Linux System Administration - FIU School of Computing and

Linux System Administration - FIU School of Computing and
Linux System Administration
Compiled by
Raimund Ege
Draft
Draft
Copyright © 2004, 2005 O'Reilly Media.
Printed in the United States of America.
Published by O.Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472.
The copyrights in individual elements of this work are owned by their respective publishers, authors or others, as the case may be, and the
prior written permission of the copyright owner is required for reuse in any form or medium of any individual element.
O.Reilly books may be purchased for educational, business or sales promotional use. Online editions are also available for most titles (http://safari.oreilly.com/). For more information, contact our corporate/institutional sales department: (800)998.9938 or [email protected]
Nutshell Handbook, the Nutshell Handbook logo, and the O.Reilly logo are registered trademarks of O.Reilly Media, Inc. Microsoft, the .NET
logo, Virtual C#, Visual Basic, Visual Studio, and Windows are registered trademarks or trademarks of Microsoft Corporation. Many of the
designations used by manufacturers and sellers to distinguish their products are claimed as trademarks. Where those designations appear in
this book, and O.Reilly Media, Inc. was aware of a trademark claim, the designations have been printed in caps or initial caps. The association
between the image of the African crowned crane and the topic of C# is a trademark of O.Reilly Media, Inc.
While reasonable care has been exercised in the preparation of this book, the publisher and the authors assume no responsibility for errors or
omissions, or for damages resulting from the use of the information contained herein.
Draft
Draft
Table of Contents
Installation Overview .......................................................................................................................... 1
Step-by-Step Installation ..................................................................................................................... 17
System Administration: Core Concepts ................................................................................................. 47
Administration Tasks ......................................................................................................................... 95
Configuring a LAN .......................................................................................................................... 121
iptables: Setting Up a Firewall ............................................................................................................ 131
Downloading and Installing Software ................................................................................................. 153
Apache (httpd): Setting Up a Web Server ............................................................................................. 183
OpenSSH: Secure Network Communication ........................................................................................ 227
Rebuilding the Linux Kernel .............................................................................................................. 245
Printing with CUPS ......................................................................................................................... 257
sendmail: Setting Up Mail Clients, Servers, and More ............................................................................ 277
FTP: Transferring Files Across a Network ............................................................................................ 303
Files, Directories, and Filesystems ....................................................................................................... 325
NFS: Sharing Filesystems .................................................................................................................. 349
Samba: Integrating Linux and Windows .............................................................................................. 369
Index ............................................................................................................................................. 395
iii
iv
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 2 - Installation Overview
Installation Overview
IN THIS CHAPTER
More Information
Planning the Installation
Setting Up the Hard Disk
How the Installation Works
The Medium: Where Is the Source Data?
Downloading, Burning, and Installing a CD Set (FEDORA)
Using BitTorrent to Download the ISO Image Files
Rescue CD
Collecting Information About the System
Installing Red Hat Linux is the process of copying operating system files from media to the local system and setting
up configuration files so that Linux runs properly on the local hardware. You can install Linux from many types
of media, including CDs or a hard disk and files on other systems that you access over a network. There are also
different types of installations including fresh installations, upgrades from older versions of Red Hat Linux, and
dual-boot installations. You can perform the installation manually or set up Kickstart to install Red Hat Linux
automatically.
This chapter discusses the installation: planning, how to divide the hard disk into partitions, obtaining the files for
the installation, burning CDs if necessary, and collecting information about the system you will need when you
install the system. Chapter 3 (page 43) covers the actual installation.
Red Hat developed Anaconda, an installation tool that performs an interactive installation using a GUI, to automate
and make friendlier the process of installing Linux. If you have standard hardware, you can typically insert the first
installation CD and boot a system, press RETURN a few times, change CDs a few times, and Linux is installed.
However, you may want to customize your system and/or have nonstandard hardware: Anaconda gives you many
choices as the installation process unfolds. Refer to “Booting the System: The boot: Prompt” (page 44) and “The
Anaconda Installer” (page 46) for information about customizing Red Hat Linux installation.
More Information
Web .
SELinux FAQ .
people.redhat.com/kwade/fedora-docs/selinux-faq-en
X.org release notes .
freedesktop.org/~xorg/X11R6.7.0/doc/RELNOTES.html
1
Draft
Planning the Installation
Memtest .
www.memtest86.com [http://www.memtest86.com]
Hardware compatibility .
LVM .
Draft
hardware.redhat.com
www.sistina.com/products_lvm.htm [http://www.sistina.com/products_lvm.htm]
BitTorrent .
bitconjurer.org/BitTorrent
Downloads .
FEDORA .
Red Hat’s site; includes server, mirrors, and BitTorrent fedora.redhat.com/download
FEDORA .
download.fedora.redhat.com/pub/fedora/linux/core
FEDORA .
mirrors fedora.redhat.com/download/mirrors.html
FEDORA .
BitTorrent tracker and software torrent.dulug.duke.edu
RHEL .
ftp://ftp.redhat.com/pub/redhat/linux/enterprise
RHEL .
mirrors www.redhat.com/download/mirror.html [http://www.redhat.com/download/mirror.html]
Planning the Installation
The biggest parts of planning an installation are determining how to divide the hard disk into partitions or, in the
case of a dual-boot system, where to put the Linux partition, and deciding which software packages to install. In
addition to these topics, this section discusses hardware requirements for Red Hat Linux, Fedora Core versus Red
Hat Enterprise Linux, fresh installations versus upgrading, and classes of installations (server, workstation, and so
on).
Considerations
SELinux . If you are going to want to use SELinux, turn it on when you install Linux (page 44). Because SELinux
sets extended attributes on files, it is an involved process to turn on SELinux after you install Linux.
GUI . On most installations except servers, you probably want to install a graphical interface. The X Window
System is required for any graphical interface. In addition, select GNOME, KDE, or both.
Chapter 4, “Introduction to Red Hat Linux,” uses examples from KDE to introduce the graphical desktop. Install
KDE if you want to follow these examples. You can remove KDE later if you like.
On a server, you normally dedicate as many resources to the server and as few to anything not required by the
server. For this reason, servers do not usually include a graphical interface.
Installation Requirements
Hardware . Red Hat Linux can run on many different types of hardware. This section details installation on
Intel and compatible platforms such as AMD, Cyrix, and VIA. Within these platforms, Red Hat Linux runs on
much of the available hardware. You can view Red Hat’s hardware list, from certified to unsupported, at hardware.redhat.com. There are many Internet sites that discuss Linux hardware; use Google to search on linux
hardware or linux and the specific hardware you want more information on (for example, linux sata or linux
nf7s). There are many HOWTOs that cover specific hardware and a Linux Hardware Compatibility HOWTO,
although it becomes dated rather quickly. Red Hat Linux usually runs on systems that Windows runs on, unless
you have a very new or unusual component in the system.
2
Draft
What Are You Installing: Fedora Core
or Red Hat Enterprise Linux?
Draft
Memory (RAM) . You need a minimum of 64 megabytes of RAM for a system that runs in text mode (no GUI)
and 192-256 megabytes for graphical system. Because Linux makes good use of extra memory, the more memory
your system has, the faster it will run. Additional memory is one of the most cost-effective ways you can speed up
a Linux system.
CPU . Red Hat Linux requires a minimum of a 200 megahertz Pentium-class processor or the equivalent AMD
or other processor for text mode and at least a 400 megahertz Pentium II processor or the equivalent for graphical
mode.
