Reference Manual

Mandrakelinux 10.0

http://www.mandrakesoft.com
Reference Manual: Mandrakelinux 10.0
Published March 2004
Copyright © 2004 Mandrakesoft SA
by Camille Bégnis, Christian Roy, Fabian Mandelbaum, Roberto Rosselli del Turco, Marco De Vitis, Alice
Lafox, John Rye, Patricia Pichardo Bégnis, Wolfgang Bornath, Joël Wardenski, Debora Rejnharc Mandelbaum,
Daniel Gueysset, Mickael Scherer, Jean-Michel Dault, Funda Wang, Lunas Moon, Céline Harrand, Fred
Lepied, Pascal Rigaux, Thierry Vignaud, Giuseppe Ghibò, and Stew Benedict

Legal Notice

This manual (except for the parts listed in the table below) is protected under Mandrakesoft intellectual property rights. By reproducing,
duplicating or distributing this manual in whole or in part, you explicitly agree to conform to the terms and conditions of this license
agreement.
This manual (except for the chapters listed in the table below) may be freely reproduced, duplicated and distributed either as such or as
part of a bundled package in electronic and/or printed format provided however that the following conditions are fulfilled :

• That this copyright notice appears clearly and distinctively on all reproduced, duplicated and distributed
copies.
• That the “front cover texts” below, About Mandrakelinux, page 1 and the section stating the names of authors
and contributors are attached to the reproduced, duplicated or distributed version and remain unchanged.
• That this manual, specifically for the printed format, is reproduced and/or distributed for noncommercial
use only.
The express authorisation of Mandrakesoft SA must be obtained prior to any other use of any manual or part thereof.
“Mandrake”, “Mandrakesoft”, “DrakX” and “Linux-Mandrake” are registered Trademarks in US and/or other countries. The related
“Star logo” is also registered. All rights reserved. All other copyrights embodied in this document remain the property of their respective
owners.

Front-cover texts
Mandrakesoft May 2004

©
http://www.mandrakesoft.com/
Copyright 1999-2004 by Mandrakesoft S.A. and Mandrakesoft Inc.

The chapters listed in the table below are protected by a different

license. Consult the table and links for more details about these
licenses.

Original Copyright License

“Building and Installing Free Software”, page by Benjamin Drieu , APRIL GNU General Public License GPL (http:
71 (http://www.april.org/) //www.gnu.org/copyleft/gpl.html)

Tools Used in The Making of This Manual

This manual was written in XML DocBook. The set of files involved were managed using Borges
(http://linux-mandrake.com/en/doc/project/Borges/). The XML source files were processed by xsltproc, openjade and jadetex
using a customized version of Norman Walsh’s stylesheets. Screen shots were taken using xwd or GIMP and converted with convert. All
these programs are free and are available in your Mandrakelinux distribution.
Table of Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1. About Mandrakelinux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1. Contacting the Mandrakelinux Community . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. Join the Club . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3. Purchasing Mandrakesoft Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.4. Contribute to Mandrakelinux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3. Note from the Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Conventions Used in this Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4.1. Typing Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4.2. General Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
I. The Linux System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1. Basic UNIX System Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1. Users and Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2. File Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3. Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4. A Short Introduction to the Command Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2. Disks and Partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1. Structure of a Hard Disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2. Conventions for Naming Disks and Partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3. Introduction to the Command Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1. File-Handling Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2. Handling File Attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3. Shell Globbing Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4. Redirections and Pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5. Command-Line Completion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.6. Starting and Handling Background Processes: Job Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.7. A Final Word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4. Text Editing: Emacs and VI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1. Emacs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2. Vi: the ancestor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3. A last word... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5. Command-Line Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.1. File Operations and Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2. find: Find Files According to Certain Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3. Commands Startup Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.4. Archiving and Data Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.5. Many, many more... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6. Process Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.1. More About Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.2. Information on Processes: ps and pstree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.3. Sending Signals to Processes: kill, killall and top . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.4. Setting Priority to Processes: nice, renice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
II. Linux in Depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7. File-Tree Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.1. Shareable/Unshareable, Static/Variable Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.2. The root Directory: / . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.3. /usr/: The Big One . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.4. /var/: Data Modifiable During Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.5. /etc/: Configuration Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
8. File Systems and Mount Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.1. Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.2. Partitioning a Hard Disk, Formatting a Partition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.3. The mount and umount Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
9. The Linux File System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
9.1. Comparison of a Few File Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
9.2. Everything is a File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
9.3. Links . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

iii
 9.4. “Anonymous” Pipes and Named Pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
9.5. “Special” Files: Character Mode and Block Mode Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
9.6. Symbolic Links, Limitation of “Hard” Links . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.7. File Attributes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
10. The /proc Filesystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10.1. Information About Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10.2. Information on The Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
10.3. The /proc/sys Sub-Directory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
11. The Start-Up Files: init sysv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.1. In the Beginning Was init . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.2. Runlevels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
III. Advanced Uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
12. Building and Installing Free Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
12.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
12.2. Decompression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
12.3. Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
12.4. Compilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
12.5. Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
12.6. Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
12.7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
13. Compiling and Installing New Kernels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
13.1. Upgrading a Kernel Using Binary Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
13.2. From The Kernel Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
13.3. Unpacking Sources, Patching the Kernel (if Necessary) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
13.4. Configuring The Kernel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
13.5. Saving, Reusing your Kernel Configuration Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
13.6. Compiling Kernel and Modules, Installing the Beast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
13.7. Installing the New Kernel Manually. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89
A. Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

iv
List of Tables
9-1. File System Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

List of Figures
1-1. Graphical Mode Login Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1-2. The Terminal Icon on the KDE Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2-1. First Example of Partition Naming under GNU/Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2-2. Second Example of Partition Naming under GNU/Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4-1. Editing Two Files at Once . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4-2. Emacs, before copying the text block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4-3. Copying Text with emacs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4-4. Starting position in VIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4-5. VIM, before copying the text block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4-6. VIM, after having copied the text block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6-1. Monitoring Processes with top . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
8-1. A Not Yet Mounted File System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8-2. File System Is Now Mounted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

v
vi
Preface

1. About Mandrakelinux
Mandrakelinux is a GNU/Linux distribution supported by Mandrakesoft S.A. which was born on the Internet in
1998. Its main goal was and still is to provide an easy-to-use and friendly GNU/Linux system. Mandrakesoft’s
two pillars are open source and collaborative work.

1.1. Contacting the Mandrakelinux Community

The following are various Internet links pointing you to various Mandrakelinux-related sources. If you wish
to know more about the Mandrakesoft company, connect to our web site (http://www.mandrakesoft.com/).
You can also check out the Mandrakelinux distribution web site (http://www.mandrakelinux.com/) and all its
derivatives.
Mandrakeexpert (http://www.mandrakeexpert.com/) is Mandrakesoft’s help platform. It offers a new expe-
rience based on trust and the pleasure of rewarding others for their contributions.
We also invite you to subscribe to the various mailing lists (http://www.mandrakelinux.com/en/flists.
php3), where the Mandrakelinux community demonstrates its vivacity and keenness.
Please also remember to connect to Mandrakesecure (http://www.mandrakesecure.net/). It gathers all
security-related material about Mandrakelinux distributions. You will find security and bug advisories, as well
as security and privacy-related articles. A must for any server administrator or user concerned about security.

1.2. Join the Club

Mandrakesoft offer a wide range of advantages though its Mandrakelinux Users Club (http://www.
mandrakelinux.com/en/club/):

• download commercial software normally only available in retail packs, such as special hardware drivers,
commercial applications, freeware, and demo versions;
• vote and propose new software through a volunteer-run RPM voting system;
• access more than 50,000 RPM packages for all Mandrakelinux distributions;
• obtain discounts for products and services on Mandrakestore (http://www.mandrakestore.com);
• access a better mirror list, exclusive to Club members;
• read multilingual forums and articles.
By financing Mandrakesoft through the Mandrakeclub you will directly enhance the Mandrakelinux distribution
and help us provide the best possible GNU/Linux desktop to our users.

1.3. Purchasing Mandrakesoft Products

Mandrakelinux users may purchase products on-line through the Mandrakestore (http://www.mandrakestore.
com/). You will not only find Mandrakelinux software, operating systems and “live” boot CDs (such as Man-
drakemove), but also special subscription offers, support, third-party software and licenses, documentation,
GNU/Linux-related books, as well as other Mandrakesoft goodies.

1.4. Contribute to Mandrakelinux

The skills of the many talented folks who use Mandrakelinux can be very useful in the making of the Mandrake-
linux system:

• Packaging. A GNU/Linux system is mainly made of programs picked up on the Internet. They have to be
packaged in order to work together.

1
Preface

• Programming. There are many, many projects directly supported by Mandrakesoft: find the one which most
appeals to you and offer your help to the main developer(s).
• Internationalization. You can help us in the translation of web pages, programs and their respective docu-
mentation.
• Documentation. Last but not least, the manual you are currently reading requires a lot of work to stay
up-to-date in regards to the rapid evolution of the system.
Consult the development projects (http://www.mandrakesoft.com/labs/) page to learn more about how you
can contribute to the evolution of Mandrakelinux.

2. Introduction
This manual is aimed at people wishing to dive into the depths of their GNU/Linux system, and who want to
exploit its huge capabilities. It is made up of three parts:

• In The Linux System, we introduce you to the command line and its various uses. We also discuss text-editing
basics, which are essential under GNU/Linux.
In the first chapter (“Basic UNIX System Concepts”, page 5) we introduce the UNIX paradigm while speaking
more specifically of the GNU/Linux world. It discusses the standard file-manipulation utilities as well as so-
me useful features provided by the shell. Then comes a complementary chapter (“Disks and Partitions”, page
13) which discusses how hard disks are managed under GNU/Linux, as well as partitioning. It is very im-
portant that you fully understand the concepts discussed in these chapters before going on to “Introduction
to the Command Line”, page 17.
The next chapter covers text editing (“Text Editing: Emacs and VI”, page 25). As most UNIX configuration files
are text files, you will eventually want or need to edit them in a text editor. You will learn how to use two of
the most famous text editors in the UNIX and GNU/Linux worlds: the mighty Emacs and the modern (!) Vi.
You should then be able to perform basic maintenance on your system. The following two chapters pre-
sent practical uses of the command line (“Command-Line Utilities”, page 33), and process control (“Process
Control”, page 43) in general.

• In Linux in Depth, we touch upon the Linux kernel and the file-system architecture.
We explore the organization of the file tree in “File-Tree Organization”, page 47. UNIX systems tend to grow
very large, but every file has its place in a specific directory. After reading this chapter, you will know where
to look for files depending on their role in the system.
Then we cover the topics of file systems and mount points (“File Systems and Mount Points”, page 51). We
define both of these terms as well as explain them with practical examples.
The next chapter deals with file systems (“The Linux File System”, page 55). After presenting the available
file systems, we discuss file types and some additional concepts and utilities such as inodes and pipes. The
following chapter (“The /proc Filesystem”, page 63) introduces a special GNU/Linux file system called /proc.
“The Start-Up Files: init sysv”, page 69 presents the Mandrakelinux boot-up procedure, and how to use it
efficiently.

• In Advanced Uses, we finish up with topics which only brave or very skilled users will want to put into practi-
ce. We will guide you through the necessary steps to build and install free software from sources in “Building
and Installing Free Software”, page 71. Reading through this chapter should encourage you to try it out, even
though it might look intimidating at first. Finally, the information provided in the last chapter (“Compiling
and Installing New Kernels”, page 85) will help you acquire total GNU/Linux autonomy. After reading and
applying the theory explained in this chapter, you can start converting Windows users to GNU/Linux (if you
haven’t started yet!).

2
 Preface

3. Note from the Editor

In the open-source philosophy, contributors are always welcome! You could provide help to this documenta-
tion project in many different ways. If you have a lot of time, you can write a whole chapter. If you speak a
foreign language, you can help us translate our manuals. If you have ideas on how to improve the content, let
us know. You can event alert us if you find typos!
For any information about the Mandrakelinux documentation project, please contact the documentation ad-
ministrator (mailto:[email protected]) or visit the Mandrakelinux Documentation Project
(http://linux-mandrake.com/en/doc/project/) web page.

4. Conventions Used in this Book

4.1. Typing Conventions

In order to clearly differentiate special words from the text flow, we use different renderings. The following
table shows examples of each special word or group of words with its actual rendering, as well as its signifi-
cation.

Formatted Example Meaning

inode Used to emphasize a technical term.

ls -lta Used for commands and their arguments. Also used for options and file names (see
Commands Synopsis, page 4).
ls(1) Reference to a man page. To read the page in a shell (or command line), simply type
man 1 ls.
$ ls *.pid Formatting used for text snapshots of what you may see on your screen including
computer interactions, program listings, etc.
localhost Literal data which does not generally fit in any of the previously defined categories.
For example, a key word taken from a configuration file.
Apache Defines application names. The example used (“Apache”) is not a command name.
However, in some contexts, the application and command name may be the same but
formatted differently.
Files Indicates menu entries or graphical interface labels. The underlined letter informs you
of a keyboard shortcut, if applicable.
SCSI-Bus Denotes a computer part or a computer itself.
Le petit chaperon Identifies foreign language words.
rouge
Warning! Reserved for special warnings in order to emphasize the importance of words. Read
out loud :-)

Highlights a note. Generally, it gives additional information about

a specific context.

Represents a tip. It can be general advice on how to perform a

particular action, or about nice feature which could make your life
easier.

Be very careful when you see this icon. It always means that very
important information about a specific subject will be dealt with.

3
Preface

4.2. General Conventions

4.2.1. Commands Synopsis

The example below shows the symbols you will see when the writer describes the arguments of a command:
command <non literal argument> [--option={arg1,arg2,arg3}]
[optional arg. ...]

These conventions are standard and you may find them elsewhere such as in the man pages.
The “<” (less than) and “>” (greater than) symbols denote a mandatory argument not to be copied verbatim,
which should be replaced according to your needs. For example, <filename> refers to the actual name of a
file. If this name is foo.txt, you should type foo.txt, not <foo.txt> or <filename>.
The square brackets (“[ ]”) denote optional arguments, which you may or may not include in the command.
The ellipsis (“...”) means an arbitrary number of items can be included.
The curly brackets (“{ }”) contain the arguments authorized at this specific place. One of them is to be placed
here.

4.2.2. Special Notations

From time to time, you will be asked to press, for example, the keys Ctrl-R, which means you need to press
and hold the Ctrl key and tap the R character as well. The same applies for the Alt and Shift keys.
Also, regarding menus, going to menu item File→Reload user config (Ctrl-R) means: click on the File text dis-
played on the menu (generally located in the upper-left of the window). Then in the pull-down menu, click
on the Reload user config item. Furthermore you are informed that you can use the Ctrl-R key combination (as
described above) to get the same result.

4.2.3. System-Generic Users

Whenever possible, we use two generic users in our examples:

Queen Pingusa This user is created at installation time.

Peter Pingus This user is created afterwards by the system administrator.

4
Chapter 1. Basic UNIX System Concepts
The name “UNIX” may be familiar to some of you. You may even use a UNIX system at work, in which case
this chapter may not be very interesting.
For those of you who have never used a UNIX system, reading this chapter is absolutely necessary. Unders-
tanding the concepts which will be introduced here will answer a surprisingly large number of questions
commonly asked by beginners in the GNU/Linux world. Similarly some of these concepts will likely answer
most of the problems you may encounter in the future.

1.1. Users and Groups

Since they have a direct influence on all other concepts, this chapter will introduce the concepts of users and
groups which are extremely important.
Linux is a true multiuser system, and in order to use your GNU/Linux machine, you must have an account on
the machine. When you created a user during installation, you actually created a user account. In case you
don’t remember, you were prompted for the following items:

• the user’s “real name” (which could actually be whatever you want)
• a login name
• and a password.
The two most important parameters here are the login name (commonly abbreviated to login) and password.
You must have both of these in order to access the system.
When you create a user, a default group is also created. Later on, we will see that groups are useful when
you want to share files with other people. A group may contain as many users as you wish, and it is very
common to see such a separation in large systems. For example, at a university, you could have one group per
department, another group for teachers, and so on. The opposite is also true: a user may be a member of one
or more groups. A math teacher, for example, could be a member of the teachers’ group and also of his math
students’ group.
Now that we’ve covered the background information, let us look at how to actually log in.
If you chose to have X automatically start on boot up, your start-up screen will look similar to figure 1-1.

Figure 1-1. Graphical Mode Login Session

In order to log in, you must first select your account from the list. A new dialog will be displayed, prompting
you for your password . Note that you will have to type in your password blindly, because the characters will
be echoed on screen as stars * instead of the characters you type in the password field. You may also choose
your session type (window manager). Once you’re ready, press the Login button.
If you are in console or “text” mode, you will be presented with something similar to the following:
Mandrakelinux Release 10.0 (CodeName) for i586
Kernel 2.6.3-4mdk on an i686 / tty1
[machine_name] login:

5
Chapter 1. Basic UNIX System Concepts

To log in, type your login name at the Login: prompt and press Enter. Next, the login program (login) will
display a Password: prompt and wait for you to enter your password. Like the graphic mode login, the console
login will not echo the characters you are typing on the screen.
Note that you can log in several times with the same account on additional consoles and under X. Each session
you open is independent to the others, and it is even possible to open several X sessions at the same time. By
default, Mandrakelinux has six virtual consoles in addition to the one reserved for the graphical interface. You
can switch to any of them by pressing the Ctrl-Alt-F<n> key sequence, where <n> is the number of the console
that you want to switch to. By default, the graphical interface is on console number 7. Therefore, to switch to
the second console, you would simultaneously press the Ctrl, Alt and F2 keys.
During the installation, DrakX also prompted you for the password of a very special user: root. root is the
system administrator which will most likely be yourself. For your system’s security, it is very important for
the root account to be always protected by a good and hard-to-guess password!
If you regularly log in as root, it can be very easy to make a mistake which could render your system unusable:
one single mistake can do it. In particular, if you did not set a password for the root account, then any user
can alter any part of your system (and even other operating systems on your machine!). Obviously this is not
a good idea.
It’s worth mentioning that internally, the system does not identify you by your login name. Instead, it uses a
unique number assigned to the name: the User ID (UID for short) . Similarly every group is identified by its
Group ID (GID) and not by its name.

1.2. File Basics

Compared to Windows and most other operating systems, files are handled very differently under GNU/Linux.
In this section we will cover the most obvious differences. For more information, please read “The Linux File
System”, page 55.
The major differences result directly from the fact that Linux is a multiuser system: every file is the exclusive
property of one user and one group. One thing we didn’t mention about users and groups is that every one of
them possesses a personal directory (called the home directory). The user is the owner of this directory and of
all files created there.
However, this would not be very useful if that were the only notion of file ownership. As the file owner, a user
may set permissions on files. These permissions distinguish between three user categories: the owner of the
file, every user who is a member of the group associated with the file (also called the owner group) but who
is not the owner, and others, which includes every other user who is neither the owner nor a member of the
owner’s group.
There are three different permissions:

1. Read permission (r): enables a user to read the contents of a file. For a directory, the user can list its contents
(i.e. the files in this directory).
2. Write permission (w): allows the modification of a file’s contents. For a directory, the write permission
allows a user to add or remove files from this directory, even if he is not the owner of these files.
3. eXecute permission (x): enables a file to be executed (normally only executable files have this permission
set). For a directory, it allows a user to traverse it, which means going into or through that directory. Note
that this is different from the read access: you may be able to traverse a directory but still be unable to read
its contents!
Every permission combination is possible. For example, you can allow only yourself to read the file and forbid
access to all other users. You can even do the opposite, even if isn’t very logical at first glance... As the file
owner, you can also change the owner group (if and only if you’re a member of the new group), and even
deprive yourself of the file (that is, change its owner). Of course, if you deprive yourself of the file, you will
also lose all of your rights over it.
Let’s take the example of a file and a directory. The display below represents entering the ls -l command
from the command line:

$ ls -l
total 1
-rw-r----- 1 queen users 0 Jul 8 14:11 a_file

6
 Chapter 1. Basic UNIX System Concepts

drwxr-xr-- 2 peter users 1024 Jul 8 14:11 a_directory/

$

The results of the ls -l command are (from left to right):

• The first ten characters represent the file’s type and the permissions associated with it. The first character is
the file’s type: if it’s a regular file, it will contain a dash (-). If it’s a directory, the leftmost character will be a
d. There are other file types, which we’ll discuss later on. The next nine characters represent the permissions
associated with that file. The nine characters are actually three groups of three permissions. The first group
represents the rights associated with the file owner; the next three apply to all users belonging to the same
group but who are not the owner; and the last three apply to others. A dash (-) means that the permission
is not set.
• Next comes the number of links for the file. Later on we’ll see that the unique identifier of a file is not its
name, but a number (the inode number), and that it’s possible for one file on disk to have several names. For
a directory, the number of links has a special meaning, which will also be discussed a bit further.
• The next piece of information is the name of the file owner and the name of the owner group.
• Finally, the size of the file (in bytes) and its last modification time are displayed, with the name of the file or
directory itself as the last item on the line.
Let’s take a closer look at the permissions associated with each of these files. First of all, we must strip off the
first character representing the type, and for the file a_file, we get the following rights: rw-r-----. Here’s a
breakdown of the permissions.

• the first three characters (rw-) are the rights of the owner, which in this case is queen. Therefore, queen has
the right to read the file (r), to modify its contents (w) but not to execute it (-).
• the next three characters (r--) apply to any user who is not queen but who is a member of the users group.
They will be able to read the file (r), but will not be able to write nor execute it (--).
• the last three characters (---) apply to any user who is not queen and is not a member of the users group.
Those users do not have any rights on the file at all.
For the directory a_directory, the rights are rwxr-xr--, so:

• peter, as the directory owner, can list files contained inside (r), add to or remove files from that directory (w),
and may traverse it (x).
• Each user who isn’t peter, but is a member of the users group, will be able to list files in this directory (r),
but not remove or add files (-), and will be able to traverse it (x).
• Every other user will only be able to list the contents of this directory (r). Because they do not have wx
permissions, they will not be able to write files or enter the directory.
There is one exception to these rules: root. root can change attributes (permissions, owner and group owner)
of all files, even if he’s not the owner, and could therefore grant himself ownership of the file! He can read
files on which he has no read permission, traverse directories which he would normally have no access to,
and so on. And if he lacks a permission, he only has to add it. root has complete control of the system, which
involves a certain amount of trust in the person wielding the root password.
Lastly, it’s worth noting the differences between file names in the UNIX and the Windows worlds. For one, UNIX
allows for a much greater flexibility and has fewer limitations.

• A file name may contain any character, including non-printable ones, except for the ASCII character 0,
which denotes the end of a string, and /, which is the directory separator. Moreover, because UNIX is case
sensitive, the files readme and Readme are different, because r and R are considered two different characters
in UNIX-based systems.
• As you may have noticed, a file name does not have to include an extension, unless that’s the way you
prefer to name your files. File extensions do not identify the content of files under GNU/Linux or almost any
other operating system. So-called “file extensions” are quite convenient though. The period (.) under UNIX
is just one character among others, but it also has one special meaning. Under UNIX, file names beginning
with a period are “hidden files”, which also includes directories whose names start with a .

7
Chapter 1. Basic UNIX System Concepts

However it’s worth noting that many graphical applications (file

managers, office applications, etc.) actually use file extensions to
recognize their files. It is therefore a good idea to use file-name
extensions for those applications which support them.

1.3. Processes
A process defines an instance of a program being executed and its environment. We will only mention the
most important differences between GNU/Linux and Windows here (please refer to “Process Control”, page 43 for
more information).
The most important difference is directly related to the user concept: each process is executed with the rights
of the user who launched it. Internally, the system identifies processes with a unique number, called the process
ID, or PID. From this PID, the system knows who, (that is, which user), has launched the process and a number
of other pieces of information, and the system only needs to verify the process’ validity. So, if we take our
a_file example, a process launched by peter will be able to open this file in read-only mode, but not in read-
write mode because the permissions associated with the file forbid it. Once again the exception to this rule is
root.
Because to this, GNU/Linux is virtually immune to viruses. In order to operate, viruses must infect executable
files. As a user, you do not have write access to vulnerable system files, so the risk is greatly reduced. Generally
speaking, viruses are very rare in the UNIX world. There are less than a dozen known viruses for Linux, and
they are harmless when executed by a normal user. Only one user can damage a system by activating these
viruses: root.
Interestingly enough, anti-virus software does exist for GNU/Linux, but mostly for DOS/Windows files. Why are
there anti-virus programs running on GNU/Linux which focus on DOS/Windows? More and more often, you
will see GNU/Linux systems acting as file servers for Windows machines with the help of the Samba software
package (see the Sharing Files and Printers chapter of the Server Administration Guide).
Linux makes it easy to control processes. One way is through “signals”, which allow you to suspend or kill a
process by sending the corresponding signal to the process. However, you are limited to sending signals to
your own processes. With the exception of root, UNIX does not allow you to send signals to a process launched
by any other user. In “Process Control”, page 43, you will learn how to obtain the PID of a process and to send
it signals.

1.4. A Short Introduction to the Command Line

The command line is the most direct way to send commands to your machine. If you use the GNU/Linux
command line, you’ll soon find that it is much more powerful and capable than other command prompts
you may have encountered previously. This power is available because you have access, not only to all X
applications, but also to thousands of other utilities in console mode (as opposed to graphical mode) which
do not have graphical equivalents, with their many options and possible combinations, which would be hard
to access in the form of buttons or menus.
Admittedly, most people require a little help to get started. If you are not already working in console mode and
are using the graphical interface, the first thing to do is to launch a terminal emulator. Access the main menu
of GNOME, KDE or any other window manager you might be using and you’ll find a number of emulators in
the System+Terminals sub-menu. Choose the one you want, for example Konsole or XTerm. Depending on your
user interface, there may also be an icon which clearly identifies it on the panel (figure 1-2).

Figure 1-2. The Terminal Icon on the KDE Panel

When you launch this terminal emulator, you are actually using a shell. This is the name of the program which
you interact with. You’ll find yourself in front of the prompt:
[queen@localhost queen]$

8
 Chapter 1. Basic UNIX System Concepts

This assumes that your user name is queen and that your machine’s name is localhost (which is the case if
your machine is not part of an existing network). Following the prompt there is space for you to type your
commands. Note that when you’re root, the prompt’s $ character becomes a # (this is true only in the default
configuration, since you may customize all such details in GNU/Linux). In order to become root, type su after
launching a shell.
# Enter the root password; (it will not appear on the screen)
[queen@localhost queen]$ su
Password:
# exit (or Ctrl-D) will take you back to your normal user account
[root@localhost queen]# exit
[queen@localhost queen]$

Everywhere else in this book, the prompt will be symbolically represented by a $, whether you are a normal
user or root. You will be told when you have to be root to execute a command, so please remember the su
command.
When you launch a shell for the first time, you normally find yourself in your home directory. To display the
name of the directory you are currently in, type pwd (which stands for Print Working Directory):

$ pwd /home/queen

Next we’ll look at a few basic commands which are very useful.

1.4.1. cd: Change Directory

The cd command is just like the DOS one, with extras. It does just what its acronym states, changes the working
directory. You can use . and .., which respectively stand for the current and parent directories. Typing cd
alone will take you back to your home directory. Typing cd - will take you back to the last directory you
visited. And lastly, you can specify peter’s home directory by typing cd ~peter (~ on its own means your
own home directory). Note that as a normal user, you cannot usually get into another user’s home directory
(unless they explicitly authorized it or if this is the default configuration on the system), unless you are root,
so let’s become root and practice:

$ pwd
/root
$ cd /usr/share/doc/HOWTO
$ pwd
/usr/share/doc/HOWTO
$ cd ../FAQ-Linux
$ pwd
/usr/share/doc/FAQ-Linux
$ cd ../../../lib
$ pwd
/usr/lib
$ cd ~peter
$ pwd
/home/peter
$ cd
$ pwd
/root

Now, go back to being a normal user again by typing exit.

9
Chapter 1. Basic UNIX System Concepts

1.4.2. Some Environment Variables and the echo Command

All processes have their environment variables and the shell allows you to view them directly with the echo
command. Some interesting variables are:

1. HOME: this environment variable contains a string which represents your home directory.
2. PATH: this variable contains the list of all directories in which the shell should look for executables when
you type a command. Note that unlike DOS, by default, a shell will not look for commands in the current
directory!
3. USERNAME: this variable contains your login name.
4. UID: this one contains your user ID.
5. PS1: this variable determines what your prompt will display, and is often a combination of special sequen-
ces. You may read the bash(1) manual page for more information.
To have the shell print a variable’s value, you must put a $ in front of its name. Here, the echo command will
give an example:

$ echo Hello
Hello
$ echo $HOME
/home/queen
$ echo $USERNAME
queen
$ echo Hello $USERNAME
Hello queen
$ cd /usr
$ pwd
/usr
$ cd $HOME
$ pwd
/home/queen

As you can see, the shell substitutes the variable’s value before it executes the command. Otherwise, our cd
$HOME example would not have worked. In fact, the shell first replaced $HOME by its value, /home/queen, so the
line became cd /home/queen, which is what we wanted. The same thing happened with the echo $USERNAME
example.

1.4.3. cat: Print the Contents of One or More Files to the Screen
Nothing much to say, this command does just that: it prints the contents of one or more files to the standard
output, normally the screen:

$ cat /etc/fstab
/dev/hda5 / ext2 defaults 1 1
/dev/hda6 /home ext2 defaults 1 2
/dev/hda7 swap swap defaults 0 0
/dev/hda8 /usr ext2 defaults 1 2
/dev/fd0 /mnt/floppy auto sync,user,noauto,nosuid,nodev 0 0
none /proc proc defaults 0 0
none /dev/pts devpts mode=0620 0 0
/dev/cdrom /mnt/cdrom auto user,noauto,nosuid,exec,nodev,ro 0 0
$ cd /etc
$ cat modules.conf shells
alias parport_lowlevel parport_pc
pre-install plip modprobe parport_pc ; echo 7 > /proc/parport/0/irq
#pre-install pcmcia_core /etc/rc.d/init.d/pcmcia start
#alias char-major-14 sound
alias sound esssolo1
keep
/bin/zsh
/bin/bash
/bin/sh
/bin/tcsh
/bin/csh

10
 Chapter 1. Basic UNIX System Concepts

/bin/ash
/bin/bsh
/usr/bin/zsh

1.4.4. less: a Pager

The name is a play on words related to the first pager ever used under UNIX, called more. A pager is a program
which allows a user to view long files page by page (more accurately, screen by screen). The reason that we
discuss less rather than more is that less is more intuitive. You should use less to view large files which will
not fit on a single screen. For example:

less /etc/termcap

To browse through this file, use the up and down arrow keys. Press Q to quit. less can do far more than just
that: press H for help to display the various options available.

1.4.5. ls: Listing Files

The ls (LiSt) command is equivalent to the dir command in DOS, but it can do much much more. In fact, this
is largely because files can do more too. The command syntax for ls is:

ls [options] [file|directory] [file|directory...]

