Linux Guide/How Linux Works
The Linux Philosophy
Linux is built with a certain set of unifying principles in mind. Understanding these principles is very helpful in understanding how the system works as a whole. They are known as the "Linux Way", which is derived from the philosophy behind the UNIX system.
The Linux Way can be summarized as:
- Use programs that do only one task, but do it well.
- To accomplish complex tasks, use several programs linked together.
- Store information in human-readable plain text files whenever it is possible.
- There is no "one true way" to do anything.
- Prefer commandline tools over graphical tools.
Most traits of Linux are a consequence of these principles. In accordance with them, a Linux system is built out of small, replaceable components. We will examine the most important of them in more detail. Those are: the boot loader, the kernel, the shell, the X window server, the window manager and the desktop environment. After that, we will have a look at the file system in Linux. Finally, we will discuss the security of a computer running Linux.
Core components of a Linux system
This is the part of the system that is executed first. When you have only one operating system installed, it simply loads the kernel. If you happen to have multiple operating systems or multiple versions of the Linux kernel installed, it allows you to choose which one you want to start. The most popular bootloaders are GRUB (GRand Unified Bootloader) and Lilo (LInux LOader). Most users don't need to care about the boot loader, because it is installed and configured automatically. Actually, the boot Loader creates the boot sequence that the Linux Kernel requires.
The kernel is the central component of the system that communicates directly with the hardware. In fact, the name "Linux" properly refers to a particular kind of this piece of software. It allows programs to ignore the differences between various computers. The kernel allocates system resources like memory, processor time, hard disk space and external devices to the programs running on the computer. It separates each program from the others, so that when one of them encounters an error, others are not affected. Most users don't need to worry about the kernel in day-to-day use, but certain software or hardware will require or perform better with certain kernel versions.
The shell implements a textual interface that allows you to run programs and control the system by entering commands from the keyboard. Without a shell (or something that can replace it, like a desktop environment) making your system actually do something would be difficult. The shell is just a program; there are several different shells for Linux, each of which offering somewhat different features. Most Linux systems use the Bourne Again Shell (Bash). Linux shells support multitasking (running several programs at once).
initramfs, the initial ram file system, is a successor to initrd. It is a cpio archive of the file system that is loaded into memory during the Linux startup process. The Linux kernel mounts initramfs as a root file system and starts the init process from it. This completes certain tasks before the real root file system is loaded. initramfs needs to contain all of the device drivers and tools needed to mount the real root filesystem. The initramfs is one solution to the chicken-and-egg problem of some mass storage devices and some file systems, especially cryptographic file systems, which require complex device drivers to read them but normally store those drivers inside the file system in the mass storage device.
X Window Server
The X window server is a graphical replacement for the command shell. It is responsible for drawing graphics and processing input from the keyboard, mouse, tablets and other devices. The X server is network transparent; that is, it allows you to work in a graphical environment both on your own computer and on a remote computer to which you connect across a network. The X server that is most used today is X.Org. Most graphical programs need only the X server to run, so they can be used under any window manager and desktop environment.
The window manager is a program that communicates with the X server. Its task is managing windows. It is responsible for drawing the window borders, bringing a window to the front when you click it, moving it on the screen and hiding it when you minimize its program. Examples of popular window managers are:
- Metacity - GNOME Desktop Environment window manager
- KWin - KDE window manager
- Xfwm - Xfce window manager, a lightweight manager designed to consume as little resources as possible without compromising usability
- Compiz Fusion - an advanced window manager with lots of eye candy like customizable window animations, multiple desktops placed on a cube that you can rotate with your mouse, transparent window borders, wobbling windows while dragging them, etc.
Desktop environments such as GNOME Desktop Environment, KDE and Xfce are collections of programs designed to present a consistent user interface for most common tasks. They are what most people mean when they say "operating system" even though they are only a piece of the whole operating system. Replacing the desktop environment is sufficiently complex that most new users should stay with the default environment offered by their distribution.
There are several file systems that Linux-based distributions use. They are BTRFS, EXT3/4, VFS, NILFS, and SquashFS.
The hard drive of your computer has a rather simple interface. It only accepts commands like "read block no. 550923 and put it in memory address 0x0021A400". Suppose you are editing a piece of text and want to save it on the disk. Using block numbers (addresses) to identify pieces of data, like your text, is awkward: not only would you have to tell your program where to save the file using raw block numbers, you would have to make sure that these blocks aren't already being used for family photos, your music collection, or even your system's kernel. To solve this, files were introduced. A file is an area of the disk which stores data and which has a name (like "example.txt"). Files are organized in collections called directories. Directories can contain other directories, in a tree-like structure. Each file can be uniquely identified by a "path," which describes its place in the directory hierarchy. For the remainder of this section, it will be assumed that you are familiar with files, directories, and paths.
