Chapter 10 : Filesystems and mount points

The best way to understand "how it works" is to look at a practical case, which is what we are going to do here. Suppose you have just bought a brand new hard disk, still with no partitions on it. Your Linux-Mandrake partition is full to bursting, and rather than starting again from scratch, you decide to move a whole section of the tree structure to your new hard disk. As this new disk is very big, you decide to move your biggest directory to it: /usr. But first, a bit of theory.

Principles

As we already mentioned in the Install guide, every hard disk is divided into several partitions, and each of these partitions contains a filesystem. While Windows gives a letter to each of these filesystems (or actually, only to those it recognizes), Linux has a unique tree structure of files, and each filesystem is mounted at one location in the tree structure.

Just as Windows needs a "C: drive", Linux has to be able to mount the root of its file tree (/) somewhere, in fact on a partition which contains the root filesystem. Once the root is mounted, you can mount other filesystems in the tree structure, at different mount points in the tree structure. Any directory below the root one can act as a mount point.

This allows great flexibility in configuration. In the case of a web server, for example, it is common to dedicate a whole partition to the directory which hosts the web server data. The directory which contains them is generally /home/httpd, which will therefore act as the mounting point for the partition. You can see in figures 56.1 and 19.2 the situation of the system before and after mounting the filesystem.

As you can imagine, this offers a number of advantages: the tree structure will always be the same, whether if extends over a single filesystem or several dozens[20], and it is always possible to move a bulk key part of the tree structure physically to another partition when space becomes lacking, which is what we are going to do here.

There are two things you need to know about mount points:

• the directory which acts as a mount point must exist,
• this directory may not be empty: if a directory chosen as a mount point already contains files and subdirectories, these will simply be "hidden" by the newly mounted filesystem, but they will not be deleted.

Partitioning a hard disk and formatting a partition

Regarding the principles referred to above and as far as we are concerned in this section, there are two things to note: a hard disk is divided into partitions and each of these partitions hosts a filesystem. Now, at present, your brand new hard disk has neither, so that is where you have to start, beginning with the partitioning. For that you must be root.

First, you have to know the "name" of your hard disk, i.e. what file designates it. Suppose you set it up as a slave on your primary IDE interface, it will then be /dev/hdb[21].

The mount and umount commands

Now that the filesystem has been created, you can mount the partition. Initially, it will of course be empty. The command to mount filesystems is the mount command, and its syntax is as follows:

mount [options] <-t type> [-mount options] <device> <mounting point>

In this case, we want to mount our partition on /mnt (or any other mount point you have chosen -- don't forget that it must exist); the command for mounting our newly created partition is as follows:

$ mount -t ext2 /dev/hdb1 /mnt

The -t option is used to specify what type of filesystem the partition is supposed to host. Among the filesystems you will encounter most frequently, are ext2 (the Linux filesystem), vfat (for all DOS/Windows partitions: FAT 12, 16 or 32) and iso9660 (CDROM filesystem).

The -o option is used to specify one or more mounting options. These options depend on the filesystem used. Refer to the mount(8) manual page for more details.

Now that you have mounted your new partition, you need to copy the whole directory /usr into it:

$ (cd /usr && tar cf - .) | (cd /mnt && tar xpvf -)

Now that the files have been copied, we can unmount our partition. To do this the command is umount. The syntax is simple:

umount <mounting point|device>

So, to unmount our new partition, we can type:

$ umount /mnt

or else:

$ umount /dev/hdb1

As this partition is going to "become" our /usr directory, we need to tell the system. To do this, we fill in:

The /etc/fstab file

The /etc/fstab file makes it possible to automate mounting of certain filesystems, especially at system startup. It contains a series of lines describing the filesystems, their mount points and other options. Here is an example of a /etc/fstab file:

/dev/hda1   /           ext2    defaults        1 1
/dev/hda5   /home       ext2    defaults        1 2
/dev/hda6   swap        swap    defaults        0 0
/dev/fd0    /mnt/floppy auto    sync,user,noauto,nosuid,nodev,unhide 0 0
/dev/cdrom  /mnt/cdrom  auto    user,noauto,nosuid,exec,nodev,ro 0 0
none        /proc       proc    defaults        0 0
none        /dev/pts    devpts  mode=0622       0 0

A line contains, in order:

• the device hosting the filesystem,
• the mount point,
• the type of filesystem,
• the mounting options,
• the dump utility backup flag,
• the order of checking by fsck (FileSystem ChecK).

Surprise, surprise, there is always an entry for the root filesystem. The swap partitions are special since they are not visible in the tree structure, and the mount point field for these partitions contains the keyword swap. We will return to /proc in greater detail.

Let's get back to the subject. You have moved the whole /usr hierarchy to /dev/hdb1 and so you want this partition to be mounted at boot time. In that case you need to add an entry to the file:

/dev/hdb1 /usr ext2 defaults 1 2

Now the partition will be mounted at each boot. It will also be checked if necessary.

There are two special options: noauto and user. The noauto option specifies that the filesystem should not be mounted at startup but is to be mounted explicitly. The user option specifies that any user can mount and unmount the filesystem. As you can see, these two options are logically used for the CDROM drive and floppy drive. There are other options, and /etc/fstab even has its own manual page: fstab(5).

Last but not least of the advantages of this file is that it simplifies the mount command syntax. To mount a filesystem referenced in it, you can either reference the mount point or the device. So, to mount a floppy disk, you can type:

$ mount /mnt/floppy

or:

$ mount /dev/fd0

To finish with our example of moving a partition: we have copied the /usr hierarchy and completed /etc/fstab so that the new partition is mounted at startup. But for the moment the old /usr files are still there! We therefore need to delete them to free up space (which was, after all, our initial aim). To do this, then, you need to:

• delete all the files in the /usr directory (i.e. the "old" one, since the "new" one is not yet mounted): rm -Rf /usr/*;
• mount the "new" /usr: mount /usr

and you are finished.