Complete Handbook/Users and the Linux file system

From Gentoo Wiki
Jump to: navigation, search

Why multi-user?

Separation of privileges

One of the advantages of a multi-user operating system like Linux is that the privileges are separated. Each process runs with specific privileges and can only execute a limited number of tasks. As long as the process does not run as the root user (the almighty administrator privilege) it can only deal with files and tasks that are assigned to that particular user.

This separation of privileges provides a small but working security wall: as long as all your users use the system with their user account and not with the root account, the worst that can happen is that a user removes their own files - the system itself is left untouched.

For this reason, you will always hear not to use the root account.

System accounts

To enforce the separation of privileges, specific system accounts are created for each task. If you run a mail server on your system, that mail server will have a user account on your system.

These accounts are not usable by regular users: you can not log in on your system using those accounts. They exist only to allow the specific processes to run with their own permissions and privileges.

Users and permissions

The user ID

To identify a system account, a unique user identification is used: the UID. This is a number used by the Linux kernel and other applications as numbers are easier to deal with than names (strings). However, Linux is intelligent enough to immediately translate the UID to a user name and vice versa, so in most cases you will only see or use the user name instead of the UID.

The process ID

When your Linux system is up and running, it will have started various processes already. Each process is an application (or part of an application) and receives a unique process identifier (which is also a number): the PID.

PIDs play an important role in the administration of a running Linux system: you need the PID of a specific process to be able to terminate it (in case it behaves badly), change priorities or receive specific system usage statistics regarding a particular application.

For regular Linux usage however, the PID is less important: you still need to understand what a PID is, but you will probably not encounter any use of it until you administer your system.


Each process obtains privileges based on the user account it uses. By default, a process runs with the privileges of the user that started the process. For instance, if you start firefox it will run with your privileges.

However, some processes have a specific flag set that tells the Linux kernel not to run the process as the user that executed it, but as a specific user instead. This flag is called the set user id (SetUID or SUID) and tells the Linux kernel to run this application with the privileges of the owner of that application instead of the executor.

Most tools that have the SUID bit set are owned by the root user and therefore start running with the privileges of the root user. Because this is a security threat (remember, running things as root can be dangerous) most tools have a feature called privilege separation: when they are started, they first run the tasks they have to run as root after which they automatically decrease their own privileges to a less powerful state.

Linux file system hierarchy

Structure of a file system

The most pertinent change Linux users will have to be comfortable with is the file system structure which is quite different from the file system structure operating systems like Microsoft Windows use.

In Linux, the entire file system is structured as one huge tree. You start with the root of the tree and traverse down until you reach your goal. The next code listing shows you the first depth of a Linux file system:

/        ## (The root)
+- bin/  ## (Executable programs needed to get the system up and running)
+- boot/ ## (Files related to the boot loader and Linux kernel)
+- dev/  ## (Device files)
+- etc/  ## (Configuration files)
+- home/ ## (User home directories)
+- lib/  ## (Libraries needed to get the system up and running)
+- mnt/  ## (Location for mount points)
+- opt/  ## (Contains large package installations not part of a regular install)
+- proc/ ## (Kernel-provided information)
+- root/ ## (Home directory for the root user)
+- sbin/ ## (System administration executables to get the system up and running)
+- sys/  ## (Kernel-provided information)
+- tmp/  ## (Temporary files)
+- usr/  ## (Applications for day-to-day system usage)
`- var/  ## (Variable information like log-files, caches, ...)

Suppose you want to navigate to the CUPS error logs (CUPS is a printing service frequently used on Linux systems) which are located inside /var/log/cups. You will find the following tree:

+- bin/
## (...)
+- usr/
`- var/
   +- cache/
   +- db/
   +- lock/
   +- log/
   |  +- cups/
   |  |  +- access_log
   |  |  +- error_log
   |  |  `- page_log
   |  +- dmesg
   |  +- emerge.log
   |  +- lastlog
   |  `- messages
   +- run/
   +- spool/
   +- state/
   `- tmp/

Each location has its purpose as defined in the Linux File System Hierarchy Standard. As said before, Linux builds upon standards and the file system structure is no exception. Each Linux distribution adheres to this standard (although a few deviations are known). If you want to learn more about the file system structure, please read this standard. A short summary can also be found on your Linux system in the hier man page ("hier" is short for "hierarchy").

Scattered files

One often frowned upon result of this file system structure is that applications scatter their files around on the system. Indeed, the locations for executable files, data files, documentation files, configuration files, ... are defined in the hierarchy standard but result in files being scattered throughout the file system instead of at a single location.

For instance, for a regular application the executable files will be stored in /usr/bin, data files in /usr/share/<application name>/, documentation files in /usr/share/man (for the manuals) or in the data file location, configuration files in /etc, libraries in /usr/lib, etc.

It is up to the distribution to keep track of the files that belong to a particular package. The software management system of a distribution is therefore a very important tool and is often the application that distinguishes one distribution from the others. For Gentoo, the software management system is called Portage.

System administration versus system usage

When you are using your Linux system for daily tasks you should be logged on as a regular user. This user will only have write-access to his personal home directory, located in /home, and have read access to most other places on the system (except where sensitive information is stored).

This user will be able to execute most applications that are stored in the regular executable locations (/bin, /usr/bin and a few other places). Whenever the user wants to run an application, the system will search through those directories for a matching application: it will not search through the entire system.

The administrative user (root) however has access to every location on the system. When they want to execute an application, the system will search through system administration locations such as /sbin and /usr/sbin as well. Those locations contain tools that should only be ran by the root user. The root user can also read and write to every location on the system (although particular kernel projects exist that allow for more access control, limiting even the root user's capabilities).

The role of hardware

Within the tree structure there does not seem to be any room for the hardware (like disks, CD-ROMs, USB sticks or network mounts). Of course, hardware is important - where else would you store your files on if you do not have a hard disk? The use of such hardware however happens transparent to the user.

Storing files in Linux happens in a layered structure. At the bottom of the layer, you have the actual storage (most likely the partition or removable media). On top of the actual storage you have the file system. A file system can be spanned across several storage devices but most users will have one partition per file system. The file system is mounted in the Linux file system structure. Such mount always happens at a certain directory.

By default, you will have at least one file system for the root of your file system. If you only want to use a single file system, you can have your entire Linux system on a single partition. If you want to use several partitions, you need to think about what directory (and its subdirectories) you want to store on a different file system.

For instance, you might want to have /home stored on a separate file system which allows you to have all the users' data on a single partition (or drive). What happens is that you create a file system on that partition (or drive) and then mount this at /home.

If you do not mount it at /home, the /home and all its contents will be stored on the file system that contains the root of the file system. If you do mount it at /home, /home and all its contents will be stored on the other file system.

This mounting does have an important implication: if you forget to mount a file system at a certain location, the Linux file system structure will look as if that location contains no files. You will be able to add files to that location of course, but they will then be stored on the root file system instead of on the file system you forgot to mount.

In the next code listing we show you an example layered approach. The root file system is stored on /dev/sda1 which represents the first partition on the first IDE disk in your system. The /home location is stored on a separate file system (which happens to be the same kind of file system: an ext3 one). This file system is stored on a meta device (a device that actually consists of multiple devices - in this case two partitions).

|   / (root)         |   /home (home directories) |  ## <- location
|   ext3 instance    |   ext3 instance            |  ## <- file system
|                    |        /dev/md/0           |
|   /dev/sda1        +--------------+-------------+  ## <- devices
|                    |  /dev/sdb1   |  /dev/sdc1  |