%dbcent; %commondata; %popcon; %datadate1; %pkgsize; %datadate2; ]> Osamu Aoki OA This book is free; you may redistribute it and/or modify it under the terms of the GNU General Public License of any version compliant to the Debian Free Software Guidelines (DFSG). 2007-2009 Osamu Aoki This Debian Reference (v2) (&build-date;) is intended to provide a broad overview of the Debian system as a post-installation user's guide. It covers many aspects of system administration through shell-command examples for non-developers. This Debian Reference (version 2) (&build-date;) is intended to provide a broad overview of Debian system administration as a post-installation user guide. The target reader is someone who is willing to learn shell scripts but who is not ready to read all the C sources to figure out how the GNU/Linux system works.

All warranties are disclaimed. All trademarks are property of their respective trademark owners. The Debian system itself is a moving target. This makes its documentation difficult to be current and correct. Although the current unstable version of Debian system was used as the basis for writing this, some contents may be already outdated by the time you read this. Please treat this document as the secondary reference. This document does not replace any authoritative guides. The author and contributors do not take responsibility for consequences of errors, omissions or ambiguity in this document.

The Debian Project is an association of individuals who have made common cause to create a free operating system. It's distribution is characterized by the following. Commitment to the software freedom: Debian Social Contract and Debian Free Software Guidelines (DFSG) Internet based distributed unpaid volunteer effort: http://www.debian.org Large number of pre-compiled high quality softwares Focus on stability and security with easy access to the security updates Focus on smooth upgrade to latest softwares with unstable and testing archives Large number of supported hardware architectures Free Software pieces in Debian come from GNU, Linux, BSD, X, ISC, Apache, Ghostscript, Common Unix Printing System , Samba, GNOME, KDE, Mozilla, OpenOffice.org, Vim, TeX, LaTeX, DocBook, Perl, Python, Tcl, Java, Ruby, PHP, Berkeley DB, MySQL, PostgreSQL, Exim, Postfix, Mutt, FreeBSD, OpenBSD, Plan 9 and many more independent free software projects. Debian integrates this diversity of Free Software into one system.

Following guiding rules are followed while compiling this document. Provide overview and skip corner cases. (Big Picture) Keep It Short and Simple. (KISS) Do not reinvent the wheel. (Use pointers to the existing references) Focus on non-GUI tools and consoles. (Use shell examples) Be objective. (Use popcon etc.) I tried to elucidate hierarchical aspects and lower levels of the system.

You are expected to make good efforts to seek answers by yourself beyond this documentation. This document only gives efficient starting points. You must seek solution by yourself from primary sources. The Debian site at http://www.debian.org for the general information The documentation under the "/usr/share/doc/<package_name>" directory The Unix style manpage: "dpkg -L <package_name> |grep '/man/man.*/'" The GNU style info page: "dpkg -L <package_name> |grep '/info/'" The bug report: http://bugs.debian.org/<package_name> The Debian Wiki at http://wiki.debian.org/ for the moving and specific topics The HOWTOs from The Linux Documentation Project (TLDP) at http://tldp.org/ The Single UNIX Specification from the Open Group's The UNIX System Home Page at http://www.unix.org/ The free encyclopedia from Wikipedia at http://wikipedia.org/ For detailed documentation, you may need to install the corresponding documentation package named with "-doc" as its suffix.

This document provides information through the following simplified presentation style with bash(1) shell command examples. # <command in root account> $ <command in user account> These shell prompts distinguish account used and correspond to set environment variables as: "PS1='\$'" and "PS2=' '". These values are chosen for the sake of readability of this document and are not typical on actual installed system. See the meaning of the "$PS1" and "$PS2" environment variables in bash(1). Action required by the system administrator is written in the imperative sentence, e.g. "Type Enter-key after typing each command string to the shell." The description column and similar ones in the table may contain a noun phrase following the package short description convention which drops leading articles such as "a" and "the". They may alternatively contain an infinitive phrase as a noun phrase without leading "to" following the short command description convention in manpages. These may look funny to some people but are my intentional choices of style to keep this documentation as simple as possible. These Noun phrases do not capitalize their starting nor end with periods following these short description convention. Proper nouns including command names keeps their case irrespective of their location. A command snippet quoted in a text paragraph is referred by the typewriter font between double quotation marks, such as "aptitude safe-upgrade". A text data from a configuration file quoted in a text paragraph is referred by the typewriter font between double quotation marks, such as "deb-src". A command is referred by its name in the typewriter font optionally followed by its manpage section number in parenthesis, such as bash(1). You are encouraged to obtain information by typing the following. $ man 1 bash A manpage is referred by its name in the typewriter font followed by its manpage section number in parenthesis, such as sources.list(5). You are encouraged to obtain information by typing the following. $ man 5 sources.list An info page is referred by its command snippet in the typewriter font between double quotation marks, such as "info make". You are encouraged to obtain information by typing the following. $ info make A filename is referred by the typewriter font between double quotation marks, such as "/etc/passwd". For configuration files, you are encouraged to obtain information by typing the following. $ sensible-pager "/etc/passwd" A directory name is referred by the typewriter font between double quotation marks, such as "/etc/init.d/". You are encouraged to explore its contents by typing the following. $ mc "/etc/init.d/" A package name is referred by its name in the typewriter font, such as vim. You are encouraged to obtain information by typing the following. $ dpkg -L vim $ apt-cache show vim $ aptitude show vim A documentation may indicate its location by the filename in the typewriter font between double quotation marks, such as "/usr/share/doc/sysv-rc/README.runlevels.gz" and "/usr/share/doc/base-passwd/users-and-groups.html"; or by its URL, such as http://www.debian.org. You are encouraged to read the documentation by typing the following. $ zcat "/usr/share/doc/sysv-rc/README.runlevels.gz" | sensible-pager $ sensible-browser "/usr/share/doc/base-passwd/users-and-groups.html" $ sensible-browse "http://www.debian.org" An environment variable is referred by its name with leading "$" in the typewriter font between double quotation marks, such as "$TERM". You are encouraged to obtain its current value by typing the following. $ echo "$TERM"

Astarisk "*" placed right after each package name is linked to Debian bug tracking system (BTS) of each package.

The popcon data is presented as the objective measure for the popularity of each package. It was downloaded on &pop-date; and contains the total submission of &pop-submissions; reports over &pop-packages; binary packages and &pop-architectures; architectures. Please note that the &arch; unstable archive contains only &all-packages; packages currently. The popcon data contains reports from many old system installations. The popcon number preceded with "V:" for "votes" is calculated by "100 * (the popcon submissions for the package executed recently on the PC)/(the total popcon submissions)". The popcon number preceded with "I:" for "installs" is calculated by "100 * (the popcon submissions for the package installed on the PC)/(the total popcon submissions)". The popcon figures should not be considered as absolute measures of the importance of packages. There are many factors which can skew statistics. For example, some system participating popcon may have mounted directories such as "/bin" with "noatime" option for system performance improvement and effectively disabled "vote" from such system.

The package size data is also presented as the objective measure for each package. It is based on the "Installed-Size:" reported by "apt-cache show" or "aptitude show" command (currently on &arch; architecture for the unstable release). The reported size is in KiB (Kibibyte = unit for 1024 bytes). A package with a small numerical package size may indicate that the package in the unstable release is a dummy package which installs other packages with significant contents by the dependency. The dummy package enables a smooth transition or split of the package. A package size followed by "(*)" indicates that the package in the unstable release is missing and the package size for the experimental release is used instead.

Please file bug reports on the debian-reference package using reportbug(1) if you find any issues on this document. Please include correction suggestion by "diff -u" to the plain text version or to the source.

Here are some interesting quotes from the Debian mailing list which may help enlighten new users. "This is Unix. It gives you enough rope to hang yourself." --- Miquel van Smoorenburg <miquels at cistron.nl> "Unix IS user friendly… It's just selective about who its friends are." --- Tollef Fog Heen <tollef at add.no>

I think learning a computer system is like learning a new foreign language. Although tutorial books and documentation are helpful, you have to practice it yourself. In order to help you get started smoothly, I elaborate a few basic points. The powerful design of Debian GNU/Linux comes from the Unix operating system, i.e., a multiuser, multitasking operating system. You must learn to take advantage of the power of these features and similarities between Unix and GNU/Linux. Don't shy away from Unix oriented texts and don't rely solely on GNU/Linux texts, as this robs you of much useful information. "Rute User's Tutorial and Exposition", in the Debian non-free archive as the rutebook package (popcon: &pop-rutebook;), provides a good online resource to the generic system administration. If you have been using any Unix-like system for a while with command line tools, you probably know everything I explain here. Please use this as a reality check and refresher.

Upon starting the system, you are presented with the character based login screen if you did not install X Window System with the display manager such as gdm. Suppose your hostname is foo, the login prompt looks as follows. foo login: If you did install a GUI environment such as GNOME or KDE, then you can get to a login prompt by Ctrl-Alt-F1, and you can return to the GUI environment via Alt-F7 (see below for more). At the login prompt, you type your username, e.g. penguin, and press the Enter-key, then type your password and press the Enter-key again. Following the Unix tradition, the username and password of the Debian system are case sensitive. The username is usually chosen only from the lowercase. The first user account is usually created during the installation. Additional user accounts can be created with adduser(8) by root. The system starts with the greeting message stored in "/etc/motd" (Message Of The Day) and presents a command prompt. Debian GNU/Linux lenny/sid foo tty1 foo login: penguin Password: Last login: Sun Apr 22 09:29:34 2007 on tty1 Linux snoopy 2.6.20-1-amd64 #1 SMP Sun Apr 15 20:25:49 UTC 2007 x86_64 The programs included with the Debian GNU/Linux system are free software; the exact distribution terms for each program are described in the individual files in /usr/share/doc/*/copyright. Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent permitted by applicable law. foo:~$ Here, the main part of the greeting message can be customized by editing the "/etc/motd.tail" file. The first line is generated from the system information using "uname -snrvm". Now you are in the shell. The shell interprets your commands.

If you installed X Window System with a display manager such as GNOME's gdm by selecting "Desktop environment" task during the installation, you are presented with the graphical login screen upon starting your system. You type your username and your password to login to the non-privileged user account. Use tab to navigate between username and password, or use the mouse and primary click. You can gain the shell prompt under X by starting a x-terminal-emulator program such as gnome-terminal(1), rxvt(1) or xterm(1). Under the GNOME Desktop environment, clicking "Applications" → "Accessories" → "Terminal" does the trick. You can also see the section below . Under some other Desktop systems (like fluxbox), there may be no obvious starting point for the menu. If this happens, just try (right) clicking the center of the screen and hope for a menu to pop-up.

The root account is also called superuser or privileged user. From this account, you can perform the following system administration tasks. Read, write, and remove any files on the system irrespective of their file permissions Set file ownership and permissions of any files on the system Set the password of any non-privileged users on the system Login to any accounts without their passwords This unlimited power of root account requires you to be considerate and responsible when using it. Never share the root password with others. File permissions of a file (including hardware devices such as CD-ROM etc. which are just another file for the Debian system) may render it unusable or inaccessible by non-root users. Although the use of root account is a quick way to test this kind of situation, its resolution should be done through proper setting of file permissions and user's group membership (see ).

Here are a few basic methods to gain the root shell prompt by using the root password. Type root at the character based login prompt. Click "Applications" → "Accessories" → "Root Terminal", under the GNOME Desktop environment. Type "su -l" from any user shell prompt. This does not preserve the environment of the current user. Type "su" from any user shell prompt. This preserves some of the environment of the current user.

When your desktop menu does not start GUI system administration tools automatically with the appropriate privilege, you can start them from the root shell prompt of the X terminal emulator, such as gnome-terminal(1), rxvt(1), or xterm(1). See and . Never start the X display/session manager under the root account by typing in root to the prompt of the display manager such as gdm(1). Never run untrusted remote GUI program under X Window when critical information is displayed since it may eavesdrop your X screen.

In the default Debian system, there are six switchable VT100-like character consoles available to start the command shell directly on the Linux host. Unless you are in a GUI environment, you can switch between the virtual consoles by pressing the Left-Alt-key and one of the F1 — F6 keys simultaneously. Each character console allows independent login to the account and offers the multiuser environment. This multiuser environment is a great Unix feature, and very addictive. If you are under the X Window System, you gain access to the character console 1 by pressing Ctrl-Alt-F1 key, i.e., the left-Ctrl-key, the left-Alt-key, and the F1-key are pressed together. You can get back to the X Window System, normally running on the virtual console 7, by pressing Alt-F7. You can alternatively change to another virtual console, e.g. to the console 1, from the commandline. # chvt 1

You type Ctrl-D, i.e., the left-Ctrl-key and the d-key pressed together, at the command prompt to close the shell activity. If you are at the character console, you return to the login prompt with this. Even though these control characters are referred as "control D" with the upper case, you do not need to press the Shift-key. The short hand expression, ^D, is also used for Ctrl-D. Alternately, you can type "exit". If you are at x-terminal-emulator(1), you can close x-terminal-emulator window with this.

Just like any other modern OS where the file operation involves caching data in memory for improved performance, the Debian system needs the proper shutdown procedure before power can safely be turned off. This is to maintain the integrity of files, by forcing all changes in memory to be written to disk. If the software power control is available, the shutdown procedure automatically turns off power of the system. (Otherwise, you may have to press power button for few seconds after the shutdown procedure.) You can shutdown the system under the normal multiuser mode from the commandline. # shutdown -h now You can shutdown the system under the single-user mode from the commandline. # poweroff -i -f Alternatively, you may type Ctrl-Alt-Delete (The left-Ctrl-key, the left-Alt-Key, and the Delete are pressed together) to shutdown if "/etc/inittab" contains "ca:12345:ctrlaltdel:/sbin/shutdown -t1 -a -h now" in it. See inittab(5) for details. See .

When the screen goes berserk after doing some funny things such as "cat <some-binary-file>", type "reset" at the command prompt. You may not be able to see the command echoed as you type. You may also issue "clear" to clean up the screen.

Although even the minimal installation of the Debian system without any desktop environment tasks provides the basic Unix functionality, it is a good idea to install few additional commandline and curses based character terminal packages such as mc and vim with aptitude(8) for beginners to get started by the following. # aptitude update ... # aptitude install mc vim sudo ... If you already had these packages installed, no new packages are installed. package popcon size description DUMMY DUMMY DUMMY A text-mode full-screen file manager DUMMY DUMMY DUMMY A program to allow limited root privileges to users DUMMY DUMMY DUMMY Unix text editor Vi IMproved, a programmers text editor (standard version) DUMMY DUMMY DUMMY Unix text editor Vi IMproved, a programmers text editor (compact version) DUMMY DUMMY DUMMY GNU project Emacs, the Lisp based extensible text editor (version 22) DUMMY DUMMY DUMMY GNU project Emacs, the Lisp based extensible text editor (version 23) DUMMY DUMMY DUMMY Text-mode WWW browsers DUMMY DUMMY DUMMY The Unix style cut-and-paste on the text console (daemon) It may be a good idea to read some informative documentations. package popcon size description DUMMY DUMMY DUMMY Debian Project documentation, (Debian FAQ) and other documents DUMMY DUMMY DUMMY Debian Policy Manual and related documents DUMMY DUMMY DUMMY Guidelines and information for Debian developers DUMMY DUMMY DUMMY Debian New Maintainers' Guide DUMMY DUMMY DUMMY History of the Debian Project DUMMY DUMMY DUMMY Debian FAQ DUMMY DUMMY DUMMY Linux HOWTOs and FAQ (text) DUMMY DUMMY DUMMY Linux HOWTOs and FAQ (html) DUMMY DUMMY DUMMY The Linux System Administrators' Guide DUMMY DUMMY DUMMY Linux: Rute User's Tutorial and Exposition (non-free) You can install some of these packages by the following. # aptitude install package_name

If you do not want to use your main user account for the following training activities, you can create a training user account, e.g. fish by the following. # adduser fish Answer all questions. This creates a new account named as fish. After your practice, you can remove this user account and its home directory by the following. # deluser --remove-home fish

For the typical single user workstation such as the desktop Debian system on the laptop PC, it is common to deploy simple configuration of sudo(8) as follows to let the non-privileged user, e.g. penguin, to gain administrative privilege just with his user password but without the root password. # echo "penguin ALL=(ALL) ALL" >> /etc/sudoers This trick should only be used for the single user workstation which you administer and where you are the only user. Do not set up accounts of regular users on multiuser workstation like this because it would be very bad for system security. The password and the account of the penguin in the above example requires as much protection as the root password and the root account. Administrative privilege in this context belongs to someone authorized to perform the system administration task on the workstation. Never give some manager in the Admin department of your company or your boss such privilege unless they are authorized and capable. For providing access privilege to limited devices and limited files, you should consider to use group to provide limited access instead of using the root privilege via sudo(8). With more thoughtful and careful configuration, sudo(8) can grant limited administrative privileges to other users on a shared system without sharing the root password. This can help with accountability with hosts with multiple administrators so you can tell who did what. On the other hand, you might not want anyone else to have such privileges.

Now you are ready to play with the Debian system without risks as long as you use the non-privileged user account. This is because the Debian system is, even after the default installation, configured with proper file permissions which prevent non-privileged users from damaging the system. Of course, there may still be some holes which can be exploited but those who worry about these issues should not be reading this section but should be reading Securing Debian Manual. We learn the Debian system as a Unix-like system with the following. (basic concept) (survival method) (basic method) (shell mechanism) (text processing method)

In GNU/Linux and other Unix-like operating systems, files are organized into directories. All files and directories are arranged in one big tree rooted at "/". It's called a tree because if you draw the filesystem, it looks like a tree but it is upside down. These files and directories can be spread out over several devices. mount(8) serves to attach the filesystem found on some device to the big file tree. Conversely, umount(8) detaches it again. On recent Linux kernels, mount(8) with some options can bind part of a file tree somewhere else or can mount filesystem as shared, private, slave, or unbindable. Supported mount options for each filesystem are available in "/share/doc/linux-doc-2.6.*/Documentation/filesystems/". Directories on Unix systems are called folders on some other systems. Please also note that there is no concept for drive such as "A:" on any Unix system. There is one filesystem, and everything is included. This is a huge advantage compared to Windows.

Here are some Unix file basics. Filenames are case sensitive. That is, "MYFILE" and "MyFile" are different files. The root directory means root of the filesystem referred as simply "/". Don't confuse this with the home directory for the root user: "/root". Every directory has a name which can contain any letters or symbols except "/". The root directory is an exception; its name is "/" (pronounced "slash" or "the root directory") and it cannot be renamed. Each file or directory is designated by a fully-qualified filename, absolute filename, or path, giving the sequence of directories which must be passed through to reach it. The three terms are synonymous. All fully-qualified filenames begin with the "/" directory, and there's a "/" between each directory or file in the filename. The first "/" is the top level directory, and the other "/"'s separate successive subdirectories, until we reach the last entry which is the name of the actual file. The words used here can be confusing. Take the following fully-qualified filename as an example: "/usr/share/keytables/us.map.gz". However, people also refers to its basename "us.map.gz" alone as a filename. The root directory has a number of branches, such as "/etc/" and "/usr/". These subdirectories in turn branch into still more subdirectories, such as "/etc/init.d/" and "/usr/local/". The whole thing viewed collectively is called the directory tree. You can think of an absolute filename as a route from the base of the tree ("/") to the end of some branch (a file). You also hear people talk about the directory tree as if it were a family tree: thus subdirectories have parents, and a path shows the complete ancestry of a file. There are also relative paths that begin somewhere other than the root directory. You should remember that the directory "../" refers to the parent directory. This terminology also applies to other directory like structures, such as hierarchical data structures. There's no special directory path name component that corresponds to a physical device, such as your hard disk. This differs from RT-11, CP/M, OpenVMS, MS-DOS, AmigaOS, and Microsoft Windows, where the path contains a device name such as "C:\". (However, directory entries do exist that refer to physical devices as a part of the normal filesystem. See .) While you can use almost any letters or symbols in a file name, in practice it is a bad idea to do so. It is better to avoid any characters that often have special meanings on the command line, including spaces, tabs, newlines, and other special characters: { } ( ) [ ] ' ` " \ / > < | ; ! # & ^ * % @ $ . If you want to separate words in a name, good choices are the period, hyphen, and underscore. You could also capitalize each word, "LikeThis". Experienced Linux users tend to avoid spaces in filenames. The word "root" can mean either "root user" or "root directory". The context of their usage should make it clear. The word path is used not only for fully-qualified filename as above but also for the command search path. The intended meaning is usually clear from the context. The detailed best practices for the file hierarchy are described in the Filesystem Hierarchy Standard ("/usr/share/doc/debian-policy/fhs/fhs-2.3.txt.gz" and hier(7)). You should remember the following facts as the starter. directory usage of the directory / the root directory /etc/ system wide configuration files /var/log/ system log files /home/ all the home directories for all non-privileged users

Following the Unix tradition, the Debian GNU/Linux system provides the filesystem under which physical data on hard disks and other storage devices reside, and the interaction with the hardware devices such as console screens and remote serial consoles are represented in an unified manner under "/dev/". Each file, directory, named pipe (a way two programs can share data), or physical device on a Debian GNU/Linux system has a data structure called an inode which describes its associated attributes such as the user who owns it (owner), the group that it belongs to, the time last accessed, etc. If you are really interested, see "/usr/include/linux/fs.h" for the exact definition of "struct inode" in the Debian GNU/Linux system. The idea of representing just about everything in the filesystem was a Unix innovation, and modern Linux kernels have developed this idea ever further. Now, even information about processes running in the computer can be found in the filesystem. This abstract and unified representation of physical entities and internal processes is very powerful since this allows us to use the same command for the same kind of operation on many totally different devices. It is even possible to change the way the kernel works by writing data to special files that are linked to running processes. If you need to identify the correspondence between the file tree and the physical entity, execute mount(8) with no arguments.