Disk space . The amount of disk space you need depends on which version of Red Hat Linux you install, which
packages you install, how many languages you install, and how much space you need for user data (your files).
Table 2-1 shows the minimum amount of disk space required.
Table 2-1. Minimum Required Disk Space
Class of Installation
Minimum Disk Space Required
Personal Desktop
2.3 gigabytes
Workstation
3.0 gigabytes
Server
1.1–1.5 gigabytes (no graphical environment) 6.9 gigabytes (everything including graphical environment)
Custom
0.62–6.9 gigabytes
BIOS setup . Modern computers can be set to boot from a CD, floppy diskette, or hard disk. The BIOS determines
the order in which the system tries to boot from each device. You may need to change this order: Make sure the
BIOS is set up to try booting from the CD before it tries to boot from the hard drive. See “CMOS,” next.
CMOS . CMOS is the persistent memory that stores system configuration information. To change the BIOS
setup, you need to edit the information stored in CMOS. When the system boots, it displays a brief message about
how to enter System Setup or CMOS Setup mode. Usually, you need to press Del or F2 while the system is
booting. Press the key that is called for, and move the cursor to the screen and line that deal with booting the system.
Generally, there is a list of three or four devices that the system tries to boot from; failing the first, the system tries
the second and so on. Manipulate the list so the CD is the first choice, save your choices, and reboot. Refer to the
hardware/BIOS manual for more information.
What Are You Installing: Fedora Core or Red Hat Enterprise Linux?
This book describes two products: Fedora Core and Red Hat Enterprise Linux. This section briefly explains the
differences between these products.
FEDORA . The Fedora Project is sponsored by Red Hat and supported by the open-source community. With
releases, called Fedora Core, coming out about every six months, it is a Linux distribution that tests cutting edge
code; it is not a supported Red Hat product and is not recommended for production environments where stability
is important. Fedora aims at staying similar to the upstream projects it incorporates, including the kernel, while
Red Hat Enterprise Linux includes many changes introduced by Red Hat.
RHEL . Red Hat Enterprise Linux is sold by annual subscription that includes Red Hat Network (page 467)
and technical support. It is more stable but less cutting edge than Fedora Core Linux.
Red Hat Enterprise Linux AS and Red Hat Enterprise Linux ES function identically and are designed to run servers.
ES is licensed for x86 compatible systems with one or two CPUs and up to 8 gigabytes of memory. AS is licensed
for servers of any size and architecture.
3
Draft
Installing a Fresh Copy or Upgrading
an Existing Red Hat System?
Draft
Red Hat Enterprise Linux WS (workstation) supports x86, Itanium, and AMD64 architectures on the desktop/client
side, running office productivity and software development applications. WS does not include all the server applications that come with AS and ES and is not designed for a server environment.
Installing a Fresh Copy or Upgrading an Existing Red Hat System?
An upgrade replaces the Linux kernel and utilities on an already installed version of Red Hat Linux with newer
versions. During an upgrade, the installation program attempts to preserve both system and user data files. An
upgrade brings utilities that are present in the old version up-to-date but does not install new utilities (you can install
them later if you like). Existing configuration files are preserved by renaming with a .rpmsave extension. A log of
the upgrade is kept in /root/upgrade.log. Before upgrading a system, back up all files on the system.
Caution: A Fresh Installation Yields a More Stable System Than an
Upgrade
For better system stability, Red Hat recommends that you back up your data and perform a fresh install
rather than upgrade.
An installation writes all fresh data to a disk. The installation program overwrites all system programs and data as
well as the kernel. You can preserve some user data during an installation depending on where it is located and
how you format/partition the disk.
Types of Installations (FEDORA)
With Fedora, there are four classes of Red Hat Linux you can install. The class you select determines which software
packages are installed. You can modify this list of packages during the installation. Under Red Hat Enterprise Linux,
you determine the “class” of installation by selecting the packages you want to install.
Personal Desktop Good if you are new to Linux. This installation gives you a GUI and installs many of the programs
you are likely to need.
Workstation For software developers and system administrators. This installation gives you a GUI and installs
many software development and system administration tools.
Server For administrators who want to deploy a Linux-based server without doing heavy customization. This installation has more services and fewer end-user programs. Although you do not normally bring up a GUI on a
server, GNOME and KDE are available.
Custom Presents you with all the programs and services that Red Hat supports and lets you decide which you
want to install. This installation gives you full control over which packages you install. Recommended for experienced
users only.
Regardless of the type of installation you choose, you can always add or remove programs once the system is up
and running. For more information, refer to “system-config-packages: Adds and Removes Software Packages” on
page 453.
Graphical or Textual Installation?
There are several ways to install Red Hat Linux. You can choose a graphical installation, which displays graphics
on the screen and allows you to use the mouse, window buttons, and scroll lists to choose how you want to configure
the system. The Anaconda utility controls a graphical installation. If you have a smaller system with less memory
or a system with an unsupported graphics board, you can use the text-mode installation. This mode performs the
4
Draft
Setting Up the Hard Disk
Draft
same functions as the graphical mode, but it uses plain text to prompt you with questions about how you want to
configure the system.
Setting Up the Hard Disk
Formatting and free space . Hard disks must be prepared in several ways so an operating system can write to
and read from them. Formatting is the first step in preparing a disk for use. Normally, you will not need to format
a hard disk as this is done at the factory. The next step in preparing a disk for use is to divide it into partitions. The
area of the disk that is not occupied by partitions is called free space. A new disk has no partitions: It is all free
space. Under DOS/Windows, the term formatting means writing a filesystem on a partition; see “Filesystems,”
following.
Partitions . A partition, or slice, is a section of a hard disk that has a name, such as /dev/hda1, so you can address
it separately from other sections. During installation you can use Disk Druid (page 53) to create partitions on the
hard disk. After installation you can use fdisk (page 58) to manipulate partitions.
Filesystems . Before most programs can write to a partition, a data structure (page 966), called a filesystem,
needs to be written on a partition. The mkfs (make filesystem) utility, similar to the DOS/Windows format utility,
writes a filesystem on a partition. There are many types of filesystems. Red Hat Linux typically creates ext3
filesystems for data, while Windows uses FAT and NTFS filesystems. Apple uses HFS (Hierarchical Filesystem)
and HFS+. OS X uses either HFS+ or UFS. Under DOS/Windows, filesystems are labeled C:, D:, and so on
(sometimes a whole disk is a single partition). Under Linux, typical filesystem names are /boot, /var, and /usr.
You can have different types of partitions on the same drive, including both Windows and Linux partitions. Under
Linux, the fsck (filesystem check) utility (page 447) checks the integrity of filesystem data structures.
Filesystem independence . The state of one filesystem does not affect other filesystems: One filesystem on a
drive may be corrupt and unreadable while other filesystems function normally. One filesystem may be full so you
cannot write to it while others have plenty of room for more data.
Primary and Extended Partitions
Partitioning allows you to divide a disk into a maximum of eight separate partitions, or subdisks. You can use each
partition independently for swap devices, filesystems, databases, other functions, and even other operating systems.
Unfortunately, disk partitions follow the template established for DOS machines a long time ago. There are two
types of partitions: primary and extended. At most, a disk can hold four primary and extended partitions in any
combination. Extended partitions can each be subdivided into additional partitions; primary partitions cannot. If
you want more than four partitions on a drive, and you usually do, at least one of the four partitions must be an
extended partition.