If no file or directory is specified on the command line, ls will list files in the current directory. Its options are
numerous, so we’ll only describe a few of them:

• -a: lists all files, including hidden files. Remember that in UNIX, hidden files are those whose names begin
with a .; the -A option lists “almost” all files, which means every file the -a option would print except for
“.” and “..”
• -R: lists recursively, i.e. all files and subdirectories of directories entered on the command line.
• -s: prints the file size in kilobytes next to each file.
• -l: prints additional information about the files such as the permissions associated to it, the owner and
owner group, the file’s size and the last-access time.
• -i: prints the inode number (the file’s unique number in the file system, see “The Linux File System”, page
55) next to each file.
• -d: treats directories on the command line as if they were normal files rather than listing their contents.
Here are some examples:

• ls -R: recursively lists the contents of the current directory.

• ls -is images/ ..: lists the inode number and the size in kilobytes for each file in the images/ directory
as well as in the parent of the current directory.
• ls -l images/*.png: lists all files in the images/ directory whose names end with .png, including the file
.png, if it exists.

11
Chapter 1. Basic UNIX System Concepts

1.4.6. Useful Keyboard Shortcuts

There are a number of shortcuts available, with the primary advantage being that they save you a lot of
typing time. This section assumes you’re using the default shell provided with Mandrakelinux, bash, but these
keystrokes might work with other shells too.
First: the arrow keys. bash maintains a history of previous commands which you can view with the up and
down arrow keys. You can scroll up to a maximum number of lines defined in the HISTSIZE environment va-
riable. In addition, the history is persistent from one session to another, so you will not lose all the commands
you typed in previous sessions.
The left and right arrow keys move the cursor left and right on the current line, allowing you to edit your com-
mands. But there’s more to editing than just moving one character at a time: Ctrl-A and Ctrl-E, for example,
will take you to the beginning and the end of the current line. The Backspace and Del keys work as expected.
Backspace and Ctrl-H are equivalent. Del and Ctrl-D can also be used interchangeably. Ctrl-K will delete
from the position of the cursor to the end of line, and Ctrl-W will delete the word before the cursor (so will
Alt-Backspace).
Typing Ctrl-D on a blank line will let you close the current session, which is much shorter than having to type
exit. Ctrl-C will interrupt the currently running command, except if you were in the process of editing your
command line, in which case it will cancel the editing and get you back to the prompt. Ctrl-L clears the screen.
Ctrl-Z temporarily stops a task, it suspends it. This shortcut is very useful when you forget to type the “&”
character after typing a command. For instance:
$ xpdf MyDocument.pdf

Hence you cannot use your shell anymore since the foreground task is allocated to the xpdf process. To put
that task in the background and restore your shell, simply type bg.
Finally, there are Ctrl-S and Ctrl-Q, which are used to suspend and restore output to the screen. They are not
used often, but you might type Ctrl-S by mistake (after all, S and D are close to each other on the keyboard).
So, if you get into the situation where you’re typing but you can’t see any characters appearing on the Terminal,
try Ctrl-Q. Note that all the characters you typed between the unwanted Ctrl-S and Ctrl-Q will be printed to
the screen all at once.

12
Chapter 2. Disks and Partitions

This chapter contains information for those who simply wish to know more about the technical details underl-
ying their system. It will give a complete description of the PC partitioning scheme. Therefore it will be most
useful if you plan to manually configure your hard drive partitions. Since the installer can partition your hard
disk automatically, it’s not critical to understand everything if you perform a standard installation. However
if you plan on modifying the partition scheme, the information in this chapter will be important.

2.1. Structure of a Hard Disk

A disk is physically divided into sectors. A sequence of sectors can form a partition. Roughly speaking, you
can create as many partitions as you wish, up to 67 (3 primary partitions and a secondary partition containing
up to 64 logical partitions inside): each of them is regarded as a single hard drive.

2.1.1. Sectors
To simplify, a hard disk is merely a sequence of sectors, which are the smallest data unit on a hard disk. The
typical size of a sector is 512 bytes. The sectors on a hard disk of “n” sectors are numbered from “0” to “n-1”.

2.1.2. Partitions
The use of multiple partitions enables you to create many virtual hard drives on your real physical drive. This
has many advantages:

• Different operating systems use different disk structures (called file systems): this is the case with Windows
and GNU/Linux. Having multiple partitions on a hard drive allows you to install various operating systems
on the same physical drive.
• For performance reasons, a single operating system may prefer different drives with various file systems on
them because they may be used for completely different things. One example is GNU/Linux which requires
a second partition called Swap. The latter is used by the virtual memory manager as virtual memory.
• Even if all of your partitions use the same file system, it may prove useful to separate the different parts
of your OS into different partitions. A simple configuration example would be to split your files into two
partitions: one for your personal data, and another one for your programs. This allows you to update your
OS, completely erasing the partition containing the programs while keeping the data partition safe.
• Physical errors on a hard disk are generally located on adjacent sectors, not scattered across the disk. Distri-
buting your files across different partitions could limit data loss if your hard disk is physically damaged.
Normally, the partition type specifies the file system which the partition is supposed to contain. Each operating
system might recognize some partition types, but not others. Please see “File Systems and Mount Points”, page
51, and “The Linux File System”, page 55, for more information.

2.1.3. Defining the Structure of Your Disk

2.1.3.1. The Simplest Way

This scenario would imply only two partitions: one for the Swap space, the other one for the files1.

1. the default file system under Mandrakelinux is called ext3

13
Chapter 2. Disks and Partitions

A rule of thumb is to set the swap partition size to twice the size
of your RAM memory (i.e.: if you have 128 MB of RAM memory
the swap size should be of 256 MB). However for large memory
configurations (>512 MB), this rule isn’t critical, and smaller sizes
are acceptable.

2.1.3.2. Another Common Scheme

Separate data from programs. To be even more efficient, one usually defines a third partition called root and
labelled as /. It will contain the programs required to start your system and to perform basic maintenance.
Therefore we could define four partitions:

Swap
A Swap partition whose size is roughly twice the amount of physical RAM.

Root: /
The most important partition. Not only does it contain critical data and programs for the system, it also
acts as a mount point for other partitions (see “File Systems and Mount Points”, page 51).
The needs of the root partition in terms of size are very limited, 400MB is generally enough. However, if
you plan to install commercial applications, which are most often located in the /opt/ directory, you will
need to increase the size of the root partition. Alternatively, you may create a separate partition for /opt/.

Static data: /usr/

Most packages install the majority of their executables and data files under the /usr/ directory. The ad-
vantage of creating a separate partition is that it allows you to easily share it with other machines over a
network.
The recommended size depends on the packages you wish to install, and can vary from 100MB for a very
lightweight installation, to several GB for a full installation. A compromise of two or three GB (depending
on your disk size) is usually sufficient.

Home directories: /home/

This directory contains the personal directories for all of the users hosted on your machine. The partition
size depends on the number of users hosted and their needs.
Another solution is to not create a separate partition for the /usr/ files: /usr/ will simply be a directory inside
the root (/) partition.

2.1.3.3. Exotic Configurations

When setting up your machine for specific uses –– such as a web server or a firewall –– the needs are radically
different than for a standard desktop machine. For example, a FTP server will probably need a large sepa-
rate partition for /var/ftp/, while the /usr/ directory could be relatively small. In these situations, you’re
encouraged to carefully think about your needs before even beginning the installation process.

If you need to resize your partitions or use a different partition

scheme, note that it is possible to resize most partitions without
the need to reinstall your system and without losing your data.
Please consult Managing Your Partitions of the Starter Guide .

With some practice, you’ll even be able to move a crowded

partition to a brand new hard drive.

14
 Chapter 2. Disks and Partitions

2.2. Conventions for Naming Disks and Partitions

GNU/Linux uses a logical method to name partitions. First, when numbering the partitions, it ignores the file-
system type of each partition you may have. Second, it names the partitions according to the disk on which
they are located. This is how the disks are named:

• The primary master and primary slave IDE devices (whether they be hard disks, CD-ROM drives or
anything else) are called /dev/hda and /dev/hdb respectively.
• On the secondary interface, the master is called /dev/hdc and the slave is /dev/hdd.
• If your computer contains other IDE interfaces (for example, the IDE interface present on some Soundblaster
cards), the disks will be called /dev/hde, /dev/hdf, etc. You may also have additional IDE interfaces if you
have RAID controllers.
• SCSI disks are called /dev/sda, /dev/sdb, etc., in the order of their appearance on the SCSI chain (depending
on the increasing IDs). The SCSI CD-ROM drives are called /dev/scd0, /dev/scd1, always in the order of
their appearance on the SCSI chain.
The partitions are named after the disk on which they’re found, in the following way (in our example, we’ve
used partitions on a primary master IDE disk):

• The primary (or extended) partitions are named /dev/hda1 through /dev/hda4, when present.
• Logical partitions, if any, are named /dev/hda5, /dev/hda6, etc. in the order of their appearance in the table
of logical partitions.
So GNU/Linux will name the partitions as follows:

Figure 2-1. First Example of Partition Naming under GNU/Linux

15
Chapter 2. Disks and Partitions

Figure 2-2. Second Example of Partition Naming under GNU/Linux

With this knowledge in hand, you should be able to name your various partitions and hard disks when you
need to manipulate them. You’ll also see that GNU/Linux names the partitions even if it doesn’t know how to
manage them initially (it ignores the fact that they’re not native GNU/Linux partitions).

For 2.4 and 2.6 kernels, Mandrakelinux uses the Linux DevFS
(Device File System) (http://www.atnf.csiro.au/~rgooch/
linux/docs/devfs.html). It ensures full compatibility with the
scheme described above, but this compatibility may disappear in
the future. Each device is dynamically added to the system as soon
as it becomes available or needed.

For example, the first IDE hard drive now becomes

/dev/ide/host0/bus0/target0/lun0/disc instead of /dev/hda.

16
Chapter 3. Introduction to the Command Line
In the chapter “Basic UNIX System Concepts”, page 5, you were shown how to launch a shell. In this chapter,
we will show you how to work with it.
The shell’s main asset is the number of existing utilities: there are thousands of them, and each one is devoted
to a particular task. We will only look at a (very) small number of them here. One of UNIX’s greatest assets is
the ability to combine these utilities, as we shall see later.

3.1. File-Handling Utilities

In this context, file handling means copying, moving and deleting files. Later, we will look at ways of changing
file attributes (owner, permissions).

3.1.1. mkdir, touch: Creating Empty Directories and Files

mkdir (MaKe DIRectory) is used to create directories. Its syntax is simple:
mkdir [options] <directory> [directory ...]

Only one option is worth noting: the -p option. It does two things:

1. it will create parent directories if they did not exist before. Without this option, mkdir would just fail,
complaining that the said parent directories do not exist;
2. it will return silently if the directory you wanted to create already exists. Similarly, if you did not specify
the -p option, mkdir will send back an error message, complaining that the directory already exists.
Here are some examples:

• mkdir foo: creates a directory foo in the current directory;

• mkdir -p images/misc docs: creates the misc directory in the images directory. First, it creates the latter if
it does not exist (-p); it also creates a directory named docs in the current directory.
Initially, the touch command was not intended for creating files but for updating file access and modification
times1. However, touch will create the files listed as empty files if they do not exist. The syntax is:
touch [options] file [file...]

So running the command:

touch file1 images/file2

will create an empty file called file1 in the current directory and an empty file file2 in directory images, if
the files did not previously exist.

3.1.2. rm: Deleting Files or Directories

The rm command (ReMove) replaces the DOS commands del and deltree, and adds more options. Its syntax
is as follows:
rm [options] <file|directory> [file|directory...]

Options include:

• -r, or -R: delete recursively. This option is mandatory for deleting a directory, empty or not. However, you
can also use rmdir to delete empty directories.

1. In UNIX, there are three distinct timestamps for each file: the last file access date (atime), i.e. the last date when the
file was opened for read or write; the last date when the inode attributes were modified (mtime); and finally, the last date
when the contents of the file were modified (ctime).

17
Chapter 3. Introduction to the Command Line

• -i: request confirmation before each deletion. Note that by default in Mandrakelinux, rm is an alias to rm
-i, for safety reasons (similar aliases exist for cp and mv). Your mileage may vary as to the usefulness of
these aliases. If you want to remove them, you can create an empty ~/.alias file which will prevent setting
system wide aliases. Alternatively you can edit your ~/.bashrc file to disable some of the system wide
aliases by adding this line: unalias rm cp mv
• -f, the opposite of -i, forces deletion of the files or directories, even if the user has no write access on the
files2.
Some examples:

• rm -i images/*.jpg file1: deletes all files with names ending in .jpg in the images directory and deletes
file1 in the current directory, requesting confirmation for each file. Answer y to confirm deletion, n to
cancel.
• rm -Rf images/misc/ file*: deletes, without requesting confirmation, the whole directory misc/ in the
images/ directory, together with all files in the current directory whose names begin with file.

Using rm deletes files irrevocably. There is no way to restore them3!

Don’t hesitate to use the -i option to ensure that you do not delete
something by mistake.

3.1.3. mv: Moving or Renaming Files

The syntax of the mv (MoVe) command is as follows:
mv [options] <file|directory> [file|directory ...] <destination>

Some options:

• -f: forces operation –– no warning if an existing file is overwritten.

• -i: the opposite. Asks the user for confirmation before overwriting an existing file.
• -v: verbose mode, report all changes and activity.
Some examples:

• mv -i /tmp/pics/*.png .: move all files in the /tmp/pics/ directory whose names end with .png to the
current directory (.), but request confirmation before overwriting any files already there.
• mv foo bar: rename file foo to bar. If a bar directory already existed, the effect of this command would be
to move file foo or the whole directory (the directory itself plus all files and directories in it, recursively)
into the bar directory.
• mv -vf file* images/ trash/: move, without requesting confirmation, all files in the current directory
whose names begin with file, together with the entire images/ directory to the trash/ directory, and show
each operation carried out.

3.1.4. cp: Copying Files and Directories

cp (CoPy) replaces the DOS commands copy and xcopy and adds more options. Its syntax is as follows:
cp [options] <file|directory> [file|directory ...] <destination>

cp has a lot of options. Here are the most common:

• -R: recursive copy; mandatory for copying a directory, even an empty directory.
• -i: request confirmation before overwriting any files which might be overwritten.

2. It is enough for the user to have write access to the directory to be able to delete files in it, even if he is not the owner
of the files.

18
 Chapter 3. Introduction to the Command Line

• -f: the opposite of -i, replaces any existing files without requesting confirmation.
• -v: verbose mode, displays all actions performed by cp.
Some examples:

• cp -i /timages/* images/: copies all files in the /timages/ directory to the images/ directory located in
the current directory. It requests confirmation if a file is going to be overwritten.
• cp -vR docs/ /shared/mp3s/* mystuff/: copies the whole docs directory, plus all files in the /shared/
mp3s directory to the mystuff directory.
• cp foo bar: makes a copy of the foo file with the name bar in the current directory.

3.2. Handling File Attributes

The series of commands shown here are used to change the owner or owner group of a file or its permissions.
We looked at the different permissions in chapter Basic UNIX System Concepts.

3.2.1. chown, chgrp: Change the Owner and Group of One or More Files
The syntax of the chown (CHange OWNer) command is as follows:
chown [options] <user[:group]> <file|directory> [file|directory...]

The options include:

• -R: recursive. To change the owner of all files and subdirectories in a given directory.
• -v: verbose mode. Displays all actions performed by chown; reports which files have changed ownership as
a result of the command and which files have not been changed.
• -c: like -v, but only reports which files have been changed.
Some examples:

• chown nobody /shared/book.tex: changes the owner of the /shared/book.tex file to nobody.
• chown -Rc queen:music *.mid concerts/: changes the ownership of all files in the current directory who-
se name ends with .mid and all files and subdirectories in the concerts/ directory to user queen and group
music, reporting only files affected by the command.
The chgrp (CHange GRouP) command lets you change the group ownership of a file (or files); its syntax is very
similar to that of chown:
chgrp [options] <group> <file|directory> [file|directory...]

The options for this command are the same as for chown, and it is used in a very similar way. Thus, the
command:
chgrp disk /dev/hd*

changes the ownership of all files in directory /dev/ with names beginning with hd to group disk.

19
Chapter 3. Introduction to the Command Line

3.2.2. chmod: Changing Permissions on Files and Directories

The chmod (CHange MODe) command has a very distinct syntax. The general syntax is:
chmod [options] <change mode> <file|directory> [file|directory...]

but what distinguishes it is the different forms that the mode change can take. It can be specified in two ways:

1. in octal. The owner user permissions then correspond to figures with the form <x>00, where <x> corres-
ponds to the permission assigned: 4 for read permission, 2 for write permission and 1 for execute permis-
sion. Similarly, the owner group permissions take the form <x>0 and permissions for “others” the form
<x>. Then, all you need to do is add together the assigned permissions to get the right mode. Thus, the
permissions rwxr-xr-- correspond to 400+200+100 (owner permissions, rwx) +40+10 (group permissions,
r-x) +4 (others’ permissions, r--) = 754; in this way, the permissions are expressed in absolute terms. This
means that previous permissions are unconditionally replaced;
2. with expressions. Here permissions are expressed by a sequence of expressions separated by commas.
Hence an expression takes the following form: [category]<+|-|=><permissions>.
The category may be one or more of:

• u (User, permissions for owner);

• g (Group, permissions for owner group);
• o (Others, permissions for “others”).
If no category is specified, changes will apply to all categories. A + sets a permission, a - removes the
permission and a = sets the permission. Finally, the permission is one or more of the following:

• r (Read);
• w (Write) or;
• x (eXecute).

The main options are quite similar to those of chown and chgrp:

• -R: changes permissions recursively.

• -v: verbose mode. Displays the actions carried out for each file.
• -c: like -v but only shows files affected by the command.
Examples:

• chmod -R o-w /shared/docs: recursively removes write permission for others on all files and subdirecto-
ries in the /shared/docs/ directory.
• chmod -R og-w,o-x private/: recursively removes write permission for group and others for the whole
private/ directory, and removes the execution permission for others.
• chmod -c 644 misc/file*: changes permissions of all files in the misc/ directory whose names begin with
file to rw-r--r-- (i.e. read permission for everyone and write permission only for the owner), and reports
only files affected by this command.

20
 Chapter 3. Introduction to the Command Line

3.3. Shell Globbing Patterns

You probably already use globbing characters without knowing it. When you specify a file in Windows or when
you look for a file, you use * to match a random string. For example, *.txt matches all files with names ending
with .txt. We also used it heavily in the last section. But there is more to globbing than just *.
When you type a command like ls *.txt and press Enter, the task of finding which files match the *.txt
pattern is not done by the ls command, but by the shell itself. This requires a little explanation about how a
command line is interpreted by the shell. When you type:
$ ls *.txt
readme.txt recipes.txt

the command line is first split into words (ls and *.txt in this example). When the shell sees a * in a word,
it will interpret the whole word as a globbing pattern and will replace it with the names of all matching files.
Therefore, the command, just before the shell executes it, has become ls readme.txt recipe.txt, which
gives the expected result. Other characters make the shell react this way too:

• ?: matches one and only one character, regardless of what that character is;
• [...]: matches any character found in the brackets. Characters can be referred to either as a range of
characters (i.e. 1-9) or discrete values, or even both. Example: [a-zBE5-7] will match all characters between
a and z, a B, an E, a 5, a 6 or a 7;
• [!...]: matches any character not found in the brackets. [!a-z], for example, will match any character
which is not a lowercase letter4;
• {c1,c2}: matches c1 or c2, where c1 and c2 are also globbing patterns, which means you can write {[0-
9]*,[acr]} for example.
Here are some patterns and their meanings:

• /etc/*conf: all files in the /etc directory with names ending in conf. It can match /etc/inetd.conf, /etc/
conf.linuxconf, and also /etc/conf if such a file exists. Remember that * can also match an empty string.
• image/{cars,space[0-9]}/*.jpg: all file names ending with .jpg in directories image/cars, image/
space0, (...), image/space9, if those directories exist.
• /usr/share/doc/*/README: all files named README in all of /usr/share/doc’s immediate subdirectories.
This will make /usr/share/doc/mandrake/README match, for example, but not /usr/share/doc/myprog/
doc/README.
• *[!a-z]: all files with names which do not end with a lowercase letter in the current directory.

3.4. Redirections and Pipes

3.4.1. A Little More About Processes

To understand the principle of redirections and pipes, we need to explain a notion about processes which has
not yet been introduced. Most UNIX processes (this also includes graphical applications but excludes most
daemons) use a minimum of three file descriptors: standard input, standard output and standard error. Their
respective numbers are 0, 1 and 2. In general, these three descriptors are associated with the terminal from
which the process was started, with the input being the keyboard. The aim of redirections and pipes is to
redirect these descriptors. The examples in this section will help you better understand this concept.

4. Beware! While this is true for most languages, this may not be true for your own language setting (locale). This
depends on the collating order. On some language configurations, [a-z] will match a, A, b, B, (...), z. And we do not even
mention the fact that some languages have accentuated characters...

21
Chapter 3. Introduction to the Command Line

3.4.2. Redirections
Imagine, for example, that you wanted a list of files ending with .png5 in the images directory. This list is very
long, so you may want to store it in a file in order to look through it at your leisure. You can enter the following
command:
$ ls images/*.png 1>file_list

This means that the standard output of this command (1) is redirected (>) to the file named file_list. The >
operator is the output redirection operator. If the redirection file does not exist, it is created, but if it exists, its
previous contents are overwritten. However, the default descriptor redirected by this operator is the standard
output and does not need to be specified on the command line. So you can write more simply:
$ ls images/*.png >file_list

and the result will be exactly the same. Then you could look at the file using a text file viewer such as less.
Now, imagine you want to know how many of these files exist. Instead of counting them by hand, you can
use the utility called wc (Word Count) with the -l option, which writes on the standard output the number of
lines in the file. One solution is as follows:
wc -l 0<file_list

and this gives the desired result. The < operator is the input redirection operator, and the default redirected
descriptor is the standard input one, i.e. 0, and you simply need to write the line:
wc -l <file_list

Now suppose you want to remove all the file “extensions” and put the result in another file. One tool for doing
this is sed (Stream EDitor). You simply redirect the standard input of sed to the file_list file and redirect its
output to the result file, i.e. the_list:
sed -e ’s/\.png$//g’ <file_list >the_list

and your list is created, ready for you to view at your leisure with any viewer.
It can also be useful to redirect standard errors. For example, you want to know which directories in
/shared you cannot access: one solution is to list this directory recursively and to redirect the errors to a
file, while not displaying the standard output:
ls -R /shared >/dev/null 2>errors

which means that the standard output will be redirected (>) to /dev/null, a special file in which everything
you write is discarded (i.e. the standard output is not displayed) and the standard error channel (2) is redirec-
ted (>) to the errors file.

3.4.3. Pipes
Pipes are in some ways a combination of input and output redirections. The principle is that of a physical
pipe, hence the name: one process sends data into one end of the pipe and another process reads the data at
the other end. The pipe operator is |. Let us go back to the file list example above. Suppose you want to find
out directly how many corresponding files there are without having to store the list in a temporary file, you
would then use the following command:
ls images/*.png | wc -l

which means that the standard output of the ls command (i.e. the list of files) is redirected to the standard
input of the wc command. This then gives you the desired result.
You can also directly put together a list of files “without extensions” using the following command:
ls images/*.png | sed -e ’s/\.png$//g’ >the_list

5. You might think it is crazy to say “files ending with .png” rather than “PNG images”. However, once again, files
under UNIX only have an extension by convention: extensions in no way define a file type. A file ending with .png could
perfectly well be a JPEG image, an application file, a text file or any other type of file. The same is true under Windows as
well!

22
 Chapter 3. Introduction to the Command Line

or, if you want to consult the list directly without storing it in a file:
ls images/*.png | sed -e ’s/\.png$//g’ | less

Pipes and redirections are not restricted solely to text which can be read by human beings. For example, the
following command sent from a Terminal:
xwd -root | convert - ~/my_desktop.png

will send a screen shot of your desktop to the my_desktop.png 6 file in your personal directory.

3.5. Command-Line Completion

Completion is a very handy function, and all modern shells (including bash) have it. Its role is to give the user
as little work to do as possible. The best way to illustrate completion is to give an example.

3.5.1. Example
Suppose your personal directory contains the file_with_very_long_name_impossible_to_type file, and you
want to look at it. Suppose you also have, in the same directory, another file called file_text. You are in your
personal directory, so type the following sequence:
$ less fi<TAB>

(i.e., type less fi and then press the TAB key). The shell will then expand the command line for you:
$ less file_

and also give the list of possible choices (in its default configuration, which can be customized). Then type the
following key sequence:
less file_w<TAB>

and the shell will extend the command line to give you the result you want:
less file_with_very_long_name_impossible_to_type

All you need to do then is press the Enter key to confirm and read the file.

3.5.2. Other Completion Methods

The TAB key is not the only way to activate completion, although it is the most common one. As a general
rule, the word to be completed will be a command name for the first word of the command line (nsl<TAB>
will give nslookup), and a file name for all the others, unless the word is preceded by a “magic” character like
~, @ or $, in which case the shell will try to complete a user name, a machine name or an environment variable
name respectively7. There is also a magic character for completing a file name (/) and a command to recall a
command from the history (!).
The other two ways to activate completion are the sequences Esc-<x> and Ctrl+x <x>, where <x> is one of
the magic characters already mentioned. Esc-<x> will attempt to come up with a unique completion. If it fails,
it will complete the word with the largest possible substring in the choice list. A beep means either that the
choice is not unique, or simply that there is no corresponding choice. The sequence Ctrl+x <x> displays the
list of possible choices without attempting any completion. Pressing the TAB key is the same as successively
pressing Esc-<x> and Ctrl+x <x>, where the magic character depends on the context.
Thus, one way to see all the environment variables defined is to type the sequence Ctrl+x $ on a blank line.
Another example: if you want to see the man page for the nslookup command, you simply type man nsl then
Esc-!, and the shell will automatically complete the command to man nslookup.

6. Yes, it will indeed be a PNG image (however, the ImageMagick package needs to be installed...).
7. Remember: UNIX differentiates between uppercase and lowercase. The HOME environment variable and the home va-
riable are not the same.

23
Chapter 3. Introduction to the Command Line

3.6. Starting and Handling Background Processes: Job Control

You have probably noticed that when you enter a command from a Terminal, you normally have to wait for
the command to finish before the shell returns control to you. This means that you have sent the command in
the foreground. However, there are occasions when this is not desirable.
Suppose, for example, that you decide to copy a large directory recursively to another. You have also decided
to ignore errors, so you redirect the error channel to /dev/null:
cp -R images/ /shared/ 2>/dev/null

Such a command can take several minutes until it is fully executed. You then have two solutions: the first one
is violent, and means stopping (killing) the command and then doing it again when you have the time. To do
this, press Ctrl+c: this will terminate the process and take you back to the prompt. But wait, don’t do it yet!
Read on.
Suppose you want the command to run while you do something else. The solution is then to put the process
into the background. To do this, press Ctrl+z to suspend the process:
$ cp -R images/ /shared/ 2>/dev/null
# Type C-z here
[1]+ Stopped cp -R images/ /shared/ 2>/dev/null
$

and there you are again at the prompt. The process is then on standby, waiting for you to restart it (as shown
by the Stopped keyword). That, of course, is what you want to do, but in the background. Type bg (for Back-
Ground) to get the desired result:
$ bg
[1]+ cp -R images/ /shared/ 2>/dev/null &
$

The process will then start running again as a background task, as indicated by the & (ampersand) sign at the
end of the line. You will then be back at the prompt and able to continue working. A process which runs as a
background task, or in the background, is called a background job.
Of course, you can start processes directly as background tasks by adding an & character at the end of the
command. For example, you can start the command to copy the directory in the background by writing:
cp -R images/ /shared/ 2>/dev/null &

If you want, you can also restore this process to the foreground and wait for it to finish by typing fg
(ForeGround). To put it into the background again, type the sequence Ctrl+z, bg.
You can start several jobs this way: each command will then be given a job number. The shell command jobs
lists all the jobs associated with the current shell. The job preceded by a + sign indicates the last process begun
as a background task. To restore a particular job to the foreground, you can then type fg <n> where <n> is the
job number, i.e. fg 5.
Note that you can also suspend or start full-screen applications this way, such as less or a text editor like Vi,
and restore them to the foreground when you want.

3.7. A Final Word

As you can see, the shell is very comprehensive and using it effectively is a matter of practice. In this relatively
long chapter, we have only mentioned a few of the available commands: Mandrakelinux has thousands of
utilities, and even the most experienced users do not make use of them all.
There are utilities for all tastes and purposes: you have utilities for handling images (like convert mentioned
above, but also GIMP batch mode and all pixmap handling utilities), sound (Ogg Vorbis encoders, audio CD
players), CD writing, e-mail clients, FTP clients and even web browsers (like lynx or links), not to mention all
the administration tools.
Even if graphical applications with equivalent functions do exist, they are usually graphical interfaces built
around these very same utilities. In addition, command-line utilities have the advantage of being able to
operate in non-interactive mode: you can start writing a CD and then log off the system with the confidence
that the writing will take place (see the nohup(1) man page or the screen(1) man page).

24
Chapter 4. Text Editing: Emacs and VI
As stated in the introduction, text editing1 is a fundamental feature when using a UNIX system. The two editors
we are going to take a quick look at are a little difficult to use initially, but once you understand the basics,
each one can prove to be a powerful tool.

4.1. Emacs
Emacs is probably the most powerful text editor in existence. It can do absolutely everything and is infinitely
extensible through its built-in lisp-based programming language. With Emacs, you can move around the web,
read your mail, take part in Usenet newsgroups, make coffee, and so on. This is not to say that you will learn
how to do all of that in this chapter, but you will get a good start with opening Emacs, editing one or more
files, saving them and quitting Emacs.

4.1.1. Short Presentation

Invoking Emacs is done as follows:
emacs [file] [file...]

Emacs will open every file entered as an argument into a separate buffer, with a maximum of two buffers
visible at a time. If you start Emacs without specifying any files on the command line you will be placed into
a buffer called *scratch*. If you are in X, menus will be available, but in this chapter we will concentrate on
working strictly with the keyboard.

4.1.2. Getting Started

It’s now time to get some hands-on experience. For our example, let us start by opening two files, file1 and
file2. If these files do not exist, they will be created as soon as you write something in them:
$ emacs file1 file2

By typing that command, the following window will be displayed:

Figure 4-1. Editing Two Files at Once

As you can see, one buffer has been created. A third one is also present at the bottom of the screen (where you
see (New file)). That is the mini-buffer. You cannot access this buffer directly. You must be invited by Emacs
during interactive entries. To change the current buffer, type Ctrl+x o. You type text just as in a “normal”
editor, deleting characters with the DEL or Backspace key.

1. “To edit text” means to modify the contents of a file containing only letters, digits, and punctuation symbols. Such files
may be e-mail messages, source code, documents, or even configuration files.

25
Chapter 4. Text Editing: Emacs and VI

To move around, you can use the arrow keys, or you could use the following key combinations: Ctrl+a to go
to the beginning of the line, Ctrl+e to go to the end of the line, Alt+< to go to the beginning of the buffer and
Alt+> to go to the end of the buffer. There are many other combinations, even ones for each of the arrow keys
2
.
Once you are ready to save your changes to disk, type Ctrl+x Ctrl+s, or if you want to save the contents of
the buffer to another file, type Ctrl+x Ctrl+w. Emacs will ask you for the name of the file that the contents of
the buffer should be written to. You can use completion to do this.