Filesystem permissions of Unix-like system are defined for three categories of affected users. The user who owns the file (u) Other users in the group which the file belongs to (g) All other users (o) also referred to as "world" and "everyone" For the file, each corresponding permission allows following actions. The read (r) permission allows owner to examine contents of the file. The write (w) permission allows owner to modify the file. The execute (x) permission allows owner to run the file as a command. For the directory, each corresponding permission allows following actions. The read (r) permission allows owner to list contents of the directory. The write (w) permission allows owner to add or remove files in the directory. The execute (x) permission allows owner to access files in the directory. Here, the execute permission on a directory means not only to allow reading of files in that directory but also to allow viewing their attributes, such as the size and the modification time. ls(1) is used to display permission information (and more) for files and directories. When it is invoked with the "-l" option, it displays the following information in the order given. Type of file (first character) Access permission of the file (nine characters, consisting of three characters each for user, group, and other in this order) Number of hard links to the file Name of the user who owns the file Name of the group which the file belongs to Size of the file in characters (bytes) Date and time of the file (mtime) Name of the file character meaning - normal file d directory l symlink c character device node b block device node p named pipe s socket chown(1) is used from the root account to change the owner of the file. chgrp(1) is used from the file's owner or root account to change the group of the file. chmod(1) is used from the file's owner or root account to change file and directory access permissions. Basic syntax to manipulate a foo file is the following. # chown <newowner> foo # chgrp <newgroup> foo # chmod [ugoa][+-=][rwxXst][,...] foo For example, you can make a directory tree to be owned by a user foo and shared by a group bar by the following. # cd /some/location/ # chown -R foo:bar . # chmod -R ug+rwX,o=rX . There are three more special permission bits. The set user ID bit (s or S instead of user's x) The set group ID bit (s or S instead of group's x) The sticky bit (t or T instead of other's x) Here the output of "ls -l" for these bits is capitalized if execution bits hidden by these outputs are unset. Setting set user ID on an executable file allows a user to execute the executable file with the owner ID of the file (for example root). Similarly, setting set group ID on an executable file allows a user to execute the executable file with the group ID of the file (for example root). Because these settings can cause security risks, enabling them requires extra caution. Setting set group ID on a directory enables the BSD-like file creation scheme where all files created in the directory belong to the group of the directory. Setting the sticky bit on a directory prevents a file in the directory from being removed by a user who is not the owner of the file. In order to secure contents of a file in world-writable directories such as "/tmp" or in group-writable directories, one must not only reset the write permission for the file but also set the sticky bit on the directory. Otherwise, the file can be removed and a new file can be created with the same name by any user who has write access to the directory. Here are a few interesting examples of file permissions. $ ls -l /etc/passwd /etc/shadow /dev/ppp /usr/sbin/exim4 crw------- 1 root root 108, 0 2007-04-29 07:00 /dev/ppp -rw-r--r-- 1 root root 1427 2007-04-16 00:19 /etc/passwd -rw-r----- 1 root shadow 943 2007-04-16 00:19 /etc/shadow -rwsr-xr-x 1 root root 700056 2007-04-22 05:29 /usr/sbin/exim4 $ ls -ld /tmp /var/tmp /usr/local /var/mail /usr/src drwxrwxrwt 10 root root 4096 2007-04-29 07:59 /tmp drwxrwsr-x 10 root staff 4096 2007-03-24 18:48 /usr/local drwxrwsr-x 4 root src 4096 2007-04-27 00:31 /usr/src drwxrwsr-x 2 root mail 4096 2007-03-28 23:33 /var/mail drwxrwxrwt 2 root root 4096 2007-04-29 07:11 /var/tmp There is an alternative numeric mode to describe file permissions with chmod(1). This numeric mode uses 3 to 4 digit wide octal (radix=8) numbers. digit meaning 1st optional digit sum of set user ID (=4), set group ID (=2), and sticky bit (=1) 2nd digit sum of read (=4), write (=2), and execute (=1) permissions for user 3rd digit ditto for group 4th digit ditto for other This sounds complicated but it is actually quite simple. If you look at the first few (2-10) columns from "ls -l" command output and read it as a binary (radix=2) representation of file permissions ("-" being "0" and "rwx" being "1"), the last 3 digit of the numeric mode value should make sense as an octal (radix=8) representation of file permissions to you. For example, try the following $ touch foo bar $ chmod u=rw,go=r foo $ chmod 644 bar $ ls -l foo bar -rw-r--r-- 1 penguin penguin 17 2007-04-29 08:22 bar -rw-r--r-- 1 penguin penguin 12 2007-04-29 08:22 foo If you need to access information displayed by "ls -l" in shell script, you should use pertinent commands such as test(1), stat(1) and readlink(1). The shell builtin such as "[" or "test" may be used too.

What permissions are applied to a newly created file or directory is restricted by the umask shell builtin command. See dash(1), bash(1), and builtins(7). (file permissions) = (requested file permissions) & ~(umask value) umask file permissions created directory permissions created usage 0022 -rw-r--r-- -rwxr-xr-x writable only by the user 0002 -rw-rw-r-- -rwxrwxr-x writable by the group The Debian system uses a user private group (UPG) scheme as its default. A UPG is created whenever a new user is added to the system. A UPG has the same name as the user for which it was created and that user is the only member of the UPG. UPG scheme makes it is safe to set umask to 0002 since every user has their own private group. (In some Unix variants, it is quite common to setup all normal users belonging to a single users group and is good idea to set umask to 0022 for security in such cases.)

In order to make group permissions to be applied to a particular user, that user needs to be made a member of the group using "sudo vigr". Alternatively, you may dynamically add users to groups during the authentication process by adding "auth optional pam_group.so" line to "/etc/pam.d/common-auth" and setting "/etc/security/group.conf". (See .) The hardware devices are just another kind of file on the Debian system. If you have problems accessing devices such as CD-ROM and USB memory stick from a user account, you should make that user a member of the relevant group. Some notable system-provided groups allow their members to access particular files and devices without root privilege. group description for accessible files and devices dialout full and direct access to serial ports ("/dev/ttyS[0-3]") dip limited access to serial ports for Dialup IP connection to trusted peers cdrom CD-ROM, DVD+/-RW drives audio audio device video video device scanner scanner(s) adm system monitoring logs staff some directories for junior administrative work: "/usr/local", "/home" You need to belong to the dialout group to reconfigure modem, dial anywhere, etc. But if root creates pre-defined configuration files for trusted peers in "/etc/ppp/peers/", you only need to belong to the dip group to create Dialup IP connection to those trusted peers using pppd(8), pon(1), and poff(1) commands. Some notable system-provided groups allow their members to execute particular commands without root privilege. group accessible commands sudo execute sudo without their password lpadmin execute commands to add, modify, and remove printers from printer databases plugdev execute pmount(1) for removable devices such as USB memories For the full listing of the system provided users and groups, see the recent version of the "Users and Groups" document in "/usr/share/doc/base-passwd/users-and-groups.html" provided by the base-passwd package. See passwd(5), group(5), shadow(5), newgrp(1), vipw(8), vigr(8), and pam_group(8) for management commands of the user and group system.

There are three types of timestamps for a GNU/Linux file. type meaning mtime the file modification time (ls -l) ctime the file status change time (ls -lc) atime the last file access time (ls -lu) ctime is not file creation time. Overwriting a file changes all of the mtime, ctime, and atime attributes of the file. Changing ownership or permission of a file changes the ctime and atime attributes of the file. Reading a file changes the atime of the file. Even simply reading a file on the Debian system normally causes a file write operation to update atime information in the inode. Mounting a filesystem with "noatime" or "relatime" option makes the system skip this operation and results in faster file access for the read. This is often recommended for laptops, because it reduces hard drive activity and saves power. See mount(8). Use touch(1) command to change timestamps of existing files. For timestamps, the ls command outputs different strings under the modern English locale ("en_US.UTF-8") from under the old one ("C"). $ LANG=en_US.UTF-8 ls -l foo -rw-r--r-- 1 penguin penguin 3 2008-03-05 00:47 foo $ LANG=C ls -l foo -rw-r--r-- 1 penguin penguin 3 Mar 5 00:47 foo See to customize "ls -l" output.

There are two methods of associating a file "foo" with a different filename "bar". Hard link Duplicate name for an existing file "ln foo bar" Symbolic link or symlink Special file that points to another file by name "ln -s foo bar" See the following example for changes in link counts and the subtle differences in the result of the rm command. $ echo "Original Content" > foo $ ls -li foo 2398521 -rw-r--r-- 1 penguin penguin 17 2007-04-29 08:15 foo $ ln foo bar # hard link $ ln -s foo baz # symlink $ ls -li foo bar baz 2398521 -rw-r--r-- 2 penguin penguin 17 2007-04-29 08:15 bar 2398538 lrwxrwxrwx 1 penguin penguin 3 2007-04-29 08:16 baz -> foo 2398521 -rw-r--r-- 2 penguin penguin 17 2007-04-29 08:15 foo $ rm foo $ echo "New Content" > foo $ ls -li foo bar baz 2398521 -rw-r--r-- 1 penguin penguin 17 2007-04-29 08:15 bar 2398538 lrwxrwxrwx 1 penguin penguin 3 2007-04-29 08:16 baz -> foo 2398540 -rw-r--r-- 1 penguin penguin 12 2007-04-29 08:17 foo $ cat bar Original Content $ cat baz New Content The hardlink can be made within the same filesystem and shares the same inode number which the "-i" option with ls(1) reveals. The symlink always has nominal file access permissions of "rwxrwxrwx", as shown in the above example, with the effective access permissions dictated by permissions of the file that it points to. It is generally good idea not to create complicated symbolic links or hardlinks at all unless you have a very good reason. It may cause nightmares where the logical combination of the symbolic links results in loops in the filesystem. It is generally preferable to use symbolic links rather than hardlinks unless you have a good reason for using a hardlink. The "." directory links to the directory that it appears in, thus the link count of any new directory starts at 2. The ".." directory links to the parent directory, thus the link count of the directory increases with the addition of new subdirectories. If you are just moving to Linux from Windows, it soon becomes clear how well-designed the filename linking of Unix is, compared with the nearest Windows equivalent of "shortcuts". Because it is implemented in the filesystem, applications can't see any difference between a linked file and the original. In the case of hardlinks, there really is no difference.

A named pipe is a file that acts like a pipe. You put something into the file, and it comes out the other end. Thus it's called a FIFO, or First-In-First-Out: the first thing you put in the pipe is the first thing to come out the other end. If you write to a named pipe, the process which is writing to the pipe doesn't terminate until the information being written is read from the pipe. If you read from a named pipe, the reading process waits until there is nothing to read before terminating. The size of the pipe is always zero --- it does not store data, it just links two processes like the shell "|". However, since this pipe has a name, the two processes don't have to be on the same command line or even be run by the same user. Pipes were a very influential innovation of Unix. For example, try the following $ cd; mkfifo mypipe $ echo "hello" >mypipe & # put into background [1] 8022 $ ls -l mypipe prw-r--r-- 1 penguin penguin 0 2007-04-29 08:25 mypipe $ cat mypipe hello [1]+ Done echo "hello" >mypipe $ ls mypipe mypipe $ rm mypipe

Sockets are used extensively by all the Internet communication, databases, and the operating system itself. It is similar to the named pipe (FIFO) and allows processes to exchange information even between different computers. For the socket, those processes do not need to be running at the same time nor to be running as the children of the same ancestor process. This is the endpoint for the inter process communication (IPC). The exchange of information may occur over the network between different hosts. The two most common ones are the Internet socket and the Unix domain socket. "netstat -an" provides a very useful overview of sockets that are open on a given system.

Device files refer to physical or virtual devices on your system, such as your hard disk, video card, screen, or keyboard. An example of a virtual device is the console, represented by "/dev/console". There are 2 types of device files. Character device Accessed one character at a time 1 character = 1 byte E.g. keyboard device, serial port, … Block device accessed in larger units called blocks 1 block > 1 byte E.g. hard disk, … You can read and write device files, though the file may well contain binary data which may be an incomprehensible-to-humans gibberish. Writing data directly to these files is sometimes useful for the troubleshooting of hardware connections. For example, you can dump a text file to the printer device "/dev/lp0" or send modem commands to the appropriate serial port "/dev/ttyS0". But, unless this is done carefully, it may cause a major disaster. So be cautious. For the normal access to a printer, use lp(1). The device node number are displayed by executing ls(1) as the following. $ ls -l /dev/hda /dev/ttyS0 /dev/zero brw-rw---- 1 root cdrom 3, 0 2007-04-29 07:00 /dev/hda crw-rw---- 1 root dialout 4, 64 2007-04-29 07:00 /dev/ttyS0 crw-rw-rw- 1 root root 1, 5 2007-04-29 07:00 /dev/zero "/dev/hda" has the major device number 3 and the minor device number 0. This is read/write accessible by the user who belongs to cdrom group. "/dev/ttyS0" has the major device number 4 and the minor device number 64. This is read/write accessible by the user who belongs to dialout group. "/dev/zero" has the major device number 1 and the minor device number 5. This is read/write accessible by anyone. In the Linux 2.6 system, the filesystem under "/dev/" is automatically populated by the udev(7) mechanism.

There are some special device files. device file action description of response /dev/null read return "end-of-file (EOF) character" /dev/null write return nothing (a bottomless data dump pit) /dev/zero read return "the \0 (NUL) character" (not the same as the number zero ASCII) /dev/random read return random characters from a true random number generator, delivering real entropy (slow) /dev/urandom read return random characters from a cryptographically secure pseudorandom number generator /dev/full write return the disk-full (ENOSPC) error These are frequently used in conjunction with the shell redirection (see ).

The procfs and sysfs mounted on "/proc" and "/sys" are the pseudo-filesystem and expose internal data structures of the kernel to the userspace. In other word, these entries are virtual, meaning that they act as a convenient window into the operation of the operating system. The directory "/proc" contains (among other things) one subdirectory for each process running on the system, which is named after the process ID (PID). System utilities that access process information, such as ps(1), get their information from this directory structure. The directories under "/proc/sys/" contain interface to change certain kernel parameters at run time. (You may do the same through specialized sysctl(8) command or its preload/configuration file "/etc/sysctrl.conf".) The Linux kernel may complain "Too many open files". You can fix this by increasing "file-max" value to a larger value from the root shell, e.g., "echo "65536" > /proc/sys/fs/file-max" (This was needed on older kernels). People frequently panic when they notice one file in particular - "/proc/kcore" - which is generally huge. This is (more or less) a copy of the content of your computer's memory. It's used to debug the kernel. It is a virtual file that points to computer memory, so don't worry about its size. The directory under "/sys" contains exported kernel data structures, their attributes, and their linkages between them. It also contains interface to change certain kernel parameters at run time. See "proc.txt(.gz)", "sysfs.txt(.gz)" and other related documents in the Linux kernel documentation ("/usr/share/doc/linux-doc-2.6.*/Documentation/filesystems/*") provided by the linux-doc-2.6.* package.

Midnight Commander (MC) is a GNU "Swiss army knife" for the Linux console and other terminal environments. This gives newbie a menu driven console experience which is much easier to learn than standard Unix commands. You may need to install the Midnight Commander package which is titled "mc" by the following. $ sudo aptitude install mc Use the mc(1) command to explore the Debian system. This is the best way to learn. Please explore few interesting locations just using the cursor keys and Enter key. "/etc" and its subdirectories "/var/log" and its subdirectories "/usr/share/doc" and its subdirectories "/sbin" and "/bin"

In order to make MC to change working directory upon exit and cd to the directory, I suggest to modify "~/.bashrc" to include a script provided by the mc package. . /usr/share/mc/bin/mc.sh See mc(1) (under the "-P" option) for the reason. (If you do not understand what exactly I am talking here, you can do this later.)

MC can be started by the following. $ mc MC takes care of all file operations through its menu, requiring minimal user effort. Just press F1 to get the help screen. You can play with MC just by pressing cursor-keys and function-keys. In some consoles such as gnome-terminal(1), key strokes of function-keys may be stolen by the console program. You can disable these features by "Edit" → "Keyboard Shortcuts" for gnome-terminal. If you encounter character encoding problem which displays garbage characters, adding "-a" to MC's command line may help prevent problems. If this doesn't clear up your display problems with MC, see .

The default is two directory panels containing file lists. Another useful mode is to set the right window to "information" to see file access privilege information, etc. Following are some essential keystrokes. With the gpm(8) daemon running, one can use a mouse on Linux character consoles, too. (Make sure to press the shift-key to obtain the normal behavior of cut and paste in MC.) key key binding F1 help menu F3 internal file viewer F4 internal editor F9 activate pull down menu F10 exit Midnight Commander Tab move between two windows Insert or Ctrl-T mark file for a multiple-file operation such as copy Del delete file (be careful---set MC to safe delete mode) Cursor keys self-explanatory

cd command changes the directory shown on the selected screen. Ctrl-Enter or Alt-Enter copies a filename to the command line. Use this with cp(1) and mv(1) commands together with command-line editing. Alt-Tab shows shell filename expansion choices. One can specify the starting directory for both windows as arguments to MC; for example, "mc /etc /root". Esc + n-key → Fn (i.e., Esc + 1 → F1, etc.; Esc + 0 → F10) Pressing Esc before the key has the same effect as pressing the Alt and the key together.; i.e., type Esc + c for Alt-C. Esc is called meta-key and sometimes noted as "M-".

The internal editor has an interesting cut-and-paste scheme. Pressing F3 marks the start of a selection, a second F3 marks the end of selection and highlights the selection. Then you can move your cursor. If you press F6, the selected area is moved to the cursor location. If you press F5, the selected area is copied and inserted at the cursor location. F2 saves the file. F10 gets you out. Most cursor keys work intuitively. This editor can be directly started on a file using one of the following commands. $ mc -e filename_to_edit $ mcedit filename_to_edit This is not a multi-window editor, but one can use multiple Linux consoles to achieve the same effect. To copy between windows, use Alt-F<n> keys to switch virtual consoles and use "File→Insert file" or "File→Copy to file" to move a portion of a file to another file. This internal editor can be replaced with any external editor of choice. Also, many programs use the environment variables "$EDITOR" or "$VISUAL" to decide which editor to use. If you are uncomfortable with vim(1) or nano(1) initially, you may set these to "mcedit" by adding the following lines to "~/.bashrc". export EDITOR=mcedit export VISUAL=mcedit I do recommend setting these to "vim" if possible. If you are uncomfortable with vim(1), you can keep using mcedit(1) for most system maintenance tasks.

MC is a very smart viewer. This is a great tool for searching words in documents. I always use this for files in the "/usr/share/doc" directory. This is the fastest way to browse through masses of Linux information. This viewer can be directly started using one of the following commands. $ mc -v path/to/filename_to_view $ mcview path/to/filename_to_view

Press Enter on a file, and the appropriate program handles the content of the file (see ). This is a very convenient MC feature. file type reaction to enter key executable file execute command man file pipe content to viewer software html file pipe content to web browser "*.tar.gz" and "*.deb" file browse its contents as if subdirectory In order to allow these viewer and virtual file features to function, viewable files should not be set as executable. Change their status using chmod(1) or via the MC file menu.

MC can be used to access files over the Internet using FTP. Go to the menu by pressing F9, then type "p" to activate the FTP virtual filesystem. Enter a URL in the form "username:passwd@hostname.domainname", which retrieves a remote directory that appears like a local one. Try "[http.us.debian.org/debian]" as the URL and browse the Debian archive.

Although MC enables you to do almost everything, it is very important for you to learn how to use the command line tools invoked from the shell prompt and become familiar with the Unix-like work environment.

You can select your login shell with chsh(1). package popcon size POSIX shell description DUMMY DUMMY DUMMY Yes Bash: the GNU Bourne Again SHell (de facto standard) DUMMY DUMMY DUMMY No TENEX C Shell: an enhanced version of Berkeley csh DUMMY DUMMY DUMMY Yes Debian Almquist Shell, good for shell script DUMMY DUMMY DUMMY Yes Z shell: the standard shell with many enhancements DUMMY DUMMY DUMMY Yes public domain version of the Korn shell DUMMY DUMMY DUMMY No OpenBSD C Shell, a version of Berkeley csh DUMMY DUMMY DUMMY Yes Stand-alone shell with builtin commands (Not meant for standard "/bin/sh") DUMMY DUMMY DUMMY Yes the real, AT&T version of the Korn shell DUMMY DUMMY DUMMY No implementation of the AT&T Plan 9 rc shell DUMMY DUMMY DUMMY Yes Policy-compliant Ordinary SHell (pdksh derivative) In this tutorial chapter, the interactive shell always means bash.