One way to set up a disk is to divide it into three primary partitions and one extended partition. The three primary
partitions are the sizes you want the final partitions to be. The extended partition occupies the rest of the disk.
Once you establish the extended partition, you can subdivide it into additional partitions that are each the size you
want.
Partitioning a Disk
During installation, Anaconda calls Disk Druid to set up disk partitions. This section discusses how to plan partition
sizes. For more information, refer to “Using Disk Druid to Partition the Disk” on page 53.
Tip: Under Red Hat, Druid Means Wizard
Red Hat uses the term druid at the ends of names of programs that guide you through a task-driven
chain of steps. Other operating systems call these types of programs wizards.
5
Draft
Setting Up the Hard Disk
Draft
Planning Partitions
It can be difficult to plan partition sizes appropriately if you are new to Linux. For this reason many people choose
to have only two partitions. Partition 1 is reserved for the swap partition, which can be any size from 0 to many
hundreds of megabytes or, rarely, gigabytes. Partition 0 is designated as root (/) and contains the remainder of the
disk space. This setup makes managing space quite easy. But if a program runs amok, or your system receives a
DoS attack (page 968), the entire disk can fill up, and system accounting and logging information (which may
contain data that can tell you what went wrong) may be lost.
When Disk Druid sets up partitions automatically, it sets up the two partitions described earlier and a /boot partition
that holds a copy of the operating system and other information the system needs at boot time.
Older disks used a constant number of sectors per track. This made positioning the disk head easier but resulted
in a lot of wasted space on the outer tracks. Newer disks use an encoding named ZBR (zone-bit recording): The
sector size of the disk stays constant, which means that the outer cylinders have more sectors. Consequently, more
information passes under the disk head per disk revolution, and data access is faster on these sectors. These outer
cylinders correspond to the lower cylinder numbers (cylinders starting at zero). Put your speed-critical data on the
lower-numbered cylinders.
Partition Suggestions
The following paragraphs discuss additional partitions you may want to create. Consider setting up LVM (page
32) before you create partitions; LVM allows you to change partition sizes easily after the system is installed.
(swap) . Linux temporarily stores programs and data on this part of the disk when it does not have enough
RAM to hold all the information it is processing (page 435). Although it is not required, most systems perform
better with a swap partition.
/boot . This partition holds the kernel and other data the system needs when it boots. Red Hat recommends
that the /boot partition be 100 megabytes. Although you can omit the /boot partition, it is useful in many cases.
Some older BIOSs require the /boot partition (or the root [/] partition if there is no /boot partition) to be near the
start of the disk.
/var . The name var is short for variable: The data in this partition changes frequently. Because it holds the
bulk of system logs, package information, and accounting data, making /var a separate partition is a good idea. If
a user runs a job that uses up all the disk space, the logs will not be affected. A good size for the /var partition is
anything from about 500 megabytes up to several gigabytes for extremely active systems with many verbose daemons
and a lot of printer activity (files in the print queue are stored on /var). Systems serving as license servers for licensed
software often fall into the category of extremely active systems.
/home . It is a common strategy to put user home directories in their own disk or partition. If you do not have
a separate disk for the home directories, put them in their own partition. Common partition names are /home and
/usr/home.
Tip: Set Up Partitions to Aid in Making Backups
Plan your partitions around what data you want to back up and how often you want to back it up.
One very large partition can be more difficult to back up than several smaller ones.
/ (root) . Some administrators choose to separate the root (/), /boot, and /usr partitions. By itself, the root
partition usually consumes less than 30 megabytes of disk space. On occasion, you may install a special program
that has many kernel drivers that consume a lot of space in the root partition. If you expect to run into this situation,
you need to adjust the space allotted to the root partition accordingly.
6
Draft
Raid
Draft
/usr . Separating the /usr partition can be useful if you want to export /usr to another machine and want the
security that a separate partition can give. The size of /usr depends largely on the number of packages you install.
Tip: Where to Put the /boot Partition
On older systems, the /boot partition must reside completely below cylinder 1,023 of the disk. When
you have more than one hard drive, the /boot partition must also reside on a drive on
•
Multiple IDE or EIDE drives: the primary controller
•
Multiple SCSI drives: ID 0 or ID 1
•
Multiple IDE and SCSI drives: the primary IDE controller or SCSI ID 0
/usr/local and /opt . Finally, /usr/local and /opt are other candidates for separation. If you plan to install
many packages in addition to Red Hat Linux, you may want to keep them on a separate partition. If you install the
additional software in the same partition as the users’ home directories, for example, it may start to encroach on
the users’ disk space. Many sites keep all the /usr/local or /opt software on one server and export it to others. If
you choose to create a /usr/local or /opt partition, its size should be appropriate to the software you plan to install.
Table 2-2 gives guidelines for minimum sizes for partitions used by Linux. Size other partitions, such as /home,
/opt, and /usr/local, according to need and the size of the drive. If you are not sure how you are going to use additional disk space, you can create extra partitions using whatever names you like (for example, /b01, /b02, and
so on).
Table 2-2. Example Partition Sizes
/boot
100 megabytes
/(root)
500 megabytes
(swap)
Two times the amount of RAM (memory) in the system
with a minimum of 500 megabytes
/home
As large as necessary; depends on the number of users
and the type of work they do
/tmp
Minimum of 500 megabytes
/usr
Minimum of 1.7–5.5 gigabytes, depending on which Red
Hat Linux programs you install. These figures assume
/usr/local is a separate partition. Refer to “Disk space”
on page 25 for more information on the size of Red Hat
Linux.
/var
Minimum of 500 megabytes
Raid
RAID (Redundant Array of Inexpensive/Independent Disks) employs two or more hard disk drives or partitions
in combination to improve fault tolerance and/or performance. Applications and utilities see the multiple
drives/partitions as a single logical device. RAID, which can be implemented in hardware (faster) or software (Red
Hat gives you this option), spreads data across multiple disks. Depending on which level you choose, RAID can
provide data redundancy to protect data in the case of hardware failure. Although it can also improve disk performance by increasing read/write speed and throughput, RAID uses quite a bit of CPU time, which may be a consideration in some situations.
7
Draft
LVM (Logical Volume Manager)
Draft
Caution: Do Not Replace Backups with RAID
Do not use RAID as a replacement for regular backups. If your system undergoes a catastrophic failure,
RAID will be useless. Earthquake, fire, theft, and so on may leave your entire system inaccessible (if
your hard drives are destroyed or missing). RAID does not take care of something as simple as replacing
a file when you delete it by accident. In these cases, a backup on removable media (that has been removed) is the only way you will be able to restore a filesystem.
Disk Druid gives you the choice of implementing RAID level 0, 1, or 5:
•
RAID Level 0 (striping) . Improves performance but offers no redundancy. The storage capacity of the
RAID device is equal to that of the member partitions or disks.
•
RAID Level 1 (mirroring) . Provides simple redundancy, improving data reliability, and can improve the
performance of read-intensive applications. The storage capacity of the RAID device is equal to one of the
member partitions or disks.
•
RAID Level 5 (disk striping with parity). Provides redundancy and improves (most notably, read) performance. The storage capacity of the RAID device is equal to that of the member partitions or disks, minus
one of the partitions or disks (assuming they are all the same size).