4.1.3. Handling buffers

If you want, you can switch to displaying a single buffer on the screen. There are two ways of doing this:

• If you are in the buffer that you want to hide: type Ctrl+x 0
• If you are in the buffer which you want to keep on the screen: type Ctrl+x 1.
There are two ways of restoring a buffer back to the screen:

• type Ctrl+x b and enter the name of the buffer you want, or
• type Ctrl+x Ctrl+b. This will open a new buffer called *Buffer List*. You can move around this buffer
using the sequence Ctrl+x o, then select the buffer you want and press the Enter key, or else type the name
of the buffer in the mini-buffer. The buffer *Buffer List* returns to the background once you have made
your choice.
If you have finished with a file and you want to get rid of the associated buffer, type Ctrl+x k. Emacs will then
ask you which buffer it should close. By default, this will be the buffer you are currently in. If you want to get
rid of a buffer other than the one suggested, enter its name directly or press TAB: Emacs will open yet another
buffer called *Completions* giving the list of possible choices. Confirm the choice with the Enter key.
You can also restore two visible buffers to the screen at any time. To do this type Ctrl+x 2. By default, the new
buffer created will be a copy of the current buffer (which enables you, for example, to edit a large file in several
places “at the same time”). To move between buffers, use the commands that were previously mentioned.
You can open other files at any time, using Ctrl+x Ctrl+f. Emacs will prompt you for the file name and you
can again use completion if you find it more convenient.

4.1.4. Copy, Cut, Paste, Search

Suppose you find yourself in the following situation: figure 4-2.

Figure 4-2. Emacs, before copying the text block

2. Emacs has been designed to work on a great variety of machines, some of which do not have arrow keys on their
keyboards. This is even more true of Vi.

26
 Chapter 4. Text Editing: Emacs and VI

First off, you will need to select the text you want to copy. In this example we want to copy the entire sentence.
The first step is to place a mark at beginning of the area. Assuming the cursor is in the position where it is
in figure 4-2, the command sequence would be Ctrl+ SPACE (Control + space bar). Emacs will display the
message Mark set in the mini-buffer. Next, move to the beginning of the line with Ctrl+a. The area selected
for copying or cutting is the entire area located between the mark and the cursor’s current position, so in this
case it will be the entire line of text. There are two command sequences available: Alt+w (to copy) or Ctrl+w
(to cut). If you copy, Emacs will briefly return to the mark position so that you can view the selected area.
Finally, go to the buffer where you want the text to end up and type Ctrl+y. This will give you the following
result:

Figure 4-3. Copying Text with emacs

In fact, what you’ve done is copy text to Emacs’s kill ring. This kill ring contains all of the areas copied or cut
since Emacs was started. Any area just copied or cut is placed at the top of the kill ring. The Ctrl+y sequence
only “pastes” the area at the top. If you want to access any of the other areas, press Ctrl+y then Alt+y until
you get to the desired text.
To search for text, go to the desired buffer and type Ctrl+s. Emacs will ask you what string it should search for.
To continue a search with the same string in the current buffer, just type Ctrl+s again. When Emacs reaches
the end of the buffer and finds no more occurrences, you can type Ctrl+s again to restart the search from the
beginning of the buffer. Pressing the Enter key ends the search.
To search and replace, type Alt+%. Emacs asks you what string to search for, what to replace it with, and asks
for confirmation for each occurrence it finds.
To Undo, type Ctrl+x u which will undo the previous operation. You can undo as many operations as you
want.

4.1.5. Quit emacs

The shortcut to quit Emacs is Ctrl+x Ctrl+c. If you have not saved your changes, Emacs will ask you whether
you want to save your buffers or not.

4.2. Vi: the ancestor

Vi was the first full-screen editor in existence. It is one of the main programs UNIX detractors point to, but also
one of the better arguments of its defenders: while it is complicated to learn, it is also an extremely powerful
tool once you get into the habit of using it. With a few keystrokes, a Vi user can move mountains, and other
than Emacs, few text editors can make the same claims.

27
Chapter 4. Text Editing: Emacs and VI

The version supplied with Mandrakelinux is in fact vim, for VI iMproved, but we will refer to it as Vi throughout
this chapter.

4.2.1. Insert Mode, Command Mode, ex Mode...

To begin using Vi we use the same sort of command line as we did with Emacs. So let us go back to our two
files and type:
$ vi file1 file2

At this point, you will find yourself looking at a window resembling the following one:

Figure 4-4. Starting position in VIM

You are now in command mode in front of the first opened file. In this mode you cannot insert text into a file.
To do so you must switch to insert mode.
Here are some shortcuts to inserting text:

• a and i: to insert text after and before the cursor (A and I insert text at the end and at the beginning of the
current line);
• o and O : to insert text below and above the current line.
In insert mode, you will see the string --INSERT-- appear at the bottom of the screen (so you know which
mode you are in). This is the only mode which will allow you to insert text. To return to command mode,
press the Esc key.
In insert mode, you can use the Backspace and DEL keys to delete text as you go along. The arrow keys will
allow you to move around within the text in Command mode and Insert mode. In command mode, there are
also other key combinations which we will look at later.
ex mode is accessed by pressing the : key in command mode. A : will appear at the bottom left of the screen
with the cursor positioned after it. Vi will consider everything you type up to the Enter key as an ex command.
If you delete the command and the : you typed in, you will be returned to command mode and the cursor will
go back to its original position in the text.
To save changes to a file type :w in command mode. If you want to save the contents of the buffer to another
file, type :w <file_name>.

4.2.2. Handling Buffers

To move, in the same buffer, between the files whose names were passed on the command line, type :next
to move to the next file and :prev to move to the previous file. You can also use :e <file_name>, which
allows you to either change to the desired file if it is already open, or to open another file. You may also use
completion.

28
 Chapter 4. Text Editing: Emacs and VI

As with Emacs, you can have several buffers displayed on the screen. To do this, use the :split command.
To change buffers, type Ctrl+w j to go to the buffer below or Ctrl+w k to go to the buffer above. You can also
use the up and down arrow keys instead of j or k. The :close command hides a buffer and the :q command
closes it.
You should be aware that if you try to hide or close a buffer without saving the changes, the command will
not be carried out and Vi will display this message:
No write since last change (use ! to override)
In this case, do as you are told and type :q! or :close!.

4.2.3. Editing Text and Move Commands

Apart from the Backspace and DEL keys in edit mode, Vi has many other commands for deleting, copying,
pasting, and replacing text in command mode. All the commands shown hereafter are in fact separated into
two parts: the action to be performed and its effect. The action may be:

• c: to replace (Change). The editor deletes the requested text and goes back into insert mode after this com-
mand.
• d: to delete (Delete);
• y: to copy (“Yank”). We will look at this in the next section.
• .: repeats last action.
The effect defines which group of characters the command acts upon.

• h, j, k, l: one character left, down, up, right3

• e, b, w: to the end, beginning of the current word and the beginning of the next word.
• ^, 0, $: to the first non-blank character of the current line, the beginning of the current line, and the end of
current line.
• f<x>: go to next occurrence of character <x>. For example, fe moves the cursor to the next occurrence of the
character e.
• /<string>, ?<string>: to the next and previous occurrence of string or regexp <string>. For example,
/foobar moves the cursor to the next occurrence of the word foobar.
• {, }: to the beginning, to the end of current paragraph;
• G, H: to end of file, to beginning of screen.
Each of these “effect” characters or move commands can be preceded by a repetition number. For G, (“Go”)
this references the line number in the file. Based on this information, you can make all sorts of combinations.
Here are some examples:

• 6b: moves 6 words backwards;

• c8fk: delete all text until the eighth occurrence of the character k then go into insert mode;
• 91G: goes to line 91 of the file;
• d3$: deletes up to the end of the current line plus the next two lines.
While many of these commands are not very intuitive, the best method to remember the commands is to
practice them. But you can see that the expression “move mountains with a few keys” is not much of an
exaggeration.

3. A shortcut for dl (delete one character forward) is x; a shortcut for dh is X; dd deletes the current line.

29
Chapter 4. Text Editing: Emacs and VI

4.2.4. Cut, Copy, Paste

Vi contains a command which we have already seen for copying text: the y command. To cut text, simply use
the d command. There are 27 memories or buffers for storing text: an anonymous memory and 26 memories
named after the 26 lowercase letters of the alphabet.
To use the anonymous memory you enter the command “as is”. So the command y12w will copy the 12 words
after the cursor into anonymous memory4. Use d12w if you want to cut this area.
To use one of the 26 named memories, enter the sequence "<x> before the command, where <x> gives the
name of the memory. Therefore, to copy the same 12 words into the memory k, you would write "ky12w, or
"kd12w to cut them.
To paste the contents of the anonymous memory, use the commands p or P (for Paste), to insert text after or
before the cursor. To paste the contents of a named memory, use "<x>p or "<x>P in the same way (for example
"dp will paste the contents of memory d after the cursor).
Let us look at an example:.

Figure 4-5. VIM, before copying the text block

To carry out this action, we will:

• recopy the first 6 words of the sentence into memory r (for example): "ry6w5;
• go into the buffer file2, which is located below: Ctrl+w j;
• paste the contents of memory r before the cursor: "rp.
We get the expected result, as shown in figure 4-6.

4. But only if the cursor is positioned at the beginning of the first word!
5. y6w literally means: “Yank 6 words”.

30
 Chapter 4. Text Editing: Emacs and VI

Figure 4-6. VIM, after having copied the text block

Searching for text is very simple: in command mode, you simply type / followed by the string to search for,
and then press the Enter key. For example, /party will search for the string party from the current cursor
position. Pressing n takes you to the next occurrence, and if you reach the end of the file, the search will start
again from the beginning. To search backwards, use ? instead of /.

4.2.5. Quit Vi
The command to quit is :q (in fact, this command actually closes the active buffer, as we have already seen,
but if it is the only buffer open, you will quit Vi). There is a shortcut: most of the time you edit only one file.
So to quit, you will use:

• :wq to save changes and quit (a quicker solution is ZZ), or

• :q! to quit without saving.
You should note that if you have several buffers, :wq will only write the active buffer and then close it.

4.3. A last word...

Of course, we have said much more here than was necessary (after all, the first aim was to edit a text file), but
hopefully we have been able to show you some of the possibilities of each of these editors. There is a great
deal more to be said about them, as witnessed by the number of books dedicated to each of these editors.
Take the time to absorb all this information, opt for one of them, or learn only as much as you think necessary.
But at least you know that when you want to go further, you can.

31
Chapter 4. Text Editing: Emacs and VI

32
Chapter 5. Command-Line Utilities
The purpose of this chapter is to introduce a number of command-line tools that may prove useful for every-
day use. Of course, you may skip this chapter if you only intend to use a graphical environment, but a quick
glance may change your opinion.
Each command will be illustrated by an example, but this chapter is meant as an exercise in order for you to
fully grasp their function and use.

5.1. File Operations and Filtering

Most command-line work is done on files. In this section we discuss how to watch and filter file content, take
required information from files using a single command, and to sort files.

5.1.1. cat, tail, head, tee: File Printing Commands

These commands have almost the same syntax:
command_name [option(s)] [file(s)]

and may be used in a pipe. All of them are used to print part of a file according to certain criteria.
The cat utility concatenates files and prints to the standard output. This is one of the most widely used com-
mands. You can use:
# cat /var/log/mail/info

to print, for example, the content of a mailer daemon log file to standard output1. The cat command has a
very useful option (-n) which allows you to print numbers of all output lines.
Some files, like daemon log files (if they are running) are usually huge in size2 and printing them completely on
the screen is not very useful. Often you need to see only some lines of the file. You can use the tail command
to see the last lines. So the following command will print the last 10 lines of the file /var/log/mail/info:
# tail /var/log/mail/info

Value 10 is the default if no single parameter was specified. If you want to display only the last 2 lines you can
use the -n option:
# tail -n2 /var/log/mail/info

The head command is similar to tail, but it prints the first lines of a file. Used without options it prints the
first 10 lines of the specified file:
# head /var/log/mail/info

As with tail you can print the first 2 lines of this file using the -n option:
# head -n2 /var/log/mail/info

You can also use these commands together. For example, if you wish to display only lines 9 and 10, you can
use a command where the head command will select the first 10 lines from a file and pass them through a pipe
to the tail command.
# head /var/log/mail/info | tail -n2

The last part will then select the last 2 lines and will print them to the screen. In the same way you can select
line number 20 from the end of a file:
# tail -n20 /var/log/mail/info |head -n1

1. Some examples in this section are based on real work with log files of some servers (services, daemons). Make sure the
syslogd is running (allows daemon’s logging), corresponding daemon (in our case Postfix) and you work as root. Anyway
you can always apply our examples to other files.
2. For example, the /var/log/mail/info file contains information about all sent mails, messages about fetching mail by
users with the POP protocol, etc.

33
Chapter 5. Command-Line Utilities

In this example we tell tail to select the file’s last 20 lines and pass them through a pipe to head. Then the
head command prints to the screen the first line from the obtained data.
Let’s suppose we want to print the result of the last example to the screen and save it to the file results.txt
at the same time. The tee utility can help us. Its syntax is:
tee [option(s)] [file]

Now we can change the previous command this way:

# tail -n20 /var/log/mail/info |head -n1|tee results.txt

Let’s take yet another example. We want to select the last 20 lines, save them to results.txt, but print on
screen only the first of the 20 selected lines. Then we should type:
# tail -n20 /var/log/mail/info |tee results.txt |head -n1

The tee command has a useful option (-a) which allows you to append received data to an existing file.
Let’s go back to the tail command. Files such as logs usually vary dynamically because the daemon cons-
tantly add actions and events to the log file. So, if you want to interactively watch the changes to the log file
you can take advantage of one more of tail’s useful options: -f:
# tail -f /var/log/mail/info

In this case all changes in the /var/log/mail/info file will be printed on screen immediately. Using the tail
command with option -f is very helpful when you want to know how your system works. For example, loo-
king through the /var/log/messages log file, you can keep up with system messages and various daemons.
Also you can use tail with the -f option to see changes in any other files.
In the next section we will see how we can use grep as a filter to separate Postfix messages from messages
coming from other services.

5.1.2. grep: Locate Strings in Files

Neither the name nor the acronym (“General Regular Expression Parser”) is very intuitive, but what it does
and its use are simple: grep looks for a pattern given as an argument in one or more files. Its syntax is:
grep [options] <pattern> [one or more file(s)]

If several files are mentioned, their names will precede each matching line displayed in the result. Use the -h
option to prevent the display of these names; use the -l option to get nothing but the matching filenames. The
pattern is a regular expression, even though most of the time it consists of a simple word. The most frequently
used options are the following:

• -i: make a case insensitive search (i.e. ignore differences between lower and uppercase);
• -v: invert search. display lines which do not match the pattern;
• -n: display the line number for each line found;
• -w: tells grep that the pattern should match a whole word.
So let’s go back to analyze the mailer daemon’s log file. We want to find all lines in the file /var/log/mail/
info which contain the “postfix” pattern. Then we type this command:
# grep postfix /var/log/mail/info

The grep command can be used in a pipe. Thus we can get the same result as in the previous example by doing
this:
# cat /var/log/mail/info | grep postfix

If we want to find all lines not containing the “postfix” pattern, we would use the -v option:
# grep -v postfix /var/log/mail/info

34
 Chapter 5. Command-Line Utilities

Let’s suppose we want to find all messages about successfully sent mails. In this case we have to filter all
lines which were added into the log file by the mailer daemon (contains the “postfix” pattern) and they must
contain a message about successful sending (“status=sent”):
# grep postfix /var/log/mail/info |grep status=sent

In this case grep is used twice. It is allowed, but not very elegant. We can get the same result by using the fgrep
utility. First, we need to create a file containing patterns written out in a column. Such a file can be created this
way (we use patterns.txt as the file name):
# echo -e ’status=sent\npostfix’ >./patterns.txt

Then we call the next command where we use the patterns.txt file with a list of patterns and the fgrep
utility instead of the “double calling” of grep:
# fgrep -f ./patterns.txt /var/log/mail/info

The file ./patterns.txt may contain as many patterns as you wish. Each of them has to be typed as a single
line. For example, to select messages about successfully sent mails to [email protected], it would be
enough to add this email into our ./patterns.txt file by running this command:
# echo ’[email protected]’ >>./patterns.txt

It is clear that you can combine grep with tail and head. If we want to find messages about the last but one
email sent to [email protected] we type:
# fgrep -f ./patterns.txt /var/log/mail/info | tail -n2 | head -n1

Here we apply the filter described above and place the result in a pipe for the tail and head commands. They
select the last but one value from received data.

5.1.3. wc: Count Elements in Files

The wc command (Word Count) is used to count the number of strings and words in files. It is also helpful to
count bytes, characters and the length of the longest line. Its syntax:
wc [option(s)] [file(s)]

The following options are useful:

• -l: print the number of new lines;

• -w: print the number of words;
• -m: print the total number of characters;
• -c: print the number of bytes;
• -L: print the length of the longest line in the obtained text.
The wc command prints the number of newlines, words and characters by default. Here some usage examples:
If we want to find the number of users in our system, we can type:
$wc -l /etc/passwd

If we want to know the number of CPU’s in our system, we write:

$grep "model name" /proc/cpuinfo |wc -l

In the previous section we obtained a list of messages about successfully sent mails to e-mail addresses listed
in our ./patterns.txt file. If we want to know the number of such messages, we can redirect our filter’s
results in a pipe for the wc command:
# fgrep -f ./patterns.txt /var/log/mail/info | wc -l

35
Chapter 5. Command-Line Utilities

5.1.4. sort: Sorting File Content

Here is the syntax of this powerful sorting utility3:
sort [option(s)] [file(s)]

Let’s consider sorting on part of the /etc/passwd file. As you can see this file is not sorted:
$ cat /etc/passwd

If we want to sort it by login field. Then we type:

$ sort /etc/passwd

The sort command sorts data in ascending order starting by the first field (in our case, the login field) by
default. If we want to sort data in descending order, we use the option -r:
$ sort -r /etc/passwd

Every user has his own UID written in the /etc/passwd file. Let’s sort a file in ascending order using the UID
field:
$ sort /etc/passwd -t":" -k3 -n

Here we use the following sort options:

• -t":": tells sort that the field separator is the ":" symbol;
• -k3: means that sorting must be done on the third column;
• -n: says that the sort is to occur on numerical data, not alphabetical.
The same can be done in reverse:
$ sort /etc/passwd -t":" -k3 -n -r

Note that sort has two other important options:

• -u: perform a strict ordering: duplicate sort fields are discarded;

• -f: ignore case (treat lowercase as uppercase characters).
Finally, if we want to find the user with the highest UID we can use this command:
$ sort /etc/passwd -t":" -k3 -n |tail -n1

where we sort the /etc/passwd file in ascending order according to the UID column, and redirect the result
through a pipe to the tail command which will print out the first value of the sorted list.

5.2. find: Find Files According to Certain Criteria

find is a long-standing UNIX utility. Its role is to recursively scan one or more directories and find files which
match a certain set of criteria in those directories. Even though it is very useful, the syntax is truly obscure,
and using it requires a little work. The general syntax is:
find [options] [directories] [criterion] [action]

If you do not specify any directory, find will search the current directory. If you do not specify criteria, this is
equivalent to “true”, thus all files will be found. The options, criteria and actions are so numerous that we will
only mention a few of each here. Let’s start with options:

• -xdev: do not search on directories located on other file systems

3. We only discuss sort briefly here because whole books can be written about its features.

36
 Chapter 5. Command-Line Utilities

• -mindepth <n>: descend at least <n> levels below the specified directory before searching for files
• -maxdepth <n>: search for files which are located at most n levels below the specified directory
• -follow: follow symbolic links if they link to directories. By default, find does not follow links.
• -daystart: when using tests related to time (see below), take the beginning of current day as a timestamp
instead of the default (24 hours before current time).
A criteria may be one or more of several atomic tests. Some useful tests are:

• -type <type>: search for a given type of file. <type> can be one of: f (regular file), d (directory), l (symbolic
link), s (socket), b (block mode file), c (character mode file) or p (named pipe).
• -name <pattern>: find files whose names match the given <pattern>. With this option, <pattern> is treated
as a shell globbing pattern (see chapter Shell Globbing Patterns, page 20);
• -iname <pattern>: like -name, but ignore case
• -atime <n>, -amin <n>: find files that have last been accessed <n> days ago (-atime) or <n> minutes ago
(-amin). You can also specify +<n> or -<n>, in which case the search will be done for files accessed at most
or at least <n> days/minutes ago;
• -anewer <file>: find files which have been accessed more recently than file <file>
• -ctime <n>, -cmin <n>, -cnewer <file>: same as for -atime, -amin and -anewer, but applies to the last
time that the contents of the file were modified
• -regex <pattern>: same as -name, but pattern is treated as a regular expression
• -iregex <pattern>: same as -regex, but ignore case.
There are many other tests, refer to find(1) for more details. To combine tests, you can use one of:

• <c1> -a <c2>: true if both <c1> and <c2> are true; -a is implicit, therefore you can type <c1> <c2> <c3> ... if
you want all tests <c1>, <c2>, ... to match
• <c1> -o <c2>: true if either <c1> or <c2> are true, or both. Note that -o has a lower precedence than -a,
therefore if you want to match files which match criteria <c1> or <c2> and match criterion <c3>, you will
have to use parentheses and write ( <c1> -o <c2> ) -a <c3>. You must escape (deactivate) parentheses, as
otherwise they will be interpreted by the shell!
• -not <c1>: inverts test <c1>, therefore -not <c1> is true if <c1> is false.
Finally, you can specify an action for each file found. The most frequently used are:

• -print: just prints the name of each file on the standard output. This is the default action.
• -ls: prints on the standard output the equivalent of ls -ilds for each file found.
• -exec <command>: execute command <command> on each file found. The command line <command> must end
with a ;, which you must escape so that the shell does not interpret it; the file position is marked with {}.
See the usage examples.
• -ok <command>: same as -exec but asks for confirmation for each command.
The best way to consolidate all of the options and parameters is with some examples. Let’s say you want to
find all directories in the /usr/share directory. You would type:

find /usr/share -type d

Suppose you have an HTTP server, all your HTML files are in /var/www/html, which is also your current
directory. You want to find all files whose contents have not been modified for a month. Because you have
pages from several writers, some files have the html extension and some have the htm extension. You want to
link these files in directory /var/www/obsolete. You would type4:

find \( -name "*.htm" -o -name "*.html" \) -a -ctime -30 \

-exec ln {} /var/www/obsolete \;

4. Note that this example requires that /var/www and /var/www/obsolete be on the same file system!

37
Chapter 5. Command-Line Utilities

This is a fairly complex example, and requires a little explanation. The criterion is this:

\( -name "*.htm" -o -name "*.html" \) -a -ctime -30

which does what we want: it finds all files whose names end either in .htm or .html “ \( -name "*.htm" -o
-name "*.html" \) ”, and (-a) which have not been modified in the last 30 days, which is roughly a month (-
ctime -30). Note the parentheses: they are necessary here, because -a has a higher precedence. If there weren’t
any, all files ending with .htm would have been found, plus all files ending with .html and which haven’t been
modified for a month, which is not what we want. Also note that parentheses are escaped from the shell: if
we had put ( .. ) instead of \( .. \), the shell would have interpreted them and tried to execute -name
"*.htm" -o -name "*.html" in a sub-shell... Another solution would have been to put parentheses between
double quotes or single quotes, but a backslash here is preferable as we only have to isolate one character.
And finally, there is the command to be executed for each file:
-exec ln {} /var/www/obsolete \;

Here too, you have to escape the ; from the shell, because otherwise the shell interprets it as a command
separator. If you happen to forget, find will complain that -exec is missing an argument.
A last example: you have a huge directory (/shared/images) containing all kinds of images. Regularly, you
use the touch command to update the times of a file named stamp in this directory, so that you have a time
reference. You want to find all JPEG images which are newer than the stamp file, but because you got the ima-
ges from various sources, these files have extensions jpg, jpeg, JPG or JPEG. You also want to avoid searching
in the old directory. You want this file list to be mailed to you, and your user name is peter:

find /shared/images -cnewer \

/shared/images/stamp \
-a -iregex ".*\.jpe?g" \
-a -not -regex ".*/old/.*" \
| mail peter -s "New images"

Of course, this command is not very useful if you have to type it each time, and you would like it to be
executed regularly. A simple way to have the command run periodically is:

5.3. Commands Startup Scheduling

5.3.1. crontab: reporting or editing your crontab file

crontab is a command which allows you to execute commands at regular time intervals, with the added bonus
that you don’t have to be logged in. crontab will have the output of your command mailed to you. You can
specify the intervals in minutes, hours, days, and even months. Depending on the options, crontab will act
differently:

• -l: Print your current crontab file.

• -e: Edit your crontab file.
• -r: Remove your current crontab file.
• -u <user>: Apply one of the above options for user <user>. Only root can do this.
Let’s start by editing a crontab. If you type crontab -e, you will be in front of your favorite text editor if
you have set the EDITOR or VISUAL environment variable, otherwise Vi will be used. A line in a crontab file is
made of six fields. The first five fields are time intervals for minutes, hours, days in the month, months and
days in the week. The sixth field is the command to be executed. Lines beginning with a # are considered to be
comments and will be ignored by crond (the program which is responsible for executing crontab files). Here
is an example of crontab:

38
 Chapter 5. Command-Line Utilities

in order to print this out in a readable font, we had to break up long

lines. Therefore, some chunks must be typed on a single line. When
the \ character ends a line, this means this line is to be continued.
This convention works in Makefile files and in the shell as well as
in other contexts.

# If you don’t want to be sent mail, just comment

# out the following line
#MAILTO=""
#
# Report every 2 days about new images at 2 pm,
# from the example above - after that, "retouch"
# the "stamp" file. The "%" is treated as a
# newline, this allows you to put several
# commands in a same line.
0 14 */2 * * find /shared/images \
-cnewer /shared/images/stamp \
-a -iregex ".*\.jpe?g" \
-a -not -regex \
".*/old/.*"%touch /shared/images/stamp
#
# Every Christmas, play a melody :)
0 0 25 12 * mpg123 $HOME/sounds/merryxmas.mp3
#
# Every Tuesday at 5pm, print the shopping list...
0 17 * * 2 lpr $HOME/shopping-list.txt

There are several ways to specify intervals other than the ones shown in this example. For example, you can
specify a set of discrete values separated by commas (1,14,23) or a range (1-15), or even combine both of
them (1-10,12-20), optionally with a step (1-12,20-27/2). Now it’s up to you to find useful commands to
put in it!

5.3.2. at: schedule a command, but only once

You may also want to launch a command at a given day, but not regularly. For example, you want to be
reminded of an appointment, today at 6pm. You run X, and you would like to be notified at 5:30pm, for
example, that you must go. at is what you want here:
$ at 5:30pm
# You’re now in front of the "at" prompt
at> xmessage "Time to go now! Appointment at 6pm"
# Type CTRL-d to exit
at> <EOT>
$

You can specify the time in different ways:

• now +<interval>: Means, well, now, plus an interval (optionally. No interval specified means just now).
The syntax for the interval is <n> (minutes|hours|days|weeks|months). For example, you can specify now
+ 1 hour (an hour from now), now + 3 days (three days from now) and so on.
• <time> <day>: Fully specify the date. The <time> parameter is mandatory. at is very liberal in what it
accepts: you can for example type 0100, 04:20, 2am, 0530pm, 1800, or one of three special values: noon,
teatime (4pm) or midnight. The <day> parameter is optional. You can specify this in different ways as well:
12/20/2001 for example, which stands for December 20th, 2001, or, the European way, 20.12.2001. You
may omit the year, but then only the European notation is accepted: 20.12. You can also specify the month
in full letters: Dec 20 or 20 Dec are both valid.
at also accepts different options:

• -l: Prints the list of currently queued jobs; the first field is the job number. This is equivalent to the atq
command.

39
Chapter 5. Command-Line Utilities

• -d <n>: Remove job number <n> from the queue. You can obtain job numbers from atq. This is equivalent
to atrm <n>.
As usual, see the at(1) manpage for more options.

5.4. Archiving and Data Compression

5.4.1. tar: Tape ARchiver

Although we have already seen a use for tar in the “Building and Installing Free Software”, page 71 chapter, we
have not explained how it works, which we will do in this section. Just like find, tar is a long standing UNIX
utility, so its syntax is a bit special. The syntax is:

tar [options] [files...]

Here is a list of some options. Note that all of them have an equivalent long option, but you will have to refer
to the manual page for this as we will not list them here.

the initial dash (-) of short options is now deprecated with tar,
except after a long option.

• c: this option is used in order to create new archives

• x: this option is used in order to extract files from an existing archive
• t: lists files from an existing archive
• v: lists the files which are added to an archive or extracted from an archive, or, in conjunction with the t
option (see above), it outputs a long listing of files instead of a short one
• f <file>: create archive with name <file>, extract from archive <file> or list files from archive <file>.
If this parameter is omitted, the default file will be /dev/rmt0, which is generally the special file associated
with a streamer. If the file parameter is - (a dash), the input or output (depending on whether you create an
archive or extract from one) will be associated to the standard input or standard output
• z: tells tar that the archive to create should be compressed with gzip, or that the archive to extract from is
compressed with gzip
• j: same as z, but the program used for compression is bzip2
• p: when extracting files from an archive, preserve all file attributes, including ownership, last access time
and so on. Very useful for file system dumps.
• r: append the list of files given on the command line to an existing archive. Note that the archive to which
you want to append files should not be compressed!
• A: append archives given on the command line to the one submitted with the f option. Similar to r, the
archives should not be compressed in order for this to work.
There are many, many, many other options, so you may want to refer to the tar(1) manual page for the entire
list. See, for example, the d option. Let’s proceed with an example. Say you want to create an archive of
all images in /shared/images, compressed with bzip2, named images.tar.bz2 and located in your home
directory. You would then type:

#
# Note: you must be in the directory from which
# you want to archive files!
#
$ cd /shared
$ tar cjf ~/images.tar.bz2 images/

40
 Chapter 5. Command-Line Utilities

As you can see, we used three options here: c told tar we wanted to create an archive, j to compress it with
bzip2, and f ~/images.tar.bz2 that the archive was to be created in our home directory, and its name will
be images.tar.bz2. We may want to check if the archive is valid now. We can do this by listing its files:

#
# Get back to our home directory
#
$ cd
$ tar tjvf images.tar.bz2

Here, we told tar to list (t) files from archive images.tar.bz2 (f images.tar.bz2), warned that this archive
was compressed with bzip2 (j), and that we wanted a long listing (v). Now, say you have erased the images
directory. Fortunately, your archive is intact, and you now want to extract it back to its original place, in
/shared. But as you don’t want to break your find command for new images, you need to preserve all file
attributes:

#
# cd to the directory where you want to extract
#
$ cd /shared
$ tar jxpf ~/images.tar.bz2

And here you are!

Now, let’s say you want to extract the directory images/cars from the archive, and nothing else. Then you can
type this:

$ tar jxf ~/images.tar.bz2 images/cars

If you try to back up special files, tar will take them as what they are, special files, and will not dump their
contents. So yes, you can safely put /dev/mem in an archive. It also deals correctly with links, so do not worry
about this either. For symbolic links, also look at the h option in the manpage.

