Unix Lesson 8: Device and Links

This lesson covers two kinds of special files in Unix: Devices, which are files that correspond to hardware devices, or special facilities provided by the Operating System, and Links, a special kind of file that can make files appear in two or more places.

Devices

Devices typically live in the /dev directory, and as mentioned correspond to a hardware device, such as a disk, or a special Operating System facility, such as the random number generator. They are created with a special program called mknod, but more commonly these days, they are created by special services that react to hardware events, such as inserting a USB device.

A disk is one of the major ways you will interact with devices. A CD-ROM, or a USB thumb drive has to be added to the filesystem. You accomplish this with the mount command. It takes the device, which has a special name (not the one shown here, they will always be different), and the directory you want to mount it into.

$ mount /dev/sda1 /mnt

Mounts "shadow", or temporarily hide all files that were previously in the directory you were mounting. Unmount to restore them with the umount command (note, no "n"!).

$ umount /mnt

Places where devices are mounted are called "mount points", to examine these, as well as how much storage is in use, use df, for "disk free".

Other Devices

As previously mentioned, the operating system provides a number of special facilities through files. Here is a short list of several of them. All these files live in /dev.

• null: this is "nothing". Redirecting I/O here means it will be discarded, as well as redirecting from it will immediately close the I/O for programs that expect input.
• zero: This file repeatedly produces 0 in ASCII, the standard for byte-encoding text in computers, which is called the "null" character -- do not confuse it with the null device above! The null character is a byte that takes space, just can't be seen. This file is used when you want to send a lot of pointless data somewhere in a hurry, like for example, when you want to overwrite a disk you previously had sensitive data on.
• /dev/fd/0, /dev/fd/1, /dev/fd/2, etc: These correspond to file descriptors for the current running program. Useful in scripts.
• stdin, stdout, stderr: correspond to the different types of standard I/O. Sometimes these are symbolically linked (keep reading) to /dev/fd/0, /dev/fd/1, and /dev/fd/2 respectively.
• random and urandom: access to the random number generator. random will spit out random data until the generator is exhausted, and will continue when it is "seeded", or has more. urandom will "fake it". The details of the faking are quite complex.

Links

There are two types of links: "hard", and "soft" or "symbolic". Links are a special kind of file which corresponds to another file's data on the filesystem.

Symbolic links are simply files which encode the "real" file or directory name in the link. The operating system handles this specially, and transparently. To use any kind of link, we use the ln command. To use a symbolic link, you must provide the -s flag, like so:

$ ln -s from to

Hard links are different. Each file system's mount point has a special table of numbers that correspond to the data on your computer's physical disk. These numbers are mapped to the name of the files in your filesystem. These numbers are called "inodes", and it is not important how they work.

A hard link is simply a second mapping of the file's inode to a new name. As a result:

• Hard links cannot span mount points
• Hard links may not be directories

Assignment

Create an empty file with echo foo >file.txt, then create a symbolic link to it named file2.txt. Then, rm and rewrite the file with echo bar >file.txt, and cat file2.txt. Describe what happened in your own words, and explain why, too. Then do it with a hard link; did anything change? Why?