LVM (Logical Volume Manager)
The Logical Volume Manager (LVM) allows you to change the size of logical volumes (LV, the LVM equivalent of
partitions) on the fly: Using LVM, if you find you made a mistake in setting up logical volumes or your needs
change, you can make LVs smaller or larger easily without affecting user data. You must choose to use LVM at the
time you install the system or add a hard disk; you cannot retroactively apply it to a disk full of information. LVM
supports IDE and SCSI drives as well as multiple devices such as those found in RAID partitions.
LVM groups disk components (hard disks or storage device arrays), called physical volumes (PVs), into a storage
pool, or virtual disk, called a volume group (VG). See Figure 2-1. You allocate a portion of a VG to create a logical
volume (LV).
Figure 2-1. LVM: Logical Volume Manager
An LV is similar in function to a traditional disk partition in that you can create a filesystem on an LV. It is much
easier, however, to change and move LVs than partitions: When you run out of space on a filesystem on an LV,
you can grow (expand) the LV and its filesystem, into empty or new disk space, or you can move the filesystem to
a larger LV. LVM’s disk space manipulation is transparent to users; service is not interrupted.
8
Draft
How the Installation Works
Draft
LVM also eases the burdens of storage migration. When you outgrow or need to upgrade PVs, LVM can move data
to new PVs, again without interrupting users.
For specifics, refer to www.sistina.com/products_lvm.htm [http://www.sistina.com/products_lvm.htm], the LVM
HOWTO, the lvm man page, and the “See also” man pages listed at the bottom of the lvm man page.
How the Installation Works
The following steps outline the process of installing Red Hat Linux from CDs using Anaconda. Installation from
other media follows similar steps. See Chapter 3 for the specifics of how to perform the installation.
1.
Insert the first installation CD in the computer and turn on or reset the computer.
2.
After going through computer-specific hardware diagnostics, the computer displays the initial install screen
with the boot: prompt at the bottom (page 44).
3.
You can enter commands and press RETURN following the boot: prompt, press RETURN without entering
commands, or wait for a minute without entering anything; the computer boots Red Hat Linux from the installation CD.
4.
As part of the boot process, Red Hat Linux creates multiple RAM disks (page 991) that it uses in place of a
hard disk used for a normal boot operation. Tools used during the installation are copied to the RAM disks.
The use of RAM disks allows the installation process to run through the specification and design phases
without writing to the hard disk and enables you to opt out of the installation at any point before the system
warns you it is about to write to the hard disk. If you opt out before this point, the system is left in its original
state. For more information, refer to “About to Install” on page 52.
5.
You can check the installation media at this point.
6.
Anaconda starts, usually probing the hardware before starting the X Window System for a graphical installation.
7.
Anaconda collects information about how you want to install Red Hat Linux.
8.
When Anaconda is finished collecting information, it warns you that it is About to Install and writes the
operating system files to the hard disk.
9.
When you reboot the system, installation scripts ask you some questions to complete the installation (page
52).
10. The Red Hat Linux system is ready to use.
The Medium: Where Is the Source Data?
When you install Red Hat Linux, you copy operating system files from a source, frequently a CD, to the target
computer’s hard disk. There are two formats and many possible sources for the files.
Formats
Red Hat Linux operating system files can be stored as directory hierarchies on CDs or a hard disk or as CD images
on a hard disk (called ISO images after ISO9660, the standard defining the CD filesystem). Although the format is
different, the content is the same. You can install Red Hat Linux or burn a CD from either format, although most
people use the ISO images to burn CDs as it is more convenient.
9
Draft
Sources
Draft
Sources
This chapter details installing Red Hat Linux from CDs and ISO image files. It does not cover installing from directory hierarchies; you use exactly the same techniques to install from a directory hierarchy as from an ISO image.
Directory hierarchies are more cumbersome to work with than ISO images because they contain many files; each
ISO image is a single file.
You can automate the installation using Kickstart (page 57).
Following is a list of possible locations for the source files for an installation:
CD
You can purchase or burn CDs for installing Fedora Core. Or, you can use the CDs included with this book.
RHEL .
Red Hat Enterprise Linux CDs are sold by Red Hat and its distributors.
Although Red Hat does not provide ISO images for Red Hat Enterprise Linux, it does provide source RPM files
(SRPMS), a different set for each of the architectures Red Hat supports. Go to ftp.redhat.com and cd to
/pub/redhat/linux/enterprise/rel/en/arch/SRPMS, where rel is the release number and arch is the architecture
you want to download files for (AMD64, i386, ia64, and so on), to download the SRPMS files. Or you can look at
www.redhat.com/download/mirror.html [http://www.redhat.com/download/mirror.html] for a list of sites that
provide these files. Because Red Hat does not provide ISO images for Red Hat Enterprise Linux, this book does
not discuss how to install Red Hat Enterprise Linux from downloaded files. It does cover how to install Red Hat
Enterprise Linux from the CDs you buy from Red Hat.
FEDORA . This book includes the CDs necessary for installing Fedora Core. Alternatively, Fedora CDs can be
purchased from third-party vendors or you can download the Fedora ISO images and install from the images or
burn your own CDs (next section).
Hard Disk
FEDORA . You can store ISO image files on the target system’s hard disk if it is already running Linux. You
need to burn only the first installation CD or the rescue CD (page 39) for a hard disk installation.
Network
FEDORA . You can use ISO image files from a server system that the target system can connect to over a network
during installation. Network installations can be done using FTP, NFS, or HTTP. Unless you have a fast Internet
connection, it is not advisable to perform an installation over the Internet as it can take a very long time; downloading
ISO files is a more reliable and possibly less frustrating option. You need to burn only the first installation CD or
the rescue CD (page 39) for a network installation.
Downloading, Burning, and Installing a CD Set (FEDORA)
You can download and burn Fedora Core CDs. Although you will not get the customer support that comes with
Red Hat Enterprise Linux, you will not pay Red Hat for the software. One of the beauties of free software (Appendix
D) is that it is always available for free. Red Hat makes it easy to obtain and use Fedora Core by providing ISO
images of its CDs online. These files are large, over 600 megabytes each, and there are four of them, so they take
days to download using a 56K modem and hours using a broadband connection.
This section tells you how to find the files you need, download them using a couple of different techniques, check
that they downloaded correctly, and burn them to CDs.
10
Draft
Finding a Site to Download From
Draft
Caution: You Can Download and Burn the CDs on Any Operating
System
You can download and burn the CDs on any computer that is connected to the Internet, has a browser,
has enough space on the hard disk to hold the ISO files (about 2.5 gigabytes), and can burn a CD.
If you do not have enough space on the hard disk for all four CDs, you can download and burn them
one at a time (each takes just over 600 megabytes). You can use ftp in place of a browser to download
the files. For more information, refer to “JumpStart: Downloading Files Using ftp ” on page 585.
Finding a Site to Download From
The Fedora Web site maintains the ISO images you need. Other (mirror) sites also maintain these packages. You
can use a Browser or ftp to download the files from one of these sites. Alternatively, you can use BitTorrent to
download the ISO images, see page 38.
To conserve network bandwidth, try to download from a mirror site that is close to you. Failing that, you can
download from the Red Hat site.
Mirror sites .
Locate a mirror site by pointing a browser at the following URL:
fedora.redhat.com/download/mirrors.html
Scroll through the list of mirror sites to find a site near you and click that site’s URL. The display will be similar to
that shown in Figure 2-2. (FTP and HTTP sites look a little different from each other.)