5.4.2. bzip2 and gzip: Data Compression Programs

You can see that we have already talked of these two programs when dealing with tar. Unlike WinZip in
Windows, archiving and compressing are done using two separate utilities –– tar for archiving, and the two
programs which we will now introduce for compressing data: bzip2 and gzip. You might also use a different
compression tool, programs like zip, arj or rar also exist for GNU/Linux (but they are rarely used).
At first, bzip2 was written as a replacement for gzip. Its compression ratios are generally better, but on the
other hand, it requires more RAM while working. The reason that gzip is still used is that it is still more
widespread than bzip2.
Both commands have a similar syntax:

gzip [options] [file(s)]

If no filename is given, both gzip and bzip2 will wait for data from the standard input and send the result to
the standard output. Therefore, you can use both programs in pipes. Both programs also have a set of common
options:

• -1, ..., -9: set the compression ratio. The higher the number, the better the compression, but better also means
slower.
• -d: uncompress file(s). This is equivalent to using gunzip or bunzip2.

41
Chapter 5. Command-Line Utilities

• -c: dump the result of compression/decompression of files given as parameters to the standard output.

By default, both gzip and bzip2 erase the file(s) that they have
compressed (or uncompressed) if you don’t use the -c option. You
can avoid doing this in bzip2 by using the -k option. gzip has no
equivalent option.

Now some examples. Let’s say you want to compress all files ending with .txt in the current directory using
bzip2. You would type:
$ bzip2 -9 *.txt

Let’s say you want to share your image archives with someone, but he does not have bzip2, only gzip. You
don’t need to uncompress the archive and re-compress it, you can just uncompress to the standard output, use
a pipe, compress from standard input and redirect the output to the new archive. Like this:
bzip2 -dc images.tar.bz2 | gzip -9 >images.tar.gz

You could have typed bzcat instead of bzip2 -dc. There is an equivalent for gzip but its name is zcat, not
gzcat. You also have bzless for bzip2 file and zless for gzip if you want to view compressed files directly
instead of having to uncompress them first. As an exercise, try and find the command you would have to type
in order to view compressed files without uncompressing them, and without using bzless or zless.

5.5. Many, many more...

There are so many commands that a comprehensive book about them would be the size of an encyclopedia.
This chapter hasn’t even covered a tenth of the subject, yet you can do much with what you learned here.
If you wish, you may read some manual pages: sort(1), sed(1) and zip(1) (yes, that’s what you think: you can
extract or make .zip archives with GNU/Linux), convert(1), and so on. The best way to get accustomed to these
tools is to practice and experiment with them, and you will probably find a lot of uses for them, even quite
unexpected ones. Have fun!

42
Chapter 6. Process Control

6.1. More About Processes

It is possible to monitor processes and to “ask” them to terminate, pause, continue, etc. To understand the
examples we’re going to examine, it is helpful to know a bit more about processes.

6.1.1. The Process Tree

As with files, all processes that run on a GNU/Linux system are organized in the form of a tree. The root of this
tree is init, a system-level process which is started at boot time. Each process is assigned a number by the
system to uniquely identify it (its PID, Process ID), and also inherits the PID of its parent process (PPID, Parent
Process ID). The PID and PPID of init is 1: init is its own father.

6.1.2. Signals
Every process in UNIX can react to signals sent to it. There are 64 different signals which are identified either
by their number (starting from 1) or by their symbolic names (SIGx, where x is the signal’s name). The 32
“higher” signals (33 to 64) are real-time signals and are beyond the scope of this chapter. For each of these
signals, the process can define its own behavior, except for two signals: signal number 9 (KILL) and number
19 (STOP).
Signal 9 terminates a process irrevocably without giving it the time to terminate properly. This is the signal
you send to a process which is stuck or exhibits other problems. A full list of signals is available using the kill
-l command.

6.2. Information on Processes: ps and pstree

These two commands display a list of processes currently running on the system, according to criteria you set.

6.2.1. ps
Sending this command without an argument will show only processes initiated by you and attached to the
terminal you are using:

$ ps
PID TTY TIME CMD
18614 pts/3 00:00:00 bash
20173 pts/3 00:00:00 ps

As with many UNIX utilities, ps has a handful of options, the most common of which are:

• a: displays processes started by other users;

• x: displays processes with no control terminal or with a control terminal different to the one you are using;
• u: displays for each process the name of the user who started it and the time at which it was started.
There are many other options. Refer to the ps(1) manual page for more information.
The output of this command is divided into different fields: the one that will interest you the most is the PID
field which contains the process identifier. The CMD field contains the name of the executed command. A very
common way of invoking ps is as follows:
$ ps ax | less

This gives you a list of all processes currently running so that you can identify one or more processes which
are causing problems, and subsequently terminate them.

43
Chapter 6. Process Control

6.2.2. pstree
The pstree command displays processes in the form of a tree structure. One advantage is that you can imme-
diately see which is the parent process of what: when you want to kill a whole series of processes and if they
are all parents and children, you can simply kill the parent. You will want to use the -p option to display the
PID of each process, and the -u option to show the name of the user who started the process. Because the tree
structure is generally quite long, you will want to invoke pstree in the following way:
$ pstree -up | less

This gives you an overview of the whole process tree structure.

6.3. Sending Signals to Processes: kill, killall and top

6.3.1. kill, killall

These two commands are used to send signals to processes. The kill command requires a process number as
an argument, while killall requires a process name.
Both of these commands can optionally receive the signal number of the signal to be sent as an argument. By
default, they both send the signal 15 (TERM) to the relevant process(es). For example, if you want to kill the
process with PID 785, you enter the command:
$ kill 785

If you want to send it signal 19 (STOP), you enter:

$ kill -19 785

Suppose instead that you want to kill a process where you know the command name. Instead of finding the
process number using ps, you can kill the process by its name:
$ killall -9 mozilla

Whatever happens, you will only kill your own processes (unless you are root) so you do not need to worry
about your “neighbor’s” processes if you’re running on a multi-user system since they will not be affected.

6.3.2. Mixing ps and kill: top

top is a program that simultaneously fulfills the functions of ps and kill and is also used to monitor processes
in real-time giving information about CPU and memory usage, running time, etc. as shown in figure 6-1.

Figure 6-1. Monitoring Processes with top

The top utility is entirely keyboard controlled. You can access help by pressing h. Its most useful commands
are the following:

44
 Chapter 6. Process Control

• k: this command is used to send a signal to a process. top will then ask you for the process’ PID followed
by the number of the signal to be sent (TERM — or 15 — by default);
• M: this command is used to sort processes by the amount of memory they take up (field %MEM);
• P: this command is used to sort processes by the CPU time they take up (field %CPU; this is the default sorting
method);
• u: this one is used to display a given user’s processes, top will ask you which one. You need to enter the
user’s name, not his UID. If you do not enter any name, all processes will be displayed;
• i: by default, all processes, even sleeping ones, are displayed. This command ensures that only processes
currently running are displayed (processes whose STAT field shows R, Running) and not the others. Using
this command again takes you back to showing all processes.

6.4. Setting Priority to Processes: nice, renice

Every process in the system is running with defined priorities (also called “nice value”). This value may vary
from -20 to +20. The maximum priority value for processes is -20. If it is not defined, every process will run
with a default priority of 0 (the “base” scheduling priority). Processes with maximum priority (any negative
value up to -20) use more system resources than others. Processes with minimal priority (+20) will work when
the system is not used by other tasks. Users other than the super-user may only lower the priority of processes
they own within a range of 0 to 20. The super-user (root) may set the priority of any process to any value.

6.4.1. renice
If one or more processes use too many system resources, you can change their priorities instead of killing
them. For such tasks the renice command can be used. Its syntax is as follows:
renice priority [[-p] pid ...] [[-g] pgrp ...] [[-u] user ...]

where priority is the value of the priority, pid (use option -p for multiple processes) is the process ID, pgrp
(introduced by -g if various) is the process group ID, and user (-u for more than one) is the user name of the
process owner.
Let’s suppose you have run a process with PID 785, which makes a long scientific operation, and while it is
working you want to play a game. You would type:
$ renice +15 785

In this case your process could potentially take longer to complete but will not take CPU time from other
processes.
If you are the system administrator and you see that some user is running too many processes and they use
too many system resources, you can change that user’s process priority with a single command:
# renice +20 -u peter

After this, all of peter’s processes will have the lowest priority and will not obstruct any other user’s processes.

6.4.2. nice
Now that you know that you can change the priority of processes, you may wish to run a command with a
defined priority. For this, use the nice command.
In this case you need to specify your command as an option to nice. By default nice sets a priority of 10. The
“niceness” range goes from -20 (highest priority) to 19 (lowest). Option -n is used to set priority value.
For example, you want to create an ISO image of a Mandrakelinux installation CD-ROM:
$ dd if=/dev/cdrom of=~/mdk1.iso

45
Chapter 6. Process Control

On some systems with a standard IDE CD-ROM, the process of copying large volumes of information can use
too many system resources. To prevent the copying from blocking all other processes, you can start the process
with a lowered priority by using this command:
$ nice -n 19 dd if=/dev/cdrom of=~/mdk1.iso

and continue with what you were doing.

To change a process’ priority you also can use the above described top utility. Use the r command within top’s
interface to change the priority of selected process.

46
Chapter 7. File-Tree Organization
Nowadays, a UNIX system is big, very big. This is especially true of GNU/Linux: the amount of software avai-
lable would make for an unmanageable system if there weren’t any guidelines for the location of files in the
tree.
The acknowledged standard is the FHS (Filesystem Hierarchy Standard), which was at version 2.3 at press time.
The document which describes the standard is available on the Internet in different formats on The Pathname
web site (http://www.pathname.com/fhs/). This chapter will only provide a brief summary, but it should be
enough to show you which directory is likely to contain a given file, or where a given file should be placed.

7.1. Shareable/Unshareable, Static/Variable Data

Data on a UNIX system can be classified according to the following criteria: shareable data may be common
to several computers in a network, while unshareable cannot. Static data must not be modified in normal use,
while variable data may be. As we explore the tree structure, we’ll classify the different directories into each
of these categories.
Note that these classifications are only a recommendation. It’s not mandatory to follow them, but adopting
these guidelines will greatly help you manage your system. Also, bear in mind that the static/variable dis-
tinction only applies to general system usage, not its configuration. If you install a program, you’ll obviously
have to modify “normally” static directories, such as /usr/.

7.2. The root Directory: /

The root directory contains the entire system hierarchy. It cannot be classified since its subdirectories may or
may not be static or shareable. Here is a list of the main directories and subdirectories, with their classifications:

• /bin/: essential binary files. It contains the basic commands which will be used by all users and which are
necessary for the operation of the system: ls, cp, login, etc. Static, unshareable.
• /boot/: contains the files required by the GNU/Linux bootloader (GRUB or LILO for Intel, yaboot for PPC,
etc). It may or may not contain the kernel, but if the kernel isn’t located in this directory then it must be in
the root directory. Static, unshareable.
• /dev/: system device files (dev for DEVices). Static, unshareable.
• /etc/: contains all configuration files specific to the computer. Static, unshareable.
• /home/: where all the personal directories of the system’s users are located. This directory may or may not
be shared (some large networks make it shareable via NFS). Variable, shareable.
• /lib/: it contains libraries which are essential to the system; it also stores kernel modules in the /lib/
modules/KERNEL_VERSION/ subdirectory. All libraries required by the binaries in the /bin/ and /sbin/ di-
rectories must be located here, together with the ld.so linker. Static, unshareable.
• /mnt/: directory containing the mount points for temporarily-mounted file systems. Variable, unshareable.
• /opt/: holds packages not essential for system operation. It’s recommended that static files (binaries, li-
braries, manual pages, etc.) to be placed in /opt/package_name and the specific configuration files in
/etc/opt/.
• /root/: home directory for root. Variable, unshareable.
• /usr/: explained in more detail in /usr/: The Big One, page 47. Static, shareable.
• /sbin/: contains system binaries essential for system start-up. Most of its files can only be executed by root.
A normal user may run them, but they won’t do anything for a regular user. Static, unshareable.
• /tmp/: directory intended to contain temporary files which certain programs may create. Variable, unsha-
reable.
• /var/: location for data which may be modified in real time by programs (such as mail servers, audit pro-
grams, print servers, etc.). Variable. Its various subdirectories may be shareable or unshareable.

47
Chapter 7. File-Tree Organization

7.3. /usr/: The Big One

The /usr/ directory is the main application-storage directory. The binary files in this directory are not required
for system start-up or maintenance, so the /usr/ hierarchy may be, and often is, located on a separate file
system. Because of its (usually) large size, /usr/ has its own hierarchy of subdirectories. We will mention just
a few:

• /usr/X11R6: the entire X Window System hierarchy. All binaries and libraries required for the operation of X
(including the X servers) must be located here. The /usr/X11R6/lib/X11 directory contains all aspects of X’s
configuration which do not vary from one computer to another. Specific configurations for each computer
should go in /etc/X11.
• /usr/bin/: contains the large majority of the system’s binaries. Any binary program which isn’t neces-
sary for the maintenance of the system and isn’t a system administration program must be located in this
directory. The only exceptions are programs you compile and install yourself, which must be located in
/usr/local/.
• /usr/lib/: contains all the necessary libraries to run programs located in /usr/bin/ and /usr/sbin/. The-
re is also a /usr/lib/X11 symbolic link pointing to /usr/X11R6/lib, the directory which contains the X
Window System libraries (but only if X is installed).
• /usr/local/: this is where you must install applications you compile from source. The installation program
should create the necessary hierarchy: lib/, bin/, etc.
• /usr/share/: this directory contains all architecture-independent data required by applications in /usr/.
Among other things, you’ll find zone and location information (zoneinfo and locale).
Let’s also mention the /usr/share/doc and /usr/share/man directories, which respectively contain applica-
tion documentation and the system’s manual pages.

7.4. /var/: Data Modifiable During Use

The /var directory contains all operative data for programs running on the system. Unlike the working data
in /tmp/, this data must be kept intact in the event of a reboot. There are many subdirectories, and some are
very useful:

• /var/log/: contains the system’s log files.

• /var/spool/: contains the working files of the system’s daemons. For example, /var/spool/cups/ con-
tains the print server’s working files, while /var/spool/mail/ contains the mail server’s working files (for
example, all mail arriving on and leaving your system).
• /var/run/: used to keep track of all processes utilized by the system, enabling you to act on them in the
event of a system change runlevel (see “The Start-Up Files: init sysv”, page 69).

7.5. /etc/: Configuration Files

/etc/ is one of UNIX systems’ most essential directories because it contains all the basic system configuration
files. Never delete it to save space! Likewise, if you want to extend your tree structure over several partitions,
remember that /etc/ must not be put on a separate partition: it is needed for system initialization and must
be on the root partition at boot time.
Here are some important files:

• passwd and shadow: these are text files which contain all system users and their encrypted passwords. You
will only see shadow if shadow passwords are used, which happens to be the default installation option for
security reasons.
• inittab: this is the configuration file for init which plays a fundamental role in starting up the system.
• services: this file contains a list of existing network services.
• profile: this is the shell configuration file, although certain shells use other files. For example, bash uses
bashrc.

48
 Chapter 7. File-Tree Organization

• crontab: the configuration file for cron, the program responsible for periodic execution of commands.
Certain subdirectories exist for programs which require a large number of configuration files. This applies to
the X Window System, for example, which stores all of its configuration files in the /etc/X11/ directory.

49
Chapter 7. File-Tree Organization

50
Chapter 8. File Systems and Mount Points
The best way to understand “how it works” is to look at a practical case, which is what we’re going to do here.
Suppose you just purchased a brand new hard disk with no partitions on it. Your Mandrakelinux partition is
completely full, and rather than starting again from scratch, you decide to move a whole section of the tree
structure to your new hard disk. Because your new disk has a lot of capacity, you decide to move your biggest
directory to it: /usr. But first, a bit of theory.

8.1. Principles
Every hard disk is divided into several partitions, and each of these contains a file system. While Windows
assigns a letter to each of these file systems (well, actually only to those it recognizes), GNU/Linux has a unique
tree structure of files, and each file system is mounted at one location in that tree structure.
Just as Windows needs a C:drive, GNU/Linux must be able to mount the root of its file tree (/) on a partition
which contains the root file system. Once the root is mounted, you can mount other file systems in the tree
structure at various mount points within the tree. Any directory below the root structure can act as a mount
point, and you can mount the same file system several times on different mount points.
This allows great configuration flexibility. For example, if you were to configure a web server, it’s fairly com-
mon to dedicate an entire partition to the directory which hosts the web server’s data. The directory which
usually contains the data is /var/www and acts as the mounting point for the partition. You can see in figure
8-1 and figure 8-2 how the system looks before and after mounting the file system.

/home /usr /var

/cgi−bin /html /icons

/var/www

Root file system File system containing files

(already mounted) of directory "/var/www"
(not yet mounted)

Figure 8-1. A Not Yet Mounted File System

/home /usr /var

/var/www

/var/www/cgi−bin /var/www/html /var/www/icons

Figure 8-2. File System Is Now Mounted

51
Chapter 8. File Systems and Mount Points

As you can imagine, this offers a number of advantages: the tree structure will always be the same, whether
it’s on a single file system or extended over several dozen.1 This flexibility allows you to move a key part of
the tree structure to another partition when space becomes scarce, which is what we are going to do here.
There are two things you need to know about mount points:

1. The directory which acts as a mount point must exist.

2. And this directory should preferably be empty: if a directory chosen as a mount point already contains
files and subdirectories, these will simply be “hidden” by the newly mounted file system. The files will
not be deleted, but they will not be accessible until you free the mount point.

It’s actually possible to access the data“hidden”by the newly moun-

ted file system. You simply need to mount the hidden directory with
the --bind option. For example, if you just mounted a directory
in /hidden/directory/ and want to access its original content in
/new/directory, you would have to run:

mount --bind /hidden/directory/ /new/directory

8.2. Partitioning a Hard Disk, Formatting a Partition

There are two things to keep in mind as you read through this section: a hard disk is divided into partitions,
and each of these partitions hosts a file system. Your brand new hard disk has neither, so we begin with
partitioning. In order to proceed, you must be root.
First, you have to know the hard disk’s “name” (i.e.: what file designates it). Suppose the new drive is set up as
the slave on your primary IDE interface. In that case, it will be known as /dev/hdb.2. Please refer to the Starter
Guide’s Managing Your Partitions section, which will explain how to partition a disk. DiskDrake will also create
the file systems for you, so once the partitioning and file system creation steps are complete, we can proceed.

8.3. The mount and umount Commands

Now that the file system has been created, you can mount the partition. Initially, it will be empty, since the
system hasn’t had access to the file system for files to have been added to it. The command to mount file
systems is mount , and its syntax is as follows:

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

In this case, we want to temporarily mount our partition on /mnt (or any other mount point you have chosen:
remember that the mount point must exist). The command for mounting our newly created partition is:
$ mount -t ext2 /dev/hdb1 /mnt

The -t option is used to specify what type of file system the partition is supposed to host. The file systems
you will most frequently encounter are ext2FS (the GNU/Linux file system) or ext3FS (an improved version
of ext2FS with journaling capabilities), VFAT (for all DOS/Windows partitions: FAT 12, 16 or 32) and ISO9660
(CD-ROM file system). If you don’t specify any type, mount will try and guess which file system is hosted by
the partition by reading the superblock.
The -o option is used to specify one or more mounting options. The options appropriate for a particular file
system will depend on the file system being used. Refer to the mount(8) man page for more details.
Now that you’ve mounted your new partition, it’s time to copy the entire /usr directory onto it:

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

1. GNU/Linux can manage up to 64 simultaneously mounted file systems.

2. Determining the name of a disk is explained in the Installation Guide.

52
 Chapter 8. File Systems and Mount Points

Now that the files are copied, we can unmount our partition. To do this, use the umount command. The syntax
is simple:
umount <mount point|device>

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

$ umount /mnt

or:
$ umount /dev/hdb1

Since this partition is going to “become” our /usr directory, we need to tell the system. To do this, we edit
the /etc/fstab file. It makes it possible to automate the mounting of certain file systems, especially at system
start-up. It contains a series of lines describing the file systems, their mount points and other options. Here’s
an example of such a file:
/dev/hda1 / ext2 defaults 1 1
/dev/hda5 /home ext2 defaults 1 2
/dev/hda6 swap swap defaults 0 0
none /mnt/cdrom supermount dev=/dev/scd0,fs=udf:iso9660,ro,-- 0 0
none /mnt/floppy supermount dev=/dev/fd0,fs=ext2:vfat,--,sync,umask=0 0 0
none /proc proc defaults 0 0
none /dev/pts devpts mode=0622 0 0

Each line consists of:

• the device hosting the file system;

• the mount point;
• the type of file system;
• the mounting options;
• the dump utility backup flag;
• fsck’s (FileSystem ChecK) checking order.
There is always an entry for the root file system. The Swap partitions are special since they’re not visible in the
tree structure, and the mount point field for those partitions contains the swap keyword. As for the /proc file
system, it will be described in more detail in “The /proc Filesystem”, page 63. Another special file system is
/dev/pts.
At this point, we have moved the entire /usr hierarchy to /dev/hdb1 and we want this partition to be mounted
as /usr at boot time. To accomplish this, add the following entry to the /etc/fstab file:
/dev/hdb1 /usr ext2 defaults 1 2

Now the partition will be mounted at each boot, and will be checked for errors if necessary.
There are two special options: noauto and user. The noauto option specifies that the file system should not
be mounted at start-up, and is mounted only when you tell it to. The user option specifies that any user can
mount and unmount the file system. These two options are typically used for the CD-ROM and floppy drives.
There are other options, and /etc/fstab has a man page (fstab(5)) you can read for more information.
One advantage of using /etc/fstab is that it simplifies the mount command syntax. To mount a file system
described in the file, you can either reference the mount point or the device. To mount a floppy disk, you can
type:
$ mount /mnt/floppy

or:
$ mount /dev/fd0

To finish our partition moving example, let’s review what we’ve already done. We copied the /usr hierarchy
and modified /etc/fstab so that the new partition will be mounted at start-up. But for the moment, the old

53
Chapter 8. File Systems and Mount Points

/usr files are still in their original place on the drive, so we need to delete them to free up space (which was,
after all, our initial goal). To do so, you first need to switch to single user mode by issuing the telinit 1
command on the command line.

• Next, we delete all files in the /usr directory. Remember that we are still referring to the “old” directory,
since the newer, larger one, is not yet mounted. rm -Rf /usr/*.
• Finally, we mount the new /usr directory: mount /usr/.
And that’s it. Now, go back to multi-user mode (telinit 3 for standard text mode or telinit 5 for the X
Window System), and if there’s no further administrative work left, you should now log off from the root
account.

54
Chapter 9. The Linux File System
Naturally, your GNU/Linux system is contained on your hard disk within a file system. Here, we will discuss
the various aspects of file systems available on GNU/Linux, as well as the possibilities they offer.

9.1. Comparison of a Few File Systems

During installation, you can choose different file systems for your partitions, so they will be formatted using
different algorithms.
Unless you are a specialist, choosing a file system is not obvious. We will take a quick look at a few current file
systems, all of which are available with Mandrakelinux.

9.1.1. Different Usable File Systems

9.1.1.1. Ext2
The Second Extended File System (its abbreviated form is Ext2 or simply ext2) has been GNU/Linux’s default
file system for many years. It replaced the Extended File System (that’s where the “Second” comes from). The
“new” file system corrected certain problems and limitations of its predecessor.
Ext2 respects the usual standards for Unix-type file systems. Since its inception, it was destined to evolve while
still offering great robustness and good performance.

9.1.1.2. Ext3
Like its name suggests, the Third Extended File System is Ext2’s successor. It is compatible with the latter but
enhanced by incorporating journaling.
One of the major flaws of “traditional” file systems like Ext2 is their low tolerance to abrupt system break-
downs (power failure or crashing software). Generally speaking, once the system is restarted, these sorts of
events involve a very long examination of the file system’s structure and attempts to correct errors, which
sometimes result in an extended corruption. This corruption could cause partial or total loss of saved data.
Journaling answers this problem. To simplify, let’s say that what we are doing is storing the actions (such as
the saving of a file) before really doing it. We could compare this functionality to that of a boat captain who
uses a log book to note daily events. The result: an always coherent file system. And if problems occur, the
verification is very rapid and the eventual repairs, very limited. The time spent in verifying a file system is
thus proportional to its actual use and not related to its size.
So, Ext3 offers journal file system technology while keeping Ext2’s structure, ensuring excellent compatibility.
This makes it very easy to switch from Ext2 to Ext3 and back again.

9.1.1.3. ReiserFS
Unlike Ext3, ReiserFS was written from scratch. It is a journalized file system like Ext3, but its internal struc-
ture is radically different because it uses binary-tree concepts inspired by database software.

9.1.1.4. JFS
JFS is the journalized file system designed and used by IBM. Proprietary and closed at first, IBM decided to
open it to access by the free software movement. Its internal structure is similar to that of ReiserFS.

55
Chapter 9. The Linux File System

9.1.1.5. XFS
XFS is the journalized file system designed by SGI and used in ins IRIX operating system. Proprietary and
closed at first, SGI decided to open it to access by the free software movement. Its internal structure has lots of
different features, such as support for real-time bandwidth, extents, and clustered file systems (but not in the
free version).

9.1.2. Differences Between the File Systems

Ext2 Ext3 ReiserFS JFS XFS

Stability Excellent Good Good Medium Good
Tools to restore Yes (complex) Yes (complex) No No No
erased files
Reboot time Long, even very Fast Very fast Very fast Very fast
after crash long
Status of the Generally Very good Mediuma Very good Very good
data in case of a speaking, good,
crash but high risk of
partial or total
data loss
ACL support Yes Yes No No Yes
Notes:
a. It is possible to improve results on crash recovery by journaling the data and not just the metadata,
adding the option data=journal to /etc/fstab.
Table 9-1. File System Characteristics

The maximum size of a file depends on many parameters (e.g. the block size for ext2/ext3), and is likely
to evolve depending on the kernel version and architecture. According to the file system limits, the cu-
rrent maximum size is currently near or greater than 2 TeraBytes (TB, 1 TB=1024 GB) and for JFS can go
up to 4 PetaBytes (PB, 1 Pb=1024 TB). Unfortunately, these values are also limited to maximum block de-
vice size, which in the current 2.4.X kernel is limited (for X86 arch only) to 2TB1 even in RAID mode. In
kernel 2.6.X this block device limit could be extended using a kernel compiled with Large Block Device
support enabled (CONFIG_LBD=y). For more information, consult Adding Support for Arbitrary File Si-
zes to the Single UNIX Specification (http://ftp.sas.com/standards/large.file/x_open.20Mar96.html),
Large File Support in Linux (http://www.suse.com/~aj/linux_lfs.html), and Large Block Devices (http:
//www.gelato.unsw.edu.au/IA64wiki/LargeBlockDevices).

9.1.3. And Performance Wise?

It is always very difficult to compare performance between file systems. All tests have their limitations and
the results must be interpreted with caution. Nowadays, Ext2 is very mature but its development is slow; Ext3
and ReiserFS are quite mature at this point. New features for ReiserFS are included in ReiserFS42. In the other
hand XFS has a lot of features, and as time passes more of the advanced features work better on linux. JFS took
a different approach here, and they are integrating on linux feature by feature. This makes the process slower,
but they are also finishing with a very clean code base. Comparisons done a couple of months or weeks ago
are already too old. Let us not forget that today’s material (specially concerning hard drive capacities) has
greatly leveraged the differences between them. XFS has the advantage that just now it is the better performer
with large streaming files.

1. You may wonder how to achieve such capacities with hard drives that barely store 320-400GB. Using for instance one
RAID card with 8 * 250GB drives in RAID-striping, you can achieve 2TB of storage. Combining the storage of several RAID
cards using Linux software RAID, or using LVM (logical volume manager) it should be possible to go even beyond (block
size permitting) the 2TB limit.
2. At the time of writing, ReiserFS4 was not included in kernel 2.6.X

56
 Chapter 9. The Linux File System

Each system offers advantages and disadvantages. In fact, it all depends on how you use your machine. A
simple desktop machine will be happy with Ext2. For a server, a journalized file system like Ext3 is preferred.
ReiserFS, perhaps because of its genesis, is more suited to a database server. JFS is preferred in cases were file
system throughput is the main issue. XFS is interesting if you need any of its advanced features.
For “normal” use, the four file systems give approximately the same results. ReiserFS allows you to access
small files rapidly, but it is fairly slow in manipulating large files (many megabytes). In most cases, the advan-
tages brought by ReiserFS’ journaling capabilities outweigh its drawbacks. Notice that by default ReiserFS is
mounted with the notail option. That means that there is no optimization for small files and that big files run
at normal speed.

9.2. Everything is a File

The Starter Guide introduced the file ownership and permissions access concepts, but really understanding the
UNIX file system (and this also applies to Linux’ file systems) requires that we redefine the concept of “What is
a file.”.
Here, “everything” really means everything. A hard disk, a partition on a hard disk, a parallel port, a connec-
tion to a web site, an Ethernet card: all these are files. Even directories are files. Linux recognizes many types
of files in addition to the standard files and directories. Note that by file type here, we do not mean the type
of content of a file: for GNU/Linux and any UNIX system, a file, whether it be a PNG image, a binary file or
whatever, is just a stream of bytes. Differentiating files according to their contents is left to applications.

9.2.1. The Different File Types

If you remember, when you do ls -l, the character before the access rights identifies the type of a file. We
have already seen two types of files: regular files (-) and directories (d). You can also find other types if you
wander through the file tree and list the contents of directories:

1. Character mode files: these files are either special system files (such as /dev/null, which we have already
discussed), or peripherals (serial or parallel ports), which share the trait that their contents (if they have
any) are not buffered (meaning they are not kept in memory). Such files are identified by the letter c.
2. Block mode files: these files are peripherals, and unlike character files, their contents are buffered. For
example, some files in this category are: hard disks, partitions on a hard disk, floppy drives, CD-ROM
drives and so on. Files /dev/hda, /dev/sda5 are example of block mode files. In ls -l output, these are
identified by the letter b.
3. Symbolic links: these files are very common, and heavily used in the Mandrakelinux system startup proce-
dure (see chapter “The Start-Up Files: init sysv”, page 69). As their name implies, their purpose is to link
files in a symbolic way, which means that they are files whose content is the path to a different file. They
may not point to an existing file. They are very frequently called “soft links”, and are identified by an
’l’.
4. Named pipes: in case you were wondering, yes, these are very similar to pipes used in shell commands, but
with the difference that these actually have names. Read on to learn more. They are very rare, however,
and it is not likely that you will see one during your journey into the file tree. Just in case you do, the letter
identifying them is p. To learn more, have a look at “Anonymous” Pipes and Named Pipes, page 59.
5. Sockets: this is the file type for all network connections, but only a few of them have names. What’s more,
there are different types of sockets and only one can be linked, but this is way beyond the scope of this
book. Such files are identified by the letter s.
Here is a sample of each file:
$ ls -l /dev/null /dev/sda /etc/rc.d/rc3.d/S20random /proc/554/maps \
/tmp/ssh-queen/ssh-510-agent
crw-rw-rw- 1 root root 1, 3 May 5 1998 /dev/null
brw-rw---- 1 root disk 8, 0 May 5 1998 /dev/sda
lrwxrwxrwx 1 root root 16 Dec 9 19:12 /etc/rc.d/rc3.d/
S20random -> ../init.d/random*
pr--r--r-- 1 queen queen 0 Dec 10 20:23 /proc/554/maps|
srwx------ 1 queen queen 0 Dec 10 20:08 /tmp/ssh-queen/
ssh-510-agent=

57
Chapter 9. The Linux File System

9.2.2. Inodes
Inodes are, along with the “Everything Is a File” paradigm, a fundamental part of any UNIX file system. The
word “inode” is short for Information NODE.
Inodes are stored on disk in an inode table. They exist for all types of files which may be stored on a file
system, including directories, named pipes, character mode files and so on. Which leads to this other famous
sentence: “The inode is the file”. Inodes are how UNIX identifies a file in a unique way.
No, you did not misread that: in UNIX, you do not identify a file by its name, but by its inode number. 3 The
reason for this is that the same file may have several names, or even no name. A file name, in UNIX, is just an
entry in a directory inode. Such an entry is called a link. Let us look at links in more detail.