You can customize bash(1) behavior by "~/.bashrc". For example, try the following. # CD upon exiting MC . /usr/share/mc/bin/mc.sh # set CDPATH to good one CDPATH=.:/usr/share/doc:~/Desktop/src:~/Desktop:~ export CDPATH PATH="${PATH}":/usr/sbin:/sbin # set PATH so it includes user's private bin if it exists if [ -d ~/bin ] ; then PATH=~/bin:"${PATH}" fi export PATH EDITOR=vim export EDITOR You can find more bash customization tips, such as , in .

In the Unix-like environment, there are few key strokes which have special meanings. Please note that on a normal Linux character console, only the left-hand Ctrl and Alt keys work as expected. Here are few notable key strokes to remember. key description of key binding Ctrl-U erase line before cursor Ctrl-H erase a character before cursor Ctrl-D terminate input (exit shell if you are using shell) Ctrl-C terminate a running program Ctrl-Z temporarily stop program by moving it to the background job Ctrl-S halt output to screen Ctrl-Q reactivate output to screen Ctrl-Alt-Del reboot/halt the system, see inittab(5) Left-Alt-key (optionally, Windows-key) meta-key for Emacs and the similar UI Up-arrow start command history search under bash Ctrl-R start incremental command history search under bash Tab complete input of the filename to the command line under bash Ctrl-V Tab input Tab without expansion to the command line under bash The terminal feature of Ctrl-S can be disabled using stty(1).

Unix style mouse operations are based on the 3 button mouse system. action response Left-click-and-drag mouse select and copy to the clipboard Left-click select the start of selection Right-click select the end of selection and copy to the clipboard Middle-click paste clipboard at the cursor The center wheel on the modern wheel mouse is considered middle mouse button and can be used for middle-click. Clicking left and right mouse buttons together serves as the middle-click under the 2 button mouse system situation. In order to use a mouse in Linux character consoles, you need to have gpm(8) running as daemon.

less(1) is the enhanced pager (file content browser). Hit "h" for help. It can do much more than more(1) and can be supercharged by executing "eval $(lesspipe)" or "eval $(lessfile)" in the shell startup script. See more in "/usr/share/doc/lessf/LESSOPEN". The "-R" option allows raw character output and enables ANSI color escape sequences. See less(1).

You should become proficient in one of variants of Vim or Emacs programs which are popular in the Unix-like system. I think getting used to Vim commands is the right thing to do, since Vi-editor is always there in the Linux/Unix world. (Actually, original vi or new nvi are programs you find everywhere. I chose Vim instead for newbie since it offers you help through F1 key while it is similar enough and more powerful.) If you chose either Emacs or XEmacs instead as your choice of the editor, that is another good choice indeed, particularly for programming. Emacs has a plethora of other features as well, including functioning as a newsreader, directory editor, mail program, etc. When used for programming or editing shell scripts, it intelligently recognizes the format of what you are working on, and tries to provide assistance. Some people maintain that the only program they need on Linux is Emacs. Ten minutes learning Emacs now can save hours later. Having the GNU Emacs manual for reference when learning Emacs is highly recommended. All these programs usually come with tutoring program for you to learn them by practice. Start Vim by typing "vim" and press F1-key. You should at least read the first 35 lines. Then do the online training course by moving cursor to "|tutor|" and pressing Ctrl-]. Good editors, such as Vim and Emacs, can be used to handle UTF-8 and other exotic encoding texts correctly with proper option in the x-terminal-emulator on X under UTF-8 locale with proper font settings. Please refer to their documentation on multibyte text.

Debian comes with a number of different editors. We recommend to install the vim package, as mentioned above. Debian provides unified access to the system default editor via command "/usr/bin/editor" so other programs (e.g., reportbug(1)) can invoke it. You can change it by the following. $ sudo update-alternatives --config editor The choice "/usr/bin/vim.basic" over "/usr/bin/vim.tiny" is my recommendation for newbies since it supports syntax highlighting. Many programs use the environment variables "$EDITOR" or "$VISUAL" to decide which editor to use (see and ). For the consistency on Debian system, set these to "/usr/bin/editor". (Historically, "$EDITOR" was "ed" and "$VISUAL" was "vi".)

You can customize vim(1) behavior by "~/.vimrc". For example, try the following " ------------------------------- " Local configuration " set nocompatible set nopaste set pastetoggle=<f2> syn on if $USER == "root" set nomodeline set noswapfile else set modeline set swapfile endif " filler to avoid the line above being recognized as a modeline " filler " filler

The output of the shell command may roll off your screen and may be lost forever. It is good practice to log shell activities into the file for you to review them later. This kind of record is essential when you perform any system administration tasks. The basic method of recording the shell activity is to run it under script(1). For example, try the following $ script Script started, file is typescript Do whatever shell commands under script. Press Ctrl-D to exit script. $ vim typescript See .

Let's learn basic Unix commands. Here I use "Unix" in its generic sense. Any Unix clone OSs usually offer equivalent commands. The Debian system is no exception. Do not worry if some commands do not work as you wish now. If alias is used in the shell, its corresponding command outputs are different. These examples are not meant to be executed in this order. Try all following commands from the non-privileged user account. command description pwd display name of current/working directory whoami display current user name id display current user identity (name, uid, gid, and associated groups) file <foo> display a type of file for the file "<foo>" type -p <commandname> display a file location of command "<commandname>" which <commandname> , , type <commandname> display information on command "<commandname>" apropos <key-word> find commands related to "<key-word>" man -k <key-word> , , whatis <commandname> display one line explanation on command "<commandname>" man -a <commandname> display explanation on command "<commandname>" (Unix style) info <commandname> display rather long explanation on command "<commandname>" (GNU style) ls list contents of directory (non-dot files and directories) ls -a list contents of directory (all files and directories) ls -A list contents of directory (almost all files and directories, i.e., skip ".." and ".") ls -la list all contents of directory with detail information ls -lai list all contents of directory with inode number and detail information ls -d list all directories under the current directory tree display file tree contents lsof <foo> list open status of file "<foo>" lsof -p <pid> list files opened by the process ID: "<pid>" mkdir <foo> make a new directory "<foo>" in the current directory rmdir <foo> remove a directory "<foo>" in the current directory cd <foo> change directory to the directory "<foo>" in the current directory or in the directory listed in the variable "$CDPATH" cd / change directory to the root directory cd change directory to the current user's home directory cd /<foo> change directory to the absolute path directory "/<foo>" cd .. change directory to the parent directory cd ~<foo> change directory to the home directory of the user "<foo>" cd - change directory to the previous directory </etc/motd pager display contents of "/etc/motd" using the default pager touch <junkfile> create a empty file "<junkfile>" cp <foo> <bar> copy a existing file "<foo>" to a new file "<bar>" rm <junkfile> remove a file "<junkfile>" mv <foo> <bar> rename an existing file "<foo>" to a new name "<bar>" ("<bar>" must not exist) mv <foo> <bar> move an existing file "<foo>" to a new location "<bar>/<foo>" (the directory "<bar>" must exist) mv <foo> <bar>/<baz> move an existing file "<foo>" to a new location with a new name "<bar>/<baz>" (the directory "<bar>" must exist but the directory "<bar>/<baz>" must not exist) chmod 600 <foo> make an existing file "<foo>" to be non-readable and non-writable by the other people (non-executable for all) chmod 644 <foo> make an existing file "<foo>" to be readable but non-writable by the other people (non-executable for all) chmod 755 <foo> make an existing file "<foo>" to be readable but non-writable by the other people (executable for all) find . -name <pattern> find matching filenames using shell "<pattern>" (slower) locate -d . <pattern> find matching filenames using shell "<pattern>" (quicker using regularly generated database) grep -e "<pattern>" *.html find a "<pattern>" in all files ending with ".html" in current directory and display them all top display process information using full screen, type "q" to quit ps aux | pager display information on all the running processes using BSD style output ps -ef | pager display information on all the running processes using Unix system-V style output ps aux | grep -e "[e]xim4*" display all processes running "exim" and "exim4" ps axf | pager display information on all the running processes with ASCII art output kill <1234> kill a process identified by the process ID: "<1234>" gzip <foo> compress "<foo>" to create "<foo>.gz" using the Lempel-Ziv coding (LZ77) gunzip <foo>.gz decompress "<foo>.gz" to create "<foo>" bzip2 <foo> compress "<foo>" to create "<foo>.bz2" using the Burrows-Wheeler block sorting text compression algorithm, and Huffman coding (better compression than gzip) bunzip2 <foo>.bz2 decompress "<foo>.bz2" to create "<foo>" tar -xvf <foo>.tar extract files from "<foo>.tar" archive tar -xvzf <foo>.tar.gz extract files from gzipped "<foo>.tar.gz" archive tar -xvf -j <foo>.tar.bz2 extract files from "<foo>.tar.bz2" archive tar -cvf <foo>.tar <bar>/ archive contents of folder "<bar>/" in "<foo>.tar" archive tar -cvzf <foo>.tar.gz <bar>/ archive contents of folder "<bar>/" in compressed "<foo>.tar.gz" archive tar -cvjf <foo>.tar.bz2 <bar>/ archive contents of folder "<bar>/" in "<foo>.tar.bz2" archive zcat README.gz | pager display contents of compressed "README.gz" using the default pager zcat README.gz > foo create a file "foo" with the decompressed content of "README.gz" zcat README.gz >> foo append the decompressed content of "README.gz" to the end of the file "foo" (if it does not exist, create it first) Unix has a tradition to hide filenames which start with ".". They are traditionally files that contain configuration information and user preferences. For cd command, see builtins(7). The default pager of the bare bone Debian system is more(1) which cannot scroll back. By installing the less package using command line "aptitude install less", less(1) becomes default pager and you can scroll back with cursor keys. The "[" and "]" in the regular expression of the "ps aux | grep -e "[e]xim4*"" command above enable grep to avoid matching itself. The "4*" in the regular expression means 0 or more repeats of character "4" thus enables grep to match both "exim" and "exim4". Although "*" is used in the shell filename glob and the regular expression, their meanings are different. Learn the regular expression from grep(1). Please traverse directories and peek into the system using the above commands as training. If you have questions on any of console commands, please make sure to read the manual page. For example, try the following $ man man $ man bash $ man builtins $ man grep $ man ls The style of man pages may be a little hard to get used to, because they are rather terse, particularly the older, very traditional ones. But once you get used to it, you come to appreciate their succinctness. Please note that many Unix-like commands including ones from GNU and BSD display brief help information if you invoke them in one of the following ways (or without any arguments in some cases). $ <commandname> --help $ <commandname> -h

Now you have some feel on how to use the Debian system. Let's look deep into the mechanism of the command execution in the Debian system. Here, I have simplified reality for the newbie. See bash(1) for the exact explanation. A simple command is a sequence of components. Variable assignments (optional) Command name Arguments (optional) Redirections (optional: > , >> , < , << , etc.) Control operator (optional: && , || , <newline> , ; , & , ( , ) )

Values of some environment variables change the behavior of some Unix commands. Default values of environment variables are initially set by the PAM system and then some of them may be reset by some application programs. The display manager such as gdm resets environment variables. The shell in its start up codes resets environment variables in "~/bash_profile" and "~/.bashrc".

The full locale value given to "$LANG" variable consists of 3 parts: "xx_YY.ZZZZ". locale value meaning xx ISO 639 language codes (lower case) such as "en" YY ISO 3166 country codes (upper case) such as "US" ZZZZ codeset, always set to "UTF-8" For language codes and country codes, see pertinent description in the "info gettext". For the codeset on the modern Debian system, you should always set it to UTF-8 unless you specifically want to use the historic one with good reason and background knowledge. For fine details of the locale configuration, see . The "LANG=en_US" is not "LANG=C" nor "LANG=en_US.UTF-8". It is "LANG=en_US.ISO-8859-1" (see ). locale recommendation Language (area) en_US.UTF-8 English(USA) en_GB.UTF-8 English(Great_Britain) fr_FR.UTF-8 French(France) de_DE.UTF-8 German(Germany) it_IT.UTF-8 Italian(Italy) es_ES.UTF-8 Spanish(Spain) ca_ES.UTF-8 Catalan(Spain) sv_SE.UTF-8 Swedish(Sweden) pt_BR.UTF-8 Portuguese(Brazil) ru_RU.UTF-8 Russian(Russia) zh_CN.UTF-8 Chinese(P.R._of_China) zh_TW.UTF-8 Chinese(Taiwan_R.O.C.) ja_JP.UTF-8 Japanese(Japan) ko_KR.UTF-8 Korean(Republic_of_Korea) vi_VN.UTF-8 Vietnamese(Vietnam) Typical command execution uses a shell line sequence as the following. $ date Sun Jun 3 10:27:39 JST 2007 $ LANG=fr_FR.UTF-8 date dimanche 3 juin 2007, 10:27:33 (UTC+0900) Here, the program date(1) is executed with different values of the environment variable "$LANG". For the first command, "$LANG" is set to the system default locale value "en_US.UTF-8". For the second command, "$LANG" is set to the French UTF-8 locale value "fr_FR.UTF-8". Most command executions usually do not have preceding environment variable definition. For the above example, you can alternatively execute as the following. $ LANG=fr_FR.UTF-8 $ date dimanche 3 juin 2007, 10:27:33 (UTC+0900) As you can see here, the output of command is affected by the environment variable to produce French output. If you want the environment variable to be inherited to subprocesses (e.g., when calling shell script), you need to export it instead by the following. $ export LANG When filing a bug report, running and checking the command under "LANG=en_US.UTF-8" is good idea if you use non-English environment. See locale(5) and locale(7) for "$LANG" and related environment variables. I recommend you to configure the system environment just by the "$LANG" variable and to stay away from "$LC_*" variables unless it is absolutely needed.

When you type a command into the shell, the shell searches the command in the list of directories contained in the "$PATH" environment variable. The value of the "$PATH" environment variable is also called the shell's search path. In the default Debian installation, the "$PATH" environment variable of user accounts may not include "/sbin" and "/usr/sbin". For example, the ifconfig command needs to be issued with full path as "/sbin/ifconfig". (Similar ip command is located in "/bin".) You can change the "$PATH" environment variable of Bash shell by "~/.bash_profile" or "~/.bashrc" files.

Many commands stores user specific configuration in the home directory and changes their behavior by their contents. The home directory is identified by the environment variable "$HOME". value of "$HOME" program execution situation / program run by the init process (daemon) /root program run from the normal root shell /home/<normal_user> program run from the normal user shell /home/<normal_user> program run from the normal user GUI desktop menu /home/<normal_user> program run as root with "sudo program" /root program run as root with "sudo -H program" Shell expands "~/" to current user’s home directory, i.e., "$HOME/". Shell expands "~foo/" to foo's home directory, i.e., "/home/foo/".

Some commands take arguments. Arguments starting with "-" or "--" are called options and control the behavior of the command. $ date Mon Oct 27 23:02:09 CET 2003 $ date -R Mon, 27 Oct 2003 23:02:40 +0100 Here the command-line argument "-R" changes date(1) behavior to output RFC2822 compliant date string.

Often you want a command to work with a group of files without typing all of them. The filename expansion pattern using the shell glob, (sometimes referred as wildcards), facilitate this need. shell glob pattern description of match rule * filename (segment) not started with "." .* filename (segment) started with "." ? exactly one character […] exactly one character with any character enclosed in brackets [a-z] exactly one character with any character between "a" and "z" [^…] exactly one character other than any character enclosed in brackets (excluding "^") For example, try the following $ mkdir junk; cd junk; touch 1.txt 2.txt 3.c 4.h .5.txt ..6.txt $ echo *.txt 1.txt 2.txt $ echo * 1.txt 2.txt 3.c 4.h $ echo *.[hc] 3.c 4.h $ echo .* . .. .5.txt ..6.txt $ echo .*[^.]* .5.txt ..6.txt $ echo [^1-3]* 4.h $ cd ..; rm -rf junk See glob(7). Unlike normal filename expansion by the shell, the shell pattern "*" tested in find(1) with "-name" test etc., matches the initial "." of the filename. (New POSIX feature) BASH can be tweaked to change its glob behavior with its shopt builtin options such as "dotglob", "noglob", "nocaseglob", "nullglob", "nocaseglob", "extglob", etc. See bash(1).

Each command returns its exit status (variable: "$?") as the return value. command exit status numeric return value logical return value success zero, 0 TRUE error non-zero, -1 FALSE For example, try the following. $ [ 1 = 1 ] ; echo $? 0 $ [ 1 = 2 ] ; echo $? 1 Please note that, in the logical context for the shell, success is treated as the logical TRUE which has 0 (zero) as its value. This is somewhat non-intuitive and needs to be reminded here.

Let's try to remember following shell command idioms typed in one line as a part of shell command. command idiom description command & background execution of command in the subshell command1 | command2 pipe the standard output of command1 to the standard input of command2 (concurrent execution) command1 2>&1 | command2 pipe both standard output and standard error of command1 to the standard input of command2 (concurrent execution) command1 ; command2 execute command1 and command2 sequentially command1 && command2 execute command1; if successful, execute command2 sequentially (return success if both command1 and command2 are successful) command1 || command2 execute command1; if not successful, execute command2 sequentially (return success if command1 or command2 are successful) command > foo redirect standard output of command to a file foo (overwrite) command 2> foo redirect standard error of command to a file foo (overwrite) command >> foo redirect standard output of command to a file foo (append) command 2>> foo redirect standard error of command to a file foo (append) command > foo 2>&1 redirect both standard output and standard error of command to a file "foo" command < foo redirect standard input of command to a file foo command << delimiter redirect standard input of command to the following lines until "delimiter" is met (here document) command <<- delimiter redirect standard input of command to the following lines until "delimiter" is met (here document, the leading tab characters are stripped from input lines) The Debian system is a multi-tasking system. Background jobs allow users to run multiple programs in a single shell. The management of the background process involves the shell builtins: jobs, fg, bg, and kill. Please read sections of bash(1) under "SIGNALS", and "JOB CONTROL", and builtins(1). For example, try the following $ </etc/motd pager $ pager </etc/motd $ pager /etc/motd $ cat /etc/motd | pager Although all 4 examples of shell redirections display the same thing, the last example runs an extra cat command and wastes resources with no reason. The shell allows you to open files using the exec builtin with an arbitrary file descriptor. $ echo Hello >foo $ exec 3<foo 4>bar # open files $ cat <&3 >&4 # redirect stdin to 3, stdout to 4 $ exec 3<&- 4>&- # close files $ cat bar Hello Here, "n<&-" and "n>&-" mean to close the file descriptor "n". The file descriptor 0-2 are predefined. device description file descriptor stdin standard input 0 stdout standard output 1 stderr standard error 2

You can set an alias for the frequently used command. For example, try the following $ alias la='ls -la' Now, "la" works as a short hand for "ls -la" which lists all files in the long listing format. You can list any existing aliases by alias (see bash(1) under "SHELL BUILTIN COMMANDS"). $ alias ... alias la='ls -la' You can identity exact path or identity of the command by type (see bash(1) under "SHELL BUILTIN COMMANDS"). For example, try the following $ type ls ls is hashed (/bin/ls) $ type la la is aliased to ls -la $ type echo echo is a shell builtin $ type file file is /usr/bin/file Here ls was recently searched while "file" was not, thus "ls" is "hashed", i.e., the shell has an internal record for the quick access to the location of the "ls" command. See .

In Unix-like work environment, text processing is done by piping text through chains of standard text processing tools. This was another crucial Unix innovation.

There are few standard text processing tools which are used very often on the Unix-like system. No regular expression is used: cat(1) concatenates files and outputs the whole content. tac(1) concatenates files and outputs in reverse. cut(1) selects parts of lines and outputs. head(1) outputs the first part of files. tail(1) outputs the last part of files. sort(1) sorts lines of text files. uniq(1) removes duplicate lines from a sorted file. tr(1) translates or deletes characters. diff(1) compares files line by line. Basic regular expression (BRE) is used: grep(1) matches text with patterns. ed(1) is a primitive line editor. sed(1) is a stream editor. vim(1) is a screen editor. emacs(1) is a screen editor. (somewhat extended BRE) Extended regular expression (ERE) is used: egrep(1) matches text with patterns. awk(1) does simple text processing. tcl(3tcl) can do every conceivable text processing: re_syntax(3). Often used with tk(3tk). perl(1) can do every conceivable text processing. perlre(1). pcregrep(1) from the pcregrep package matches text with Perl Compatible Regular Expressions (PCRE) pattern. python(1) with the re module can do every conceivable text processing. See "/usr/share/doc/python/html/index.html". If you are not sure what exactly these commands do, please use "man command" to figure it out by yourself. Sort order and range expression are locale dependent. If you wish to obtain traditional behavior for a command, use C locale instead of UTF-8 ones by prepnding command with "LANG=C" (see and ). Perl regular expressions (perlre(1)), Perl Compatible Regular Expressions (PCRE), and Python regular expressions offered by the re module have many common extensions to the normal ERE.