Figure 2-2. The top-level Fedora download directory at redhat.com
The Red Hat site . To download files from the Red Hat Fedora site, point a browser at the following URL,
which locates the Red Hat Web page at the top of the directory hierarchy that holds the Fedora Core files (Figure
2-2):
11
Draft
Finding the Right Files and Downloading Them
Draft
download.fedora.redhat.com/pub/fedora/linux/core/
When you have located a site to download from, continue with the next section.
Finding the Right Files and Downloading Them
The pathname of the Fedora Core files differs from site to site; see the following examples:
/pub/linux/fedora/core/2/i386/iso
/fedora/core/2/i386/iso
/fedora/linux/core/2/i386/iso
All sites share a common path following the core directory (bold in the preceding list). Table 2-3 shows the hierarchy
below the core/2 directory, the directory hierarchy that holds the Fedora Core 2 release. If you are downloading
other than Fedora Core 2, go to the directory hierarchy below the appropriately numbered directory (3, 4, and so
on). These hierarchies will parallel the Fedora Core 2 directory.
Table 2-3. Relative Locations of Fedora Core Files (Fedora Core 2 Shown)
Location in the* /fedora/linux
Directory Hierarchy
Contains
core/2
Fedora Core Linux version 2 directory hierarchy.
core/2/SRPMS
Fedora source RPM files.
core/2/i386
Fedora files for the x86 architecture.
core/2/i386/debug
Fedora debugging programs.
core/2/i386/iso/disc[1–4].iso
Fedora installation CDs 1 through 4 (compiled).
core/2/i386/iso/rescuecd.iso
Fedora rescue CD (page 39).
core/2/i386/iso/MD5SUM
MD5 sums for the ISO files in the same directory.
core/2/i386/os
Individual Fedora rpm packages.
core/2/i386/os/RELEASE-NOTES-en
Fedora release notes in English. (Also RELEASE-NOTESen.html)
core/2/x86_64
Fedora files for the AMD64 architecture. This directory
hierarchy is almost the same as the i386 hierarchy.
Click (open) directories until you get to the iso directory for the release you want to download (Figure 2-3). Click
on the five files listed following, one at a time, to download them. Replace FC2 with the name of the release you
are downloading.
12
Draft
Using BitTorrent to Download the ISO
Image Files
Draft
Figure 2-3. The iso directory for Fedora Core 2 at redhat.com
FC2-i386-disc1.iso
FC2-i386-disc2.iso
FC2-i386-disc3.iso
FC2-i386-disc4.iso
MD5SUM
The four large *.iso files hold the ISO images of the Fedora Core CDs. The short MD5SUM file holds the MD5
checksums that you can use to make sure the downloaded files are correct (page 38). You may want to download
the rescue CD image (FC2-i386-rescuecd.iso) as well (page 39). The FC2-i386-SRPMS-disc*.iso files hold the
source code for Fedora Core; you do not normally need these files to install Fedora.
Depending on the speed of the Internet connection and how busy the site you are downloading from is, it may be
better to wait until one download finishes before starting the next. (Using ftp [page 585], you can queue the
downloads so they proceed sequentially without intervention.)
Once you have downloaded the four installation ISO files, the MD5 checksum file, and optionally the rescue CD
ISO file, the next step is to check that the files are correct. See “Checking the Files,” following the next section.
Using BitTorrent to Download the ISO Image Files
You can use BitTorrent (page 478) to obtain the ISO images; BitTorrent is available for Windows and Mac OS X
(bitconjurer.org/BitTorrent) so you can download and burn the Fedora CDs on a Windows machine or a Mac. See
“BitTorrent” on page 478 for information on using BitTorrent from Linux. You can obtain BitTorrent for Fedora
and the tracker for the ISO files from torrent.dulug.duke.edu.
13
Draft
Checking the Files
Draft
Checking the Files
The MD5SUM file contains the MD5 (page 982) sums for each of the ISO files. When you process a file using the
md5sum utility, md5sum generates a number based on the file. If that number matches the corresponding number
in the MD5SUM file, the downloaded file is correct:
$ grep i386-disc1 MD5SUM; md5sum FC2-i386-disc1.iso
c366d585853768283dac6cdcefcd3a2d FC2-i386-disc1.iso
c366d585853768283dac6cdcefcd3a2d FC2-i386-disc1.iso
Check each of the ISO images you downloaded in the same manner. Computing an MD5 sum for a large file takes
a while. The two long strings that the preceding command displays must be identical: If they are not, you must
download the file again.
Tip: Test the ISO Files and Test the CDs
It is a good idea to test the ISO image files when they are downloaded and the burned CDs before you
use them to install Red Hat Linux. A bad file on a CD may not show up until you have finished installing
Red Hat Linux and have it running. At that point, it may be difficult and time-consuming to figure
out where the problem is. Testing the files and CDs takes a few minutes, but can save you hours if
something is not right. If you want to do only one test, test the CD.
Burning the CDs
An ISO image file is an exact image of what needs to be on the CD. Putting that image on a CD involves a different
process than copying files to a CD. The CD burning software you use has a special selection for burning an ISO
image. It will be labeled something similar to Record CD from CD Image or Burn CD Image. Refer to the instructions for the software you are using for information on how to burn an ISO image file to a CD.
Tip: You Need Only to Burn the Rescue CD for a Hard Disk or
Network Installation
If you are installing Linux from files on a hard disk on the target system or from files on another system
on a network using FTP, NFS, or HTTP, you need a way to boot the system to begin the installation.
The rescue CD (following) or the first installation CD can serve that purpose. Once the system is
booted, you have no need for the CDs.
Tip: Make Sure the Software Is Set Up to Burn an ISO Image
Burning an ISO image is not the same as copying files to a CD. As discussed in the text, you must
make sure the CD burning software is set up to burn an ISO image. If you simply copy the ISO file to
the CD, it will not work to install Fedora.
Rescue CD
The rescue CD cannot do anything the first installation CD cannot do. However, it holds less information so you
can download and burn it more quickly than the first installation CD.
Rescue mode . You can use the rescue CD to bring a system up in rescue mode. Bringing a system up and
working in rescue mode are discussed on page 377.
14
Draft
Collecting Information About the System
Draft
Hard disk or network installation . You can use the rescue CD the same way you use the first installation CD
to boot the system to begin a hard disk or network installation: After booting from either CD, give the command
linux askmethod in response to the boot: prompt. See page 45 for more information.
Collecting Information About the System
It is not difficult to install and bring up a Linux system, but the more you know about the process before you start,
the easier it is. The installation software collects information about the system and can help you make decisions.
However, the system will work better when you know how you want your disk partitioned rather than letting the
installation program partition it without your input. The screen will be easier to use if you know what resolution
you want. There are many details, and the more details you take control of, the happier you will be with the finished
product. Finding the information that this section asks for will help ensure that you end up with a system you understand and know how to change when you need to.
Some of the information is trivial to obtain: When the installation program asks what kind of mouse you have, you
can look at it and perhaps turn it over to read the label. More and more, the installation software probes the hardware
and figures out what you have. Newer equipment is more likely to report on itself than older equipment is.