9.3. Links
The best way to understand what links are is to look at an example. Let us create a (regular) file:
$ pwd
/home/queen/example
$ ls
$ touch a
$ ls -il a
32555 -rw-rw-r-- 1 queen queen 0 Dec 10 08:12 a

The -i option of the ls command prints the inode number, which is the first field on the output. As you can
see, before we created file a, there were no files in the directory. The other field of interest is the third one,
which is the number of file links (well, inode links, in fact).
The touch a command can be separated into two distinct actions:

• creation of an inode, to which the operating system has given the number 32555, and whose type is the one
of a regular file;
• and creation of a link to this inode, named a, in the current directory, /home/queen/example. Therefore, the
/home/queen/example/a file is a link to the inode numbered 32555, and it is currently the only one: the link
counter shows 1.
But now, if we type:
$ ln a b
$ ls -il a b
32555 -rw-rw-r-- 2 queen queen 0 Dec 10 08:12 a
32555 -rw-rw-r-- 2 queen queen 0 Dec 10 08:12 b
$

we create another link to the same inode. As you can see, we did not create a file named b. Instead, we just
added another link to the inode numbered 32555 in the same directory, and attributed the name b to this new
link. You can see on the ls -l output that the link counter for the inode is now 2 rather than 1.
Now, if we do:
$ rm a
$ ls -il b
32555 -rw-rw-r-- 1 queen queen 0 Dec 10 08:12 b
$

3. Important: note that inode numbers are unique per file system, which means that an inode with the same number can
exist on another file system. This leads to the difference between on-disk inodes and in-memory inodes. While two on-disk
inodes may have the same number if they are on two different file systems, in-memory inodes have a unique number right
across the system. One solution to obtain uniqueness, for example, is to hash the on-disk inode number against the block
device identifier.

58
 Chapter 9. The Linux File System

we see that even though we deleted the “original file”, the inode still exists. But now, the only link to it is the
file named /home/queen/example/b.
Therefore, a file in UNIX has no name; instead, it has one or more link(s) in one or more directory(ies).
Directories themselves are also stored in inodes, their link count coincides with the number of subdirectories
within them. This is due to the fact that there are at least two links per directory: the directory itself (.) and its
parent directory (..).
Typical examples of files which are not linked (i.e.: have no name) are network connections; you will never see
the file corresponding to your connection to the Mandrakelinux web site (www.mandrakelinux.com) in your file
tree, no matter which directory you look in. Similarly, when you use a pipe in the shell, the inode corresponding
to the pipe exists, but it is not linked. Other use of inodes without names is temporary files. You create a
temporary file, open it, and then remove it. The file exists while it is open, but nobody else can open it (as there
is no name for opening it). This way, if the application crashes, the temporary file is removed automatically.

9.4. “Anonymous” Pipes and Named Pipes

Let’s get back to the example of pipes, as it is quite interesting and is also a good illustration of the links
notion. When you use a pipe in a command line, the shell creates the pipe for you and operates so that the
command before the pipe writes to it, while the command after the pipe reads from it. All pipes, whether they
be anonymous (like the ones used by the shells) or named (see below) act like FIFOs (First In, First Out). We
have already seen examples of how to use pipes in the shell, but let’s take another look for the sake of our
demonstration:
$ ls -d /proc/[0-9] | head -5
/proc/1/
/proc/2/
/proc/3/
/proc/4/
/proc/5/

One thing that you will not notice in this example (because it happens too fast for one to see) is that writes on
pipes are blocking. This means that when the ls command writes to the pipe, it is blocked until a process at
the other end reads from the pipe. In order to visualize the effect, you can create named pipes, which unlike
the pipes used by shells, have names (i.e.: they are linked, whereas shell pipes are not). 4 The command to create
a named pipe is mkfifo:
$ mkfifo a_pipe
$ ls -il
total 0
169 prw-rw-r-- 1 queen queen 0 Dec 10 14:12 a_pipe|
#
# You can see that the link counter is 1, and that the output shows
# that the file is a pipe (’p’).
#
# You can also use ln here:
#
$ ln a_pipe the_same_pipe
$ ls -il
total 0
169 prw-rw-r-- 2 queen queen 0 Dec 10 15:37 a_pipe|
169 prw-rw-r-- 2 queen queen 0 Dec 10 15:37 the_same_pipe|
$ ls -d /proc/[0-9] >a_pipe
#
# The process is blocked, as there is no reader at the other end.
# Type C-z to suspend the process...
#
zsh: 3452 suspended ls -d /proc/[0-9] > a_pipe
#
# ...Then put in into the background:
#
$ bg
[1] + continued ls -d /proc/[0-9] > a_pipe
#
# now read from the pipe...
#

4. Other differences exist between the two kinds of pipes, but they are beyond the scope of this book.

59
Chapter 9. The Linux File System

$ head -5 <the_same_pipe
#
# ...the writing process terminates
#
[1] + 3452 done ls -d /proc/[0-9] > a_pipe
/proc/1/
/proc/2/
/proc/3/
/proc/4/
/proc/5/
#

Similarly, reads are also blocking. If we execute the above commands in the reverse order, we will see that
head blocks, waiting for some process to give it something to read:
$ head -5 <a_pipe
#
# Program blocks, suspend it: C-z
#
zsh: 741 suspended head -5 < a_pipe
#
# Put it into the background...
#
$ bg
[1] + continued head -5 < a_pipe
#
# ...And give it some food :)
#
$ ls -d /proc/[0-9] >the_same_pipe
$ /proc/1/
/proc/2/
/proc/3/
/proc/4/
/proc/5/
[1] + 741 done head -5 < a_pipe
$

You can also see an undesired effect in the previous example: the ls command has terminated before the
head command took over. The consequence is that you were immediately returned to the prompt, but head
executed later and you only saw its output after returning.

9.5. “Special” Files: Character Mode and Block Mode Files

As already stated, such files are either files created by the system or peripherals on your machine. We also
mentioned that the contents of block mode character files were buffered, while character mode files were not.
In order to illustrate this, insert a floppy into the drive and type the following command twice:
$ dd if=/dev/fd0 of=/dev/null

You should have observed the following: the first time the command was launched, the entire content of the
floppy was read. The second time you executed the command, there was no access to the floppy drive at all.
This is because the content of the floppy was buffered the first time you launched the command – and you did
not change anything on the floppy between the two instances.
But now, if you want to print a big file this way (yes it will work):
$ cat /a/big/printable/file/somewhere >/dev/lp0

the command will take as much time, whether you launch it once, twice or fifty times. This is because /dev/
lp0 is a character mode file, and its contents are not buffered.
The fact that block mode files are buffered has a nice side effect: not only are reads buffered, but writes are
buffered too. This allows for writes to the disks to be asynchronous: when you write a file on disk, the write
operation itself is not immediate. It will only occur when the Linux kernel decides to execute the write to the
hardware.
Finally, each special file has a major and minor number. On a ls -l output, they appear in place of the size,
as the size for such files is irrelevant:

60
 Chapter 9. The Linux File System

ls -l /dev/hda /dev/lp0
brw-rw---- 1 root disk 3, 0 May 5 1998 /dev/hda
crw-rw---- 1 root daemon 6, 0 May 5 1998 /dev/lp0

Here, the major and minor of /dev/hda are 3 and 0, whereas for /dev/lp0, they are 6 and 0. Note that these
numbers are unique per file category, which means that there can be a character mode file with major 3 and
minor 0 (this file actually exists: /dev/ttyp0), and similarly, there can be a block mode file with major 6 and
minor 0. These numbers exist for a simple reason: it allows the kernel to associate the correct operations to
these files (that is, to the peripherals these files refer to): you do not handle a floppy drive the same way as,
say, a SCSI hard drive.

9.6. Symbolic Links, Limitation of “Hard” Links

Here we have to face a very common misconception, even among UNIX users, which is mainly due to the fact
that links as we have seen them so far (wrongly called “hard” links) are only associated with regular files (and
we have seen that it is not the case – since even symbolic links are “linked”). But this requires that we first
explain what symbolic links (“soft” links, or even more often “symlinks”) are.
Symbolic links are files of a particular type whose sole content is an arbitrary string, which may or may not
point to an existing file. When you mention a symbolic link on the command line or in a program, in fact, you
access the file it points to, if it exists. For example:
$ echo Hello >myfile
$ ln -s myfile mylink
$ ls -il
total 4
169 -rw-rw-r-- 1 queen queen 6 Dec 10 21:30 myfile
416 lrwxrwxrwx 1 queen queen 6 Dec 10 21:30 mylink
-> myfile
$ cat myfile
Hello
$ cat mylink
Hello

You can see that the file type for mylink is l, for symbolic Link. The access rights for a symbolic link are not
significant: they will always be rwxrwxrwx. You can also see that it is a different file from myfile, as its inode
number is different. But it refers to it symbolically, therefore when you type cat mylink, you will in fact print
the contents of the myfile file. To demonstrate that a symbolic link contains an arbitrary string, we can do the
following:
$ ln -s "I’m no existing file" anotherlink
$ ls -il anotherlink
418 lrwxrwxrwx 1 queen queen 20 Dec 10 21:43 anotherlink
-> I’m no existing file
$ cat anotherlink
cat: anotherlink: No such file or directory
$

But symbolic links exist because they overcome several limitations encountered by normal (“hard”) links:

• You cannot create a link to an inode in a directory which is on a different file system than the said inode.
The reason is simple: the link counter is stored in the inode itself, and inodes cannot be shared between file
systems. Symlinks allow this;
• You cannot link directories to avoid creating cycles in the file system. But you can make a symlink point to
a directory and use it as if it were actually a directory.
Symbolic links are therefore very useful in several circumstances, and very often, people tend to use them to
link files together even when a normal link could be used instead. One advantage of normal linking, though,
is that you do not lose the file if you delete the “original one”.
Lastly, if you observed carefully, you know what the size of a symbolic link is: it is simply the size of the string.

61
Chapter 9. The Linux File System

9.7. File Attributes

The same way that FAT has file attributes (archive, system file, invisible), a GNU/Linux file systems has its own,
but they are different. We will briefly go over them here for the sake of completeness, but they are very seldom
used. However, if you really want a secure system, read on.
There are two commands for manipulating file attributes: lsattr(1) and chattr(1). You probably guessed
it, lsattr LiSts attributes, whereas chattr CHanges them. These attributes can only be set on directories and
regular files. The following attributes are possible:

1. A (no Access time): if a file or directory has this attribute set, whenever it is accessed, either for reading or
for writing, its last access time will not be updated. This can be useful, for example, on files or directories
which are often accessed for reading, especially since this parameter is the only one which changes on an
inode when it is open read-only.
2. a ( append only): if a file has this attribute set and is open for writing, the only operation possible will be to
append data to its previous contents. For a directory, this means that you can only add files to it, but not
rename or delete any existing file. Only root can set or clear this attribute.
3. d (no dump): dump (8) is the standard UNIX utility for backups. It dumps any file system for which the
dump counter is 1 in /etc/fstab (see chapter “File Systems and Mount Points”, page 51). But if a file or
directory has this attribute set, unlike others, it will not be taken into account when a dump is in progress.
Note that for directories, this also includes all subdirectories and files under it.
4. i ( immutable): a file or directory with this attribute set can not be modified at all: it can not be renamed,
no further link can be created to it 5 and it cannot be removed. Only root can set or clear this attribute.
Note that this also prevents changes to access time, therefore you do not need to set the A attribute when
i is set.
5. s ( secure deletion): when a file or directory with this attribute is deleted, the blocks it was occupying on
disk are overwritten with zeroes.
6. S ( Synchronous mode): when a file or directory has this attribute set, all modifications on it are synchronous
and written to the disk immediately.
For example, you may want to set the i attribute on essential system files in order to avoid bad surprises. Also,
consider the A attribute on man pages: this prevents a lot of disk operations and, in particular, can save some
battery life on laptops.

5. Be sure to understand what “adding a link” means, both for a file and a directory :-)

62
Chapter 10. The /proc Filesystem
The /proc file system is specific to GNU/Linux. It is a virtual file system, so the files that you will find in
this directory do not actually take up any space on your hard drive. It is a very convenient way to obtain
information about the system, especially since most files in this directory are human readable (well, with a
little help). Many programs actually gather information from files in /proc, format it in their own way and
then display the results. There are a few programs which display information about processes (top, ps and
friends) which do exactly that. /proc is also a good source of information about your hardware, and just like
the programs which display processes, quite a few programs are just interfaces to the information contained
in /proc.
There is also a special subdirectory, /proc/sys. It allows you to display kernel parameters and to change them,
with the changes taking effect immediately.

10.1. Information About Processes

If you list the contents of the /proc directory, you will see many directories where the name of the directory is
a number. These are the directories containing information on all processes currently running on the system:
$ ls -d /proc/[0-9]*
/proc/1/ /proc/302/ /proc/451/ /proc/496/ /proc/556/ /proc/633/
/proc/127/ /proc/317/ /proc/452/ /proc/497/ /proc/557/ /proc/718/
/proc/2/ /proc/339/ /proc/453/ /proc/5/ /proc/558/ /proc/755/
/proc/250/ /proc/385/ /proc/454/ /proc/501/ /proc/559/ /proc/760/
/proc/260/ /proc/4/ /proc/455/ /proc/504/ /proc/565/ /proc/761/
/proc/275/ /proc/402/ /proc/463/ /proc/505/ /proc/569/ /proc/769/
/proc/290/ /proc/433/ /proc/487/ /proc/509/ /proc/594/ /proc/774/
/proc/3/ /proc/450/ /proc/491/ /proc/554/ /proc/595/

Note that as a user, you can (logically) only display information related to your own processes, but not those
of other users. So, let’s be root and see what information is available from process 127:
$ su
Password:
$ cd /proc/127
$ ls -l
total 0-9
-r--r--r-- 1 root root 0 Dec 14 19:53 cmdline
lrwx------ 1 root root 0 Dec 14 19:53 cwd -> //
-r-------- 1 root root 0 Dec 14 19:53 environ
lrwx------ 1 root root 0 Dec 14 19:53 exe -> /usr/sbin/apmd*
dr-x------ 2 root root 0 Dec 14 19:53 fd/
pr--r--r-- 1 root root 0 Dec 14 19:53 maps|
-rw------- 1 root root 0 Dec 14 19:53 mem
lrwx------ 1 root root 0 Dec 14 19:53 root -> //
-r--r--r-- 1 root root 0 Dec 14 19:53 stat
-r--r--r-- 1 root root 0 Dec 14 19:53 statm
-r--r--r-- 1 root root 0 Dec 14 19:53 status
$

Each directory contains the same entries. Here is a brief description of some of the entries:

1. cmdline: this (pseudo-)file contains the entire command line used to invoke the process. It is not formatted:
there is no space between the program and its arguments, and there is no newline at the end of the line. In
order to view it, you can use: perl -ple ’s,\00, ,g’ cmdline.
2. cwd: this symbolic link points to the current working directory (hence the name) of the process.
3. environ This file contains all the environment variables defined for this process, in the form VARIA-
BLE=value. Similar to cmdline, the output is not formatted at all: no newlines separate the different varia-
bles, and there is no newline at the end. One solution to view it: perl -pl -e ’s,\00,\n,g’ environ.
4. exe: this is a symlink pointing to the executable file corresponding to the process being run.
5. fd: this subdirectory contains the list of file descriptors currently opened by the process. See below.
6. maps: when you print the content of this named pipe (with cat for example), you can see the parts of
the process’ address space which are currently mapped to a file. From left to right, the fields are: the
address space associated to this mapping, the permissions associated to this mapping, the offset from

63
Chapter 10. The /proc Filesystem

the beginning of the file where the mapping starts, the major and minor number (in hexadecimal) of the
device on which the mapped file is located, the inode number of the file, and finally the name of the file
itself. When the device is 0 and there’s no inode number or filename, this is an anonymous mapping. See
mmap(2).
7. root: this is a symbolic link which points to the root directory used by the process. Usually, it will be /,
but see chroot(2).
8. status: this file contains various information about the process: the name of the executable, its current
state, its PID and PPID, its real and effective UID and GID, its memory usage, and other information.
Note that files stat and statm are now obsolete. The information they contained is now stored in status.
If we list the contents of directory fd for process 127, we obtain this:
$ ls -l fd
total 0
lrwx------ 1 root root 64 Dec 16 22:04 0 -> /dev/console
l-wx------ 1 root root 64 Dec 16 22:04 1 -> pipe:[128]
l-wx------ 1 root root 64 Dec 16 22:04 2 -> pipe:[129]
l-wx------ 1 root root 64 Dec 16 22:04 21 -> pipe:[130]
lrwx------ 1 root root 64 Dec 16 22:04 3 -> /dev/apm_bios
lr-x------ 1 root root 64 Dec 16 22:04 7 -> pipe:[130]
lrwx------ 1 root root 64 Dec 16 22:04 9 ->
/dev/console
$

In fact, this is the list of file descriptors opened by the process. Each opened descriptor is shown by a symbolic
link, where the name is the descriptor number, and which points to the file opened by this descriptor1. Note the
permissions on the symlinks: this is the only place where they make sense, as they represent the permissions
with which the file corresponding to the descriptor has been opened.

10.2. Information on The Hardware

Apart from the directories associated to the different processes, /proc also contains a myriad of information
on the hardware present in your machine. A list of files from the /proc directory shows the following:
$ ls -d [a-z]*
apm dma interrupts loadavg mounts rtc swaps
bus/ fb ioports locks mtrr scsi/ sys/
cmdline filesystems kcore meminfo net/ self/ tty/
cpuinfo fs/ kmsg misc partitions slabinfo uptime
devices ide/ ksyms modules pci stat version
$

For example, if we look at the contents of /proc/interrupts, we can see that it contains the list of interrupts
currently used by the system, along with the peripheral which uses them. Similarly, ioports contains the list
of input/output address ranges currently busy, and lastly dma does the same for DMA channels. Therefore, in
order to chase down a conflict, look at the contents of these three files:
$ cat interrupts
CPU0
0: 127648 XT-PIC timer
1: 5191 XT-PIC keyboard
2: 0 XT-PIC cascade
5: 1402 XT-PIC xirc2ps_cs
8: 1 XT-PIC rtc
10: 0 XT-PIC ESS Solo1
12: 2631 XT-PIC PS/2 Mouse
13: 1 XT-PIC fpu
14: 73434 XT-PIC ide0
15: 80234 XT-PIC ide1
NMI: 0
$ cat ioports
0000-001f : dma1
0020-003f : pic1
0040-005f : timer
0060-006f : keyboard

1. If you remember what was described in section Redirections and Pipes, page 21, you know what descriptors 0, 1 and 2
stand for.

64
 Chapter 10. The /proc Filesystem

0070-007f : rtc
0080-008f : dma page reg
00a0-00bf : pic2
00c0-00df : dma2
00f0-00ff : fpu
0170-0177 : ide1
01f0-01f7 : ide0
0300-030f : xirc2ps_cs
0376-0376 : ide1
03c0-03df : vga+
03f6-03f6 : ide0
03f8-03ff : serial(auto)
1050-1057 : ide0
1058-105f : ide1
1080-108f : ESS Solo1
10c0-10cf : ESS Solo1
10d4-10df : ESS Solo1
10ec-10ef : ESS Solo1
$ cat dma
4: cascade
$

Or, more simply, use the lsdev command, which gathers information from these files and sorts them by pe-
ripheral, which is undoubtedly more convenient.2:
$ lsdev
Device DMA IRQ I/O Ports
------------------------------------------------
cascade 4 2
dma 0080-008f
dma1 0000-001f
dma2 00c0-00df
ESS 1080-108f 10c0-10cf 10d4-10df 10ec-10ef
fpu 13 00f0-00ff
ide0 14 01f0-01f7 03f6-03f6 1050-1057
ide1 15 0170-0177 0376-0376 1058-105f
keyboard 1 0060-006f
Mouse 12
pic1 0020-003f
pic2 00a0-00bf
rtc 8 0070-007f
serial 03f8-03ff
Solo1 10
timer 0 0040-005f
vga+ 03c0-03df
xirc2ps_cs 5 0300-030f
$

An exhaustive listing of files would take too long, but here’s the description of some of them:

• cpuinfo: this file contains, as its name says, information on the processor(s) present in your machine.
• modules: this file contains the list of modules currently used by the kernel, along with the usage count for
each one. In fact, this is the same information as that reported by the lsmod command.
• meminfo: this file contains information on memory usage at the time you print its contents. The free com-
mand. will display the same information in a easier-to-read format.
• apm: if you have a laptop, displaying the contents of this file allows you to see the state of your battery. You
can see whether the AC is plugged in, the charge level of your battery, and if the APM BIOS of your laptop
supports it (unfortunately this is not the case for all), the remaining battery life in minutes. The file isn’t very
readable by itself, therefore you want to use the apm command instead, which gives the same information
in a human readable format.
Note that modern computers now provide ACPI support instead of APM. See below.

• bus: this subdirectory contains information on all peripherals found on different buses in your machine.
The information is usually not readable, and for the most part it is reformatted with external utilities: lsp-
cidrake, lspnp, etc.

2. lsdev is part of the procinfo package.

65
Chapter 10. The /proc Filesystem

• acpi: Several of the files provided in this directory are interesting especially for laptops, where you can
select several power-saving options. Note that it is easier to modify those options through a higher level
application , such as the ones included in the acpid and kacpi packages.
The most interesting entries are:

battery
shows how many batteries are in the laptop, and related information as current remaining life, maxi-
mum capacity, etc.

button
Allows you to control actions associated to “special” buttons such as power, sleep, lid, etc.

fan
Displays the state of the fans on your computer, whether they are running or not, and enables you
to start/stop them according to certain criteria. The amount of control of the fans in your machine
depends of the motherboard.

processor
There is one subdirectory for each of the CPUs in your machine. Control options vary from one proces-
sor to another. Mobile processors have more features enabled, including:
• possibility to use several power states, balancing between performance and power consumption.
• possibility to use clock rate change in order to reduce the amount of CPU power consumption.
Note that there are several processors that do not offer these possibilities.

thermal_zone
Information about how hot your system/processor is running.

10.3. The /proc/sys Sub-Directory

The role of this subdirectory is to report different kernel parameters, and to allow you to interactively change
some of them. As opposed to all other files in /proc, some files in this directory can be written to, but only by
root.
A list of directories and files would take too long to describe, mostly because the content of the directories are
system-dependent and that most files will only be useful for very specialized applications. However, here are
three common uses of this subdirectory:

1. Allow routing: Even if the default kernel from Mandrakelinux is able to route, you must explicitly allow it
to do so. For this, you just have to type the following command as root:
$ echo 1 >/proc/sys/net/ipv4/ip_forward

Replace the 1 by a 0 if you want to forbid routing.

2. Prevent IP spoofing: IP spoofing consists of making one believe that a packet coming from the outside
world comes from the interface by which it arrives. This technique is very commonly used by crackers 3,
but you can make the kernel prevent this kind of intrusion. Type:
$ echo 1 >/proc/sys/net/ipv4/conf/all/rp_filter

and this kind of attack becomes impossible.

3. And not hackers!

66
 Chapter 10. The /proc Filesystem

3. Increase the size of the table of open files and the inode table: The size of the table of open files and the
inode table is dynamic in GNU/Linux. The default values are usually sufficient for normal use, but they
may be too conservative if your machine is a huge server (a database server for example). You will know
that you need to increase the size of the table if you get messages that processes cannot open any more
files because the table is full. If you increase the size of the open file table, then do not forget that the size
of the inode table has to be increased as well. These two lines will solve the problem:
$ echo 8192 >/proc/sys/fs/file-max
$ echo 16384 >/proc/sys/fs/inode-max

These changes will only remain in effect while the system is running. If the system is rebooted, then the values
will go back to their defaults. To reset the values to something other than the default at boot time, you can take
the commands that you typed at the shell prompt and add them to /etc/rc.d/rc.local so that you avoid
typing them each time. Another solution is to modify /etc/sysctl.conf, see sysctl.conf (5).

67
Chapter 10. The /proc Filesystem

68
Chapter 11. The Start-Up Files: init sysv
The “System V” startup scheme, inherited from AT&T UNIX is one of the UNIX traditional system-startup
schemes. This system is responsible for starting or stopping services to bring the system in one of the default
system states. Services range here from basic users authentication to local graphical server or Internet services.

The Mandrakelinux tool allowing you to manually start or stop

services is drakxservices.

11.1. In the Beginning Was init

When the system starts, and after the kernel has configured everything and mounted the root file system, it
executes /sbin/init 1. init is the father of all of the system’s processes and is responsible for taking the
system to the desired runlevel. We will look at runlevels later on (see Runlevels, page 69).
The init configuration file is called /etc/inittab. This file has its own manual page (inittab(5)), so we will
only document a few of the possible configuration values.
The first line –– which should be the focus of your attention –– is this one:
si::sysinit:/etc/rc.d/rc.sysinit

This line tells init that /etc/rc.d/rc.sysinit is to be run once the system has been initialized (si stands
for System Init). To determine the default runlevel, init will then look for the line containing the initdefault
keyword:
id:5:initdefault:

In this case, init knows that the default runlevel is 5. It also knows that to enter level 5, it must run the
following command:
l5:5:wait:/etc/rc.d/rc 5

As you can see, the syntax for each runlevel is similar.

init is also responsible for restarting (respawn) some programs which cannot be started by any other process.
For example, each of the login programs which run on the six virtual consoles are started by init.2. The second
virtual console is identified this way:
2:2345:respawn:/sbin/mingetty tty2

11.2. Runlevels
All files related to system startup are located in the /etc/rc.d directory. Here is the list of the files:
$ ls /etc/rc.d
init.d/ rc0.d/ rc2.d/ rc4.d/ rc6.d/ rc.local* rc.sysinit*
rc* rc1.d/ rc3.d/ rc5.d/ rc.alsa_default* rc.modules*

As already stated, rc.sysinit is the first file run by the system. It is responsible for setting up the basic
machine configuration: keyboard type, configuration of certain devices, file system checking, etc.
Then the rc script is run, with the desired runlevel as an argument. As we have seen, the runlevel is a simple
integer, and for each runlevel <x> defined, there must be a corresponding rc<x>.d directory. In a typical
Mandrakelinux installation, you might therefore see that that there are six runlevels:

• 0: complete machine stop.

1. Which is why putting /sbin on a file system other than the root one would be a very bad idea. The kernel has not
mounted any other partitions at this point, and therefore would not be able to find /sbin/init.
2. If you do not want six virtual consoles, you may add or remove them by modifying this file. If you wish to increase
the number of consoles, you can have up to a maximum of 64. But do not forget that X also runs on a virtual console, so
leave at least one console free for X.

69
Chapter 11. The Start-Up Files: init sysv

• 1: single-user mode. To be used in the event of major problems or system recovery.

• 2: multi-user mode, with no network.
• 3: Multi-user mode, with networking
• 4: unused.
• 5: like runlevel 3, but also launches the graphical login interface.
• 6: restart.
Let us take a look at the content of the rc5.d directory:
$ ls rc5.d
K15postgresql@ K60atd@ S15netfs@ S60lpd@ S90xfs@
K20nfs@ K96pcmcia@ S20random@ S60nfs@ S99linuxconf@
K20rstatd@ S05apmd@ S30syslog@ S66yppasswdd@ S99local@
K20rusersd@ S10network@ S40crond@ S75keytable@
K20rwhod@ S11portmap@ S50inet@ S85gpm@
K30sendmail@ S12ypserv@ S55named@ S85httpd@
K35smb@ S13ypbind@ S55routed@ S85sound@

As you can see, all the files in this directory are symbolic links, and they all have a very specific form. Their
general form is:
<S|K><order><service_name>

The S means Start service, and K means Kill (stop) service. The scripts are run in ascending numerical order,
and if two scripts have the same number, the ascending alphabetical order will apply. We can also see that
each symbolic link points to a given script located in the /etc/rc.d/init.d directory (other than the local
script which is responsible for controlling a specific service.)
When the system goes into a given runlevel, it starts by running the K links in order: the rc command looks
at where the link is pointing, then calls up the corresponding script with a single argument: stop. It then runs
the S scripts using the same method, except that the scripts are called with a start parameter.
So, without discussing all the scripts, we can see that when the system goes into runlevel 5, it first runs
the K15postgresql command, (i.e. /etc/rc.d/init.d/postgresql stop). Then K20nfs, then K20rstatd,
etc., until it has run the last command. Next, it runs all the S scripts: first S05apmd, which in turn calls
/etc/rc.d/init.d/apmd start, and so on.
Armed with this information, you can create your own complete runlevel in a few minutes (using runlevel 4,
for instance), or prevent a service from starting or stopping by deleting the corresponding symbolic link. You
can also use a number of interface programs to do this, notably drakxservices (see DrakXServices: Configuring
Start-Up Services in the Starter Guide) which uses a graphical program interface, or chkconfig for text-mode
configuration.

You can also use the chkconfig command to list, add or remove
services from a specific runlevel. See chkconfig(8).

70
Chapter 12. Building and Installing Free Software

We are often asked how to install free software from sources. Compiling software yourself is really easy be-
cause most of the steps to follow are the same no matter what the software you are installing is. The aim of
this document is to guide the beginner step by step and explain to him the meaning of each move. We assume
that the reader has a minimal knowledge of the UNIX system (ls or mkdir for instance).
This guide is only a guide, not a reference manual. That is why several links are given at the end to answer
any remaining questions. This guide can probably be improved, so we appreciate receiving any remarks or
corrections on its contents.

12.1. Introduction
What makes the difference between free software and proprietary software is the access to the source code of
the software1. This means that free software is distributed as archives of source code files. It may be unfamiliar
to beginners because users of free software must compile source code for themselves before they can use the
software.
There are compiled versions of most of the existing free software. The user in a hurry only has to install
these pre-compiled binaries. Some free software is not distributed in this form, or the earlier versions are not
distributed in binary form. Furthermore, if you use an exotic operating system or an exotic architecture, a lot
of software will not be compiled for you. More importantly, compiling software for yourself allows you to
enable only the interesting options or to extend the functionality of the software by adding extensions in order
to obtain a program which exactly fits your needs.

12.1.1. Requirements
To build software, you need:

• a computer with a working operating system,

• general knowledge of the operating system you use,
• some space on your disk,
• a compiler (usually for the C language) and an archiver (tar),
• some food (in difficult cases, it may last a long time). A real hacker eats pizzas –– not quiches.
• something to drink (for the same reason). A real hacker drinks soda –– for caffeine.
• the phone number of that techie friend who recompiles his kernel every week,
• specially patience, and a lot of it!
Compiling from source does not generally present a lot of problems, but if you are not used to it, the smallest
snag can throw you. The aim of this document is to show you how to escape from such a situation.

