information. The /proc filesystem
does not have a device node under /dev. Also, you cannot delete
or rename files in /proc - they are just virtual files.
The line with the swap partition has been commented out right now
because you have not yet set up a swap partition. We will do this
as soon as the first version of your fstab file is working.
You will now have to create such an fstab file of your own. It
might be a good idea to take the above excerpt and use it as a
skeleton for your customized version. Comment out anything you do not
have, and insert lines you need as required.
When you are done, put your fstab file on an AmigaOS partition, mount
Linux again with a RAM filesystem as you have been doing before,
mount the AmigaOS partition with the fstab on it under e.g. /mnt,
mount your root partition e.g. under /mnt2 (you might have to do
mkdir /mnt2 again, as the changes you have done last time have
not been saved anywhere - remember this is only a RAM filesystem!)
and copy the fstab file from the AmigaOS partition to the
etc directory on your root partition (in our case, /mnt2/etc).
Important
The list of device nodes in /dev on your new root partition is not as
complete as the list of nodes in /dev of the RAM filesystem. This
means that you might not be able to access a couple of partitions if you do
not extend the list of nodes in /dev on the root partition (e.g.
it does not contain any nodes for IDE partitions (hda, hdb)). So you will have
to create those nodes yourself.
Mount your root partition e.g. on /mnt and go to /mnt/dev
(this is the dev/ directory of your root partition). Here, issue
the mknod command with corresponding parameters. The mknod
command is not on your RAM filesystem, so you will have to use the one on
your root partition:
cd /mnt/dev
/mnt/bin/mknod <parameters>
Here are a few examples for nodes you might want to create if they do not yet exist:
/mnt/bin/mknod sda9 b 8 9 ; for a ninth partition on SCSI disk sda
/mnt/bin/mknod sda10 b 8 10 ; for a tenth partition on SCSI disk sda
(There is a limit of 15 partitions per harddisk, numbered 1...15).
/mnt/bin/mknod hda1 b 3 1 ; for the first IDE partition on hda
/mnt/bin/mknod hda2 b 3 2 ; for the first IDE partition on hda
The second parameter is the name of the node you want to create. The third is always "b" for a block special file (harddisks, CD-ROMs). The fourth is the major device number. For SCSI this is 8, for IDE 3. The last number is the minor number. It starts at 1 for the first partition on the first disk, with a maximum of 15 partitions per harddisk. So, SCSI partitions sda1...15 have a minor number of 1...15. The second harddisk starts with minor number 16. The formula to calculate minor numbers is thus:
partition number + (harddisk number * 16)
Ok, so you should now verify that all nodes required in /dev (on
the root partition) are present, and if not, create them now. If you forget
one or more, you will get strange error messages about "corrupt filesystems"
or "invalid superblocks" the first time you try to start the kernel with
a root filesystem on a harddisk partition.
After this, issue the cd /; /sbin/shutdown command again, and reboot
once more when it has finished.
You are done !
So, here's the exciting moment: Boot Linux with your / and usr
filesystems from harddisk. In Ingo's case, this is:
amiboot-<x.y> -k vmlinux-<version> root=/dev/sdb1 ro video=pal,inverse
You will have to replace the sdb1 with whatever is your root partition.
Also, you might want to use another video mode (like video=ntsc,inverse
or the like).
The ro argument tells Linux that the root filesystem should be mounted
read-only for now. It is usually re-mounted read/write when the
/etc/rc file is executed. Without this option, you will get a
message like
warning: mounting unchecked file system
every time the system is started.
If everything goes well, the system should come up. If there are warnings about missing files or directories, ignore them for the moment.
You should be able to login as root (no password is set initially), look
around in the filesystem, try a few programs, and execute the
/sbin/shutdown when you want to turn off the power.
Ok, let's hope you managed to get everything running up to this point. Still missing is the creation of a swap partition. So that's what we'll do now.