It is critical to have certain pieces of information before you start. When you are upgrading Linux from an earlier
version, you need to have a list of the disk partitions and where each is mounted so you do not overwrite existing
data. One thing Linux can never figure out is all the relevant names and IP addresses (unless you are running DHCP,
in which case most of the addresses are set up for you).
Following is a list of items you may need information about. Get as much information on each item as you can:
manufacturer, model number, size (megabytes, gigabytes, and so forth), number of buttons, chipset (for boards),
and so on. Some items, such as the network interface card (NIC), may be built into the motherboard.
•
Hard disks
•
CD
•
Memory (you don’t need it for installation, but it is good to know)
•
SCSI interface card
•
Network interface card (NIC)
•
Video interface card (including the amount of video RAM/memory)
•
Sound card and compatibility with standards, such as SoundBlaster
•
Mouse (PS/2, USB, AT, and number of buttons)
•
Monitor (size, maximum resolution)
•
IP addresses and names, unless you are using DHCP (page 408), in which case the IP addresses for the system
are dynamically assigned. Most of this information comes from the system administrator or ISP.
•
System hostname (anything you like)
•
System address
•
Network mask (netmask)
•
Gateway address (the connecting point to the network/Internet) or a phone number when you use a dialup connection
15
Draft
Chapter Summary
•
Addresses for your nameservers, also called DNS addresses
•
Domain name (not required)
Draft
Chapter Summary
When you install Red Hat Linux, you copy operating system files from media to the local system and set up configuration files so that Linux runs properly on the local hardware. You can install Linux from many types of media,
including CDs or hard disks, and files on other systems that you access over a network. Operating system files can
be stored as directory hierarchies on CDs or a hard disk or as CD (ISO) images on a hard disk. You can use a
browser, ftp, or BitTorrent to obtain the ISO images. It is a good idea to test the ISO image files when they are
downloaded and the burned CDs before you use them to install Red Hat Linux.
The biggest parts of planning an installation are determining how to divide the hard disk into partitions and deciding
which software packages to install. If you are going to want to use SELinux, turn it on when you install Linux (page
44). Because SELinux sets extended attributes on files, it is an involved process to turn on SELinux after you install
Linux.
The Fedora Project is sponsored by Red Hat and supported by the open-source community. Fedora Core is a Linux
release that tests cutting-edge code; it is not recommended for production environments. Red Hat Enterprise Linux
is more stable than Fedora Core.
Exercises
1. Briefly, what does the process of installing an operating system such as Red Hat Linux involve?
2. What is Anaconda?
3. Would you set up a GUI on a server system? Why or why not?
4. A system boots from the hard disk. In order to install Linux, you need it to boot from a CD. How can you get
the system to boot from a CD?
5. What is free space on a hard disk? What is a filesystem?
6. What is an ISO image? How do you burn an ISO image to a CD?
Advanced Exercises
7. List two reasons why you should not use RAID in place of backups.
8. What are RAM disks and how are they used during installation?
9. What is MD5? How does it work to ensure that an ISO image file you download is correct?
16
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 3 - Step-by-Step Installation
Step-by-Step Installation
IN THIS CHAPTER
Booting the System: The boot: Prompt
The Anaconda Installer
Using Disk Druid to Partition the Disk
Setting Up a Dual-Boot System
The X Window System
system-config-display: Configuring the Display
More Information
Chapter 2 covered planning the installation: requirements, an upgrade versus a clean installation, classes of installations, planning the layout of the hard disk, how to obtain the files you need for the installation including how to
download and burn ISO (CD) images, and collecting the information about the system you will need during installation. This chapter steps through the process of installing either Red Hat Enterprise Linux or Fedora Core. Frequently,
the installation is quite simple, especially if you have done a good job of planning. Sometimes you may run into a
problem or have a special circumstance; this chapter gives you the tools to use in these cases.
Installing Red Hat Linux
To begin most installations, insert the first installation CD into the CD drive and turn on or reset the system. For
hard disk and network-based installations, you can use the rescue CD (page 39) in place of the first installation
CD.
The system boots from the CD and displays a screen of instructions with a boot: prompt at the bottom. Refer to
“BIOS setup” on page 25 if the system does not boot from the CD.
You cannot boot from a floppy diskette . Because most kernels have grown too large to fit on a floppy diskette,
you cannot boot from a floppy. You cannot fit a standard Fedora Core 2 (and later) kernel on a floppy diskette.
You may be able to fit some Red Hat Enterprise Linux kernels on a diskette. Fedora gives you the option of booting
from a USB pen drive using the diskboot.img file.
Tip: Disable SATA Adapters If You Are Not Using SATA
Many newer motherboards come with SATA (Serial ATA) adapters ready for use. If you do not have
SATA devices, the presence of these adapters can dramatically lengthen the time it takes the system
to boot. Before starting the installation, disable SATA adapters in the BIOS. For more information,
refer to “CMOS” on page 25.
17
Draft
Booting the System: The boot: Prompt
Draft
Booting the System: The boot: Prompt
Normal installation . You can give many different commands at a boot: prompt. If you are installing from
CDs, you can generally press RETURN without entering a command to start installing Red Hat Linux. Or you can
just wait; if you do not type anything for a minute, the installation proceeds as though you pressed RETURN.
Display problems . If you have problems with the display during installation, give the following command,
which turns off video memory, in response to the boot: prompt:
boot: linux nofb
SELinux . By default, Fedora is installed with SELinux (page 379) turned off. Unless you need a very secure
system, you do not need to turn this feature on. Give the following command to install Fedora with SELinux turned
on:
boot: linux selinux
When you give this command to install Fedora, the Firewall screen allows you to set up SELinux (Figure 3-5, page
51).
Non-CD installations . If you are installing from other than CDs, that is, if you are installing from files on the
local hard disk or from files on another system using FTP, NFS, or HTTP, give the following command in response
to the boot: prompt:
boot: linux askmethod
Booting . As the system boots, text scrolls on the monitor, pausing occasionally. After a while (up to a few
minutes, depending on the speed of the system), the installer displays a graphical or pseudographical display, depending on the system you are installing and the commands you gave at the boot: prompt.
The balance of this section covers the commands you can give in response to the boot: prompt. Unless you are
having problems with the installation or have special requirements, you can skip to the next section, “The Anaconda
Installer” on page 46.
Boot Commands
All the commands (except for memtest86) you can give in response to the boot: prompt consist of the word linux
followed by an argument that is passed to the Anaconda installer. Many of the commands can be combined. For
example, to install Linux in text mode using a terminal running at 115,200 baud, no parity, 8 bits, connected to
the first serial device, give the following command (the ,115200n8 is optional):
boot: linux text console=ttyS0,115200n8
The next command installs Red Hat Linux in graphical mode (by default) on a monitor with a resolution of
1024x768, without probing for any devices. The installation program asks you to specify the source of the installation
data (CD, FTP site, or other).
boot: linux resolution=1024x768 noprobe askmethod
18
Draft
Boot Commands
Draft
Following are some of the commands you can give at the boot: prompt. Each command must be terminated with
RETURN.
RETURN . Without entering a command, press RETURN in response to the boot: prompt to perform a
graphical installation from CDs. This installation probes the computer to determine as much as possible about the
hardware.
memtest86 . FEDORA Calls memtest86 when you boot from a CD only. The GPL-licensed memtest86 utility
is a standalone memory test for x86-based computers. Press C to configure the test, ESCAPE to exit. See
www.memtest86.com [http://www.memtest86.com] for more information.
linux askmethod .