12.1.2. Compilation

12.1.2.1. Principle
In order to translate source code into a binary file, a compilation must be done (usually from C or C++ sources,
which are the most widespread languages used by the (UNIX) free software community). Some free software
is written in languages which do not require compilation (for instance perl or the shell), but they still require
some configuration.
C compilation is logically done by a C compiler, usually gcc, the free compiler written by the GNU project
(http://www.gnu.org/). Compiling a complete software package is a complex task, which goes through the

1. This is not completely true since some proprietary software also provides source code. But unlike what happens with
free software, the end user is not allowed to use or modify the code as he wants.

71
Chapter 12. Building and Installing Free Software

successive compilations of different source files (it is easier for various reasons for the programmer to put the
different parts of his work in separate files). In order to make it easier on you, these repetitive operations are
handled by a utility named make.

12.1.2.2. The four steps of compilation

To understand how compilation works (in order to be able to solve possible problems), you have to know the
steps involved. The objective is to little by little convert a text file written in a language that is comprehensible
to a trained human being (i.e. C language), into a language that is comprehensible to a machine (or a very well
trained human being in some cases). gcc executes four programs one after the other, each of which carries out
one step:

1. cpp: The first step consists of replacing directives (preprocessors) by pure C instructions. Typically, this
means inserting a header (#include) or defining a macro (#define). At the end of this stage, pure C code
has been generated.
2. cc1: This step consists of converting C into assembly language. The generated code depends on the target
architecture.
3. as: This step consists of generating object code (or binary code) from the assembly language. At the end
of this stage, a .o file is generated.
4. ld: The last step (linkage) links all the object files (.o) and the associated libraries together and produces
an executable file.

12.1.3. Structure of a distribution

A correctly structured free software distribution always has the same organization:

• An INSTALL file, which describes the installation procedure.

• A README file, which contains general information related to the program (short description, author, URL
where to fetch it, related documentation, useful links, etc). If the INSTALL file is missing, the README file
usually contains a brief installation procedure.
• A COPYING file, which contains the license or describes the distribution conditions of the software. Someti-
mes a LICENSE file is used instead, with the same contents.
• A CONTRIB or CREDITS file, which contains a list of people related to the software (active participation,
pertinent comments, third-party programs, etc).
• A CHANGES file (or less frequently, a NEWS file), which contains recent improvements and bug fixes.
• A Makefile file (see the section Make, page 77), which controls compilation of the software (it is a necessary
file for make). If this file does not exist at the beginning, then it is generated by a pre-compilation configura-
tion process.
• Quite often, a configure or Imakefile file, which allows one to generate a new Makefile file customized
for a particular system (see Configuration, page 74).
• A directory which contains the sources, and where the binary file is usually stored at the end of the compi-
lation. Its name is often src.
• A directory which contains the documentation related to the program (usually in man or Texinfo format),
whose name is often doc.
• Sometimes, a directory which contains data specific to the software (typically configuration files, example
of produced data, or resources files).

72
 Chapter 12. Building and Installing Free Software

12.2. Decompression

12.2.1. A tar.gz archive

The standard2 compression format under UNIX systems is the gzip format, developed by the GNU project,
and is considered as one of the best general compression tools.
gzip is often associated with a utility named tar. tar is a survivor of antediluvian times, when computerists
stored their data on tapes. Nowadays, floppy disks and CD-ROM have replaced tapes, but tar is still being
used to create archives. All the files in a directory can be appended in a single file for instance. This file can
then be easily compressed with gzip.
This is the reason why much free software is available as tar archives, compressed with gzip. So, their exten-
sions are .tar.gz (or also .tgz to shorten).

12.2.2. The use of GNU Tar

To decompress this archive, gzip and then tar can be used. But the GNU version of tar (gtar) allows us to
use gzip “on-the-fly”, and to decompress an archive file without noticing each step (and without the need
for the extra disk space).
The use of tar follows this format:
tar <file options> <.tar.gz file> [files]

The <files> option is not required. If it is omitted, processing will be made on the whole archive. This argu-
ment does not need to be specified to extract the contents of a .tar.gz archive.
For instance:
$ tar xvfz guile-1.3.tar.gz
-rw-r--r-- 442/1002 10555 1998-10-20 07:31 guile-1.3/Makefile.in
-rw-rw-rw- 442/1002 6668 1998-10-20 06:59 guile-1.3/README
-rw-rw-rw- 442/1002 2283 1998-02-01 22:05 guile-1.3/AUTHORS
-rw-rw-rw- 442/1002 17989 1997-05-27 00:36 guile-1.3/COPYING
-rw-rw-rw- 442/1002 28545 1998-10-20 07:05 guile-1.3/ChangeLog
-rw-rw-rw- 442/1002 9364 1997-10-25 08:34 guile-1.3/INSTALL
-rw-rw-rw- 442/1002 1223 1998-10-20 06:34 guile-1.3/Makefile.am
-rw-rw-rw- 442/1002 98432 1998-10-20 07:30 guile-1.3/NEWS
-rw-rw-rw- 442/1002 1388 1998-10-20 06:19 guile-1.3/THANKS
-rw-rw-rw- 442/1002 1151 1998-08-16 21:45 guile-1.3/TODO
...

Among the options of tar:

• v makes tar verbose. This means it will display all the files it finds in the archive on the screen. If this option
is omitted, the processing will be silent.
• f is a required option. Without it, tar tries to use a tape instead of an archive file (i.e., the /dev/rmt0 device).
• z allows you to process a “gziped” archive (with a .gz extension). If this option is forgotten, tar will produce
an error. Conversely, this option must not be used with an uncompressed archive.
tar allows you to perform several actions on an archive (extract, read, create, add...). An option defines which
action is used:

• x: allows you to extract files from the archive.

• t: lists the contents of the archive.
• c: allows you to create an archive. You may use it to backup your personal files, for instance.
• r: allows you to add files at the end of the archive. It cannot be used on an already compressed archive.

2. On GNU/Linux nowadays the standard is the bzip2 format. bzip2 is more efficient on text files at the cost of more
computing power. Please refer to Bzip2, page 73, which deals specifically with this program.

73
Chapter 12. Building and Installing Free Software

12.2.3. Bzip2
A compression format named bzip2 has already replaced gzip in general use, though some software is still
distributed in gzip format, mainly for compatibility reasons with older systems. Almost all free software is
now distributed in archives which use the .tar.bz2 extension.
bzip2 is used like gzip by means of the tar command. The only change is to replace the letter z by the letter
j. For instance:
$ tar xvjf foo.tar.bz2

Some distributions may still use the option I instead:

$ tar xvfI foo.tar.bz2

Another way (which seems to be more portable, but is longer to type!):

$ tar --use-compress-program=bzip2 -xvf foo.tar.bz2

bzip2 must be installed on the system in a directory included in your PATH environment variable before you
run tar.

12.2.4. Just do it!

12.2.4.1. The easiest way

Now that you are ready to decompress the archive, do not forget to do it as administrator (root). You will
need to do things that an ordinary user is not allowed to do, and even if you can perform some of them as a
regular user, it is simpler to just be root the whole time, even if it might not be very secure.
The first step is to be in the /usr/local/src directory and copy the archive there. You should then always be
able to find the archive if you lose the installed software. If you do not have a lot of space on your disk, save
the archive on a floppy disk after having installed the software. You can also delete it but be sure that you can
find it on the web whenever you need it.
Normally, decompressing a tar archive should create a new directory (you can check that beforehand thanks
to the t option). Go then in that directory. You are now ready to proceed to the next step.

12.2.4.2. The safest way

UNIX systems (of which GNU/Linux and FreeBSD are examples) can be secure systems. This means that normal
users cannot perform operations which may endanger the system (format a disk, for instance) or alter other
users’ files. It also immunizes the system against viruses.
On the other hand, root can do everything - even run a malicious program. Having the source code available
allows you to examine it for malicious code (viruses or Trojans). It is better to be cautious in this regard3.
The idea is to create a user dedicated to administration (free or admin for example) by using the adduser
command. This user must be allowed to write in the following directories: /usr/local/src, /usr/local/
bin and /usr/local/lib, as well as all the sub-trees of /usr/share/man (he may also need to be able to copy
files elsewhere). We recommend that you make this user owner of the necessary directories or create a group
for him and make the directories writable by the group.
Once these precautions are taken, you can follow the instructions in The easiest way, page 74.

3. A proverb from the BSD world says: “Never trust a package you don’t have the sources for.”

74
 Chapter 12. Building and Installing Free Software

12.3. Configuration
For purely technical interest, the fact that authors create the sources is for the porting of the software. Free
software developed for a UNIX system may be used on all of the existing UNIX systems (whether they are free
or proprietary), with few or no changes. This requires configuration of the software just before compiling it.
Several configuration systems exist. You have to use the one the author of the software wants (sometimes,
several are needed). Usually, you can:

• Use AutoConf (see Autoconf , page 75) if a file named configure exists in the parent directory of the distribu-
tion.
• Use imake (see Imake, page 76) if a file named Imakefile exists in the parent directory of the distribution.
• Run a shell script (for instance install.sh) according to the contents of the INSTALL file (or the README file).

12.3.1. Autoconf

12.3.1.1. Principle
AutoConf is used to correctly configure software. It creates the files required by the compilation (Makefile for
instance), and sometimes directly changes the sources (for instance by using a config.h.in file).
The principle of AutoConf is simple:

• The programmer of the software knows which tests are required to configure his software (eg: “which
version of this library do you use?”). He writes them in a file named configure.in, using a precise syntax.
• He executes AutoConf, which generates a configuration script named configure from the configure.in file.
This script makes the tests required when the program is configured.
• The end-user runs the script, and AutoConf configures everything that is needed by the compilation.

12.3.1.2. Example
An example of the use of AutoConf:
$ ./configure
loading cache ./config.cache
checking for gcc... gcc
checking whether the C compiler (gcc ) works... yes
checking whether the C compiler (gcc ) is a cross-compiler... no
checking whether we are using GNU C... yes
checking whether gcc accepts -g... yes
checking for main in -lX11... yes
checking for main in -lXpm... yes
checking for main in -lguile... yes
checking for main in -lm... yes
checking for main in -lncurses... yes
checking how to run the C preprocessor... gcc -E
checking for X... libraries /usr/X11R6/lib, headers /usr/X11R6/include
checking for ANSI C header files... yes
checking for unistd.h... yes
checking for working const... yes
updating cache ./config.cache
creating ./config.status
creating lib/Makefile
creating src/Makefile
creating Makefile

To have better control of what configure generates, some options may be added by the way of the command
line or environment variables. Example:
$ ./configure --with-gcc --prefix=/opt/GNU

or (with bash):
$ export CC=‘which gcc‘

75
Chapter 12. Building and Installing Free Software

$ export CFLAGS=-O2
$ ./configure --with-gcc

or:
$ CC=gcc CFLAGS=-O2 ./configure

12.3.1.3. What if... it does not work?

Typically, it is an error that looks like configure: error: Cannot find library guile (most of the errors
of the configure script look like this).
This means that the configure script was not able to find a library (the guile library in the example). The
principle is that the configure script compiles a short test program, which uses this library. If it does not
succeed in compiling this program, it will not be able to compile the software. Then an error occurs.

• Look for the reason of the error by looking at the end of the config.log file, which contains a track of all
the steps of the configuration. The C compiler is clear enough with its error messages. This will usually help
you in solving the problem.
• Check that the said library is properly installed. If not, install it (from the sources or a compiled binary file)
and run configure again. An efficient way to check it is to search for the file that contains the symbols of
the library; which is always lib<name>.so. For instance,
$ find / -name ’libguile*’

or else:
$ locate libguile

• Check that the library is accessible by the compiler. That means it is in a directory among: /usr/lib, /lib,
/usr/X11R6/lib (or among those specified by the environment variable LD_LIBRARY_PATH, explained What
if... it does not work?, page 78 number 5.b). Check that this file is a library by typing file libguile.so.
• Check that the headers corresponding to the library are installed in the right place (usually /usr/include or
/usr/local/include or /usr/X11R6/include). If you do not know which headers you need, check that you
have installed the development version of the required library (for instance, libgtk+2.0-devel instead of
libgtk+2.0). The development version of the library provides the “include” files necessary for the compi-
lation of software using this library.
• Check that you have enough space on your disk (the configure script needs some space for temporary files).
Use the command df -h to display the partitions of your system, and note the full or nearly full partitions.
If you do not understand the error message stored in the config.log file, do not hesitate to ask for help from
the free software community (see section Technical support, page 83).
Furthermore, check whether configure answers by 100% of No or whether it answers No while you are sure
that a library exists. For instance, it would be very strange that there is no curses library on your system). In
that case, the LD_LIBRARY_PATH variable is probably wrong!

12.3.2. Imake
imake allows you to configure free software by creating a Makefile file from simple rules. These rules determi-
ne which files need to be compiled to build the binary file, and imake generates the corresponding Makefile.
These rules are specified in a file named Imakefile.
The interesting thing about imake is that it uses information which is site-dependent (architecture-dependent).
It is quite handy for applications using X Window System. But imake is also used for many other applications.
The easiest use of imake is to go into the main directory of the decompressed archive, and then to run the
xmkmf script, which calls the imake program:
$ xmkmf -a
$ imake -DUseInstalled -I/usr/X11R6/lib/X11/config
$ make Makefiles

76
 Chapter 12. Building and Installing Free Software

If the site is not correctly installed, recompile and install X11R6!

12.3.3. Various shell scripts

Read the INSTALL or README files for more information. Usually, you have to run a file called install.sh or
configure.sh. Then, either the installation script is non-interactive (and determines itself what it needs) or it
asks you information on your system (paths, for instance).
If you cannot manage to determine the file you must run, you can type ./ (under bash), and then press the
TAB key (tabulation key) twice. bash automatically (in its default configuration) completes with a possible exe-
cutable file from the directory (and therefore, a possible configuration script). If several files may be executed,
it shows you a list. You then only have to choose the right file.
A particular case is the installation of perl modules (but not only). The installation of such modules is made by
the execution of a configuration script, which is written in perl. The command to execute is usually:
$ perl Makefile.PL

12.3.4. Alternatives
Some free software distributions are badly organized, especially during the first stages of development (but
the user is warned!). They sometimes require you to change “by hand” some configuration files. Usually, these
files are a Makefile file (see section Make, page 77) and a config.h file (this name is only conventional).
We advise against these manipulations except for users who really do know what they are doing. This requires
real knowledge and some motivation to succeed, but practice makes perfect.

12.4. Compilation
Now that the software is correctly configured, all that remains is for it to be compiled. This stage is usually
easy, and does not pose serious problems.

12.4.1. Make
The favorite tool of the free software community to compile sources is make. It has two advantages:

• The developer saves time, because it allows him to efficiently manage the compilation of his project.
• The end-user can compile and install the software in a few command lines, even if he has no previous
knowledge of development.
Actions which must be executed to obtain a compiled version of the sources are stored in a file often named
Makefile or GNUMakefile. Actually, when make is called, it reads this file – if it exists – in the current directory.
If not, the file may be specified by using make’s -f option.

12.4.2. Rules
make operates in accordance with a system of dependencies, so compiling a binary file (“ target ”) requires
going through several stages (“dependencies”). For instance, to create the (imaginary) glloq binary file, the
main.o and init.o object files (intermediate files of the compilation) must be compiled and then linked. These
object files are also targets, whose dependencies are their corresponding source files.
This text is only a minimal introduction to survive in the merciless world of make. For an exhaustive docu-
mentation, refer to Managing Projects with Make, 2nd edition, O’Reilly, by Andrew Oram and Steve Talbott.

77
Chapter 12. Building and Installing Free Software

12.4.3. Go, go, go!

Usually, the use of make follows several conventions. For instance:

• make without argument just executes the compilation of the program, without installation.
• make install compiles the program (but not always), and then installs the required files at the right place
in the file system. Some files are not always correctly installed (man, info), they may have to be copied by the
user himself. Sometimes, make install has to be executed again in sub-directories. Usually, this happens
with modules developed by third parties.
• make clean clears all the temporary files created by the compilation, and also the executable file in most
cases.
The first stage is to compile the program, and therefore to type (imaginary example):
$ make
gcc -c glloq.c -o glloq.o
gcc -c init.c -o init.o
gcc -c main.c -o main.o
gcc -lgtk -lgdk -lglib -lXext -lX11 -lm glloq.o init.o main.o -o glloq

Excellent, the binary file is correctly compiled. We are ready to go to the next stage, which is the installation of
the files of the distribution (binary files, data files, etc). See Installation, page 82.

12.4.4. Explanations
If you are curious enough to look in the Makefile file, you will find known commands (rm, mv, cp, etc), but
also strange strings, looking like $(CFLAGS).
They are variables which are strings which are usually set at the beginning of the Makefile file, and then
replaced by the value they are associated with. It is quite useful when you want to use the same compilation
options several times in a row.
For instance, to print the string “foo” on the screen using make all:
TEST = foo
all:
echo $(TEST)

Most of the time, the following variables are set:

1. CC: This is the compiler. Usually, it is cc, which is in most free systems synonymous with gcc. When in
doubt, put gcc here.
2. LD: This is the program used to ensure the final compilation stage (see section The four steps of compilation,
page 72). By default, this is ld.
3. CFLAGS: These are the additional arguments that are given to the compiler during the first compilation
stages. Among them:
• -I<path>: Specifies to the compiler where to search for some additional headers (eg: -I/usr/X11R6/include
allows inclusion of the header files that are in directory /usr/X11R6/include).
• -D<symbol>: Defines an additional symbol, useful for programs whose compilation depends on the
defined symbols (ex: use the string.h file if HAVE_STRING_H is defined).
There are often compilation lines like:
$(CC) $(CFLAGS) -c foo.c -o foo.o

4. LDFLAGS (or LFLAGS): These are arguments used during the final compilation stage. Among them:
• -L<path>: Specifies an additional path to search for libraries (eg: -L/usr/X11R6/lib).
• -l<library>: Specifies an additional library to use during the final compilation stage.

78
 Chapter 12. Building and Installing Free Software

12.4.5. What if... it does not work?

Do not panic, it can happen to anyone. Among the most common causes:

1. glloq.c:16: decl.h: No such file or directory

The compiler did not manage to find the corresponding header. Yet, the software configuration step should
have anticipated this error. Here is how to solve this problem:

• Check that the header really exists on the disk in one of the following directories: /usr/include,
/usr/local/include, /usr/X11R6/include or one of their sub-directories. If not, look for it on the
whole disk (with find or locate), and if you still do not find it, check that you have installed the li-
brary corresponding to this header. You can find examples of the find and locate commands in their
respective manual pages.
• Check that the header is really readable (type less <path>/<file>.h to test this)
• If it is in a directory like /usr/local/include or /usr/X11R6/include, you sometimes have to add a
new argument to the compiler. Open the corresponding Makefile (be careful to open the right file, those
in the directory where the compilation fails 4) with your favorite text editor (Emacs, Vi, etc). Look for the
faulty line, and add the string -I<path> – where <path> is the path where the header in question can be
found – just after the call of the compiler (gcc, or sometimes $(CC)). If you do not know where to add
this option, add it at the beginning of the file, after CFLAGS=<something> or after CC=<something>.
• Launch make again, and if it still does not work, check that this option (see the previous point) is added
during compilation on the faulty line.
• If it still does not work, call for help from your local guru or call for help from the free software com-
munity to solve your problem (see section Technical support, page 83).

2. glloq.c:28: ‘struct foo’ undeclared (first use this function)

The structures are special data types that all programs use. A lot of them are defined by the system in hea-
ders. That means the problem is certainly caused by a missing or misused header. The correct procedure
for solving the problem is:

• try to check whether the structure in question is defined by the program or by the system. A solution is
to use the command grep in order to see whether the structure is defined in one of the headers.
For instance, when you are in the root of the distribution:
$ find . -name ’*.h’| xargs grep ’struct foo’ | less

Many lines may appear on the screen (each time that a function using this type of structure is defined
for instance). If it exists, pick out the line where the structure is defined by looking at the header file
obtained by the use of grep.
The definition of a structure is:
struct foo {
<contents of the structure>
};

Check if it corresponds with what you have. If so, this means that the header is not included in the
faulty .c file. There are two solutions:

• add the line #include "<filename>.h" at the beginning of the faulty .c file.
• or copy-paste the definition of the structure at the beginning of this file (it is not really neat, but at
least it usually works).

• If not, do the same thing with the system header files (which are usually in directories /usr/include,
/usr/X11R6/include, or /usr/local/include). But this time, use the line #include <<filename>.h>.

4. Analyze the error message returned by make. Normally, the last lines should contain a directory (a message like ma-
ke[1]: Leaving directory ‘/home/queen/Project/foo’). Pick out the one with the highest number. To check that it is
the good one, go to that directory and execute make again to obtain the same error.

79
Chapter 12. Building and Installing Free Software

• If this structure still does not exist, try to find out which library (i.e. set of functions put together in a
single package) it should be defined in (look in the INSTALL or README file to see which libraries are
used by the program and their required versions). If the version that the program needs is not the one
installed on your system, you will need to update this library.
• If it still does not work, check that the program properly works with your architecture (some programs
have not been ported yet on all the UNIX systems). Check also that you have correctly configured the
program (when configure ran, for instance) for your architecture.

3. parse error
This is a problem that is quite complicated to solve, because it often is an error that appears at a certain
line, but after the compiler has met it. Sometimes, it is simply a data type that is not defined. If you meet
an error message like:
main.c:1: parse error before ‘glloq_t
main.c:1: warning: data definition has no type or storage class

then the problem is that the glloq_t type is not defined. The solution to solve the problem is more or less
the same as that in the previous problem.

there may be a parse error in the old curses libraries if my

memory serves me right.

4. no space left on device

This problem is easy to solve: there is not enough space on the disk to generate a binary file from the source
file. The solution consists of making free space on the partition which contains the install directory (delete
temporary files or sources, uninstall any programs you do not use). If you decompressed it in /tmp, rather
than in /usr/local/src, which avoids needlessly saturating the /tmp partition. Check whether there are
core files5 on your disk. If so, delete them or make them get deleted if they belong to another user.

5. /usr/bin/ld: cannot open -lglloq: No such file or directory

That clearly means that the ld program (used by gcc during the last compilation stage) does not manage
to find a library. To include a library, ld searches for a file whose name is in the arguments of type -
l<library>. This file is lib<library>.so. If ld does not manage to find it, it produces an error message.
To solve the problem, follow the steps below:

a. Check that the file exists on the hard disk by using the locate command. Usually, the graphic libraries
can be found in /usr/X11R6/lib. For instance:
$ locate libglloq

If the search is unrewarding, you can make a search with the find command (eg: find /usr -name
"libglloq.so*"). If you still cannot find the library, you will have to install it.

b. Once the library is located, check that it is accessible by ld: the /etc/ld.so.conf file specifies where to
find these libraries. Add the incriminate directory at the end (you may have to reboot your computer
for this to be taken into account). You can also add this directory by changing the contents of the
environment variable LD_LIBRARY_PATH. For instance, if the directory to add is /usr/X11R6/lib, type:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/X11R6/lib

(if your shell is bash).

c. If it still does not work, check that the library format is an executable file (or ELF) with the file
command. If it is a symbolic link, check that the link is good and does not point at an non-existent

5. File generated by the system when a process tries to access a memory location that it is not allowed to. These files are
used to analyze the reason of such a behavior to correct the problem.

80
 Chapter 12. Building and Installing Free Software

file (for instance, with nm libglloq.so). The permissions may be wrong (if you use an account other
than root and if the library is protected against reading, for example).

6. glloq.c(.text+0x34): undefined reference to ‘glloq_init’

It is a problem of a symbol that was not solved during the last compilation stage. Usually, it is a library
problem. There may be several causes:

• the first thing to do is to find out whether the symbol is supposed to be in a library. For instance, if it is
a symbol beginning by gtk, it belongs to the gtk library. If the name of the library is easily identifiable
(frobnicate_foobar), you may list the symbols of a library with the nm command. For example,
$ nm libglloq.so
0000000000109df0 d glloq_message_func
000000000010a984 b glloq_msg
0000000000008a58 t glloq_nearest_pow
0000000000109dd8 d glloq_free_list
0000000000109cf8 d glloq_mem_chunk

Adding the -o option to nm allows you to print the library name on each line, which makes the search
easier. Let’s imagine that we are searching for the symbol bulgroz_max, a crude solution is to make a
search like:
$ nm /usr/lib/lib*.so | grep bulgroz_max
$ nm /usr/X11R6/lib/lib*.so | grep bulgroz_max
$ nm /usr/local/lib/lib*.so | grep bulgroz_max
/usr/local/lib/libfrobnicate.so:000000000004d848 T bulgroz_max

Wonderful! The symbol bulgroz_max is defined in the frobnicate library (the capital letter T is before
its name). Then, you only have to add the string -lfrobnicate in the compilation line by editing the
Makefile file: add it at the end of the line where LDFLAGS or LFGLAGS (or CC at worst) are defined, or on
the line corresponding to the creation of the final binary file.

• the compilation is being made with a version of the library which is not the one allowed for by the
software. Read the README or INSTALL files of the distribution to see which version must be used.
• not all the object files of the distribution are correctly linked. The file where this function is defined is
lacking. Type nm -o *.o to see which one it is and add the corresponding .o file on the compilation line
if it is missing.
• the problem function or variable may be non-existent. Try to delete it: edit the problem source file (its
name is specified at the beginning of the error message). It is a desperate solution whose consequence
will certainly be a chaotic execution of the program with a (segfault at startup, etc).

7. Segmentation fault (core dumped)

Sometimes, the compiler hangs immediately and produces this error message. I have no advise except
suggesting that you install a more recent version of your compiler.

8. no space on /tmp
Compilation needs temporary workspace during the different stages; if it does not have space, it fails. So,
you may have to clean the partition, but be careful some programs being executed (X server, pipes, etc)
can hang if some files are deleted. You must know what you are doing! If /tmp is part of a partition that
does not only contain it (for example the root), search and delete some possible core files.

9. make/configure in infinite recursion

It is often a problem of time in your system. Indeed, make needs to know the date in the computer and
the date of the files it checks. It compares the dates and uses the result to know whether the target is more
recent than the dependence.

81
Chapter 12. Building and Installing Free Software

Some date problems may cause make to endlessly build itself (or to build and build again a sub-tree in
infinite recursion). In such a case, the use of touch (whose use here is to set the files in question to the
current time) usually solves the problem.
For instance:
$ touch *

Or also (cruder, but efficient):

$ find . | xargs touch

12.5. Installation

12.5.1. With Make

Now that all is compiled, you have to copy the built files to an appropriate place (usually in one of the sub-
directories of /usr/local).
make can usually perform this task. A special target is the target install. So, using make install carries out
the installation of the required files.
Usually, the procedure is described in the INSTALL or README file. But sometimes, the developer has forgotten
to provide one. In that case, you must install everything yourself.
Copy then:

• The executable files (programs) into the /usr/local/bin directory.

• The libraries (lib*.so files) into the /usr/local/lib directory.
• The headers (*.h files) into the /usr/local/include directory (be careful not to delete the originals).
• The data files usually go in /usr/local/share. If you do not know the installation procedure, you can try
to run the programs without copying the data files, and to put them at the right place when it asks you for
them (in an error message such as Cannot open /usr/local/share/glloq/data.db for example).
• The documentation is a little bit different:
• The man files are usually put in one of the sub-directories of /usr/local/man. Usually, these files are
in troff (or groff) format, and their extension is a figure. Their name is the name of a command (for
instance, echo.1). If the figure is n, copy the file in /usr/local/man/man<n>.
• The info files are put in the directory /usr/info or /usr/local/info

You are finished! Congratulations! You now are ready to compile an entire operating system!

12.5.2. Problems
If you have just installed free software, GNU tar for instance, and if, when you execute it, another program is
started or it does not work like it did when you tested it directly from the src directory, it is a PATH problem,
which finds the programs in a directory before the one where you have installed the new software. Check by
executing type -a <program>.
The solution is to put the installation directory higher in the PATH and/or to delete or rename the files that
were executed when they were not asked for, and/or rename your new programs (into gtar in this example)
so that there is no more confusion.
You can also make an alias if the shell allows it (for instance, say that tar means /usr/local/bin/gtar).

82
 Chapter 12. Building and Installing Free Software

12.6. Support

12.6.1. Documentation
Several documentation sources:

• HOWTOs, short documents on precise points (usually far from what we need here, but sometimes useful).
Look on your disk in /usr/share/doc/HOWTO (not always, they are sometimes elsewhere; check that out
with the command locate HOWTO),
• The manual pages. Type man <command> to get documentation on the command <command>,
• Specialized literature. Several large publishers have begun publishing books about free systems (especially
on GNU/Linux). It is often useful if you are a beginner and if you do not understand all the terms of the
present documentation.

12.6.2. Technical support

If you have bought an “official” Mandrakelinux distribution, you can ask the technical support staff for infor-
mation on your system.
You can also rely on help from the free software community:

• newsgroups (on Usenet) comp.os.linux.* (news:comp.os.linux.*) answer all the questions about GNU/Linux.
Newsgroups matching comp.os.bsd.* deal with BSD systems. There may be other newsgroups dealing
with other UNIX systems. Remember to read them for some time prior to writing to them.
• Several associations or groups of enthusiasts in the free software community offer voluntary support. The
best way to find the ones closest to you, is to check out the lists on specialized web sites, or to read the
relevant newsgroups for a while.
• Several IRC channels offer a real time (but blind) assistance by gurus. See for instance the #linux channel
on most of the IRC network, or #linuxhelp on IRCNET.
• As a last resort, ask the developer of the software (if he mentioned his name and his email address in a file
of the distribution) if you are sure that you have found a bug (which may be due only to your architecture,
but after all, free software is supposed to be portable).

12.6.3. How to find free software

To find free software, a lot of links may help you:

• the huge FTP site sunsite.unc.edu (sunsite.unc.edu) or one of its mirrors

• the following web sites make a catalog of many free software programs which can be used on UNIX plat-
forms (but one can also find proprietary software on these):
• FreshMeat (http://www.freshmeat.net/) is probably the most complete site,
• SourceForge.net (http://sourceforge.net/) is the world’s largest Open Source software development
web site, with the largest repository of Open Source code and applications available on the Internet.
• GNU Software (http://www.gnu.org/software/) for an exhaustive list of all of GNU software. Of course,
all of them are free and most are licensed under the GPL,

• you can also perform a search with a search engine such as Google/ (http://www.google.com/) and Lycos/
(http://www.lycos.com/) and make a request like: +<software>+download or "download software".