Regular expressions are used in many text processing tools. They are analogous to the shell globs, but they are more complicated and powerful. The regular expression describes the matching pattern and is made up of text characters and metacharacters. The metacharacter is just a character with a special meaning. There are 2 major styles, BRE and ERE, depending on the text tools as described above. BRE ERE description of the regular expression \ . [ ] ^ $ * \ . [ ] ^ $ * common metacharacters \+ \? \( \) \{ \} \|   BRE only "\" escaped metacharacters   + ? ( ) { } | ERE only non-"\" escaped metacharacters c c match non-metacharacter "c" \c \c match a literal character "c" even if "c" is metacharacter by itself . . match any character including newline ^ ^ position at the beginning of a string $ $ position at the end of a string \< \< position at the beginning of a word \> \> position at the end of a word \[abc…\] [abc…] match any characters in "abc…" \[^abc…\] [^abc…] match any characters except in "abc…" r* r* match zero or more regular expressions identified by "r" r\+ r+ match one or more regular expressions identified by "r" r\? r? match zero or one regular expressions identified by "r" r1\|r2 r1|r2 match one of the regular expressions identified by "r1" or "r2" \(r1\|r2\) (r1|r2) match one of the regular expressions identified by "r1" or "r2" and treat it as a bracketed regular expression The regular expression of emacs is basically BRE but has been extended to treat "+"and "?" as the metacharacters as in ERE. Thus, there are no needs to escape them with "\" in the regular expression of emacs. grep(1) can be used to perform the text search using the regular expression. For example, try the following $ egrep 'GNU.*LICENSE|Yoyodyne' /usr/share/common-licenses/GPL GNU GENERAL PUBLIC LICENSE GNU GENERAL PUBLIC LICENSE Yoyodyne, Inc., hereby disclaims all copyright interest in the program See .

For the replacement expression, some characters have special meanings. replacement expression description of the text to replace the replacement expression & what the regular expression matched (use \& in emacs) \n what the n-th bracketed regular expression matched ("n" being number) For Perl replacement string, "$n" is used instead of "\n" and "&" has no special meaning. For example, try the following $ echo zzz1abc2efg3hij4 | \ sed -e 's/\(1[a-z]*\)[0-9]*\(.*\)$/=&=/' zzz=1abc2efg3hij4= $ echo zzz1abc2efg3hij4 | \ sed -e 's/\(1[a-z]*\)[0-9]*\(.*\)$/\2===\1/' zzzefg3hij4===1abc $ echo zzz1abc2efg3hij4 | \ perl -pe 's/(1[a-z]*)[0-9]*(.*)$/$2===$1/' zzzefg3hij4===1abc $ echo zzz1abc2efg3hij4 | \ perl -pe 's/(1[a-z]*)[0-9]*(.*)$/=&=/' zzz=&= Here please pay extra attention to the style of the bracketed regular expression and how the matched strings are used in the text replacement process on different tools. These regular expressions can be used for cursor movements and text replacement actions in some editors too. The back slash "\" at the end of line in the shell commandline escapes newline as a white space character and continues shell command line input to the next line. Please read all the related manual pages to learn these commands.

The ed(1) command can replace all instances of "FROM_REGEX" with "TO_TEXT" in "file". $ ed file <<EOF ,s/FROM_REGEX/TO_TEXT/g w q EOF The sed(1) command can replace all instances of "FROM_REGEX" with "TO_TEXT" in "file". $ sed file 's/FROM_REGEX/TO_TEXT/g' | sponge file The sponge(8) command is a non-standard Unix tool offered by the moreutils package. This is quite useful when you wish to overwrite original file. The vim(1) command can replace all instances of "FROM_REGEX" with "TO_TEXT" in "file" by using ex(1) commands. $ vim '+%s/FROM_REGEX/TO_TEXT/gc' '+w' '+q' file The "c" flag in the above ensures interactive confirmation for each substitution. Multiple files ("file1", "file2", and "file3") can be processed with regular expressions similarly with vim(1) or perl(1). $ vim '+argdo %s/FROM_REGEX/TO_TEXT/ge|update' '+q' file1 file2 file3 The "e" flag in the above prevents the "No match" error from breaking a mapping. $ perl -i -p -e 's/FROM_REGEX/TO_TEXT/g;' file1 file2 file3 In the perl(1) example, "-i" is for in-place editing, "-p" is for implicit loop over files. Use of argument "-i.bak" instead of "-i" keeps each original file by adding ".bak" to its filename. This makes recovery from errors easier for complex substitutions. ed(1) and vim(1) are BRE; perl(1) is ERE.

Let's consider a text file called "DPL" in which some pre-2004 Debian project leader's names and their initiation days are listed in a space-separated format. Ian Murdock August 1993 Bruce Perens April 1996 Ian Jackson January 1998 Wichert Akkerman January 1999 Ben Collins April 2001 Bdale Garbee April 2002 Martin Michlmayr March 2003 See "A Brief History of Debian" for the latest Debian leadership history. Awk is frequently used to extract data from these types of files. For example, try the following $ awk '{ print $3 }' <DPL # month started August April January January April April March $ awk '($1=="Ian") { print }' <DPL # DPL called Ian Ian Murdock August 1993 Ian Jackson January 1998 $ awk '($2=="Perens") { print $3,$4 }' <DPL # When Perens started April 1996 Shells such as Bash can be also used to parse this kind of file. For example, try the following $ while read first last month year; do echo $month done <DPL ... same output as the first Awk example Here, the read builtin command uses characters in "$IFS" (internal field separators) to split lines into words. If you change "$IFS" to ":", you can parse "/etc/passwd" with shell nicely. $ oldIFS="$IFS" # save old value $ IFS=':' $ while read user password uid gid rest_of_line; do if [ "$user" = "bozo" ]; then echo "$user's ID is $uid" fi done < /etc/passwd bozo's ID is 1000 $ IFS="$oldIFS" # restore old value (If Awk is used to do the equivalent, use "FS=':'" to set the field separator.) IFS is also used by the shell to split results of parameter expansion, command substitution, and arithmetic expansion. These do not occur within double or single quoted words. The default value of IFS is <space>, <tab>, and <newline> combined. Be careful about using this shell IFS tricks. Strange things may happen, when shell interprets some parts of the script as its input. $ IFS=":," # use ":" and "," as IFS $ echo IFS=$IFS, IFS="$IFS" # echo is a Bash builtin IFS= , IFS=:, $ date -R # just a command output Sat, 23 Aug 2003 08:30:15 +0200 $ echo $(date -R) # sub shell --> input to main shell Sat 23 Aug 2003 08 30 36 +0200 $ unset IFS # reset IFS to the default $ echo $(date -R) Sat, 23 Aug 2003 08:30:50 +0200

The following scripts do nice things as a part of a pipe. script snippet (type in one line) effect of command find /usr -print find all files under "/usr" seq 1 100 print 1 to 100 | xargs -n 1 <command> run command repeatedly with each item from pipe as its argument | xargs -n 1 echo split white-space-separated items from pipe into lines | xargs echo merge all lines from pipe into a line | grep -e <regex_pattern> extract lines from pipe containing <regex_pattern> | grep -v -e <regex_pattern> extract lines from pipe not containing <regex_pattern> | cut -d: -f3 - extract third field from pipe separated by ":" (passwd file etc.) | awk '{ print $3 }' extract third field from pipe separated by whitespaces | awk -F'\t' '{ print $3 }' extract third field from pipe separated by tab | col -bx remove backspace and expand tabs to spaces | expand - expand tabs | sort| uniq sort and remove duplicates | tr 'A-Z' 'a-z' convert uppercase to lowercase | tr -d '\n' concatenate lines into one line | tr -d '\r' remove CR | sed 's/^/# /' add "#" to the start of each line | sed 's/\.ext//g' remove ".ext" | sed -n -e 2p print the second line | head -n 2 - print the first 2 lines | tail -n 2 - print the last 2 lines One-line shell script can loop over many files using find(1) and xargs(1) to perform quite complicated tasks. See and . When using the shell interactive mode becomes too complicated, please consider to write a shell script (see ).

This chapter is written assuming the latest stable release is codename: &codename-stable;. Debian is a volunteer organization which builds consistent distributions of pre-compiled binary packages of free software and distributes them from its archive. The Debian archive is offered by many remote mirror sites for access through HTTP and FTP methods. It is also available as CD-ROM/DVD. The Debian package management system, when used properly, offers the user to install consistent sets of binary packages to the system from the archive. Currently, there are &all-packages; packages available for the &arch; architecture. The Debian package management system has a rich history and many choices for the front end user program and back end archive access method to be used. Currently, we recommend aptitude(8) as the main front end program for the Debian package management activity. package popcon size description DUMMY DUMMY DUMMY terminal-based package manager (current standard, front-end for apt) DUMMY DUMMY DUMMY Advanced Packaging Tool (APT), front-end for dpkg providing "http", "ftp", and "file" archive access methods (apt-get/apt-cache commands included) DUMMY DUMMY DUMMY tool for selecting tasks for installation on Debian system (front-end for APT) DUMMY DUMMY DUMMY enhancement package for APT to enable automatic installation of security upgrades DUMMY DUMMY DUMMY terminal-based package manager (previous standard, front-end for APT and other old access methods) DUMMY DUMMY DUMMY package management system for Debian DUMMY DUMMY DUMMY older ftp method for dselect DUMMY DUMMY DUMMY graphical package manager (GNOME front-end for APT) DUMMY DUMMY DUMMY graphical package manager (KDE front-end for APT) DUMMY DUMMY DUMMY APT utility programs: apt-extracttemplates(1), apt-ftparchive(1), and apt-sortpkgs(1) DUMMY DUMMY DUMMY package change history notification tool DUMMY DUMMY DUMMY lists critical bugs before each APT installation DUMMY DUMMY DUMMY APT package searching utility — command-line interface DUMMY DUMMY DUMMY recursively lists package dependencies The annoying bug #411123 for the mixed use of aptitude(8) and apt-get(8) commands has been resolved. If this kept you from using aptitude, please reconsider.

Here are some key points for package configuration on the Debian system. The manual configuration by the system administrator is respected. In other words, the package configuration system makes no intrusive configuration for the sake of convenience. Each package comes with its own configuration script with standardized user interface called debconf(7) to help initial installation process of the package. Debian Developers try their best to make your upgrade experience flawless with package configuration scripts. Full functionalities of packaged software are available to the system administrator. But ones with security risks are disabled in the default installation. If you manually activate a service with some security risks, you are responsible for the risk containment. Esoteric configuration may be manually enabled by the system administrator. This may creates interference with popular generic helper programs for the system configuration.

Do not install packages from random mixture of suites. It probably breaks the package consistency which requires deep system management knowledge, such as compiler ABI, library version, interpreter features, etc. The newbie Debian system administrator should stay with the stable release of Debian while applying only security updates. I mean that some of the following valid actions are better avoided, as a precaution, until you understand the Debian system very well. Here are some reminders. Do not include testing or unstable in "/etc/apt/sources.list". Do not mix standard Debian with other non-Debian archives such as Ubuntu in "/etc/apt/sources.list". Do not create "/etc/apt/preferences". Do not change default behavior of package management tools through configuration files without knowing their full impacts. Do not install random packages by "dpkg -i <random_package>". Do not ever install random packages by "dpkg --force-all -i <random_package>". Do not erase or alter files in "/var/lib/dpkg/". Do not overwrite system files by installing software programs directly compiled from source. Install them into "/usr/local" or "/opt", if needed. The non-compatible effects caused by above actions to the Debian package management system may leave your system unusable. The serious Debian system administrator who runs mission critical servers, should use extra precautions. Do not install any packages including security updates from Debian without thoroughly testing them with your particular configuration under safe conditions. You as the system administrator are responsible for your system in the end. The long stability history of Debian system is no guarantee by itself.

Despite my warnings above, I know many readers of this document wish to run the testing or unstable suites of Debian as their main system for self-administered Desktop environments. This is because they work very well, are updated frequently, and offer the latest features. For your production server, the stable suite with the security updates is recommended. The same can be said for desktop PCs on which you can spend limited administration efforts, e.g. for your mother's PC. It takes no more than simply setting the distribution string in the "/etc/apt/sources.list" to the suite name: "testing" or "unstable"; or the codename: "&codename-testing;" or "&codename-unstable;". This makes you live the life of eternal upgrades. The use of testing or unstable is a lot of fun but comes with some risks. Even though the unstable suite of Debian system looks very stable for most of the times, there have been some package problems on the testing and unstable suite of Debian system and a few of them were not so trivial to resolve. It may be quite painful for you. Sometimes, you may have a broken package or missing functionality for a few weeks. Here are some ideas to ensure quick and easy recovery from bugs in Debian packages. Make the system dual bootable by installing the stable suite of Debian system to another partition Make the installation CD handy for the rescue boot Consider installing apt-listbugs to check the Debian Bug Tracking System (BTS) information before the upgrade Learn the package system infrastructure enough to work around the problem Create a chroot or similar environment and run the latest system in it in advance (optional) (If you can not do any one of these precautionary actions, you are probably not ready for the testing and unstable suites.) Enlightenment with the following saves a person from the eternal karmic struggle of upgrade hell and let him reach Debian nirvana.

Let's look into the Debian archive from a system user's perspective. Official policy of the Debian archive is defined at Debian Policy Manual, Chapter 2 - The Debian Archive. For the typical HTTP access, the archive is specified in the "/etc/apt/sources.list" file as the following, e.g. for the current stable = &codename-stable; system. deb http://ftp.XX.debian.org/debian/ &codename-stable; main contrib non-free deb-src http://ftp.XX.debian.org/debian/ &codename-stable; main contrib non-free deb http://security.debian.org/ &codename-stable;/updates main contrib deb-src http://security.debian.org/ &codename-stable;/updates main contrib Please note "ftp.XX.debian.org" must be replaced with appropriate mirror site URL for your location, for USA "ftp.us.debian.org", which can be found in the list of Debian worldwide mirror sites. The status of these servers can be checked at Debian Mirror Checker site. Here, I tend to use codename "&codename-stable;" instead of suite name "stable" to avoid surprises when the next stable is released. The meaning of "/etc/apt/sources.list" is described in sources.list(5) and key points are followings. The "deb" line defines for the binary packages. The "deb-src" line defines for the source packages. The 1st argument is the root URL of the Debian archive. The 2nd argument is the distribution name: either the suite name or the codename. The 3rd and following arguments are the list of valid archive component names of the Debian archive. The "deb-src" lines can safely be omitted (or commented out by placing "#" at the start of the line) if it is just for aptitude which does not access source related meta data. It speeds up the updates of the archive meta data. The URL can be "http://", "ftp://", "file://", …. If "sid" is used in the above example instead of "&codename-stable;", the "deb: http://security.debian.org/ …" line for security updates in the "/etc/apt/sources.list" is not required. Security updates are only available for stable and testing (i.e., &codename-stable; and &codename-testing;). Here is the list of URL of the Debian archive sites and suite name or codename used in the configuration file. archive URL suite name (codename) purpose http://ftp.XX.debian.org/debian/ stable (&codename-stable;) stable (&codename-stable;) release http://ftp.XX.debian.org/debian/ testing (&codename-testing;) testing (&codename-testing;) release http://ftp.XX.debian.org/debian/ unstable (&codename-unstable;) unstable (&codename-unstable;) release http://ftp.XX.debian.org/debian/ experimental experimental pre-release (optional, only for developer) http://ftp.XX.debian.org/debian/ stable-proposed-updates Updates for the next stable point release (optional) http://security.debian.org/ stable/updates security updates for stable release (important) http://security.debian.org/ testing/updates security updates for testing release (important) http://volatile.debian.org/debian-volatile/ volatile compatible updates for spam filter, IM clients, etc. http://volatile.debian.org/debian-volatile/ volatile-sloppy non-compatible updates for spam filter, IM clients, etc. http://backports.org/debian/ &codename-stable;-backports newer backported packages for &codename-stable; (non-official, optional) Only pure stable release with security updates provides the best stability. Running mostly stable release mixed with some packages from testing or unstable release is riskier than running pure unstable release. If you really need the latest version of some programs under stable release, please use packages from the debian-volatile project and backports.org (see ) services. These services must be used with extra care. You should basically list only one of stable, testing, or unstable suites in the "deb" line. If you list any combination of stable, testing, and unstable suites in the "deb" line, APT programs slow down while only the latest archive is effective. Multiple listing makes sense for these when the "/etc/apt/preferences" file is used with clear objectives (see ). For the Debian system with the stable and testing suites, it is a good idea to include lines with "http://security.debian.org/" in the "/etc/apt/sources.list" to enable security updates as in the example above. Each Debian archive consists of 3 components. Components are alternatively called categories in "Debian Policy" or areas in "Debian Social Contract". The component is grouped by the compliance to "The Debian Free Software Guidelines" (DFSG). component number of packages criteria of package main &main-packages; DSFG compliant and no dependency to non-free contrib &contrib-packages; DSFG compliant but having dependency to non-free non-free &non-free-packages; not DSFG compliant Here the number of packages in the above is for the &arch; architecture. Strictly speaking, only the main component archive shall be considered as the Debian system. The Debian archive organization can be studied best by pointing your browser to the each archive URL appended with dists or pool. The distribution is referred by two ways, the suite or codename. The word distribution is alternatively used as the synonym to the suite in many documentations. The relationship between the suite and the codename can be summarized as the following. Timing suite = stable suite = testing suite = unstable after the &codename-stable; release codename = &codename-stable; codename = &codename-testing; codename = sid after the &codename-testing; release codename = &codename-testing; codename = &codename-nexttesting; codename = sid The history of codenames are described in Debian FAQ: 6.3.1 Which other codenames have been used in the past? In the stricter Debian archive terminology, the word "section" is specifically used for the categorization of packages by the application area. (Although, the word "main section" may sometimes be used to describe the Debian archive section which provides the main component.) Every time a new upload is done by the Debian developer (DD) to the unstable archive (via incoming processing), DD is required to ensure uploaded packages to be compatible with the latest set of packages in the latest unstable archive. If DD breaks this compatibility intentionally for important library upgrade etc, there is usually announcement to the debian-devel mailing list etc. Before a set of packages are moved by the Debian archive maintenance script from the unstable archive to the testing archive, the archive maintenance script not only checks the maturity (about 10 days old) and the status of the RC bug reports for the packages but also tries to ensure them to be compatible with the latest set of packages in the testing archive. This process makes the testing archive very current and usable. Through the gradual archive freeze process led by the release team, the testing archive is matured to make it completely consistent and bug free with some manual interventions. Then the new stable release is created by assigning the codename for the old testing archive to the new stable archive and creating the new codename for the new testing archive. The initial contents of the new testing archive is exactly the same as that of the newly released stable archive. Both the unstable and the testing archives may suffer temporary glitches due to several factors. Broken package upload to the archive (mostly for unstable) Delay of accepting the new packages to the archive (mostly for unstable) Archive synchronization timing issue (both for testing and unstable) Manual intervention to the archive such as package removal (more for testing) etc. So if you ever decide to use these archives, you should be able to fix or work around these kinds of glitches. For about few months after a new stable release, most desktop users should use the stable archive with its security updates even if they usually use unstable or testing archives. For this transition period, both unstable and testing archives are not good for most people. Your system is difficult to keep in good working condition with the unstable archive since it suffers surges of major upgrades for core packages. The testing archive is not useful either since it contains mostly the same content as the stable archive without its security support (Debian testing-security-announce 2008-12). After a month or so, the unstable archive may be usable if you are careful. When tracking the testing archive, problem caused by a removed package is usually worked around by installing corresponding package from the unstable archive which is uploaded for bug fix. See Debian Policy Manual for archive definitions. "Sections" "Priorities" "Base system" "Essential packages"

The Debian system offers a consistent set of binary packages through its versioned binary dependency declaration mechanism in the control file fields. Here is a bit over simplified definition for them. "Depends" This declares an absolute dependency and all of the packages listed in this field must be installed at the same time or in advance. "Pre-Depends" This is like Depends, except that it requires completed installation of the listed packages in advance. "Recommends" This declares a strong, but not absolute, dependency. Most users would not want the package unless all of the packages listed in this field are installed. "Suggests" This declares a weak dependency. Many users of this package may benefit from installing packages listed in this field but can have reasonable functions without them. "Enhances" This declares a week dependency like Suggests but works in the opposite direction. "Conflicts" This declares an absolute incompatibility. All of the packages listed in this field must be removed to install this package. "Replaces" This is declared when files installed by this package replace files in the listed packages. "Provides" This is declared when this package provide all of the files and functionality in the listed packages. Please note that defining, "Provides", "Conflicts" and "Replaces" simultaneously to an virtual package is the sane configuration. This ensures that only one real package providing this virtual package can be installed at any one time. The official definition including source dependency can be found in the Policy Manual: Chapter 7 - Declaring relationships between packages.