NFS, FTP, or HTTP.
Gives you a choice of installation sources: local CD or hard drive or over a network using
•
Local CD Displays the CD Found screen, which allows you to test the installation media (the same as if you
had just pressed RETURN).
•
Hard drive Prompts for the partition and directory that contains the ISO images of the installation CDs.
•
NFS , FTP , or HTTP Displays the Configure TCP/IP screen from which you can select DHCP or enter the
system’s IP address, netmask, default gateway (IP), and primary nameserver.
linux lowres .
Runs the installation program at a resolution of 640x480. See also linux resolution.
linux mem= xxx M .
in the computer.
Overrides the detected memory size. Replace xxx with the number of megabytes of RAM
linux mediacheck . Tests one or more installation CDs using an MD5 sum. This option works with the CD,
DVD, hard drive ISO, and NFS ISO installation methods. For more information, refer to “Check the CDs” on page
47.
FEDORA This test is performed automatically during a normal CD installation.
linux nofb . no framebuffer Turns off the framebuffer (video memory). Useful when problems occur when
the graphical phase of the installation starts. Particularly useful for systems with LCD displays.
linux noprobe . Disables hardware probing for all devices, including network cards (NICs), graphics cards,
and the monitor. Forces you to select devices from a list. You must know exactly which cards or chips the system
uses when you use this command. Use when probing causes the installation to hang or otherwise fail. This command
allows you to give arguments for each device driver you specify.
Linux rescue .
Puts the system in rescue mode; see page 377 for details.
linux resolution= WxH . Specifies the resolution of the monitor you are using for a graphical installation. For
example, resolution=1024x768 specifies a monitor with a resolution of 1024 by 768 pixels.
linux selinux . FEDORA Enables SELinux (page 379) on the system you are installing. When you give this
command to install Fedora, the Firewall screen allows you to set up SELinux (Figure 3-5, page 51).
linux skipddc .
Allows you to configure the monitor manually; see linux noprobe for more information.
linux text . Installs Linux using pseudographical mode. Although the images on the screen appear to be
graphical, they are composed entirely of text characters.
linux vnc . FEDORA Installs over a VNC (Virtual Network Computing) remote desktop session. After
providing an IP address, you can control the installation remotely using a VNC client from a remote computer.
19
Draft
The Anaconda Installer
Draft
You can download the VNC client, which runs on several platforms, from www.realvnc.com
[http://www.realvnc.com].
The Anaconda Installer
Anaconda, the program that installs Red Hat Linux, is written in Python and C, identifies the hardware, builds the
filesystems, and installs or upgrades the Red Hat Linux operating system. Anaconda can run in textual or graphical
(default) interactive mode or in batch mode (see “Using the Kickstart Configurator” on page 57).
Tip: Anaconda Does Not Write to the Hard Disk Until It Displays
the About to Install Screen
While you are installing Red Hat Linux, until Anaconda displays the About to Install screen, you can
press CONTROL-ALT-DEL to abort the installation process and reboot without making any changes
to the hard disk.
Exactly which screens Anaconda displays depends on whether you are installing Fedora or Red Hat Enterprise
Linux and which command(s) you specified following the boot: prompt (preceding). With some exceptions, notably
if you are running a text-mode installation, Anaconda starts by probing the video card, monitor, and mouse and
by starting a native X server with a log in /tmp/X.log. (This log is not preserved unless you complete the installation.)
While it is running, Anaconda opens the virtual consoles (page 103) shown in Table 3-1. You can display a virtual
console by pressing CONTROL-ALT-F x where F x is the function key that corresponds to the virtual console
number.
Table 3-1. Virtual Console Assignments During Installation
1
Installation dialog
2
Shell
3
Installation log
4
System messages
5
Miscellaneous messages
7
GUI interactive installation
At any time during the installation, you can go virtual console 2 (CONTROL-ALT-F2) and give commands to see
what is going on. Do not give any commands that change any part of the installation procedure.
Using Anaconda
Anaconda provides a NEXT button at the lower-right of each of the installation screens and a BACK button next
to it on most screens. When you have completed the entries on an installation screen, click NEXT , or, from a text
install, press the TAB key until the NEXT button is highlighted and then press RETURN. Select BACK to return
to the previous screen.
Initially, each of the graphical installation screens is divided into two columns: a narrow help column on the left
and information and prompts about the installation on the right. Select the Hide Help button (lower-left) to remove
the help column. Select Release Notes (next to Hide Help) to display the release notes for the version of Red Hat
Linux you are installing.
20
Draft
The Anaconda Installer
Draft
Caution: Check the CDs
FEDORA Because Red Hat does not manufacture Fedora disks, during a CD-based Fedora installation,
Anaconda displays the pseudographical CD Found screen before starting the installation. From this
screen, you can verify that the installation CDs do not have any errors. Checking the CDs takes a few
minutes and can save you hours of aggravation if the installation fails due to bad media.
RHEL+FEDORA You can force the display of the CD Found screen by giving the command linux
mediacheck in response to the boot: prompt (page 45).
Anaconda Screens
Following is a list of screens that Anaconda displays during a default installation. It may display different screens
depending on what commands you give and choices you make.
CD Found . FEDORA Allows you to check as many installation CDs as you like, in any order. Choose OK to
test the media, Skip to bypass the test. See the preceding TIP. This screen is displayed in pseudotext mode and
does not appear in two columns. Use the TAB key to move between choices; press RETURN to select the highlighted
choice.
During a graphical installation, when you leave the CD Found screen, Anaconda displays messages telling you it
is probing for the devices it will use during installation. After probing, it starts the X server.
Welcome . Displayed after Anaconda obtains enough information to start the X Window System. There is
nothing for you to do on this screen except hide the help panel and display the release notes. Select NEXT.
Language Selection . Select the language you want to use for the installation. This language is not necessarily
the same language that the installed system displays.
Keyboard Configuration .
Select the type of keyboard attached to the system.
Mouse . Displayed only if the type of mouse cannot be determined. Select the type of mouse attached to the
system. Mark the Emulate 3 buttons box if you have a two-button mouse and want the system to respond as
though you had pressed the middle button when you press the two mouse buttons at the same time.
Monitor . FEDORA Displayed only if the monitor cannot be probed successfully. Select the brand and model
of the monitor attached to the system. Select a generic LCD or CRT display if the monitor is not listed. You can
specify the Sync frequencies in place of the monitor brand and model, but be careful: Specifying the wrong values
can ruin some older hardware.
Upgrade Examine . If it detects a version of Red Hat Linux on the hard disk that it can upgrade, Anaconda
gives you the choice of upgrading the existing installation or overwriting the existing installation with a new one
(Figure 3-1). Refer to “Installing a Fresh Copy or Upgrading an Existing Red Hat System?” on page 26 for help in
making this selection.
21
Draft
The Anaconda Installer
Draft
Figure 3-1. Upgrade Examine screen
Installation Type . FEDORA Select the type of installation you want (Figure 3-2): Personal Desktop, Workstation,
Server, or Custom. Refer to “Types of Installations (FEDORA)” on page 27 for help in answering this question.
22
Draft
The Anaconda Installer
Draft
Figure 3-2. Installation Type screen
Disk Partitioning Setup . Select Automatically partition or Manually partition with Disk Druid. The automatic
selection gives you the option of reviewing and changing the partitions Anaconda sets up. For more information,
refer to “Using Disk Druid to Partition the Disk” on page 53.
