/ / Language: English / Genre:comp_osnet,

Linux Network Administrator Guide Second Edition

Olaf Kirch

This book was written to provide a single reference for network administration in a Linux environment. Beginners and experienced users alike should find the information they need to cover nearly all important administration activities required to manage a Linux network configuration. The possible range of topics to cover is nearly limitless, so of course it has been impossible to include everything there is to say on all subjects. We've tried to cover the most important and common ones. We've found that beginners to Linux networking, even those with no prior exposure to Unix-like operating systems, have found this book good enough to help them successfully get their Linux network configurations up and running and get them ready to learn more. There are many books and other sources of information from which you can learn any of the topics covered in this book (with the possible exception of some of the truly Linux-specific features, such as the new Linux firewall interface, which is not well documented elsewhere) in greater depth. We've provided a bibliography for you to use when you are ready to explore more.

Olaf Kirch & Terry Dawson

Linux Network Administrator's Guide, 2nd Edition


The Internet is now a household term in many countries. With otherwise serious people beginning to joyride along the Information Superhighway, computer networking seems to be moving toward the status of TV sets and microwave ovens. The Internet has unusually high media coverage, and social science majors are descending on Usenet newsgroups, online virtual reality environments, and the Web to conduct research on the new "Internet Culture."

Of course, networking has been around for a long time. Connecting computers to form local area networks has been common practice, even at small installations, and so have long-haul links using transmission lines provided by telecommunications companies. A rapidly growing conglomerate of world-wide networks has, however, made joining the global village a perfectly reasonable option for even small non-profit organizations of private computer users. Setting up an Internet host with mail and news capabilities offering dialup and ISDN access has become affordable, and the advent of DSL (Digital Subscriber Line) and Cable Modem technologies will doubtlessly continue this trend.

Talking about computer networks often means talking about Unix. Of course, Unix is not the only operating system with network capabilities, nor will it remain a frontrunner forever, but it has been in the networking business for a long time, and will surely continue to be for some time to come.

What makes Unix particularly interesting to private users is that there has been much activity to bring free Unix-like operating systems to the PC, such as 386BSD, FreeBSD, and Linux.

Linux is a freely distributable Unix clone for personal computers. It currently runs on a variety of machines that includes the Intel family of processors, but also Motorola 680x0 machines, such as the Commodore Amiga and Apple Macintosh; Sun SPARC and Ultra-SPARC machines; Compaq Alphas; MIPS; PowerPCs, such as the new generation of Apple Macintosh; and StrongARM, like the rebel.com Netwinder and 3Com Palm machines. Linux has been ported to some relatively obscure platforms, like the Fujitsu AP-1000 and the IBM System 3/90. Ports to other interesting architectures are currently in progress in developers' labs, and the quest to move Linux into the embedded controller space promises success.

Linux was developed by a large team of volunteers across the Internet. The project was started in 1990 by Linus Torvalds, a Finnish college student, as an operating systems course project. Since that time, Linux has snowballed into a full-featured Unix clone capable of running applications as diverse as simulation and modeling programs, word processors, speech recognition systems, World Wide Web browsers, and a horde of other software, including a variety of excellent games. A great deal of hardware is supported, and Linux contains a complete implementation of TCP/IP networking, including SLIP, PPP, firewalls, a full IPX implementation, and many features and some protocols not found in any other operating system. Linux is powerful, fast, and free, and its popularity in the world beyond the Internet is growing rapidly.

The Linux operating system itself is covered by the GNU General Public License, the same copyright license used by software developed by the Free Software Foundation. This license allows anyone to redistribute or modify the software (free of charge or for a profit) as long as all modifications and distributions are freely distributable as well. The term "free software" refers to freedom of application, not freedom of cost.

Purpose and Audience for This Book

This book was written to provide a single reference for network administration in a Linux environment. Beginners and experienced users alike should find the information they need to cover nearly all important administration activities required to manage a Linux network configuration. The possible range of topics to cover is nearly limitless, so of course it has been impossible to include everything there is to say on all subjects. We've tried to cover the most important and common ones. We've found that beginners to Linux networking, even those with no prior exposure to Unix-like operating systems, have found this book good enough to help them successfully get their Linux network configurations up and running and get them ready to learn more.

There are many books and other sources of information from which you can learn any of the topics covered in this book (with the possible exception of some of the truly Linux-specific features, such as the new Linux firewall interface, which is not well documented elsewhere) in greater depth. We've provided a bibliography for you to use when you are ready to explore more.

Sources of Information

If you are new to the world of Linux, there are a number of resources to explore and become familiar with. Having access to the Internet is helpful, but not essential.

Linux Documentation Project guides

The Linux Documentation Project is a group of volunteers who have worked to produce books (guides), HOWTO documents, and manual pages on topics ranging from installation to kernel programming. The LDP works include:

Linux Installation and Getting Started

By Matt Welsh, et al. This book describes how to obtain, install, and use Linux. It includes an introductory Unix tutorial and information on systems administration, the X Window System, and networking.

Linux System Administrators Guide

By Lars Wirzenius and Joanna Oja. This book is a guide to general Linux system administration and covers topics such as creating and configuring users, performing system backups, configuration of major software packages, and installing and upgrading software.

Linux System Adminstration Made Easy

By Steve Frampton. This book describes day-to-day administration and maintenance issues of relevance to Linux users.

Linux Programmers Guide

By B. Scott Burkett, Sven Goldt, John D. Harper, Sven van der Meer, and Matt Welsh. This book covers topics of interest to people who wish to develop application software for Linux.

The Linux Kernel

By David A. Rusling. This book provides an introduction to the Linux Kernel, how it is constructed, and how it works. Take a tour of your kernel.

The Linux Kernel Module Programming Guide

By Ori Pomerantz. This guide explains how to write Linux kernel modules.

More manuals are in development. For more information about the LDP you should consult their World Wide Web server at http://www.linuxdoc.org/ or one of its many mirrors.

HOWTO documents

The Linux HOWTOs are a comprehensive series of papers detailing various aspects of the system - such as installation and configuration of the X Window System software, or how to write in assembly language programming under Linux. These are generally located in the HOWTO subdirectory of the FTP sites listed later, or they are available on the World Wide Web at one of the many Linux Documentation Project mirror sites. See the Bibliography at the end of this book, or the file HOWTO-INDEX for a list of what's available.

You might want to obtain the Installation HOWTO, which describes how to install Linux on your system; the Hardware Compatibility HOWTO, which contains a list of hardware known to work with Linux; and the Distribution HOWTO, which lists software vendors selling Linux on diskette and CD-ROM.

The bibliography of this book includes references to the HOWTO documents that are related to Linux networking.

Linux Frequently Asked Questions

The Linux Frequently Asked Questions with Answers (FAQ) contains a wide assortment of questions and answers about the system. It is a must-read for all newcomers.

Documentation Available via FTP

If you have access to anonymous FTP, you can obtain all Linux documentation listed above from various sites, including metalab.unc.edu:/pub/Linux/docs and tsx-11.mit.edu:/pub/linux/docs.

These sites are mirrored by a number of sites around the world.

Documentation Available via WWW

There are many Linux-based WWW sites available. The home site for the Linux Documentation Project can be accessed at http://www.linuxdoc.org/.

The Open Source Writers Guild (OSWG) is a project that has a scope that extends beyond Linux. The OSWG, like this book, is committed to advocating and facilitating the production of OpenSource documentation. The OSWG home site is at http://www.oswg.org:8080/oswg.

Both of these sites contain hypertext (and other) versions of many Linux related documents.

Documentation Available Commercially

A number of publishing companies and software vendors publish the works of the Linux Documentation Project. Two such vendors are:

Specialized Systems Consultants, Inc. (SSC)


P.O. Box 55549 Seattle, WA 98155-0549


1-206-782-7191 (FAX)



Linux Systems Labs


18300 Tara Drive

Clinton Township, MI 48036


1-810-987-3562 (FAX)


Both companies sell compendiums of Linux HOWTO documents and other Linux documentation in printed and bound form.

O'Reilly& Associates publishes a series of Linux books. This one is a work of the Linux Documentation Project, but most have been independently authored. Their range includes:

Running Linux

An installation and user guide to the system describing how to get the most out of personal computing with Linux.

Learning Debian GNU/Linux

Learning Red Hat Linux

More basic than Running Linux, these books contain popular distributions on CD-ROM and offer robust directions for setting them up and using them.

Linux in a Nutshell

Another in the successful "in a Nutshell" series, this book focuses on providing a broad reference text for Linux.

Linux Journal and Linux Magazine

Linux Journal and Linux Magazine are monthly magazines for the Linux community, written and published by a number of Linux activists. They contain articles ranging from novice questions and answers to kernel programming internals. Even if you have Usenet access, these magazines are a good way to stay in touch with the Linux community.

Linux Journal is the oldest magazine and is published by S.S.C. Incorporated, for which details were listed previously. You can also find the magazine on the World Wide Web at http://www.linuxjournal.com/.

Linux Magazine is a newer, independent publication. The home web site for the magazine is http://www.linuxmagazine.com/.

Linux Usenet Newsgroups

If you have access to Usenet news, the following Linux-related newsgroups are available:


A moderated newsgroup containing announcements of new software, distributions, bug reports, and goings-on in the Linux community. All Linux users should read this group. Submissions may be mailed to linux-announce@news.ornl.gov.


General questions and answers about installing or using Linux.


Discussions relating to systems administration under Linux.


Discussions relating to networking with Linux.


Discussions about developing the Linux kernel and system itself.


A catch-all newsgroup for miscellaneous discussions that don't fall under the previous categories.

There are also several newsgroups devoted to Linux in languages other than English, such as fr.comp.os.linux in French and de.comp.os.linux in German.

Linux Mailing Lists

There is a large number of specialist Linux mailing lists on which you will find many people willing to help with questions you might have.

The best-known of these are the lists hosted by Rutgers University. You may subscribe to these lists by sending an email message formatted as follows:

To: majordomo@vger.rutgers.edu Subject: anything at all Body: subscribe listname

Some of the available lists related to Linux networking are:


Discussion relating to Linux networking


Discussion relating to the Linux PPP implementation


Discussion relating to Linux kernel development

Online Linux Support

There are many ways of obtaining help online, where volunteers from around the world offer expertise and services to assist users with questions and problems.

The OpenProjects IRC Network is an IRC network devoted entirely to Open Projects - Open Source and Open Hardware alike. Some of its channels are designed to provide online Linux support services. IRC stands for Internet Relay Chat, and is a network service that allows you to talk interactively on the Internet to other users. IRC networks support multiple channels on which groups of people talk. Whatever you type in a channel is seen by all other users of that channel.

There are a number of active channels on the OpenProjects IRC network where you will find users 24 hours a day, 7 days a week who are willing and able to help you solve any Linux problems you may have, or just chat. You can use this service by installing an IRC client like irc-II, connecting to servername irc.openprojects.org:6667, and joining the #linpeople channel.

Linux User Groups

Many Linux User Groups around the world offer direct support to users. Many Linux User Groups engage in activities such as installation days, talks and seminars, demonstration nights, and other completely social events. Linux User Groups are a great way of meeting other Linux users in your area. There are a number of published lists of Linux User Groups. Some of the better-known ones are:

Groups of Linux Users Everywhere


LUG list project


LUG registry


Obtaining Linux

There is no single distribution of the Linux software; instead, there are many distributions, such as Debian, RedHat, Caldera, Corel, SuSE, and Slackware. Each distribution contains everything you need to run a complete Linux system: the kernel, basic utilities, libraries, support files, and applications software.

Linux distributions may be obtained via a number of online sources, such as the Internet. Each of the major distributions has its own FTP and web site. Some of these sites are:













Many of the popular general FTP archive sites also mirror various Linux distributions. The best-known of these sites are:





Many of the modern distributions can be installed directly from the Internet. There is a lot of software to download for a typical installation, though, so you'd probably want to do this only if you have a high-speed, permanent network connection, or if you just need to update an existing installation.[1]

Linux may be purchased on CD-ROM from an increasing number of software vendors. If your local computer store doesn't have it, perhaps you should ask them to stock it! Most of the popular distributions can be obtained on CD-ROM. Some vendors produce products containing multiple CD-ROMs, each of which provides a different Linux distribution. This is an ideal way to try a number of different distributions before you settle on your favorite one.

File System Standards

In the past, one of the problems that afflicted Linux distributions, as well as the packages of software running on Linux, was the lack of a single accepted filesystem layout. This resulted in incompatibilities between different packages, and confronted users and administrators with the task of locating various files and programs.

To improve this situation, in August 1993, several people formed the Linux File System Standard Group (FSSTND). After six months of discussion, the group created a draft that presents a coherent file sytem structure and defines the location of the most essential programs and configuration files.

This standard was supposed to have been implemented by most major Linux distributions and packages. It is a little unfortunate that, while most distributions have made some attempt to work toward the FSSTND, there is a very small number of distributions that has actually adopted it fully. Throughout this book, we will assume that any files discussed reside in the location specified by the standard; alternative locations will be mentioned only when there is a long tradition that conflicts with this specification.

The Linux FSSTND continued to develop, but was replaced by the Linux File Hierarchy Standard (FHS) in 1997. The FHS addresses the multi-architecture issues that the FSSTND did not. The FHS can be obtained from the Linux documentation directory of all major Linux FTP sites and their mirrors, or at its home site at http://www.pathname.com/fhs/. Daniel Quinlan, the coordinator of the FHS group, can be reached at quinlan@transmeta.com.

Standard Linux Base

The vast number of different Linux distributions, while providing lots of healthy choice for Linux users, has created a problem for software developers - particularly developers of non-free software.

Each distribution packages and supplies certain base libraries, configuration tools, system applications, and configuration files. Unfortunately, differences in their versions, names, and locations make it very difficult to know what will exist on any distribution. This makes it hard to develop binary applications that will work reliably on all Linux distribution bases.

To help overcome this problem, a new project sprang up called the "Linux Standard Base." It aims to describe a standard base distribution that complying distributions will use. If a developer designs an application to work against the standard base platform, the application will work, and be portable to, any complying Linux distribution.

You can find information on the status of the Linux Standard Base project at its home web site at http://www.linuxbase.org/.

If you're concerned about interoperability, particularly of software from commercial vendors, you should ensure that your Linux distribution is making an effort to participate in the standardization project.

About This Book

When Olaf joined the Linux Documentation Project in 1992, he wrote two small chapters on UUCP and smail, which he meant to contribute to the System Administrator's Guide. Development of TCP/IP networking was just beginning, and when those "small chapters" started to grow, he wondered aloud whether it would be nice to have a Networking Guide. "Great!" everyone said. "Go for it!" So he went for it and wrote the first version of the Networking Guide, which was released in September 1993.

Olaf continued work on the Networking Guide and eventually produced a much enhanced version of the guide. Vince Skahan contributed the original sendmail mail chapter, which was completely replaced in this edition because of a new interface to the sendmail configuration.

The version of the guide that you are reading now is a revision and update prompted by O'Reilly& Associates and undertaken by Terry Dawson.[2] Terry has been an amateur radio operator for over 20 years and has worked in the telecommunications industry for over 15 of those. He was co-author of the original NET-FAQ, and has since authored and maintained various networking-related HOWTO documents. Terry has always been an enthusiastic supporter of the Network Administrators Guide project, and added a few new chapters to this version describing features of Linux networking that have been developed since the first edition, plus a bunch of changes to bring the rest of the book up to date.

The exim chapter was contributed by Philip Hazel,[3] who is a lead developer and maintainer of the package.

The book is organized roughly along the sequence of steps you have to take to configure your system for networking. It starts by discussing basic concepts of networks, and TCP/IP-based networks in particular. It then slowly works its way up from configuring TCP/IP at the device level to firewall, accounting, and masquerade configuration, to the setup of common applications such as rlogin and friends, the Network File System, and the Network Information System. This is followed by a chapter on how to set up your machine as a UUCP node. Most of the remaining sections is dedicated to two major applications that run on top of TCP/IP and UUCP: electronic mail and news. A special chapter has been devoted to the IPX protocol and the NCP filesystem, because these are used in many corporate environments where Linux is finding a home.

The email part features an introduction to the more intimate parts of mail transport and routing, and the myriad of addressing schemes you may be confronted with. It describes the configuration and management of exim, a mail transport agent ideal for use in most situations not requiring UUCP, and sendmail, which is for people who have to do more complicated routing involving UUCP.

The news part gives you an overview of how Usenet news works. It covers INN and C News, the two most widely used news transport software packages at the moment, and the use of NNTP to provide newsreading access to a local network. The book closes with a chapter on the care and feeding of the most popular newsreaders on Linux.

Of course, a book can never exhaustively answer all questions you might have. So if you follow the instructions in this book and something still does not work, please be patient. Some of your problems may be due to mistakes on our part (see the section ", later in this Preface), but they also may be caused by changes in the networking software. Therefore, you should check the listed information resources first. There's a good chance that you are not alone with your problems, so a fix or at least a proposed workaround is likely to be known. If you have the opportunity, you should also try to get the latest kernel and network release from one of the Linux FTP sites or a BBS near you. Many problems are caused by software from different stages of development, which fail to work together properly. After all, Linux is a "work in progress."

The Official Printed Version

In Autumn 1993, Andy Oram, who had been around the LDP mailing list from almost the very beginning, asked Olaf about publishing this book at O'Reilly& Associates. He was excited about this book, never having imagined that it would become this successful. He and Andy finally agreed that O'Reilly would produce an enhanced Official Printed Version of the Networking Guide, while Olaf retained the original copyright so that the source of the book could be freely distributed. This means that you can choose freely: you can get the various free forms of the document from your nearest Linux Documentation Project mirror site and print it out, or you can purchase the official printed version from O'Reilly.

Why, then, would you want to pay money for something you can get for free? Is Tim O'Reilly out of his mind for publishing something everyone can print and even sell themselves?[4] Is there any difference between these versions?

The answers are "it depends," "no, definitely not," and "yes and no." O'Reilly& Associates does take a risk in publishing the Networking Guide, and it seems to have paid off for them (they've asked us to do it again). We believe this project serves as a fine example of how the free software world and companies can cooperate to produce something both can benefit from. In our view, the great service O'Reilly is providing to the Linux community (apart from the book becoming readily available in your local bookstore) is that it has helped Linux become recognized as something to be taken seriously: a viable and useful alternative to other commercial operating systems. It's a sad technical bookstore that doesn't have at least one shelf stacked with O'Reilly Linux books.

Why are they publishing it? They see it as their kind of book. It's what they'd hope to produce if they contracted with an author to write about Linux. The pace, level of detail, and style fit in well with their other offerings.

The point of the LDP license is to make sure no one gets shut out. Other people can print out copies of this book, and no one will blame you if you get one of these copies. But if you haven't gotten a chance to see the O'Reilly version, try to get to a bookstore or look at a friend's copy. We think you'll like what you see, and will want to buy it for yourself.

So what about the differences between the printed and online versions? Andy Oram has made great efforts at transforming our ramblings into something actually worth printing. (He has also reviewed a few other books produced by the Linux Documentation Project, contributing whatever professional skills he can to the Linux community.)

Since Andy started reviewing the Networking Guide and editing the copies sent to him, the book has improved vastly from its original form, and with every round of submission and feedback it improves again. The opportunity to take advantage of a professional editor's skill is one not to be wasted. In many ways, Andy's contribution has been as important as that of the authors. The same is also true of the copyeditors, who got the book into the shape you see now. All these edits have been fed back into the online version, so there is no difference in content.

Still, the O'Reilly version will be different. It will be professionally bound, and while you may go to the trouble to print the free version, it is unlikely that you will get the same quality result, and even then it is more unlikely that you'll do it for the price. Secondly, our amateurish attempts at illustration will have been replaced with nicely redone figures by O'Reilly's professional artists. Indexers have generated an improved index, which makes locating information in the book a much simpler process. If this book is something you intend to read from start to finish, you should consider reading the official printed version.


Chapter 1, Introduction to Networking, discusses the history of Linux and covers basic networking information on UUCP, TCP/IP, various protocols, hardware, and security. The next few chapters deal with configuring Linux for TCP/IP networking and running some major applications. We examine IP a little more closely in Chapter 2, Issues of TCP/IP Networking, before getting our hands dirty with file editing and the like. If you already know how IP routing works and how address resolution is performed, you can skip this chapter.

Chapter 3, Configuring the Networking Hardware, deals with very basic configuration issues, such as building a kernel and setting up your Ethernet card. The configuration of your serial ports is covered separately in Chapter 4, Configuring the Serial Hardware, because the discussion does not apply to TCP/IP networking only, but is also relevant for UUCP.

Chapter 5, Configuring TCP/IP Networking, helps you set up your machine for TCP/IP networking. It contains installation hints for standalone hosts with loopback enabled only, and hosts connected to an Ethernet. It also introduces you to a few useful tools you can use to test and debug your setup. Chapter 6, Name Service and Resolver Configuration, discusses how to configure hostname resolution and explains how to set up a name server.

Chapter 7, Serial Line IP, explains how to establish SLIP connections and gives a detailed reference for dip, a tool that allows you to automate most of the necessary steps. Chapter 8, The Point-to-Point Protocol, covers PPP and pppd, the PPP daemon.

Chapter 9, TCP/IP Firewall, extends our discussion on network security and describes the Linux TCP/IP firewall and its configuration tools: ipfwadm, ipchains, and iptables. IP firewalling provides a means of controlling who can access your network and hosts very precisely.

Chapter 10, IP Accounting, explains how to configure IP Accounting in Linux so you can keep track of how much traffic is going where and who is generating it.

Chapter 11, IP Masquerade and Network Address Translation, covers a feature of the Linux networking software called IP masquerade, which allows whole IP networks to connect to and use the Internet through a single IP address, hiding internal systems from outsiders in the process.

Chapter 12, Important Network Features, gives a short introduction to setting up some of the most important network applications, such as rlogin, ssh, etc. This chapter also covers how services are managed by the inetd superuser, and how you may restrict certain security-relevant services to a set of trusted hosts.

Chapter 13, The Network Information System, and Chapter 14, The Network File System, discuss NIS and NFS. NIS is a tool used to distribute administative information, such as user passwords in a local area network. NFS allows you to share filesystems between several hosts in your network.

In Chapter 15, IPX and the NCP Filesystem, we discuss the IPX protocol and the NCP filesystem. These allow Linux to be integrated into a Novell NetWare environment, sharing files and printers with non-Linux machines.

Chapter 16, Managing Taylor UUCP, gives you an extensive introduction to the administration of Taylor UUCP, a free implementation of the UUCP suite.

The remainder of the book is taken up by a detailed tour of electronic mail and Usenet news. Chapter 17, Electronic Mail, introduces you to the central concepts of electronic mail, like what a mail address looks like, and how the mail handling system manages to get your message to the recipient.

Chapter 18, Sendmail, and Chapter 19, Getting Exim Up and Running, cover the configuration of sendmail and exim, two mail transport agents you can use for Linux. This book explains both of them, because exim is easier to install for the beginner, while sendmail provides support for UUCP.

Chapter 20, Netnews, through Chapter 23, Internet News, explain the way news is managed in Usenet and how you install and use C News, nntpd, and INN: three popular software packages for managing Usenet news. After the brief introduction in Chapter 20, you can read Chapter 21, C News, if you want to transfer news using C News, a traditional service generally used with UUCP. The following chapters discuss more modern alternatives to C News that use the Internet-based protocol NNTP (Network News Transfer Protocol). Chapter 22, NNTP and the nntpd Daemon covers how to set up a simple NNTP daemon, nntpd, to provide news reading access for a local network, while Chapter 23 describes a more robust server for more extensive NetNews transfers, the InterNet News daemon (INN). And finally, Chapter 24, Newsreader Configuration, shows you how to configure and maintain various newsreaders.

Conventions Used in This Book

All examples presented in this book assume you are using a sh compatible shell. The bash shell is sh compatible and is the standard shell of all Linux distributions. If you happen to be a csh user, you will have to make appropriate adjustments.

The following is a list of the typographical conventions used in this book:


Used for file and directory names, program and command names, command-line options, email addresses and pathnames, URLs, and for emphasizing new terms.


Used for machine names, hostnames, site names, usernames and IDs, and for occasional emphasis.

Constant Width

Used in examples to show the contents of code files or the output from commands and to indicate environment variables and keywords that appear in code.

Constant Width Italic

Used to indicate variable options, keywords, or text that the user is to replace with an actual value.

Constant Width Bold

Used in examples to show commands or other text that should be typed literally by the user.

WARNING: Text appearing in this manner offers a warning. You can make a mistake here that hurts your system or is hard to recover from.

Submitting Changes

We have tested and verified the information in this book to the best of our ability, but you may find that features have changed (or even that we have made mistakes!). Please let us know about any errors you find, as well as your suggestions for future editions, by writing to:

O'Reilly& Associates, Inc.

101 Morris Street

Sebastopol, CA 95472

1-800-998-9938 (in the U.S. or Canada)

1-707-829-0515 (international or local)

1-707-829-0104 (FAX)

You can send us messages electronically. To be put on the mailing list or request a catalog, send email to:


To ask technical questions or comment on the book, send email to:


We have a web site for the book, where we'll list examples, errata, and any plans for future editions. You can access this page at:


For more information about this book and others, see the O'Reilly web site:



This edition of the Networking Guide owes almost everything to the outstanding work of Olaf and Vince. It is difficult to appreciate the effort that goes into researching and writing a book of this nature until you've had a chance to work on one yourself. Updating the book was a challenging task, but with an excellent base to work from, it was an enjoyable one.

This book owes very much to the numerous people who took the time to proof-read it and help iron out many mistakes, both technical and grammatical (never knew that there was such a thing as a dangling participle). Phil Hughes, John Macdonald, and Erik Ratcliffe all provided very helpful (and on the whole, quite consistent) feedback on the content of the book.

We also owe many thanks to the people at O'Reilly we've had the pleasure to work with: Sarah Jane Shangraw, who got the book into the shape you can see now; Maureen Dempsey, who copyedited the text; Rob Romano, Rhon Porter, and Chris Reilley, who created all the figures; Hanna Dyer, who designed the cover; Alicia Cech, David Futato, and Jennifer Niedherst for the internal layout; Lars Kaufman for suggesting old woodcuts as a visual theme; Judy Hoer for the index; and finally, Tim O'Reilly for the courage to take up such a project.

We are greatly indebted to Andres Sepúlveda, Wolfgang Michaelis, Michael K. Johnson, and all developers who spared the time to check the information provided in the Networking Guide. Phil Hughes, John MacDonald, and Eric Ratcliffe contributed invaluable comments on the second edition. We also wish to thank all those who read the first version of the Networking Guide and sent corrections and suggestions. You can find a hopefully complete list of contributors in the file Thanks in the online distribution. Finally, this book would not have been possible without the support of Holger Grothe, who provided Olaf with the Internet connectivity he needed to make the original version happen.

Olaf would also like to thank the following groups and companies that printed the first edition of the Networking Guide and have donated money either to him or to the Linux Documentation Project as a whole: Linux Support Team, Erlangen, Germany; S.u.S.E. GmbH, Fuerth, Germany; and Linux System Labs, Inc., Clinton Twp., United States, RedHat Software, North Carolina, United States.

Terry thanks his wife, Maggie, who patiently supported him throughout his participation in the project despite the challenges presented by the birth of their first child, Jack. Additionally, he thanks the many people of the Linux community who either nurtured or suffered him to the point at which he could actually take part and actively contribute. "I'll help you if you promise to help someone else in return."

The Hall of Fame

Besides those we have already mentioned, a large number of people have contributed to the Networking Guide, by reviewing it and sending us corrections and suggestions. We are very grateful.

Here is a list of those whose contributions left a trace in our mail folders.

Al Longyear, Alan Cox, Andres Sepúlveda, Ben Cooper, Cameron Spitzer, Colin McCormack, D.J. Roberts, Emilio Lopes, Fred N. van Kempen, Gert Doering, Greg Hankins, Heiko Eissfeldt, J.P. Szikora, Johannes Stille, Karl Eichwalder, Les Johnson, Ludger Kunz, Marc van Diest, Michael K. Johnson, Michael Nebel, Michael Wing, Mitch D'Souza, Paul Gortmaker, Peter Brouwer, Peter Eriksson, Phil Hughes, Raul Deluth Miller, Rich Braun, Rick Sladkey, Ronald Aarts, Swen Thüemmler, Terry Dawson, Thomas Quinot, and Yury Shevchuk.

Chapter 1. Introduction to Networking


The idea of networking is probably as old as telecommunications itself. Consider people living in the Stone Age, when drums may have been used to transmit messages between individuals. Suppose caveman A wants to invite caveman B over for a game of hurling rocks at each other, but they live too far apart for B to hear A banging his drum. What are A's options? He could 1) walk over to B's place, 2) get a bigger drum, or 3) ask C, who lives halfway between them, to forward the message. The last option is called networking.

Of course, we have come a long way from the primitive pursuits and devices of our forebears. Nowadays, we have computers talk to each other over vast assemblages of wires, fiber optics, microwaves, and the like, to make an appointment for Saturday's soccer match.[5] In the following description, we will deal with the means and ways by which this is accomplished, but leave out the wires, as well as the soccer part.

 We will describe three types of networks in this guide. We will focus on TCP/IP most heavily because it is the most popular protocol suite in use on both Local Area Networks (LANs) and Wide Area Networks (WANs), such as the Internet. We will also take a look at UUCP and IPX. UUCP was once commonly used to transport news and mail messages over dialup telephone connections. It is less common today, but is still useful in a variety of situations. The IPX protocol is used most commonly in the Novell NetWare environment and we'll describe how to use it to connect your Linux machine into a Novell network. Each of these protocols are networking protocols and are used to carry data between host computers. We'll discuss how they are used and introduce you to their underlying principles.

We define a network as a collection of hosts that are able to communicate with each other, often by relying on the services of a number of dedicated hosts that relay data between the participants. Hosts are often computers, but need not be; one can also think of X terminals or intelligent printers as hosts. Small agglomerations of hosts are also called sites.

Communication is impossible without some sort of language or code. In computer networks, these languages are collectively referred to as protocols. However, you shouldn't think of written protocols here, but rather of the highly formalized code of behavior observed when heads of state meet, for instance. In a very similar fashion, the protocols used in computer networks are nothing but very strict rules for the exchange of messages between two or more hosts.

TCP/IP Networks

Modern networking applications require a sophisticated approach to carrying data from one machine to another. If you are managing a Linux machine that has many users, each of whom may wish to simultaneously connect to remote hosts on a network, you need a way of allowing them to share your network connection without interfering with each other. The approach that a large number of modern networking protocols uses is called packet-switching. A packet is a small chunk of data that is transferred from one machine to another across the network. The switching occurs as the datagram is carried across each link in the network. A packet-switched network shares a single network link among many users by alternately sending packets from one user to another across that link.

The solution that Unix systems, and subsequently many non-Unix systems, have adopted is known as TCP/IP. When talking about TCP/IP networks you will hear the term datagram, which technically has a special meaning but is often used interchangeably with packet. In this section, we will have a look at underlying concepts of the TCP/IP protocols.

Introduction to TCP/IP Networks

TCP/IP traces its origins to a research project funded by the United States Defense Advanced Research Projects Agency (DARPA) in 1969. The ARPANET was an experimental network that was converted into an operational one in 1975 after it had proven to be a success.

In 1983, the new protocol suite TCP/IP was adopted as a standard, and all hosts on the network were required to use it. When ARPANET finally grew into the Internet (with ARPANET itself passing out of existence in 1990), the use of TCP/IP had spread to networks beyond the Internet itself. Many companies have now built corporate TCP/IP networks, and the Internet has grown to a point at which it could almost be considered a mainstream consumer technology. It is difficult to read a newspaper or magazine now without seeing reference to the Internet; almost everyone can now use it.

For something concrete to look at as we discuss TCP/IP throughout the following sections, we will consider Groucho Marx University (GMU), situated somewhere in Fredland, as an example. Most departments run their own Local Area Networks, while some share one and others run several of them. They are all interconnected and hooked to the Internet through a single high-speed link.

Suppose your Linux box is connected to a LAN of Unix hosts at the Mathematics department, and its name is erdos. To access a host at the Physics department, say quark, you enter the following command:

$ rlogin quark.physics

Welcome to the Physics Department at GMU

(ttyq2) login:

At the prompt, you enter your login name, say andres, and your password. You are then given a shell[6] on quark, to which you can type as if you were sitting at the system's console. After you exit the shell, you are returned to your own machine's prompt. You have just used one of the instantaneous, interactive applications that TCP/IP provides: remote login.

While being logged into quark, you might also want to run a graphical user interface application, like a word processing program, a graphics drawing program, or even a World Wide Web browser. The X windows system is a fully network-aware graphical user environment, and it is available for many different computing systems. To tell this application that you want to have its windows displayed on your host's screen, you have to set the DISPLAY environment variable:

$ DISPLAY=erdos.maths:0.0

$ export DISPLAY

If you now start your application, it will contact your X server instead of quark's, and display all its windows on your screen. Of course, this requires that you have X11 runnning on erdos. The point here is that TCP/IP allows quark and erdos to send X11 packets back and forth to give you the illusion that you're on a single system. The network is almost transparent here.

Another very important application in TCP/IP networks is NFS, which stands for Network File System. It is another form of making the network transparent, because it basically allows you to treat directory hierarchies from other hosts as if they were local file systems and look like any other directories on your host. For example, all users' home directories can be kept on a central server machine from which all other hosts on the LAN mount them. The effect is that users can log in to any machine and find themselves in the same home directory. Similarly, it is possible to share large amounts of data (such as a database, documentation or application programs) among many hosts by maintaining one copy of the data on a server and allowing other hosts to access it. We will come back to NFS in Chapter 14, The Network File System.

Of course, these are only examples of what you can do with TCP/IP networks. The possibilities are almost limitless, and we'll introduce you to more as you read on through the book.

We will now have a closer look at the way TCP/IP works. This information will help you understand how and why you have to configure your machine. We will start by examining the hardware, and slowly work our way up.


The most common type of LAN hardware is known as Ethernet. In its simplest form, it consists of a single cable with hosts attached to it through connectors, taps, or transceivers. Simple Ethernets are relatively inexpensive to install, which together with a net transfer rate of 10, 100, or even 1,000 Megabits per second, accounts for much of its popularity.

Ethernets come in three flavors: thick, thin, and twisted pair. Thin and thick Ethernet each use a coaxial cable, differing in diameter and the way you may attach a host to this cable. Thin Ethernet uses a T-shaped "BNC" connector, which you insert into the cable and twist onto a plug on the back of your computer. Thick Ethernet requires that you drill a small hole into the cable, and attach a transceiver using a "vampire tap." One or more hosts can then be connected to the transceiver. Thin and thick Ethernet cable can run for a maximum of 200 and 500 meters respectively, and are also called 10base-2 and 10base-5. The "base" refers to "baseband modulation" and simply means that the data is directly fed onto the cable without any modem. The number at the start refers to the speed in Megabits per second, and the number at the end is the maximum length of the cable in hundreds of metres. Twisted pair uses a cable made of two pairs of copper wires and usually requires additional hardware known as active hubs. Twisted pair is also known as 10base-T, the "T" meaning twisted pair. The 100 Megabits per second version is known as 100base-T.

To add a host to a thin Ethernet installation, you have to disrupt network service for at least a few minutes because you have to cut the cable to insert the connector. Although adding a host to a thick Ethernet system is a little complicated, it does not typically bring down the network. Twisted pair Ethernet is even simpler. It uses a device called a "hub," which serves as an interconnection point. You can insert and remove hosts from a hub without interrupting any other users at all.

Many people prefer thin Ethernet for small networks because it is very inexpensive; PC cards come for as little as US $30 (many companies are literally throwing them out now), and cable is in the range of a few cents per meter. However, for large-scale installations, either thick Ethernet or twisted pair is more appropriate. For example, the Ethernet at GMU's Mathematics Department originally chose thick Ethernet because it is a long route that the cable must take so traffic will not be disrupted each time a host is added to the network. Twisted pair installations are now very common in a variety of installations. The Hub hardware is dropping in price and small units are now available at a price that is attractive to even small domestic networks. Twisted pair cabling can be significantly cheaper for large installations, and the cable itself is much more flexible than the coaxial cables used for the other Ethernet systems. The network administrators in GMU's mathematics department are planning to replace the existing network with a twisted pair network in the coming finanical year because it will bring them up to date with current technology and will save them significant time when installing new host computers and moving existing computers around.

One of the drawbacks of Ethernet technology is its limited cable length, which precludes any use of it other than for LANs. However, several Ethernet segments can be linked to one another using repeaters, bridges, or routers. Repeaters simply copy the signals between two or more segments so that all segments together will act as if they are one Ethernet. Due to timing requirements, there may not be more than four repeaters between any two hosts on the network. Bridges and routers are more sophisticated. They analyze incoming data and forward it only when the recipient host is not on the local Ethernet.

Ethernet works like a bus system, where a host may send packets (or frames) of up to 1,500 bytes to another host on the same Ethernet. A host is addressed by a six-byte address hardcoded into the firmware of its Ethernet network interface card (NIC). These addresses are usually written as a sequence of two-digit hex numbers separated by colons, as in aa:bb:cc:dd:ee:ff.

A frame sent by one station is seen by all attached stations, but only the destination host actually picks it up and processes it. If two stations try to send at the same time, a collision occurs. Collisions on an Ethernet are detected very quickly by the electronics of the interface cards and are resolved by the two stations aborting the send, each waiting a random interval and re-attempting the transmission. You'll hear lots of stories about collisions on Ethernet being a problem and that utilization of Ethernets is only about 30 percent of the available bandwidth because of them. Collisions on Ethernet are a normal phenomenon, and on a very busy Ethernet network you shouldn't be surprised to see collision rates of up to about 30 percent. Utilization of Ethernet networks is more realistically limited to about 60 percent before you need to start worrying about it.[7]

Other Types of Hardware

In larger installations, such as Groucho Marx University, Ethernet is usually not the only type of equipment used. There are many other data communications protocols available and in use. All of the protocols listed are supported by Linux, but due to space constraints we'll describe them briefly. Many of the protocols have HOWTO documents that describe them in detail, so you should refer to those if you're interested in exploring those that we don't describe in this book.