Here is a summary of the simplified event flow of the package management by APT. Update ("aptitude update" or "apt-get update"): Fetch archive metadata from remote archive Reconstruct and update local metadata for use by APT Upgrade ("aptitude safe-upgrade" and "aptitude full-upgrade", or "apt-get upgrade" and "apt-get dist-upgrade"): Chose candidate version which is usually the latest available version for all installed packages (see for exception) Make package dependency resolution Fetch selected binary packages from remote archive if candidate version is different from installed version Unpack fetched binary packages Run preinst script Install binary files Run postinst script Install ("aptitude install …" or "apt-get install …"): Chose packages listed on the command line Make package dependency resolution Fetch selected binary packages from remote archive Unpack fetched binary packages Run preinst script Install binary files Run postinst script Remove ("aptitude remove …" or "apt-get remove …"): Chose packages listed on the command line Make package dependency resolution Run prerm script Remove installed files except configuration files Run postrm script Purge ("aptitude purge …" or "apt-get purge …"): Chose packages listed on the command line Make package dependency resolution Run prerm script Remove installed files including configuration files Run postrm script Here, I intentionally skipped technical details for the sake of big picture.

You should read the fine official documentation. The first document to read is the Debian specific "/usr/share/doc/<package_name>/README.Debian". Other documentation in "/usr/share/doc/<package_name>/" should be consulted too. If you set shell as , type the following. $ cd <package_name> $ pager README.Debian $ mc You may need to install the corresponding documentation package named with "-doc" suffix for detailed information. If you are experiencing problems with a specific package, make sure to check out the Debian bug tracking system (BTS) sites, first. web site command Home page of the Debian bug tracking system (BTS) sensible-browser "http://bugs.debian.org/" The bug report of a known package name sensible-browser "http://bugs.debian.org/<package_name>" The bug report of known bug number sensible-browser "http://bugs.debian.org/<bug_number>" Search Google with search words including "site:debian.org", "site:wiki.debian.org", "site:lists.debian.org", etc. When you file a bug report, please use reportbug(1) command.

Aptitude is the current preferred package management tool for the Debian system. It can be used as the commandline alternative to apt-get / apt-cache and also as the full screen interactive package management tool. For the package management operation which involves package installation or updates package metadata, you need to have root privilege.

Here are basic package management operations with commandline using aptitude(8) and apt-get(8) /apt-cache(8). aptitude syntax apt-get/apt-cache syntax description aptitude update apt-get update update package archive metadata aptitude install foo apt-get install foo install candidate version of "foo" package with its dependencies aptitude safe-upgrade apt-get upgrade install candidate version of installed packages without removing any other packages aptitude full-upgrade apt-get dist-upgrade <package> install candidate version of installed packages while removing other packages if needed aptitude remove foo apt-get remove foo remove "foo" package while leaving its configuration files N/A apt-get autoremove remove auto-installed packages which is no longer required aptitude purge foo apt-get purge foo purge "foo" package with its configuration files aptitude clean apt-get clean clear out the local repository of retrieved package files completely aptitude autoclean apt-get autoclean clear out the local repository of retrieved package files for outdated packages aptitude show foo apt-cache show <package> display detailed information about "foo" package aptitude search <regex> apt-cache search <regex> search packages which match <regex> aptitude why <regex> N/A explain the reason why <regex> matching packages should be installed aptitude why-not <regex> N/A explain the reason why <regex> matching packages can not be installed Although it is now safe to mix different package tools on the Debian system, it is best to continue using aptitude as much as possible. The difference between "safe-upgrade"/"upgrade" and "full-upgrade"/"dist-upgrade" only appears when new versions of packages stand in different dependency relationships from old versions of those packages. The "aptitude safe-upgrade" command does not install new packages nor remove installed packages. The "aptitude why <regex>" can list more information by "aptitude -v why <regex>". Similar information can be obtained by "apt-cache rdepends <package>". When aptitude command is started in the commandline mode and faces some issues such as package conflicts, you can switch to the full screen interactive mode by pressing "e"-key later at the prompt. You may provide command options right after "aptitude". command option description -s simulate the result of the command -d download only but no install/upgrade -D show brief explanations before the automatic installations and removals See aptitude(8) and "aptitude user's manual" at "/usr/share/doc/aptitude/README" for more. The dselect package is still available and was the preferred full screen interactive package management tool in previous releases.

For the interactive package management, you start aptitude in interactive mode from the console shell prompt as follows. $ sudo aptitude -u Password: This updates the local copy of the archive information and display the package list in the full screen with menu. Aptitude places its configuration at "~/.aptitude/config". If you want to use root's configuration instead of user's one, use "sudo -H aptitude …" instead of "sudo aptitude …" in the above expression. Aptitude automatically sets pending actions as it is started interactively. If you do not like it, you can reset it from menu: "Action" → "Cancel pending actions".

Notable key strokes to browse status of packages and to set "planned action" on them in this full screen mode are the following. key key binding F10 or Ctrl-t menu ? display help for keystroke (more complete listing) F10 → Help → User's Manual display User's Manual u update package archive information + mark the package for the upgrade or the install - mark the package for the remove (keep configuration files) _ mark the package for the purge (remove configuration files) = place the package on hold U mark all upgradable packages (function as full-upgrade) g start downloading and installing selected packages q quit current screen and save changes x quit current screen and discard changes Enter view information about a package C view a package's changelog l change the limit for the displayed packages / search for the first match \ repeat the last search The file name specification of the command line and the menu prompt after pressing "l" and "//" take the aptitude regex as described below. Aptitude regex can explicitly match a package name using a string started by "~n and followed by the package name. You need to press "U" to get all the installed packages upgraded to the candidate version in the visual interface. Otherwise only the selected packages and certain packages with versioned dependency to them are upgraded to the candidate version.

In the interactive full screen mode of aptitude(8), packages in the package list are displayed as the next example. idA libsmbclient -2220kB 3.0.25a-1 3.0.25a-2 Here, this line means from the left as the following. The "current state" flag (the first letter) The "planned action" flag (the second letter) The "automatic" flag (the third letter) The Package name The change in disk space usage attributed to "planned action" The current version of the package The candidate version of the package The full list of flags are given at the bottom of Help screen shown by pressing "?". The candidate version is chosen according to the current local preferences (see apt_preferences(5) and ). Several types of package views are available under the menu "Views". view status description of view Package View Good see (default) Audit Recommendations Good list packages which are recommended by some installed packages but not yet installed are listed Flat Package List Good list packages without categorization (for use with regex) Debtags Browser Very usable list packages categorized according to their debtags entries Categorical Browser Deprecated list packages categorized according to their category (use Debtags Browser, instead) Please help us improving tagging packages with debtags! The standard "Package View" categorizes packages somewhat like dselect with few extra features. category description of view Upgradable Packages list packages organized as section → component → package New Packages , , Installed Packages , , Not Installed Packages , , Obsolete and Locally Created Packages , , Virtual Packages list packages with the same function Tasks list packages with different functions generally needed for a task Tasks view can be used to cherry pick packages for your task.

Aptitude offers several options for you to search packages using its regex formula. Shell commandline: "aptitude search '<aptitude_regex>'" to list installation status, package name and short description of matching packages "aptitude show '<package_name>'" to list detailed description of the package Interactive full screen mode: "l" to limit package view to matching packages "/" for search to a matching package "\" for backward search to a matching package "n" for find-next "N" for find-next (backward) The string for <package_name> is treated as the exact string match to the package name unless it is started explicitly with "~" to be the regex formula.

The aptitude regex formula is mutt-like extended ERE (see ) and the meanings of the aptitude specific special match rule extensions are as follows. description of the extended match rule regex formula match on package name ~n<regex_name> match on description ~d<regex_description> match on task name ~t<regex_task> match on debtag ~G<regex_debtag> match on maintainer ~m<regex_maintainer> match on package section ~s<regex_section> match on package version ~V<regex_version> match archive ~A{sarge,etch,sid} match origin ~O{debian,…} match priority ~p{extra,important,optional,required,standard} match essential packages ~E match virtual packages ~v match new packages ~N match with pending action ~a{install,upgrade,downgrade,remove,purge,hold,keep} match installed packages ~i match installed packages with A-mark (auto installed package) ~M match installed packages without A-mark (administrator selected package) ~i!~M match installed and upgradable packages ~U match removed but not purged packages ~c match removed, purged or can-be-removed packages ~g match packages with broken relation ~b match packages with broken depends/predepends/conflict ~B<type> match packages from which relation <type> is defined to <term> package ~D[<type>:]<term> match packages from which broken relation <type> is defined to <term> package ~DB[<type>:]<term> match packages to which the <term> package defines relation <type> ~R[<type>:]<term> match packages to which the <term> package defines broken relation <type> ~RB[<type>:]<term> match packages to which some other installed packages depend on ~R~i match packages to which no other installed packages depend on !~R~i match packages to which some other installed packages depend or recommend on ~R~i|~Rrecommends:~i match <term> package with filtered version ~S filter <term> match all packages (true) ~T match no packages (false) ~F The regex part is the same ERE as the one used in typical Unix-like text tools using "^", ".*", "$" etc. as in egrep(1), awk(1) and perl(1). The relation <type> is one of (depends, predepends, recommends, suggests, conflicts, replaces, provides). The default relation type is "depends". When <regex_pattern> is a null string, place "~T" immediately after the command. Here are some short cuts. "~P<term>" == "~Dprovides:<term>" "~C<term>" == "~Dconflicts:<term>" "…~W term" == "(…|term)" Users familiar with mutt pick up quickly, as mutt was the inspiration for the expression syntax. See "SEARCHING, LIMITING, AND EXPRESSIONS" in the "User's Manual" "/usr/share/doc/aptitude/README". With the lenny version of aptitude(8), the new long form syntax such as "?broken" may be used for regex matching in place for its old short form equivalent "~b". Now space character " " is considered as one of the regex terminating character in addition to tilde character "~". See "User's Manual" for the new long form syntax.

The selection of a package in aptitude not only pulls in packages which are defined in its "Depends:" list but also defined in the "Recommends:" list if the menu "F10 → Options → Dependency handling" is set accordingly. These auto installed packages are removed automatically if they are no longer needed under aptitude. Before the lenny release, apt-get and other standard APT tools did not offer the autoremove functionality.

You can check package activity history in the log files. file content /var/log/dpkg.log Log of dpkg level activity for all package activities /var/log/apt/term.log Log of generic APT activity /var/log/aptitude Log of aptitude command activity In reality, it is not so easy to get meaningful understanding quickly out from these logs. See for easier way.

Aptitude has advantages over other APT based packaging systems (apt-get, apt-cache, synaptic, …). aptitude removes unused auto installed packages automatically using its own extra layer of package state file (/var/lib/aptitude/pkgstates). (For new "lenny", other APT does the same.) aptitude makes it easy to resolve package conflicts and to add recommended packages. aptitude makes it easy to keep track of obsolete software by listing under "Obsolete and Locally Created Packages". aptitude gives a log of its history in "/var/log/aptitude". aptitude offers access to all versions of the package if available. aptitude includes a fairly powerful regex based system for searching particular packages and limiting the package display. aptitude in the full screen mode has su functionality embedded and can be run from normal user until you really need administrative privileges. For the old etch release version, synaptic also gives you the history log; apt-get did not but you can rely on the log of dpkg. Anyway, aptitude is nice for interactive console use.

Here are few examples of aptitude(8) operations.

The following command lists packages with regex matching on package names. $ aptitude search '~n(pam|nss).*ldap' p libnss-ldap - NSS module for using LDAP as a naming service p libpam-ldap - Pluggable Authentication Module allowing LDAP interfaces This is quite handy for you to find the exact name of a package.

The regex "~dipv6" in the "New Flat Package List" view with "l" prompt, limits view to packages with the matching description and let you browse their information interactively.

You can purge all remaining configuration files of removed packages. Check results of the following command. # aptitude search '~c' If you think listed packages are OK to be purged, execute the following command. # aptitude purge '~c' You may want to do the similar in the interactive mode for fine grained control. You provide the regex "~c" in the "New Flat Package List" view with "l" prompt. This limits the package view only to regex matched packages, i.e., "removed but not purged". All these regex matched packages can be shown by pressing "[" at top level headings. Then you press "_" at top level headings such as "Installed Packages". Only regex matched packages under the heading are marked to be purged by this. You can exclude some packages to be purged by pressing "=" interactively for each of them. This technique is quite handy and works for many other command keys.

Here is how I tidy auto/manual install status for packages (after using non-aptitude package installer etc.). Start aptitude in interactive mode as root. Type "u", "U", "f" and "g" to update and upgrade package list and packages. Type "l" to enter the package display limit as "~i(~R~i|~Rrecommends:~i)" and type "M" over "Installed Packages" as auto installed. Type "l" to enter the package display limit as "~prequired|~pimportant|~pstandard|~E" and type "m" over "Installed Packages" as manual installed. Type "l" to enter the package display limit as "~i!~M" and remove unused package by typing "-" over each of them after exposing them by typing "[" over "Installed Packages". Type "l" to enter the package display limit as "~i" and type "m" over "Tasks" as manual installed. Exit aptitude. Start "apt-get -s autoremove|less" as root to check what are not used. Restart aptitude in interactive mode and mark needed packages as "m". Restart "apt-get -s autoremove|less" as root to recheck REMOVED contain only expected packages. Start "apt-get autoremove|less" as root to autoremove unused packages. The "m" action over "Tasks" is an optional one to prevent mass package removal situation in future.

When moving to a new release etc, you should consider to perform a clean installation of new system even though Debian is upgradable as described below. This provides you a chance to remove garbages collected and exposes you to the best combination of latest packages. Of course, you should make a full backup of system to a safe place (see ) before doing this. I recommend to make a dual boot configuration using different partition to have the smoothest transition. You can perform system wide upgrade to a newer release by changing contents of the "/etc/apt/sources.list" file pointing to a new release and running the "aptitude update; aptitude full-upgrade" command. To upgrade from stable to testing or unstable, you replace "&codename-stable;" in the "/etc/apt/sources.list" example of with "&codename-testing;" or "sid". In reality, you may face some complications due to some package transition issues, mostly due to package dependencies. The larger the difference of the upgrade, the more likely you face larger troubles. For the transition from the old stable to the new stable after its release, you can read its new Release Notes and follow the exact procedure described in it to minimize troubles. When you decide to move from stable to testing before its formal release, there are no Release Notes to help you. The difference between stable and testing could have grown quite large after the previous stable release and makes upgrade situation complicated. You should make precautionary moves for the full upgrade while gathering latest information from mailing list and using common senses. Read previous "Release Notes". Backup entire system (especially data and configuration information). Have bootable media handy for broken bootloader. Inform users on the system well in advance. Record upgrade activity with script(1). Apply "unmarkauto" to required packages, e.g., "aptitude unmarkauto vim", to prevent removal. Minimize installed packages to reduce chance of package conflicts, e.g., remove desktop task packages. Remove the "/etc/apt/preferences" file (disable apt-pinning). Try to upgrade step wise: oldstable → stable → testing → unstable. Update the "/etc/apt/sources.list" file to point to new archive only and run "aptitude update". Install, optionally, new core packages first, e.g., "aptitude install perl". Run the "aptitude full-upgrade -s" command to assess impact. Run the "aptitude full-upgrade" command at last. It is not wise to skip major Debian release when upgrading between stable releases. In previous "Release Notes", GCC, Linux Kernel, initrd-tools, Glibc, Perl, APT tool chain, etc. have required some special attention for system wide upgrade. For daily upgrade in unstable, see .

Here are list of other package management operations for which aptitude is too high-level or lacks required functionalities. command action COLUMNS=120 dpkg -l <package_name_pattern> list status of an installed package for the bug report dpkg -L <package_name> list contents of an installed package dpkg -L <package_name> | egrep '/usr/share/man/man.*/.+' list manpages for an installed package dpkg -S <file_name_pattern> list installed packages which have matching file name apt-file search <file_name_pattern> list packages in archive which have matching file name apt-file list <package_name_pattern> list contents of matching packages in archive dpkg-reconfigure <package_name> reconfigure the exact package dpkg-reconfigure -p=low <package_name> reconfigure the exact package with the most detailed question configure-debian reconfigure packages from the full screen menu dpkg --audit audit system for partially installed packages dpkg --configure -a configure all partially installed packages apt-cache policy <binary_package_name> show available version, priority, and archive information of a binary package apt-cache madison <package_name> show available version, archive information of a package apt-cache showsrc <binary_package_name> show source package information of a binary package apt-get build-dep <package_name> install required packages to build package apt-get source <package_name> download a source (from standard archive) dget <URL for dsc file> download a source packages (from other archive) dpkg-source -x <package_name>_<version>-<debian_version>.dsc build a source tree from a set of source packages ("*.tar.gz" and "*.diff.gz") debuild binary build package(s) from a local source tree make-kpkg kernel_image build a kernel package from a kernel source tree make-kpkg --initrd kernel_image build a kernel package from a kernel source tree with initramfs enabled dpkg -i <package_name><version>-<debian_version><arch>.deb install a local package to the system debi <package_name><version>-<debian_version><arch>.dsc install local package(s) to the system dpkg --get-selection '*' >selection.txt save dpkg level package selection state information dpkg --set-selection <selection.txt set dpkg level package selection state information Lower level package tools such as "dpkg -i …" and "debi …" should be carefully used by the system administrator. It does not automatically take care required package dependencies. Dpkg's commandline options "--force-all" and similar (see dpkg(1)) are intended to be used by experts only. Using them without fully understanding their effects may break your whole system. Please note the following. All system configuration and installation commands require to be run from root. Unlike aptitude which uses regex (see ), other package management commands use pattern like shell glob (see ). apt-file(1) provided by the apt-file package must run "apt-file update" in advance. configure-debian(8) provided by the configure-debian package runs dpkg-reconfigure(8) as its backend. dpkg-reconfigure(8) runs package scripts using debconf(1) as its backend. "apt-get build-dep", "apt-get source" and "apt-cache showsrc" commands require "deb-src" entry in "/etc/apt/sources.list". dget(1), debuild(1), and debi(1) require devscripts package. See (re)packaging procedure using "apt-get source" in . make-kpkg command requires the kernel-package package (see ). See for general packaging. The source package format described here as a set of source packages ("*.tar.gz" and "*.diff.gz") is format 1.0 which is still popular. See more on dpkg-source(1) for other newer formats.

The installation of debsums enables verification of installed package files against MD5sum values in the "/var/lib/dpkg/info/*.md5sums" file with debsums(1). See for how MD5sum works. Because MD5sum database may be tampered by the intruder, debsums(1) is of limited use as a security tool. It is only good for checking local modifications by the administrator or damage due to media errors.

Many users prefer to follow the unstable release of the Debian system for its new features and packages. This makes the system more prone to be hit by the critical package bugs. The installation of the apt-listbugs package safeguards your system against critical bugs by checking Debian BTS automatically for critical bugs when upgrading with APT system. The installation of the apt-listchanges package provides important news in "NEWS.Debian" when upgrading with APT system.

Although visiting Debian site http://packages.debian.org/ facilitates easy ways to search on the package meta data these days, let's look into more traditional ways. The grep-dctrl(1), grep-status(1), and grep-available(1) commands can be used to search any file which has the general format of a Debian package control file. The "dpkg -S <file_name_pattern>" can be used search package names which contain files with the matching name installed by dpkg. But this overlooks files created by the maintainer scripts. If you need to make more elaborate search on the dpkg meta data, you need to run "grep -e regex_pattern *" command in the "/var/lib/dpkg/info/" directory. This makes you search words mentioned in package scripts and installation query texts. If you wish to look up package dependency recursively, you should use apt-rdepends(8).

Let's learn how the Debian package management system works internally. This should help you to create your own solution to some package problems.

Meta data files for each distribution are stored under "dist/<codename>" on each Debian mirror sites, e.g., "http://ftp.us.debian.org/debian/". Its archive structure can be browsed by the web browser. There are 6 types of key meta data. file location content Release top of distribution archive description and integrity information Release.gpg top of distribution signature file for the "Release" file signed with the archive key Contents-<architecture> top of distribution list of all files for all the packages in the pertinent archive Release top of each distribution/component/architecture combination archive description used for the rule of apt_preferences(5) Packages top of each distribution/component/binary-architecture combination concatenated debian/control for binary packages Sources top of each distribution/component/source combination concatenated debian/control for source packages In the recent archive, these meta data are stored as the compressed and differential files to reduce network traffic.