83
Chapter 12. Building and Installing Free Software

12.7. Acknowledgments

• Proof-reading and disagreeable comments (in alphabetical order): Sébastien Blondeel, Mathieu Bois, Xavier
Renaut and Kamel Sehil.
• Beta-testing: Laurent Bassaler
• English translation: Fanny Drieu; English editing: Hoyt Duff

84
Chapter 13. Compiling and Installing New Kernels
Along with file system mounting and building from sources, compiling the kernel is undoubtedly the subject
which causes the most problems for beginners. Compiling a new kernel is not generally necessary, since the
kernel installed by Mandrakelinux contains support for a significant number of devices (in fact, more devices
than you will ever need or even think of), along with a set of trusted patches and so on. But...
It may be that you want to do it, for no other reason than to see “what it does”. Apart from making your PC
and your coffee machine work a bit harder than usual, not a lot. The reasons why you should want to compile
your own kernel range from deactivating an option to rebuilding a brand new experimental kernel. Anyway,
the aim of this chapter is to ensure that your coffee machine still works after compilation.
There are other valid reasons for recompiling the kernel. For example, you have read that the kernel you are
using has a security bug, which has been fixed in a more recent version, or that a new kernel includes support
for a device you need. Of course, in these cases, you do have the choice of waiting for binary upgrades, but
updating the kernel sources and recompiling the new kernel yourself makes for a faster solution.
Whatever you do, stock up with coffee.

13.1. Upgrading a Kernel Using Binary Packages

Before diving deep into kernel compilation from sources, we will detail a simple procedure when you just
want/need to update your kernel using binary RPM packages compiled for your version of Mandrakelinux.
The example will assume the new kernel is kernel-2.6.3-5mdk and the old (current) one is kernel-2.6.3-
1mdk.

1. Install the New Kernel. Issue the command: urpmi kernel-2.6.3-5mdk in a terminal window.
2. Verify it Works. A new entry will be available in the LILO menu (named something like 263-5). Reboot
your computer and select that entry to boot with the new kernel. Perform all the tests you consider neces-
sary to make sure the new kernel works correctly.
3. Uninstall the Old Kernel. Once you are sure the new kernel works in your computer, you can remove
the files related to the old kernel. Issue urpme kernel-2.6.3-1mdk in a terminal window and edit /etc/
lilo.conf removing the 263-5 section and issue lilo -v to make the new kernel the default one.

13.2. From The Kernel Sources

You can basically get the sources from two places:

1. Official Mandrake Linux Kernel. In the SRPMS directory of any of the Cooker mirrors (http://www.
mandrakelinux.com/en/cookerdevel.php3), you will find the following packages:

kernel-2.6.??.?mdk-?-?mdk.src.rpm
The kernel sources for compiling the kernel used in the distribution. It is highly modified for additio-
nal functions.

kernel2.6-linus-2.6.??-?mdk.src.rpm
The stock kernel as published by the maintainer of the GNU/Linux kernel.
If you choose this option (recommended), just download the source RPM, install it (as root) and jump to
Configuring The Kernel, page 86.
2. The Official Linux Kernel Repository. The main kernel source host site is ftp.kernel.org (ftp://ftp.
kernel.org), but there are a large number of mirrors, all named ftp.xx.kernel.org, where xx represents
the ISO country code. Following the official announcement of the availability of the kernel, you should
allow at least two hours for all the mirrors to be updated.
On all of these FTP servers, the kernel sources are in the /pub/linux/kernel directory. Next, go to the
directory with the series that interests you: it will undoubtedly be v2.6. Nothing prevents you from trying
the experimental versions or using the old 2.0/2.2/2.4 versions. The file containing the kernel sources is
called linux-<kernel_version>.tar.bz2, e.g. linux-2.6.3.tar.bz2.

85
Chapter 13. Compiling and Installing New Kernels

You can also apply patches to kernel sources in order to upgrade them incrementally: thus, if you already
have kernel sources version 2.6.1 and want to upgrade to kernel 2.6.3, you do not need to download the
whole 2.6.3 source, you can simply download the patches patch-2.6.2.bz2 and patch-2.6.3.bz2. As a
general rule, this is a good idea, since sources currently take up dozens of MB.

13.3. Unpacking Sources, Patching the Kernel (if Necessary)

Kernel sources should be placed in /usr/src. So you should go into this directory then unpack the sources
there:
$ cd /usr/src
$ mv linux linux.old
$ tar xjf /path/to/linux-2.6.1.tar.bz2

The command mv linux linux.old is required: this is because you may already have sources of another
version of the kernel. This command will ensure that you do not overwrite them. Once the archive is unpacked,
you have a linux-<version> directory (where <version> is the version of the kernel) with the new kernel’s
sources. You can make a link (ln -s linux-<version> linux) for convenience’s sake.
Now, the patches. We will assume that you do want to patch from version 2.6.1 to 2.6.3 and have downloa-
ded the patches needed to do this: go to the newly created linux directory, then apply the patches:
$ cd linux
$ bzcat /path/to/patch-2.6.2.bz2 | patch -p1
$ bzcat /path/to/patch-2.6.3.bz2 | patch -p1
$ cd ..

Generally speaking, moving from a version 2.6.x to a version 2.6.y requires you to apply all the patches
numbered 2.6.x+1, 2.6.x+2, ..., 2.6.y in this order. To revert from 2.6.y to 2.6.x, repeat exactly the same
procedure but applying the patches in reverse order and with option -R from patch (R stands for Reverse). So,
to go back from kernel 2.6.3 to kernel 2.6.1, you would do:
$ bzcat /path/to/patch-2.6.3.bz2 | patch -p1 -R
$ bzcat /path/to/patch-2.6.2.bz2 | patch -p1 -R

If you wish to test if a patch will correctly apply before actually

applying it, add the --dry-run option to the patch command.

Next, for the sake of clarity (and so you know where you are), you can rename linux to reflect the kernel
version and create a symbolic link:
$ mv linux linux-2.6.3
$ ln -s linux-2.6.3 linux

It is now time to move on to configuration. For this, you have to be in the source directory:
$ cd linux

86
 Chapter 13. Compiling and Installing New Kernels

13.4. Configuring The Kernel

To start, go into the /usr/src/linux directory.
First, a little trick: you can, if you want, customize the version of your kernel. The kernel version is determined
by the four first lines of the Makefile:
$ head -4 Makefile
VERSION = 2
PATCHLEVEL = 6
SUBLEVEL = 3
EXTRAVERSION = -1mdkcustom

Further on in the Makefile, you can see that the kernel version is built as:
KERNELRELEASE=$(VERSION).$(PATCHLEVEL).$(SUBLEVEL)$(EXTRAVERSION)

All you have to do is modify one of these fields in order to change your version. Preferably, you will only
change EXTRAVERSION. Say you set it to -foo, for example. Your new kernel version will then become 2.6.3-
foo. Do not hesitate to change this field each time you recompile a new kernel with different versions, so that
you can test different options while keeping the previous versions.
Now, on to configuration. You can choose between:

• make xconfig for a graphical interface;

• make menuconfig for an interface based on ncurses;
• make config for the most rudimentary interface, line by line, section by section;
• make oldconfig the same as above, but based on your former configuration. See Saving, Reusing your Kernel
Configuration Files, page 87.
You will go through the configuration section by section, but you can skip sections and jump to the ones that
interest you if you are using menuconfig or xconfig. The options are y for Yes (functionality hard-compiled
into the kernel), m for Module (functionality compiled as a module), or n for No (do not include it in the kernel).
Both make xconfig and make menuconfig have the options bundled in hierarchical groups. For example, Pro-
cessor family goes under Processor type and features.
For xconfig, the button Main Menu is used to return to the main menu when in a hierarchical group; Next goes
to the next group of options; and Prev returns to the previous group. For menuconfig, use the Enter key to
select a section, and switch options with y, m or n to change the options status, or else, press the Enter key and
make your choice for the multiple choice options. Exit will take you out of a section or out of configuration if
you are in the main menu. And there is also Help.
We are not going to enumerate all options here, as there are several hundreds of them. Furthermore, if you
have reached this chapter, you probably know what you are doing anyway. So you are left to browse through
the kernel configuration and set/unset whichever options you see fit. However, here is some advice to avoid
ending up with an unusable kernel:

1. unless you use an initial ramdisk, never compile the drivers necessary to mount your root file system
(hardware drivers and file-system drivers) as modules! And if you use an initial ramdisk, say Y to ext2FS
support, as this is the file system used for ramdisks. You will also need the initrd support;
2. if you have network cards on your system, compile their drivers as modules. Hence, you can define which
card will be the first one, which will be the second, and so on, by putting appropriate aliases in /etc/
modules.conf. If you compile the drivers into the kernel, the order in which they will be loaded will
depend on the linking order, which may not be the order you want;
3. and finally: if you don’t know what an option is about, read the help! If the help text still doesn’t inspire
you, just leave the option as it was. (for config and oldconfig targets, press the ? key to access the help).
Et voilà! Configuration is finally over. Save your configuration and quit.

87
Chapter 13. Compiling and Installing New Kernels

13.5. Saving, Reusing your Kernel Configuration Files

The kernel configuration is saved in the /usr/src/linux/.config file. There’s a backup for it in /boot/
config-<version>, it is a good to keep it as a reference. But also save your own configurations for different
kernels, as this is just a matter of giving different names to configuration files.
One possibility is to name configuration files after the kernel version. Say you modified your kernel version
as shown in Configuring The Kernel, page 86, then you can do:
$ cp .config /root/config-2.6.3-foo

If you decide to upgrade to 2.6.4 (for example), you will be able to reuse this file, as the differences between
the configuration of these two kernels will be very small. Just use the backup copy:
$ cp /root/config-2.6.3-foo .config

But copying the back file does not mean that the kernel is ready to be compiled just yet. You have to invo-
ke make menuconfig (or whatever else you chose to use) again, because some files needed in order for the
compilation to succeed are created and/or modified by these commands.
However, apart from the chore of going through all the menus again, you could possibly miss some interesting
new option(s). You can avoid this by using make oldconfig. It has two advantages:

1. it is fast;
2. if a new option appears in the kernel and was not present in your configuration file, it will stop and wait
for you to enter your choice.

After you have copied your .config to the root home, as suggested
above, run make mrproper. It will ensure that nothing remains from
the old configuration and you will get a clean kernel.

Next, time for compilation.

13.6. Compiling Kernel and Modules, Installing the Beast

A small point to begin with: if you are recompiling a kernel with exactly the same version as the one already
present on your system, the old modules must be deleted first. For example, if you are recompiling 2.6.3, you
must delete the /lib/modules/2.6.3 directory.
Compiling the kernel and modules, and then installing modules, is done with the following lines:
make clean bzImage modules
make modules_install install

A little vocabulary: bzImage, modules, etc., as well as oldconfig and others which we used above, are called
targets. If you specify several targets to make as shown above, they will be executed in the order of appearance.
But if one target fails, make will not go any further1.
Let us look at the different targets and see what they do:

• bzImage: this builds the kernel. Note that this target is only valid for Intel processors. This target also gene-
rates the System.map for this kernel. We will see later what this file is used for;
• modules: this target will generate modules for the kernel you have just built. If you have chosen not to have
modules, this target will do nothing;
• modules_install: this will install modules. By default, modules will be installed in the /lib/modules/
<kernel-version> directory. This target also computes module dependencies (unlike in older kernels);
• install: this last target will finally copy the kernel and modules to the right places and modify the boot
loader’s configurations in order for the new kernel to be available at boot time. Do not use it if you prefer to
perform a manual installation as described in Installing the New Kernel Manually, page 89.

1. In this case, if the compilation fails, it means that there is a bug in the kernel... If this is the case, please report it!

88
 Chapter 13. Compiling and Installing New Kernels

It is important to respect the target order modules_install ins-

tall so that modules actually get installed first.

At this point, everything is now compiled and correctly installed, ready to be tested! Just reboot your machine
and choose the new kernel in the boot menu. Note that the old kernel remains available so that you can use
it if you experience problems with the new one. However, you can choose to manually install the kernel and
change the boot menus by hand. We will explain that in the next section.

The old zImage target is now obsolete, it is deprecated, and you

should not use it.

13.7. Installing the New Kernel Manually

The kernel is located in arch/i386/boot/bzImage . The standard directory in which kernels are installed is
/boot. You also need to copy the System.map file to ensure that some programs (top is just one example) will
work correctly. Good practice again: name these files after the kernel version. Let us assume that your kernel
version is 2.6.3-foo. The sequence of commands you will have to type is:
$ cp arch/i386/boot/bzImage /boot/vmlinuz-2.6.3-foo
$ cp System.map /boot/System.map-2.6.3-foo

Now you need to tell the boot loader about your new kernel. There are two bootloaders: GRUB or LILO. Note
that Mandrakelinux is configured with LILO by default.

13.7.1. Updating LILO

The simplest way of updating LILO is to use drakboot (see chapter Change Your Boot-up Configuration in the
Starter Guide). Alternatively, you can manually edit the configuration file as follows.
The LILO configuration file is /etc/lilo.conf. This is what a typical lilo.conf looks like:
boot=/dev/hda
map=/boot/map
default="linux"
keytable=/boot/es-latin1.klt
prompt
nowarn
timeout=50
message=/boot/message
menu-scheme=wb:bw:wb:bw
image=/boot/vmlinuz
label="linux"
root=/dev/hda1
initrd=/boot/initrd.img
append="devfs=mount acpi=ht resume=/dev/hda5"
read-only
other=/dev/fd0
label="floppy"
unsafe

A lilo.conf file consists of a main section, followed by a section for each operating system. In the example of
the file above, the main section is made up of the following directives:
boot=/dev/hda
map=/boot/map
default="linux"
keytable=/boot/es-latin1.klt
prompt
nowarn
timeout=50
message=/boot/message
menu-scheme=wb:bw:wb:bw

89
Chapter 13. Compiling and Installing New Kernels

The boot= directive tells LILO where to install its boot sector; in this case, it is the MBR (Master Boot Record)
of the first IDE hard disk. If you want to make a LILO floppy disk, simply replace /dev/hda with /dev/fd0.
The prompt directive asks LILO to show the menu on start-up. As a timeout is set, LILO will start the default
image after 5 seconds (timeout=50). If you remove the timeout directive , LILO will wait until you have typed
something.
Then comes a linux section:
image=/boot/vmlinuz
label="linux"
root=/dev/hda1
initrd=/boot/initrd.img
append="devfs=mount acpi=ht resume=/dev/hda5"
read-only

A section to boot a GNU/Linux kernel always starts with an image= directive, followed by the full path to a
valid GNU/Linux kernel. Like any section, it contains a label= directive as a unique identifier, here linux. The
root= directive tells LILO which partition hosts the root file system for this kernel. It may be different in your
configuration... The read-only directive tells LILO that it should mount the root file system as read-only on
start-up: if this directive is not there, you will get a warning message. The append line specifies options to pass
to the kernel during booting.
Then comes the floppy section:
other=/dev/fd0
label="floppy"
unsafe

In fact, a section beginning with other= is used by LILO to start any operating system other than GNU/Linux:
the argument of this directive is the location of this system’s boot sector, and in this case, it is to boot from a
floppy disk.
Now, it’s time to add a section for our new kernel. You can put this section anywhere after the main section,
but do not enclose it within another section. Here is what it should look like:
image=/boot/vmlinuz-2.6.3-foo
label="foo"
root=/dev/hda1
read-only
append="devfs=mount acpi=ht resume=/dev/hda5"

Needless to say: adapt the entry to your system’s configuration.

So this is what our lilo.conf looks like after modification, decorated with a few additional comments (all the
lines beginning with #), which will be ignored by LILO:
#
# Main section
#
boot=/dev/hda
map=/boot/map
install=/boot/boot.b
message=/boot/message
# What should be booted by default. Let’s put our own new kernel:
default="foo"
# Show prompt...
prompt
# ... wait 5 seconds
timeout=50
#
# Our new kernel: default image
#
image=/boot/vmlinuz-2.6.3-foo
label="foo"
root=/dev/hda1
read-only
append="devfs=mount acpi=ht resume=/dev/hda5"
#
# The original kernel
#
image=/boot/vmlinuz
label="linux"

90
 Chapter 13. Compiling and Installing New Kernels

root=/dev/hda1
read-only
append="devfs=mount acpi=ht resume=/dev/hda5"
#
# Floppy Section
#
other=/dev/floppy
label="floppy"
unsafe

This could well be what your lilo.conf will look like... but remember, again, adapt it to your own configura-
tion.
Now that the file has been modified appropriately, but unlike GRUB which does not need it, you must tell
LILO to change the boot sector:

$ lilo
Added foo *
Added linux
Added floppy
$

In this way, you can compile as many kernels as you want, by adding as many sections as necessary. All you
need to do now is restart your machine to test your new kernel. Notice that if LILO shows any errors during
installation, it means that it has changed nothing. LILO only modifies your configuration if it finds no errors
in the process.

13.7.2. Updating Grub

Obviously, retain the option of starting your current kernel! The simplest way of updating GRUB is to use drak-
boot (see chapter Change Your Boot-up Configuration in the Starter Guide). Alternatively, you can manually
edit the configuration file as follows.
You need to edit the /boot/grub/menu.lst file. This is what a typical menu.lst looks like, after you have
installed your Mandrakelinux distribution and before modification:
timeout 5
color black/cyan yellow/cyan
i18n (hd0,4)/boot/grub/messages
keytable (hd0,4)/boot/fr-latin1.klt
default 0

title linux
kernel (hd0,4)/boot/vmlinuz root=/dev/hda5

title failsafe
kernel (hd0,4)/boot/vmlinuz root=/dev/hda5 failsafe

title Windows
root (hd0,0)
makeactive
chainloader +1

title floppy
root (fd0)
chainloader +1

This file is made of two parts: the header with common options (the five first lines), and the images, each
one corresponding to a different GNU/Linux kernel or another OS. timeout 5 defines the time (in seconds) for
which GRUB will wait for input before it loads the default image (this is defined by the default 0 directive
in common options, i.e. the first image in this case). The keytable directive, if present, defines where to find
the keymap for your keyboard. In this example, this is a French layout. If none are present, the keyboard is
assumed to be a plain QWERTY keyboard. All of the hd(x,y) which you see refer to partition number y on disk
number x as seen by the BIOS.
Then come the different images. In this example, four images are defined: linux, failsafe, windows, and
floppy.

91
Chapter 13. Compiling and Installing New Kernels

• The linux section starts by telling GRUB about the kernel which is to be loaded (kernel hd(0,4)/boot/vmlinuz),
followed by the options to pass to the kernel. In this case, root=/dev/hda5 will tell the kernel that the root
file system is located on /dev/hda5. In fact, /dev/hda5 is the equivalent of GRUB’s hd(0,4), but nothing
prevents the kernel from being on a different partition to the one containing the root file system;
• The failsafe section is very similar to the previous one, except that we will pass an argument to the kernel
(failsafe) which tells it to enter “single” or “rescue” mode;
• The Windows section tells GRUB to simply load the first partition’s boot sector, which probably contains a
Windows boot sector;
• The floppy section simply boots your system from the floppy disk in the first floppy drive, whatever the
system installed on it. It can be a Windows boot disk, or even a GNU/Linux kernel on a floppy;

Depending on the security level you use on your system, some of

the entries described here may be absent from your file.

Now to the point. We need to add another section to tell GRUB about our new kernel. In this example, it will
be placed before the other entries, but nothing prevents you from putting it somewhere else:
title foo
kernel (hd0,4)/boot/vmlinuz-2.6.3-foo root=/dev/hda5

Do not forget to adapt the file to your configuration! The GNU/Linux root file system here is /dev/hda5, but it
can be somewhere else on your system.
And that’s it. Unlike LILO, as we will see below, there is nothing else to do. Just restart your computer and
your newly defined entry will just appear. Just select it from the menu and your new kernel will boot.
If you compiled your kernel with the framebuffer, you will probably want to use it: in this case, you need to
add a directive to the kernel which tells it what resolution you want to start in. The list of modes is available
in the /usr/src/linux/Documentation/fb/vesafb.txt file (but only in the case of the VESA framebuffer!
Otherwise, refer to the corresponding file). For the 800x600 mode in 32 bits2, the mode number is 0x315, so
you need to add the directive:
vga=0x315

and your entry now resembles:

title foo
kernel (hd0,4)/boot/vmlinuz-2.6.3-foo root=/dev/hda5 vga=0x315

For more information, please refer to the info pages about GRUB (info grub).

2. 8 bits means 28 colors, i.e. 256; 16 bits means 216 colors, i.e. 64k, i.e. 65536; in 24 bits as in 32 bits, color is coded on 24
bits, i.e. 224 possible colors, in other words exactly 16M, or a bit more than 16 million.

92
Appendix A. Glossary
account
on a UNIX system, the combination of a name, a personal directory, a password and a shell which allows
a user to connect to this system.

alias
a mechanism used in a shell in order to make it substitute one string for another before executing the
command. You can see all aliases defined in the current session by typing alias at the prompt.

APM
Advanced Power Management. A feature used by some BIOSes in order to make the machine enter a
standby state after a given period of inactivity. On laptops, APM is also responsible for reporting the
battery status and (if supported) the estimated remaining battery life.

ARP
Address Resolution Protocol. The Internet protocol used to dynamically map an Internet address to a phy-
sical (hardware) address on a local area network. This is limited to networks that support hardware
broadcasting.

ASCII
American Standard Code for Information Interchange. The standard code used for storing characters, inclu-
ding control characters, on a computer. Many 8-bit codes (such as ISO 8859-1, the Linux default character
set) contain ASCII as their lower half.
See Also: ISO 8859.
assembly language
is the programming language that is closest to the computer, which is why it’s called a “low level” pro-
gramming language. Assembly has the advantage of speed since assembly programs are written in terms
of processor instructions so little or no translation is needed when generating executables. Its main di-
sadvantage is that it is processor (or architecture) dependent. Writing complex programs is very time-
consuming as well. So, assembly is the fastest programming language, but it isn’t portable between
architectures.
ATAPI
(“AT Attachment Packet Interface”) An extension to the ATA specification (“Advanced Technology At-
tachment”, more commonly known as IDE, Integrated Drive Electronics) which provides additional com-
mands to control CD-ROM drives and magnetic tape drives. IDE controllers equipped with this extension
are also referred to as EIDE (Enhanced IDE) controllers.

ATM
This is an acronym for Asynchronous Transfer Mode. An ATM network packages data into standard
size blocks (53 bytes: 48 for the data and 5 for the header) that it can be conveyed efficiently from point to
point. ATM is a circuit switched packet network technology oriented towards high speed (multi-megabit)
optical networks.

atomic
a set of operations is said to be atomic when they execute all at once and cannot be preempted.

background
in shell context, a process is running in the background if you can type commands that are captured by
the process while it is running.
See Also: job, foreground.

backup
is a means of saving your important data to a safe medium and location. Backups should be done regu-
larly, especially with more critical information and configuration files (the most important directories to
backup are /etc, /home and /usr/local). Traditionally, many people use tar with gzip or bzip2 to bac-
kup directories and files. You can use these tools or programs like dump and restore, along with many
other free or commercial backup solutions.

batch
is a processing mode where jobs which are submitted to the CPU are executed sequentially until all the
jobs have been processed.

93
Appendix A. Glossary

beep
is the little noise your computer’s speaker emits to warn you of some ambiguous situation when you’re
using command completion and, for example, there’s more than one possible choice for completion.
There might be other programs that make beeps to let you know of some particular situation.

beta testing
is the name given to the process of testing the beta version of a program. Programs usually get released
in alpha and beta states for testing prior to final release.

binary
in the context of programming, binaries are the compiled, executable code.

bit
stands for BInary digiT. A single digit which can take the values 0 or 1, because calculation is done in base
two.
block mode files
files whose contents are buffered. All read/write operations for such files go through buffers, which
allow for asynchronous writes on the underlying hardware, and for reads, which allows the system to
avoid disk access if the data is already in a buffer.
See Also: buffer, buffer cache, character mode files.

boot
the procedure taking place when a computer is switched on, where peripherals are recognized sequen-
tially and where the operating system is loaded into memory.

bootdisk
a bootable floppy disk containing the code necessary to load the operating system from the hard disk
(sometimes it is self-sufficient).

bootloader
is a program which starts the operating system. Many bootloaders give you the opportunity to load more
than one operating system by allowing you choose between them from a menu. Bootloaders like GRUB
and LILO are popular because of this feature and are very useful in dual- or multi-boot systems.

BSD
Berkeley Software Distribution. A UNIX variant developed at the Berkeley University computing depart-
ment. This version has always been considered more technically advanced than the others, and has
brought many innovations to the computing world in general and to UNIX in particular.

buffer
a small portion of memory with a fixed size, which can be associated with a block mode file, a system
table, a process and so on. The buffer cache maintains coherency of all buffers.
See Also: buffer cache.
buffer cache
a crucial part of an operating system kernel, it is in charge of keeping all buffers up-to-date, shrinking
the cache when needed, clearing unneeded buffers and more.
See Also: buffer.
bug
illogical or incoherent behavior of a program in a special case, or a behavior that does not follow the
documentation or accepted standards issued for the program. Often, new features introduce new bugs
in a program. Historically, this term comes from the old days of punch cards: a bug (the insect!) slipped
into a hole of a punch card and, as a consequence, the program misbehaved. Admiral Grace Hopper,
having discovered this, declared “It’s a bug!”, and since then the term has remained. Note that this is
only one of the many stories which attempt to explain the term bug.

byte
eight consecutive bits, which when interpreted in base ten result in a number between 0 and 255.
See Also: bit.
case
when taken in the context of strings, the case is the difference between lowercase letters and uppercase
(or capital) letters.

94
 Appendix A. Glossary

CHAP
Challenge-Handshake Authentication Protocol: protocol used by ISPs to authenticate their clients. In this
scheme, a value is sent to the client (the machine making the connection), which it uses to calculate a hash
based on the value. The client sends the hash back to the server for comparison to the hash calculated by
the server. This authentication method is different to PAP in that it re- authenticates on a periodic basis
after the initial authentication.
See Also: PAP.
character mode files
files whose content is not buffered. When associated with physical devices, all input/output on these
devices is performed immediately. Some special character devices are created by the operating system
(/dev/zero, /dev/null and others). They correspond to data flows.
See Also: block mode files.
CIFS
Common Internet File System The successor to the SMB file system, used on DOS systems.

client
program or computer which periodically connects to another program or computer to give it orders or
ask for information. In the case of peer to peer systems such as SLIP or PPP the client is taken to be the
end that initiates the connection and the remote end receiving the call is designated as the server. It is
one of the components of a client/server system.

client/server system
system or protocol consisting of a server and one or several clients.

command line
provided by a shell and which allows the user to type commands directly. Also subject of an eternal
“flame war” between its supporters and its detractors.

command mode
under Vi or one of its clones, it is the state of the program in which pressing a key will not insert the
character in the file being edited, but instead performs an action specific to the key (unless the clone has
re-mappable commands and you have customized your configuration). You may get out of it typing one
of the “back to insertion mode” commands: i, I, a, A, s, S, o, O, c, C, ...

compilation
is the process of translating source code that is human readable (well, with some training) and that is
written in some programming language (C, for example) into a binary file that is machine readable.

completion
the ability of a shell to automatically expand a substring to a filename, user name or other item, as long
as there is a match.
compression
is a way to shrink files or decrease the number of characters sent over a communications connection.
Some file compression programs include compress, zip, gzip, and bzip2.

console
is the name given to what used to be called terminals. They were the machines (a screen plus a keyboard)
connected to one big central mainframe. On PC s, the physical terminal is the keyboard and screen.
See Also: virtual console.
cookies
temporary files written on the local hard disk by a remote web server. It allows for the server to be aware
of a user’s preferences when this user connects again.

datagram
A datagram is a discrete package of data and headers which contain addresses, which is the basic unit of
transmission across an IP network. You might also hear this called a “packet”.

dependencies
are the stages of compilation that need to be satisfied before going on to other compilation stages in order
to successfully compile a program. This term is also used where one set of programs you wish to install
are dependent on other programs which may or may not be installed on your system, in which case you

95
Appendix A. Glossary

may get a message telling you that the system needs to “satisfy dependencies” in order to continue the
installation.
desktop
If you’re using the X Window System, the desktop is the place on the screen where you work and upon
which your windows and icons are displayed. It is also called the background, and is usually filled with
a simple color, a gradient color or even an image.
See Also: virtual desktops.

DHCP
Dynamic Host Configuration Protocol. A protocol designed for machines on a local network to dynamically
get an IP address from a DHCP server.

directory
Part of the file system structure. Files or other directories can be stored within a directory. Sometimes
there are sub-directories (or branches) within a directory. This is often referred to as a directory tree. If
you want to see what’s inside another directory, you will either have to list it or change to it. Files inside a
directory are referred to as leaves while sub-directories are referred to as branches. Directories follow the
same restrictions as files although the permissions mean different things. The special directories . and
.. refer to the directory itself and to the parent directory respectively.

discrete values
are values that are non-continuous. That is, there’s some kind of “spacing” between two consecutive
values.
distribution
is a term used to distinguish one GNU/Linux manufacturers product from another. A distribution is made
up of the core Linux kernel and utilities, as well as installation programs, third-party programs, and
sometimes proprietary software.

DLCI
The DLCI is the Data Link Connection Identifier and is used to identify a unique virtual point to point
connection via a Frame Relay network. The DLCI’s are normally assigned by the Frame Relay network
provider.

DMA
Direct Memory Access. A facility used in the PC architecture which allows a peripheral to read or write
from main memory without the help of the CPU. PCI peripherals use bus mastering and do not need
DMA.
DNS
Domain Name System. The distributed name and address mechanism used in the Internet. This mechanism
allows you to map a domain name to an IP address, allowing you to look up a site by domain name
without knowing the IP address of the site. DNS also allows reverse lookup, allowing you to obtain
machine’s IP address from its name.
DPMS
Display Power Management System. Protocol used by all modern monitors to manage power saving featu-
res. Monitors supporting these features are commonly called “green” monitors.

echo
occurs when the characters you type in a user name entry field, for example, are shown “as is”, instead
of showing “*” for each one you type.

editor
is a term typically used for programs that edit text files (aka text editor). The most well-known GNU/Linux
editors are the GNU Emacs (Emacs) editor and the UNIX editor Vi.

ELF
Executable and Linking Format. This is the binary format used by most GNU/Linux distributions.

email
stands for Electronic Mail. This is a way to send messages electronically between people on the same
network. Similar to regular mail (aka snail mail), email needs a destination and sender address to be

96
 Appendix A. Glossary

sent properly. The sender must have an address like “[email protected]” and the recipient must
have an address like “[email protected].” Email is a very fast method of communication and
typically only takes a few minutes to reach anyone, regardless of where in the world they are located. In
order to write email, you need an email client such as pine or mutt which are text-mode clients, or GUI
clients such as KMail.
environment
is the execution context of a process. It includes all the information that the operating system needs to
manage the process and what the processor needs to execute the process properly.
See Also: process.

environment variables
a part of a process’ environment. Environment variables are directly viewable from the shell.
See Also: process.
escape
in the shell context, is the action of surrounding a string between quotes to prevent the shell from inter-
preting that string. For example, when you need to use spaces in a command line and pipe the results
to some other command you have to put the first command between quotes (“escape” the command)
otherwise the shell will interpret it incorrectly and it won’t work as expected.

ext2
short for the “Extended 2 file system”. This is GNU/Linux’s native file system and has the characteristics
of any UNIX file system: support for special files (character devices, symbolic links, etc), file permissions
and ownership, and other features.

FAQ
Frequently Asked Questions. A document containing a series of questions and answers about a specific
topic. Historically, FAQs appeared in newsgroups, but this sort of document now appears on various web
sites, and even commercial products have FAQs. Generally, they are very good sources of information.

FAT
File Allocation Table. File system used by DOS and Windows.