At Groucho Marx University, each department's LAN is linked to the campus high-speed "backbone" network, which is a fiber optic cable running a network technology called Fiber Distributed Data Interface (FDDI). FDDI uses an entirely different approach to transmitting data, which basically involves sending around a number of tokens, with a station being allowed to send a frame only if it captures a token. The main advantage of a token-passing protocol is a reduction in collisions. Therefore, the protocol can more easily attain the full speed of the transmission medium, up to 100 Mbps in the case of FDDI. FDDI, being based on optical fiber, offers a significant advantage because its maximum cable length is much greater than wire-based technologies. It has limits of up to around 200 km, which makes it ideal for linking many buildings in a city, or as in GMU's case, many buildings on a campus.

Similarly, if there is any IBM computing equipment around, an IBM Token Ring network is quite likely to be installed. Token Ring is used as an alternative to Ethernet in some LAN environments, and offers the same sorts of advantages as FDDI in terms of achieving full wire speed, but at lower speeds (4 Mbps or 16 Mbps), and lower cost because it is based on wire rather than fiber. In Linux, Token Ring networking is configured in almost precisely the same way as Ethernet, so we don't cover it specifically.

Although it is much less likely today than in the past, other LAN technologies, such as ArcNet and DECNet, might be installed. Linux supports these too, but we don't cover them here.

Many national networks operated by Telecommunications companies support packet switching protocols. Probably the most popular of these is a standard named X.25. Many Public Data Networks, like Tymnet in the U.S., Austpac in Australia, and Datex-P in Germany offer this service. X.25 defines a set of networking protocols that describes how data terminal equipment, such as a host, communicates with data communications equipment (an X.25 switch). X.25 requires a synchronous data link, and therefore special synchronous serial port hardware. It is possible to use X.25 with normal serial ports if you use a special device called a PAD (Packet Assembler Disassembler). The PAD is a standalone device that provides asynchronous serial ports and a synchronous serial port. It manages the X.25 protocol so that simple terminal devices can make and accept X.25 connections. X.25 is often used to carry other network protocols, such as TCP/IP. Since IP datagrams cannot simply be mapped onto X.25 (or vice versa), they are encapsulated in X.25 packets and sent over the network. There is an experimental implementation of the X.25 protocol available for Linux.

A more recent protocol commonly offered by telecommunications companies is called Frame Relay. The Frame Relay protocol shares a number of technical features with the X.25 protocol, but is much more like the IP protocol in behavior. Like X.25, Frame Relay requires special synchronous serial hardware. Because of their similarities, many cards support both of these protocols. An alternative is available that requires no special internal hardware, again relying on an external device called a Frame Relay Access Device (FRAD) to manage the encapsulation of Ethernet packets into Frame Relay packets for transmission across a network. Frame Relay is ideal for carrying TCP/IP between sites. Linux provides drivers that support some types of internal Frame Relay devices.

If you need higher speed networking that can carry many different types of data, such as digitized voice and video, alongside your usual data, ATM (Asynchronous Transfer Mode) is probably what you'll be interested in. ATM is a new network technology that has been specifically designed to provide a manageable, high-speed, low-latency means of carrying data, and provide control over the Quality of Service (Q.S.). Many telecommunications companies are deploying ATM network infrastructure because it allows the convergence of a number of different network services into one platform, in the hope of achieving savings in management and support costs. ATM is often used to carry TCP/IP. The Networking-HOWTO offers information on the Linux support available for ATM.

Frequently, radio amateurs use their radio equipment to network their computers; this is commonly called packet radio. One of the protocols used by amateur radio operators is called AX.25 and is loosely derived from X.25. Amateur radio operators use the AX.25 protocol to carry TCP/IP and other protocols, too. AX.25, like X.25, requires serial hardware capable of synchronous operation, or an external device called a "Terminal Node Controller" to convert packets transmitted via an asynchronous serial link into packets transmitted synchronously. There are a variety of different sorts of interface cards available to support packet radio operation; these cards are generally referred to as being "Z8530 SCC based," and are named after the most popular type of communications controller used in the designs. Two of the other protocols that are commonly carried by AX.25 are the NetRom and Rose protocols, which are network layer protocols. Since these protocols run over AX.25, they have the same hardware requirements. Linux supports a fully featured implementation of the AX.25, NetRom, and Rose protocols. The AX25-HOWTO is a good source of information on the Linux implementation of these protocols.

Other types of Internet access involve dialing up a central system over slow but cheap serial lines (telephone, ISDN, and so on). These require yet another protocol for transmission of packets, such as SLIP or PPP, which will be described later.

The Internet Protocol

Of course, you wouldn't want your networking to be limited to one Ethernet or one point-to-point data link. Ideally, you would want to be able to communicate with a host computer regardless of what type of physical network it is connected to. For example, in larger installations such as Groucho Marx University, you usually have a number of separate networks that have to be connected in some way. At GMU, the Math department runs two Ethernets: one with fast machines for professors and graduates, and another with slow machines for students. Both are linked to the FDDI campus backbone network.

This connection is handled by a dedicated host called a gateway that handles incoming and outgoing packets by copying them between the two Ethernets and the FDDI fiber optic cable. For example, if you are at the Math department and want to access quark on the Physics department's LAN from your Linux box, the networking software will not send packets to quark directly because it is not on the same Ethernet. Therefore, it has to rely on the gateway to act as a forwarder. The gateway (named sophus) then forwards these packets to its peer gateway niels at the Physics department, using the backbone network, with niels delivering it to the destination machine. Data flow between erdos and quark is shown in Figure 1.1.

Figure 1.1: The three steps of sending a datagram from erdos to quark

This scheme of directing data to a remote host is called routing, and packets are often referred to as datagrams in this context. To facilitate things, datagram exchange is governed by a single protocol that is independent of the hardware used: IP, or Internet Protocol. In Chapter 2, Issues of TCP/IP Networking, we will cover IP and the issues of routing in greater detail.

The main benefit of IP is that it turns physically dissimilar networks into one apparently homogeneous network. This is called internetworking, and the resulting "meta-network" is called an internet. Note the subtle difference here between an internet and the Internet. The latter is the official name of one particular global internet.

Of course, IP also requires a hardware-independent addressing scheme. This is achieved by assigning each host a unique 32-bit number called the IP address. An IP address is usually written as four decimal numbers, one for each 8-bit portion, separated by dots. For example, quark might have an IP address of 0x954C0C04, which would be written as This format is also called dotted decimal notation and sometimes dotted quad notation. It is increasingly going under the name IPv4 (for Internet Protocol, Version 4) because a new standard called IPv6 offers much more flexible addressing, as well as other modern features. It will be at least a year after the release of this edition before IPv6 is in use.

You will notice that we now have three different types of addresses: first there is the host's name, like quark, then there are IP addresses, and finally, there are hardware addresses, like the 6-byte Ethernet address. All these addresses somehow have to match so that when you type rlogin quark, the networking software can be given quark's IP address; and when IP delivers any data to the Physics department's Ethernet, it somehow has to find out what Ethernet address corresponds to the IP address.

We will deal with these situations in Chapter 2. For now, it's enough to remember that these steps of finding addresses are called hostname resolution, for mapping hostnames onto IP addresses, and address resolution, for mapping the latter to hardware addresses.

IP Over Serial Lines

On serial lines, a "de facto" standard exists known as SLIP, or Serial Line IP. A modification of SLIP known as CSLIP, or Compressed SLIP, performs compression of IP headers to make better use of the relatively low bandwidth provided by most serial links. Another serial protocol is PPP, or the Point-to-Point Protocol. PPP is more modern than SLIP and includes a number of features that make it more attractive. Its main advantage over SLIP is that it isn't limited to transporting IP datagrams, but is designed to allow just about any protocol to be carried across it.

The Transmission Control Protocol

Sending datagrams from one host to another is not the whole story. If you log in to quark, you want to have a reliable connection between your rlogin process on erdos and the shell process on quark. Thus, the information sent to and fro must be split up into packets by the sender and reassembled into a character stream by the receiver. Trivial as it seems, this involves a number of complicated tasks.

A very important thing to know about IP is that, by intent, it is not reliable. Assume that ten people on your Ethernet started downloading the latest release of Netscape's web browser source code from GMU's FTP server. The amount of traffic generated might be too much for the gateway to handle, because it's too slow and it's tight on memory. Now if you happen to send a packet to quark, sophus might be out of buffer space for a moment and therefore unable to forward it. IP solves this problem by simply discarding it. The packet is irrevocably lost. It is therefore the responsibility of the communicating hosts to check the integrity and completeness of the data and retransmit it in case of error.

This process is performed by yet another protocol, Transmission Control Protocol (TCP), which builds a reliable service on top of IP. The essential property of TCP is that it uses IP to give you the illusion of a simple connection between the two processes on your host and the remote machine, so you don't have to care about how and along which route your data actually travels. A TCP connection works essentially like a two-way pipe that both processes may write to and read from. Think of it as a telephone conversation.

TCP identifies the end points of such a connection by the IP addresses of the two hosts involved and the number of a port on each host. Ports may be viewed as attachment points for network connections. If we are to strain the telephone example a little more, and you imagine that cities are like hosts, one might compare IP addresses to area codes (where numbers map to cities), and port numbers to local codes (where numbers map to individual people's telephones). An individual host may support many different services, each distinguished by its own port number.

In the rlogin example, the client application (rlogin) opens a port on erdos and connects to port 513 on quark, to which the rlogind server is known to listen. This action establishes a TCP connection. Using this connection, rlogind performs the authorization procedure and then spawns the shell. The shell's standard input and output are redirected to the TCP connection, so that anything you type to rlogin on your machine will be passed through the TCP stream and be given to the shell as standard input.

The User Datagram Protocol

Of course, TCP isn't the only user protocol in TCP/IP networking. Although suitable for applications like rlogin, the overhead involved is prohibitive for applications like NFS, which instead uses a sibling protocol of TCP called UDP, or User Datagram Protocol. Just like TCP, UDP allows an application to contact a service on a certain port of the remote machine, but it doesn't establish a connection for this. Instead, you use it to send single packets to the destination service - hence its name.

Assume you want to request a small amount of data from a database server. It takes at least three datagrams to establish a TCP connection, another three to send and confirm a small amount of data each way, and another three to close the connection. UDP provides us with a means of using only two datagrams to achieve almost the same result. UDP is said to be connectionless, and it doesn't require us to establish and close a session. We simply put our data into a datagram and send it to the server; the server formulates its reply, puts the data into a datagram addressed back to us, and transmits it back. While this is both faster and more efficient than TCP for simple transactions, UDP was not designed to deal with datagram loss. It is up to the application, a name server for example, to take care of this.

More on Ports

Ports may be viewed as attachment points for network connections. If an application wants to offer a certain service, it attaches itself to a port and waits for clients (this is also called listening on the port). A client who wants to use this service allocates a port on its local host and connects to the server's port on the remote host. The same port may be open on many different machines, but on each machine only one process can open a port at any one time.

An important property of ports is that once a connection has been established between the client and the server, another copy of the server may attach to the server port and listen for more clients. This property permits, for instance, several concurrent remote logins to the same host, all using the same port 513. TCP is able to tell these connections from one another because they all come from different ports or hosts. For example, if you log in twice to quark from erdos, the first rlogin client will use the local port 1023, and the second one will use port 1022. Both, however, will connect to the same port 513 on quark. The two connections will be distinguished by use of the port numbers used at erdos.

This example shows the use of ports as rendezvous points, where a client contacts a specific port to obtain a specific service. In order for a client to know the proper port number, an agreement has to be reached between the administrators of both systems on the assignment of these numbers. For services that are widely used, such as rlogin, these numbers have to be administered centrally. This is done by the IETF (Internet Engineering Task Force), which regularly releases an RFC titled Assigned Numbers (RFC-1700). It describes, among other things, the port numbers assigned to well-known services. Linux uses a file called /etc/services that maps service names to numbers.

It is worth noting that although both TCP and UDP connections rely on ports, these numbers do not conflict. This means that TCP port 513, for example, is different from UDP port 513. In fact, these ports serve as access points for two different services, namely rlogin (TCP) and rwho (UDP).

The Socket Library

In Unix operating systems, the software performing all the tasks and protocols described above is usually part of the kernel, and so it is in Linux. The programming interface most common in the Unix world is the Berkeley Socket Library. Its name derives from a popular analogy that views ports as sockets and connecting to a port as plugging in. It provides the bind call to specify a remote host, a transport protocol, and a service that a program can connect or listen to (using connect, listen, and accept). The socket library is somewhat more general in that it provides not only a class of TCP/IP-based sockets (the AF_INET sockets), but also a class that handles connections local to the machine (the AF_UNIX class). Some implementations can also handle other classes, like the XNS (Xerox Networking System) protocol or X.25.

In Linux, the socket library is part of the standard libc C library. It supports the AF_INET and AF_INET6 sockets for TCP/IP and AF_UNIX for Unix domain sockets. It also supports AF_IPX for Novell's network protocols, AF_X25 for the X.25 network protocol, AF_ATMPVC and AF_ATMSVC for the ATM network protocol and AF_AX25, AF_NETROM, and AF_ROSE sockets for Amateur Radio protocol support. Other protocol families are being developed and will be added in time.

UUCP Networks

Unix-to-Unix Copy (UUCP) started out as a package of programs that transferred files over serial lines, scheduled those transfers, and initiated execution of programs on remote sites. It has undergone major changes since its first implementation in the late seventies, but it is still rather spartan in the services it offers. Its main application is still in Wide Area Networks, based on periodic dialup telephone links.

UUCP was first developed by Bell Laboratories in 1977 for communication between their Unix development sites. In mid-1978, this network already connected over 80 sites. It was running email as an application, as well as remote printing. However, the system's central use was in distributing new software and bug fixes. Today, UUCP is not confined solely to the Unix environment. There are free and commercial ports available for a variety of platforms, including AmigaOS, DOS, and Atari's TOS.

One of the main disadvantages of UUCP networks is that they operate in batches. Rather than having a permanent connection established between hosts, it uses temporary connections. A UUCP host machine might dial in to another UUCP host only once a day, and then only for a short period of time. While it is connected, it will transfer all of the news, email, and files that have been queued, and then disconnect. It is this queuing that limits the sorts of applications that UUCP can be applied to. In the case of email, a user may prepare an email message and post it. The message will stay queued on the UUCP host machine until it dials in to another UUCP host to transfer the message. This is fine for network services such as email, but is no use at all for services such as rlogin.

Despite these limitations, there are still many UUCP networks operating all over the world, run mainly by hobbyists, which offer private users network access at reasonable prices. The main reason for the longtime popularity of UUCP was that it was very cheap compared to having your computer directly connected to the Internet. To make your computer a UUCP node, all you needed was a modem, a working UUCP implementation, and another UUCP node that was willing to feed you mail and news. Many people were prepared to provide UUCP feeds to individuals because such connections didn't place much demand on their existing network.

We cover the configuration of UUCP in a chapter of its own later in the book, but we won't focus on it too heavily, as it's being replaced rapidly with TCP/IP, now that cheap Internet access has become commonly available in most parts of the world.

Linux Networking

As it is the result of a concerted effort of programmers around the world, Linux wouldn't have been possible without the global network. So it's not surprising that in the early stages of development, several people started to work on providing it with network capabilities. A UUCP implementation was running on Linux almost from the very beginning, and work on TCP/IP-based networking started around autumn 1992, when Ross Biro and others created what has now become known as Net-1.

After Ross quit active development in May 1993, Fred van Kempen began to work on a new implementation, rewriting major parts of the code. This project was known as Net-2. The first public release, Net-2d, was made in the summer of 1993 (as part of the 0.99.10 kernel), and has since been maintained and expanded by several people, most notably Alan Cox.[8] Alan's original work was known as Net-2Debugged. After heavy debugging and numerous improvements to the code, he changed its name to Net-3 after Linux 1.0 was released. The Net-3 code was further developed for Linux 1.2 and Linux 2.0. The 2.2 and later kernels use the Net-4 version network support, which remains the standard official offering today. 

The Net-4 Linux Network code offers a wide variety of device drivers and advanced features. Standard Net-4 protocols include SLIP and PPP (for sending network traffic over serial lines), PLIP (for parallel lines), IPX (for Novell compatible networks, which we'll discuss in Chapter 15, IPX and the NCP Filesystem), Appletalk (for Apple networks) and AX.25, NetRom, and Rose (for amateur radio networks). Other standard Net-4 features include IP firewalling, IP accounting (discussed later in Chapter 9, TCP/IP Firewall and Chapter 10, IP Accounting), and IP Masquerade (discussed later in Chapter 11, IP Masquerade and Network Address Translation. IP tunnelling in a couple of different flavors and advanced policy routing are supported. A very large variety of Ethernet devices is supported, in addition to support for some FDDI, Token Ring, Frame Relay, and ISDN, and ATM cards.

Additionally, there are a number of other features that greatly enhance the flexibility of Linux. These features include an implementation of the SMB filesystem, which interoperates with applications like lanmanager and Microsoft Windows, called Samba, written by Andrew Tridgell, and an implementation of the Novell NCP (NetWare Core Protocol).[9]

Different Streaks of Development

There have been, at various times, varying network development efforts active for Linux.

Fred continued development after Net-2Debugged was made the official network implementation. This development led to the Net-2e, which featured a much revised design of the networking layer. Fred was working toward a standardized Device Driver Interface (DDI), but the Net-2e work has ended now.

Yet another implementation of TCP/IP networking came from Matthias Urlichs, who wrote an ISDN driver for Linux and FreeBSD. For this driver, he integrated some of the BSD networking code in the Linux kernel. That project, too is no longer being worked on.

There has been a lot of rapid change in the Linux kernel networking implementation, and change is still the watchword as development continues. Sometimes this means that changes also have to occur in other software, such as the network configuration tools. While this is no longer as large a problem as it once was, you may still find that upgrading your kernel to a later version means that you must upgrade your network configuration tools, too. Fortunately, with the large number of Linux distributions available today, this is a quite simple task.

The Net-4 network implementation is now quite mature and is in use at a very large number of sites around the world. Much work has been done on improving the performance of the Net-4 implementation, and it now competes with the best implementations available for the same hardware platforms. Linux is proliferating in the Internet Service Provider environment, and is often used to build cheap and reliable World Wide Web servers, mail servers, and news servers for these sorts of organizations. There is now sufficient development interest in Linux that it is managing to keep abreast of networking technology as it changes, and current releases of the Linux kernel offer the next generation of the IP protocol, IPv6, as a standard offering.

Where to Get the Code

It seems odd now to remember that in the early days of the Linux network code development, the standard kernel required a huge patch kit to add the networking support to it. Today, network development occurs as part of the mainstream Linux kernel development process. The latest stable Linux kernels can be found on ftp.kernel.org in /pub/linux/kernel/v2.x/, where x is an even number. The latest experimental Linux kernels can be found on ftp.kernel.org in /pub/linux/kernel/v2.y/, where y is an odd number. There are Linux kernel source mirrors all over the world. It is now hard to imagine Linux without standard network support.

Maintaining Your System

Throughout this book, we will mainly deal with installation and configuration issues. Administration is, however, much more than that - after setting up a service, you have to keep it running, too. For most services, only a little attendance will be necessary, while some, like mail and news, require that you perform routine tasks to keep your system up to date. We will discuss these tasks in later chapters.

The absolute minimum in maintenance is to check system and per-application log files regularly for error conditions and unusual events. Often, you will want to do this by writing a couple of administrative shell scripts and periodically running them from cron. The source distributions of some major applications, like inn or C News, contain such scripts. You only have to tailor them to suit your needs and preferences.

The output from any of your cron jobs should be mailed to an administrative account. By default, many applications will send error reports, usage statistics, or log file summaries to the root account. This makes sense only if you log in as root frequently; a much better idea is to forward root's mail to your personal account by setting up a mail alias as described in Chapter 19, Getting Exim Up and Running or Chapter 18, Sendmail.

However carefully you have configured your site, Murphy's law guarantees that some problem will surface eventually. Therefore, maintaining a system also means being available for complaints. Usually, people expect that the system administrator can at least be reached via email as root, but there are also other addresses that are commonly used to reach the person responsible for a specific aspect of maintenence. For instance, complaints about a malfunctioning mail configuration will usually be addressed to postmaster, and problems with the news system may be reported to newsmaster or usenet. Mail to hostmaster should be redirected to the person in charge of the host's basic network services, and the DNS name service if you run a name server.

System Security

Another very important aspect of system administration in a network environment is protecting your system and users from intruders. Carelessly managed systems offer malicious people many targets. Attacks range from password guessing to Ethernet snooping, and the damage caused may range from faked mail messages to data loss or violation of your users' privacy. We will mention some particular problems when discussing the context in which they may occur and some common defenses against them.

This section will discuss a few examples and basic techniques for dealing with system security. Of course, the topics covered cannot treat all security issues you may be faced with in detail; they merely serve to illustrate the problems that may arise. Therefore, reading a good book on security is an absolute must, especially in a networked system.

System security starts with good system administration. This includes checking the ownership and permissions of all vital files and directories and monitoring use of privileged accounts. The COPS program, for instance, will check your file system and common configuration files for unusual permissions or other anomalies. It is also wise to use a password suite that enforces certain rules on the users' passwords that make them hard to guess. The shadow password suite, for instance, requires a password to have at least five letters and to contain both upper- and lowercase numbers, as well as non-alphabetic characters.

When making a service accessible to the network, make sure to give it "least privilege"; don't permit it to do things that aren't required for it to work as designed. For example, you should make programs setuid to root or some other privileged account only when necessary. Also, if you want to use a service for only a very limited application, don't hesitate to configure it as restrictively as your special application allows. For instance, if you want to allow diskless hosts to boot from your machine, you must provide Trivial File Transfer Protocol (TFTP) so that they can download basic configuration files from the /boot directory. However, when used unrestrictively, TFTP allows users anywhere in the world to download any world-readable file from your system. If this is not what you want, restrict TFTP service to the /boot directory.[10]

You might also want to restrict certain services to users from certain hosts, say from your local network. In Chapter 12, we introduce tcpd, which does this for a variety of network applications. More sophisticated methods of restricting access to particular hosts or services will be explored later in Chapter 9.

Another important point is to avoid "dangerous" software. Of course, any software you use can be dangerous because software may have bugs that clever people might exploit to gain access to your system. Things like this happen, and there's no complete protection against it. This problem affects free software and commercial products alike.[11] However, programs that require special privilege are inherently more dangerous than others, because any loophole can have drastic consequences.[12] If you install a setuid program for network purposes, be doubly careful to check the documentation so that you don't create a security breach by accident.

Another source of concern should be programs that enable login or command execution with limited authentication. The rlogin, rsh, and rexec commands are all very useful, but offer very limited authentication of the calling party. Authentication is based on trust of the calling host name obtained from a name server (we'll talk about these later), which can be faked. Today it should be standard practice to disable the r commands completely and replace them with the ssh suite of tools. The ssh tools use a much more reliable authentication method and provide other services, such as encryption and compression, as well.

You can never rule out the possibility that your precautions might fail, regardless of how careful you have been. You should therefore make sure you detect intruders early. Checking the system log files is a good starting point, but the intruder is probably clever enough to anticipate this action and will delete any obvious traces he or she left. However, there are tools like tripwire, written by Gene Kim and Gene Spafford, that allow you to check vital system files to see if their contents or permissions have been changed. tripwire computes various strong checksums over these files and stores them in a database. During subsequent runs, the checksums are recomputed and compared to the stored ones to detect any modifications.

Chapter 2. Issues of TCP/IP Networking

In this chapter we turn to the configuration decisions you'll need to make when connecting your Linux machine to a TCP/IP network, including dealing with IP addresses, hostnames, and routing issues. This chapter gives you the background you need in order to understand what your setup requires, while the next chapters cover the tools you will use.

To learn more about TCP/IP and the reasons behind it, refer to the three-volume set Internetworking with TCP/IP, by Douglas R. Comer (Prentice Hall). For a more detailed guide to managing a TCP/IP network, see TCP/IP Network Administration by Craig Hunt (O'Reilly).

Networking Interfaces

To hide the diversity of equipment that may be used in a networking environment, TCP/IP defines an abstract interface through which the hardware is accessed. This interface offers a set of operations that is the same for all types of hardware and basically deals with sending and receiving packets.

For each peripheral networking device, a corresponding interface has to be present in the kernel. For example, Ethernet interfaces in Linux are called by such names as eth0 and eth1; PPP (discussed in Chapter 8, The Point-to-Point Protocol) interfaces are named ppp0 and ppp1; and FDDI interfaces are given names like fddi0 and fddi1. These interface names are used for configuration purposes when you want to specify a particular physical device in a configuration command, and they have no meaning beyond this use.

Before being used by TCP/IP networking, an interface must be assigned an IP address that serves as its identification when communicating with the rest of the world. This address is different from the interface name mentioned previously; if you compare an interface to a door, the address is like the nameplate pinned on it.

Other device parameters may be set, like the maximum size of datagrams that can be processed by a particular piece of hardware, which is referred to as Maximum Transfer Unit (MTU). Other attributes will be introduced later. Fortunately, most attributes have sensible defaults.

IP Addresses

As mentioned in Chapter 1, Introduction to Networking, the IP networking protocol understands addresses as 32-bit numbers. Each machine must be assigned a number unique to the networking environment.[13] If you are running a local network that does not have TCP/IP traffic with other networks, you may assign these numbers according to your personal preferences. There are some IP address ranges that have been reserved for such private networks. These ranges are listed in Table 2.1. However, for sites on the Internet, numbers are assigned by a central authority, the Network Information Center (NIC).[14]

IP addresses are split up into four eight-bit numbers called octets for readability. For example, quark.physics.groucho.edu has an IP address of 0x954C0C04, which is written as This format is often referred to as dotted quad notation.

Another reason for this notation is that IP addresses are split into a network number, which is contained in the leading octets, and a host number, which is the remainder. When applying to the NIC for IP addresses, you are not assigned an address for each single host you plan to use. Instead, you are given a network number and allowed to assign all valid IP addresses within this range to hosts on your network according to your preferences.

The size of the host part depends on the size of the network. To accommodate different needs, several classes of networks, defining different places to split IP addresses, have been defined. The class networks are described here:

Class A

Class A comprises networks through The network number is contained in the first octet. This class provides for a 24-bit host part, allowing roughly 1.6 million hosts per network.

Class B

Class B contains networks through; the network number is in the first two octets. This class allows for 16,320 nets with 65,024 hosts each.

Class C

Class C networks range from through, with the network number contained in the first three octets. This class allows for nearly 2 million networks with up to 254 hosts.

Classes D, E, and F

Addresses falling into the range of through are either experimental or are reserved for special purpose use and don't specify any network. IP Multicast, which is a service that allows material to be transmitted to many points on an internet at one time, has been assigned addresses from within this range.

If we go back to the example in Chapter 1, we find that, the address of quark, refers to host 12.4 on the class B network

You may have noticed that not all possible values in the previous list were allowed for each octet in the host part. This is because octets 0 and 255 are reserved for special purposes. An address where all host part bits are 0 refers to the network, and an address where all bits of the host part are 1 is called a broadcast address. This refers to all hosts on the specified network simultaneously. Thus, is not a valid host address, but refers to all hosts on network

A number of network addresses are reserved for special purposes. and are two such addresses. The first is called the default route, and the latter is the loopback address. The default route has to do with the way the IP routes datagrams.

Network is reserved for IP traffic local to your host. Usually, address will be assigned to a special interface on your host, the loopback interface, which acts like a closed circuit. Any IP packet handed to this interface from TCP or UDP will be returned to them as if it had just arrived from some network. This allows you to develop and test networking software without ever using a "real" network. The loopback network also allows you to use networking software on a standalone host. This may not be as uncommon as it sounds; for instance, many UUCP sites don't have IP connectivity at all, but still want to run the INN news system. For proper operation on Linux, INN requires the loopback interface.

Some address ranges from each of the network classes have been set aside and designated "reserved" or "private" address ranges. These addresses are reserved for use by private networks and are not routed on the Internet. They are commonly used by organizations building their own intranet, but even small networks often find them useful. The reserved network addresses appear in Table 2.1.

Table 2.1: IP Address Ranges Reserved for Private Use

Class Networks
A through
B through
C through

Address Resolution

Now that you've seen how IP addresses are composed, you may be wondering how they are used on an Ethernet or Token Ring network to address different hosts. After all, these protocols have their own addresses to identify hosts that have absolutely nothing in common with an IP address, don't they? Right.

A mechanism is needed to map IP addresses onto the addresses of the underlying network. The mechanism used is the Address Resolution Protocol (ARP). In fact, ARP is not confined to Ethernet or Token Ring, but is used on other types of networks, such as the amateur radio AX.25 protocol. The idea underlying ARP is exactly what most people do when they have to find Mr. X in a throng of 150 people: the person who wants him calls out loudly enough that everyone in the room can hear them, expecting him to respond if he is there. When he responds, we know which person he is.

When ARP wants to find the Ethernet address corresponding to a given IP address, it uses an Ethernet feature called broadcasting, in which a datagram is addressed to all stations on the network simultaneously. The broadcast datagram sent by ARP contains a query for the IP address. Each receiving host compares this query to its own IP address and if it matches, returns an ARP reply to the inquiring host. The inquiring host can now extract the sender's Ethernet address from the reply.

You may wonder how a host can reach an Internet address that may be on a different network halfway around the world. The answer to this question involves routing, namely finding the physical location of a host in a network. We will discuss this issue further in the next section.

Let's talk a little more about ARP. Once a host has discovered an Ethernet address, it stores it in its ARP cache so that it doesn't have to query for it again the next time it wants to send a datagram to the host in question. However, it is unwise to keep this information forever; the remote host's Ethernet card may be replaced because of technical problems, so the ARP entry becomes invalid. Therefore, entries in the ARP cache are discarded after some time to force another query for the IP address.

Sometimes it is also necessary to find the IP address associated with a given Ethernet address. This happens when a diskless machine wants to boot from a server on the network, which is a common situation on Local Area Networks. A diskless client, however, has virtually no information about itself - except for its Ethernet address! So it broadcasts a message containing a request asking a boot server to provide it with an IP address. There's another protocol for this situation named Reverse Address Resolution Protocol (RARP). Along with the BOOTP protocol, it serves to define a procedure for bootstrapping diskless clients over the network.

IP Routing

We now take up the question of finding the host that datagrams go to based on the IP address. Different parts of the address are handled in different ways; it is your job to set up the files that indicate how to treat each part.

IP Networks

When you write a letter to someone, you usually put a complete address on the envelope specifying the country, state, and Zip Code. After you put it in the mailbox, the post office will deliver it to its destination: it will be sent to the country indicated, where the national service will dispatch it to the proper state and region. The advantage of this hierarchical scheme is obvious: wherever you post the letter, the local postmaster knows roughly which direction to forward the letter, but the postmaster doesn't care which way the letter will travel once it reaches its country of destination.

IP networks are structured similarly. The whole Internet consists of a number of proper networks, called autonomous systems. Each system performs routing between its member hosts internally so that the task of delivering a datagram is reduced to finding a path to the destination host's network. As soon as the datagram is handed to any host on that particular network, further processing is done exclusively by the network itself.


This structure is reflected by splitting IP addresses into a host and network part, as explained previously. By default, the destination network is derived from the network part of the IP address. Thus, hosts with identical IP network numbers should be found within the same network.[15]

It makes sense to offer a similar scheme inside the network, too, since it may consist of a collection of hundreds of smaller networks, with the smallest units being physical networks like Ethernets. Therefore, IP allows you to subdivide an IP network into several subnets.

A subnet takes responsibility for delivering datagrams to a certain range of IP addresses. It is an extension of the concept of splitting bit fields, as in the A, B, and C classes. However, the network part is now extended to include some bits from the host part. The number of bits that are interpreted as the subnet number is given by the so-called subnet mask, or netmask. This is a 32-bit number too, which specifies the bit mask for the network part of the IP address.

The campus network of Groucho Marx University is an example of such a network. It has a class B network number of, and its netmask is therefore

Internally, GMU's campus network consists of several smaller networks, such various departments' LANs. So the range of IP addresses is broken up into 254 subnets, through For example, the department of Theoretical Physics has been assigned The campus backbone is a network in its own right, and is given These subnets share the same IP network number, while the third octet is used to distinguish between them. They will thus use a subnet mask of

Figure 2.1 shows how, the address of quark, is interpreted differently when the address is taken as an ordinary class B network and when used with subnetting.

Figure 2.1: Subnetting a class B network

It is worth noting that subnetting (the technique of generating subnets) is only an internal division of the network. Subnets are generated by the network owner (or the administrators). Frequently, subnets are created to reflect existing boundaries, be they physical (between two Ethernets), administrative (between two departments), or geographical (between two locations), and authority over each subnet is delegated to some contact person. However, this structure affects only the network's internal behavior, and is completely invisible to the outside world.


Subnetting is not only a benefit to the organization; it is frequently a natural consequence of hardware boundaries. The viewpoint of a host on a given physical network, such as an Ethernet, is a very limited one: it can only talk to the host of the network it is on. All other hosts can be accessed only through special-purpose machines called gateways. A gateway is a host that is connected to two or more physical networks simultaneously and is configured to switch packets between them.

Figure 2.2 shows part of the network topology at Groucho Marx University (GMU). Hosts that are on two subnets at the same time are shown with both addresses.

Figure 2.2: A part of the net topology at Groucho Marx University

Different physical networks have to belong to different IP networks for IP to be able to recognize if a host is on a local network. For example, the network number is reserved for hosts on the mathematics LAN. When sending a datagram to quark, the network software on erdos immediately sees from the IP address that the destination host is on a different physical network, and therefore can be reached only through a gateway (sophus by default).

sophus itself is connected to two distinct subnets: the Mathematics department and the campus backbone. It accesses each through a different interface, eth0 and fddi0, respectively. Now, what IP address do we assign it? Should we give it one on subnet, or on

The answer is: "both." sophus has been assigned the address for use on the network and address for use on the network. A gateway must be assigned one IP address for each network it belongs to. These addresses - along with the corresponding netmask - are tied to the interface through which the subnet is accessed. Thus, the interface and address mapping for sophus would look like this:

Interface Address Netmask

The last entry describes the loopback interface lo, which we talked about earlier.

Generally, you can ignore the subtle difference between attaching an address to a host or its interface. For hosts that are on one network only, like erdos, you would generally refer to the host as having this-and-that IP address, although strictly speaking, it's the Ethernet interface that has this IP address. The distinction is really important only when you refer to a gateway.

The Routing Table

We now focus our attention on how IP chooses a gateway to use to deliver a datagram to a remote network.

We have seen that erdos, when given a datagram for quark, checks the destination address and finds that it is not on the local network. erdos therefore sends the datagram to the default gateway sophus, which is now faced with the same task. sophus recognizes that quark is not on any of the networks it is connected to directly, so it has to find yet another gateway to forward it through. The correct choice would be niels, the gateway to the Physics department. sophus thus needs information to associate a destination network with a suitable gateway.

IP uses a table for this task that associates networks with the gateways by which they may be reached. A catch-all entry (the default route) must generally be supplied too; this is the gateway associated with network All destination addresses match this route, since none of the 32 bits are required to match, and therefore packets to an unknown network are sent through the default route. On sophus, the table might look like this:

Network Netmask Gateway Interface - fddi0 fddi0 fddi0 - eth0 fddi0 fddi0

If you need to use a route to a network that sophus is directly connected to, you don't need a gateway; the gateway column here contains a hyphen.

The process for identifying whether a particular destination address matches a route is a mathematical operation. The process is quite simple, but it requires an understanding of binary arithmetic and logic: A route matches a destination if the network address logically ANDed with the netmask precisely equals the destination address logically ANDed with the netmask.

Translation: a route matches if the number of bits of the network address specified by the netmask (starting from the left-most bit, the high order bit of byte one of the address) match that same number of bits in the destination address.

When the IP implementation is searching for the best route to a destination, it may find a number of routing entries that match the target address. For example, we know that the default route matches every destination, but datagrams destined for locally attached networks will match their local route, too. How does IP know which route to use? It is here that the netmask plays an important role. While both routes match the destination, one of the routes has a larger netmask than the other. We previously mentioned that the netmask was used to break up our address space into smaller networks. The larger a netmask is, the more specifically a target address is matched; when routing datagrams, we should always choose the route that has the largest netmask. The default route has a netmask of zero bits, and in the configuration presented above, the locally attached networks have a 24-bit netmask. If a datagram matches a locally attached network, it will be routed to the appropriate device in preference to following the default route because the local network route matches with a greater number of bits. The only datagrams that will be routed via the default route are those that don't match any other route.

You can build routing tables by a variety of means. For small LANs, it is usually most efficient to construct them by hand and feed them to IP using the route command at boot time (see Chapter 5, Configuring TCP/IP Networking). For larger networks, they are built and adjusted at runtime by routing daemons; these daemons run on central hosts of the network and exchange routing information to compute "optimal" routes between the member networks.

Depending on the size of the network, you'll need to use different routing protocols. For routing inside autonomous systems (such as the Groucho Marx campus), the internal routing protocols are used. The most prominent one of these is the Routing Information Protocol (RIP), which is implemented by the BSD routed daemon. For routing between autonomous systems, external routing protocols like External Gateway Protocol (EGP) or Border Gateway Protocol (BGP) have to be used; these protocols, including RIP, have been implemented in the University of Cornell's gated daemon.

Metric Values

We depend on dynamic routing to choose the best route to a destination host or network based on the number of hops. Hops are the gateways a datagram has to pass before reaching a host or network. The shorter a route is, the better RIP rates it. Very long routes with 16 or more hops are regarded as unusable and are discarded.

RIP manages routing information internal to your local network, but you have to run gated on all hosts. At boot time, gated checks for all active network interfaces. If there is more than one active interface (not counting the loopback interface), it assumes the host is switching packets between several networks and will actively exchange and broadcast routing information. Otherwise, it will only passively receive RIP updates and update the local routing table.

When broadcasting information from the local routing table, gated computes the length of the route from the so-called metric value associated with the routing table entry. This metric value is set by the system administrator when configuring the route, and should reflect the actual route cost.[16] Therefore, the metric of a route to a subnet that the host is directly connected to should always be zero, while a route going through two gateways should have a metric of two. You don't have to bother with metrics if you don't use RIP or gated.