The top level "Release" file is used for signing the archive under the secure APT system. Each suite of the Debian archive has a top level "Release" file, e.g., "http://ftp.us.debian.org/debian/dists/unstable/Release", as follows. Origin: Debian Label: Debian Suite: unstable Codename: sid Date: Sat, 26 Jan 2008 20:13:58 UTC Architectures: alpha amd64 arm hppa hurd-i386 i386 ia64 m68k mips mipsel powerpc s390 sparc Components: main contrib non-free Description: Debian x.y Unstable - Not Released MD5Sum: e9f11bc50b12af7927d6583de0a3bd06 22788722 main/binary-alpha/Packages 43524d07f7fa21b10f472c426db66168 6561398 main/binary-alpha/Packages.gz ... Here, you can find my rationale to use the "suite", "codeneme", and "components" in . The "distribution" is used when referring to both "suite" and "codeneme". The integrity of the top level "Release" file is verified by cryptographic infrastructure called the secure apt. The cryptographic signature file "Release.gpg" is created from the authentic top level "Release" file and the secret Debian archive key. The public Debian archive key can be seeded into "/etc/apt/trusted.gpg"; automatically by installing the keyring with the latest base-files package, or manually by gpg or apt-key tool with the latest public archive key posted on the ftp-master.debian.org . The secure APT system verifies the integrity of the downloaded top level "Release" file cryptographically by this "Release.gpg" file and the public Debian archive key in "/etc/apt/trusted.gpg". The integrity of all the "Packages" and "Sources" files are verified by using MD5sum values in its top level "Release" file. The integrity of all package files are verified by using MD5sum values in the "Packages" and "Sources" files. See debsums(1) and . Since the cryptographic signature verification is very CPU intensive process than the MD5sum value calculation, use of MD5sum value for each package while using cryptographic signature for the top level "Release" file provides the good security with the performance (see ).

The archive level "Release" files are used for the rule of apt_preferences(5). There are archive level "Release" files for all archive locations specified by "deb" line in "/etc/apt/sources.list", such as "http://ftp.us.debian.org/debian/dists/unstable/main/binary-amd64/Release" or "http://ftp.us.debian.org/debian/dists/sid/main/binary-amd64/Release" as follows. Archive: unstable Component: main Origin: Debian Label: Debian Architecture: amd64 For "Archive:" stanza, suite names ("stable", "testing", "unstable", …) are used in the Debian archive while codenames ("dapper", "feisty", "gutsy", "hardy", "intrepid", …) are used in the Ubuntu archive. For some archives, such as experimental, volatile-sloppy, and &codename-stable;-backports, which contain packages which should not be installed automatically, there is an extra line, e.g., "http://ftp.us.debian.org/debian/dists/experimental/main/binary-amd64/Release" as follows. Archive: experimental Component: main Origin: Debian Label: Debian NotAutomatic: yes Architecture: amd64 Please note that for normal archives without "NotAutomatic: yes", the default Pin-Priority value is 500, while for special archives with "NotAutomatic: yes", the default Pin-Priority value is 1 (see apt_preferences(5) and ).

When APT tools, such as aptitude, apt-get, synaptic, apt-file, auto-apt…, are used, we need to update the local copies of the meta data containing the Debian archive information. These local copies have following file names corresponding to the specified distribution, component, and architecture names in the "/etc/apt/sources.list" (see ). "/var/lib/apt/lists/ftp.us.debian.org_debian_dists_<distribution>_Release" "/var/lib/apt/lists/ftp.us.debian.org_debian_dists_<distribution>_Release.gpg" "/var/lib/apt/lists/ftp.us.debian.org_debian_dists_<distribution>_<component>_binary-<architecture>_Packages" "/var/lib/apt/lists/ftp.us.debian.org_debian_dists_<distribution>_<component>_source_Sources" "/var/cache/apt/apt-file/ftp.us.debian.org_debian_dists_<distribution>_Contents-<architecture>.gz" (for apt-file) First 4 types of files are shared by all the pertinent APT commands and updated from command line by "apt-get update" and "aptitude update". The "Packages" meta data are updated if there is the "deb" line in "/etc/apt/sources.list". The "Sources" meta data are updated if there is the "deb-src" line in "/etc/apt/sources.list". The "Packages" and "Sources" meta data contain "Filename:" stanza pointing to the file location of the binary and source packages. Currently, these packages are located under the "pool/" directory tree for the improved transition over the releases. Local copies of "Packages" meta data can be interactively searched with the help of aptitude. The specialized search command grep-dctrl(1) can search local copies of "Packages" and "Sources" meta data. Local copy of "Contents-<architecture>" meta data can be updated by "apt-file update" and its location is different from other 4 ones. See apt-file(1). (The auto-apt uses different location for local copy of "Contents-<architecture>.gz" as default.)

In addition to the remotely fetched meta data, the APT tool after lenny stores its locally generated installation state information in the "/var/lib/apt/extended_states" which is used by all APT tools to track all auto installed packages.

In addition to the remotely fetched meta data, the aptitude command stores its locally generated installation state information in the "/var/lib/aptitude/pkgstates" which is used only by it.

All the remotely fetched packages via APT mechanism are stored in the "/var/cache/apt/packages" until they are cleaned.

Debian package files have particular name structures. package type name structure The binary package (a.k.a deb) <package-name>_<epoch>:<upstream-version>-<debian.version>-<architecture>.deb The binary package for the debian-installer (a.k.a udeb) <package-name>_<epoch>:<upstream-version>-<debian.version>-<architecture>.udeb The source package (upstream source) <package-name>_<epoch>:<upstream-version>-<debian.version>.tar.gz The source package (Debian changes) <package-name>_<epoch>:<upstream-version>-<debian.version>.diff.gz The source package (description) <package-name>_<epoch>:<upstream-version>-<debian.version>.dsc name component usable characters (regex) existence <package-name> [a-z,A-Z,0-9,.,,-] required <epoch>: [0-9]+: optional <upstream-version> [a-z,A-Z,0-9,.,,-,:] required <debian.version> [a-z,A-Z,0-9,.,,~] optional You can check package version order by dpkg(1), e.g., "dpkg --compare-versions 7.0 gt 7.~pre1 ; echo $?" . The debian-installer (d-i) uses udeb as the file extension for its binary package instead of normal deb. An udeb package is a stripped down deb package which removes few non-essential contents such as documentation to save space while relaxing the package policy requirements. Both deb and udeb packages share the same package structure. The "u" stands for micro.

dpkg(1) is the lowest level tool for the Debian package management. This is very powerful and needs to be used with care. While installing package called "<package_name>", dpkg process it in the following order. Unpack the deb file ("ar -x" equivalent) Execute "<package_name>.preinst" using debconf(1) Install the package content to the system ("tar -x" equivalent) Execute "<package_name>.postinst" using debconf(1) The debconf system provides standardized user interaction with I18N and L10N () supports. file description of contents /var/lib/dpkg/info/<package_name>.conffiles list of configuration files. (user modifiable) /var/lib/dpkg/info/<package_name>.list list of files and directories installed by the package /var/lib/dpkg/info/<package_name>.md5sums list of MD5 hash values for files installed by the package /var/lib/dpkg/info/<package_name>.preinst package script run before the package installation /var/lib/dpkg/info/<package_name>.postinst package script run after the package installation /var/lib/dpkg/info/<package_name>.prerm package script run before the package removal /var/lib/dpkg/info/<package_name>.postrm package script run after the package removal /var/lib/dpkg/info/<package_name>.config package script for debconf system /var/lib/dpkg/alternatives/<package_name> the alternative information used by the update-alternatives command /var/lib/dpkg/available the availability information for all the package /var/lib/dpkg/diversions the diversions information used by dpkg(1) and set by`dpkg-divert`(8) /var/lib/dpkg/statoverride the stat override information used by dpkg(1) and set by`dpkg-statoverride`(8) /var/lib/dpkg/status the status information for all the packages /var/lib/dpkg/status-old the first-generation backup of the "var/lib/dpkg/status" file /var/backups/dpkg.status* the second-generation backup and older ones of the "var/lib/dpkg/status" file The "status" file is also used by the tools such as dpkg(1), "dselect update" and "apt-get -u dselect-upgrade". The specialized search command grep-dctrl(1) can search the local copies of "status" and "available" meta data. In the debian-installer environment, the udpkg command is used to open udeb packages. The udpkg command is a stripped down version of the dpkg command.

The Debian system has mechanism to install somewhat overlapping programs peacefully using update-alternatives(8). For example, you can make the vi command select to run vim while installing both vim and nvi packages. $ ls -l $(type -p vi) lrwxrwxrwx 1 root root 20 2007-03-24 19:05 /usr/bin/vi -> /etc/alternatives/vi $ sudo update-alternatives --display vi ... $ sudo update-alternatives --config vi Selection Command ---------------------------------------------- 1 /usr/bin/vim *+ 2 /usr/bin/nvi Enter to keep the default[*], or type selection number: 1 The Debian alternatives system keeps its selection as symlinks in "/etc/alternatives/". The selection process uses corresponding file in "/var/lib/dpkg/alternatives/".

Stat overrides provided by the dpkg-statoverride(8) command are a way to tell dpkg(1) to use a different owner or mode for a file when a package is installed. If "--update" is specified and file exists, it is immediately set to the new owner and mode. The direct alteration of owner or mode for a file owned by the package using chmod or chown commands by the system administrator is reset by the next upgrade of the package. I use the word file here, but in reality this can be any filesystem object that dpkg handles, including directories, devices, etc.

File diversions provided by the dpkg-divert(8) command are a way of forcing dpkg(1) not to install a file into its default location, but to a diverted location. The use of dpkg-divert is meant for the package maintenance scripts. Its casual use by the system administrator is deprecated.

When running unstable system, the administrator is expected to recover from broken package management situation. Some methods described here are high risk actions. You have been warned!

If a desktop GUI program experienced instability after significant upstream version upgrade, you should suspect interferences with old local configuration files created by it. If it is stable under newly created user account, this hypothesis is confirmed. (This is a bug of packaging and usually avoided by the packager.) To recover stability, you should move corresponding local configuration files and restart the GUI program. You may need to read old configuration file contents to recover configuration information later. (Do not erase them too quickly.)

Archive level package management systems, such as aptitude(8) or apt-get(1), do not even try to install packages with overlapped files using package dependencies (see ). Errors by the package maintainer or deployment of inconsistently mixed source of archives (see ) by the system administrator may create situation with incorrectly defined package dependencies. When you install a package with overlapped files using aptitude(8) or apt-get(1) under such situation, dpkg(1) which unpacks package ensures to return error to the calling program without overwriting existing files. The use of third party packages introduces significant system risks via maintainer scripts which are run with root privilege and can do anything to your system. The dpkg(1) command only protects against overwriting by the unpacking. You can work around such broken installation by removing the old offending package, <old-package>, first. $ sudo dpkg -P <old-package>

When a command in the package script returns error for some reason and the script exits with error, the package management system aborts their action and ends up with partially installed packages. When a package contains bugs in its removal scripts, the package may become impossible to remove and quite nasty. For the package script problem of "<package_name>", you should look into following package scripts. "/var/lib/dpkg/info/<package_name>.preinst" "/var/lib/dpkg/info/<package_name>.postinst" "/var/lib/dpkg/info/<package_name>.prerm" "/var/lib/dpkg/info/<package_name>.postrm" Edit the offending package script from the root using following techniques. disable the offending line by preceding "#" force to return success by appending the offending line with "|| true" Configure all partially installed packages with the following command. # dpkg --configure -a

Since dpkg is very low level package tool, it can function under the very bad situation such as unbootable system without network connection. Let's assume foo package was broken and needs to be replaced. You may still find cached copies of older bug free version of foo package in the package cache directory: "/var/cache/apt/archives/". (If not, you can download it from archive of http://snapshot.debian.net/ or copy it from package cache of a functioning machine.) If you can boot the system, you may install it by the following command. # dpkg -i /path/to/foo_<old_version>_<arch>.deb If system breakage is minor, you may alternatively downgrade the whole system as using the higher level APT system. If your system is unbootable from hard disk, you should seek other ways to boot it. Boot the system using the debian-installer CD in rescue mode. Mount the unbootable system on the hard disk to "/target". Install older version of foo package by the following. # dpkg --root /target -i /path/to/foo_<old_version>_<arch>.deb This example works even if the dpkg command on the hard disk is broken. Any GNU/Linux system started by another system on hard disk, live GNU/Linux CD, bootable USB-key drive, or netboot can be used similarly to rescue broken system. If attempting to install a package this way fails due to some dependency violations and you really need to do this as the last resort, you can override dependency using dpkg's "--ignore-depends", "--force-depends" and other options. If you do this, you need to make serious effort to restore proper dependency later. See dpkg(8) for details. When your system is seriously broken, you should make a full backup of system to a safe place (see ) and should perform a clean installation. This is less time consuming and produces better results in the end.

If "/var/lib/dpkg/status" becomes corrupt for any reason, the Debian system loses package selection data and suffers severely. Look for the old "/var/lib/dpkg/status" file at "/var/lib/dpkg/status-old" or "/var/backups/dpkg.status.*". Keeping "/var/backups/" in a separate partition may be a good idea since this directory contains lots of important system data. For serious breakage, I recommend to make fresh re-install after making backup of the system. Even if everything in "/var/" is gone, you can still recover some information from directories in "/usr/share/doc/" to guide your new installation. Reinstall minimal (desktop) system. # mkdir -p /path/to/old/system Mount old system at "/path/to/old/system/". # cd /path/to/old/system/usr/share/doc # ls -1 >~/ls1.txt # cd /usr/share/doc # ls -1 >>~/ls1.txt # cd # sort ls1.txt | uniq | less Then you are presented with package names to install. (There may be some non-package names such as "texmf".)

You can seek packages which satisfy your needs with aptitude from the package description or from the list under "Tasks". When you encounter more than 2 similar packages and wonder which one to install without "trial and error" efforts, you should use some common sense. I consider following points are good indications of preferred packages. Essential: yes > no Component: main > contrib > non-free Priority: required > important > standard > optional > extra Tasks: packages listed in tasks such as "Desktop environment" Packages selected by the dependency package (e.g., python2.4 by python) Popcon: higher in the vote and install number Changelog: regular updates by the maintainer BTS: No RC bugs (no critical, no grave, and no serious bugs) BTS: responsive maintainer to bug reports BTS: higher number of the recently fixed bugs BTS: lower number of remaining non-wishlist bugs Debian being a volunteer project with distributed development model, its archive contains many packages with different focus and quality. You must make your own decision what to do with them.

Installing packages from mixed source of archives is not supported by the official Debian distribution except for officially supported particular combinations of archives such as stable with security updates and volatile updates. Here is an example of operations to include specific newer upstream version packages found in unstable while tracking testing for single occasion. Change the "/etc/apt/sources.list" file temporarily to single "unstable" entry. Run "aptitude update". Run "aptitude install <package-name>". Recover the original "/etc/apt/sources.list" file for testing. Run "aptitude update". You do not create the "/etc/apt/preferences" file nor need to worry about apt-pinning with this manual approach. But this is very cumbersome. When using mixed source of archives, you must ensure compatibility of packages by yourself since the Debian does not guarantee it. If package incompatibility exists, you may break system. You must be able to judge these technical requirements. The use of mixed source of random archives is completely optional operation and its use is not something I encourage you to use. General rules for installing packages from different archives are followings. Non-binary packages ("Architecture: all") are safer to install. documentation packages: no special requirements interpreter program packages: compatible interpreter must be available Binary packages (non "Architecture: all") usually face many road blocks and unsafe to install. library version compatibility (including "libc") related utility program version compatibility Kernel ABI compatibility C++ ABI compatibility … In order to make a package to be safer to install, some commercial non-free binary program packages may be provided with completely statically linked libraries. You should still check ABI compatibility issues etc. for them. Except to avoid broken package for a short term, installing binary packages from officially unsupported archives is generally bad idea. This is true even if you use apt-pinning (see ). You should consider chroot or similar techniques (see ) to run programs from different archives.

In lenny, aptitude(8) has a bug for handling "/etc/apt/preferences" file. (Bug#514930) Without the "/etc/apt/preferences" file, APT system choses the latest available version as the candidate version using the version string. This is the normal state and most recommended usage of APT system. All officially supported combinations of archives do not require the "/etc/apt/preferences" file since some archives which should not be used as the automatic source of upgrades are marked as NotAutomatic and dealt properly. The version string comparison rule can be verified with, e.g., "dpkg --compare-versions ver1.1 gt ver1.1~1; echo $?" (see dpkg(1)). When you install packages from mixed source of archives (see ) regularly, you can automate these complicated operations by creating the "/etc/apt/preferences" file with proper entries and tweaking the package selection rule for candidate version as described in apt_preferences(5). This is called apt-pinning. Use of apt-pinning by a novice user is sure call for major troubles. You must avoid using apt-pinning except when you absolutely need it. When using apt-pinning, you must ensure compatibility of packages by yourself since the Debian does not guarantee it. The apt-pinning is completely optional operation and its use is not something I encourage you to use. Archive level Release files (see ) are used for the rule of apt_preferences(5). Thus apt-pinning works only with "suite" name for normal Debian archives and security Debian archives. (This is different from Ubuntu archives). For example, you can do "Pin: release a=unstable" but can not do "Pin: release a=sid" in the "/etc/apt/preferences" file. When you use non-Debian archive as a part of apt-pinning, you should check what they are intended for and also check their credibility. For example, Ubuntu and Debian are not meant to be mixed. Even if you do not create the "/etc/apt/preferences" file, you can do fairly complex system operations (see and ) without apt-pinning. Here is a simplified explanation of apt-pinning technique. APT system choses highest Pin-Priority upgrading package from available package sources defined in the "/etc/apt/sources.list" file as the candidate version package. If the Pin-Priority of the package is larger than 1000, this version restriction for upgrading is dropped to enable downgrading (see ). Pin-Priority value of each package is defined by "Pin-Priority" entries in the "/etc/apt/preferences" file or uses its default value. default Pin-Priority package source type 990 target release archive 500 normal archive 100 installed package 1 NotAutomatic archive The target release archive can be set by several methods. "/etc/apt/apt.conf" configuration file with "APT::Default-Release "stable";" line command line option, e.g., "apt-get install -t testing some-package" The NotAutomatic archive is set by archive server having its archive level Release file (see ) containing "NotAutomatic: yes". The apt-pinning situation of <package> from multiple archive sources is displayed by "apt-cache policy <package>". A line started with "Package pin:" lists the package version of pin if association just with <package> is defined, e.g., "Package pin: 0.190". No line with "Package pin:" exists if no association just with <package> is defined. The Pin-Priority value associated just with <package> is listed right side of all version strings, e.g., "0.181 700". "0" is listed right side of all version strings if no association just with <package> is defined, e.g., "0.181 0". The Pin-Priority values of archives (defined as "Package: *" in the "/etc/apt/preferences" file) are listed left side of all archive paths, e.g., "200 http://backports.org etch-backports/main Packages". Here is an example of apt-pinning technique to include specific newer upstream version packages found in unstable regularly upgraded while tracking testing. You list all required archives in the "/etc/apt/sources.list" file as the following. deb http://ftp.us.debian.org/debian/ testing main contrib non-free deb http://ftp.us.debian.org/debian/ unstable main contrib non-free deb http://security.debian.org/ testing/updates main contrib Set the "/etc/apt/preferences" file as as the following. Package: * Pin: release a=testing Pin-Priority: 500 Package: * Pin: release a=unstable Pin-Priority: 200 When you wish to install a package named "<package-name>" with its dependencies from unstable archive under this configuration, you issue the following command which switches target release with "-t" option (Pin-Priority of unstable becomes 990.). $ sudo apt-get install -t unstable <package-name> With this configuration, usual execution of "apt-get upgrade" and "apt-get dist-upgrade" (or "aptitude safe-upgrade" and "aptitude full-upgrade" for squeeze) upgrades packages which were installed from testing archive using current testing archive and packages which were installed from unstable archive using current unstable archive. Be careful not to remove "testing" entry from the "/etc/apt/sources.list" file. Without "testing" entry in it, APT system upgrades packages using newer unstable archive. I usually edit the "/etc/apt/sources.list" file to comment out "unstable" archive entry right after above operation. This avoids slow update process of having too many entries in the "/etc/apt/sources.list" file although this prevents upgrading packages which were installed from unstable archive using current unstable archive. If "Pin-Priority: 20" is used instead of "Pin-Priority: 200" for the "/etc/apt/preferences" file, already installed packages having Pin-Priority value of 100 are not upgraded by unstable archive even if "testing" entry in the "/etc/apt/sources.list" file is removed. If you wish to track particular packages in unstable automatically without initial "-t unstable" installation, you must create the "/etc/apt/preferences" file and explicitly list all those packages at the top of it as the following. Package: <package-1> Pin: release a=unstable Pin-Priority: 700 Package: <package-2> Pin: release a=unstable Pin-Priority: 700 These set Pin-Priority value for each specific package. For example, in order to track the latest unstable version of this "Debian Reference" in English, you should have following entries in the "/etc/apt/preferences" file. Package: debian-reference-en Pin: release a=unstable Pin-Priority: 700 Package: debian-reference-common Pin: release a=unstable Pin-Priority: 700 This apt-pinning technique is valid even when you are tracking stable archive. Documentation packages have been always safe to install from unstable archive in my experience, so far. Here is another example of apt-pinning technique to include specific newer upstream version packages found in experimental while tracking unstable. You list all required archives in the "/etc/apt/sources.list" file as the following. deb http://ftp.us.debian.org/debian/ unstable main contrib non-free deb http://ftp.us.debian.org/debian/ experimental main contrib non-free deb http://security.debian.org/ testing/updates main contrib The default Pin-Priority value for experimental archive is always 1 (<<100) since it is NotAutomatic archive (see ). There is no need to set Pin-Priority value explicitly in the "/etc/apt/preferences" file just to use experimental archive unless you wish to track particular packages in it automatically for next upgrading.