FDDI
Fiber Distributed Digital Interface. A high-speed network physical layer, which uses optical fiber for com-
munication. Mostly used on large networks, mainly because of its price. It is rarely seen as a means of
connection between a PC and a network switch.
FHS
File system Hierarchy Standard. A document containing guidelines for a coherent file tree organization on
UNIX systems. Mandrakelinux complies with this standard in most aspects.

FIFO
First In, First Out. A data structure or hardware buffer where items are taken out in the order they were
put in. UNIX pipes are the most common examples of FIFO s.

filesystem
scheme used to store files on a physical media (hard drive, floppy) in a consistent manner. Examples of
file systems are FAT, GNU/Linux’ ext2fs, ISO9660 (used by CD-ROMs) and so on. An example of a virtual
filesystem is the /proc filesystem.

firewall
a machine or a dedicated piece of hardware that in the topology of a local network is the single connection
point to the outside network, and which filters, controls the activity on some ports, or makes sure that
only some specific interfaces may have access to the outside world.

flag
is an indicator (usually a bit) that is used to signal some condition to a program. For example, a filesystem
has, among others, a flag indicating if it has to be dumped in a backup, so when the flag is active the
filesystem gets backed up, and when it’s inactive it doesn’t.

focus
the state for a window to receive keyboard events (such as key-presses, key-releases and mouse clicks)
unless they are trapped by the window manager.

97
Appendix A. Glossary

foreground
in shell context, the process in the foreground is the one that is currently running. You have to wait for
such a process to finish in order to be able to type commands again.
See Also: job, background.

Frame Relay
Frame Relay is a network technology ideally suited to carrying traffic which is of bursty or sporadic
nature. Network costs are reduced by having many Frame Relay customers sharing the same network
capacity and relying on them wanting to make use of the network at slightly different times.

framebuffer
projection of a video card’s RAM into the machine’s address space. This allows for applications to access
the video RAM without the chore of having to talk to the card. All high-end graphical workstations use
frame buffers.
FTP
File Transfer Protocol. This is the standard Internet protocol used to transfer files from one machine to
another.
full-screen
This term is used to refer to applications that take up the entire visible area of your display.

gateway
link connecting two IP networks.

GFDL
The GNU Free Documentation License. The license which applies to all Mandrakelinux documentation.

GIF
Graphics Interchange Format. An image file format, widely used on the web. GIF images may be compres-
sed or animated. Due to copyright problems it is a bad idea to use them, so the recommended solution is
to replace them as much as possible by the PNG format.

globbing
in the shell, the ability to group a certain set of filenames with a globbing pattern.
See Also: globbing pattern.

globbing pattern
a string made of normal characters and special characters. Special characters are interpreted and expan-
ded by the shell.

GNU
GNU’s Not Unix. The GNU project was initiated by Richard Stallman at the beginning of the 1980s, and
aimed at developing a free operating system (“free” as in “free speech”). Currently, all tools are there,
except... the kernel. The GNU project kernel, Hurd, is not rock solid yet. Linux borrows, among others,
two things from GNU: its C compiler, gcc, and its license, the GPL.
See Also: GPL.
GPL
General Public License. The license of the GNU/Linux kernel, it goes the opposite way of all proprietary
licenses in that it applies no restrictions as to copying, modifying and redistributing the software, as long
as the source code is made available. The only restriction is that the persons to whom you redistribute it
must also benefit from the same rights.

GUI
Graphical User Interface. Interface to a computer consisting of windows with menus, buttons, icons and so
on. A great majority of users prefer a GUI to a CLI (Command Line Interface) for ease of use, even though
the latter is far more versatile.
guru
An expert. Used to qualify someone particularly skilled, but also of valuable help for others.

hardware address
This is a number that uniquely identifies a host in a physical network at the media access layer. Examples
of this are Ethernet Addresses and AX.25 Addresses.

98
 Appendix A. Glossary

hidden file
is a file which can’t be “seen” when doing a ls command with no options. Hidden files’ filenames be-
gin with a . and are used to store the user’s personal preferences and configurations for the different
programs (s)he uses. For example, bash’s command history is saved into .bash_history, a hidden file.

home directory
often abbreviated as “home”, this is the name for the personal directory of a given user.
See Also: account.
host
refers to a computer and is commonly used when talking about computers that are connected to a net-
work.
HTML
HyperText Markup Language. The language used to create web documents.

HTTP
HyperText Transfer Protocol. The protocol used to connect to web sites and retrieve HTML documents or
files.
icon
is a little drawing (normally sized 16x 16, 32x 32, 48x 48 and sometimes 64x 64 pixels) which in a graphical
environment represents a document, a file or a program.

IDE
Integrated Drive Electronics. The most widely used bus on today’s PC s for hard disks. An IDE bus may
contain up to two devices, and the speed of the bus is limited by the device on the bus with the slower
command queue (and not the slower transfer rate!).
See Also: ATAPI.
IP masquerading
This is a technique where a firewall is used to hide your computer’s true IP address from the outside.
Typically, any outside network connections you make through the firewall will inherit the firewall’s IP
address. This is useful in situations where you may have a fast Internet connection with only one IP
address but wish to use more than one computer on your internal network.

inode
entry point leading to the contents of a file on a UNIX-like filesystem. An inode is identified in a unique
way by a number, and contains meta-information about the file it refers to, such as its access times, its
type, its size, but not its name!

insert mode
under Vi or any of its clones, it is the state of the program in which pressing a key will insert that character
in the file being edited (except pathological cases like the completion of an abbreviation, right justify at
the end of the line, ...). One gets out of it pressing the key Esc (or Ctrl-[).

Internet
is a huge network that connects computers around the world.

IP address
is a numeric address consisting of four parts which identifies your computer on the Internet. IP addresses
are structured in a hierarchical manner, with top level and national domains, domains, sub-domains and
each machine’s personal address. An IP address will look something like 192.168.0.1. A machine’s
personal address can be one of two types: static or dynamic. Static IP addresses are addresses which
never change, they are permanently assigned.. Dynamic IP addresses mean that an IP address will change
with each new connection to the network. Dial-up and cable modem users typically have dynamic IP
addresses while some DSL and other high-speed connections provide static IP addresses.

IRC
Internet Relay Chat. One of the few Internet standards for live speech. It allows for channel creation,
private talks and file exchange. It also allows servers to connect to each other, which is why several IRC
networks exist today: Undernet, DALnet, EFnet to name a few.

99
Appendix A. Glossary

IRC channels
are the “places” inside IRC servers where you can chat with other people. Channels are created in IRC
servers and users join those channels so they can communicate with each other. Messages written on one
channel are only visible to the people connected to that channel. Two or more users can create a “private”
channel so they don’t get disturbed by other users. Channel names begin with a #.

ISA
Industry Standard Architecture. The very first bus used on PC s, it is slowly being abandoned in favor of
the PCI bus. ISA is still commonly found on SCSI cards supplied with scanners, CD writers and some
other older hardware.
ISDN
Integrated Services Digital Network. A set of communication standards for voice, digital network services
and video. It has been designed to eventually replace the current phone system, known as PSTN (Public
Switched Telephone Network) or POTS (Plain Ole Telephone Service). ISDN is known as a circuit switched
data network.
ISO
International Standards Organization. A group of companies, consultants, universities and other sources
which enumerates standards in various disciplines, including computing. The papers describing stan-
dards are numbered. The standard number iso9660, for example, describes the file system used on CD-
ROMs.
ISO 8859
The ISO 8859 standard includes several 8-bit extensions to the ASCII character set. Especially important
is ISO 8859-1, the “Latin Alphabet No. 1”, which has become widely implemented and may already be
seen as the de facto standard ASCII replacement.
ISO 8859-1 supports the following languages: Afrikaans, Basque, Catalan, Danish, Dutch, English, Faroe-
se, Finnish, French, Galician, German, Icelandic, Irish, Italian, Norwegian, Portuguese, Scottish, Spanish,
and Swedish.
Note that the ISO 8859-1 characters are also the first 256 characters of ISO 10646 (Unicode). However, it
lacks the EURO symbol and does not fully cover Finnish and French. ISO 8859-15 is a modification of
ISO 8859-1 to covers these needs.
See Also: ASCII.
ISP
Internet Service Provider. A company which sells Internet access to its customers, either over telephone
lines or high-bandwidth circuits such as dedicated T-1 circuits, DSL or cable.

JPEG
Joint Photographic Experts Group. Another very common image file format. JPEG is mostly suited for com-
pressing real-world scenes, and does not work very well on non-realistic images.

job
in a shell context, a job is a process running in the background. You can have several jobs in the same shell
and control these jobs independently.
See Also: foreground, background.

kernel
is the core of the operating system. The kernel is responsible for allocating resources and separating
processes from each other. It handles all of the low-level operations which allow programs to talk directly
to the hardware on your computer, manages the buffer cache and so on.

kill ring
under Emacs, it is the set of text areas cut or copied since the editor was started. The text areas may be
recalled to be inserted again, and the structure is ring-like.

LAN
Local Area Network. Generic name given to a network of machines connected to the same physical wire.

launch
is the action of invoking, or starting, a program.

100
 Appendix A. Glossary

LDP
Linux Documentation Project. A nonprofit organization that maintains GNU/Linux documentation. It’s
mostly known for documents like HOWTOs, but it also maintains FAQ s, and even a few books.

library
is a collection of procedures and functions in binary form to be used by programmers in their programs
(as long as the library’s license allows them to do so). The program in charge of loading shared libraries
at run time is called the dynamic linker.

link
reference to an inode in a directory, therefore giving a (file) name to the inode. Examples of inodes which
don’t have a link (and hence have no name) are: anonymous pipes (as used by the shell), sockets (aka
network connections), network devices and so on.

linkage
last stage of the compile process, consisting of linking together all object files in order to produce an
executable file, and matching unresolved symbols with dynamic libraries (unless a static linkage has
been requested, in which case the code of these symbols will be included in the executable).

Linux
is a UNIX-like operating system which runs on a variety of different computers, and is free for anyone to
use and modify. Linux (the kernel) was written by Linus Torvalds.

login
connection name for a user on a UNIX system, and the action to connect.

lookup table
is a table that stores corresponding codes (or tags) and their meaning. It is often a data file used by a
program to get further information about a particular item.
For example, HardDrake uses such a table to know what a manufacturer’s product code means. This is
one line from the table, giving information about item CTL0001
CTL0001 sound sb Creative Labs SB16 \
HAS_OPL3|HAS_MPU401|HAS_DMA16|HAS_JOYSTICK

loopback
virtual network interface of a machine to itself, allowing the running programs not to have to take into
account the special case where two network entities are in fact the same machine.

major
number specific to the device class.

manual page
a small document containing the definitions of a command and its usage, to be consulted with the man
command. The first thing one should (learn how to) read when learning about a command you aren’t
familiar with.
MBR
Master Boot Record. Name given to the first sector of a bootable hard drive. The MBR contains the code
used to load the operating system into memory or a bootloader (such as LILO), and the partition table of
that hard drive.
MIME
Multipurpose Internet Mail Extensions. A string of the form type/subtype describing the contents of a file
attached in an e-mail. This allows MIME -aware mail clients to define actions depending on the type of
the file.
minor
number identifying the specific device we are talking about.

MPEG
Moving Pictures Experts Group. An ISO committee which generates standards for video and audio com-
pression. MPEG is also the name of their algorithms. Unfortunately, the license for this format is very
restrictive, and as a consequence there are still no Open Source MPEG players...

101
Appendix A. Glossary

mount point
is the directory where a partition or another device is attached to the GNU/Linux filesystem. For example,
your CD-ROM is mounted in the /mnt/cdrom directory, from where you can explore the contents of any
mounted CDs.
mounted
A device is mounted when it is attached to the GNU/Linux filesystem. When you mount a device you can
browse its contents. This term is partly obsolete due to the “supermount” feature, so users do not need
to manually mount removable media.
See Also: mount point.

MSS
The Maximum Segment Size (MSS) is the largest quantity of data which can be transmitted at one time.
If you want to prevent local fragmentation MSS would equal MTU-IP header.

MTU
The Maximum Transmission Unit (MTU) is a parameter which determines the largest datagram than
can be transmitted by an IP interface without it needing to be broken down into smaller units. The MTU
should be larger than the largest datagram you wish to transmit un-fragmented. Note, this only prevents
fragmentation locally, some other link in the path may have a smaller MTU and the datagram will be
fragmented there. Typical values are 1500 bytes for an Ethernet interface, or 576 bytes for a PPP interface.

multitasking
the ability of an operating system to share CPU time between several processes. At a low level, this is also
known as multiprogramming. Switching from one process to another requires that all the current process
context be saved and restored when this process runs again. This operation is called a context switch, and
on Intel, is done 100 times per second, thereby making it fast enough so that a user has the illusion that
the operating system runs several applications at the same time. There are two types of multitasking:
in preemptive multitasking the operating system is responsible for taking away the CPU and passing
it to another process; cooperative multitasking is where the process itself gives back the CPU. The first
variant is obviously the better choice because no program can consume the entire CPU time and block
other processes. GNU/Linux performs preemptive multitasking. The policy to select which process should
be run, depending on several parameters, is called scheduling.

multiuser
is used to describe an operating system which allows multiple users to log into and use the system at the
exact same time, each being able to do their own work independent of other users. A multitasking ope-
rating system is required to provide multiuser support. GNU/Linux is both a multitasking and multiuser
operating system, as is any UNIX system for that matter.

named pipe
a UNIX pipe which is linked, as opposed to pipes used in shells.
See Also: pipe, link.

naming
a word commonly used in computing for a method to identify objects. You will often hear of “naming
conventions” for files, functions in a program and so on.

NCP
NetWare Core Protocol. A protocol defined by Novell to access Novell NetWare file and print services.

NFS
Network File System. A network file system created by Sun Microsystems in order to share files across a
network in a transparent way.
newsgroups
discussion and news areas which can be accessed by a news or USENET client to read and write mes-
sages specific to the topic of the newsgroup. For example, the newsgroup alt.os.linux.mandrake is an
alternate newsgroup (alt) dealing with the Operating System (OS) GNU/Linux, and specifically, Mandrake-
linux (mandrake). Newsgroups are broken down in this fashion to make it easier to search for a particular
topic.

102
 Appendix A. Glossary

NIC
Network Interface Controller. An adapter installed in a computer which provides a physical connection to
a network, such as an Ethernet card.

NIS
Network Information System. NIS was also known as “Yellow Pages”, but British Telecom holds a copy-
right on this name. NIS is a protocol designed by Sun Microsystems in order to share common informa-
tion across a NIS domain, which may consist of an entire LAN, or just a part of it. It can export password
databases, service databases, groups information and more.

null, character
the character or byte number 0. It is used to mark the end of a string.

object code
is the code generated by the compilation process to be linked with other object codes and libraries to
form an executable file. Object code is machine readable.
See Also: compilation, linkage.

on the fly
Something is said to be done “on the fly” when it’s done along with something else, without you noticing
it or explicitly asking for it.
open source
is the name given to free source code of a program that is made available to the development community
and public at large. The theory behind this is that allowing source code to be used and modified by
a broader group of programmers will ultimately produce a more useful product for everyone. Some
popular open source programs include Apache, sendmail and GNU/Linux.

operating system
is the interface between the applications and the underlying hardware. The tasks for any operating sys-
tem are primarily to manage all of the machine specific resources. On a GNU/Linux system, this is done
by the kernel and loadable modules. Other well-known operating systems include AmigaOS, MacOS,
FreeBSD, OS/2, UNIX, Windows NT, and Windows 9x.
owner
in the context of users and their files, the owner of a file is the user who created that file.
owner group
in the context of groups and their files, the owner group of a file is the group to which the user who
created that file belongs.

PAP
Password Authentication Protocol. A protocol used by many ISPs to authenticate their clients. In this sche-
me, the client (you) sends an identifier/password pair to the server, but none of the information is encryp-
ted. See CHAP for the description of a more secure system.
See Also: CHAP.
pager
program displaying a text file one screen at a time, and making it easy to move back and forth and search
for strings in this file. We suggest you to use less.

password
is a secret word or combination of words or letters which is used to secure something. Passwords are
used in conjunction with user logins to multi-user operating systems, web sites, FTP sites, and so forth.
Passwords should be hard-to-guess phrases or alphanumeric combinations, and should never be based
on common dictionary words. Passwords ensure that other people cannot log into a computer or site
with your account.

patch, to patch
file containing a list of corrections to issue to source code in order to add new features, to remove bugs,
or to modify it according to one’s wishes and needs. The action consisting of the application of these
corrections to the archive of source code (aka “patching”).

103
Appendix A. Glossary

path
is an assignment for files and directories to the filesystem. The different layers of a path are separated by
the "slash" or ’/’ character. There are two types of paths on GNU/Linux systems. The relative path is the
position of a file or directory in relation to the current directory. The absolute path is the position of a file
or directory in relation to the root directory.

PCI
Peripheral Components Interconnect. A bus created by Intel which today is the standard bus for PC and
other architectures. It is the successor to ISA, and it offers numerous services: device identification, con-
figuration information, IRQ sharing, bus mastering and more.

PCMCIA
Personal Computer Memory Card International Association. More and more commonly called “PC Card”
for simplicity reasons, this is the standard for external cards attached to a laptop: modems, hard disks,
memory cards, Ethernet cards, and more. The acronym is sometimes humorously expanded to People
Cannot Memorize Computer Industry Acronyms...

pipe
a special UNIX file type. One program writes data into the pipe, and another program reads the data at
the other end. UNIX pipes are FIFO s, so the data is read at the other end in the order it was sent. Very
widely used with the shell. See also named pipe.

pixmap
is an acronym for “pixel map”. It’s another way of referring to bitmap images.

plugin
add-on program used to display or play some multimedia content found on a web document. It can
usually be easily downloaded if your browser is not yet able to display or play that kind of information.

PNG
Portable Network Graphics. Image file format created mainly for web use, it has been designed as a patent-
free replacement for GIF and also has some additional features.

PnP
Plug’N’Play. First an add-on for ISA in order to add configuration information for devices, it has become
a more widespread term which groups all devices able to report their configuration parameters. All PCI
devices are Plug’N’Play.

POP
Post Office Protocol. One common protocol used for retrieving mail from an ISP. See IMAP for an example
of another remote-access mail protocol.

porting
one of two ways to run a program on a system it was not originally intended for. For example, to be able
to run a Windows-native program under GNU/Linux (natively), it must first be ported to GNU/Linux.

PPP
Point to Point Protocol. This is the protocol used to send data over serial lines. It is commonly used to send
IP packets to the Internet, but it can also be used with other protocols such as Novell’s IPX protocol.

precedence
dictates the order of evaluation of operands in an expression. For example: If you have 4 + 3 * 2 you get
10 as the result, since the multiplication has higher precedence than the addition. If you want to evaluate
the addition first, then you have to add parenthesis like this: (4 + 3) * 2. When you do this, you’ll get
14 as the result since the parenthesis have higher precedence than the addition and the multiplication, so
the operations in parenthesis get evaluated first.
preprocessors
are compilation directives which instruct the compiler to replace those directives for code in the pro-
gramming language used in the source file. Examples of C ’s preprocessors are #include, #define, etc.
process
in the operating system context, a process is an instance of a program being executed along with its
environment.

104
 Appendix A. Glossary

prompt
in a shell, this is the string before the cursor. When you see it, you can type your commands.

protocol
Protocols organize the communications between different machines across a network, either using hard-
ware or software. They define the format of transferred data, whether one machine controls another, etc.
Many well-known protocols include HTTP, FTP, TCP, and UDP.
proxy
a machine which sits between a network and the Internet, whose role is to speed up data transfers for the
most widely used protocols (for example, HTTP and FTP). It maintains a cache of previous requests, so a
machine which makes a request for something which is already cached will receive it quickly, because it
will get the information from the local cache. Proxies are very useful on low bandwidth networks (such
as modem connections). Sometimes the proxy is the only machine able to access outside the network.

pull-down menu
is a menu that is “rolled” with a button in some of its corners. When you press that button, the menu
“unrolls” itself, showing you the full menu.

quota
is a method for restricting disk usage and limits for users. Administrators can restrict the size of home
directories for a user by setting quota limits on specific file systems.

RAID
Redundant Array of Independent Disks. A project initiated at the computing science department of Berkeley
University, in which the storage of data is spread across an array of disks using different schemes. At
first, this was implemented using floppy drives, which is why the acronym originally stood for Redundant
Array of Inexpensive Disks.

RAM
Random Access Memory. Term used to identify a computer’s main memory. The “Random” here means
that any part of the memory can be directly accessed.

read-only mode
for a file means that the file cannot be written to. You can read its contents but you can’t modify them.
See Also: read-write mode.
read-write mode
for a file, it means that the file can be written to. You can read its contents and modify them.
See Also: read-only mode.

regular expression
a powerful theoretical tool which is used to search and match text strings. It lets one specify patterns
these strings must obey. Many UNIX utilities use it: sed, awk, grep, perl and others.

RFC
Request For Comments. RFC s are the official Internet standard documents, published by the IETF (Internet
Engineering Task Force). They describe all protocols, their usage, their requirements and so on. When you
want to learn how a protocol works, pick up the corresponding RFC.

root
is the superuser of any UNIX system. Typically root (aka the system administrator) is the person responsi-
ble for maintaining and supervising the UNIX system. This person also has complete access to everything
on the system.

root directory
This is the top level directory of a filesystem. This directory has no parent directory, thus ’..’ for root
points back to itself. The root directory is written as ’/’.

root filesystem
This is the top level filesystem. This is the filesystem where GNU/Linux mounts its root directory tree. It is
necessary for the root filesystem to reside in a partition of its own, as it is the basis for the whole system.
It contains the root directory.

105
Appendix A. Glossary

route
Is the path that your datagrams take through the network to reach their destination. It is the path between
one machine and another in a network.
RPM
Red Hat Package Manager. A packaging format developed by Red Hat in order to create software packages,
it is used in many GNU/Linux distributions, including Mandrakelinux.

run level
is a configuration of the system software which only allows certain selected processes to exist. Allowed
processes are defined, for each runlevel, in the file /etc/inittab. There are eight defined runlevels: 0,
1, 2, 3, 4, 5, 6, S and switching between them can only be achieved by a privileged user by means of
executing the commands init and telinit.

script
shell scripts are sequences of commands to be executed as if they were sequentially entered in the console.
shell scripts are UNIX’s (somewhat) equivalent of DOS batch files.

SCSI
Small Computers System Interface. A bus with a high throughput designed to allow for several types of pe-
ripherals to be connected to it. Unlike IDE, a SCSI bus is not limited by the speed at which the peripherals
accept commands. Only high-end machines integrate a SCSI bus directly on the motherboard, therefore
most PC s need add-on cards.
security levels
Mandrakelinux’s unique feature which allows you to set different levels of restriction according to how
secure you want to make your system. There are 6 predefined levels ranging from 0 to 5, where 5 is the
tightest security. You can also define your own security level.

segmentation fault
A segmentation fault occurs when a program tries to access memory that is not allocated to it. This
generally causes the program to stop immediately.
server
program or computer that provides a feature or service and awaits the connections from clients to execute
their orders or give them the information they ask. In the case of peer to peer systems such as SLIP or
PPP, the server is taken to be the end of the link that is called and the end calling is taken to be the client.
It is one of the components of a client/ server system.

shadow passwords
a password management suite on UNIX systems in which the file containing the encrypted passwords
is not world-readable, unlike that usually found with a normal password system. It also offers other
features such as password aging.

shell
The shell is the basic interface to the operating system kernel and provides the command line where users
enter commands to run programs and system commands. All shells provide a scripting language which
can be used to automate tasks or simplify often-used complex tasks. These shell scripts are similar to
batch files from the DOS operating system, but are much more powerful. Some example shells are bash,
sh, and tcsh.

single user
is used to describe a state of an operating system, or even an operating system itself, that only allows a
single user to log into and use the system at any time.

site dependent
means that the information used by programs like imake and make to compile some source file depends
on the site, the computer architecture, the computer’s installed libraries, and so on.

SMB
Server Message Block. Protocol used by Windows machines (9x or NT ) for file and printer sharing across a
network.
See Also: CIFS.

106
 Appendix A. Glossary

SMTP
Simple Mail Transfer Protocol. This is the common protocol for transferring email. Mail Transfer Agents
such as sendmail or postfix use SMTP. They are sometimes called SMTP servers.

socket
file type corresponding to any network connection.

soft links
See: symbolic links

standard error
the file descriptor number 2, opened by every process, used by convention to print error messages to the
terminal screen.
See Also: standard input, standard output.

standard input
the file descriptor number 0, opened by every process, used by convention as the file descriptor from
which the process receives data.
See Also: standard error, standard output.

standard output
the file descriptor number 1, opened by every process, used by convention as the file descriptor in which
the process prints its output.
See Also: standard error, standard input.

streamer
is a device which takes “streams” (not interrupted or divided in shorter chunks) of characters as its input.
A typical streamer is a tape drive.

SVGA
Super Video Graphics Array. The video display standard defined by VESA for the PC architecture. The
resolution is 800x 600 x 16 colors.
switch
Switches are used to change the behavior of programs, and are also called command-line options or
arguments. To determine if a program has optional switches which may be used, read the man pages or
try to pass the --help switch to the program (i.e.. program --help).

symbolic links
are special files, containing nothing but a string which references another file. Any access to them is the
same as accessing the file whose name is the referenced string, which may or may not exist, and the path
to which can be given in a relative or an absolute way.

target
is the object of compilation, i.e. the binary file to be generated by the compiler.

TCP
Transmission Control Protocol. This is the most common reliable protocol which uses IP to transfer network
packets. TCP adds the necessary checks on top of IP to make sure that packets are delivered. Unlike
UDP, TCP works in connected mode, which means that two machines must establish a connection before
exchanging data.

telnet
creates a connection to a remote host and allows you to log into the machine, provided you have an
account. Telnet is the most widely-used method of remote logins, however there are better and more
secure alternatives, such as ssh.

theme-able
a graphical application is theme-able if it is able to change its appearance in real time. Many window
managers are theme-able.

traverse
for a directory on a UNIX system, this means that the user is allowed to go through this directory, and
possibly to directories under it. This requires that the user has the execute permission on this directory.

107
Appendix A. Glossary

URL
Uniform Resource Locator. A string with a special format used to identify a resource on the Internet in a
unique way. The resource can be a file, a server or other item. The syntax for a URL is
protocol://server.name[:port]/path/to/resource.
When only a machine name is given and the protocol is http://, it defaults to retrieving the file index.
html on the server.
username
is a name (or more generally a word) which identifies a user on a system. Each username is attached to a
unique and single UID (user ID)
See Also: login.
variables
are strings which are used in Makefile files to be replaced by their value each time they appear. Usually
they are set at the beginning of the Makefile. They are used to simplify Makefile and source files tree
management.
More generally, variables in programming are words that refer to other entities (numbers, strings, tables,
etc.) that are likely to vary while the program is executing.
verbose
For commands, the verbose mode means that the command reports to standard (or possibly error) output
all the actions it performs and the results of those actions. Sometimes, commands have a way to define
the “verbosity level”, which means that the amount of information that the command will report can be
controlled.
VESA
Video Electronics Standards Association. An industry standards association aimed at the PC architecture.
For example, it is the author of the SVGA standard.
virtual console
is the name given to what used to be called terminals. On GNU/Linux systems, you have what are called
virtual consoles which enable you to use one screen or monitor for many independently running sessions.
By default, you have six virtual consoles that can be reached by pressing ALT-F1 through ALT-F6. There
is a seventh virtual console, ALT-F7, which will permit you to reach a running X Window System. In X,
you can reach the text console by pressing CTRL-ALT-F1 through CTRL-ALT-F6.
See Also: console.
virtual desktops
In the X Window System, the window manager may provide you several desktops. This handy feature
allows you to organize your windows, avoiding the problem of having dozens of them stacked on top of
each other. It works as if you had several screens. You can switch from one virtual desktop to another in
a manner which depends on the window manager you’re using.
See Also: window manager, desktop.
WAN
Wide Area Network. This network, although similar to a LAN, connects computers on a network which is
not physically connected to the same wires and are separated by a greater distance.
wildcard
The ’*’ and ’?’ characters are used as wildcard characters and can represent anything. The ’*’ represents
any number of characters, including no characters. The ’?’ represents exactly one character. Wildcards are
often used in regular expressions.
window
In networking, the window is the largest amount of data that the receiving end can accept at a given
point in time.
window manager
the program responsible for the “look and feel” of a graphical environment, dealing with window bars,
frames, buttons, root menus, and some keyboard shortcuts. Without it, it would be hard or impossible to
have virtual desktops, to resize windows on the fly, to move them around, ...
workspace switcher
a little applet that allows you to switch between the available virtual desktops.
See Also: virtual desktops.

108
Index environment
process, 63
.bashrc, 18 variable, 10
account, 5 FHS, 47
applications file
ImageMagick, 23 attribute, 19, 62
terminals, 23 block mode, 57
attribute block mode, 60
file, 19 character mode, 57
characters character mode, 60
globbing, 21 copying, 18
special, 23 creating, 17
collating order, 21 deleting, 17
command line, 33 find, 36
completion, 23 link, 57, 58
introduction, 17 moving, 18
commands renaming, 18
at, 39 socket, 57
bzip2, 41, 74 file system
cat, 10 Devfs, 16
cd, 9 framebuffer, 92
chgrp, 19 GID, 6
chmod, 20 globbing
chown, 19 character, 21
configure, 75 group, 5
cp, 18 change, 19
crontab, 38 GRUB, 91
find, 36 home
grep, 34 partition, 14
gzip, 41 IDE
init, 69 devices, 15
kill, killall, 44 inode, 58
less, 11, 22 internationalization, 2
ls, 11 LILO, 89
make, 77 link
mkdir, 17 hard, 61
mount, 52 symbolic, 61
mv, 18 Makefile, 72, 78
patch, 86 Mandrakeclub, 1
ps, 43 Mandrakeexpert, 1
pwd, 9 Mandrakelinux
rm, 17 mailing lists, 1
rmdir, 17 Mandrakesecure, 1
sed, 22 Mandrakestore, 1
tar, 40, 73 modules, 65
touch, 17 owner, 19
umount, 53 change, 19
wc, 22 packaging, 1
console, 5 partitions, 13, 51
development, 2 extended, 15
Device File System, 16 logical, 15
directory primary, 15
copying, 18 password, 5
creating, 17 permissions, 20
deleting, 17 Peter Pingus, 4
moving, 18 PID, 8
renaming, 18 pipe, 22
disks, 13 anonymous, 59
documentation, 2 file, 57
Mandrakelinux, 3 named, 59

109
primary
master, 15
slave, 15
process, 8, 24, 63
processes, 43
programming, 2
prompt, 6, 8
Queen Pingusa, 4
RAM memory, 14
redirection, 22
root
directory, 47, 64
partition, 14
user, 6
runlevel, 69
SCSI
disks, 15
sector, 13
shell, 8, 17
globbing patterns, 21
Soundblaster, 15
standard
error, 21
input, 21
output, 21
swap, 13
partition, 14
size, 14
text editors
vi, 27
text editors
Emacs, 25
timestamps
atime, 17
ctime, 17
mtime, 17
UID, 6
UNIX, 5
users, 5
generic, 4
usr
partition, 14
utilities
file-handling, 17
values
discrete, 21
virus, 8

110