The Internet Control Message Protocol

IP has a companion protocol that we haven't talked about yet. This is the Internet Control Message Protocol (ICMP), used by the kernel networking code to communicate error messages to other hosts. For instance, assume that you are on erdos again and want to telnet to port 12345 on quark, but there's no process listening on that port. When the first TCP packet for this port arrives on quark, the networking layer will recognize this arrival and immediately return an ICMP message to erdos stating "Port Unreachable."

The ICMP protocol provides several different messages, many of which deal with error conditions. However, there is one very interesting message called the Redirect message. It is generated by the routing module when it detects that another host is using it as a gateway, even though a much shorter route exists. For example, after booting, the routing table of sophus may be incomplete. It might contain the routes to the Mathematics network, to the FDDI backbone, and the default route pointing at the Groucho Computing Center's gateway (gcc1). Thus, packets for quark would be sent to gcc1 rather than to niels, the gateway to the Physics department. When receiving such a datagram, gcc1 will notice that this is a poor choice of route and will forward the packet to niels, meanwhile returning an ICMP Redirect message to sophus telling it of the superior route.

This seems to be a very clever way to avoid manually setting up any but the most basic routes. However, be warned that relying on dynamic routing schemes, be it RIP or ICMP Redirect messages, is not always a good idea. ICMP Redirect and RIP offer you little or no choice in verifying that some routing information is indeed authentic. This situation allows malicious good-for-nothings to disrupt your entire network traffic, or even worse. Consequently, the Linux networking code treats Network Redirect messages as if they were Host Redirects. This minimizes the damage of an attack by restricting it to just one host, rather than the whole network. On the flip side, it means that a little more traffic is generated in the event of a legitimate condition, as each host causes the generation of an ICMP Redirect message. It is generally considered bad practice to rely on ICMP redirects for anything these days.

Resolving Host Names

As described previously, addressing in TCP/IP networking, at least for IP Version 4, revolves around 32-bit numbers. However, you will have a hard time remembering more than a few of these numbers. Therefore, hosts are generally known by "ordinary" names such as gauss or strange. It becomes the application's duty to find the IP address corresponding to this name. This process is called hostname resolution.

When an application needs to find the IP address of a given host, it relies on the library functions gethostbyname(3) and gethostbyaddr(3). Traditionally, these and a number of related procedures were grouped in a separate library called the resolverlibrary; on Linux, these functions are part of the standard libc. Colloquially, this collection of functions is therefore referred to as "the resolver." Resolver name configuration is detailed in Chapter 6, Name Service and Resolver Configuration.

On a small network like an Ethernet or even a cluster of Ethernets, it is not very difficult to maintain tables mapping hostnames to addresses. This information is usually kept in a file named /etc/hosts. When adding or removing hosts, or reassigning addresses, all you have to do is update the hosts file on all hosts. Obviously, this will become burdensome with networks that comprise more than a handful of machines.

One solution to this problem is the Network Information System (NIS), developed by Sun Microsystems, colloquially called YP or Yellow Pages. NIS stores the hosts file (and other information) in a database on a master host from which clients may retrieve it as needed. Still, this approach is suitable only for medium-sized networks such as LANs, because it involves maintaining the entire hosts database centrally and distributing it to all servers. NIS installation and configuration is discussed in detail in Chapter 13, The Network Information System.

On the Internet, address information was initially stored in a single HOSTS.TXT database, too. This file was maintained at the Network Information Center (NIC), and had to be downloaded and installed by all participating sites. When the network grew, several problems with this scheme arose. Besides the administrative overhead involved in installing HOSTS.TXT regularly, the load on the servers that distributed it became too high. Even more severe, all names had to be registered with the NIC, which made sure that no name was issued twice.

This is why a new name resolution scheme was adopted in 1994: the Domain Name System. DNS was designed by Paul Mockapetris and addresses both problems simultaneously. We discuss the Domain Name System in detail in Chapter 6.

Chapter 3. Configuring the Networking Hardware

We've been talking quite a bit about network interfaces and general TCP/IP issues, but we haven't really covered what happens when the "networking code" in the kernel accesses a piece of hardware. In order to describe this accurately, we have to talk a little about the concept of interfaces and drivers.

First, of course, there's the hardware itself, for example an Ethernet, FDDI or Token Ring card: this is a slice of Epoxy cluttered with lots of tiny chips with strange numbers on them, sitting in a slot of your PC. This is what we generally call a physical device.

For you to use a network card, special functions have to be present in your Linux kernel that understand the particular way this device is accessed. The software that implements these functions is called a device driver. Linux has device drivers for many different types of network interface cards: ISA, PCI, MCA, EISA, Parallel port, PCMCIA, and more recently, USB.

But what do we mean when we say a driver "handles" a device? Let's consider an Ethernet card. The driver has to be able to communicate with the peripheral's on-card logic somehow: it has to send commands and data to the card, while the card should deliver any data received to the driver.

In IBM-style personal computers, this communication takes place through a cluster of I/O addresses that are mapped to registers on the card and/or through shared or direct memory transfers. All commands and data the kernel sends to the card have to go to these addresses. I/O and memory addresses are generally described by providing the starting or base address. Typical base addresses for ISA bus Ethernet cards are 0x280 or 0x300. PCI bus network cards generally have their I/O address automatically assigned.

Usually you don't have to worry about any hardware issues such as the base address because the kernel makes an attempt at boot time to detect a card's location. This is called auto probing, which means that the kernel reads several memory or I/O locations and compares the data it reads there with what it would expect to see if a certain network card were installed at that location. However, there may be network cards it cannot detect automatically; this is sometimes the case with cheap network cards that are not-quite clones of standard cards from other manufacturers. Also, the kernel will normally attempt to detect only one network device when booting. If you're using more than one card, you have to tell the kernel about the other cards explicitly.

Another parameter that you might have to tell the kernel about is the interrupt request line. Hardware components usually interrupt the kernel when they need to be taken care of - for example, when data has arrived or a special condition occurs. In an ISA bus PC, interrupts may occur on one of 15 interrupt channels numbered 0, 1, and 3 through 15. The interrupt number assigned to a hardware component is called its interrupt request number (IRQ).[17]

As described in Chapter 2, Issues of TCP/IP Networking, the kernel accesses a piece of network hardware through a software construct called an interface. Interfaces offer an abstract set of functions that are the same across all types of hardware, such as sending or receiving a datagram.

Interfaces are identified by means of names. In many other Unix-like operating systems, the network interface is implemented as a special device file in the /dev/ directory. If you type the ls -las /dev/ command, you will see what these device files look like. In the file permissions (second) column you will see that device files begin with a letter rather than the hyphen seen for normal files. This character indicates the device type. The most common device types are b, which indicates the device is a block device and handles whole blocks of data with each read and write, and c, which indicates the device is a character device and handles data one character at a time. Where you would normally see the file length in the ls output, you instead see two numbers, called the major and minor device numbers. These numbers indicate the actual device with which the device file is associated.

Each device driver registers a unique major number with the kernel. Each instance of that device registers a unique minor number for that major device. The tty interfaces,/dev/tty*, are a character mode device indicated by the " c ", and each have a major number of 4, but /dev/tty1 has a minor number of 1, and /dev/tty2 has a minor number of 2. Device files are very useful for many types of devices, but can be clumsy to use when trying to find an unused device to open.

Linux interface names are defined internally in the kernel and are not device files in the /dev directory. Some typical device names are listed later in "A Tour of Linux Network Devices"." The assignment of interfaces to devices usually depends on the order in which devices are configured. For instance, the first Ethernet card installed will become eth0, and the next will be eth1. SLIP interfaces are handled differently from others because they are assigned dynamically. Whenever a SLIP connection is established, an interface is assigned to the serial port.

Figure 3.1 illustrates the relationship between the hardware, device drivers, and interfaces.

Figure 3.1: The relationship between drivers, interfaces, and hardware

When booting, the kernel displays the devices it detects and the interfaces it installs. The following is an excerpt from typical boot messages:


. This processor honors the WP bit even when in supervisor mode./


Swansea University Computer Society NET3.035 for Linux 2.0

NET3: Unix domain sockets 0.13 for Linux NET3.035.

Swansea University Computer Society TCP/IP for NET3.034


Swansea University Computer Society IPX 0.34 for NET3.035

IPX Portions Copyright (c) 1995 Caldera, Inc.

Serial driver version 4.13 with no serial options enabled

tty00 at 0x03f8 (irq = 4) is a 16550A

tty01 at 0x02f8 (irq = 3) is a 16550A

CSLIP: code copyright 1989 Regents of the University of California

PPP: Version 2.2.0 (dynamic channel allocation)

PPP Dynamic channel allocation code copyright 1995 Caldera, Inc.

PPP line discipline registered.

eth0: 3c509 at 0x300 tag 1, 10baseT port, address 00 a0 24 0e e4 e0,/

    IRQ 10.

3c509.c:1.12 6/4/97 becker@cesdis.gsfc.nasa.gov

Linux Version 2.0.32 (root@perf) (gcc Version

#1 Tue Oct 21 15:30:44 EST 1997



This example shows that the kernel has been compiled with TCP/IP enabled, and it includes drivers for SLIP, CSLIP, and PPP. The third line from the bottom says that a 3C509 Ethernet card was detected and installed as interface eth0. If you have some other type of network card - perhaps a D-Link pocket adaptor, for example - the kernel will usually print a line starting with its device name - dl0 in the D-Link example - followed by the type of card detected. If you have a network card installed but don't see any similar message, the kernel is unable to detect your card properly. This situation will be discussed later in the section "Ethernet Autoprobing."

Kernel Configuration

Most Linux distributions are supplied with boot disks that work for all common types of PC hardware. Generally, the supplied kernel is highly modularized and includes nearly every possible driver. This is a great idea for boot disks, but is probably not what you'd want for long-term use. There isn't much point in having drivers cluttering up your disk that you will never use. Therefore, you will generally roll your own kernel and include only those drivers you actually need or want; that way you save a little disk space and reduce the time it takes to compile a new kernel.

In any case, when running a Linux system, you should be familiar with building a kernel. Think of it as a right of passage, an affirmation of the one thing that makes free software as powerful as it is - you have the source. It isn't a case of, "I have to compile a kernel," rather it's a case of, "I can compile a kernel." The basics of compiling a Linux kernel are explained in Matt Welsh's book, Running Linux (O'Reilly). Therefore, we will discuss only configuration options that affect networking in this section.

One important point that does bear repeating here is the way the kernel version numbering scheme works. Linux kernels are numbered in the following format: 2.2.14. The first digit indicates the major version number. This digit changes when there are large and significant changes to the kernel design. For example, the kernel changed from major 1 to 2 when the kernel obtained support for machines other than Intel machines. The second number is the minor version number. In many respects, this number is the most important number to look at. The Linux development community has adopted a standard at which even minor version numbers indicate production, or stable, kernels and odd minor version numbers indicate development, or unstable, kernels. The stable kernels are what you should use on a machine that is important to you, as they have been more thoroughly tested. The development kernels are what you should use if you are interested in experimenting with the newest features of Linux, but they may have problems that haven't yet been found and fixed. The third number is simply incremented for each release of a minor version.[18]

When running make menuconfig, you are presented with a text-based menu that offers lists of configuration questions, such as whether you want kernel math emulation. One of these queries asks you whether you want TCP/IP networking support. You must answer this with y to get a kernel capable of networking.

Kernel Options in Linux 2.0 and Higher

After the general option section is complete, the configuration will go on to ask whether you want to include support for various features, such as SCSI drivers or sound cards. The prompt will indicate what options are available. You can press? to obtain a description of what the option is actually offering. You'll always have the option of yes (y) to statically include the component in the kernel, or no (n) to exclude the component completely. You'll also see the module (m) option for those components that may be compiled as a run-time loadable module. Modules need to be loaded before they can be used, and are useful for drivers of components that you use infrequently.

The subsequent list of questions deal with networking support. The exact set of configuration options is in constant flux due to ongoing development. A typical list of options offered by most kernel versions around 2.0 and 2.1 looks like this:


* Network device support


Network device support (CONFIG_NETDEVICES) [Y/n/?]

You must answer this question with y if you want to use any type of networking devices, whether they are Ethernet, SLIP, PPP, or whatever. When you answer the question with y, support for Ethernet-type devices is enabled automatically. You must answer additional questions if you want to enable support for other types of network drivers:

PLIP (parallel port) support (CONFIG_PLIP) [N/y/m/?] y

PPP (point-to-point) support (CONFIG_PPP) [N/y/m/?] y


* CCP compressors for PPP are only built as modules.


SLIP (serial line) support (CONFIG_SLIP) [N/y/m/?] m

  CSLIP compressed headers (CONFIG_SLIP_COMPRESSED) [N/y/?] (NEW) y

  Keepalive and linefill (CONFIG_SLIP_SMART) [N/y/?] (NEW) y

  Six bit SLIP encapsulation (CONFIG_SLIP_MODE_SLIP6) [N/y/?] (NEW) y

These questions concern the various link layer protocols that Linux supports. Both PPP and SLIP allow you to transport IP datagrams across serial lines. PPP is actually a suite of protocols used to send network traffic across serial lines. Some of the protocols that form PPP manage the way that you authenticate yourself to the dial-in server, while others manage the way certain protocols are carried across the link - PPP is not limited to carrying TCP/IP datagrams; it may also carry other protocol such as IPX.

If you answer y or m to SLIP support, you will be prompted to answer the three questions that appear below it. The compressed header option provides support for CSLIP, a technique that compresses TCP/IP headers to as little as three bytes. Note that this kernel option does not turn on CSLIP automatically; it merely provides the necessary kernel functions for it. The Keepalive and linefill option causes the SLIP support to periodically generate activity on the SLIP line to avoid it being dropped by an inactivity timer. The Six bit SLIP encapsulation option allows you to run SLIP over lines and circuits that are not capable of transmitting the whole 8-bit data set cleanly. This is similar to the uuencoding or binhex technique used to send binary files by electronic mail.

PLIP provides a way to send IP datagrams across a parallel port connection. It is mostly used to communicate with PCs running DOS. On typical PC hardware, PLIP can be faster than PPP or SLIP, but it requires much more CPU overhead to perform, so while the transfer rate might be good, other tasks on the machine may be slow.

The following questions address network cards from various vendors. As more drivers are being developed, you are likely to see questions added to this section. If you want to build a kernel you can use on a number of different machines, or if your machine has more than one type of network card installed, you can enable more than one driver:



Ethernet (10 or 100Mbit) (CONFIG_NET_ETHERNET) [Y/n/?]


3c501 support (CONFIG_EL1) [N/y/m/?]

3c503 support (CONFIG_EL2) [N/y/m/?]

3c509/3c579 support (CONFIG_EL3) [Y/m/n/?]

3c590/3c900 series (592/595/597/900/905) "Vortex/Boomerang" support  (CONFIG_VORTEX) [N/y/m/?]

AMD LANCE and PCnet (AT1500 and NE2100) support (CONFIG_LANCE) [N/y/?]

AMD PCInet32 (VLB and PCI) support (CONFIG_LANCE32) [N/y/?] (NEW)

Western Digital/SMC cards (CONFIG_NET_VENDOR_SMC) [N/y/?]

WD80*3 support (CONFIG_WD80x3) [N/y/m/?] (NEW)

SMC Ultra support (CONFIG_ULTRA) [N/y/m/?] (NEW)

SMC Ultra32 support (CONFIG_ULTRA32) [N/y/m/?] (NEW)

SMC 9194 support (CONFIG_SMC9194) [N/y/m/?] (NEW)

Other ISA cards (CONFIG_NET_ISA) [N/y/?]

Cabletron E21xx support (CONFIG_E2100) [N/y/m/?] (NEW)

DEPCA, DE10x, DE200, DE201, DE202, DE422 support (CONFIG_DEPCA) [N/y/m/?] (NEW)

EtherWORKS 3 (DE203, DE204, DE205) support (CONFIG_EWRK3) [N/y/m/?] (NEW)

EtherExpress 16 support (CONFIG_EEXPRESS) [N/y/m/?] (NEW)

HP PCLAN+ (27247B and 27252A) support (CONFIG_HPLAN_PLUS) [N/y/m/?] (NEW)

HP PCLAN (27245 and other 27xxx series) support (CONFIG_HPLAN) [N/y/m/?] (NEW)

HP 10/100VG PCLAN (ISA, EISA, PCI) support (CONFIG_HP100) [N/y/m/?] (NEW)

NE2000/NE1000 support (CONFIG_NE2000) [N/y/m/?] (NEW)

SK_G16 support (CONFIG_SK_G16) [N/y/?] (NEW)

EISA, VLB, PCI and on card controllers (CONFIG_NET_EISA) [N/y/?]

Apricot Xen-II on card ethernet (CONFIG_APRICOT) [N/y/m/?] (NEW)

Intel EtherExpress/Pro 100B support (CONFIG_EEXPRESS_PRO100B) [N/y/m/?] (NEW)

DE425, DE434, DE435, DE450, DE500 support (CONFIG_DE4X5) [N/y/m/?] (NEW)

DECchip Tulip (dc21x4x) PCI support (CONFIG_DEC_ELCP) [N/y/m/?] (NEW)

Digi Intl. RightSwitch SE-X support (CONFIG_DGRS) [N/y/m/?] (NEW)

Pocket and portable adaptors (CONFIG_NET_POCKET) [N/y/?]

AT-LAN-TEC/RealTek pocket adaptor support (CONFIG_ATP) [N/y/?]

(NEW) D-Link DE600 pocket adaptor support (CONFIG_DE600) [N/y/m/?] (NEW)

D-Link DE620 pocket adaptor support (CONFIG_DE620) [N/y/m/?] (NEW)

Token Ring driver support (CONFIG_TR) [N/y/?]

IBM Tropic chipset based adaptor support (CONFIG_IBMTR) [N/y/m/?] (NEW)

FDDI driver support (CONFIG_FDDI) [N/y/?]

Digital DEFEA and DEFPA adapter support (CONFIG_DEFXX) [N/y/?] (NEW)

ARCnet support (CONFIG_ARCNET) [N/y/m/?]

Enable arc0e (ARCnet "Ether-Encap" packet format) (CONFIG_ARCNET_ETH) [N/y/?] (NEW)

Enable arc0s (ARCnet RFC1051 packet format) (CONFIG_ARCNET_1051) [N/y/?] (NEW)



Finally, in the file system section, the configuration script will ask you whether you want support for NFS, the networking file system. NFS lets you export file systems to several hosts, which makes the files appear as if they were on an ordinary hard disk attached to the host:

NFS file system support (CONFIG_NFS_FS) [y]

We describe NFS in detail in Chapter 14, The Network File System.

Kernel Networking Options in Linux 2.0.0 and Higher

Linux 2.0.0 marked a significant change in Linux Networking. Many features were made a standard part of the Kernel, such as support for IPX. A number of options were also added and made configurable. Many of these options are used only in very special circumstances and we won't cover them in detail. The Networking HOWTO probably addresses what is not covered here. We'll list a number of useful options in this section, and explain when you'd want to use each one:


To use TCP/IP networking, you must answer this question with y. If you answer with n, however, you will still be able to compile the kernel with IPX support:

Networking options --›

     [*] TCP/IP networking


You have to enable this option if your system acts as a gateway between two networks or between a LAN and a SLIP link, etc. It doesn't hurt to enable this by default, but you may want to disable it to configure a host as a so-called firewall. Firewalls are hosts that are connected to two or more networks, but don't route traffic between them. They're commonly used to provide users with Internet access at minimal risk to the internal network. Users are allowed to log in to the firewall and use Internet services, but the company's machines are protected from outside attacks because incoming connections can't cross the firewall (firewalls are covered in detail in Chapter 9, TCP/IP Firewall):

[*] IP: forwarding/gatewaying

Virtual hosting

These options together allow to you configure more than one IP address onto an interface. This is sometimes useful if you want to do "virtual hosting," through which a single machine can be configured to look and act as though it were actually many separate machines, each with its own network personality. We'll talk more about IP aliasing in a moment:

[*] Network aliasing

    ‹*› IP: aliasing support


This option enables you to collect data on the volume of IP traffic leaving and arriving at your machine (we cover this is detail in Chapter 10, IP Accounting):

[*] IP: accounting

PC hug

This option works around an incompatibility with some versions of PC/TCP, a commercial TCP/IP implementation for DOS-based PCs. If you enable this option, you will still be able to communicate with normal Unix machines, but performance may be hurt over slow links:

- - (it is safe to leave these untouched)

[*] IP: PC/TCP compatibility mode

Diskless booting

This function enables Reverse Address Resolution Protocol (RARP). RARP is used by diskless clients and X terminals to request their IP address when booting. You should enable RARP if you plan to serve this sort of client. A small program called rarp, included with the standard networking utilities, is used to add entries to the kernel RARP table:

‹*› IP: Reverse ARP


When sending data over TCP, the kernel has to break up the stream into blocks of data to pass to IP. The size of the block is called the Maximum Transmission Unit, or MTU. For hosts that can be reached over a local network such as an Ethernet, it is typical to use an MTU as large as the maximum length of an Ethernet packet - 1,500 bytes. When routing IP over a Wide Area Network like the Internet, it is preferable to use smaller-sized datagrams to ensure that they don't need to be further broken down along the route through a process called IP fragmentation.[19] The kernel is able to automatically determine the smallest MTU of an IP route and to automatically configure a TCP connection to use it. This behavior is on by default. If you answer y to this option this feature will be disabled.

If you do want to use smaller packet sizes for data sent to specific hosts (because, for example, the data goes through a SLIP link), you can do so using the mss option of the route command, which is briefly discussed at the end of this chapter:

[] IP: Disable Path MTU Discovery (normally enabled)

Security feature

The IP protocol supports a feature called Source Routing. Source routing allows you to specify the route a datagram should follow by coding the route into the datagram itself. This was once probably useful before routing protocols such as RIP and OSPF became commonplace. But today it's considered a security threat because it can provide clever attackers with a way of circumventing certain types of firewall protection by bypassing the routing table of a router. You would normally want to filter out source routed datagrams, so this option is normally enabled:

[*] IP: Drop source routed frames

Novell support

This option enables support for IPX, the transport protocol Novell Networking uses. Linux will function quite happily as an IPX router and this support is useful in environments where you have Novell fileservers. The NCP filesystem also requires IPX support enabled in your kernel; if you wish to attach to and mount your Novell filesystems you must have this option enabled (we'll dicuss IPX and the NCP filesystem in Chapter 15, IPX and the NCP Filesystem):

‹*› The IPX protocol

Amateur radio

These three options select support for the three Amateur Radio protocols supported by Linux: AX.25, NetRom and Rose (we don't describe them in this book, but they are covered in detail in the AX25 HOWTO):

‹*› Amateur Radio AX.25 Level 2

‹*› Amateur Radio NET/ROM

‹*› Amateur Radio X.25 PLP (Rose)

Linux supports another driver type: the dummy driver. The following question appears toward the start of the device-driver section:

‹*› Dummy net driver support

The dummy driver doesn't really do much, but it is quite useful on standalone or PPP/SLIP hosts. It is basically a masqueraded loopback interface. On hosts that offer PPP/SLIP but have no other network interface, you want to have an interface that bears your IP address all the time. This is discussed in a little more detail in "The Dummy Interface" in Chapter 5, Configuring TCP/IP Networking. Note that today you can achieve the same result by using the IP alias feature and configuring your IP address as an alias on the loopback interface.

A Tour of Linux Network Devices

The Linux kernel supports a number of hardware drivers for various types of equipment. This section gives a short overview of the driver families available and the interface names they use.

There is a number of standard names for interfaces in Linux, which are listed here. Most drivers support more than one interface, in which case the interfaces are numbered, as in eth0 and eth1:


This is the local loopback interface. It is used for testing purposes, as well as a couple of network applications. It works like a closed circuit in that any datagram written to it will immediately be returned to the host's networking layer. There's always one loopback device present in the kernel, and there's little sense in having more.

eth0, eth1,…

These are the Ethernet card interfaces. They are used for most Ethernet cards, including many of the parallel port Ethernet cards.

tr0, tr1,…

These are the Token Ring card interfaces. They are used for most Token Ring cards, including non-IBM manufactured cards.

sl0, sl1,…

These are the SLIP interfaces. SLIP interfaces are associated with serial lines in the order in which they are allocated for SLIP.

ppp0, ppp1,…

These are the PPP interfaces. Just like SLIP interfaces, a PPP interface is associated with a serial line once it is converted to PPP mode.

plip0, plip1,…

These are the PLIP interfaces. PLIP transports IP datagrams over parallel lines. The interfaces are allocated by the PLIP driver at system boot time and are mapped onto parallel ports. In the 2.0.x kernels there is a direct relationship between the device name and the I/O port of the parallel port, but in later kernels the device names are allocated sequentially, just as for SLIP and PPP devices.

ax0, ax1,…

These are the AX.25 interfaces. AX.25 is the primary protocol used by amateur radio operators. AX.25 interfaces are allocated and mapped in a similar fashion to SLIP devices.

There are many other types of interfaces available for other network drivers. We've listed only the most common ones.

During the next few sections, we will discuss the details of using the drivers described previously. The Networking HOWTO provides details on how to configure most of the others, and the AX25 HOWTO explains how to configure the Amateur Radio network devices.

Ethernet Installation

The current Linux network code supports a large variety of Ethernet cards. Most drivers were written by Donald Becker, who authored a family of drivers for cards based on the National Semiconductor 8390 chip; these have become known as the Becker Series Drivers. Many other developers have contributed drivers, and today there are few common Ethernet cards that aren't supported by Linux. The list of supported Ethernet cards is growing all the time, so if your card isn't supported yet, chances are it will be soon.

Sometime earlier in Linux's history we would have attempted to list all supported Ethernet cards, but that would now take too much time and space. Fortunately, Paul Gortmaker maintains the Ethernet HOWTO, which lists each of the supported cards and provides useful information about getting each of them running under Linux.[20] It is posted monthly to the comp.os.linux.answers newsgroup, and is also available on any of the Linux Documentation Project mirror sites.

Even if you are confident you know how to install a particular type of Ethernet card in your machine, it is often worthwhile taking a look at what the Ethernet HOWTO has to say about it. You will find information that extends beyond simple configuration issues. For example, it could save you a lot of headaches to know the behavior of some DMA-based Ethernet cards that use the same DMA channel as the Adaptec 1542 SCSI controller by default. Unless you move one of them to a different DMA channel, you will wind up with the Ethernet card writing packet data to arbitrary locations on your hard disk.

To use any of the supported Ethernet cards with Linux, you may use a precompiled kernel from one of the major Linux distributions. These generally have modules available for all of the supported drivers, and the installation process usually allows you to select which drivers you want loaded. In the long term, however, it's better to build your own kernel and compile only those drivers you actually need; this saves disk space and memory.

Ethernet Autoprobing

Many of the Linux Ethernet drivers are smart enough to know how to search for the location of your Ethernet card. This saves you having to tell the kernel where it is manually. The Ethernet HOWTO lists whether a particular driver uses autoprobing and in which order it searches the I/O address for the card.

There are three limitations to the autoprobing code. First, it may not recognize all cards properly. This is especially true for some of the cheaper clones of common cards. Second, the kernel won't autoprobe for more than one card unless specifically instructed. This was a conscious design decision, as it is assumed you will want to have control over which card is assigned to which interface. The best way to do this reliably is to manually configure the Ethernet cards in your machine. Third, the driver may not probe at the address that your card is configured for. Generally speaking, the drivers will autoprobe at the addresses that the particular device is capable of being configured for, but sometimes certain addresses are ignored to avoid hardware conflicts with other types of cards that commonly use that same address.

PCI network cards should be reliably detected. But if you are using more than one card, or if the autoprobe should fail to detect your card, you have a way to explicitly tell the kernel about the card's base address and name.

At boot time you can supply arguments and information to the kernel that any of the kernel components may read. This mechanism allows you to pass information to the kernel that Ethernet drivers can use to locate your Ethernet hardware without making the driver probe.

If you use lilo to boot your system, you can pass parameters to the kernel by specifying them through the append option in the lilo.conf file. To inform the kernel about an Ethernet device, you can pass the following parameters:


The first four parameters are numeric, while the last is the device name. The irq, base_addr, and name parameters are required, but the two param parameters are optional. Any of the numeric values may be set to zero, which causes the kernel to determine the value by probing.

The first parameter sets the IRQ assigned to the device. By default, the kernel will try to autodetect the device's IRQ channel. The 3c503 driver, for example, has a special feature that selects a free IRQ from the list 5, 9, 3, 4 and configures the card to use this line. The base_addr parameter gives the I/O base address of the card; a value of zero tells the kernel to probe the addresses listed above.

Different drivers use the next two parameters differently. For shared-memory cards, such as the WD80x3, they specify starting and ending addresses of the shared memory area. Other cards commonly use param1 to set the level at which debugging information is displayed. Values of 1 through 7 denote increasing levels of verbosity, while 8 turns them off altogether; 0 denotes the default. The 3c503 driver uses param2 to choose between the internal transceiver (default) or an external transceiver (a value of 1). The former uses the card's BNC connector; the latter uses its AUI port. The param arguments need not be included at all if you don't have anything special to configure.

The first non-numeric argument is interpreted by the kernel as the device name. You must specify a device name for each Ethernet card you describe.

If you have two Ethernet cards, you can have Linux autodetect one card and pass the second card's parameters with lilo, but you'll probably want to manually configure both cards. If you decide to have the kernel probe for one and manually configure the second, you must make sure the kernel doesn't accidentally find the second card first, or else the other one won't be registered at all. You do this by passing lilo a reserve option, which explicitly tells the kernel to avoid probing the I/O space taken up by the second card. For instance, to make Linux install a second Ethernet card at 0x300 as eth1, you would pass the following parameters to the kernel:

reserve=0x300,32 ether=0,0x300,eth1

The reserve option makes sure no driver accesses the second card's I/O space when probing for some device. You may also use the kernel parameters to override autoprobing for eth0:

reserve=0x340,32 ether=0,0x340,eth0

You can turn off autoprobing altogether. You might do this, for example, to stop a kernel probing for an Ethernet card you might have temporarily removed. Disabling autoprobing is as simple as specifying a base_addr argument of -1:


To supply these parameters to the kernel at boot time, you enter the parameters at the lilo "boot:" prompt. To have lilo give you the "boot:" at the prompt, you must press any one of the Control, Alt or Shift keys while lilo is booting. If you press the Tab key at the prompt, you will be presented with a list of kernels that you may boot. To boot a kernel with parameters supplied, enter the name of the kernel you wish to boot, followed by a space, then followed by the parameters you wish to supply. When you press the Enter key, lilo will load that kernel and boot it with the parameters you've supplied.

To make this change occur automatically on each reboot, enter the parameters into the /etc/lilo.conf using the append= keyword. An example might look like this:









label=2.2.14 read-only

After you've edited lilo.conf, you must rerun the lilo command to activate the change.

The PLIP Driver

Parallel Line IP (PLIP) is a cheap way to network when you want to connect only two machines. It uses a parallel port and a special cable, achieving speeds of 10 kilobytes per second to 20 kilobytes per second.

PLIP was originally developed by Crynwr, Inc. Its design at the time was rather ingenious (or, if you prefer, a hack), because the original parallel ports on IBM PCs were designed to spend their time being unidirectional printer ports; the eight data lines could be used only to send data from the PC to the peripheral device, but not the other way around.[21] The Cyrnwr PLIP design worked around this limitation by using the port's five status lines for input, which limited it to transferring all data as nibbles (half bytes) only, but allowed for bidirectional transfer. This mode of operation was called PLIP "mode 0." Today, the parallel ports supplied on PC hardware cater to full bidirectional 8-bit data transfer, and PLIP has been extended to accomodate this with the addition of PLIP "mode 1."

Linux kernels up to and including Version 2.0 support PLIP mode 0 only, and an enhanced parallel port driver exists as a patch against the 2.0 kernel and as a standard part of the 2.2 kernel code to provide PLIP mode 1 operation, too.[22] Unlike earlier versions of the PLIP code, the driver now attempts to be compatible with the PLIP implementations from Crynwr, as well as the PLIP driver in NCSA telnet.[23] To connect two machines using PLIP, you need a special cable sold at some shops as a Null Printer or Turbo Laplink cable. You can, however, make one yourself fairly easily; Appendix B, Useful Cable Configurations shows you how.

The PLIP driver for Linux is the work of almost countless persons. It is currently maintained by Niibe Yutaka.[24] If compiled into the kernel, it sets up a network interface for each of the possible printer ports, with plip0 corresponding to parallel port lp0, plip1 corresponding to lp1, etc. The mapping of interfaces to ports differs in the 2.0 kernels and the 2.2 kernels. In the 2.0 kernels, the mapping was hardwired in the drivers/net/Spacd.c file in the kernel source. The default mappings in this file are:

Interface I/O Port IRQ
plip0 0x3BC 7
plip1 0x378 7
plip2 0x278 5

If you configured your printer port in a different way, you must change these values in drivers/net/Space.c in the Linux kernel source and build a new kernel.

In the 2.2 kernels, the PLIP driver uses the "parport" parallel port sharing driver developed by Philip Blundell.[25] The new driver allocates the PLIP network device names serially, just as for the Ethernet or PPP drivers, so the first PLIP device created is plip0, the second is plip1, and so on. The physical parallel port hardware is also allocated serially. By default, the parallel port driver will attempt to detect your parallel port hardware with an autoprobe routine, recording the physical device information in the order found. It is better practice to explicitly tell the kernel the physical I/O parameters. You can do this by supplying arguments to the parport_pc.o module as you load it, or if you have compiled the driver into your kernel, using lilo to supply arguments to the kernel at boot time. The IRQ setting of any device may be changed later by writing the new IRQ value to the related /proc/parport/*/irq file.

Configuring the physical I/O parameters in a 2.2 kernel when loading the module is straightforward. For instance, to tell the driver that you have two PC-style parallel ports at I/O addresses 0x278 and 0c378 and IRQs 5 and 7, respectively, you would load the module with the following arguments:

modprobe parport_pc io=0x278,0x378 irq=5,7

The corresponding arguments to pass to the kernel for a compiled-in driver are:

parport=0x278,5 parport=0x378,7

You would use the lilo append keyword to have these arguments passed to the kernel automatically at boot time.

When the PLIP driver is initialized, either at boot time if it is built-in, or when the plip.o module is loaded, each of the parallel ports will have a plip network device associated with it. plip0 will be assigned to the first parallel port device, plip1 the second, and so on. You can manually override this automatic assignment using another set of kernel arguments. For instance, to assign parport0 to network device plip0, and parport1 to network device plip1, you would use kernel arguments of:

plip=parport1 plip=parport0

This mapping does not mean, however, that you cannot use these parallel ports for printing or other purposes. The physical parallel port devices are used by the PLIP driver only when the corresponding interface is configured up.

The PPP and SLIP Drivers

Point-to-Point Protocol (PPP) and Serial Line IP (SLIP) are widely used protocols for carrying IP packets over a serial link. A number of institutions offer dialup PPP and SLIP access to machines that are on the Internet, thus providing IP connectivity to private persons (something that's otherwise hardly affordable).

No hardware modifications are necessary to run PPP or SLIP; you can use any serial port. Since serial port configuration is not specific to TCP/IP networking, we have devoted a separate chapter to this. Please refer to Chapter 4, Configuring the Serial Hardware, for more information. We cover PPP in detail in Chapter 8, The Point-to-Point Protocol, and SLIP in Chapter 7, Serial Line IP.

Other Network Types

Most other network types are configured similarly to Ethernet. The arguments passed to the loadable modules will be different and some drivers may not support more than one card, but just about everything else is the same. Documentation for these cards is generally available in the /usr/src/linux/Documentation/networking/ directory of the Linux kernel source.

Chapter 4. Configuring the Serial Hardware

The Internet is growing at an incredible rate. Much of this growth is attributed to Internet users who can't afford high-speed permanent network connections and who use protocols such as SLIP, PPP, or UUCP to dial in to a network provider to retrieve their daily dose of email and news.

This chapter is intended to help all people who rely on modems to maintain their link to the outside world. We won't cover the mechanics of how to configure your modem (the manual that came with it will tell you more about it than we can), but we will cover most of the Linux-specific aspects of managing devices that use serial ports. Topics include serial communications software, creating the serial device files, serial hardware, and configuring serial devices using the setserial and stty commands. Many other related topics are covered in the Serial HOWTO by David Lawyer.[26]

Communications Software for Modem Links

There are a number of communications packages available for Linux. Many of these packages are terminal programs, which allow a user to dial in to another computer as if she were sitting in front of a simple terminal. The traditional terminal program for Unix-like environments is kermit. It is, however, fairly ancient now, and would probably be considered difficult to use. There are more comfortable programs available that support features, like telephone-dialing dictionaries, script languages to automate dialing and logging in to remote computer systems, and a variety of file exchange protocols. One of these programs is minicom, which was modeled after some of the most popular DOS terminal programs. X11 users are accommodated, too. seyon is a fully featured X11-based communications program.

Terminal programs aren't the only type of serial communication programs available. Other programs let you connect to a host and download news and email in a single bundle, to read and reply later at your leisure. This can save a lot of time, and is especially useful if you are unfortunate enough to live in an area where your local calls are time-charged. All of the reading and replying time can be spent offline, and when you are ready, you can redial and upload your responses in a single bundle. This all consumes a bit more hard disk because all of the messages have to be stored to your disk before you can read them, but this could be a reasonable trade-off at today's hard drive prices.

UUCP epitomizes this communication software style. It is a program suite that copies files from one host to another and executes programs on a remote host. It is frequently used to transport mail or news in private networks. Ian Taylor's UUCP package, which also runs under Linux, is described in detail in Chapter 16, Managing Taylor UUCP. Other noninteractive communications software is used throughout networks such as Fidonet. Fidonet application ports like ifmail are also available, although we expect that not many people still use them.

PPP and SLIP are in between, allowing both interactive and noninteractive use. Many people use PPP or SLIP to dial in to their campus network or other Internet Service Provider to run FTP and read web pages. PPP and SLIP are also, however, commonly used over permanent or semipermanent connections for LAN-to-LAN coupling, although this is really only interesting with ISDN or other high-speed network connections.