You should have reserved a partition for swapping when you re-partitioned
your harddisk, and you should already have created an entry in /dev
for it. If not, read the previous section about creation of missing device
nodes. E.g. if your swap partition is to be on the 11th partition on the
first SCSI harddisk (as is the case for me), your command would be:
cd /dev
mknod sda11 b 8 11
Next, you need to initialize this partition - which is somewhat comparable to quick-formatting a partition under AmigaOS. Take the size of this partition from AmigaOS (you should have written it down), convert it to 1 KByte blocks (HDToolBox lists the size in 512 byte blocks, so divide the number by 2) and enter the command:
mkswap /dev/<device> <size_in_blocks>
In my case, this had been:
mkswap /dev/sda11 <size_in_blocks>
If you are uncertain about the size_in_blocks parameter, double-check
it by looking into HDToolBox again. The parameter you give should be
the size in KBytes, so if you have a 16 MByte swap partition, the
value should be somewhere in the 16000 area.
If the mkswap command was successful, you can now remove the comment from
the /etc/fstab file that shows the swap partition (don't forget to
correct the device name here if it's not correct for your system).
Next, you can enable swapping by typing in:
swapon -a
This means "Enable Swapping on all partitions listed for swapping in the
/etc/fstab file". If you should ever feel the need for several swap
partitions, you can add them to the /etc/fstab file.
Alternatively, you can give the swapon command a device as parameter
so that swapping is only started on this particular swap partition, like
in:
swapon /dev/sda11
When swapping is to be turned off, the command to do that is swapoff.
Its use is identical to swapon: Either use
swapoff -a
to stop swapping on all partitions, or give a device name like in
swapoff /dev/sda11
The swapon -a command should be executed automatically when your system
is booting. Look into the system start file, /etc/rc, and search for
such a line. If it's not yet there, add it somewhere at the top -
swapping should be enabled as soon as possible.
Similarly, swapping should be automatically disabled when you want to
bring down the system. Some init programs (SysV-init in particular)
do this via a script file like /etc/rc.d/rc.halt. but the init you
will probably be using (simpleinit) does not do that, so you might want to
write a small script for shutting down which does just this.
A word of warning about swap partitions:
It is possible to use a swap partition under both AmigaOS (for instance
with VMM) and Linux. I don't yet have any experience with VMM, but if
you used a partition under AmigaOS (or NetBSD or whatever) and want to
use it again for swapping under Linux, you will have to issue the mkswap
command on that partition under Linux again. If you don't do that, Linux
will complain with a message like
Unable to find swap signature
when trying to enable swapping. However, I have personally had a situation where this did not happen. Ok, I'll explain:
I wanted to use /dev/sda11 under AmigaOS for a small test.
I re-enabled it using HDToolBox and quick-formatted it (I did not write
anything else on that partition!). When I restarted Linux, it did not
complain about not being able to find the swap signature - probably because
it was not overwritten when quick-formatting. However, 2 days later I got
very strange "SCSI I/O error on device 080b" messages on my console which
made me believe my harddisk was going to die (imagine the panic!). But
then I saw that in fact Linux tried to access blocks outside the
partition - so the size information on the partition was somehow mangled.
I did a new mkswap on the swap partition, and everything worked again.
(People told me swap partitions should be at the beginning of a harddisk
to make swapping as fast as possible, but in this case I was very happy
that my swap partition was on the innermost cylinders - this way there
was no other data behind the partition that might have become trashed).
Now you might think it could be a good idea to automatically do the
mkswap each time the system is coming up - e.g. in the /etc/rc
file. But this can also be dangerous: This way you have to keep both
partition-dependent files in mind (/etc/rc and /etc/fstab).
If you ever repartition the harddisk again so that the Linux partitions
get shifted to new positions, you have to correct both files.
An example: For a long time, my root partition was on /dev/sda8
and the swap partition on /dev/sda9. Then I split a lower
partition into two halves so that root was then on /dev/sda9
and swap on /dev/sda10. If I had the command mkswap /dev/sda9 <size_of_blocks> in my /etc/rc file, I would have destroyed
my root partition on the next Linux boot. Be careful!
Next Chapter, Previous Chapter
Table of contents of this chapter, General table of contents
Top of the document, Beginning of this Chapter