There are debian-volatile project and backports.org archives which provide updgrade packages for stable. Do not use all packages available in the NotAutomatic archives such as &codename-stable;-backports and volatile-sloppy. Use only selected packages which fits your needs. backports.org is a non-Debian archive, although its packages are signed by Debian developers. Archive level Release files (see ) are used for the rule of apt_preferences(5). Thus apt-pinning works only with "code" name for volatile Debian archives. This is different from other Debian archives. For example, you can do "Pin: release a=&codename-stable;" but can not do "Pin: release a=stable" in the "/etc/apt/preferences" file for volatile Debian archives. Here is an example of apt-pinning technique to include specific newer upstream version packages found in &codename-stable;-backports while tracking &codename-stable; and volatile. You list all required archives in the "/etc/apt/sources.list" file as the following. deb http://ftp.us.debian.org/debian/ &codename-stable; main contrib non-free deb http://security.debian.org/ &codename-stable;/updates main contrib deb http://volatile.debian.org/debian-volatile/ &codename-stable;/volatile main contrib non-free deb http://volatile.debian.org/debian-volatile/ &codename-stable;/volatile-sloppy main contrib non-free deb http://backports.org/debian/ &codename-stable;-backports main contrib non-free The default Pin-Priority value for backports.org and volatile-sloppy archives are always 1 (<<100) since they are NotAutomatic archive (see ). There is no need to set Pin-Priority value explicitly in the "/etc/apt/preferences" file just to use for backports.org and volatile-sloppy archive unless you wish to track packages automatically for next upgrading. So whenever you wish to install a package named "<package-name>" with its dependency from &codename-stable;-backports archive, you use following command while switching target release with "-t" option. $ sudo apt-get install -t &codename-stable;-backports <package-name> If you wish to upgrade particular packages, you must create the "/etc/apt/preferences" file and explicitly lists all packages in it as the following. Package: <package-1> Pin: release o=Backports.org archive Pin-Priority: 700 Package: <package-2> Pin: release o=volatile.debian.org Pin-Priority: 700 Alternatively, with the "/etc/apt/preferences" file as the following. Package: * Pin: release a=stable , o=Debian Pin-Priority: 500 Package: * Pin: release a=&codename-stable;, o=volatile.debian.org Pin-Priority: 500 Package: * Pin: release a=&codename-stable;-backports, o=Backports.org archive Pin-Priority: 200 Package: * Pin: release a=&codename-stable;-sloppy, o=volatile.debian.org Pin-Priority: 200 Execution of "apt-get upgrade" and "apt-get dist-upgrade" (or "aptitude safe-upgrade" and "aptitude full-upgrade" for squeeze) upgrades packages which were installed from stable archive using current stable archive and packages which were installed from other archives using current corresponding archive for all archives in the "/etc/apt/sources.list" file.

The apt package comes with its own cron script "/etc/cron.daily/apt" to support the automatic download of packages. This script can be enhanced to perform the automatic upgrade of packages by installing the unattended-upgrades package. These can be customized by parameters in "/etc/apt/apt.conf.d/02backup" and "/etc/apt/apt.conf.d/50unattended-upgrades" as described in "/usr/share/doc/unattended-upgrades/README". The unattended-upgrades package is mainly intended for the security upgrade for the stable system. If the risk of breaking an existing stable system by the automatic upgrade is smaller than that of the system broken by the intruder using its security hole which has been closed by the security update, you should consider using this automatic upgrade with configuration parameters as the following. APT::Periodic::Update-Package-Lists "1"; APT::Periodic::Download-Upgradeable-Packages "1"; APT::Periodic::Unattended-Upgrade "1"; If you are running an unstable system, you do not want to use the automatic upgrade since it certainly breaks system some day. Even for such unstable case, you may still want to download packages in advance to save time for the interactive upgrade with configuration parameters as the following. APT::Periodic::Update-Package-Lists "1"; APT::Periodic::Download-Upgradeable-Packages "1"; APT::Periodic::Unattended-Upgrade "0";

If you want to limit the download bandwidth for APT to e.g. 800Kib/sec (=100kiB/sec), you should configure APT with its configuration parameter as the following. APT::Acquire::http::Dl-Limit "800";

Downgrading is not officially supported by the Debian by design. It should be done only as a part of emergency recovery process. Despite of this situation, it is known to work well in many incidents. For critical systems, You should backup all important data on the system after the recovery operation and re-install the new system from the scratch. You may be lucky to downgrade from newer archive to older archive to recover from broken system upgrade by manipulating candidate version (see ). This is lazy alternative to tedious actions of many "dpkg -i <broken-package>_<old-version>.deb" commands (see ). Search lines in the "/etc/apt/sources.list" file tracking unstable as the following. deb http://ftp.us.debian.org/debian/ &codename-unstable; main contrib non-free Replace it with the following to track testing. deb http://ftp.us.debian.org/debian/ &codename-testing; main contrib non-free Set the "/etc/apt/preferences" file as the following. Package: * Pin: release a=testing Pin-Priority: 1010 Run "apt-get dist-upgrade" to force downgrading of packages across the system. Remove this special "/etc/apt/preferences" file after this emergency downgrading. It is good idea to remove (not purge!) as much packages to minimize dependency problems. You may need to manually remove and install some packages to get system downgraded. Linux kernel, bootloader, udev, PAM, APT, and networking related packages and their configuration files require special attention.

Although the maintainer name listed in "/var/lib/dpkg/available" and "/usr/share/doc/package_name/changelog" provide some information on "who is behind the packaging activity", the actual uploader of the package is somewhat obscure. who-uploads(1) in the devscripts package identifies the actual uploader of Debian source packages.

If you are to compile a program from source to replace the Debian package, it is best to make it into a real local debianized package (*.deb) and use private archive. If you chose to compile a program from source and to install them under "/usr/local" instead, you may need to use equivs as a last resort to satisfy the missing package dependency. Package: equivs Priority: extra Section: admin Description: Circumventing Debian package dependencies This is a dummy package which can be used to create Debian packages, which only contain dependency information.

For partial upgrades of the stable system, rebuilding a package within its environment using the source package is desirable. This avoids massive package upgrades due to their dependencies. Add the following entries to the "/etc/apt/sources.list" of a stable system. deb-src http://http.us.debian.org/debian unstable main contrib non-free Install required packages for the compilation and download the source package as the following. # apt-get update # apt-get dist-upgrade # apt-get install fakeroot devscripts build-essential $ apt-get build-dep foo $ apt-get source foo $ cd foo* Adjust installed packages if needed. Execute the following. $ dch -i Bump package version, e.g. one appended with "+bp1" in "debian/changelog" Build packages and install them to the system as the following. $ debuild $ cd .. # debi foo*.changes

Since mirroring whole subsection of Debian archive wastes disk space and network bandwidth, deployment of a local proxy server for APT is desirable consideration when you administer many systems on LAN. APT can be configure to use generic web (http) proxy servers such as squid (see ) as described in apt.conf(5) and in "/usr/share/doc/apt/examples/configure-index.gz". The "$http_proxy" environment variable can be used to override proxy server setting in the "/etc/apt/apt.conf" file. There are proxy tools specially for Debian archive. You should check BTS before using them. package popcon size description DUMMY DUMMY DUMMY caching proxy server for Debian archive files (compiled OCaml program) DUMMY DUMMY DUMMY Debian archive proxy and partial mirror builder (Python program) DUMMY DUMMY DUMMY Caching proxy for Debian package and source files (Perl program) DUMMY DUMMY DUMMY Caching proxy for distribution of software packages (compiled C++ program) DUMMY DUMMY DUMMY Bittorrent proxy for downloading Debian packages (Python program) When Debian reorganizes its archive structure, these specialized proxy tools tend to require code rewrites by the package maintainer and may not be functional for a while. On the other hand, generic web (http) proxy servers are more robust and easier to cope with such changes.

Here is an example for creating a small public package archive compatible with the modern secure APT system (see ). Let's assume few things. Account name: "foo" Host name: "www.example.com" Required packages: apt-utils, gnupg, and other packages URL: "http://www.example.com/~foo/" ( → "/home/foo/public_html/index.html") Architecture of packages: "amd64" Create an APT archive key of Foo on your server system as the following. $ ssh foo@www.example.com $ gpg --gen-key ... $ gpg -K ... sec 1024D/3A3CB5A6 2008-08-14 uid Foo (ARCHIVE KEY) <foo@www.example.com> ssb 2048g/6856F4A7 2008-08-14 $ gpg --export -a 3A3CB5A6 >foo.public.key Publish the archive key file "foo.public.key" with the key ID "3A3CB5A6" for Foo Create an archive tree called "Origin: Foo" as the following. $ umask 022 $ mkdir -p ~/public_html/debian/pool/main $ mkdir -p ~/public_html/debian/dists/unstable/main/binary-amd64 $ mkdir -p ~/public_html/debian/dists/unstable/main/source $ cd ~/public_html/debian $ cat > dists/unstable/main/binary-amd64/Release << EOF Archive: unstable Version: 4.0 Component: main Origin: Foo Label: Foo Architecture: amd64 EOF $ cat > dists/unstable/main/source/Release << EOF Archive: unstable Version: 4.0 Component: main Origin: Foo Label: Foo Architecture: source EOF $ cat >aptftp.conf <<EOF APT::FTPArchive::Release { Origin "Foo"; Label "Foo"; Suite "unstable"; Codename "sid"; Architectures "amd64"; Components "main"; Description "Public archive for Foo"; }; EOF $ cat >aptgenerate.conf <<EOF Dir::ArchiveDir "."; Dir::CacheDir "."; TreeDefault::Directory "pool/"; TreeDefault::SrcDirectory "pool/"; Default::Packages::Extensions ".deb"; Default::Packages::Compress ". gzip bzip2"; Default::Sources::Compress "gzip bzip2"; Default::Contents::Compress "gzip bzip2"; BinDirectory "dists/unstable/main/binary-amd64" { Packages "dists/unstable/main/binary-amd64/Packages"; Contents "dists/unstable/Contents-amd64"; SrcPackages "dists/unstable/main/source/Sources"; }; Tree "dists/unstable" { Sections "main"; Architectures "amd64 source"; }; EOF You can automate repetitive updates of APT archive contents on your server system by configuring dupload. Place all package files into "~foo/public_html/debian/pool/main/" by executing "dupload -t foo changes_file" in client while having "~/.dupload.conf" containing the following. $cfg{'foo'} = { fqdn => "www.example.com", method => "scpb", incoming => "/home/foo/public_html/debian/pool/main", # The dinstall on ftp-master sends emails itself dinstall_runs => 1, }; $cfg{'foo'}{postupload}{'changes'} = " echo 'cd public_html/debian ; apt-ftparchive generate -c=aptftp.conf aptgenerate.conf; apt-ftparchive release -c=aptftp.conf dists/unstable >dists/unstable/Release ; rm -f dists/unstable/Release.gpg ; gpg -u 3A3CB5A6 -bao dists/unstable/Release.gpg dists/unstable/Release'| ssh foo@www.example.com 2>/dev/null ; echo 'Package archive created!'"; The postupload hook script initiated by dupload(1) creates updated archive files for each upload. You can add this small public archive to the apt-line of your client system by the following. $ sudo bash # echo "deb http://www.example.com/~foo/debian/ unstable main" \ >> /etc/apt/sources.list # apt-key add foo.public.key If the archive is located on the local filesystem, you can use "deb file:///home/foo/debian/ …" instead.

You can make a local copy of the package and debconf selection states by the following. # dpkg --get-selections '*' > selection.dpkg # debconf-get-selections > selection.debconf Here, "*" makes "selection.dpkg" to include package entries for "purge" too. You can transfer these 2 files to another computer, and install there with the following. # dselect update # debconf-set-selections < myselection.debconf # dpkg --set-selections < myselection.dpkg # apt-get -u dselect-upgrade # or dselect install If you are thinking about managing many servers in a cluster with practically the same configuration, you should consider to use specialized package such as fai to manage the whole system.

alien(1) enables the conversion of binary packages provided in Red Hat rpm, Stampede slp, Slackware tgz, and Solaris pkg file formats into a Debian deb package. If you want to use a package from another Linux distribution than the one you have installed on your system, you can use alien to convert it from your preferred package format and install it. alien also supports LSB packages. alien(1) should not be used to replace essential system packages, such as sysvinit, libc6, libpam-modules, etc. Practically, alien(1) should only used for non-free binary-only packages which are LSB compliant or statically linked. For free softwares, you should use their source packages to make real Debian packages.

The current "*.deb" package contents can be extracted without using dpkg(1) on any Unix-like environment using standard ar(1) and tar(1). # ar x /path/to/dpkg_<version>_<arch>.deb # ls total 24 -rw-r--r-- 1 bozo bozo 1320 2007-05-07 00:11 control.tar.gz -rw-r--r-- 1 bozo bozo 12837 2007-05-07 00:11 data.tar.gz -rw-r--r-- 1 bozo bozo 4 2007-05-07 00:11 debian-binary # mkdir control # mkdir data # tar xvzf control.tar.gz -C control # tar xvzf data.tar.gz -C data You can also browse package content using the mc command.

You can learn more on the package management from following documentations. Primary documentations on the package management: aptitude(8), dpkg(1), tasksel(8), apt-get(8), apt-config(8), apt-key(8), sources.list(5), apt.conf(5), and apt_preferences(5); "/usr/share/doc/apt-doc/guide.html/index.html" and "/usr/share/doc/apt-doc/offline.html/index.html" from the apt-doc package; and "/usr/share/doc/aptitude/html/en/index.html" from the aptitude-doc-en package. Official and detailed documentations on the Debian archive: "Debian Policy Manual Chapter 2 - The Debian Archive", "Debian Developer's Reference, Chapter 4 - Resources for Debian Developers 4.6 The Debian archive", and "The Debian GNU/Linux FAQ, Chapter 5 - The Debian FTP archives". Tutorial for building of a Debian package for Debian users: "Debian New Maintainers' Guide".

It is wise for you as the system administrator to know roughly how the Debian system is started and configured. Although the exact details are in the source files of the packages installed and their documentations, it is a bit overwhelming for most of us. I did my best to provide a quick overview of the key points of the Debian system and their configuration for your reference, based on the current and previous knowledge of mine and others. Since the Debian system is a moving target, the situation over the system may have been changed. Before making any changes to the system, you should refer to the latest documentation for each package.

The computer system undergoes several phases of boot strap processes from the power-on event until it offers the fully functional operating system (OS) to the user. For simplicity, I limit discussion to the typical PC platform with the default installation. The typical boot strap process is like a four-stage rocket. Each stage rocket hands over the system control to the next stage one. Of course, these can be configured differently. For example, if you compiled your own kernel, you may be skipping the step with the mini-Debian system. So please do not assume this is the case for your system until you check it yourself. For non-legacy PC platform such as the SUN or the Macintosh system, the BIOS on ROM and the partition on the disk may be quite different (). Please seek the platform specific documentations elsewhere for such a case.

The BIOS is the 1st stage of the boot process which is started by the power-on event. The BIOS residing on the read only memory (ROM) is executed from the particular memory address to which the program counter of CPU is initialized by the power-on event. This BIOS performs the basic initialization of the hardware (POST: power on self test) and hands the system control to the next step which you provide. The BIOS is usually provided with the hardware. The BIOS startup screen usually indicates what key(s) to press to enter the BIOS setup screen to configure the BIOS behavior. Popular keys used are F1, F2, F10, Esc, Ins, and Del. If your BIOS startup screen is hidden by a nice graphics screen, you may press some keys such as Esc to disable this. These keys are highly dependent on the hardware. The hardware location and the priority of the code started by the BIOS can be selected from the BIOS setup screen. Typically, the first few sectors of the first found selected device (hard disk, floppy disk, CD-ROM, …) are loaded to the memory and this initial code is executed. This initial code can be any one of the following. The boot loader code The kernel code of the stepping stone OS such as FreeDOS The kernel code of the target OS if it fits in this small space Typically, the system is booted from the specified partition of the primary hard disk partition. First 2 sectors of the hard disk on legacy PC contain the master boot record (MBR). The disk partition information including the boot selection is recorded at the end of this MBR. The first boot loader code executed from the BIOS occupies the rest of this MBR.

The boot loader is the 2nd stage of the boot process which is started by the BIOS. It loads the system kernel image and the initrd image to the memory and hands control over to them. This initrd image is the root filesystem image and its support depends on the bootloader used. The Debian system normally uses the Linux kernel as the default system kernel. The initrd image for the current 2.6 Linux kernel is technically the initramfs (initial RAM filesystem) image. The initramfs image is a gzipped cpio archive of files in the root filesystem. The default install of the Debian system places first-stage GRUB boot loader code into the MBR for the PC platform. There are many boot loaders and configuration options available. bootloader package popcon size initrd description GRUB Legacy DUMMY DUMMY DUMMY Supported This is smart enough to understand disk partitions and filesystems such as vfat, ext3, …. (lenny default) GRUB 2 DUMMY DUMMY DUMMY Supported This is smart enough to understand disk partitions and filesystems such as vfat, ext3, …. GRUB 2 DUMMY DUMMY DUMMY Supported This is GRUB 2 bootable rescue images (CD and floppy) (PC/BIOS version) Lilo DUMMY DUMMY DUMMY Supported This relies on the sector locations of data on the hard disk. (Old) Isolinux DUMMY DUMMY DUMMY Supported This understands the ISO9660 filesystem. This is used by the boot CD. Syslinux DUMMY DUMMY DUMMY Supported This understands the MSDOS filesystem (FAT). This is used by the boot floppy. Loadlin DUMMY DUMMY DUMMY Supported New system is started from the FreeDOS/MSDOS system. MBR by Neil Turton DUMMY DUMMY DUMMY Not supported This is free software which substitutes MSDOS MBR. This only understands disk partitions. Do not play with boot loaders without having bootable rescue media (CD or floppy) created from images in the grub-rescue-pc package. It makes you boot your system even without functioning bootloader on the hard disk. For GRUB Legacy, the menu configuration file is located at "/boot/grub/menu.lst". For example, it has entries as the following. title Debian GNU/Linux root (hd0,2) kernel /vmlinuz root=/dev/hda3 ro initrd /initrd.img For GRUB 2, the menu configuration file is located at "/boot/grub/grub.cfg". It is automatically generated by "/usr/sbin/update-grub" using templates from "/etc/grub.d/*" and settings from "/etc/default/grub". For example, it has entries as the following. menuentry "Debian GNU/Linux" { set root=(hd0,3) linux /vmlinuz root=/dev/hda3 initrd /initrd.img } For these examples, these GRUB parameters mean the following. GRUB parameter meaning root use 3rd partition on the primary disk by setting it as "(hd0,2)" in GRUB legacy or as "(hd0,3)" in GRUB 2 kernel use kernel located at "/vmlinuz" with kernel parameter: "root=/dev/hda3 ro" initrd use initrd/initramfs image located at "/initrd.img" The value of the partition number used by GRUB legacy program is one less than normal one used by Linux kernel and utility tools. GRUB 2 program fixes this problem. UUID (see ) may be used to identify a block special device instead of its file name such as "/dev/hda3", e.g."root=UUID=81b289d5-4341-4003-9602-e254a17ac232 ro". You can start a boot loader from another boot loader using techniques called chain loading. See "info grub" and grub-install(8).

The mini-Debian system is the 3rd stage of the boot process which is started by the boot loader. It runs the system kernel with its root filesystem on the memory. This is an optional preparatory stage of the boot process. The term "the mini-Debian system" is coined by the author to describe this 3rd stage boot process for this document. This system is commonly referred as the initrd or initramfs system. Similar system on the memory is used by the Debian Installer. The "/init" script is executed as the first program in this root filesystem on the memory. It is a shell script program which initializes the kernel in user space and hands control over to the next stage. This mini-Debian system offers flexibility to the boot process such as adding kernel modules before the main boot process or mounting the root filesystem as an encrypted one. You can interrupt this part of the boot process to gain root shell by providing "break=init" etc. to the kernel boot parameter. See the "/init" script for more break conditions. This shell environment is sophisticated enough to make a good inspection of your machine's hardware. Commands available in this mini-Debian system are stripped down ones and mainly provided by a GNU tool called busybox(1). You need to use "-n" option for mount command when you are on the readonly root filesystem.