Introduction to Serial Devices

The Unix kernel provides devices for accessing serial hardware, typically called tty devices (pronounced as it is spelled: T-T-Y). This is an abbreviation for Teletype device, which used to be one of the major manufacturers of terminal devices in the early days of Unix. The term is used now for any character-based data terminal. Throughout this chapter, we use the term to refer exclusively to the Linux device files rather than the physical terminal.

Linux provides three classes of tty devices: serial devices, virtual terminals (all of which you can access in turn by pressing Alt-F1 through Alt-Fnn on the local console), and pseudo-terminals (similar to a two-way pipe, used by applications such as X11). The former were called tty devices because the original character-based terminals were connected to the Unix machine by a serial cable or telephone line and modem. The latter two were named after the tty device because they were created to behave in a similar fashion from the programmer's perspective.

SLIP and PPP are most commonly implemented in the kernel. The kernel doesn't really treat the tty device as a network device that you can manipulate like an Ethernet device, using commands such as ifconfig. However, it does treat tty devices as places where network devices can be bound. To do this, the kernel changes what is called the "line discipline" of the tty device. Both SLIP and PPP are line disciplines that may be enabled on tty devices. The general idea is that the serial driver handles data given to it differently, depending on the line discipline it is configured for. In its default line discipline, the driver simply transmits each character it is given in turn. When the SLIP or PPP line discipline is selected, the driver instead reads a block of data, wraps a special header around it that allows the remote end to identify that block of data in a stream, and transmits the new data block. It isn't too important to understand this yet; we'll cover both SLIP and PPP in later chapters, and it all happens automatically for you anyway.

Accessing Serial Devices

Like all devices in a Unix system, serial ports are accessed through device special files, located in the /dev directory. There are two varieties of device files related to serial drivers, and there is one device file of each type for each port. The device will behave slightly differently, depending on which of its device files we open. We'll cover the differences because it will help you understand some of the configurations and advice that you might see relating to serial devices, but in practice you need to use only one of these. At some point in the future, one of them may even disappear completely.

The most important of the two classes of serial device has a major number of 4, and its device special files are named ttyS0, ttyS1, etc. The second variety has a major number of 5, and was designed for use when dialing out (calling out) through a port; its device special files are called cua0, cua1, etc. In the Unix world, counting generally starts at zero, while laypeople tend to start at one. This creates a small amount of confusion for people because COM1: is represented by /dev/ttyS0, COM2: by /dev/ttyS1, etc. Anyone familiar with IBM PC-style hardware knows that COM3: and greater were never really standardized anyway.

The cua, or "callout," devices were created to solve the problem of avoiding conflicts on serial devices for modems that have to support both incoming and outgoing connections. Unfortunately, they've created their own problems and are now likely to be discontinued. Let's briefly look at the problem.

Linux, like Unix, allows a device, or any other file, to be opened by more than one process simultaneously. Unfortunately, this is rarely useful with tty devices, as the two processes will almost certainly interfere with each other. Luckily, a mechanism was devised to allow a process to check if a tty device had already been opened by another device before opening it. The mechanism uses what are called lock files. The idea was that when a process wanted to open a tty device, it would check for the existence of a file in a special location, named similarly to the device it intends to open. If the file does not exist, the process creates it and opens the tty device. If the file does exist, the process assumes another process already has the tty device open and takes appropriate action. One last clever trick to make the lock file management system work was writing the process ID (pid) of the process that had created the lock file into the lock file itself; we'll talk more about that in a moment.

The lock file mechanism works perfectly well in circumstances in which you have a defined location for the lock files and all programs know where to find them. Alas, this wasn't always the case for Linux. It wasn't until the Linux Filesystem Standard defined a standard location for lock files when tty lock files began to work correctly. At one time there were at least four, and possibly more locations chosen by software developers to store lock files: /usr/spool/locks/, /var/spool/locks/, /var/lock/, and /usr/lock/. Confusion caused chaos. Programs were opening lock files in different locations that were meant to control a single tty device; it was as if lock files weren't being used at all.

The cua devices were created to provide a solution to this problem. Rather than relying on the use of lock files to prevent clashes between programs wanting to use the serial devices, it was decided that the kernel could provide a simple means of arbitrating who should be given access. If the ttyS device were already opened, an attempt to open the cua would result in an error that a program could interpret to mean the device was already being used. If the cua device were already open and an attempt was made to open the ttyS, the request would block; that is, it would be put on hold and wait until the cua device was closed by the other process. This worked quite well if you had a single modem that you had configured for dial-in access and you occasionally wanted to dial out on the same device. But it did not work very well in environments where you had multiple programs wanting to call out on the same device. The only way to solve the contention problem was to use lock files! Back to square one.

Suffice it to say that the Linux Filesystem Standard came to the rescue and now mandates that lock files be stored in the /var/lock directory, and that by convention, the lock file name for the ttyS1 device, for instance, is LCK…ttyS1. The cua lock files should also go in this directory, but use of cua devices is now discouraged.

The cua devices will probably still be around for some time to provide a period of backward compatibility, but in time they will be retired. If you are wondering what to use, stick to the ttyS device and make sure that your system is Linux FSSTND compliant, or at the very least that all programs using the serial devices agree on where the lock files are located. Most software dealing with serial tty devices provides a compile-time option to specify the location of the lock files. More often than not, this will appear as a variable called something like LOCKDIR in the Makefile or in a configuration header file. If you're compiling the software yourself, it is best to change this to agree with the FSSTND-specified location. If you're using a precompiled binary and you're not sure where the program will write its lock files, you can use the following command to gain a hint:

strings binaryfile | grep lock

If the location found does not agree with the rest of your system, you can try creating a symbolic link from the lock directory that the foreign executable wants to use back to /var/lock/. This is ugly, but it will work.

The Serial Device Special Files

Minor numbers are identical for both types of serial devices. If you have your modem on one of the ports COM1: through COM4:, its minor number will be the COM port number plus 63. If you are using special serial hardware, such as a high-performance multiple port serial controller, you will probably need to create special device files for it; it probably won't use the standard device driver. The Serial-HOWTO should be able to assist you in finding the appropriate details.

Assume your modem is on COM2:. Its minor number will be 65, and its major number will be 4 for normal use. There should be a device called ttyS1 that has these numbers. List the serial ttys in the /dev/ directory. The fifth and sixth columns show the major and minor numbers, respectively:

$ ls -l /dev/ttyS*

0 crw-rw-- 1 uucp dialout 4, 64 Oct 13 1997 /dev/ttyS0

0 crw-rw-- 1 uucp dialout 4, 65 Jan 26 21:55 /dev/ttyS1

0 crw-rw-- 1 uucp dialout 4, 66 Oct 13 1997 /dev/ttyS2

0 crw-rw-- 1 uucp dialout 4, 67 Oct 13 1997 /dev/ttyS3

If there is no device with major number 4 and minor number 65, you will have to create one. Become the superuser and type:

# mknod -m 666 /dev/ttyS1 c 4 65

# chown uucp.dialout /dev/ttyS1

The various Linux distributions use slightly differing strategies for who should own the serial devices. Sometimes they will be owned by root, and other times they will be owned by another user, such as uucp in our example. Modern distributions have a group specifically for dial-out devices, and any users who are allowed to use them are added to this group.

Some people suggest making /dev/modem a symbolic link to your modem device so that casual users don't have to remember the somewhat unintuitive ttyS1. However, you cannot use modem in one program and the real device file name in another. Their lock files would have different names and the locking mechanism wouldn't work.

Serial Hardware

RS-232 is currently the most common standard for serial communications in the PC world. It uses a number of circuits for transmitting single bits, as well as for synchronization. Additional lines may be used for signaling the presence of a carrier (used by modems) and for handshaking. Linux supports a wide variety of serial cards that use the RS-232 standard.

Hardware handshake is optional, but very useful. It allows either of the two stations to signal whether it is ready to receive more data, or if the other station should pause until the receiver is done processing the incoming data. The lines used for this are called "Clear to Send" (CTS) and "Ready to Send" (RTS), respectively, which explains the colloquial name for hardware handshake: "RTS/CTS." The other type of handshake you might be familiar with is called "XON/XOFF" handshaking. XON/XOFF uses two nominated characters, conventionally Ctrl-S and Ctrl-Q, to signal to the remote end that it should stop and start transmitting data, respectively. While this method is simple to implement and okay for use by dumb terminals, it causes great confusion when you are dealing with binary data, as you may want to transmit those characters as part of your data stream, and not have them interpreted as flow control characters. It is also somewhat slower to take effect than hardware handshake. Hardware handshake is clean, fast, and recommended in preference to XON/XOFF when you have a choice.

In the original IBM PC, the RS-232 interface was driven by a UART chip called the 8250. PCs around the time of the 486 used a newer version of the UART called the 16450. It was slightly faster than the 8250. Nearly all Pentium-based machines have been supplied with an even newer version of the UART called the 16550. Some brands (most notably internal modems equipped with the Rockwell chip set) use completely different chips that emulate the behavior of the 16550 and can be treated similarly. Linux supports all of these in its standard serial port driver.[27]

The 16550 was a significant improvement over the 8250 and the 16450 because it offered a 16-byte FIFO buffer. The 16550 is actually a family of UART devices, comprising the 16550, the 16550A, and the 16550AFN (later renamed PC16550DN). The differences relate to whether the FIFO actually works; the 16550AFN is the one that is sure to work. There was also an NS16550, but its FIFO never really worked either.

The 8250 and 16450 UARTs had a simple 1-byte buffer. This means that a 16450 generates an interrupt for every character transmitted or received. Each interrupt takes a short period of time to service, and this small delay limits 16450s to a reliable maximum bit speed of about 9,600 bps in a typical ISA bus machine.

In the default configuration, the kernel checks the four standard serial ports, COM1: through COM4:. The kernel is also able to automatically detect what UART is used for each of the standard serial ports, and will make use of the enhanced FIFO buffer of the 16550, if it is available.

Using the Configuration Utilities

Now let's spend some time looking at the two most useful serial device configuration utilities: setserial and stty.

The setserial Command

The kernel will make its best effort to correctly determine how your serial hardware is configured, but the variations on serial device configuration makes this determination difficult to achieve 100 percent reliably in practice. A good example of where this is a problem is the internal modems we talked about earlier. The UART they use has a 16-byte FIFO buffer, but it looks like a 16450 UART to the kernel device driver: unless we specifically tell the driver that this port is a 16550 device, the kernel will not make use of the extended buffer. Yet another example is that of the dumb 4-port cards that allow sharing of a single IRQ among a number of serial devices. We may have to specifically tell the kernel which IRQ port it's supposed to use, and that IRQs may be shared.

setserial was created to configure the serial driver at runtime. The setserial command is most commonly executed at boot time from a script called 0setserial on some distributions, and rc.serial on others. This script is charged with the responsibility of initializing the serial driver to accommodate any nonstandard or unusual serial hardware in the machine.

The general syntax for the setserial command is:

setserial device [parameters]

in which the device is one of the serial devices, such as ttyS0.

The setserial command has a large number of parameters. The most common of these are described in Table 4.1. For information on the remainder of the parameters, you should refer to the setserial manual page.

Table 4.1: setserial Command-Line Parameters

Parameter Description
port port_number Specify the I/O port address of the serial device. Port numbers should be specified in hexadecimal notation, e.g., 0x2f8.
irq num Specify the interrupt request line the serial device is using.
uart uart_type Specify the UART type of the serial device. Common values are 16450, 16550, etc. Setting this value to none will disable this serial device.
fourport Specifying this parameter instructs the kernel serial driver that this port is one port of an AST Fourport card.
spd_hi Program the UART to use a speed of 57.6 kbps when a process requests 38.4 kbps.
spd_vhi Program the UART to use a speed of 115 kbps when a process requests 38.4 kbps.
spd_normal Program the UART to use the default speed of 38.4 kbps when requested. This parameter is used to reverse the effect of a spd_hi or spd_vhi performed on the specified serial device.
auto_irq This parameter will cause the kernel to attempt to automatically determine the IRQ of the specified device. This attempt may not be completely reliable, so it is probably better to think of this as a request for the kernel to guess the IRQ. If you know the IRQ of the device, you should specify that it use the irq parameter instead.
autoconfig This parameter must be specified in conjunction with the port parameter. When this parameter is supplied, setserial instructs the kernel to attempt to automatically determine the UART type located at the supplied port address. If the auto_irq parameter is also supplied, the kernel attempts to automatically determine the IRQ, too.
skip_test This parameter instructs the kernel not to bother performing the UART type test during auto-configuration. This is necessary when the UART is incorrectly detected by the kernel.

A typical and simple rc file to configure your serial ports at boot time might look something like that shown in Example 4.1. Most Linux distributions will include something slightly more sophisticated than this one.

Example 4.1: Example rc.serial setserial Commands

# /etc/rc.serial - serial line configuration script.


# Configure serial devices

/sbin/setserial /dev/ttyS0 auto_irq skip_test autoconfig

/sbin/setserial /dev/ttyS1 auto_irq skip_test autoconfig

/sbin/setserial /dev/ttyS2 auto_irq skip_test autoconfig

/sbin/setserial /dev/ttyS3 auto_irq skip_test autoconfig


# Display serial device configuration

/sbin/setserial -bg /dev/ttyS*

The -bg /dev/ttyS* argument in the last command will print a neatly formatted summary of the hardware configuration of all active serial devices. The output will look like that shown in Example 4.2.

Example 4.2: Output of setserial -bg /dev/ttyS Command

/dev/ttyS0 at 0x03f8 (irq = 4) is a 16550A

/dev/ttyS1 at 0x02f8 (irq = 3) is a 16550A

The stty Command

The name stty probably means "set tty," but the stty command can also be used to display a terminal's configuration. Perhaps even more so than setserial, the stty command provides a bewildering number of characteristics you can configure. We'll cover the most important of these in a moment. You can find the rest described in the stty manual page.

The stty command is most commonly used to configure terminal parameters, such as whether characters will be echoed or what key should generate a break signal. We explained earlier that serial devices are tty devices and the stty command is therefore equally applicable to them.

One of the more important uses of the stty for serial devices is to enable hardware handshaking on the device. We talked briefly about hardware handshaking earlier. The default configuration for serial devices is for hardware handshaking to be disabled. This setting allows "three wire" serial cables to work; they don't support the necessary signals for hardware handshaking, and if it were enabled by default, they'd be unable to transmit any characters to change it.

Surprisingly, some serial communications programs don't enable hardware handshaking, so if your modem supports hardware handshaking, you should configure the modem to use it (check your modem manual for what command to use), and also configure your serial device to use it. The stty command has a crtscts flag that enables hardware handshaking on a device; you'll need to use this. The command is probably best issued from the rc.serial file (or equivalent) at boot time using commands like those shown in Example 4.3.

Example 4.3: Example rc.serial stty Commands


stty crtscts ‹ /dev/ttyS0

stty crtscts ‹ /dev/ttyS1

stty crtscts ‹ /dev/ttyS2

stty crtscts ‹ /dev/ttyS3


The stty command works on the current terminal by default, but by using the input redirection ("‹") feature of the shell, we can have stty manipulate any tty device. It's a common mistake to forget whether you are supposed to use "‹" or "›"; modern versions of the stty command have a much cleaner syntax for doing this. To use the new syntax, we'd rewrite our sample configuration to look like that shown in Example 4.4.

Example 4.4: Example rc.serial stty Commands Using Modern Syntax


stty crtscts -F /dev/ttyS0

stty crtscts -F /dev/ttyS1

stty crtscts -F /dev/ttyS2

stty crtscts -F /dev/ttyS3


We mentioned that the stty command can be used to display the terminal configuration parameters of a tty device. To display all of the active settings on a tty device, use:

$ stty -a -F /dev/ttyS1

The output of this command, shown in Example 4.5, gives you the status of all flags for that device; a flag shown with a preceding minus, as in -crtscts, means that the flag has been turned off.

Example 4.5: Output of stty -a Command

speed 19200 baud; rows 0; columns 0; line = 0;

intr = ^C; quit = ^\; erase = ^?; kill = ^U; eof = ^D; eol = ‹undef›;

 eol2 = ‹undef›; start = ^Q; stop = ^S; susp = ^Z; rprnt = ^R;

 werase = ^W; lnext = ^V; flush = ^O; min = 1; time = 0;

-parenb -parodd cs8 hupcl -cstopb cread clocal -crtscts

-ignbrk -brkint -ignpar -parmrk -inpck -istrip -inlcr -igncr -icrnl -ixon

 -ixoff -iuclc -ixany -imaxbel

-opost -olcuc -ocrnl onlcr -onocr -onlret -ofill -ofdel nl0 cr0 tab0

 bs0 vt0 ff0

-isig -icanon iexten echo echoe echok -echonl -noflsh -xcase -tostop

 -echoprt echoctl echoke

A description of the most important of these flags is given in Table 4.2. Each of these flags is enabled by supplying it to stty and disabled by supplying it to stty with the - character in front of it. Thus, to disable hardware handshaking on the ttyS0 device, you would use:

$ stty -crtscts -F /dev/ttyS0

Table 4.2: stty Flags Most Relevant to Configuring Serial Devices

Flags Description
N Set the line speed to N bits per second.
crtsdts Enable/Disable hardware handshaking.
ixon Enable/Disable XON/XOFF flow control.
clocal Enable/Disable modem control signals such as DTR/DTS and DCD. This is necessary if you are using a "three wire" serial cable because it does not supply these signals.
cs5 cs6 cs7 cs8 Set number of data bits to 5, 6, 7, or 8, respectively.
parodd Enable odd parity. Disabling this flag enables even parity.
parenb Enable parity checking. When this flag is negated, no parity is used.
cstopb Enable use of two stop bits per character. When this flag is negated, one stop bit per character is used.
echo Enable/Disable echoing of received characters back to sender. 

The next example combines some of these flags and sets the ttyS0 device to 19,200 bps, 8 data bits, no parity, and hardware handshaking with echo disabled:

$ stty 19200 cs8 -parenb crtscts -echo -F /dev/ttyS0

Serial Devices and the login: Prompt

It was once very common that a Unix installation involved one server machine and many "dumb" character mode terminals or dial-up modems. Today that sort of installation is less common, which is good news for many people interested in operating this way, because the "dumb" terminals are now very cheap to acquire. Dial-up modem configurations are no less common, but these days they would probably be used to support a SLIP or PPP login (discussed in Chapter 7, Serial Line IP and Chapter 8, The Point-to-Point Protocol) than to be used for a simple login. Nevertheless, each of these configurations can make use of a simple program called a getty program.

The term getty is probably a contraction of "get tty." A getty program opens a serial device, configures it appropriately, optionally configures a modem, and waits for a connection to be made. An active connection on a serial device is usually indicated by the Data Carrier Detect (DCD) pin on the serial device being raised. When a connection is detected, the getty program issues a login: prompt, and then invokes the login program to handle the actual system login. Each of the virtual terminals (e.g., /dev/tty1) in Linux has a getty running against it.

There are a number of different getty implementations, each designed to suit some configurations better than others. The getty that we'll describe here is called mgetty. It is quite popular because it has all sorts of features that make it especially modem-friendly, including support for automatic fax programs and voice modems. We'll concentrate on configuring mgetty to answer conventional data calls and leave the rest for you to explore at your convenience.

Configuring the mgetty Daemon

The mgetty daemon is available in source form from ftp://alpha.greenie.net/pub/mgetty/source/, and is available in just about all Linux distributions in prepackaged form. The mgetty daemon differs from most other getty implementations in that it has been designed specifically for Hayes-compatible modems. It still supports direct terminal connections, but is best suited for dialup applications. Rather than using the DCD line to detect an incoming call, it listens for the RING message generated by modern modems when they detect an incoming call and are not configured for auto-answer.

The main executable program is called /usr/sbin/mgetty, and its main configuration file is called /etc/mgetty/mgetty.config. There are a number of other binary programs and configuration files that cover other mgetty features.

For most installations, configuration is a matter of editing the /etc/mgetty/ mgetty.config file and adding appropriate entries to the /etc/inittab file to execute mgetty automatically.

Example 4.6 shows a very simple mgetty configuration file. This example configures two serial devices. The first, /dev/ttyS0, supports a Hayes-compatible modem at 38,400 bps. The second, /dev/ttyS0, supports a directly connected VT100 terminal at 19,200 bps.

Example 4.6: Sample /etc/mgetty/mgetty.config File


# mgetty configuration file


# this is a sample configuration file, see mgetty.info for details


# comment lines start with a "#", empty lines are ignored


# ____________________ global section ____________________


# In this section, you put the global defaults, per-port stuff is below


# access the modem(s) with 38400 bps

speed 38400


# set the global debug level to "4" (default from policy.h)

debug 4


# ____________________ port specific section ____________________


# Here you can put things that are valid only for one line, not the others



# Hayes modem connected to ttyS0: don't do fax, less logging


port ttyS0

 debug 3

 data-only y


# direct connection of a VT100 terminal which doesn't like DTR drops


port ttyS1

 direct y

 speed 19200

 toggle-dtr n


The configuration file supports global and port-specific options. In our example we used a global option to set the speed to 38,400 bps. This value is inherited by the ttyS0 port. Ports we apply mgetty to use this speed setting unless it is overwritten by a port-specific speed setting, as we have done in the ttyS1 configuration.

The debug keyword controls the verbosity of mgetty logging. The data-only keyword in the ttyS0 configuration causes mgetty to ignore any modem fax features, to operate just as a data modem. The direct keyword in the ttyS1 configuration instructs mgetty not to attempt any modem initialization on the port. Finally, the toggle-dtr keyword instructs mgetty not to attempt to hang up the line by dropping the DTR (Data Terminal Ready) pin on the serial interface; some terminals don't like this to happen.

You can also choose to leave the mgetty.config file empty and use command-line arguments to specify most of the same parameters. The documentation accompanying the application includes a complete description of the mgetty configuration file parameters and command-line arguments. See the following example.

We need to add two entries to the /etc/inittab file to activate this configuration. The inittab file is the configuration file of the Unix System V init command. The init command is responsible for system initialization; it provides a means of automatically executing programs at boot time and re-executing them when they terminate. This is ideal for the goals of running a getty program.

T0:23:respawn:/sbin/mgetty ttyS0

T1:23:respawn:/sbin/mgetty ttyS1

Each line of the /etc/inittab file contains four fields, separated by colons. The first field is an identifier that uniquely labels an entry in the file; traditionally it is two characters, but modern versions allow four. The second field is the list of run levels at which this entry should be active. A run level is a means of providing alternate machine configurations and is implemented using trees of startup scripts stored in directories called /etc/rc1.d, /etc/rc2.d, etc. This feature is typically implemented very simply, and you should model your entries on others in the file or refer to your system documentation for more information. The third field describes when to take action. For the purposes of running a getty program, this field should be set to respawn, meaning that the command should be re-executed automatically when it dies. There are several other options, as well, but they are not useful for our purposes here. The fourth field is the actual command to execute; this is where we specify the mgetty command and any arguments we wish to pass it. In our simple example we're starting and restarting mgetty whenever the system is operating at either of run levels two or three, and are supplying as an argument just the name of the device we wish it to use. The mgetty command assumes the /dev/, so we don't need to supply it.

This chapter was a quick introduction to mgetty and how to offer login prompts to serial devices. You can find more extensive information in the Serial-HOWTO.

After you've edited the configuration files, you need to reload init to make the changes take effect. Simply send a hangup signal to the init process; it always has a process ID of one, so you can use the following command safely:

# kill -HUP 1

Chapter 5. Configuring TCP/IP Networking

In this chapter, we walk you through all the necessary steps to set up TCP/IP networking on your machine. Starting with the assignment of IP addresses, we slowly work our way through the configuration of TCP/IP network interfaces and introduce a few tools that come in handy when hunting down network installation problems.

Most of the tasks covered in this chapter will generally have to be done only once. Afterward, you have to touch most configuration files only when adding a new system to your network or when you reconfigure your system entirely. Some of the commands used to configure TCP/IP, however, have to be executed each time the system is booted. This is usually done by invoking them from the system /etc/rc* scripts.

Commonly, the network-specific part of this procedure is contained in a script. The name of this script varies in different Linux distributions. In many older Linux distributions, it is known as rc.net or rc.inet. Sometimes you will also see two scripts named rc.inet1 and rc.inet2; the former initializes the kernel part of networking and the latter starts basic networking services and applications. In modern distributions, the rc files are structured in a more sophisticated arrangement; here you may find scripts in the /etc/init.d/ (or /etc/rc.d/init.d/) directory that create the network devices and other rc files that run the network application programs. This book's examples are based on the latter arrangement.

This chapter discusses parts of the script that configure your network interfaces, while applications will be covered in later chapters. After finishing this chapter, you should have established a sequence of commands that properly configure TCP/IP networking on your computer. You should then replace any sample commands in your configuration scripts with your commands, make sure the script is executed from the basic rc script at startup time, and reboot your machine. The networking rc scripts that come along with your favorite Linux distribution should provide a solid example from which to work.

Mounting the /proc Filesystem

Some of the configuration tools of the Linux NET-2 and NET-3 release rely on the /proc filesystem for communicating with the kernel. This interface permits access to kernel runtime information through a filesystem-like mechanism. When mounted, you can list its files like any other filesystem, or display their contents. Typical items include the loadavg file, which contains the system load average, and meminfo, which shows current core memory and swap usage.

To this, the networking code adds the net directory. It contains a number of files that show things like the kernel ARP tables, the state of TCP connections, and the routing tables. Most network administration tools get their information from these files.

The proc filesystem (or procfs, as it is also known) is usually mounted on /proc at system boot time. The best method is to add the following line to /etc/fstab:

# procfs mount point:

none /proc proc defaults

Then execute mount /proc from your /etc/rc script.

The procfs is now configured into most kernels by default. If the procfs is not in your kernel, you will get a message such as: mount: fs type procfs not supported by kernel. You will then have to recompile the kernel and answer "yes" when asked for procfs support.

Installing the Binaries

If you are using one of the prepackaged Linux distributions, it will contain the major networking applications and utilities along with a coherent set of sample files. The only case in which you might have to obtain and install new utilities is when you install a new kernel release. As they occasionally involve changes in the kernel networking layer, you will need to update the basic configuration tools. This update at least involves recompiling, but sometimes you may also be required to obtain the latest set of binaries. These binaries are available at their official home site at ftp.inka.de/pub/comp/Linux/networking/NetTools/, packaged in an archive called net-tools-XXX.tar.gz, where XXX is the version number. The release matching Linux 2.0 is net-tools-1.45.

If you want to compile and install the standard TCP/IP network applications yourself, you can obtain the sources from most Linux FTP servers. All modern Linux distributions include a fairly comprehensive range of TCP/IP network applications, such as World Wide Web browsers, telnet and ftp programs, and other network applications, such as talk. If you do find something that you do need to compile yourself, the chances are good that it will compile under Linux from source quite simply if you follow the instructions included in the source package.

Setting the Hostname

Most, if not all, network applications rely on you to set the local host's name to some reasonable value. This setting is usually made during the boot procedure by executing the hostname command. To set the hostname to name, enter:

# hostname name

It is common practice to use the unqualified hostname without specifying the domain name. For instance, hosts at the Virtual Brewery (described in Appendix A, Example Network: The Virtual Brewery) might be called vale.vbrew.com or vlager.vbrew.com. These are their official fully qualified domain names (FQDNs). Their local hostnames would be the first component of the name, such as vale. However, as the local hostname is frequently used to look up the host's IP address, you have to make sure that the resolver library is able to look up the host's IP address. This usually means that you have to enter the name in /etc/hosts.

Some people suggest using the domainname command to set the kernel's idea of a domain name to the remaining part of the FQDN. This way you could combine the output from hostname and domainname to get the FQDN again. However, this is at best only half correct. domainname is generally used to set the host's NIS domain, which may be entirely different from the DNS domain to which your host belongs. Instead, to ensure that the short form of your hostname is resolvable with all recent versions of the hostname command, either add it as an entry in your local Domain Name Server or place the fully qualified domain name in the /etc/hosts file. You may then use the -fqdn argument to the hostname command, and it will print the fully qualifed domain name.

Assigning IP Addresses

If you configure the networking software on your host for standalone operation (for instance, to be able to run the INN Netnews software), you can safely skip this section, because the only IP address you will need is for the loopback interface, which is always

Things are a little more complicated with real networks like Ethernets. If you want to connect your host to an existing network, you have to ask its administrators to give you an IP address on this network. When setting up a network all by yourself, you have to assign IP addresses yourself.

Hosts within a local network should usually share addresses from the same logical IP network. Hence, you have to assign an IP network address. If you have several physical networks, you have to either assign them different network numbers, or use subnetting to split your IP address range into several subnetworks. Subnetting will be revisited in the next section, "Creating Subnets".

When picking an IP network number, much depends on whether you intend to get on the Internet in the near future. If so, you should obtain an official IP address now. Ask your network service provider to help you. If you want to obtain a network number, just in case you might get on the Internet someday, request a Network Address Application Form from hostmaster@internic.net, or your country's own Network Information Center, if there is one.

If your network is not connected to the Internet and won't be in the near future, you are free to choose any legal network address. Just make sure no packets from your internal network escape to the real Internet. To make sure no harm can be done even if packets did escape, you should use one of the network numbers reserved for private use. The Internet Assigned Numbers Authority (IANA) has set aside several network numbers from classes A, B, and C that you can use without registering. These addresses are valid only within your private network and are not routed between real Internet sites. The numbers are defined by RFC 1597 and are listed in Table 2.1 in Chapter 2, Issues of TCP/IP Networking. Note that the second and third blocks contain 16 and 256 networks, respectively.

Picking your addresses from one of these network numbers is not only useful for networks completely unconnected to the Internet; you can still implement a slightly more restricted access using a single host as a gateway. To your local network, the gateway is accessible by its internal IP address, while the outside world knows it by an officially registered address (assigned to you by your provider). We come back to this concept in connection with the IP masquerade facility in Chapter 11, IP Masquerade and Network Address Translation.

Throughout the remainder of the book, we will assume that the brewery's network manager uses a class B network number, say Of course, a class C network number would definitely suffice to accommodate both the Brewery's and the Winery's networks. We'll use a class B network here for the sake of simplicity; it will make the subnetting examples in the next section of this chapter a little more intuitive.

Creating Subnets

To operate several Ethernets (or other networks, once a driver is available), you have to split your network into subnets. Note that subnetting is required only if you have more than one broadcast network - point-to-point links don't count. For instance, if you have one Ethernet, and one or more SLIP links to the outside world, you don't need to subnet your network. This is explained in more detail in Chapter 7, Serial Line IP.

To accommodate the two Ethernets, the Brewery's network manager decides to use 8 bits of the host part as additional subnet bits. This leaves another 8 bits for the host part, allowing for 254 hosts on each of the subnets. She then assigns subnet number 1 to the brewery, and gives the winery number 2. Their respective network addresses are thus and The subnet mask is

vlager, which is the gateway between the two networks, is assigned a host number of 1 on both of them, which gives it the IP addresses and, respectively.

Note that in this example we are using a class B network to keep things simple, but a class C network would be more realistic. With the new networking code, subnetting is not limited to byte boundaries, so even a class C network may be split into several subnets. For instance, you could use two bits of the host part for the netmask, giving you 4 possible subnets with 64 hosts on each.[28]

Writing hosts and networks Files

After you have subnetted your network, you should prepare for some simple sort of hostname resolution using the /etc/hosts file. If you are not going to use DNS or NIS for address resolution, you have to put all hosts in the hosts file.

Even if you want to run DNS or NIS during normal operation, you should have some subset of all hostnames in /etc/hosts. You should have some sort of name resolution, even when no network interfaces are running, for example, during boot time. This is not only a matter of convenience, but it allows you to use symbolic hostnames in your network rc scripts. Thus, when changing IP addresses, you only have to copy an updated hosts file to all machines and reboot, rather than edit a large number of rc files separately. Usually you put all local hostnames and addresses in hosts, adding those of any gateways and NIS servers used.[29]

You should make sure your resolver only uses information from the hosts file during initial testing. Sample files that come with your DNS or NIS software may produce strange results. To make all applications use /etc/hosts exclusively when looking up the IP address of a host, you have to edit the /etc/host.conf file. Comment out any lines that begin with the keyword order by preceding them with a hash sign, and insert the line:

order hosts

The configuration of the resolver library is covered in detail in Chapter 6, Name Service and Resolver Configuration.

The hosts file contains one entry per line, consisting of an IP address, a hostname, and an optional list of aliases for the hostname. The fields are separated by spaces or tabs, and the address field must begin in the first column. Anything following a hash sign (#) is regarded as a comment and is ignored.

Hostnames can be either fully qualified or relative to the local domain. For vale, you would usually enter the fully qualified name, vale.vbrew.com, and vale by itself in the hosts file, so that it is known by both its official name and the shorter local name.

This is an example how a hosts file at the Virtual Brewery might look. Two special names are included, vlager-if1 and vlager-if2, which give the addresses for both interfaces used on vlager:


# Hosts file for Virtual Brewery/Virtual Winery


# IP FQDN aliases

# localhost

# vlager.vbrew.com vlager vlager-if1 1 vstout.vbrew.com vstout vale.vbrew.com vale

# vlager-if2 vbeaujolais.vbrew.com vbeaujolais vbardolino.vbrew.com vbardolino vchianti.vbrew.com vchianti

Just as with a host's IP address, you should sometimes use a symbolic name for network numbers, too. Therefore, the hosts file has a companion called /etc/networks that maps network names to network numbers, and vice versa. At the Virtual Brewery, we might install a networks file like this:[30]

# /etc/networks for the Virtual Brewery



Interface Configuration for IP

After setting up your hardware as explained in Chapter 4, Configuring the Serial Hardware, you have to make these devices known to the kernel networking software. A couple of commands are used to configure the network interfaces and initialize the routing table. These tasks are usually performed from the network initialization script each time you boot the system. The basic tools for this process are called ifconfig (where "if" stands for interface) and route.

ifconfig is used to make an interface accessible to the kernel networking layer. This involves the assignment of an IP address and other parameters, and activation of the interface, also known as "bringing up" the interface. Being active here means that the kernel will send and receive IP datagrams through the interface. The simplest way to invoke it is with:

ifconfig interface ip-address

This command assigns ip-address to interface and activates it. All other parameters are set to default values. For instance, the default network mask is derived from the network class of the IP address, such as for a class B address. ifconfig is described in detail in the section "All About ifconfig".

route allows you to add or remove routes from the kernel routing table. It can be invoked as:

route [add|del] [-net|-host] target [if]

The add and del arguments determine whether to add or delete the route to target. The -net and -host arguments tell the route command whether the target is a network or a host (a host is assumed if you don't specify). The if argument is again optional, and allows you to specify to which network interface the route should be directed - the Linux kernel makes a sensible guess if you don't supply this information. This topic will be explained in more detail in succeeding sections.

The Loopback Interface

The very first interface to be activated is the loopback interface:

# ifconfig lo

Occasionally, you will see the dummy hostname localhost being used instead of the IP address. ifconfig will look up the name in the hosts file, where an entry should declare it as the hostname for

# Sample /etc/hosts entry for localhost


To view the configuration of an interface, you invoke ifconfig, giving it only the interface name as argument:

$ ifconfig lo

lo Link encap:Local Loopback

   inet addr: Mask:


   RX packets:0 errors:0 dropped:0 overruns:0 frame:0

   TX packets:0 errors:0 dropped:0 overruns:0 carrier:0


As you can see, the loopback interface has been assigned a netmask of, since is a class A address.

Now you can almost start playing with your mini-network. What is still missing is an entry in the routing table that tells IP that it may use this interface as a route to destination This is accomplished by using:

# route add

Again, you can use localhost instead of the IP address, provided you've entered it into your /etc/hosts.

Next, you should check that everything works fine, for example by using ping. ping is the networking equivalent of a sonar device.[31] The command is used to verify that a given address is actually reachable, and to measure the delay that occurs when sending a datagram to it and back again. The time required for this process is often referred to as the "round-trip time":

# ping localhost

PING localhost ( 56 data bytes

64 bytes from icmp_seq=0 ttl=255 time=0.4 ms

64 bytes from icmp_seq=1 ttl=255 time=0.4 ms

64 bytes from icmp_seq=2 ttl=255 time=0.4 ms


 -- localhost ping statistics -- 3 packets transmitted, 3 packets received, 0% packet loss

round-trip min/avg/max = 0.4/0.4/0.4 ms


When you invoke ping as shown here, it will continue emitting packets forever, unless interrupted by the user. The ^C marks the place where we pressed Ctrl-C.

The previous example shows that packets for are properly delivered and a reply is returned to ping almost instantaneously. This shows that you have successfully set up your first network interface.

If the output you get from ping does not resemble that shown in the previous example, you are in trouble. Check any errors if they indicate that some file hasn't been installed properly. Check that the ifconfig and route binaries you use are compatible with the kernel release you run, and above all, that the kernel has been compiled with networking enabled (you see this from the presence of the /proc/net directory). If you get an error message saying "Network unreachable," you probably got the route command wrong. Make sure you use the same address you gave to ifconfig.

The steps previously described are enough to use networking applications on a standalone host. After adding the lines mentioned earlier to your network initialization script and making sure it will be executed at boot time, you may reboot your machine and try out various applications. For instance, telnet localhost should establish a telnet connection to your host, giving you a login: prompt.

However, the loopback interface is useful not only as an example in networking books, or as a test bed during development, but is actually used by some applications during normal operation.[32] Therefore, you always have to configure it, regardless of whether your machine is attached to a network or not.

Ethernet Interfaces

Configuring an Ethernet interface is pretty much the same as the loopback interface; it just requires a few more parameters when you are using subnetting.

At the Virtual Brewery, we have subnetted the IP network, which was originally a class B network, into class C subnetworks. To make the interface recognize this, the ifconfig incantation would look like this:

# ifconfig eth0 vstout netmask

This command assigns the eth0 interface the IP address of vstout ( If we omitted the netmask, ifconfig would deduce the netmask from the IP network class, which would result in an incorrect netmask of Now a quick check shows:

# ifconfig eth0

eth0 Link encap 10Mps Ethernet HWaddr 00:00:C0:90:B3:42

   inet addr Bcast Mask


   RX packets 0 errors 0 dropped 0 overrun 0

   TX packets 0 errors 0 dropped 0 overrun 0

You can see that ifconfig automatically sets the broadcast address (the Bcast field) to the usual value, which is the host's network number with all the host bits set. Also, the maximum transmission unit (the maximum size of IP datagrams the kernel will generate for this interface) has been set to the maximum size of Ethernet packets: 1,500 bytes. The defaults are usually what you will use, but all these values can be overidden if required, with special options that will be described under "All About ifconfig".