In Linux, the top-level directory is called the root directory. Every file and directory in the system must be a descendant of the root directory. (It is common to talk about directories using the terminology of family relations, like "parent," "child," "descendant," "ancestor," "sibling," and so forth.) Names of files and directories can contain all characters except the null character (which is impossible to enter from the keyboard) and the "/" character. An example path would be:
This path refers to a file called "error.log" which is found in a directory called "apache," which is a subdirectory of a directory called "logs," which is subdirectory of a directory called "var," which is a subdirectory of the root directory. The root directory doesn't have a name like the other; it is just denoted with a "/" character at the beginning of the path.
The root directory usually contains only a small number of subdirectories. The most important are:
- bin - programs needed to perform basic tasks, i.e. change a directory or copy a file
- dev - special files that represent hardware devices
- etc - configuration files
- home - contains private directories of users
- media or mnt - Mount point for external drives connected to this computer, i.e. CDs or USB keys
- tmp - temporary files
- usr - programs installed on the computer
- var - variable data produced by programs, like error logs
Devices as files
Just as files can be written to and read, devices in the computer system may send and receive data. Because of this, Linux represents the devices connected to the system as files in the /dev directory. These files can not be renamed or moved (they are not stored on any disk). This approach greatly simplifies application programming. If you want to send something to another computer through a serial port, you don't even need another program - you simply write to the file /dev/ttyS0, which represents a serial port. In the same manner the file representing the sound card (/dev/dsp) can be read to capture the sound from an attached microphone, or written to in order to produce sound through the speakers.
Where are the drive letters?
If you have used Windows, you might be surprised that there are no drive letters in Linux. The root directory represents the drive on which the system is installed (C: in Windows). Other drives can be "mounted" or "unmounted" in any directory (preferably, an empty one) in the file system. By mounting a disk, you attach the root directory of this disk to a directory in the file system. After that, you can access the disk like it were a part of your system disk. For example: if you have a disk that contains a directory text, which in turn contains a file called linux-intro.tex and you mount this drive in the directory /media/usbkey, you can access the file linux-intro.tex through the path /media/usbkey/text/linux-intro.tex.
In most Linux distributions, USB keys and CDs are automatically mounted when they are inserted or attached, and the default mount directory is a subdirectory of /media or /mnt. For example, your first CD-ROM drive might be mounted at /media/cdrom0, while the contents of a USB key might be accessible through /media/usb0. You may manually change the mount directory, but you will have to learn two shell commands and know the device file that represents your drive to do that (the one we talked about in the preceding section - disks also get their file representation in the /dev directory). We will cover this subject later.
The user is a metaphor for somebody or something interacting with the system. Users are identified by a user name and a password. Internally, each user has a unique number assigned, which is called a user ID, or UID for short. You only need to know your UID in some rare situations. Users can additionally be organized in groups. There is one special user in all Linux systems, which has the user name "root" and UID 0. It is also called the superuser. The superuser can do anything and is not controlled in any way by the security mechanisms. Having such a user account is very useful for administrative tasks and configuring the system. In some distributions (like Ubuntu) direct access to the root account is disabled and other mechanisms are used instead.
If you have more than one user account on a Linux system, you do not need to log out and back again to switch impersonations. There are special shell commands that allow you to access files and execute programs as other users, provided you know their user names and passwords. Thanks to this mechanism, you can spend most of the time as a user with low-privileges and switch to a higher-privileged account only if you need to.
The advantage of running as a non-privileged user is that any mistakes you happen to make are very unlikely to damage the system. System-critical components can only be altered by the root user.
Users exist to control the extent to which people and programs using the system can control it. This is accomplished by a system of file permissions. Each file belongs to one of the users - that is, each file has an owner. Additionally, a file can be assigned to a group of users, but the owner must be a member of that group. Each file has three kinds of permissions: read, write and execute. These permissions can be assigned to three kinds of owner relations: owner, group and other. Other includes all users who are not the owner of the file and do not belong to the group which owns the file. Only the file owner or the superuser (root) can change the permissions or ownership of a file.
This system allows precise control over who can do what on a given computer. Users can be prevented from modifying system files by removing the "write" permission from them, or from executing certain commands by removing the "execute" permission. Notice that users may be allowed to execute programs but not alter them. This is very important, since most Linux systems include a compiler that allows you to create your own programs.
File permissions are usually given as three octal digits (each from 0 to 7). The digits represent the permissions for, respectively, owner, group and other users. Each digit is the sum of permission codes: 1 for execute, 2 for write and 4 for read. For example, "755" allows everyone to read or execute the file, but only its owner can write it. "400" allows the owner to read the file, and no one else is allowed to do anything. "540" allows the owner to read or execute the file, group members to only read the file and other users to do nothing.