This section describes classical System V style boot system on lenny. Debian is moving to the event driven boot system. See The future of the boot system in Debian and Dependency based boot sequence. The normal Debian system is the 4th stage of the boot process which is started by the mini-Debian system. The system kernel for the mini-Debian system continues to run in this environment. The root filesystem is switched from the one on the memory to the one on the real hard disk filesystem. The "/sbin/init" program is executed as the first program and performs the main boot process. The Debian normally uses the traditional sysvinit scheme with the sysv-rc package. See init(8), inittab(5), and "/usr/share/doc/sysv-rc/README.runlevels.gz" for the exact explanation. This main boot process essentially goes through the following. The Debian system goes into runlevel N (none) to initialize the system by following the "/etc/inittab" description. The Debian system goes into runlevel S to initialize the system under the single-user mode to complete hardware initialization etc. The Debian system goes into one of the specified multi-user runlevels (2 to 5) to start the system services. The initial runlevel used for multi-user mode is specified with the "init=" kernel boot parameter or in the "initdefault" line of the "/etc/inittab". The Debian system as installed starts at the runlevel 2. All actual script files executed by the init system are located in the directory "/etc/init.d/". For alternative boot mechanism to the sysv-rc package using a single configuration file "/etc/runlevel.conf", see the file-rc package. Both mechanisms are compatible through "/etc/init.d/rc", "/etc/init.d/rcS", "/usr/sbin/update-rc.d", and "/usr/sbin/invoke-rc.d" scripts.

Each runlevel uses a directory for its configuration and has specific meaning as the following. runlevel directory description of runlevel usage N none system bootup (NONE) level (no "/etc/rcN.d/" directory) 0 /etc/rc0.d/ halt the system S /etc/rcS.d/ single-user mode on boot (alias: "s") 1 /etc/rc1.d/ single-user mode switched from multi-user mode 2 /etc/rc2.d/ multi-user mode 3 /etc/rc3.d/ ,, 4 /etc/rc4.d/ ,, 5 /etc/rc5.d/ ,, 6 /etc/rc6.d/ reboot the system 7 /etc/rc7.d/ valid multi-user mode but not normally used 8 /etc/rc8.d/ ,, 9 /etc/rc9.d/ ,, You can change the runlevel from the console to, e.g., 4 by the following. $ sudo telinit 4 The Debian system does not pre-assign any special meaning differences among the runlevels between 2 and 5. The system administrator on the Debian system may change this. (I.e., Debian is not Red Hat Linux nor Solaris by Sun Microsystems nor HP-UX by Hewlett Packard nor AIX by IBM nor …) The Debian system does not populate directories for the runlevels between 7 and 9 when the package is installed. Traditional Unix variants don’t use these runlevels.

The name of the symlink in each runlevel directory has the form "S<2-digit-number><original-name>" or "K<2-digit-number><original-name>". The 2-digit-number is used to determine the order in which to run the scripts. "S" is for "Start" and "K" is for "Kill". When init(8) or telinit(8) commands goes into the runlevel to "<n>", it execute following scripts. The script names starting with a "K" in "/etc/rc<n>.d/" are executed in alphabetical order with the single argument "stop". (killing services) The script names starting with an "S" in "/etc/rc<n>.d/" are executed in alphabetical order with the single argument "start". (starting services) For example, if you had the links "S10sysklogd" and "S20exim4" in a runlevel directory, "S10sysklogd" which is symlinked to "../init.d/sysklogd" would run before "S20exim4" which is symlinked to "../init.d/exim4". It is not advisable to make any changes to symlinks in "/etc/rcS.d/" unless you know better than the maintainer.

For example, let's set up runlevel system somewhat like Red Hat Linux as the following. init starts the system in runlevel=3 as the default. init does not start gdm(1) in runlevel=(0,1,2,6). init starts gdm(1) in runlevel=(3,4,5). This can be done by using editor on the "/etc/inittab" file to change starting runlevel and using user friendly runlevel management tools such as sysv-rc-conf or bum to edit the runlevel. If you are to use command line only instead, here is how you do it (after the default installation of the gdm package and selecting it to be the choice of display manager). # cd /etc/rc2.d ; mv S21gdm K21gdm # cd /etc ; perl -i -p -e 's/^id:.:/id:3:/' inittab Please note the "/etc/X11/default-display-manager" file is checked when starting the display manager daemons: xdm, gdm, kdm, and wdm. You can still start X from any console shell with the startx(1) command.

The default parameter for each init script in "/etc/init.d/" is given by the corresponding file in "/etc/default/" which contains environment variable assignments only. This choice of directory name is specific to the Debian system. It is roughly the equivalent of the "/etc/sysconfig" directory found in Red Hat Linux and other distributions. For example, "/etc/default/cron" can be used to control how "/etc/init.d/cron" works. The "/etc/default/rcS" file can be used to customize boot-time defaults for motd(5), sulogin(8), etc. If you cannot get the behavior you want by changing such variables then you may modify the init scripts themselves. These are configuration files editable by system administrators.

The kernel maintains the system hostname. The init script in runlevel S which is symlinked to "/etc/init.d/hostname.sh" sets the system hostname at boot time (using the hostname command) to the name stored in "/etc/hostname". This file should contain only the system hostname, not a fully qualified domain name. To print out the current hostname run hostname(1) without an argument.

Although the root filesystem is mounted by the kernel when it is started, other filesystems are mounted in the runlevel S by the following init scripts. "`/etc/init.d/mountkernfs.sh" for kernel filesystems in "/proc", "/sys", etc. "`/etc/init.d/mountdevsubfs.sh" for virtual filesystems in "/dev" "`/etc/init.d/mountall.sh" for normal filesystems using "/etc/fstab" "`/etc/init.d/mountnfs.sh" for network filesystems using"/etc/fstab" The mount options of the filesystem are set in "/etc/fstab". See . The actual mounting of network filesystems waits for the start of the network interface. After mounting all the filesystems, temporary files in "/tmp", "/var/lock", and "/var/run" are cleaned for each boot up.

Network interfaces are initialized in runlevel S by the init script symlinked to "/etc/init.d/ifupdown-clean" and "/etc/init.d/ifupdown". See for how to configure them.

Many network services (see ) are started under multi-user mode directly as daemon processes at boot time by the init script, e.g., "/etc/rc2.d/S20exim4" (for RUNLEVEL=2) which is a symlink to "/etc/init.d/exim4". Some network services can be started on demand using the super-server inetd (or its equivalents). The inetd is started at boot time by "/etc/rc2.d/S20inetd" (for RUNLEVEL=2) which is a symlink to "/etc/init.d/inetd". Essentially, inetd allows one running daemon to invoke several others, reducing load on the system. Whenever a request for service arrives at super-server inetd , its protocol and service are identified by looking them up in the databases in "/etc/protocols" and "/etc/services". inetd then looks up a normal Internet service in the "/etc/inetd.conf" database, or a Open Network Computing Remote Procedure Call (ONC RPC)/Sun RPC based service in "/etc/rpc.conf". Sometimes, inetd does not start the intended server directly but starts the TCP wrapper program, tcpd(8), with the intended server name as its argument in "/etc/inetd.conf". In this case, tcpd runs the appropriate server program after logging the request and doing some additional checks using "/etc/hosts.deny" and "/etc/hosts.allow". For system security, disable as much network service programs as possible. See . See inetd(8), inetd.conf(5), protocols(5), services(5), tcpd(8), hosts_access(5), hosts_options(5), rpcinfo(8), portmap(8), and "/usr/share/doc/portmap/portmapper.txt.gz".

The system message can be customized by "/etc/default/syslogd" and "/etc/syslog.conf" for both the log file and on-screen display. See syslogd(8) and syslog.conf(5). See also .

The kernel message can be customized by "/etc/default/klogd" for both the log file and on-screen display. Set "KLOGD='-c 3'" in this file and run "/etc/init.d/klogd restart". See klogd(8). You may directly change the error message level by the following. # dmesg -n3 error level value error level name meaning 0 KERN_EMERG system is unusable 1 KERN_ALERT action must be taken immediately 2 KERN_CRIT critical conditions 3 KERN_ERR error conditions 4 KERN_WARNING warning conditions 5 KERN_NOTICE normal but significant condition 6 KERN_INFO informational 7 KERN_DEBUG debug-level messages

For Linux kernel 2.6, the udev system provides mechanism for the automatic hardware discovery and initialization (see udev(7)). Upon discovery of each device by the kernel, the udev system starts a user process which uses information from the sysfs filesystem (see ), loads required kernel modules supporting it using the modprobe(8) program (see ), and creates corresponding device nodes. If "/lib/modules/<kernel-version>/modules.dep" was not generated properly by depmod(8) for some reason, modules may not be loaded as expected by the udev system. Execute "depmod -a" to fix it. The name of device nodes can be configured by udev rule files in "/etc/udev/rules.d/". Current default rules tend to create dynamically generated names resulting non-static device names except for cd and network devices. By adding your custom rules similar to what cd and network devices do, you can generate static device names for other devices such as USB memory sticks, too. See "Writing udev rules" or "/usr/share/doc/udev/writing_udev_rules/index.html". Since the udev system is somewhat a moving target, I leave details to other documentations and describe the minimum information here. For mounting rules in "/etc/fstab", device nodes do not need to be static ones. You can use UUID to mount devices instead of device names such as "/dev/sda". See .

The modprobe(8) program enables us to configure running Linux kernel from user process by adding and removing kernel modules. The udev system (see ) automates its invocation to help the kernel module initialization. There are non-hardware modules and special hardware driver modules as the following which need to be pre-loaded by listing them in the "/etc/modules" file (see modules(5)). TUN/TAP modules providing virtual Point-to-Point network device (TUN) and virtual Ethernet network device (TAP), netfilter modules providing netfilter firewall capabilities (iptables(8), ), and watchdog timer driver modules. The configuration files for the modprobe(8) program are located under the "/etc/modprobes.d/" directory as explained in modprobe.conf(5). (If you want to avoid some kernel modules to be auto-loaded, consider to blacklist them in the "/etc/modprobes.d/blacklist" file.) The "/lib/modules/<version>/modules.dep" file generated by the depmod(8) program describes module dependencies used by the modprobe(8) program. If you experience module loading issues with boot time module loading or with modprobe(8), "depmod -a" may resolve these issues by reconstructing "modules.dep". The modinfo(8) program shows information about a Linux kernel module. The lsmod(8) program nicely formats the contents of the "/proc/modules", showing what kernel modules are currently loaded. You can identify exact hardware on your system. See . You may configure hardware at boot time to activate expected hardware features. See . You can add support for your device by recompiling kernel. See .

When a person (or a program) requests access to the system, authentication confirms the identity to be a trusted one. Configuration errors of PAM may lock you out of your own system. You must have a rescue CD handy or setup an alternative boot partition. To recover, boot the system with them and correct things from there.

Normal Unix authentication is provided by the pam_unix(8) module under the PAM (Pluggable Authentication Modules). Its 3 important configuration files, with ":" separated entries, are the following. file permission user group description /etc/passwd -rw-r--r-- root root (sanitized) user account information /etc/shadow -rw-r----- root shadow secure user account information /etc/group -rw-r--r-- root root group information "/etc/passwd" contains the following. ... user1:x:1000:1000:User1 Name,,,:/home/user1:/bin/bash user2:x:1001:1001:User2 Name,,,:/home/user2:/bin/bash ... As explained in passwd(5), each ":" separated entry of this file means the following. Login name Password specification entry Numerical user ID Numerical group ID User name or comment field User home directory Optional user command interpreter The second entry of "/etc/passwd" was used for the encrypted password entry. After the introduction of "/etc/shadow", this entry is used for the password specification entry. content meaning (empty) passwordless account x the encrypted password is in "/etc/shadow" * no login for this account ! no login for this account "/etc/shadow" contains the following. ... user1:$1$Xop0FYH9$IfxyQwBe9b8tiyIkt2P4F/:13262:0:99999:7::: user2:$1$vXGZLVbS$ElyErNf/agUDsm1DehJMS/:13261:0:99999:7::: ... As explained in shadow(5), each ":" separated entry of this file means the following. Login name Encrypted password (The initial "$1$" indicates use of the MD5 encryption. The "*" indicates no login.) Days since Jan 1, 1970 that password was last changed Days before password may be changed Days after which password must be changed Days before password is to expire that user is warned … "/etc/group" contains the following. group1:x:20:user1,user2 As explained in shadow(5), each ":" separated entry of this file means the following. Group name Encrypted password (not really used) Numerical group ID "," separated list of user names "/etc/gshadow" provides the similar function as "/etc/shadow" for "/etc/group" but is not really used. The actual group membership of a user may be dynamically added if "auth optional pam_group.so" line is added to "/etc/pam.d/common-auth" and set it in "/etc/security/group.conf". See pam_group(8). The base-passwd package contains an authoritative list of the user and the group: "/usr/share/doc/base-passwd/users-and-groups.html".

Here are few notable commands to manage account information. command function getent passwd <user_name> browse account information of "<user_name>" getent shadow <user_name> browse shadowed account information of "<user_name>" getent group <group_name> browse group information of "<group_name>" passwd manage password for the account passwd -e set one-time password for the account activation chage manage password aging information You may need to have the root privilege for some functions to work. See crypt(3) for the password and data encryption. On the system set up with PAM and NSS as the Debian alioth machine, the content of local "/etc/passwd", "/etc/group" and "/etc/shadow" may not be actively used by the system. Above commands are valid even under such environment.

When creating an account during your system installation or with the passwd(1) command, you should choose a good password which consists of 6 to 8 characters including one or more characters from each of the following sets according to passwd(1). Lower case alphabetics Digits 0 through 9 Punctuation marks Do not chose guessable words for the password.

There are independent tools to generate encrypted password with salt. package popcon size command function DUMMY DUMMY DUMMY mkpasswd over-featured front end to the crypt(3) library DUMMY DUMMY DUMMY openssl passwd compute password hashes (OpenSSL). passwd(1ssl)

Modern Unix-like systems such as the Debian system provide PAM (Pluggable Authentication Modules) and NSS (Name Service Switch) mechanism to the local system administrator to configure his system. The role of these can be summarizes as the following. PAM offers a flexible authentication mechanism used by the application software thus involves password data exchange. NSS offers a flexible name service mechanism which is frequently used by the C standard library to obtain the user and group name for programs such as ls(1) and id(1). These PAM and NSS systems need to be configured consistently. The notable packages of PAM and NSS systems are the following. package popcon size description DUMMY DUMMY DUMMY Pluggable Authentication Modules (basic service) DUMMY DUMMY DUMMY Pluggable Authentication Module allowing LDAP interfaces DUMMY DUMMY DUMMY Pluggable Authentication Module to enable cracklib support DUMMY DUMMY DUMMY Pluggable Authentication Modules (documentation in html and text) DUMMY DUMMY DUMMY GNU C Library: Shared libraries which also provides "Name Service Switch" service DUMMY DUMMY DUMMY GNU C Library: Manpages DUMMY DUMMY DUMMY GNU C Library: Reference manual in info, pdf and html format (non-free) DUMMY DUMMY DUMMY NSS module for Multicast DNS name resolution DUMMY DUMMY DUMMY NSS module for using LDAP as a naming service DUMMY DUMMY DUMMY NSS module for using LDAP as a naming service (new folk of libnss-ldap) "The Linux-PAM System Administrators' Guide" in libpam-doc is essential for learning PAM configuration. "System Databases and Name Service Switch" section in glibc-doc-reference is essential for learning NSS configuration. You can see more extensive and current list by "aptitude search 'libpam-|libnss-'" command. The acronym NSS may also mean "Network Security Service" which is different from "Name Service Switch". PAM is the most basic way to initialize environment variables for each program with the system wide default value.

Here are few notable configuration files accessed by the PAM. configuration file function /etc/pam.d/<program_name> set up PAM configuration for the "<program_name>" program; see pam(7) and pam.d(5) /etc/nsswitch.conf set up NSS configuration with the entry for each service. See nsswitch.conf(5) /etc/nologin limit the user login by the pam_nologin(8) module /etc/securetty limit the tty for the root access by the pam_securetty(8) module /etc/security/access.conf set access limit by the pam_access(8) module /etc/security/group.conf set group based restraint by the pam_group(8) module /etc/security/pam_env.conf set environment variables by the pam_env(8) module /etc/environment set additional environment variables by the pam_env(8) module with the "readenv=1" argument /etc/default/locale set locale by pam_env(8) module with the "readenv=1 envfile=/etc/default/locale" argument. (Debian) /etc/security/limits.conf set resource restraint (ulimit, core, …) by the pam_linits(8) module /etc/security/time.conf set time restraint by the pam_time(8) module The limitation of the password selection is implemented by the PAM modules, pam_unix(8) and pam_cracklib(8). They can be configured by their arguments. PAM modules use suffix ".so" for their filenames.

The modern centralized system management can be deployed using the centralized Lightweight Directory Access Protocol (LDAP) server to administer many Unix-like and non-Unix-like systems on the network. The open source implementation of the Lightweight Directory Access Protocol is OpenLDAP Software. The LDAP server provides the account information through the use of PAM and NSS with libpam-ldap and libnss-ldap packages for the Debian system. Several actions are required to enable this (I have not used this setup and the following is purely secondary information. Please read this in this context.). You set up a centralized LDAP server by running program such as stand-alone LDAP daemon, slapd(8). You change the PAM configuration files in the "/etc/pam.d/" directory to use "pam_ldap.so" instead of the default "pam_unix.so". Debian uses "/etc/pam_ldap.conf" as the configuration file for libpam-ldap and "/etc/pam_ldap.secret" as the file to store the password of the root. You change the NSS configuration in the "/etc/nsswitch.conf" file to use "ldap" instead of the default ("compat" or "file"). Debian uses "/etc/libnss-ldap.conf" as the configuration file for libnss-ldap. You must make libpam-ldap to use SSL (or TLS) connection for the security of password. You may make libnss-ldap to use SSL (or TLS) connection to ensure integrity of data at the cost of the LDAP network overhead. You should run nscd(8) locally to cache any LDAP search results in order to reduce the LDAP network traffic. See documentations in pam_ldap.conf(5) and "/usr/share/doc/libpam-doc/html/" offered by the libpam-doc package and "info libc 'Name Service Switch'" offered by the glibc-doc package. Similarly, you can set up alternative centralized systems with other methods. NIS (originally called YP) or NIS+ with older Unix-like systems Winbind with Windows NT and SAMBA

This is the famous phrase at the bottom of the old "info su" page by Richard M. Stallman. Not to worry: the current su command in Debian uses PAM, so that one can restrict the ability to use su to the root group by enabling the line with "pam_wheel.so" in "/etc/pam.d/su".

Installing the libpam-cracklib package enables you to force stricter password rule, for example, by having active lines in "/etc/pam.d/common-password" as the following. For lenny: password required pam_cracklib.so retry=3 minlen=9 difok=3 password required pam_unix.so use_authtok nullok md5 For squeeze: password required pam_cracklib.so retry=3 minlen=9 difok=3 password [success=1 default=ignore] pam_unix.so use_authtok nullok md5 password requisite pam_deny.so password required pam_permit.so

See for restricting the kernel secure attention key (SAK) feature.

sudo(8) is a program designed to allow a sysadmin to give limited root privileges to users and log root activity. sudo requires only an ordinary user's password. Install sudo package and activate it by setting options in "/etc/sudoers". See configuration example at "/usr/share/doc/sudo/examples/sudoers". My usage of sudo for the single user system (see ) is aimed to protect myself from my own stupidity. Personally, I consider using sudo a better alternative to using the system from the root account all the time. For example, the following changes the owner of "<some_file>" to "<my_name>". $ sudo chown <my_name> <some_file> Of course if you know the root password (as self-installed Debian users do), any command can be run under root from any user's account using "su -c".

Security-Enhanced Linux (SELinux) is a framework to tighten privilege model tighter than the ordinary Unix-like security model with the mandatory access control (MAC) policies. The root power may be restricted under some conditions.

For system security, It is a good idea to disable as much server programs as possible. This becomes critical for network servers. Having unused servers, activated either directly as daemon or via super-server program, are considered security risks. Many programs, such as sshd(8), use PAM based access control. There are many ways to restrict access to some server services. configuration files: "/etc/default/<program_name>" runlevel configuration for daemon PAM (Pluggable Authentication Modules) "/etc/inetd.conf" for super-server "/etc/hosts.deny" and "/etc/hosts.allow" for TCP wrapper, tcpd(8) "/etc/rpc.conf" for Sun RPC "/etc/at.allow" and "/etc/at.deny" for atd(8) "/etc/cron.allow" and "/etc/cron.deny" for crontab(1) Network firewall of netfilter infrastructure See , , , , and . Sun RPC services need to be active for NFS and other RPC based programs. If you have problems with remote access in a recent Debian system, comment out offending configuration such as "ALL: PARANOID" in "/etc/hosts.deny" if it exists. (But you must be careful on security risks involved with this kind of action.)

The information here may not be sufficient for your security needs but it should be a good start.

Many popular transportation layer services communicate messages including password authentication in the plain text. It is very bad idea to transmit password in the plain text over the wild Internet where it can be intercepted. You can run these services over "Transport Layer Security" (TLS) or its predecessor, "Secure Sockets Layer" (SSL) to secure entire communication including password by the encryption. insecure service name port secure service name port www (http) 80 https 443 smtp (mail) 25 ssmtp (smtps) 465 ftp-data 20 ftps-da