Just as for the loopback interface, you now have to install a routing entry that informs the kernel about the network that can be reached through eth0. For the Virtual Brewery, you might invoke route as:

# route add -net

At first this looks a little like magic, because it's not really clear how route detects which interface to route through. However, the trick is rather simple: the kernel checks all interfaces that have been configured so far and compares the destination address ( in this case) to the network part of the interface address (that is, the bitwise AND of the interface address and the netmask). The only interface that matches is eth0.

Now, what's that -net option for? This is used because route can handle both routes to networks and routes to single hosts (as you saw before with localhost). When given an address in dotted quad notation, route attempts to guess whether it is a network or a hostname by looking at the host part bits. If the address's host part is zero, route assumes it denotes a network; otherwise, route takes it as a host address. Therefore, route would think that is a host address rather than a network number, because it cannot know that we use subnetting. We have to tell route explicitly that it denotes a network, so we give it the -net flag.

Of course, the route command is a little tedious to type, and it's prone to spelling mistakes. A more convenient approach is to use the network names we defined in /etc/networks. This approach makes the command much more readable; even the -net flag can be omitted because route knows that denotes a network:

# route add brew-net

Now that you've finished the basic configuration steps, we want to make sure that your Ethernet interface is indeed running happily. Choose a host from your Ethernet, for instance vlager, and type:

# ping vlager

PING vlager: 64 byte packets

64 bytes from icmp_seq=0. time=11. ms

64 bytes from icmp_seq=1. time=7. ms

64 bytes from icmp_seq=2. time=12. ms

64 bytes from icmp_seq=3. time=3. ms


--vstout.vbrew.com PING Statistics-- 4 packets transmitted, 4 packets received, 0

round-trip (ms) min/avg/max = 3/8/12

If you don't see similar output, something is broken. If you encounter unusual packet loss rates, this hints at a hardware problem, like bad or missing terminators. If you don't receive any replies at all, you should check the interface configuration with netstat described later in "The netstat Command". The packet statistics displayed by ifconfig should tell you whether any packets have been sent out on the interface at all. If you have access to the remote host too, you should go over to that machine and check the interface statistics. This way you can determine exactly where the packets got dropped. In addition, you should display the routing information with route to see if both hosts have the correct routing entry. route prints out the complete kernel routing table when invoked without any arguments (-n just makes it print addresses as dotted quad instead of using the hostname):

# route -n

Kernel routing table Destination Gateway Genmask Flags Metric Ref Use Iface * UH 1 0 112 lo * U 1 0 10 eth0

The detailed meaning of these fields is explained later in "The netstat Command". The Flags column contains a list of flags set for each interface. U is always set for active interfaces, and H says the destination address denotes a host. If the H flag is set for a route that you meant to be a network route, you have to reissue the route command with the -net option. To check whether a route you have entered is used at all, check to see if the Use field in the second to last column increases between two invocations of ping.

Routing Through a Gateway

In the previous section, we covered only the case of setting up a host on a single Ethernet. Quite frequently, however, one encounters networks connected to one another by gateways. These gateways may simply link two or more Ethernets, but may also provide a link to the outside world, such as the Internet. In order to use a gateway, you have to provide additional routing information to the networking layer.

The Ethernets of the Virtual Brewery and the Virtual Winery are linked through such a gateway, namely the host vlager. Assuming that vlager has already been configured, we just have to add another entry to vstout's routing table that tells the kernel it can reach all hosts on the Winery's network through vlager. The appropriate incantation of route is shown below; the gw keyword tells it that the next argument denotes a gateway:

# route add wine-net gw vlager

Of course, any host on the Winery network you wish to talk to must have a routing entry for the Brewery's network. Otherwise you would only be able to send data to the Winery network from the Brewery network, but the hosts on the Winery would be unable to reply.

This example describes only a gateway that switches packets between two isolated Ethernets. Now assume that vlager also has a connection to the Internet (say, through an additional SLIP link). Then we would want datagrams to any destination network other than the Brewery to be handed to vlager. This action can be accomplished by making it the default gateway for vstout:

# route add default gw vlager

The network name default is a shorthand for, which denotes the default route. The default route matches every destination and will be used if there is no more specific route that matches. You do not have to add this name to /etc/networks because it is built into route.

If you see high packet loss rates when pinging a host behind one or more gateways, this may hint at a very congested network. Packet loss is not so much due to technical deficiencies as to temporary excess loads on forwarding hosts, which makes them delay or even drop incoming datagrams.

Configuring a Gateway

Configuring a machine to switch packets between two Ethernets is pretty straightforward. Assume we're back at vlager, which is equipped with two Ethernet cards, each connected to one of the two networks. All you have to do is configure both interfaces separately, giving them their respective IP addresses and matching routes, and that's it.

It is quite useful to add information on the two interfaces to the hosts file as shown in the following example, so we have handy names for them, too: vlager.vbrew.com vlager vlager-if1 vlager-if2

The sequence of commands to set up the two interfaces is then:

# ifconfig eth0 vlager-if1

# route add brew-net

# ifconfig eth1 vlager-if2

# route add wine-net

If this sequence doesn't work, make sure your kernel has been compiled with support for IP forwarding enabled. One good way to do this is to ensure that the first number on the second line of /proc/net/snmp is set to 1.

The PLIP Interface

A PLIP link used to connect two machines is a little different from an Ethernet. PLIP links are an example of what are called point-to-point links, meaning that there is a single host at each end of the link. Networks like Ethernet are called broadcast networks. Configuration of point-to-point links is different because unlike broadcast networks, point-to-point links don't support a network of their own.

PLIP provides very cheap and portable links between computers. As an example, we'll consider the laptop computer of an employee at the Virtual Brewery that is connected to vlager via PLIP. The laptop itself is called vlite and has only one parallel port. At boot time, this port will be registered as plip1. To activate the link, you have to configure the plip1 interface using the following commands:[33]

# ifconfig plip1 vlite pointopoint vlager

# route add default gw vlager

The first command configures the interface, telling the kernel that this is a point-to-point link, with the remote side having the address of vlager. The second installs the default route, using vlager as gateway. On vlager, a similar ifconfig command is necessary to activate the link (a route invocation is not needed):

# ifconfig plip1 vlager pointopoint vlite

Note that the plip1 interface on vlager does not need a separate IP address, but may also be given the address Point-to-point networks don't support a network directly, so the interfaces don't require an address on any supported network. The kernel uses the interface information in the routing table to avoid any possible confusion.[34]

Now we have configured routing from the laptop to the Brewery's network; what's still missing is a way to route from any of the Brewery's hosts to vlite. One particularly cumbersome way is to add a specific route to every host's routing table that names vlager as a gateway to vlite:

# route add vlite gw vlager

Dynamic routing offers a much better option for temporary routes. You could use gated, a routing daemon, which you would have to install on each host in the network in order to distribute routing information dynamically. The easiest option, however, is to use proxy ARP (Address Resolution Protocol). With proxy ARP, vlager will respond to any ARP query for vlite by sending its own Ethernet address. All packets for vlite will wind up at vlager, which then forwards them to the laptop. We will come back to proxy ARP in the section "Checking the ARP Tables".

Current net-tools releases contain a tool called plipconfig, which allows you to set certain PLIP timing parameters. The IRQ to be used for the printer port can be set using the ifconfig command.

The SLIP and PPP Interfaces

Although SLIP and PPP links are only simple point-to-point links like PLIP connections, there is much more to be said about them. Usually, establishing a SLIP connection involves dialing up a remote site through your modem and setting the serial line to SLIP mode. PPP is used in a similar fashion. We discuss SLIP and PPP in detail in Chapter 7 and Chapter 8, The Point-to-Point Protocol.

The Dummy Interface

The dummy interface is a little exotic, but rather useful nevertheless. Its main benefit is with standalone hosts and machines whose only IP network connection is a dialup link. In fact, the latter are standalone hosts most of the time, too.

The dilemma with standalone hosts is that they only have a single network device active, the loopback device, which is usually assigned the address On some occasions, however, you must send data to the "official" IP address of the local host. For instance, consider the laptop vlite, which was disconnected from a network for the duration of this example. An application on vlite may now want to send data to another application on the same host. Looking up vlite in /etc/hosts yields an IP address of, so the application tries to send to this address. As the loopback interface is currently the only active interface on the machine, the kernel has no idea that actually refers to itself! Consequently, the kernel discards the datagram and returns an error to the application.

This is where the dummy device steps in. It solves the dilemma by simply serving as the alter ego of the loopback interface. In the case of vlite, you simply give it the address and add a host route pointing to it. Every datagram for is then delivered locally. The proper invocation is:[35]

# ifconfig dummy vlite

# route add vlite

IP Alias

New kernels support a feature that can completely replace the dummy interface and serve other useful functions. IP Alias allows you to configure multiple IP addresses onto a physical device. In the simplest case, you could replicate the function of the dummy interface by configuring the host address as an alias onto the loopback interface and completely avoid using the dummy interface. In more complex uses, you could configure your host to look like many different hosts, each with its own IP address. This configuration is sometimes called "Virtual Hosting," although technically it is also used for a variety of other techniques.[36]

To configure an alias for an interface, you must first ensure that your kernel has been compiled with support for IP Alias (check that you have a /proc/net/ip_alias file; if not, you will have to recompile your kernel). Configuration of an IP alias is virtually identical to configuring a real network device; you use a special name to indicate it's an alias that you want. For example:

# ifconfig lo:0

This command would produce an alias for the loopback interface with the address IP aliases are referred to by appending:n to the actual network device, in which "n" is an integer. In our example, the network device we are creating the alias on is lo, and we are creating an alias numbered zero for it. This way, a single physical device may support a number of aliases.

Each alias may be treated as though it is a separate device, and as far as the kernel IP software is concerned, it will be; however, it will be sharing its hardware with another interface.

All About ifconfig

There are many more parameters to ifconfig than we have described so far. Its normal invocation is this:

ifconfig interface [address [parameters]]

interface is the interface name, and address is the IP address to be assigned to the interface. This may be either an IP address in dotted quad notation or a name that ifconfig will look up in /etc/hosts.

If ifconfig is invoked with only the interface name, it displays that interface's configuration. When invoked without any parameters, it displays all interfaces you have configured so far; a -a option forces it to show the inactive ones as well. A sample invocation for the Ethernet interface eth0 may look like this:

# ifconfig eth0

eth0 Link encap 10Mbps Ethernet HWaddr 00:00:C0:90:B3:42

   inet addr Bcast Mask


   RX packets 3136 errors 217 dropped 7 overrun 26

   TX packets 1752 errors 25 dropped 0 overrun 0

The MTU and Metric fields show the current MTU and metric value for that interface. The metric value is traditionally used by some operating systems to compute the cost of a route. Linux doesn't use this value yet, but defines it for compatibility, nevertheless.

The RX and TX lines show how many packets have been received or transmitted error free, how many errors occurred, how many packets were dropped (probably because of low memory), and how many were lost because of an overrun. Receiver overruns usually occur when packets come in faster than the kernel can service the last interrupt. The flag values printed by ifconfig roughly correspond to the names of its command-line options; they will be explained later.

The following is a list of parameters recognized by ifconfig with the corresponding flag names. Options that simply turn on a feature also allow it to be turned off again by preceding the option name by a dash (-).


This option makes an interface accessible to the IP layer. This option is implied when an address is given on the command line. It may also be used to reenable an interface that has been taken down temporarily using the down option.

This option corresponds to the flags UP and RUNNING.


This option marks an interface inaccessible to the IP layer. This effectively disables any IP traffic through the interface. Note that this option will also automatically delete all routing entries that use this interface.

netmask mask

This option assigns a subnet mask to be used by the interface. It may be given as either a 32-bit hexadecimal number preceded by 0x, or as a dotted quad of decimal numbers. While the dotted quad format is more common, the hexadecimal representation is often easier to work with. Netmasks are essentially binary, and it is easier to do binary-to-hexadecimal than binary-to-decimal conversion.

pointopoint address

This option is used for point-to-point IP links that involve only two hosts. This option is needed to configure SLIP or PLIP interfaces, for example. If a point-to-point address has been set, ifconfig displays the POINTOPOINT flag.

broadcast address

The broadcast address is usually made up from the network number by setting all bits of the host part. Some IP implementations (systems derived from BSD 4.2, for instance) use a different scheme in which all host part bits are cleared instead. The broadcast option adapts to these strange environments. If a broadcast address has been set, ifconfig displays the BROADCAST flag.


This option allows you to set the IRQ line used by certain devices. This is especially useful for PLIP, but may also be useful for certain Ethernet cards.

metric number

This option may be used to assign a metric value to the routing table entry created for the interface. This metric is used by the Routing Information Protocol (RIP) to build routing tables for the network.[37] The default metric used by ifconfig is zero. If you don't run a RIP daemon, you don't need this option at all; if you do, you will rarely need to change the metric value.

mtu bytes

This sets the Maximum Transmission Unit, which is the maximum number of octets the interface is able to handle in one transaction. For Ethernets, the MTU defaults to 1,500 (the largest allowable size of an Ethernet packet); for SLIP interfaces, it is 296. (There is no constraint on the MTU of SLIP links; this value is a good compromise.)


This is an option specific to broadcast networks such as Ethernets or packet radio. It enables the use of the Address Resolution Protocol (ARP) to detect the physical addresses of hosts attached to the network. For broadcast networks, it is on by default. If ARP is disabled, ifconfig displays the NOARP flag.

- arp

This option disables the use of ARP on this interface.


This option puts the interface in promiscuous mode. On a broadcast network, this makes the interface receive all packets, regardless of whether they were destined for this host or not. This allows network traffic analysis using packet filters and such, also called Ethernet snooping. Usually, this is a good technique for hunting down network problems that are otherwise hard to detect. Tools such as tcpdump rely on this.

On the other hand, this option allows attackers to do nasty things, such as skim the traffic of your network for passwords. You can protect against this type of attack by prohibiting just anyone from plugging their computers into your Ethernet. You could also use secure authentication protocols, such as Kerberos or the secure shell login suite.[38] This option corresponds to the PROMISC flag.

- promisc

This option turns promiscuous mode off.


Multicast addresses are like Ethernet broadcast addresses, except that instead of automatically including everybody, the only people who receive packets sent to a multicast address are those programmed to listen to it. This is useful for applications like Ethernet-based videoconferencing or network audio, to which only those interested can listen. Multicast addressing is supported by most, but not all, Ethernet drivers. When this option is enabled, the interface receives and passes multicast packets for processing. This option corresponds to the ALLMULTI flag.

- allmulti

This option turns multicast addresses off.

The netstat Command

netstat is a useful tool for checking your network configuration and activity. It is in fact a collection of several tools lumped together. We discuss each of its functions in the following sections.

Displaying the Routing Table

When you invoke netstat with the -r flag, it displays the kernel routing table in the way we've been doing with route. On vstout, it produces:

# netstat -nr

Kernel IP routing table

Destination Gateway Genmask Flags MSS Window irtt Iface    * UH 0 0 0 lo  * U 0 0 0 eth0 UG 0 0 0 eth0

The -n option makes netstat print addresses as dotted quad IP numbers rather than the symbolic host and network names. This option is especially useful when you want to avoid address lookups over the network (e.g., to a DNS or NIS server).

The second column of netstat 's output shows the gateway to which the routing entry points. If no gateway is used, an asterisk is printed instead. The third column shows the "generality" of the route, i.e., the network mask for this route. When given an IP address to find a suitable route for, the kernel steps through each of the routing table entries, taking the bitwise AND of the address and the genmask before comparing it to the target of the route.

The fourth column displays the following flags that describe the route:


The route uses a gateway.


The interface to be used is up.


Only a single host can be reached through the route. For example, this is the case for the loopback entry


This route is dynamically created. It is set if the table entry has been generated by a routing daemon like gated or by an ICMP redirect message (see "The Internet Control Message Protocol" in Chapter 2).


This route is set if the table entry was modified by an ICMP redirect message.


The route is a reject route and datagrams will be dropped.

The next three columns show the MSS, Window and irtt that will be applied to TCP connections established via this route. The MSS is the Maximum Segment Size and is the size of the largest datagram the kernel will construct for transmission via this route. The Window is the maximum amount of data the system will accept in a single burst from a remote host. The acronym irtt stands for "initial round trip time." The TCP protocol ensures that data is reliably delivered between hosts by retransmitting a datagram if it has been lost. The TCP protocol keeps a running count of how long it takes for a datagram to be delivered to the remote end, and an acknowledgement to be received so that it knows how long to wait before assuming a datagram needs to retransmitted; this process is called the round-trip time. The initial round-trip time is the value that the TCP protocol will use when a connection is first established. For most network types, the default value is okay, but for some slow networks, notably certain types of amateur packet radio networks, the time is too short and causes unnecessary retransmission. The irtt value can be set using the route command. Values of zero in these fields mean that the default is being used.

Finally, the last field displays the network interface that this route will use.

Displaying Interface Statistics

When invoked with the -i flag, netstat displays statistics for the network interfaces currently configured. If the -a option is also given, it prints all interfaces present in the kernel, not only those that have been configured currently. On vstout, the output from netstat will look like this:

# netstat -i

Kernel Interface table


lo 0 0 3185 0 0 0 3185 0 0 0 BLRU

eth0 1500 0 972633 17 20 120 628711 217 0 0 BRU

The MTU and Met fields show the current MTU and metric values for that interface. The RX and TX columns show how many packets have been received or transmitted error-free (RX-OK / TX-OK) or damaged (RX-ERR / TX-ERR); how many were dropped (RX-DRP / TX-DRP); and how many were lost because of an overrun (RX-OVR / TX-OVR).

The last column shows the flags that have been set for this interface. These characters are one-character versions of the long flag names that are printed when you display the interface configuration with ifconfig:


A broadcast address has been set.


This interface is a loopback device.


All packets are received (promiscuous mode).


ARP is turned off for this interface.


This is a point-to-point connection.


Interface is running.


Interface is up.

Displaying Connections

netstat supports a set of options to display active or passive sockets. The options -t, -u, -w, and -x show active TCP, UDP, RAW, or Unix socket connections. If you provide the -a flag in addition, sockets that are waiting for a connection (i.e., listening) are displayed as well. This display will give you a list of all servers that are currently running on your system.

Invoking netstat -ta on vlager produces this output:

$ netstat -ta

Active Internet Connections Proto Recv-Q Send-Q Local Address Foreign Address (State)

tcp 0 0 *:domain *:* LISTEN

tcp 0 0 *:time *:* LISTEN

tcp 0 0 *:smtp *:* LISTEN

tcp 0 0 vlager:smtp vstout:1040 ESTABLISHED

tcp 0 0 *:telnet *:* LISTEN

tcp 0 0 localhost:1046 vbardolino:telnet ESTABLISHED

tcp 0 0 *:chargen *:* LISTEN

tcp 0 0 *:daytime *:* LISTEN

tcp 0 0 *:discard *:* LISTEN

tcp 0 0 *:echo *:* LISTEN

tcp 0 0 *:shell *:* LISTEN

tcp 0 0 *:login *:* LISTEN

This output shows most servers simply waiting for an incoming connection. However, the fourth line shows an incoming SMTP connection from vstout, and the sixth line tells you there is an outgoing telnet connection to vbardolino.[39]

Using the -a flag by itself will display all sockets from all families.

Checking the ARP Tables

On some occasions, it is useful to view or alter the contents of the kernel's ARP tables, for example when you suspect a duplicate Internet address is the cause for some intermittent network problem. The arp tool was made for situations like this. Its command-line options are:

arp [-v] [-t hwtype] -a [hostname] arp [-v] [-t hwtype] -s hostname hwaddr

arp [-v] -d hostname [hostname…]

All hostname arguments may be either symbolic hostnames or IP addresses in dotted quad notation.

The first invocation displays the ARP entry for the IP address or host specified, or all hosts known if no hostname is given. For example, invoking arp on vlager may yield:

# arp -a

IP address HW type HW address 10Mbps Ethernet 00:00:C0:5A:42:C1 10Mbps Ethernet 00:00:C0:90:B3:42 10Mbps Ethernet 00:00:C0:04:69:AA

which shows the Ethernet addresses of vlager, vstout and vale.

You can limit the display to the hardware type specified using the -t option. This may be ether, ax25, or pronet, standing for 10 Mbps Ethernet; AMPR AX.25, and IEEE 802.5 token ring equipment, respectively.

The -s option is used to permanently add hostname's Ethernet address to the ARP tables. The hwaddr argument specifies the hardware address, which is by default expected to be an Ethernet address specified as six hexadecimal bytes separated by colons. You may also set the hardware address for other types of hardware, using the -t option.

For some reason, ARP queries for the remote host sometimes fail, for instance when its ARP driver is buggy or there is another host in the network that erroneously identifies itself with that host's IP address; this problem requires you to manually add an IP address to the ARP table. Hard-wiring IP addresses in the ARP table is also a (very drastic) measure to protect yourself from hosts on your Ethernet that pose as someone else.

Invoking arp using the -d switch deletes all ARP entries relating to the given host. This switch may be used to force the interface to re-attempt obtaining the Ethernet address for the IP address in question. This is useful when a misconfigured system has broadcasted wrong ARP information (of course, you have to reconfigure the broken host first).

The -s option may also be used to implement proxy ARP. This is a special technique through which a host, say gate, acts as a gateway to another host named fnord by pretending that both addresses refer to the same host, namely gate. It does so by publishing an ARP entry for fnord that points to its own Ethernet interface. Now when a host sends out an ARP query for fnord, gate will return a reply containing its own Ethernet address. The querying host will then send all datagrams to gate, which dutifully forwards them to fnord.

These contortions may be necessary when you want to access fnord from a DOS machine with a broken TCP implementation that doesn't understand routing too well. When you use proxy ARP, it will appear to the DOS machine as if fnord was on the local subnet, so it doesn't have to know about how to route through a gateway.

Another useful application of proxy ARP is when one of your hosts acts as a gateway to some other host only temporarily, for instance, through a dial-up link. In a previous example, we encountered the laptop vlite, which was connected to vlager through a PLIP link from time to time. Of course, this application will work only if the address of the host you want to provide proxy ARP for is on the same IP subnet as your gateway. vstout could proxy ARP for any host on the Brewery subnet (, but never for a host on the Winery subnet (

The proper invocation to provide proxy ARP for fnord is given below; of course, the given Ethernet address must be that of gate:

# arp -s fnord 00:00:c0:a1:42:e0 pub

The proxy ARP entry may be removed again by invoking:

# arp -d fnord

Chapter 6. Name Service and Resolver Configuration

As we discussed in Chapter 2, Issues of TCP/IP Networking, TCP/IP networking may rely on different schemes to convert names into addresses. The simplest way is a host table stored in /etc/hosts. This is useful only for small LANs that are run by one single administrator and otherwise have no IP traffic with the outside world. The format of the hosts file has already been described in Chapter 5, Configuring TCP/IP Networking.

Alternatively, you can use the Berkeley Internet Name Domain service (BIND) for resolving hostnames to IP addresses. Configuring BIND can be a real chore, but once you've done it, you can easily make changes in the network topology. On Linux, as on many other Unixish systems, name service is provided through a program called named. At startup, it loads a set of master files into its internal cache and waits for queries from remote or local user processes. There are different ways to set up BIND, and not all require you to run a name server on every host.

This chapter can do little more than give a rough sketch of how DNS works and how to operate a name server. It should be sufficient if you have a small LAN and an Internet uplink. For the most current information, you may want to check the documentation contained in the BIND source package, which supplies manual pages, release notes, and the BIND Operator's Guide (BOG). Don't let this name scare you off; it's actually a very useful document. For a more comprehensive coverage of DNS and associated issues, you may find DNS and BIND by Paul Albitz and Cricket Liu (O'Reilly) a useful reference. DNS questions may be answered in a newsgroup called comp.protocols.tcp-ip.domains. For technical details, the Domain Name System is defined by RFC numbers 1033, 1034, and 1035.

The Resolver Library

The term resolver refers not to a special application, but to the resolver library. This is a collection of functions that can be found in the standard C library. The central routines are gethostbyname(2) and gethostbyaddr(2), which look up all IP addresses associated with a host name, and vice versa. They may be configured to simply look up the information in hosts, to query a number of DNS name servers, or to use the hosts database of Network Information Service (NIS).

The resolver functions read configuration files when they are invoked. From these configuration files, they determine what databases to query, in which order, and other details relevant to how you've configured your environment. The older Linux standard library, libc, used /etc/host.conf as its master configuration file, but Version 2 of the GNU standard library, glibc, uses /etc/nsswitch.conf. We'll describe each in turn, since both are commonly used.

The host.conf File

The /etc/host.conf tells the older Linux standard library resolver functions which services to use, and in what order.

Options in host.conf must appear on separate lines. Fields may be separated by white space (spaces or tabs). A hash sign (#) introduces a comment that extends to the next newline. The following options are available:


This option determines the order in which the resolving services are tried. Valid options are bind for querying the name server, hosts for lookups in /etc/hosts, and nis for NIS lookups. Any or all of them may be specified. The order in which they appear on the line determines the order in which the respective services are tried.


multi takes on or off as options. This determines if a host in /etc/hosts is allowed to have several IP addresses, which is usually referred to as being "multi-homed." The default is off. This flag has no effect on DNS or NIS queries.


As we'll explain in the section "Reverse Lookups", DNS allows you to find the hostname belonging to an IP address by using the in-addr.arpa domain. Attempts by name servers to supply a false hostname are called spoofing. To guard against this, the resolver can be configured to check whether the original IP address is in fact associated with the obtained hostname. If not, the name is rejected and an error is returned. This behavior is turned on by setting nospoof on.


This option takes on or off as arguments. If it is turned on, any spoof attempts will cause the resolver to log a message to the syslog facility.


This option takes an argument specifying a domain name that will be removed from hostnames before lookup. This is useful for hosts entries, for which you might only want to specify hostnames without a local domain. If you specify your local domain name here, it will be removed from a lookup of a host with the local domain name appended, thus allowing the lookup in /etc/hosts to succeed. The domain name you add must end with the (.) character (e.g.,: linux.org.au.) if trim is to work correctly.

trim options accumulate; you can consider your host as being local to several domains.

A sample file for vlager is shown in Example 6.1.

Example 6.1: Sample host.conf File

# /etc/host.conf

# We have named running, but no NIS (yet)

order bind,hosts

# Allow multiple addrs

multi on

# Guard against spoof attempts

nospoof on

# Trim local domain (not really necessary).

trim vbrew.com.

Resolver environment variables

The settings from host.conf may be overridden using a number of environment variables:


This variable specifies a file to be read instead of /etc/host.conf.


This variable overrides the order option given in host.conf. Services are given as hosts, bind, and nis, separated by a space, comma, colon, or semicolon.


This variable determines the measures taken against spoofing. It is completely disabled by off. The values warn and warn off enable spoof checking by turning logging on and off, respectively. A value of * turns on spoof checks, but leaves the logging facility as defined in host.conf.


This variable uses a value of on or off to override the multi options from host.conf.


This variable specifies a list of trim domains that override those given in host.conf. Trim domains were explained earlier when we discussed the trim keyword.


This variable specifies a list of trim domains that are added to those given in host.conf.

The nsswitch.conf File

Version 2 of the GNU standard library includes a more powerful and flexible replacement for the older host.conf mechanism. The concept of the name service has been extended to include a variety of different types of information. Configuration options for all of the different functions that query these databases have been brought back into a single configuration file; the nsswitch.conf file.

The nsswitch.conf file allows the system administrator to configure a wide variety of different databases. We'll limit our discussion to options that relate to host and network IP address resolution. You can easily find more information about the other features by reading the GNU standard library documentation.

Options in nsswitch.conf must appear on separate lines. Fields may be separated by whitespace (spaces or tabs). A hash sign (#) introduces a comment that extends to the next newline. Each line describes a particular service; hostname resolution is one of these. The first field in each line is the name of the database, ending with a colon. The database name associated with host address resolution is hosts. A related database is networks, which is used for resolution of network names into network addresses. The remainder of each line stores options that determine the way lookups for that database are performed.

The following options are available:


Use the Domain Name System (DNS) service to resolve the address. This makes sense only for host address resolution, not network address resolution. This mechanism uses the /etc/resolv.conf file that we'll describe later in the chapter.


Search a local file for the host or network name and its corresponding address. This option uses the traditional /etc/hosts and /etc/network files.

nis or nisplus

Use the Network Information System (NIS) to resolve the host or network address. NIS and NIS+ are discussed in detail in Chapter 13, The Network Information System.

The order in which the services to be queried are listed determines the order in which they are queried when attempting to resolve a name. The query-order list is in the service description in the /etc/nsswitch.conf file. The services are queried from left to right and by default searching stops when a resolution is successful.

A simple example of host and network database specification that would mimic our configuration using the older libc standard library is shown in Example 6.2.

Example 6.2: Sample nsswitch.conf File

# /etc/nsswitch.conf


# Example configuration of GNU Name Service Switch functionality.

# Information about this file is available in the `libc6-doc' package.

hosts: dns files

networks: files

This example causes the system to look up hosts first in the Domain Name System, and the /etc/hosts file, if that can't find them. Network name lookups would be attempted using only the /etc/networks file.

You are able to control the lookup behavior more precisely using "action items" that describe what action to take given the result of the previous lookup attempt. Action items appear between service specifications, and are enclosed within square brackets, []. The general syntax of the action statement is:

[[!] status = action…]

There are two possible actions:


Controls returns to the program that attempted the name resolution. If a lookup attempt was successful, the resolver will return with the details, otherwise it will return a zero result.


The resolver will move on to the next service in the list and attempt resolution using it.

The optional (!) character specifies that the status value should be inverted before testing; that is, it means "not."

The available status values on which we can act are:


The requested entry was found without error. The default action for this status is return.


There was no error in the lookup, but the target host or network could not be found. The default action for this status is continue.


The service queried was unavailable. This could mean that the hosts or networks file was unreadable for the files service or that a name server or NIS server did not respond for the dns or nis services. The default action for this status is continue.


This status means the service is temporarily unavailable. For the files files service, this would usually indicate that the relevant file was locked by some process. For other services, it may mean the server was temporarily unable to accept connections. The default action for this status is continue.

A simple example of how you might use this mechanism is shown in Example 6.3.

Example 6.3: Sample nsswitch.conf File Using an Action Statement

# /etc/nsswitch.conf


# Example configuration of GNU Name Service Switch functionality.

# Information about this file is available in the `libc6-doc' package.

hosts: dns [!UNAVAIL=return] files

networks: files

This example attempts host resolution using the Domain Name Service system. If the return status is anything other than unavailable, the resolver returns whatever it has found. If, and only if, the DNS lookup attempt returns an unavailable status, the resolver attempts to use the local /etc/hosts. This means that we should use the hosts file only if our name server is unavailable for some reason.

Configuring Name Server Lookups Using resolv.conf

When configuring the resolver library to use the BIND name service for host lookups, you also have to tell it which name servers to use. There is a separate file for this called resolv.conf. If this file does not exist or is empty, the resolver assumes the name server is on your local host.

To run a name server on your local host, you have to set it up separately, as will be explained in the following section. If you are on a local network and have the opportunity to use an existing name server, this should always be preferred. If you use a dialup IP connection to the Internet, you would normally specify the name server of your service provider in the resolv.conf file.

The most important option in resolv.conf is name server, which gives the IP address of a name server to use. If you specify several name servers by giving the name server option several times, they are tried in the order given. You should therefore put the most reliable server first. The current implementation allows you to have up to three name server statements in resolv.conf. If no name server option is given, the resolver attempts to connect to the name server on the local host.

Two other options, domain and search, let you use shortcut names for hosts in your local domain. Usually, when just telnetting to another host in your local domain, you don't want to type in the fully qualified hostname, but use a name like gauss on the command line and have the resolver tack on the mathematics.groucho.edu part.

This is just the domain statement's purpose. It lets you specify a default domain name to be appended when DNS fails to look up a hostname. For instance, when given the name gauss, the resolver fails to find gauss. in DNS, because there is no such top-level domain. When given mathematics.groucho.edu as a default domain, the resolver repeats the query for gauss with the default domain appended, this time succeeding.

That's just fine, you may think, but as soon you get out of the Math department's domain, you're back to those fully qualified domain names. Of course, you would also want to have shorthands like quark.physics for hosts in the Physics department's domain.

This is when the search list comes in. A search list can be specified using the search option, which is a generalization of the domain statement. Where the latter gives a single default domain, the former specifies a whole list of them, each to be tried in turn until a lookup succeeds. This list must be separated by blanks or tabs.

The search and domain statements are mutually exclusive and may not appear more than once. If neither option is given, the resolver will try to guess the default domain from the local hostname using the getdomainname(2) system call. If the local hostname doesn't have a domain part, the default domain will be assumed to be the root domain.

If you decide to put a search statement into resolv.conf, you should be careful about what domains you add to this list. Resolver libraries prior to BIND 4.9 used to construct a default search list from the domain name when no search list was given. This default list was made up of the default domain itself, plus all of its parent domains up to the root. This caused some problems because DNS requests wound up at name servers that were never meant to see them.

Assume you're at the Virtual Brewery and want to log in to foot.groucho.edu. By a slip of your fingers, you mistype foot as foo, which doesn't exist. GMU's name server will therefore tell you that it knows no such host. With the old-style search list, the resolver would now go on trying the name with vbrew.com and com appended. The latter is problematic because groucho.edu.com might actually be a valid domain name. Their name server might then even find foo in their domain, pointing you to one of their hosts - which clearly was not intended.

For some applications, these bogus host lookups can be a security problem. Therefore, you should usually limit the domains on your search list to your local organization, or something comparable. At the Mathematics department of Groucho Marx University, the search list would commonly be set to maths.groucho.edu and groucho.edu.

If default domains sound confusing to you, consider this sample resolv.conf file for the Virtual Brewery:

# /etc/resolv.conf # Our domain domain vbrew.com # # We use vlager as central name server: name server

When resolving the name vale, the resolver looks up vale and, failing this, vale.vbrew.com.

Resolver Robustness

When running a LAN inside a larger network, you definitely should use central name servers if they are available. The name servers develop rich caches that speed up repeat queries, since all queries are forwarded to them. However, this scheme has a drawback: when a fire destroyed the backbone cable at Olaf's university, no more work was possible on his department's LAN because the resolver could no longer reach any of the name servers. This situation caused difficulties with most network services, such as X terminal logins and printing.

Although it is not very common for campus backbones to go down in flames, one might want to take precautions against cases like this.

One option is to set up a local name server that resolves hostnames from your local domain and forwards all queries for other hostnames to the main servers. Of course, this is applicable only if you are running your own domain.

Alternatively, you can maintain a backup host table for your domain or LAN in /etc/hosts. This is very simple to do. You simply ensure that the resolver library queries DNS first, and the hosts file next. In an /etc/host.conf file you'd use "order bind,hosts", and in an /etc/nsswitch.conf file you'd use "hosts: dns files", to make the resolver fall back to the hosts file if the central name server is unreachable.

How DNS Works

DNS organizes hostnames in a domain hierarchy. A domain is a collection of sites that are related in some sense - because they form a proper network (e.g., all machines on a campus, or all hosts on BITNET), because they all belong to a certain organization (e.g., the U.S. government), or because they're simply geographically close. For instance, universities are commonly grouped in the edu domain, with each university or college using a separate subdomain, below which their hosts are subsumed. Groucho Marx University have the groucho.edu domain, while the LAN of the Mathematics department is assigned maths.groucho.edu. Hosts on the departmental network would have this domain name tacked onto their hostname, so erdos would be known as erdos.maths.groucho.edu. This is called the fully qualified domain name (FQDN), which uniquely identifies this host worldwide.

Figure 6.1 shows a section of the namespace. The entry at the root of this tree, which is denoted by a single dot, is quite appropriately called the root domain and encompasses all other domains. To indicate that a hostname is a fully qualified domain name, rather than a name relative to some (implicit) local domain, it is sometimes written with a trailing dot. This dot signifies that the name's last component is the root domain.

Figure 6.1: A part of the domain namespace

Depending on its location in the name hierarchy, a domain may be called top-level, second-level, or third-level. More levels of subdivision occur, but they are rare. This list details several top-level domains you may see frequently:

Domain Description
edu (Mostly U.S.) educational institutions like universities.
com Commercial organizations and companies.
org Non-commercial organizations. Private UUCP networks are often in this domain.
net Gateways and other administrative hosts on a network.
mil U.S. military institutions.
gov U.S. government institutions.
uucp Officially, all site names formerly used as UUCP names without domains have been moved to this domain. 

Historically, the first four of these were assigned to the U.S., but recent changes in policy have meant that these domains, named global Top Level Domains (gTLD), are now considered global in nature. Negotiations are currently underway to broaden the range of gTLDs, which may result in increased choice in the future.

Outside the U.S., each country generally uses a top-level domain of its own named after the two-letter country code defined in ISO-3166. Finland, for instance, uses the fi domain; fr is used by France, de by Germany, and au by Australia. Below this top-level domain, each country's NIC is free to organize hostnames in whatever way they want. Australia has second-level domains similar to the international top-level domains, named com.au and edu.au. Other countries, like Germany, don't use this extra level, but have slightly long names that refer directly to the organizations running a particular domain. It's not uncommon to see hostnames like ftp.informatik.uni-erlangen.de. Chalk that up to German efficiency.

Of course, these national domains do not imply that a host below that domain is actually located in that country; it means only that the host has been registered with that country's NIC. A Swedish manufacturer might have a branch in Australia and still have all its hosts registered with the se top-level domain.

Organizing the namespace in a hierarchy of domain names nicely solves the problem of name uniqueness; with DNS, a hostname has to be unique only within its domain to give it a name different from all other hosts worldwide. Furthermore, fully qualified names are easy to remember. Taken by themselves, these are already very good reasons to split up a large domain into several subdomains.

DNS does even more for you than this. It also allows you to delegate authority over a subdomain to its administrators. For example, the maintainers at the Groucho Computing Center might create a subdomain for each department; we already encountered the math and physics subdomains above. When they find the network at the Physics department too large and chaotic to manage from outside (after all, physicists are known to be an unruly bunch of people), they may simply pass control of the physics.groucho.edu domain to the administrators of this network. These administrators are free to use whatever hostnames they like and assign them IP addresses from their network in whatever fashion they desire, without outside interference.

To this end, the namespace is split up into zones, each rooted at a domain. Note the subtle difference between a zone and a domain: the domain groucho.edu encompasses all hosts at Groucho Marx University, while the zone groucho.edu includes only the hosts that are managed by the Computing Center directly; those at the Mathematics department, for example. The hosts at the Physics department belong to a different zone, namely physics.groucho.edu. In Figure 6.1, the start of a zone is marked by a small circle to the right of the domain name.

Name Lookups with DNS

At first glance, all this domain and zone fuss seems to make name resolution an awfully complicated business. After all, if no central authority controls what names are assigned to which hosts, how is a humble application supposed to know?

Now comes the really ingenious part about DNS. If you want to find the IP address of erdos, DNS says, "Go ask the people who manage it, and they will tell you."

In fact, DNS is a giant distributed database. It is implemented by so-called name servers that supply information on a given domain or set of domains. For each zone there are at least two, or at most a few, name servers that hold all authoritative information on hosts in that zone. To obtain the IP address of erdos, all you have to do is contact the name server for the groucho.edu zone, which will then return the desired data.

Easier said than done, you might think. So how do I know how to reach the name server at Groucho Marx University? In case your computer isn't equipped with an address-resolving oracle, DNS provides for this, too. When your application wants to look up information on erdos, it contacts a local name server, which conducts a so-called iterative query for it. It starts off by sending a query to a name server for the root domain, asking for the address of erdos.maths.groucho.edu. The root name server recognizes that this name does not belong to its zone of authority, but rather to one below the edu domain. Thus, it tells you to contact an edu zone name server for more information and encloses a list of all edu name servers along with their addresses. Your local name server will then go on and query one of those, for instance, a.isi.edu. In a manner similar to the root name server, a.isi.edu knows that the groucho.edu people run a zone of their own, and points you to their servers. The local name server will then present its query for erdos to one of these, which will finally recognize the name as belonging to its zone, and return the corresponding IP address.

This looks like a lot of traffic being generated for looking up a measly IP address, but it's really only miniscule compared to the amount of data that would have to be transferred if we were still stuck with HOSTS.TXT. There's still room for improvement with this scheme, however.

To improve response time during future queries, the name server stores the information obtained in its local cache. So the next time anyone on your local network wants to look up the address of a host in the groucho.edu domain, your name server will go directly to the groucho.edu name server.[40]

Of course, the name server will not keep this information forever; it will discard it after some time. The expiration interval is called the time to live, or TTL. Each datum in the DNS database is assigned such a TTL by administrators of the responsible zone.

Types of Name Servers

Name servers that hold all information on hosts within a zone are called authoritative for this zone, and sometimes are referred to as master name servers. Any query for a host within this zone will end up at one of these master name servers.

Master servers must be fairly well synchronized. Thus, the zone's network administrator must make one the primary server, which loads its zone information from data files, and make the others secondary servers, which transfer the zone data from the primary server at regular intervals.

Having several name servers distributes workload; it also provides backup. When one name server machine fails in a benign way, like crashing or losing its network connection, all queries will fall back to the other servers. Of course, this scheme doesn't protect you from server malfunctions that produce wrong replies to all DNS requests, such as from software bugs in the server program itself.

You can also run a name server that is not authoritative for any domain.[41] This is useful, as the name server will still be able to conduct DNS queries for the applications running on the local network and cache the information. Hence it is called a caching-only server.

The DNS Database

We have seen that DNS not only deals with IP addresses of hosts, but also exchanges information on name servers. DNS databases may have, in fact, many different types of entries.

A single piece of information from the DNS database is called a resource record (RR). Each record has a type associated with it describing the sort of data it represents, and a class specifying the type of network it applies to. The latter accommodates the needs of different addressing schemes, like IP addresses (the IN class), Hesiod addresses (used by MIT's Kerberos system), and a few more. The prototypical resource record type is the A record, which associates a fully qualified domain name with an IP address.

A host may be known by more than one name. For example you might have a server that provides both FTP and World Wide Web servers, which you give two names: ftp.machine.org and www.machine.org. However, one of these names must be identified as the official or canonical hostname, while the others are simply aliases referring to the official hostname. The difference is that the canonical hostname is the one with an associated A record, while the others only have a record of type CNAME that points to the canonical hostname.

We will not go through all record types here, but we will give you a brief example. Example 6.4 shows a part of the domain database that is loaded into the name servers for the physics.groucho.edu zone.

Example 6.4: An Excerpt from the named.hosts File for the Physics Department

; Authoritative Information on physics.groucho.edu.

@ IN SOA niels.physics.groucho.edu. janet.niels.physics.groucho.edu. {

 1999090200; serial no

 360000; refresh

 3600; retry

 3600000; expire

 3600; default ttl



; Name servers

 IN NS niels

 IN NS gauss.maths.groucho.edu.

gauss.maths.groucho.edu. IN A


; Theoretical Physics (subnet 12)

niels IN A


name server IN CNAME niels

otto IN A

quark IN A

down IN A

strange IN A

; Collider Lab. (subnet 14)

boson IN A

muon IN A

bogon IN A

Apart from the A and CNAME records, you can see a special record at the top of the file, stretching several lines. This is the SOA resource record signaling the Start of Authority, which holds general information on the zone the server is authoritative for. The SOA record comprises, for instance, the default time to live for all records.

Note that all names in the sample file that do not end with a dot should be interpreted relative to the physics.groucho.edu domain. The special name (@) used in the SOA record refers to the domain name by itself.

We have seen earlier that the name servers for the groucho.edu domain somehow have to know about the physics zone so that they can point queries to their name servers. This is usually achieved by a pair of records: the NS record that gives the server's FQDN, and an A record that associates an address with that name. Since these records are what holds the namespace together, they are frequently called glue records. They are the only instances of records in which a parent zone actually holds information on hosts in the subordinate zone. The glue records pointing to the name servers for physics.groucho.edu are shown in Example 6.5.

Example 6.5: An Excerpt from the named.hosts File for GMU

; Zone data for the groucho.edu zone.

@ IN SOA vax12.gcc.groucho.edu. joe.vax12.gcc.groucho.edu. {

 1999070100; serial no

 360000; refresh

 3600; retry

 3600000; expire

 3600; default ttl



; Glue records for the physics.groucho.edu zone

physics IN NS niels.physics.groucho.edu.

 IN NS gauss.maths.groucho.edu.

niels.physics IN A

gauss.maths IN A

Reverse Lookups

Finding the IP address belonging to a host is certainly the most common use for the Domain Name System, but sometimes you'll want to find the canonical hostname corresponding to an address. Finding this hostname is called reverse mapping, and is used by several network services to verify a client's identity. When using a single hosts file, reverse lookups simply involve searching the file for a host that owns the IP address in question. With DNS, an exhaustive search of the namespace is out of the question. Instead, a special domain, in-addr.arpa, has been created that contains the IP addresses of all hosts in a reversed dotted quad notation. For instance, an IP address of corresponds to the name The resource-record type linking these names to their canonical hostnames is PTR.

Creating a zone of authority usually means that its administrators have full control over how they assign addresses to names. Since they usually have one or more IP networks or subnets at their hands, there's a one-to-many mapping between DNS zones and IP networks. The Physics department, for instance, comprises the subnets,, and

Consequently, new zones in the in-addr.arpa domain have to be created along with the physics zone, and delegated to the network administrators at the department: 8.76.149.in-addr.arpa, 12.76.149.in-addr.arpa, and 14.76.149.in-addr.arpa. Otherwise, installing a new host at the Collider Lab would require them to contact their parent domain to have the new address entered into their in-addr.arpa zone file.

The zone database for subnet 12 is shown in Example 6.6. The corresponding glue records in the database of their parent zone are shown in Example 6.7.

Example 6.6: An Excerpt from the named.rev File for Subnet 12

; the 12.76.149.in-addr.arpa domain.

@ IN SOA niels.physics.groucho.edu. janet.niels.physics.groucho.edu. {

 1999090200 360000 3600 3600000 3600


2 IN PTR otto.physics.groucho.edu.

4 IN PTR quark.physics.groucho.edu.

5 IN PTR down.physics.groucho.edu.

6 IN PTR strange.physics.groucho.edu.

Example 6.7: An Excerpt from the named.rev File for Network 149.76

; the 76.149.in-addr.arpa domain.

@ IN SOA vax12.gcc.groucho.edu. joe.vax12.gcc.groucho.edu. {

 1999070100 360000 3600 3600000 3600


; subnet 4: Mathematics Dept.

1.4 IN PTR sophus.maths.groucho.edu.

17.4 IN PTR erdos.maths.groucho.edu.

23.4 IN PTR gauss.maths.groucho.edu

; subnet 12: Physics Dept, separate zone

12 IN NS niels.physics.groucho.edu.

 IN NS gauss.maths.groucho.edu.

niels.physics.groucho.edu. IN A

gauss.maths.groucho.edu. IN A

in-addr.arpa system zones can only be created as supersets of IP networks. An even more severe restriction is that these networks' netmasks have to be on byte boundaries. All subnets at Groucho Marx University have a netmask of, hence an in-addr.arpa zone could be created for each subnet. However, if the netmask were instead, creating zones for the subnet would be impossible, because there's no way to tell DNS that the 12.76.149.in-addr.arpa domain has been split into two zones of authority, with hostnames ranging from 1 through 127, and 128 through 255, respectively.

Running named

named (pronounced name-dee) provides DNS on most Unix machines. It is a server program originally developed for BSD to provide name service to clients, and possibly to other name servers. BIND Version 4 was around for some time and appeared in most Linux distributions. The new release, Version 8, has been introduced in most Linux distributions, and is a big change from previous versions.[42] It has many new features, such as support for DNS dynamic updates, DNS change notifications, much improved performance, and a new configuration file syntax. Please check the documentation contained in the source distribution for details.

This section requires some understanding of the way DNS works. If the following discussion is all Greek to you, you may want to reread the section "How DNS Works".

named is usually started at system boot time and runs until the machine goes down again. Implementations of BIND prior to Version 8 take their information from a configuration file called /etc/named.boot and various files that map domain names to addresses. The latter are called zone files. Versions of BIND from Version 8 onwards use /etc/named.conf in place of /etc/named.boot.

To run named at the prompt, enter:

# /usr/sbin/named

named will come up and read the named.boot file and any zone files specified therein. It writes its process ID to /var/run/named.pid in ASCII, downloads any zone files from primary servers, if necessary, and starts listening on port 53 for DNS queries.

The named.boot File

The BIND configuration file prior to Version 8 was very simple in structure. BIND Version 8 has a very different configuration file syntax to deal with many of the new features introduced. The name of the configuration file changed from /etc/named.boot, in older versions of BIND, to /etc/named.conf in BIND Version 8. We'll focus on configuring the older version because it is probably what most distributions are still using, but we'll present an equivalent named.conf to illustrate the differences, and we'll talk about how to convert the old format into the new one.

The named.boot file is generally small and contains little but pointers to master files containing zone information and pointers to other name servers. Comments in the boot file start with the (#) or (;) characters and extend to the next newline. Before we discuss the format of named.boot in more detail, we will take a look at the sample file for vlager given in Example 6.8.

Example 6.8: The named.boot File for vlager


; /etc/named.boot file for vlager.vbrew.com


directory /var/named


; domain file


cache  .  named.ca

primary vbrew.com named.hosts

primary 0.0.127.in-addr.arpa named.local

primary 16.172.in-addr.arpa named.rev

Let's look at each statement individually. The directory keyword tells named that all filenames referred to later in this file, zone files for example, are located in the /var/named directory. This saves a little typing.

The primary keyword shown in this example loads information into named. This information is taken from the master files specified as the last of the parameters. These files represent DNS resource records, which we will look at next.

In this example, we configured named as the primary name server for three domains, as indicated by the three primary statements. The first of these statements instructs named to act as a primary server for vbrew.com, taking the zone data from the file named.hosts.

The cache keyword is very special and should be present on virtually all machines running a name server. It instructs named to enable its cache and to load the root name server hints from the cache file specified (named.ca in our example). We will come back to the name server hints in the following list.

Here's a list of the most important options you can use in named.boot:


This option specifies a directory in which zone files reside. Names of files in other options may be given relative to this directory. Several directories may be specified by repeatedly using directory. The Linux file system standard suggests this should be /var/named.


This option takes a domain name and filename as an argument, declaring the local server authoritative for the named domain. As a primary server, named loads the zone information from the given master file.

There will always be at least one primary entry in every boot file used for reverse mapping of network, which is the local loopback network.


This statement takes a domain name, an address list, and a filename as an argument. It declares the local server a secondary master server for the specified domain.

A secondary server holds authoritative data on the domain, too, but it doesn't gather it from files; instead, it tries to download it from the primary server. The IP address of at least one primary server thus must be given to named in the address list. The local server contacts each of them in turn until it successfully transfers the zone database, which is then stored in the backup file given as the third argument. If none of the primary servers responds, the zone data is retrieved from the backup file instead.

named then attempts to refresh the zone data at regular intervals. This process is explained later in connection with the SOA resource record type.


This option takes a domain name and filename as arguments. This file contains the root server hints, which is a list of records pointing to the root name servers. Only NS and A records will be recognized. The domain should be the root domain name, a simple period (.).

This information is absolutely crucial to named; if the cache statement does not occur in the boot file, named will not develop a local cache at all. This situation/lack of development will severely degrade performance and increase network load if the next server queried is not on the local net. Moreover, named will not be able to reach any root name servers, and thus won't resolve any addresses except those it is authoritative for. An exception from this rule involves forwarding servers (see the forwarders option that follows).


This statement takes a whitespace-separated list of addresses as an argument. The IP addresses in this list specify a list of name servers that named may query if it fails to resolve a query from its local cache. They are tried in order until one of them responds to the query. Typically, you would use the name server of your network provider or another well-known server as a forwarder.


This statement makes the name server a slave server. It never performs recursive queries itself, but only forwards them to servers specified in the forwarders statement.

There are two options that we will not describe here: sortlist and domain. Two other directives may also be used inside these database files: $INCLUDE and $ORIGIN. Since they are rarely needed, we will not describe them here, either.

The BIND 8 host.conf File

BIND Version 8 introduced a range of new features, and with these came a new configuration file syntax. The named.boot, with its simple single line statements, was replaced by the named.conf file, with a syntax like that of gated and resembling C source syntax.

The new syntax is more complex, but fortunately a tool has been provided that automates conversion from the old syntax to the new syntax. In the BIND 8 source package, a perl program called named-bootconf.pl is provided that will read your existing named.boot file from stdin and convert it into the equivalent named.conf format on stdout. To use it, you must have the perl interpreter installed.

You should use the script somewhat like this:

# cd /etc

# named-bootconf.pl ‹named.boot ›named.conf

The script then produces a named.conf that looks like that shown in Example 6.9. We've cleaned out a few of the helpful comments the script includes to help show the almost direct relationship between the old and the new syntax.

Example 6.9: The BIND 8 equivalent named.conf File for vlager


// /etc/named.boot file for vlager.vbrew.com

options {

 directory "/var/named";


zone "." {

 type hint;

 file "named.ca";


zone "vbrew.com" {

 type master;

 file "named.hosts";


zone "0.0.127.in-addr.arpa" {

 type master;

 file "named.local";


zone "16.172.in-addr.arpa" {

 type master;

 file "named.rev";


If you take a close look, you will see that each of the one-line statements in named.boot has been converted into a C-like statement enclosed within {} characters in the named.conf file.

The comments, which in the named.boot file were indicated by a semicolon (;), are now indicated by two forward slashes (//).

The directory statement has been translated into an options paragraph with a directory clause.

The cache and primary statements have been converted into zone paragraphs with type clauses of hint and master, respectively.

The zone files do not need to be modified in any way; their syntax remains unchanged.

The new configuration syntax allows for many new options that we haven't covered here. If you'd like information on the new options, the best source of information is the documentation supplied with the BIND Version 8 source package.

The DNS Database Files

Master files included with named, like named.hosts, always have a domain associated with them, which is called the origin. This is the domain name specified with the cache and primary options. Within a master file, you are allowed to specify domain and host names relative to this domain. A name given in a configuration file is considered absolute if it ends in a single dot, otherwise it is considered relative to the origin. The origin by itself may be referred to using (@).

The data contained in a master file is split up in resource records(RRs). RRs are the smallest units of information available through DNS. Each resource record has a type. A records, for instance, map a hostname to an IP address, and a CNAME record associates an alias for a host with its official hostname. To see an example, look at Example 6.11, which shows the named.hosts master file for the Virtual Brewery.

Resource record representations in master files share a common format:

[domain] [ttl] [class] type rdata

Fields are separated by spaces or tabs. An entry may be continued across several lines if an opening brace occurs before the first newline and the last field is followed by a closing brace. Anything between a semicolon and a newline is ignored. A description of the format terms follows:


This term is the domain name to which the entry applies. If no domain name is given, the RR is assumed to apply to the domain of the previous RR.


In order to force resolvers to discard information after a certain time, each RR is associated a time to live (ttl). The ttl field specifies the time in seconds that the information is valid after it has been retrieved from the server. It is a decimal number with at most eight digits.

If no ttl value is given, the field value defaults to that of the minimum field of the preceding SOA record.


This is an address class, like IN for IP addresses or HS for objects in the Hesiod class. For TCP/IP networking, you have to specify IN.

If no class field is given, the class of the preceding RR is assumed.


This describes the type of the RR. The most common types are A, SOA, PTR, and NS. The following sections describe the various types of RRs.


This holds the data associated with the RR. The format of this field depends on the type of RR. In the following discussion, it will be described for each RR separately.

The following is partial list of RRs to be used in DNS master files. There are a couple more of them that we will not explain; they are experimental and of little use, generally.


This RR describes a zone of authority (SOA means "Start of Authority"). It signals that the records following the SOA RR contain authoritative information for the domain. Every master file included by a primary statement must contain an SOA record for this zone. The resource data contains the following fields:


This field is the canonical hostname of the primary name server for this domain. It is usually given as an absolute name.


This field is the email address of the person responsible for maintaining the domain, with the "@" sign replaced by a dot. For instance, if the responsible person at the Virtual Brewery were janet, this field would contain janet.vbrew.com.


This field is the version number of the zone file, expressed as a single decimal number. Whenever data is changed in the zone file, this number should be incremented. A common convention is to use a number that reflects the date of the last update, with a version number appended to it to cover the case of multiple updates occurring on a single day, e.g., 2000012600 being update 00 that occurred on January 26, 2000.

The serial number is used by secondary name servers to recognize zone information changes. To stay up to date, secondary servers request the primary server's SOA record at certain intervals and compare the serial number to that of the cached SOA record. If the number has changed, the secondary servers transfer the whole zone database from the primary server.


This field specifies the interval in seconds that the secondary servers should wait between checking the SOA record of the primary server. Again, this is a decimal number with at most eight digits.

Generally, the network topology doesn't change too often, so this number should specify an interval of roughly a day for larger networks, and even more for smaller ones.


This number determines the intervals at which a secondary server should retry contacting the primary server if a request or a zone refresh fails. It must not be too low, or a temporary failure of the server or a network problem could cause the secondary server to waste network resources. One hour, or perhaps one-half hour, might be a good choice.


This field specifies the time in seconds after which a secondary server should finally discard all zone data if it hasn't been able to contact the primary server. You should normally set this field to at least a week (604,800 seconds), but increasing it to a month or more is also reasonable.


This field is the default ttl value for resource records that do not explicitly contain one. The ttl value specifies the maximum amount of time other name servers may keep the RR in their cache. This time applies only to normal lookups, and has nothing to do with the time after which a secondary server should try to update the zone information.

If the topology of your network does not change frequently, a week or even more is probably a good choice. If single RRs change more frequently, you could still assign them smaller ttls individually. If your network changes frequently, you may want to set minimum to one day (86,400 seconds).


This record associates an IP address with a hostname. The resource data field contains the address in dotted quad notation.

For each hostname, there must be only one A record. The hostname used in this A record is considered the official or canonical hostname. All other hostnames are aliases and must be mapped onto the canonical hostname using a CNAME record. If the canonical name of our host were vlager, we'd have an A record that associated that hostname with its IP address. Since we may also want another name associated with that address, say news, we'd create a CNAME record that associates this alternate name with the canonical name. We'll talk more about CNAME records shortly.


NS records are used to specify a zone's primary server and all its secondary servers. An NS record points to a master name server of the given zone, with the resource data field containing the hostname of the name server.

You will meet NS records in two situations: The first situation is when you delegate authority to a subordinate zone; the second is within the master zone database of the subordinate zone itself. The sets of servers specified in both the parent and delegated zones should match.

The NS record specifies the name of the primary and secondary name servers for a zone. These names must be resolved to an address so they can be used. Sometimes the servers belong to the domain they are serving, which causes a "chicken and egg" problem; we can't resolve the address until the name server is reachable, but we can't reach the name server until we resolve its address. To solve this dilemma, we can configure special A records directly into the name server of the parent zone. The A records allow the name servers of the parent domain to resolve the IP address of the delegated zone name servers. These records are commonly called glue records because they provide the "glue" that binds a delegated zone to its parent.


This record associates an alias with a host's canonical hostname. It provides an alternate name by which users can refer to the host whose canonical name is supplied as a parameter. The canonical hostname is the one the master file provides an A record for; aliases are simply linked to that name by a CNAME record, but don't have any other records of their own.


This type of record is used to associate names in the in-addr.arpa domain with hostnames. It is used for reverse mapping of IP addresses to hostnames. The hostname given must be the canonical hostname.


This RR announces a mail exchanger for a domain. Mail exchangers are discussed in "Mail Routing on the Internet". The syntax of an MX record is:

[domain] [ttl] [class] MX preference host

host names the mail exchanger for domain. Every mail exchanger has an integer preference associated with it. A mail transport agent that wants to deliver mail to domain tries all hosts who have an MX record for this domain until it succeeds. The one with the lowest preference value is tried first, then the others, in order of increasing preference value.


This record provides information on the system's hardware and software. Its syntax is:

[domain] [ttl] [class] HINFO hardware software

The hardware field identifies the hardware used by this host. Special conventions are used to specify this. A list of valid "machine names" is given in the Assigned Numbers RFC (RFC-1700). If the field contains any blanks, it must be enclosed in double quotes. The software field names the operating system software used by the system. Again, a valid name from the Assigned Numbers RFC should be chosen.

An HINFO record to describe an Intel-based Linux machine should look something like:

tao 36500 IN HINFO IBM-PC LINUX2.2

and HINFO records for Linux running on Motorola 68000-based machines might look like:

cevad 36500 IN HINFO ATARI-104ST LINUX2.0

jedd 36500 IN HINFO AMIGA-3000 LINUX2.0

Caching-only named Configuration

There is a special type of named configuration that we'll talk about before we explain how to build a full name server configuration. It is called a caching-only configuration. It doesn't really serve a domain, but acts as a relay for all DNS queries produced on your host. The advantage of this scheme is that it builds up a cache so only the first query for a particular host is actually sent to the name servers on the Internet. Any repeated request will be answered directly from the cache in your local name server. This may not seem useful yet, but it will when you are dialing in to the Internet, as described in Chapter 7, Serial Line IP and Chapter 8, The Point-to-Point Protocol.

A named.boot file for a caching-only server looks like this:

; named.boot file for caching-only server

directory    /var/named

primary 0.0.127.in-addr.arpa named.local; localhost network

cache  .  named.ca; root servers

In addition to this named.boot file, you must set up the named.ca file with a valid list of root name servers. You could copy and use Example 6.10 for this purpose. No other files are needed for a caching-only server configuration.

Writing the Master Files

Example 6.10, Example 6.11, Example 6.12, and Example 6.13 give sample files for a name server at the brewery, located on vlager. Due to the nature of the network discussed (a single LAN), the example is pretty straightforward.

The named.ca cache file shown in Example 6.10 shows sample hint records for a root name server. A typical cache file usually describes about a dozen name servers. You can obtain the current list of name servers for the root domain using the nslookup tool described in the next section.[43]

Example 6.10: The named.ca File


; /var/named/named.ca Cache file for the brewery.

; We're not on the Internet, so we don't need

; any root servers. To activate these; records, remove the semicolons.





























Example 6.11: The named.hosts File


; /var/named/named.hosts Local hosts at the brewery

; Origin is vbrew.com


@ IN SOA vlager.vbrew.com. janet.vbrew.com. (

 2000012601; serial

 86400; refresh: once per day

 3600; retry: one hour

 3600000; expire: 42 days

 604800; minimum: 1 week


 IN NS vlager.vbrew.com.


; local mail is distributed on vlager

 IN MX 10 vlager


; loopback address

localhost. IN A


; Virtual Brewery Ethernet

vlager IN A

vlager-if1 IN CNAME vlager

; vlager is also news server

news IN CNAME vlager

vstout IN A

vale IN A


; Virtual Winery Ethernet

vlager-if2 IN A

vbardolino IN A

vchianti IN A

vbeaujolais IN A


; Virtual Spirits (subsidiary) Ethernet

vbourbon IN A

vbourbon-if1 IN CNAME vbourbon

Example 6.12: The named.local File


; /var/named/named.local Reverse mapping of 127.0.0

; Origin is 0.0.127.in-addr.arpa.


@ IN SOA vlager.vbrew.com. joe.vbrew.com. (

 1; serial

 360000; refresh: 100 hrs

 3600; retry: one hour

 3600000; expire: 42 days

 360000; minimum: 100 hrs


  IN NS vlager.vbrew.com.

1 IN PTR localhost.

Example 6.13: The named.rev File


; /var/named/named.rev Reverse mapping of our IP addresses

; Origin is 16.172.in-addr.arpa.


@ IN SOA vlager.vbrew.com. joe.vbrew.com. (

 16; serial

 86400; refresh: once per day

 3600; retry: one hour

 3600000; expire: 42 days

 604800; minimum: 1 week


 IN NS vlager.vbrew.com.

; brewery

1.1 IN PTR vlager.vbrew.com.

2.1 IN PTR vstout.vbrew.com.

3.1 IN PTR vale.vbrew.com.

; winery

1.2 IN PTR vlager-if2.vbrew.com.

2.2 IN PTR vbardolino.vbrew.com.

3.2 IN PTR vchianti.vbrew.com.

4.2 IN PTR vbeaujolais.vbrew.com.

Verifying the Name Server Setup

nslookup is a great tool for checking the operation of your name server setup. It can be used both interactively with prompts and as a single command with immediate output. In the latter case, you simply invoke it as:

$ nslookup


nslookup queries the name server specified in resolv.conf for hostname. (If this file names more than one server, nslookup chooses one at random.)

The interactive mode, however, is much more exciting. Besides looking up individual hosts, you may query for any type of DNS record and transfer the entire zone information for a domain.

When invoked without an argument, nslookup displays the name server it uses and enters interactive mode. At the › prompt, you may type any domain name you want to query. By default, it asks for class A records, those containing the IP address relating to the domain name.

You can look for record types by issuing:

set type=type

in which type is one of the resource record names described earlier, or ANY.

You might have the following nslookup session:

$ nslookup

Default Server: tao.linux.org.au



Server: tao.linux.org.au


Name: metalab.unc.edu Address:

The output first displays the DNS server being queried, and then the result of the query.

If you try to query for a name that has no IP address associated with it, but other records were found in the DNS database, nslookup returns with an error message saying " No type A records found." However, you can make it query for records other than type A by issuing the set type command. To get the SOA record of unc.edu, you would issue:


Server: tao.linux.org.au


*** No address (A) records available for unc.edu

set type=SOA


Server: tao.linux.org.au



 origin = ns.unc.edu

 mail addr = host-reg.ns.unc.edu

 serial = 1998111011

 refresh = 14400 (4H)

 retry = 3600 (1H)

 expire = 1209600 (2W)

 minimum ttl = 86400 (1D)

unc.edu name server = ns2.unc.edu

unc.edu name server = ncnoc.ncren.net

unc.edu name server = ns.unc.edu

ns2.unc.edu internet address =

ncnoc.ncren.net internet address =

ncnoc.ncren.net internet address =

ns.unc.edu internet address =

In a similar fashion, you can query for MX records:

set type=MX


Server: tao.linux.org.au


unc.edu preference = 0, mail exchanger = conga.oit.unc.edu

unc.edu preference = 10, mail exchanger = imsety.oit.unc.edu

unc.edu name server = ns.unc.edu

unc.edu name server = ns2.unc.edu

unc.edu name server = ncnoc.ncren.net

conga.oit.unc.edu internet address =

imsety.oit.unc.edu internet address =

ns.unc.edu internet address =

ns2.unc.edu internet address =

ncnoc.ncren.net internet address =

ncnoc.ncren.net internet address =

Using a type of ANY returns all resource records associated with a given name.

A practical application of nslookup, besides debugging, is to obtain the current list of root name servers. You can obtain this list by querying for all NS records associated with the root domain:

set type=NS


Server: tao.linux.org.au


Non-authoritative answer:

 (root) name server = A.ROOT-SERVERS.NET

 (root) name server = H.ROOT-SERVERS.NET

 (root) name server = B.ROOT-SERVERS.NET

 (root) name server = C.ROOT-SERVERS.NET

 (root) name server = D.ROOT-SERVERS.NET

 (root) name server = E.ROOT-SERVERS.NET

 (root) name server = I.ROOT-SERVERS.NET

 (root) name server = F.ROOT-SERVERS.NET

 (root) name server = G.ROOT-SERVERS.NET

 (root) name server = J.ROOT-SERVERS.NET

 (root) name server = K.ROOT-SERVERS.NET

 (root) name server = L.ROOT-SERVERS.NET

 (root) name server = M.ROOT-SERVERS.NET

Authoritative answers can be found from:

A.ROOT-SERVERS.NET internet address =

H.ROOT-SERVERS.NET internet address =

B.ROOT-SERVERS.NET internet address =

C.ROOT-SERVERS.NET internet address =

D.ROOT-SERVERS.NET internet address =

E.ROOT-SERVERS.NET internet address =

I.ROOT-SERVERS.NET internet address =

F.ROOT-SERVERS.NET internet address =

G.ROOT-SERVERS.NET internet address =

J.ROOT-SERVERS.NET internet address =

K.ROOT-SERVERS.NET internet address =

L.ROOT-SERVERS.NET internet address =

M.ROOT-SERVERS.NET internet address =

To see the complete set of available commands, use the help command in nslookup.

Other Useful Tools

There are a few tools that can help you with your tasks as a BIND administrator. We will briefly describe two of them here. Please refer to the documentation that comes with these tools for more information on how to use them.

hostcvt helps you with your initial BIND configuration by converting your /etc/hosts file into master files for named. It generates both the forward (A) and reverse mapping (PTR) entries, and takes care of aliases. Of course, it won't do the whole job for you, as you may still want to tune the timeout values in the SOA record, for example, or add MX records. Still, it may help you save a few aspirins. hostcvt is part of the BIND source, but can also be found as a standalone package on a few Linux FTP servers.

After setting up your name server, you may want to test your configuration. Some good tools that make this job much simpler: the first is called dnswalk, which is a Perl-based package. The second is called nslint. They both walk your DNS database looking for common mistakes and verify that the information they find is consistent. Two other useful tools are host and dig, which are general purpose DNS database query tools. You can use these tools to manually inspect and diagnose DNS database entries.

These tools are likely to be available in prepackaged form. dnswalk and nslint are available in source from http://www.visi.com/~barr/dnswalk/ and ftp://ftp.ee.lbl.gov/nslint.tar.Z. The host and dig source codes can be found at ftp://ftp.nikhef.nl/pub/network/ and ftp://ftp.is.co.za/networking/ip/dns/dig/.

Chapter 7. Serial Line IP

Packet protocols like IP or IPX rely upon the receiver host knowing where the start and end of each packet are in the data stream. The mechanism used to mark and detect the start and end of packets is called delimitation. The Ethernet protocol manages this mechanism in a LAN environment, and the SLIP and PPP protocols manage it for serial communications lines.

The comparatively low cost of low-speed dialup modems and telephone circuits has made the serial line IP protocols immensely popular, especially for providing connectivity to end users of the Internet. The hardware required to run SLIP or PPP is simple and readily available. All that is required is a modem and a serial port equipped with a FIFO buffer.

The SLIP protocol is very simple to implement and at one time was the more common of the two. Today almost everyone uses the PPP protocol instead. The PPP protocol adds a host of sophisticated features that contribute to its popularity today, and we'll look at the most important of these later.

Linux supports kernel-based drivers for both SLIP and PPP. The drivers have both been around for some time and are stable and reliable. In this chapter and the next, we'll discuss both protocols and how to configure them.

General Requirements

To use SLIP or PPP, you have to configure some basic networking features as described in the previous chapters. You must set up the loopback interface and configure the name resolver. When connecting to the Internet, you will want to use DNS. Your options here are the same as for PPP: you can perform your DNS queries across your serial link by configuring your Internet Service Provider's IP address into your /etc/resolv.conf file, or configure a caching-only name server as described under "Caching-only named Configuration" in Chapter 6, "Name Service and Resolver Configuration."

SLIP Operation

Dialup IP servers frequently offer SLIP service through special user accounts. After logging in to such an account, you are not dropped into the common shell; instead, a program or shell script is executed that enables the server's SLIP driver for the serial line and configures the appropriate network interface. Then you have to do the same at your end of the link.

On some operating systems, the SLIP driver is a user-space program; under Linux, it is part of the kernel, which makes it a lot faster. This speed requires, however, that the serial line be converted to the SLIP mode explicitly. This conversion is done by means of a special tty line discipline, SLIPDISC. While the tty is in normal line discipline (DISC0), it exchanges data only with user processes, using the normal read(2) and write(2) calls, and the SLIP driver is unable to write to or read from the tty. In SLIPDISC, the roles are reversed: now any user-space processes are blocked from writing to or reading from the tty, while all data coming in on the serial port is passed directly to the SLIP driver.

The SLIP driver itself understands a number of variations on the SLIP protocol. Apart from ordinary SLIP, it also understands CSLIP, which performs the so-called Van Jacobson header compression (described in RFC-1144) on outgoing IP packets. This compression improves throughput for interactive sessions noticeably. There are also six-bit versions for each of these protocols.

A simple way to convert a serial line to SLIP mode is by using the slattach tool. Assume you have your modem on /dev/ttyS3 and have logged in to the SLIP server successfully. You will then execute:

# slattach /dev/ttyS3 &

This tool switches the line discipline of ttyS3 to SLIPDISC and attaches it to one of the SLIP network interfaces. If this is your first active SLIP link, the line will be attached to sl0; the second will be attached to sl1, and so on. The current kernels support a default maximum of 256 simultaneous SLIP links.

The default line discipline chosen by slattach is CSLIP. You may choose any other discipline using the -p switch. To use normal SLIP (no compression), you use:

# slattach -p slip /dev/ttyS3 &

The disciplines available are listed in Table 7.1. A special pseudo-discipline is available called adaptive, which causes the kernel to automatically detect which type of SLIP encapsulation is being used by the remote end.

Table 7.1: Linux Slip-Line Disciplines

Disclipline Description
slip Traditional SLIP encapsulation.
cslip SLIP encapsulation with Van Jacobsen header compression.
slip6 SLIP encapsulation with six-bit encoding. The encoding method is similar to that used by the uuencode command, and causes the SLIP datagram to be converted into printable ASCII characters. This conversion is useful when you do not have a serial link that is eight bit clean.
cslip6 SLIP encapsulation with Van Jacobsen header compression and six-bit encoding.
adaptive This is not a real line discipline; instead, it causes the kernel to attempt to identify the line discipline being used by the remote machine and to match it.

Note that you must use the same encapsulation as your peer. For example, if cowslip uses CSLIP, you also have to do so. If your SLIP connection doesn't work, the first thing you should do is ensure that both ends of the link agree on whether to use header compression or not. If you are unsure what the remote end is using, try configuring your host for adaptive slip. The kernel might figure out the right type for you.

slattach lets you enable not only SLIP, but other protocols that use the serial line, like PPP or KISS (another protocol used by ham radio people). Doing this is not common, though, and there are better tools available to support these protocols. For details, please refer to the slattach(8) manual page.

After turning over the line to the SLIP driver, you must configure the network interface. Again, you do this using the standard ifconfig and route commands. Assume that we have dialed up a server named cowslip from vlager. On vlager you would execute:

# ifconfig sl0 vlager-slip pointopoint cowslip

# route add cowslip

# route add default gw cowslip

The first command configures the interface as a point-to-point link to cowslip, while the second and third add the route to cowslip and the default route, using cowslip as a gateway.

Two things are worth noting about the ifconfig invocation: The pointopoint option that specifies the address of the remote end of a point-to-point link and our use of vlager-slip as the address of the local SLIP interface.

We have mentioned that you can use the same address you assigned to vlager's Ethernet interface for your SLIP link, as well. In this case, vlager-slip might just be another alias for address However, it is also possible that you have to use an entirely different address for your SLIP link. One such case is when your network uses an unregistered IP network address, as the Brewery does. We will return to this scenario in greater detail in the next section.

For the remainder of this chapter we will always use vlager-slip to refer to the address of the local SLIP interface.

When taking down the SLIP link, you should first remove all routes through cowslip using route with the del option, then take the interface down, and send slattach the hangup signal. The you must hang up the modem using your terminal program again:

# route del default

# route del cowslip

# ifconfig sl0 down

# kill -HUP 516

Note that the 516 should be replaced with the process id (as shown in the output of ps ax) of the slattach command controlling the slip device you wish to take down.

Dealing with Private IP Networks

You will remember from Chapter 5, Configuring TCP/IP Networking, that the Virtual Brewery has an Ethernet-based IP network using unregistered network numbers that are reserved for internal use only. Packets to or from one of these networks are not routed on the Internet; if we were to have vlager dial into cowslip and act as a router for the Virtual Brewery network, hosts within the Brewery's network could not talk to real Internet hosts directly because their packets would be dropped silently by the first major router.

To work around this dilemma, we will configure vlager to act as a kind of launch pad for accessing Internet services. To the outside world, it will present itself as a normal SLIP-connected Internet host with a registered IP address (probably assigned by the network provider running cowslip). Anyone logged in to vlager can use text-based programs like ftp, telnet, or even lynx to make use of the Internet. Anyone on the Virtual Brewery LAN can therefore telnet and log in to vlager and use the programs there. For some applications, there may be solutions that avoid logging in to vlager. For WWW users, for example, we could run a so-called proxy server on vlager, which would relay all requests from your users to their respective servers.

Having to log in to vlager to make use of the Internet is a little clumsy. But apart from eliminating the paperwork (and cost) of registering an IP network, it has the added benefit of going along well with a firewall setup. Firewalls are dedicated hosts used to provide limited Internet access to users on your local network without exposing the internal hosts to network attacks from the outside world. Simple firewall configuration is covered in more detail in Chapter 9, TCP/IP Firewall. In Chapter 11, IP Masquerade and Network Address Translation, we'll discuss a Linux feature called "IP masquerade" that provides a powerful alternative to proxy servers.

Assume that the Brewery has been assigned the IP address for SLIP access. All you have to do to realize that the setup discussed above is to enter this address into your /etc/hosts file, naming it vlager-slip. The procedure for bringing up the SLIP link itself remains unchanged.

Using dip

Now that was rather simple. Nevertheless, you might want to automate the steps previously described. It would be much better to have a simple command that performs all the steps necessary to open the serial device, cause the modem to dial the provider, log in, enable the SLIP line discipline, and configure the network interface. This is what the dip command is for.

dip means Dialup IP. It was written by Fred van Kempen and has been patched very heavily by a number of people. Today there is one strain that is used by almost everyone: Version dip337p-uri, which is included with most modern Linux distributions, or is available from the metalab.unc.edu FTP archive.

dip provides an interpreter for a simple scripting language that can handle the modem for you, convert the line to SLIP mode, and configure the interfaces. The script language is powerful enough to suit most configurations.

To be able to configure the SLIP interface, dip requires root privilege. It would now be tempting to make dip setuid to root so that all users can dial up some SLIP server without having to give them root access. This is very dangerous, though, because setting up bogus interfaces and default routes with dip may disrupt routing on your network. Even worse, this action would give your users power to connect to any SLIP server and launch dangerous attacks on your network. If you want to allow your users to fire up a SLIP connection, write small wrapper programs for each prospective SLIP server and have these wrappers invoke dip with the specific script that establishes the connection. Carefully written wrapper programs can then safely be made setuid to root.[44] An alternative, more flexible approach is to give trusted users root access to dip using a program like sudo.

A Sample Script

Assume that the host to which we make our SLIP connection is cowslip, and that we have written a script for dip to run called cowslip.dip that makes our connection. We invoke dip with the script name as argument:

# dip cowslip.dip

DIP: Dialup IP Protocol Driver version 3.3.7 (12/13/93)

Written by Fred N. van Kempen, MicroWalt Corporation.

connected to cowslip.moo.com with addr


The script itself is shown in Example 7.1.

Example 7.1: A Sample dip Script

# Sample dip script for dialing up cowslip

# Set local and remote name and address

 get $local vlager-slip

 get $remote cowslip

 port ttyS3 # choose a serial port speed

 38400 # set speed to max

 modem HAYES # set modem type

 reset # reset modem and tty

 flush # flush out modem response

# Prepare for dialing.

 send ATQ0V1E1X1\r

wait OK 2

 if $errlvl != 0 goto error

 dial 41988

 if $errlvl != 0 goto error

 wait CONNECT 60

 if $errlvl != 0 goto error

# Okay, we're connected now

 sleep 3

 send \r\n\r\n

 wait ogin: 10

 if $errlvl != 0 goto error

 send Svlager\n

 wait ssword: 5

 if $errlvl != 0 goto error

 send knockknock\n

 wait running 30

 if $errlvl != 0 goto error

# We have logged in, and the remote side is firing up SLIP.

 print Connected to $remote with address $rmtip

 default # Make this link our default route

 mode SLIP # We go to SLIP mode, too

# fall through in case of error


 print SLIP to $remote failed.

After connecting to cowslip and enabling SLIP, dip will detach from the terminal and go to the background. You can then start using the normal networking services on the SLIP link. To terminate the connection, simply invoke dip with the -k option. This sends a hangup signal to dip, using the process ID dip records in /etc/dip.pid:

# dip -k

In dip 's scripting language, keywords prefixed with a dollar symbol denote variable names. dip has a predefined set of variables, which will be listed below. $remote and $local, for instance, contain the hostnames of the remote and local hosts involved in the SLIP link.

The first two statements in the sample script are get commands, which is dip 's way to set a variable. Here, the local and remote hostnames are set to vlager and cowslip, respectively.

The next five statements set up the terminal line and the modem. reset sends a reset string to the modem. The next statement flushes out the modem response so that the login chat in the next few lines works properly. This chat is pretty straightforward: it simply dials 41988, the phone number of cowslip, and logs in to the account Svlager using the password knockknock. The wait command makes dip wait for the string given as its first argument; the number given as its second argument makes the wait time out after that many seconds if no such string is received. The if commands interspersed in the login procedure check that no error occurred while executing the command.

The final commands executed after logging in are default, which makes the SLIP link the default route to all hosts, and mode, which enables SLIP mode on the line and configures the interface and routing table for you.

A dip Reference

In this section, we will give a reference for most of dip 's commands. You can get an overview of all the commands it provides by invoking dip in test mode and entering the help command. To learn about the syntax of a command, you may enter it without any arguments. Remember that this does not work with commands that take no arguments. The following example illustrates the help command:

# dip -t

DIP: Dialup IP Protocol Driver version 3.3.7p-uri (25 Dec 96)

Written by Fred N. van Kempen, MicroWalt Corporation.

Debian version 3.3.7p-2 (debian).


DIP knows about the following commands:

 beep bootp break chatkey config

 databits dec default dial echo

 flush get goto help if

 inc init mode modem netmask

 onexit parity password proxyarp print

 psend port quit reset securidfixed

 securid send shell skey sleep

 speed stopbits term timeout wait

DIP› echo

Usage: echo on|off


Throughout the following section, examples that display the DIP› prompt show how to enter a command in test mode and what output it produces. Examples lacking this prompt should be taken as script excerpts.

The modem commands

dip provides a number of commands that configure your serial line and modem. Some of these are obvious, such as port, which selects a serial port, and speed, databits, stopbits, and parity, which set the common line parameters. The modem command selects a modem type. Currently, the only type supported is HAYES (capitalization required). You have to provide dip with a modem type, or else it will refuse to execute the dial and reset commands. The reset command sends a reset string to the modem; the string used depends on the modem type selected. For Hayes-compatible modems, this string is ATZ.

The flush code can be used to flush out all responses the modem has sent so far. Otherwise, a chat script following reset might be confused because it reads the OK responses from earlier commands.

The init command selects an initialization string to be passed to the modem before dialing. The default for Hayes modems is "ATE0 Q0 V1 X1", which turns on echoing of commands and long result codes, and selects blind dialing (no checking of dial tone). Modern modems have a good factory default configuration, so this is a little unnecessary, though it does no harm.

The dial command sends the initialization string to the modem and dials up the remote system. The default dial command for Hayes modems is ATD.

The echo command

The echo command serves as a debugging aid. Calling echo on makes dip echo to the console everything it sends to the serial device. This can be turned off again by calling echo off.

dip also allows you to leave script mode temporarily and enter terminal mode. In this mode, you can use dip just like any ordinary terminal program, writing the characters you type to the serial line, reading data from the serial line, and displaying the characters. To leave this mode, enter Ctrl-].

The get command

The get command is dip 's way of setting a variable. The simplest form is to set a variable to a constant, as we did in cowslip.dip. You may, however, also prompt the user for input by specifying the keyword ask instead of a value:

DIP› get $local ask

Enter the value for $local: _

A third method is to obtain the value from the remote host. Bizarre as it seems at first, this is very useful in some cases. Some SLIP servers will not allow you to use your own IP address on the SLIP link, but will rather assign you one from a pool of addresses whenever you dial in, printing some message that informs you about the address you have been assigned. If the message looks something like " Your address: ", the following piece of dip code would let you pick up the address:

# finish login

wait address: 10

get $locip remote

The print command

This is the command used to echo text to the console from which dip was started. Any of dip 's variables may be used in print commands. Here's an example:

DIP› print Using port $port at speed $speed

Using port ttyS3 at speed 38400

Variable names

dip understands only a predefined set of variables. A variable name always begins with a dollar symbol and must be written in lowercase letters.

The $local and $locip variables contain the local host's name and IP address. When you store the canonical hostname in $local, dip will automatically attempt to resolve the hostname to an IP address and to store it in the $locip variable. A similar but backward process occurs when you assign an IP address to the $locip variable; dip will attempt to perform a reverse lookup to identify the name of the host and store it in the $local variable.

The $remote and $rmtip variables operate in the same way for the remote host's name and address. $mtu contains the MTU value for the connection.

These five variables are the only ones that may be assigned values directly using the get command. A number of other variables are set as a result of the configuration commands bearing the same name, but may be used in print statements; these variables are $modem, $port, and $speed.

$errlvl is the variable through which you can access the result of the last command executed. An error level of 0 indicates success, while a nonzero value denotes an error.

The if and goto commands

The if command is a conditional branch, rather than a full-featured programming if statement. Its syntax is:

if var op number goto label

The expression must be a simple comparison between one of the variables $errlvl, $locip, and $rmtip. var must be an integer number; the operator op may be one of -,!=, ‹, ›, ‹=, and ›=.

The goto command makes the execution of the script continue at the line following that bearing the label. A label must be the first word on the line and must be followed immediately by a colon.

send, wait, and sleep

These commands help implement simple chat scripts in dip. The send command outputs its arguments to the serial line. It does not support variables, but understands all C-style backslash character sequences, such as \n for newline and \b for backspace. The tilde character (~) can be used as an abbreviation for carriage return/newline.

The wait command takes a word as an argument and will read all input on the serial line until it detects a sequence of characters that match this word. The word itself may not contain any blanks. Optionally, you may give wait a timeout value as a second argument; if the expected word is not received within that many seconds, the command will return with an $errlvl value of 1. This command is used to detect login and other prompts.

The sleep command may be used to wait for a certain amount of time; for instance, to patiently wait for any login sequence to complete. Again, the interval is specified in seconds.

mode and default

These commands are used to flip the serial line to SLIP mode and configure the interface.

The mode command is the last command executed by dip before going into daemon mode. Unless an error occurs, the command does not return.

mode takes a protocol name as argument. dip currently recognizes SLIP, CSLIP, SLIP6, CSLIP6, PPP, and TERM as valid names. The current version of dip does not understand adaptive SLIP, however.

After enabling SLIP mode on the serial line, dip executes ifconfig to configure the interface as a point-to-point link, and invokes route to set the route to the remote host.

If, in addition, the script executes the default command before mode, dip creates a default route that points to the SLIP link.

Running in Server Mode

Setting up your SLIP client was the hard part. Configuring your host to act as a SLIP server is much easier.

There are two ways of configuring a SLIP server. Both ways require that you set up one login account per SLIP client. Assume you provide SLIP service to Arthur Dent at dent.beta.com. You might create an account named dent by adding the following line to your passwd file:

dent:*:501:60:Arthur Dent's SLIP account:/tmp:/usr/sbin/diplogin

Afterwards, you would set dent's password using the passwd utility.

The dip command can be used in server mode by invoking it as diplogin. Usually diplogin is a link to dip. Its main configuration file is /etc/diphosts, which is where you specify what IP address a SLIP user will be assigned when he or she dials in. Alternatively, you can also use the sliplogin command, a BSD-derived tool featuring a more flexible configuration scheme that lets you execute shell scripts whenever a host connects and disconnects.

When our SLIP user dent logs in, dip starts up as a server. To find out if he is indeed permitted to use SLIP, it looks up the username in /etc/diphosts. This file details the access rights and connection parameter for each SLIP user. The general format for an /etc/diphosts entry looks like:

# /etc/diphosts


Each of the fields is described in Table 7.2.

Table 7.2: /etc/diphosts Field Description

Field Description
user The username of the user invoking dip that this entry will apply to.
password Field 2 of the /etc/diphosts file is used to add an extra layer of password-based security on the connection. You can place a password in encrypted form here (just as in /etc/passwd) and diplogin will prompt for the user to enter the password before allowing SLIP access. Note that this password is used in addition to the normal login -based password the user will enter.
rem-addr The address that will be assigned to the remote machine. This address may be specified either as a hostname that will be resolved or an IP address in dotted quad notation.
loc-addr The IP address that will be used for this end of the SLIP link. This may also be specified as a resolvable hostname or in dotted quad format.
netmask The netmask that will be used for routing purposes. Many people are confused by this entry. The netmask doesn't apply to the SLIP link itself, but is used in combination with the rem-addr field to produce a route to the remote site. The netmask should be that used by the network supported by that of the remote host.
comments This field is free-form text that you may use to help document the /etc/diphosts file. It serves no other purpose.
protocol This field is where you specify what protocol or line discipline you want applied to this connection. Valid entries here are the same as those valid for the -p argument to the slattach command.
MTU The maximum transmission unit that this link will carry. This field describes the largest datagram that will be transmitted across the link. Any datagram routed to the SLIP device that is larger than the MTU will be fragmented into datagrams no larger than this value. Usually, the MTU is configured identically at both ends of the link.

A sample entry for dent could look like this:

dent::dent.beta.com:vbrew.com: Dent:CSLIP,296

Our example gives our user dent access to SLIP with no additional password required. He will be assigned the IP address associated with dent.beta.com with a netmask of His default route should be directed to the IP address of vbrew.com, and he will use the CSLIP protocol with an MTU of 296 bytes.

When dent logs in, diplogin extracts the information on him from the diphosts file. If the second field contains a value, diplogin will prompt for an "external security password." The string entered by the user is encrypted and compared to the password from diphosts. If they do not match, the login attempt is rejected. If the password field contains the string s/key, and dip was compiled with S/Key support, S/Key authentication will take place. S/Key authentication is described in the documentation that comes in the dip source package.

After a successful login, diplogin proceeds by flipping the serial line to CSLIP or SLIP mode, and sets up the interface and route. This connection remains established until the user disconnects and the modem drops the line. diplogin then returns the line to normal line discipline and exits.

diplogin requires superuser privilege. If you don't have dip running setuid root, you should make diplogin a separate copy of dip instead of a simple link. diplogin can then safely be made setuid without affecting the status of dip itself.

Chapter 8. The Point-to-Point Protocol

Like SLIP, PPP is a protocol used to send datagrams across a serial connection; however, it addresses a couple of the deficiencies of SLIP. First, it can carry a large number of protocols and is thus not limited to the IP protocol. It provides error detection on the link itself, while SLIP accepts and forwards corrupted datagrams as long as the corruption does not occur in the header. Equally important, it lets the communicating sides negotiate options, such as the IP address and the maximum datagram size at startup time, and provides client authorization. This built-in negotiation allows reliable automation of the connection establishment, while the authentication removes the need for the clumsy user login accounts that SLIP requires. For each of these capabilities, PPP has a separate protocol. In this chapter, we briefly cover these basic building blocks of PPP. This discussion of PPP is far from complete; if you want to know more about PPP, we urge you to read its RFC specification and the dozen or so companion RFCs.[45] There is also a comprehensive O'Reilly book on the topic of Using & Managing PPP, by Andrew Sun.

At the very bottom of PPP is the High-Level Data Link Control (HDLC) protocol, which defines the boundaries around the individual PPP frames and provides a 16-bit checksum.[46] As opposed to the more primitive SLIP encapsulation, a PPP frame is capable of holding packets from protocols other than IP, such as Novell's IPX or Appletalk. PPP achieves this by adding a protocol field to the basic HDLC frame that identifies the type of packet carried by the frame.

The Link Control Protocol, (LCP) is used on top of HDLC to negotiate options pertaining to the data link. For instance, the Maximum Receive Unit (MRU), states the maximum datagram size that one side of the link agrees to receive.

An important step at the configuration stage of a PPP link is client authorization. Although it is not mandatory, it is really a must for dialup lines in order to keep out intruders. Usually the called host (the server) asks the client to authorize itself by proving it knows some secret key. If the caller fails to produce the correct secret, the connection is terminated. With PPP, authorization works both ways; the caller may also ask the server to authenticate itself. These authentication procedures are totally independent of each other. There are two protocols for different types of authorization, which we will discuss further in this chapter: Password Authentication Protocol (PAP) and Challenge Handshake Authentication Protocol (CHAP).

Each network protocol that is routed across the data link (like IP and AppleTalk) is configured dynamically using a corresponding Network Control Protocol (NCP). To send IP datagrams across the link, both sides running PPP must first negotiate which IP address each of them uses. The control protocol used for this negotiation is the Internet Protocol Control Protocol (IPCP).

Besides sending standard IP datagrams across the link, PPP also supports Van Jacobson header compression of IP datagrams. This technique shrinks the headers of TCP packets to as little as three bytes. It is also used in CSLIP, and is more colloquially referred to as VJ header compression. The use of compression may be negotiated at startup time through IPCP, as well.

PPP on Linux

On Linux, PPP functionality is split into two parts: a kernel component that handles the low-level protocols (HDLC, IPCP, IPXCP, etc.) and the user space pppd daemon that handles the various higher-level protocols, such as PAP and CHAP. The current release of the PPP software for Linux contains the PPP daemon pppd and a program named chat that automates the dialing of the remote system.

The PPP kernel driver was written by Michael Callahan and reworked by Paul Mackerras. pppd was derived from a free PPP implementation[47] for Sun and 386BSD machines that was written by Drew Perkins and others, and is maintained by Paul Mackerras. It was ported to Linux by Al Longyear. chat was written by Karl Fox.[48]

Like SLIP, PPP is implemented by a special line discipline. To use a serial line as a PPP link, you first establish the connection over your modem as usual, and subsequently convert the line to PPP mode. In this mode, all incoming data is passed to the PPP driver, which checks the incoming HDLC frames for validity (each HDLC frame carries a 16-bit checksum), and unwraps and dispatches them. Currently, PPP is able to transport both the IP protocol, optionally using Van Jacobson header compression, and the IPX protocol.

pppd aids the kernel driver, performing the initialization and authentication phase that is necessary before actual network traffic can be sent across the link. pppd 's behavior may be fine-tuned using a number of options. As PPP is rather complex, it is impossible to explain all of them in a single chapter. This book therefore cannot cover all aspects of pppd, but only gives you an introduction. For more information, consult Using& Managing PPP or the pppd manual pages, and READMEs in the pppd source distribution, which should help you sort out most questions this chapter fails to discuss. The PPP-HOWTO might also be of use.

Probably the greatest help you will find in configuring PPP will come from other users of the same Linux distribution. PPP configuration questions are very common, so try your local usergroup mailing list or the IRC Linux channel. If your problems persist even after reading the documentation, you could try the comp.protocols.ppp newsgroup. This is the place where you can find most of the people involved in pppd development.

Running pppd

When you want to connect to the Internet through a PPP link, you have to set up basic networking capabilities, such as the loopback device and the resolver. Both have been covered in Chapter 5, Configuring TCP/IP Networking, and Chapter 6, Name Service and Resolver Configuration. You can simply configure the name server of your Internet Service Provider in the /etc/resolv.conf file, but this will mean that every DNS request is sent across your serial link. This situation is not optimal; the closer (network-wise) you are to your name server, the faster the name lookups will be. An alternative solution is to configure a caching-only name server at a host on your network. This means that the first time you make a DNS query for a particular host, your request will be sent across your serial link, but every subsequent request will be answered directly by your local name server, and will be much faster. This configuration is described in Chapter 6, in "Caching-only named Configuration".

As an introductory example of how to establish a PPP connection with pppd, assume you are at vlager again. First, dial in to the PPP server c3po and log in to the ppp account. c3po will execute its PPP driver. After exiting the communications program you used for dialing, execute the following command, substituting the name of the serial device you used for the ttyS3 shown here:

# pppd /dev/ttyS3 38400 crtscts defaultroute

This command flips the serial line ttyS3 to the PPP line discipline and negotiates an IP link with c3po. The transfer speed used on the serial port will be 38,400 bps. The crtscts option turns on hardware handshake on the port, which is an absolute must at speeds above 9,600 bps.

The first thing pppd does after starting up is negotiate several link characteristics with the remote end using LCP. Usually, the default set of options pppd tries to negotiate will work, so we won't go into this here. Expect to say that part of this negotiation involves requesting or assigning the IP addresses at each end of the link.

For the time being, we also assume that c3po doesn't require any authentication from us, so the configuration phase is completed successfully.

pppd will then negotiate the IP parameters with its peer using IPCP, the IP control protocol. Since we didn't specify any particular IP address to pppd earlier, it will try to use the address obtained by having the resolver look up the local hostname. Both will then announce their addresses to each other.

Usually, there's nothing wrong with these defaults. Even if your machine is on an Ethernet, you can use the same IP address for both the Ethernet and the PPP interface. Nevertheless, pppd allows you to use a different address, or even to ask your peer to use some specific address. These options are discussed later in the "IP Configuration Options" section.

After going through the IPCP setup phase, pppd will prepare your host's networking layer to use the PPP link. It first configures the PPP network interface as a point-to-point link, using ppp0 for the first PPP link that is active, ppp1 for the second, and so on. Next, it sets up a routing table entry that points to the host at the other end of the link. In the previous example, pppd made the default network route point to c3po, because we gave it the defaultroute option.[49] The default route simplifies your routing by causing any IP datagram destined to a nonlocal host to be sent to c3po; this makes sense since it is the only way they can be reached. There are a number of different routing schemes pppd supports, which we will cover in detail later in this chapter.

Using Options Files

Before pppd parses its command-line arguments, it scans several files for default options. These files may contain any valid command-line arguments spread out across an arbitrary number of lines. Hash signs introduce comments.

The first options file is /etc/ppp/options, which is always scanned when pppd starts up. Using it to set some global defaults is a good idea, because it allows you to keep your users from doing several things that may compromise security. For instance, to make pppd require some kind of authentication (either PAP or CHAP) from the peer, you add the auth option to this file. This option cannot be overridden by the user, so it becomes impossible to establish a PPP connection with any system that is not in your authentication databases. Note, however, that some options can be overridden; the connect string is a good example.

The other options file, which is read after /etc/ppp/options, is .ppprc in the user's home directory. It allows each user to specify her own set of default options.

A sample /etc/ppp/options file might look like this:

# Global options for pppd running on vlager.vbrew.com

lock # use UUCP-style device locking

auth # require authentication

usehostname # use local hostname for CHAP

domain vbrew.com # our domain name

The lock keyword makes pppd comply to the standard UUCP method of device locking. With this convention, each process that accesses a serial device, say /dev/ttyS3, creates a lock file with a name like LCK…ttyS3 in a special lock-file directory to signal that the device is in use. This is necessary to prevent signal other programs, such as minicom or uucico, from opening the serial device while it is used by PPP.

The next three options relate to authentication and, therefore, to system security. The authentication options are best placed in the global configuration file because they are "privileged" and cannot be overridden by users' ~/.ppprc options files.

Using chat to Automate Dialing

One of the things that may have struck you as inconvenient in the previous example is that you had to establish the connection manually before you could fire up pppd. Unlike dip, pppd does not have its own scripting language for dialing the remote system and logging in, but relies on an external program or shell script to do this. The command to be executed can be given to pppd with the connect command-line option. pppd will redirect the command's standard input and output to the serial line.

The pppd software package is supplied with a very simple program called chat, which is capable of being used in this way to automate simple login sequences. We'll talk about this command in some detail.

If your login sequence is complex, you will need something more powerful than chat. One useful alternative you might consider is expect, written by Don Libes. It has a very powerful language based on Tcl, and was designed exactly for this sort of application. Those of you whose login sequence requires, for example, challenge/response authentication involving calculator-like key generators will find expect powerful enough to handle the task. Since there are so many possible variations on this theme, we won't describe how to develop an appropriate expect script in this book. Suffice it to say, you'd call your expect script by specifying its name using the pppd connect option. It's also important to note that when the script is running, the standard input and output will be attached to the modem, not to the terminal that invoked pppd. If you require user interaction, you should manage it by opening a spare virtual terminal, or arrange some other means.

The chat command lets you specify a UUCP-style chat script. Basically, a chat script consists of an alternating sequence of strings that we expect to receive from the remote system, and the answers we are to send. We will call them expect and send strings, respectively. This is a typical excerpt from a chat script:

ogin: b1ff ssword: s3|‹r1t

This script tells chat to wait for the remote system to send the login prompt and return the login name b1ff. We wait only for ogin: so that it doesn't matter if the login prompt starts with an uppercase or lowercase l, or if it arrives garbled. The following string is another expect string that makes chat wait for the password prompt and send our response password.

This is basically what chat scripts are all about. A complete script to dial up a PPP server would, of course, also have to include the appropriate modem commands. Assume that your modem understands the Hayes command set, and the server's telephone number is 318714. The complete chat invocation to establish a connection with c3po would then be:

$ chat -v '' ATZ OK ATDT318714 CONNECT '' ogin: ppp word: GaGariN

By definition, the first string must be an expect string, but as the modem won't say anything before we have kicked it, we make chat skip the first expect by specifying an empty string. We then send ATZ, the reset command for Hayes-compatible modems, and wait for its response (OK). The next string sends the dial command along with the phone number to chat, and expects the CONNECT message in response. This is followed by an empty string again because we don't want to send anything now, but rather wait for the login prompt. The remainder of the chat script works exactly as described previously. This description probably looks a bit confusing, but we'll see in a moment that there is a way to make chat scripts a lot easier to understand.

The -v option makes chat log all activities to the syslog daemon local2 facility.[50]

Specifying the chat script on the command line bears a certain risk because users can view a process's command line with the ps command. You can avoid this risk by putting the chat script in a file like dial-c3po. You make chat read the script from the file instead of the command line by giving it the -f option, followed by the filename. This action has the added benefit of making our chat expect sequences easier to understand. To convert our example, our dial-c3po file would look like:

'' ATZ

OK ATDT318714


ogin: ppp

word: GaGariN

When we use a chat script file in this way, the string we expect to receive is on the left and the response we will send is on the right. They are much easier to read and understand when presented this way.

The complete pppd incantation would now look like this:

# pppd connect "chat -f dial-c3po" /dev/ttyS3 38400 -detach \

 crtscts modem defaultroute

Besides the connect option that specifies the dialup script, we have added two more options to the command line: -detach, which tells pppd not to detach from the console and become a background process, and the modem keyword, which makes it perform modem-specific actions on the serial device, like disconnecting the line before and after the call. If you don't use this keyword, pppd will not monitor the port's DCD line and will therefore not detect whether the remote end hangs up unexpectedly.

The examples we have shown are rather simple; chat allows for much more complex scripts. For instance, it can specify strings on which to abort the chat with an error. Typical abort strings are messages like BUSY or NO CARRIER that your modem usually generates when the called number is busy or doesn't answer. To make chat recognize these messages immediately rather than timing out, you can specify them at the beginning of the script using the ABORT keyword:


Similarly, you can change the timeout value for parts of the chat scripts by inserting TIMEOUT options.

Sometimes you also need to have conditional execution for parts of the chat script: when you don't receive the remote end's login prompt, you might want to send a BREAK or a carriage return. You can achieve this by appending a subscript to an expect string. The subscript consists of a sequence of send and expect strings, just like the overall script itself, which are separated by hyphens. The subscript is executed whenever the expected string it is appended to is not received in time. In the example above, we would modify the chat script as follows:

ogin:-BREAK-ogin: ppp ssword: GaGariN

When chat doesn't see the remote system send the login prompt, the subscript is executed by first sending a BREAK, and then waiting for the login prompt again. If the prompt now appears, the script continues as usual; otherwise, it will terminate with an error.

IP Configuration Options

IPCP is used to negotiate a number of IP parameters at link configuration time. Usually, each peer sends an IPCP Configuration Request packet, indicating which values it wants to change from the defaults and the new value. Upon receipt, the remote end inspects each option in turn and either acknowledges or rejects it.

pppd gives you a lot of control over which IPCP options it will try to negotiate. You can tune it through various command-line options that we will discuss in this section.

Choosing IP Addresses

All IP interfaces require IP addresses assigned to them; a PPP device always has an IP address. The PPP suite of protocols provides a mechanism that allows the automatic assignment of IP addresses to PPP interfaces. It is possible for the PPP program at one end of a point-to-point link to assign an IP address for the remote end to use, or each may use its own.

Some PPP servers that handle a lot of client sites assign addresses dynamically; addresses are assigned to systems only when calling in and are reclaimed after they have logged off again. This allows the number of IP addresses required to be limited to the number of dialup lines. While limitation is convenient for managers of the PPP dialup server, it is often less convenient for users who are dialing in. We discussed the way that hostnames are mapped to IP addresses by use of a database in Chapter 6. In order for people to connect to your host, they must know your IP address or the hostname associated with it. If you are a user of a PPP service that assigns you an IP address dynamically, this knowledge is difficult without providing some means of allowing the DNS database to be updated after you are assigned an IP address. Such systems do exist, but we won't cover them in detail here; instead, we will look at the more preferable approach, which involves you being able to use the same IP address each time you establish your network connection.[51]

In the previous example, we had pppd dial up c3po and establish an IP link. No provisions were taken to choose a particular IP address on either end of the link. Instead, we let pppd take its default action. It attempts to resolve the local hostname, vlager in our example, to an IP address, which it uses for the local end, while letting the remote machine, c3po, provide its own. PPP supports several alternatives to this arrangement.

To ask for particular addresses, you generally provide pppd with the following option:


local_addr and remote_addr may be specified either in dotted quad notation or as hostnames.[52] This option makes pppd attempt to use the first address supplied as its own IP address, and the second as the peer's. If the peer rejects either of the addresses during IPCP negotiation, no IP link will be established.[53]

If you are dialing in to a server and expect it to assign you an IP address, you should ensure that pppd does not attempt to negotiate one for itself. To do this, use the noipdefault option and leave the local_addr blank. The noipdefault option will stop pppd from trying to use the IP address associated with the hostname as the local address.

If you want to set only the local address but accept any address the peer uses, simply leave out the remote_addr part. To make vlager use the IP address instead of its own, give it on the command line. Similarly, to set the remote address only, leave the local_addr field blank. By default, pppd will then use the address associated with your hostname.

Routing Through a PPP Link

After setting up the network interface, pppd will usually set up a host route to its peer only. If the remote host is on a LAN, you certainly want to be able to connect to hosts "behind" your peer as well; in that case, a network route must be set up.

We have already seen that pppd can be asked to set the default route using the defaultroute option. This option is very useful if the PPP server you dialed up acts as your Internet gateway.

The reverse case, in which your system acts as a gateway for a single host, is also relatively easy to accomplish. For example, take some employee at the Virtual Brewery whose home machine is called oneshot. Let's also assume that we've configured vlager as a dialin PPP server. If we've configured vlager to dynamically assign an IP address that belongs to the Brewery's subnet, then we can use the proxyarp option with pppd, which will install a proxy ARP entry for oneshot. This automatically makes oneshot accessible from all hosts at the Brewery and the Winery.

However, things aren't always that simple. Linking two local area networks usually requires adding a specific network route because these networks may have their own default routes. Besides, having both peers use the PPP link as the default route would generate a loop, through which packets to unknown destinations would ping-pong between the peers until their time to live expired.

Suppose the Virtual Brewery opens a branch in another city. The subsidiary runs an Ethernet of its own using the IP network number, which is subnet 3 of the Brewery's class B network. The subsidiary wants to connect to the Brewery's network via PPP to update customer databases. Again, vlager acts as the gateway for the brewery network and will support the PPP link; its peer at the new branch is called vbourbon and has an IP address of This network is illustrated in Figure 24.2 in Appendix A, Example Network: The Virtual Brewery.

When vbourbon connects to vlager, it makes the default route point to vlager as usual. On vlager, however, we will have only the point-to-point route to vbourbon and will have to specially configure a network route for subnet 3 that uses vbourbon as its gateway. We could do this manually using the route command by hand after the PPP link is established, but this is not a very practical solution. Fortunately, we can configure the route automatically by using a feature of pppd that we haven't discussed yet - the ip-up command. This command is a shell script or program located in /etc/ppp that is executed by pppd after the PPP interface has been configured. When present, it is invoked with the following parameters:

ip-up iface device speed local_addr remote_addr

The following table summarizes the meaning of each of the arguments (in the first column, we show the number used by the shell script to refer to each argument):

Argument Name Purpose
$1 iface The network interface used, e.g., ppp0
$2 device The pathname of the serial device file used (/dev/tty, if stdin/stdout are used)
$3 speed The speed of the serial device in bits per second
$4 local_addr The IP address of the link's remote end in dotted quad notation
$5 remote_addr The IP address of the remote end of the link in dotted quad notation #!/bin/sh case $5 in # this is vbourbon route add -net gw;;… esac exit 0

In our case, the ip-up script may contain the following code fragment:[54]


case $5 in            # this is vbourbon

        route add -net gw;;



exit 0

Similarly, /etc/ppp/ip-down can be used to undo any actions of ip-up after the PPP link has been taken down again. So in our /etc/ppp/ip-down script we would have a route command that removed the route we created in the /etc/ppp/ip-up script.

However, the routing scheme is not yet complete. We have set up routing table entries on both PPP hosts, but so far none of the hosts on either network knows anything about the PPP link. This is not a big problem if all hosts at the subsidiary have their default route pointing at vbourbon, and all Brewery hosts route to vlager by default. If this is not the case, your only option is usually to use a routing daemon like gated. After creating the network route on vlager, the routing daemon broadcasts the new route to all hosts on the attached subnets.

Link Control Options

We already encountered the Link Control Protocol (LCP), which is used to negotiate link characteristics and test the link.

The two most important options negotiated by LCP are the Asynchronous Control Character Map and the Maximum Receive Unit. There are a number of other LCP configuration options, but they are far too specialized to discuss here.

The Asynchronous Control Character Map, colloquially called the async map, is used on asynchronous links, such as telephone lines, to identify control characters that must be escaped (replaced by a specific two-character sequence) to avoid them being interpreted by equipment used to establish the link. For instance, you may want to avoid the XON and XOFF characters used for software handshake because a misconfigured modem might choke upon receipt of an XOFF. Other candidates include Ctrl-l (the telnet escape character). PPP allows you to escape any of the characters with ASCII codes 0 through 31 by specifying them in the async map.

The async map is a 32-bit-wide bitmap expressed in hexadecimal. The least significant bit corresponds to the ASCII NULL character, and the most significant bit corresponds to ASCII 31 decimal. These 32 ASCII characters are the control characters. If a bit is set in the bitmap, it signals that the corresponding character must be escaped before it is transmitted across the link.

To tell your peer that it doesn't have to escape all control characters, but only a few of them, you can specify an async map to pppd using the asyncmap option. For example, if only ^S and ^Q (ASCII 17 and 19, commonly used for XON and XOFF) must be escaped, use the following option:

asyncmap 0x000A0000

The conversion is simple as long as you can convert binary to hex. Lay out 32 bits in front of you. The right-most bit corresponds to ASCII 00 (NULL), and the left-most bit corresponds to ASCII 32 decimal. Set the bits corresponding to the characters you want escaped to one, and all others to zero. To convert that into the hexadecimal number pppd expects, simply take each set of 4 bits and convert them into hex. You should end up with eight hexadecimal figures. String them all together and preprend "0x" to signify it is a hexadecimal number, and you are done.

Initially, the async map is set to 0xffffffff - that is, all control characters will be escaped. This is a safe default, but is usually much more than you need. Each character that appears in the async map results in two characters being transmitted across the link, so escaping comes at the cost of increased link utilization and a corresponding performance reduction.

In most circumstances, an async map of 0x0 works fine. No escaping is performed.

The Maximum Receive Unit (MRU), signals to the peer the maximum size of HDLC frames we want to receive. Although this may remind you of the Maximum Transfer Unit (MTU) value, these two have little in common. The MTU is a parameter of the kernel networking device and describes the maximum frame size the interface is able to transmit. The MRU is more of an advice to the remote end not to generate frames larger than the MRU; the interface must nevertheless be able to receive frames of up to 1,500 bytes.

Choosing an MRU is therefore not so much a question of what the link is capable of transferring, but of what gives you the best throughput. If you intend to run interactive applications over the link, setting the MRU to values as low as 296 is a good idea, so that an occasional larger packet (say, from an FTP session) doesn't make your cursor "jump." To tell pppd to request an MRU of 296, you give it the option mru 296. Small MRUs, however, make sense only if you have VJ header compression (it is enabled by default), because otherwise you'd waste a large amount of your bandwidth just carrying the IP header for each datagram.

pppd also understands a couple of LCP options that configure the overall behavior of the negotiation process, such as the maximum number of configuration requests that may be exchanged before the link is terminated. Unless you know exactly what you are doing, you should leave these options alone.

Finally, there are two options that apply to LCP echo messages. PPP defines two messages, Echo Request and Echo Response. pppd uses this feature to check if a link is still operating. You can enable this by using the lcp-echo-interval option together with a time in seconds. If no frames are received from the remote host within this interval, pppd generates an Echo Request and expects the peer to return an Echo Response. If the peer does not produce a response, the link is terminated after a certain number of requests are sent. This number can be set using the lcp-echo-failure option. By default, this feature is disabled altogether.

General Security Considerations

A misconfigured PPP daemon can be a devastating security breach. It can be as bad as letting anyone plug their machine into your Ethernet (and that can be very bad). In this section, we discuss a few measures that should make your PPP configuration safe.

NOTE: Root privilege is required to configure the network device and routing table. You will usually solve this by running pppd setuid root. However, pppd allows users to set various security-relevant options.

To protect against any attacks a user may launch by manipulating pppd options, you should set a couple of default values in the global /etc/ppp/options file, like those shown in the sample file in "Using Options Files", earlier in this chapter. Some of them, such as the authentication options, cannot be overridden by the user, and thus provide reasonable protection against manipulations. An important option to protect is the connect option. If you intend to allow non-root users to invoke pppd to connect to the Internet, you should always add the connect and noauth options to the global options file /etc/ppp/options. If you fail to do this, users will be able to execute arbitrary commands with root privileges by specifying the command as their connect command on the pppd line or in their personal options file.

Another good idea is to restrict which users may execute pppd by creating a group in /etc/group and adding only those users who you wish to have the ability to execute the PPP daemon. You should then change group ownership of the pppd daemon to that group and remove the world execute privileges. To do this, assuming you've called your group dialout, you could use something like:

# chown root /usr/sbin/pppd

# chgrp dialout /usr/sbin/pppd

# chmod 4750 /usr/sbin/pppd

Of course, you have to protect yourself from the systems you speak PPP with, too. To fend off hosts posing as someone else, you should always require some sort of authentication from your peer. Additionally, you should not allow foreign hosts to use any IP address they choose, but restrict them to at most a few. The following section will deal with these topics in detail.

Authentication with PPP 

With PPP, each system may require its peer to authenticate itself using one of two authentication protocols: the Password Authentication Protocol (PAP), and the Challenge Handshake Authentication Protocol (CHAP). When a connection is established, each end can request the other to authenticate itself, regardless of whether it is the caller or the callee. In the description that follows, we will loosely talk of "client" and "server" when we want to distinguish between the system sending authentication requests and the system responding to them. A PPP daemon can ask its peer for authentication by sending yet another LCP configuration request identifying the desired authentication protocol.


PAP, which is offered by many Internet Service Providers, works basically the same way as the normal login procedure. The client authenticates itself by sending a username and a (optionally encrypted) password to the server, which the server compares to its secrets database.[55] This technique is vulnerable to eavesdroppers, who may try to obtain the password by listening in on the serial line, and to repeated trial and error attacks.

CHAP does not have these deficiencies. With CHAP, the server sends a randomly generated "challenge" string to the client, along with its hostname. The client uses the hostname to look up the appropriate secret, combines it with the challenge, and encrypts the string using a one-way hashing function. The result is returned to the server along with the client's hostname. The server now performs the same computation, and acknowledges the client if it arrives at the same result.

CHAP also doesn't require the client to authenticate itself only at startup time, but sends challenges at regular intervals to make sure the client hasn't been replaced by an intruder, for instance by switching phone lines, or because of a modem configuration error that causes the PPP daemon not to notice that the original phone call has dropped out and someone else has dialed in.

pppd keeps the secret keys for PAP and CHAP in two separate files called /etc/ppp/pap-secrets and /etc/ppp/chap-secrets. By entering a remote host in one or the other file, you have fine control over whether PAP or CHAP is used to authenticate yourself with your peer, and vice versa.

By default, pppd doesn't require authentication from the remote host, but it will agree to authenticate itself when requested by the remote host. Since CHAP is so much stronger than PAP, pppd tries to use the former whenever possible. If the peer does not support it, or if pppd can't find a CHAP secret for the remote system in its chap-secrets file, it reverts to PAP. If it doesn't have a PAP secret for its peer either, it refuses to authenticate altogether. As a consequence, the connection is shut down.

You can modify this behavior in several ways. When given the auth keyword, pppd requires the peer to authenticate itself. pppd agrees to use either CHAP or PAP as long as it has a secret for the peer in its CHAP or PAP database. There are other options to turn a particular authentication protocol on or off, but I won't describe them here.

If all systems you talk to with PPP agree to authenticate themselves with you, you should put the auth option in the global /etc/ppp/options file and define passwords for each system in the chap-secrets file. If a system doesn't support CHAP, add an entry for it to the pap-secrets file. That way, you can make sure no unauthenticated system connects to your host.

The next two sections discuss the two PPP secrets files, pap-secrets and chap-secrets. They are located in /etc/ppp and contain triplets of clients, servers, and passwords, optionally followed by a list of IP addresses. The interpretation of the client and server fields is different for CHAP and PAP, and also depends on whether we authenticate ourselves with the peer, or whether we require the server to authenticate itself with us.

The CHAP Secrets File

When it has to authenticate itself with a server using CHAP, pppd searches the chap-secrets file for an entry with the client field equal to the local hostname, and the server field equal to the remote hostname sent in the CHAP challenge. When requiring the peer to authenticate itself, the roles are simply reversed: pppd then looks for an entry with the client field equal to the remote hostname (sent in the client's CHAP response), and the server field equal to the local hostname.

The following is a sample chap-secrets file for vlager:[56]

# CHAP secrets for vlager.vbrew.com


# client         server           secret                addrs


vlager.vbrew.com  c3po.lucas.com   "Use The Source Luke" vlager.vbrew.com

c3po.lucas.com    vlager.vbrew.com "arttoo! arttoo!"     c3po.lucas.com

*                 vlager.vbrew.com "TuXdrinksVicBitter"  pub.vbrew.com

When vlager establishes a PPP connection with c3po, c3po asks vlager to authenticate itself by sending a CHAP challenge. pppd on vlager then scans chap-secrets for an entry with the client field equal to vlager.vbrew.com and the server field equal to c3po.lucas.com, and finds the first line shown in the example.[57] It then produces the CHAP response from the challenge string and the secret (Use The Source Luke), and sends it off to c3po.

pppd also composes a CHAP challenge for c3po containing a unique challenge string and its fully qualified hostname, vlager.vbrew.com. c3po constructs a CHAP response in the way we discussed, and returns it to vlager. pppd then extracts the client hostname (c3po.vbrew.com) from the response and searches the chap-secrets file for a line matching c3po as a client and vlager as the server. The second line does this, so pppd combines the CHAP challenge and the secret arttoo! arttoo!, encrypts them, and compares the result to c3po's CHAP response.

The optional fourth field lists the IP addresses that are acceptable for the client named in the first field. The addresses can be given in dotted quad notation or as hostnames that are looked up with the resolver. For instance, if c3po asks to use an IP address during IPCP negotiation that is not in this list, the request is rejected, and IPCP is shut down. In the sample file shown above, c3po is therefore limited to using its own IP address. If the address field is empty, any addresses are allowed; a value of " - " prevents the use of IP with that client altogether.

The third line of the sample chap-secrets file allows any host to establish a PPP link with vlager because a client or server field of * is a wildcard matching any hostname. The only requirements are that the connecting host must know the secret and that it must use the IP address associated with pub.vbrew.com. Entries with wildcard hostnames may appear anywhere in the secrets file, since pppd will always use the best match it can find for the server/client pair.

pppd may need some help forming hostnames. As explained before, the remote hostname is always provided by the peer in the CHAP challenge or response packet. The local hostname is obtained by calling the gethostname(2) function by default. If you have set the system name to your unqualified hostname, you also have to provide pppd with the domain name using the domain option:

# pppd… domain vbrew.com

This provision appends the Brewery's domain name to vlager for all authentication related activities. Other options that modify pppd 's idea of the local hostname are usehostname and name. When you give the local IP address on the command line using local:remote and local as a name instead of a dotted quad, pppd uses this as the local hostname.

The PAP Secrets File

The PAP secrets file is very similar to CHAP's. The first two fields always contain a username and a server name; the third holds the PAP secret. When the remote host sends its authentication information, pppd uses the entry that has a server field equal to the local hostname, and a user field equal to the username sent in the request. When it is necessary for us to send our credentials to the peer, pppd uses the secret that has a user field equal to the local username and the server field equal to the remote hostname.

A sample PAP secrets file might look like this:

# /etc/ppp/pap-secrets


# user          server          secret          addrs

vlager-pap      c3po            cresspahl       vlager.vbrew.com

c3po            vlager          DonaldGNUth     c3po.lucas.com

The first line is used to authenticate ourselves when talking to c3po. The second line describes how a user named c3po has to authenticate itself with us.

The name vlager-pap in the first column is the username we send to c3po. By default, pppd picks the local hostname as the username, but you can also specify a different name by giving the user option followed by that name.

When picking an entry from the pap-secrets file to identify us to a remote host, pppd must know the remote host's name. As it has no way of finding that out, you must specify it on the command line using the remotename keyword followed by the peer's hostname. To use the above entry for authentication with c3po, for example, we must add the following option to pppd 's command line:

# pppd… remotename c3po user vlager-pap

In the fourth field of the PAP secrets file (and all following fields), you can specify what IP addresses are allowed for that particular host, just as in the CHAP secrets file. The peer will be allowed to request only addresses from that list. In the sample file, the entry that c3po will use when it dials in - the line where c3po is the client - allows it to use its real IP address and no other.

Note that PAP is a rather weak authentication method, you should use CHAP instead whenever possible. We will therefore not cover PAP in greater detail here; if you are interested in using it, you will find more PAP features in the pppd(8) manual page.

Debugging Your PPP Setup

By default, pppd logs any warnings and error messages to syslog 's daemon facility. You have to add an entry to syslog.conf that redirects these messages to a file or even the console; otherwise, syslog simply discards them. The following entry sends all messages to /var/log/ppp-log:

daemon.* /var/log/ppp-log

If your PPP setup doesn't work right away, you should look in this log file. If the log messages don't help, you can also turn on extra debugging output using the debug option. This output makes pppd log the contents of all control packets sent or received to syslog. All messages then go to the daemon facility.

Finally, the most drastic way to check a problem is to enable kernel-level debugging by invoking pppd with the kdebug option. It is followed by a numeric argument that is the sum of the following values: 1 for general debug messages, 2 for printing the contents of all incoming HDLC frames, and 4 to make the driver print all outgoing HDLC frames. To capture kernel debugging messages, you must either run a syslogd daemon that reads the /proc/kmsg file, or the klogd daemon. Either of them directs kernel debugging to the syslog kernel facility.

More Advanced PPP Configurations

While configuring PPP to dial in to a network like the Internet is the most common application, there are those of you who have more advanced requirements. In this section we'll talk about a few of the more advanced configurations possible with PPP under Linux.

PPP Server

Running pppd as a server is just a matter of configuring a serial tty device to invoke pppd with appropriate options when an incoming data call has been received. One way to do this is to create a special account, say ppp, and give it a script or program as a login shell that invokes pppd with these options. Alternatively, if you intend to support PAP or CHAP authentication, you can use the mgetty program to support your modem and exploit its "/AutoPPP/" feature.

To build a server using the login method, you add a line similar to the following to your /etc/passwd file:[58]

ppp:x:500:200:Public PPP Account:/tmp:/etc/ppp/ppplogin

If your system supports shadow passwords, you also need to add an entry to the /etc/shadow file:


Of course, the UID and GID you use depends on which user you wish to own the connection, and how you've created it. You also have to set the password for the mentioned account using the passwd command.

The ppplogin script might look like this:


# ppplogin - script to fire up pppd on login

mesg n

stty -echo

exec pppd -detach silent modem crtscts

The mesg command disables other users from writing to the tty by using, for instance, the write command. The stty command turns off character echoing. This command is necessary; otherwise, everything the peer sends would be echoed back to it. The most important pppd option given is -detach because it prevents pppd from detaching from the controlling tty. If we didn't specify this option, it would go to the background, making the shell script exit. This in turn would cause the serial line to hang up and the connection to be dropped. The silent option causes pppd to wait until it receives a packet from the calling system before it starts sending. This option prevents transmit timeouts from occurring when the calling system is slow in firing up its PPP client. The modem option makes pppd drive the modem control lines of the serial port. You should always turn this option on when using pppd with a modem. The crtscts option turns on hardware handshake.

Besides these options, you might want to force some sort of authentication, for example, by specifying auth on pppd 's command line or in the global options file. The manual page also discusses more specific options for turning individual authentication protocols on and off.

If you wish to use mgetty, all you need to do is configure mgetty to support the serial device your modem is connected to (see "Configuring the mgetty Daemon" for details), configure pppd for either PAP or CHAP authentication with appropriate options in its options file, and finally, add a section similar to the following to your /etc/mgetty/login.config file:

# Configure mgetty to automatically detect incoming PPP calls and invoke

# the pppd daemon to handle the connection.


/AutoPPP/ - ppp /usr/sbin/pppd auth -chap +pap login

The first field is a special piece of magic used to detect that an incoming call is a PPP one. You must not change the case of this string; it is case sensitive. The third column is the username that appears in who listings when someone has logged in. The rest of the line is the command to invoke. In our example, we've ensured that PAP authentication is required, disabled CHAP, and specified that the system passwd file should be used for authenticating users. This is probably similar to what you'll want. Remember, you can specify the options in the options file or on the command line if you prefer.

Here is a small checklist of tasks to perform and the sequence you should perform them to get PPP dial in working on your machine. Make sure each step works before moving on to the next:

1. Configure the modem for auto-answer mode. On Hayes-compatible modems, this is performed using a command like ATS0=3. If you're going to be using the mgetty daemon, this isn't necessary.

2. Configure the serial device with a getty type of command to answer incoming calls. A commonly used getty variant is mgetty.

3. Consider authentication. Will your callers authenticate using PAP, CHAP, or system login?

4. Configure pppd as server as described in this section.

5. Consider routing. Will you need to provide a network route to callers? Routing can be performed using the ip-up script.

Demand Dialing

When there is IP traffic to be carried across the link, demand dialing causes your telephone modem to dial and to establish a connection to a remote host. Demand dialing is most useful when you can't leave your telephone line permanently switched to your Internet provider. For example, you might have to pay timed local calls, so it might be cheaper to have the telephone line switched on only when you need it and disconnected when you aren't using the Internet.

Traditional Linux solutions have used the diald command, which worked well but was fairly tricky to configure. Versions 2.3.0 and later of the PPP daemon have built-in support for demand dialing and make it very simple to configure. You must use a modern kernel for this to work, too. Any of the later 2.0 kernels will work just fine.

To configure pppd for demand dialing, all you need to do is add options to your options file or the pppd command line. The following table summarizes the options related to demand dialing:

Option Description
demand This option specifies that the PPP link should be placed in demand dial mode. The PPP network device will be created, but the connect command will not be used until a datagram is transmitted by the local host. This option is mandatory for demand dialing to work.
active-filter expression This option allows you to specify which data packets are to be considered active traffic. Any traffic matching the specified rule will restart the demand dial idle timer, ensuring that pppd waits again before closing the link. The filter syntax has been borrowed from the tcpdump command. The default filter matches all datagrams.
holdoff n This option allows you to specify the minimum amount of time, in seconds, to wait before reconnecting this link if it terminates. If the connection fails while pppd believes it is in active use, it will be re-established after this timer has expired. This timer does not apply to reconnections after an idle timeout.
idle n If this option is configured, pppd will disconnect the link whenever this timer expires. Idle times are specified in seconds. Each new active data packet will reset the timer.

A simple demand dialing configuration would therefore look something like this:


holdoff 60

idle 180

This configuration would enable demand dialing, wait 60 seconds before re-establishing a failed connection, and drop the link if 180 seconds pass without any active data on the link.

Persistent Dialing

Persistent dialing is what people who have permanent dialup connections to a network will want to use. There is a subtle difference between demand dialing and persistent dialing. With persistent dialing, the connection is automatically established as soon as the PPP daemon is started, and the persistent aspect comes into play whenever the telephone call supporting the link fails. Persistent dialing ensures that the link is always available by automatically rebuilding the connection if it fails.

You might be fortunate to not have to pay for your telephone calls; perhaps they are local and free, or perhaps they're paid by your company. The persistent dialing option is extremely useful in this situation. If you do have to pay for your telephone calls, then you have to be a little careful. If you pay for your telephone calls on a time-charged basis, persistent dialing is almost certainly not what you want, unless you're very sure you'll be using the connection fairly steadily twenty-four hours a day. If you do pay for calls, but they are not time charged, you need to be careful to protect yourself against situations that might cause the modem to endlessly redial. The pppd daemon provides an option that can help reduce the effects of this problem.

To enable persistent dialing, you must include the persist option in one of your pppd options files. Including this option alone is all you need to have pppd automatically invoke the command specified by the connect option to rebuild the connection when the link fails. If you are concerned about the modem redialing too rapidly (in the case of modem or server fault at the other end of the connection), you can use the holdoff option to set the minimum amount of time that pppd will wait before attempting to reconnect. This option won't solve the problem of a fault costing you money in wasted phone calls, but it will at least serve to reduce the impact of one.

A typical configuration might have persistent dialing options that look like this:

persist holdoff 600

The holdoff time is specified in seconds. In our example, pppd waits a full five minutes before redialing after the call drops out.

It is possible to combine persistent dialing with demand dialing, using idle to drop the link if it has been idle for a specified period of time. We doubt many users would want to do so, but this scenario is described briefly in the pppd manual page, if you'd like to pursue it.

Chapter 9. TCP/IP Firewall

Security is increasingly important for companies and individuals alike. The Internet has provided them with a powerful tool to distribute information about themselves and obtain information from others, but it has also exposed them to dangers that they have previously been exempt from. Computer crime, information theft, and malicious damage are all potential dangers.

An unauthorized and unscrupulous person who gains access to a computer system may guess system passwords or exploit the bugs and idiosyncratic behavior of certain programs to obtain a working account on that machine. Once they are able to log in to the machine, they may have access to information that may be damaging, such as commercially sensitive information like marketing plans, new project details, or customer information databases. Damaging or modifying this type of data can cause severe setbacks to the company.

The safest way to avoid such widespread damage is to prevent unauthorized people from gaining network access to the machine. This is where firewalls come in.

WARNING: Constructing secure firewalls is an art. It involves a good understanding of technology, but equally important, it requires an understanding of the philosophy behind firewall designs. We won't cover everything you need to know in this book; we strongly recommend you do some additional research before trusting any particular firewall design, including any we present here.

There is enough material on firewall configuration and design to fill a whole book, and indeed there are some good resources that you might like to read to expand your knowledge on the subject. Two of these are:

Building Internet Firewalls by D. Chapman and E. Zwicky (O'Reilly). A guide explaining how to design and install firewalls for Unix, Linux, and Windows NT, and how to configure Internet services to work with the firewalls.

Firewalls and Internet Security by W. Cheswick and S. Bellovin (Addison Wesley). This book covers the philosophy of firewall design and implementation.

We will focus on the Linux-specific technical issues in this chapter. Later we will present a sample firewall configuration that should serve as a useful starting point in your own configuration, but as with all security-related matters, trust no one. Double check the design, make sure you understand it, and then modify it to suit your requirements. To be safe, be sure.

Methods of Attack 

As a network administrator, it is important that you understand the nature of potential attacks on computer security. We'll briefly describe the most important types of attacks so that you can better understand precisely what the Linux IP firewall will protect you against. You should do some additional reading to ensure that you are able to protect your network against other types of attacks. Here are some of the more important methods of attack and ways of protecting yourself against them:

Unauthorized access

This simply means that people who shouldn't use your computer services are able to connect and use them. For example, people outside your company might try to connect to your company accounting machine or to your NFS server.

There are various ways to avoid this attack by carefully specifying who can gain access through these services. You can prevent network access to all except the intended users.

Exploitation of known weaknesses in programs

Some programs and network services were not originally designed with strong security in mind and are inherently vulnerable to attack. The BSD remote services (rlogin, rexec, etc.) are an example.

The best way to protect yourself against this type of attack is to disable any vulnerable services or find alternatives. With Open Source, it is sometimes possible to repair the weaknesses in the software.

Denial of service

Denial of service attacks cause the service or program to cease functioning or prevent others from making use of the service or program. These may be performed at the network layer by sending carefully crafted and malicious datagrams that cause network connections to fail. They may also be performed at the application layer, where carefully crafted application commands are given to a program that cause it to become extremely busy or stop functioning.

Preventing suspicious network traffic from reaching your hosts and preventing suspicious program commands and requests are the best ways of minimizing the risk of a denial of service attack. It's useful to know the details of the attack method, so you should educate yourself about each new attack as it gets publicized.


This type of attack causes a host or application to mimic the actions of another. Typically the attacker pretends to be an innocent host by following IP addresses in network packets. For example, a well-documented exploit of the BSD rlogin service can use this method to mimic a TCP connection from another host by guessing TCP sequence numbers.

To protect against this type of attack, verify the authenticity of datagrams and commands. Prevent datagram routing with invalid source addresses. Introduce unpredictablility into connection control mechanisms, such as TCP sequence numbers and the allocation of dynamic port addresses.


This is the simplest type of attack. A host is configured to "listen" to and capture data not belonging to it. Carefully written eavesdropping programs can take usernames and passwords from user login network connections. Broadcast networks like Ethernet are especially vulnerable to this type of attack.

To protect against this type of threat, avoid use of broadcast network technologies and enforce the use of data encryption.

IP firewalling is very useful in preventing or reducing unauthorized access, network layer denial of service, and IP spoofing attacks. It not very useful in avoiding exploitation of weaknesses in network services or programs and eavesdropping.

What Is a Firewall?

A firewall is a secure and trusted machine that sits between a private network and a public network.[59] The firewall machine is configured with a set of rules that determine which network traffic will be allowed to pass and which will be blocked or refused. In some large organizations, you may even find a firewall located inside their corporate network to segregate sensitive areas of the organization from other employees. Many cases of computer crime occur from within an organization, not just from outside.

Firewalls can be constructed in quite a variety of ways. The most sophisticated arrangement involves a number of separate machines and is known as a perimeter network. Two machines act as "filters" called chokes to allow only certain types of network traffic to pass, and between these chokes reside network servers such as a mail gateway or a World Wide Web proxy server. This configuration can be very safe and easily allows quite a great range of control over who can connect both from the inside to the outside, and from the outside to the inside. This sort of configuration might be used by large organizations.

Typically though, firewalls are single machines that serve all of these functions. These are a little less secure, because if there is some weakness in the firewall machine itself that allows people to gain access to it, the whole network security has been breached. Nevertheless, these types of firewalls are cheaper and easier to manage than the more sophisticated arrangement just described. Figure 9.1 illustrates the two most common firewall configurations.

Figure 9.1: The two major classes of firewall design

The Linux kernel provides a range of built-in features that allow it to function quite nicely as an IP firewall. The network implementation includes code to do IP filtering in a number of different ways, and provides a mechanism to quite accurately configure what sort of rules you'd like to put in place. The Linux firewall is flexible enough to make it very useful in either of the configurations illustrated in Figure 9.1. Linux firewall software provides two other useful features that we'll discuss in separate chapters: IP Accounting (Chapter 10, IP Accounting) and IP masquerade (Chapter 11, IP Masquerade and Network Address Translation).

What Is IP Filtering?

IP filtering is simply a mechanism that decides which types of IP datagrams will be processed normally and which will be discarded. By discarded we mean that the datagram is deleted and completely ignored, as if it had never been received. You can apply many different sorts of criteria to determine which datagrams you wish to filter; some examples of these are:

· Protocol type: TCP, UDP, ICMP, etc.

· Socket number (for TCP/UPD)

· Datagram type: SYN/ACK, data, ICMP Echo Request, etc.

· Datagram source address: where it came from

· Datagram destination address: where it is going to

It is important to understand at this point that IP filtering is a network layer facility. This means it doesn't understand anything about the application using the network connections, only about the connections themselves. For example, you may deny users access to your internal network on the default telnet port, but if you rely on IP filtering alone, you can't stop them from using the telnet program with a port that you do allow to pass trhough your firewall. You can prevent this sort of problem by using proxy servers for each service that you allow across your firewall. The proxy servers understand the application they were designed to proxy and can therefore prevent abuses, such as using the telnet program to get past a firewall by using the World Wide Web port. If your firewall supports a World Wide Web proxy, their telnet connection will always be answered by the proxy and will allow only HTTP requests to pass. A large number of proxy-server programs exist. Some are free software and many others are commercial products. The Firewall-HOWTO discusses one popular set of these, but they are beyond the scope of this book.

The IP filtering ruleset is made up of many combinations of the criteria listed previously. For example, let's imagine that you wanted to allow World Wide Web users within the Virtual Brewery network to have no access to the Internet except to use other sites' web servers. You would configure your firewall to allow forwarding of:

· datagrams with a source address on Virtual Brewery network, a destination address of anywhere, and with a destination port of 80 (WWW)

· datagrams with a destination address of Virtual Brewery network and a source port of 80 (WWW) from a source address of anywhere

Note that we've used two rules here. We have to allow our data to go out, but also the corresponding reply data to come back in. In practice, as we'll see shortly, Linux simplifies this and allows us to specify this in one command.

Setting Up Linux for Firewalling

To build a Linux IP firewall, it is necessary to have a kernel built with IP firewall support and the appropriate configuration utility. In all production kernels prior to the 2.2 series, you would use the ipfwadm utility. The 2.2.x kernels marked the release of the third generation of IP firewall for Linux called IP Chains. IP chains use a program similar to ipfwadm called ipchains. Linux kernels 2.3.15 and later support the fourth generation of Linux IP firewall called netfilter. The netfilter code is the result of a large redesign of the packet handling flow in Linux. The netfilter is a multifaceted creature, providing direct backward-compatible support for both ipfwadm and ipchains as well as a new alternative command called iptables. We'll talk about the differences between the three in the next few sections.

Kernel Configured with IP Firewall

The Linux kernel must be configured to support IP firewalling. There isn't much more to it than selecting the appropriate options when performing a make menuconfig of your kernel.[60] We described how to do this is in Chapter 3, Configuring the Networking Hardware". In 2.2 kernels you should select the following options:

Networking options  --->

        [*] Network firewalls

        [*] TCP/IP networking

        [*] IP: firewalling

        [*] IP: firewall packet logging

In kernels 2.4.0 and later you should select this option instead:

  Networking options  --->

     [*] Network packet filtering (replaces ipchains)

         IP: Netfilter Configuration  --->


             <M> Userspace queueing via NETLINK (EXPERIMENTAL)

             <M> IP tables support (required for filtering/masq/NAT)

             <M>   limit match support

             <M>   MAC address match support

             <M>   netfilter MARK match support

             <M>   Multiple port match support

             <M>   TOS match support

             <M>   Connection state match support

             <M>   Unclean match support (EXPERIMENTAL)

             <M>   Owner match support (EXPERIMENTAL)

             <M>   Packet filtering

             <M>     REJECT target support

             <M>     MIRROR target support (EXPERIMENTAL)


             <M>   Packet mangling

             <M>     TOS target support

             <M>     MARK target support

             <M>   LOG target support

             <M> ipchains (2.2-style) support

             <M> ipfwadm (2.0-style) support

The ipfwadm Utility

The ipfwadm (IP Firewall Administration) utility is the tool used to build the firewall rules for all kernels prior to 2.2.0. Its command syntax can be very confusing because it can do such a complicated range of things, but we'll provide some common examples that will illustrate the most important variations of these.

The ipfwadm utility is included in most modern Linux distributions, but perhaps not by default. There may be a specific software package for it that you have to install. If your distribution does not include it, you can obtain the source package from ftp.xos.nl in the /pub/linux/ipfwadm/ directory, and compile it yourself.

The ipchains Utility

Just as for the ipfwadm utility, the ipchains utility can be somewhat baffling to use at first. It provides all of the flexibility of ipfwadm with a simplified command syntax, and additionally provides a "chaining" mechanism that allows you to manage multiple rulesets and link them together. We'll cover rule chaining in a separate section near the end of the chapter, because for most situations it is an advanced concept.

The ipchains command appears in most Linux distributions based on the 2.2 kernels. If you want to compile it yourself, you can find the source package from its developer's site at http://www.rustcorp.com/linux/ipchains/. Included in the source package is a wrapper script called ipfwadm-wrapper that mimics the ipfwadm command, but actually invokes the ipchains command. Migration of an existing firewall configuration is much more painless with this addition.

The iptables Utility

The syntax of the iptables utility is quite similar to that of the ipchains syntax. The changes are improvements and a result of the tool being redesigned to be extensible through shared libraries. Just as for ipchains, we'll present iptables equivalents of the examples so you can compare and contrast its syntax with the others.

The iptables utility is included in the netfilter source package available at http://www.samba.org/netfilter/. It will also be included in any Linux distribution based on the 2.4 series kernels.

We'll talk a bit about netfilter's huge step forward in a section of its own later in this chapter.

Three Ways We Can Do Filtering

Consider how a Unix machine, or in fact any machine capable of IP routing, processes IP datagrams. The basic steps, shown in Figure 9.2 are:

Figure 9.2: The stages of IP datagram processing

· The IP datagram is received. (1)

· The incoming IP datagram is examined to determine if it is destined for a process on this machine.

· If the datagram is for this machine, it is processed locally. (2)

· If it is not destined for this machine, a search is made of the routing table for an appropriate route and the datagram is forwarded to the appropriate interface or dropped if no route can be found. (3)

· Datagrams from local processes are sent to the routing software for forwarding to the appropriate interface. (4)

· The outgoing IP datagram is examined to determine if there is a valid route for it to take, if not, it is dropped.

· The IP datagram is transmitted. (5)

In our diagram, the flow 1→3→5 represents our machine routing data between a host on our Ethernet network to a host reachable via our PPP link. The flows 1→2 and 4→5 represent the data input and output flows of a network program running on our local host. The flow 4→3→2 would represent data flow via a loopback connection. Naturally data flows both into and out of network devices. The question marks on the diagram represent the points where the IP layer makes routing decisions.

The Linux kernel IP firewall is capable of applying filtering at various stages in this process. That is, you can filter the IP datagrams that come in to your machine, filter those datagrams being forwarded across your machine, and filter those datagrams that are ready to be transmitted.

In ipfwadm and ipchains, an Input rule applies to flow 1 on the diagram, a Forwarding rule to flow 3, and an Output rule to flow 5. We'll see when we discuss netfilter later that the points of interception have changed so that an Input rule is applied at flow 2, and an Output rule is applied at flow 4. This has important implications for how you structure your rulesets, but the general principle holds true for all versions of Linux firewalling.

This may seem unnecessarily complicated at first, but it provides flexibility that allows some very sophisticated and powerful configurations to be built.

Original IP Firewall (2.0 Kernels)

The first generation IP firewall support for Linux appeared in the 1.1 series kernel. It was a port of the BSD ipfw firewall support to Linux by Alan Cox. The firewall support that appeared in the 2.0 series kernels and is the second generation was enhanced by Jos Vos, Pauline Middelink, and others.

Using ipfwadm

The ipfwadm command was the configuration tool for the second generation Linux IP firewall. Perhaps the simplest way to describe the use of the ipfwadm command is by example. To begin, let's code the example we presented earlier.

A naïve example

Let's suppose that we have a network in our organization and that we are using a Linux-based firewall machine to connect our network to the Internet. Additionally, let's suppose that we wish the users of that network to be able to access web servers on the Internet, but to allow no other traffic to be passed.

We will put in place a forwarding rule to allow datagrams with a source address on our network and a destination socket of port 80 to be forwarded out, and for the corresponding reply datagrams to be forwarded back via the firewall.

Assume our network has a 24-bit network mask (Class C) and an address of The rules we might use are: The second rule sets our default forwarding policy. We tell the kernel to deny or disallow forwarding of IP datagrams. It is very important to set the default policy, because this describes what will happen to any datagrams that are not specifically handled by any other rule. In most firewall configurations, you will want to set your default policy to "deny," as shown, to be sure that only the traffic you specifically allow past your firewall is forwarded.

# ipfwadm -F -f

# ipfwadm -F -p deny

# ipfwadm -F -a accept -P tcp -S -D 0/0 80

# ipfwadm -F -a accept -P tcp -S 0/0 80 -D

The -F command-line argument tells ipfwadm that this is a forwarding rule. The first command instructs ipfwadm to "flush" all of the forwarding rules. This ensures we are working from a known state before we begin adding specific rules.

The second rule sets our default forwarding policy. We tell the kernel to deny or disallow forwarding of IP datagrams. It is very important to set the default policy, because this describes what will happen to any datagrams that are not specifically handled by any other rule. In most firewall configurations, you will want to set your default policy to "deny," as shown, to be sure that only the traffic you specifically allow past your firewall is forwarded.

The third and fourth rules are the ones that implement our requirement. The third command allows our datagrams out, and the fourth rule allows the responses back.

Let's review each of the arguments:

- F

This is a Forwarding rule.

- a accept

Append this rule with the policy set to "accept," meaning we will forward any datagrams that match this rule.

- P tcp

This rule applies to tcp datagrams (as opposed to UDP or ICMP).

- S

The Source address must have the first 24 bits matching those of the network address

- D 0/0 80

The destination address must have zero bits matching the address This is really a shorthand notation for "anything." The 80 is the destination port, in this case WWW. You may also use any entry that appears in the /etc/services file to describe the port, so -D 0/0 www would have worked just as well.

ipfwadm accepts network masks in a form with which you may not be familiar. The /nn notation is a means of describing how many bits of the supplied address are significant, or the size of the mask. The bits are always counted from left to right; some common examples are listed in Table 9.1.

Table 9.1: Common Netmask Bit Values

Netmask Bits 8 16 24 25 26 27 28 29 30

We mentioned earlier that ipfwadm implements a small trick that makes adding these sorts of rules easier. This trick is an option called -b, which makes the command a bidirectional rule.

The bidirectional flag allows us to collapse our two rules into one as follows:

# ipfwadm -F -a accept -P tcp -S -D 0/0 80 -b

An important refinement

Take a closer look at our ruleset. Can you see that there is still one method of attack that someone outside could use to defeat our firewall?

Our ruleset allows all datagrams from outside our network with a source port of 80 to pass. This will include those datagrams with the SYN bit set! The SYN bit is what declares a TCP datagram to be a connection request. If a person on the outside had privileged access to a host, they could make a connection through our firewall to any of our hosts, provided they use port 80 at their end. This is not what we intended.

Fortunately there is a solution to this problem. The ipfwadm command provides another flag that allows us to build rules that will match datagrams with the SYN bit set. Let's change our example to include such a rule:

# ipfwadm -F -a deny -P tcp -S 0/0 80 -D -y

# ipfwadm -F -a accept -P tcp -S -D 0/0 80 -b

The -y flag causes the rule to match only if the SYN flag is set in the datagram. So our new rule says: "Deny any TCP datagrams destined for our network from anywhere with a source port of 80 and the SYN bit set," or "Deny any connection requests from hosts using port 80."

Why have we placed this special rule before the main rule? IP firewall rules operate so that the first match is the rule that is used. Both rules would match the datagrams we want to stop, so we must be sure to put the deny rule before the accept rule.

Listing our rules

After we've entered our rules, we ask ipfwadm to list them for us using the command:

# ipfwadm -F -l

This command will list all of the configured forwarding rules. The output should look something like this:

# ipfwadm -F -l

IP firewall forward rules, default policy: accept

type  prot source               destination          ports

deny  tcp  anywhere          www -> any

acc   tcp        anywhere             any -> www

The ipfwadm command will attempt to translate the port number into a service name using the /etc/services if an entry exists there.

The default output is lacking in some important detail for us. In the default listing output, we can't see the effect of the -y argument. The ipfwadm command is able to produce a more detailed listing output if you specify the -e (extended output) argument too. We won't show the whole output here because it is too wide for the page, but it includes an opt (options) column that shows the -y option controlling SYN packets:

# ipfwadm -F -l -e

P firewall forward rules, default policy: accept

pkts bytes type  prot opt  tosa tosx ifname  ifaddress   source        ...  

    0     0 deny  tcp  --y- 0xFF 0x00 any     any         anywhere      ...      

    0     0 acc   tcp  b--- 0xFF 0x00 any     any ...

A More Complex Example

The previous example was a simple one. Not all network services are as simple as the WWW service to configure; in practice, a typical firewall configuration would be much more complex. Let's look at another common example, this time FTP. We want our internal network users to be able to log into FTP servers on the Internet to read and write files. But we don't want people on the Internet to be able to log into our FTP servers.

We know that FTP uses two TCP ports: port 20 (ftp-data) and port 21 (ftp), so:

# ipfwadm -a deny -P tcp -S 0/0 20 -D -y

# ipfwadm -a accept -P tcp -S -D 0/0 20 -b


# ipfwadm -a deny -P tcp -S 0/0 21 -D -y

# ipfwadm -a accept -P tcp -S -D 0/0 21 -b

Right? Well, not necessarily. FTP servers can operate in two different modes: passive mode and active mode.[61] In passive mode, the FTP server listens for a connection from the client. In active mode, the server actually makes the connection to the client. Active mode is usually the default. The differences are illustrated in Figure 9.3.

Figure 9.3: FTP server modes

Many FTP servers make their data connection from port 20 when operating in active mode, which simplifies things for us a little, but unfortunately not all do.[62]

But how does this affect us? Take a look at our rule for port 20, the FTP-data port. The rule as we have it now assumes that the connection will be made by our client to the server. This will work if we use passive mode. But it is very difficult for us to configure a satisfactory rule to allow FTP active mode, because we may not know in advance what ports will be used. If we open up our firewall to allow incoming connections on any port, we are exposing our network to attack on all services that accept connections.

The dilemna is most safely resolved by insisting that our users operate in passive mode. Most FTP servers and many FTP clients will operate this way. The popular ncftp client also supports passive mode, but it may require a small configuration change to make it default to passive mode. Many World Wide Web browsers such as the Netscape browser also support passive mode FTP, so it shouldn't be too hard to find appropriate software to use. Alternatively, you can avoid the issue entirely by using an FTP proxy server that accepts a connection from the internal network and establishes connections to the outside network.

In building your firewall, you will probably find a number of these sorts of problems. You should always give careful thought to how a service actually operates to be sure you have put in place an appropriate ruleset for it. A real firewall configuration can be quite complex.

Summary of ipfwadm Arguments

The ipfwadm has many different arguments that relate to IP firewall configuration. The general syntax is:

ipfwadm category command parameters [options]

Let's take a look at each of these.


One and only one of the following must be supplied. The category tells the firewall what sort of firewall rule you are configuring:

- I

Input rule

- O

Output rule

- F

Forwarding rule


At least one of the following must be supplied and applies only to those rules that relate to the supplied category. The command tells the firewall what action to take.

- a [policy]

Append a new rule

- i [policy]

Insert a new rule

- d [policy]

Delete an existing rule

- p policy

Set the default policy

- l

List all existing rules

- f

Flush all existing rules

The policies relevant to IP firewall and their meanings are:


Allows matching datagrams to be received, forwarded, or transmitted


Blocks matching datagrams from being received, forwarded, or transmitted


Blocks matching datagrams from being received, forwarded, or transmitted, and sends the host that sent the datagram and ICMP error message


At least one of the following must be supplied. Use the parameters to specify to which datagrams this rule applies:

- P protocol