Boot Loader Configuration . By default, Anaconda installs the grub boot loader (page 514). If you do not want
to install a boot loader, click Change boot loader and select Do not install a boot loader. When you install Red
Hat Linux on a machine that already runs another operating system, Anaconda frequently recognizes the other
operating system and sets up grub so you can boot from either operating system. Refer to “Setting Up a Dual-Boot
System” on page 61. You can manually add other operating systems to grub’s list of bootable systems by clicking
Add and specifying a label and device to boot from. For a more secure system, specify a boot loader password.
Network Configuration . Specify network configuration information (Figure 3-3). This screen has three frames:
Network Devices, Hostname, and Miscellaneous Settings. If you are using DHCP to set up the network interface,
you do not need to change anything.
23
Draft
The Anaconda Installer
Draft
Figure 3-3. Network Configuration window
The Network Devices frame lists the network devices the installer knows about. Normally, you want network
devices to become active when the system boots; remove the mark from the check box at the left of a device if you
do not want it to become active when the system boots.
To configure a network device manually (not using DHCP), highlight the device and click Edit to the right of the
list of devices. Anaconda displays the Edit Interface window (Figure 3-4). Remove the mark from the Configure
using DHCP check box and enter the IP address and netmask in the appropriate boxes before clicking OK.
24
Draft
The Anaconda Installer
Draft
Figure 3-4. Network Configuration: Edit Interface window
If you are not using DHCP, click manually under Set the hostname and enter the name of the system. When you
turn off DHCP configuration in Network Devices, Anaconda allows you to specify a gateway address, and one or
more DNS (name-server) addresses. You do not have to specify more than one DNS address, although it can be
useful to have two in case one nameserver stops working. Click Next.
Firewall . Set up a firewall (Figure 3-5). First, select No firewall or Enable firewall. If you select Enable firewall,
select the services that you want the firewall to allow to pass through to the system. Selecting WWW (HTTP) does
not allow HTTPS (secure HTTP), which is used for secure browser connections to financial institutions and when
giving credit card information, through the firewall. Specify https:tcp in Other ports to allow secure HTTP to
pass. If you have multiple network devices (one of which may be a dial-up line), you may want to consider one of
them to be trusted. Never trust a device connected to the Internet. Put a mark in the check box next to any trusted
devices in the box labeled If you would like to allow all traffic from a device, select it below. If Anaconda
displays the SELinux combo box shown in Figure 3-5, select the state you want SELinux (page 379) to be in when
you start the system. The chapter on iptables (page 737) has information on how to build a more complete and
functional firewall.
25
Draft
The Anaconda Installer
Draft
Figure 3-5. Firewall Configuration screen showing SELinux combo box
Additional Language Support . Select languages that the installed system will support. The default language
is the language the system will boot up in. Select as many other languages as you like from the list.
Time Zone . Specify the time zone the system is located in. Click on a location on the map or scroll through
the list and highlight the appropriate selection. Put a mark in the check box if your system clock is set to UTC (page
1004).
Root Password .
Specify the root password.
After you specify the root password, Anaconda pauses to read the software package information.
Package Installation Defaults . In this screen, Anaconda lists the packages it will install by default. Choose
Install default software packages to install this list of packages or Customize software packages to be installed
to modify the list. Regardless of your selection, you can change which packages are installed on a system at any
time; refer to “system-config-packages: Adds and Removes Software Packages” on page 453.
Caution: Install KDE to Follow the Examples in Chapter 4
Chapter 4, “Introduction to Red Hat Linux,” uses examples from KDE to introduce the graphical
desktop. Install KDE if you want to follow these examples. You can remove KDE later if you like.
Package Group Selection . If you choose to accept the default list of software packages (previous screen),
Anaconda skips this screen. Select the groups and packages you want to install. For more information, refer to
“system-config-packages: Adds and Removes Software Packages” on page 453.
About to Install .
on the hard disk.
26
Anaconda displays the About to Install screen just before it starts installing Red Hat Linux
Draft
When You Reboot
Draft
Caution: This Is When Anaconda Writes to the Hard Disk
You can abort the installation by pressing CONTROL-ALT-DEL at any point up to and including
this screen without making any changes to the system. Once you press NEXT in the About to Install
screen, Anaconda writes to the hard disk.
Installing . After you press NEXT, Anaconda installs Red Hat Linux. Depending on the number of software
packages you are installing, this process can take quite a while. Anaconda keeps you informed of its progress and
requests CDs (if you are installing from CDs) as it needs them.
When You Reboot
When Anaconda instructs you to do so, remove the installation CD if present and reboot the system. The system
boots Red Hat Linux and asks a few questions before allowing you to log in.
License Agreement . First, Anaconda displays the Welcome screen, followed by the License Agreement screen.
Select Yes I agree to the License Agreement if you agree with the terms of the licence agreement.
Date and time . The next screen allows you to set the date and time. Running the Network Time Protocol (NTP)
causes the system clock to reset itself periodically from a clock on the Internet. If the system is connected to the
Internet, you may want to select Enable Network Time Protocol and choose a server from the combo box.
Display . Next, Anaconda displays the Display screen, which allows you to specify the resolution and color
depth of the monitor.
User Account . The next screen allows you to set up user accounts. You can set up user accounts now or once
the system is fully operational. For more information, refer to “Configuring User and Group Accounts” on page
521.
Sound Card . The Sound Card window identifies the sound card and has a button that can play a test sound.
There is nothing that you can configure from the Sound Card window.
Additional CDs .
Finish Setup .
screen.
Next, the system asks if you have additional CDs for installing more software packages.
Finally, from the Finish Setup screen, click NEXT to complete the setup and display the login
When you leave the Finish Setup screen, you are done with the installation. You can use the system and set it up
as you desire. You may want to customize the desktop as explained in Chapters 4 and 8 or set up servers as discussed
in Part V of this book.
Initializing Databases
After booting the system, log in as, or su to, root and update the whatis database so that whatis (page 137) and
apropos (page 137) work properly. Then update the slocate database so that slocate works properly. (The slocate
[secure locate] utility allows you to index and search for files on your system quickly and securely.) Instead of updating these databases when you install the system, you can wait for cron to run them overnight, but whatis,
apropos, and slocate will not work until the next day. The best way to update these databases is to run the cron
scripts that run them daily. Working as root, give the following commands:
# /etc/cron.daily/makewhatis.cron
# /etc/cron.daily/slocate.cron
27
Draft
Installation Tasks
Draft
These utilities run for up to several minutes and may complain about not being able to find a file or two. When
you get the prompt back, your whatis and slocate data-bases are up-to-date.
Installation Tasks
This section details some common tasks you may need to perform during or after installation. It covers using Disk
Druid to partition the disk during installation, using fdisk to modify partitions after installation, using Kickstart to
automate installation, and setting up a system that will boot either Windows or Linux (a dual boot system).
Using Disk Druid to Partition the Disk
Disk Druid, a graphical disk-partitioning program that can add, delete, and modify partitions on a hard drive, is
part of the Red Hat installation system. You can use Disk Druid only while you are installing a system: It cannot
be run on its own. You can use fdisk (page 58) to manipulate partitions after you install Red Hat Linux. As explained
earlier, if you want a basic set of partitions, you can allow Disk Druid to partition the hard drive automatically.
Clone and RAID . Disk Druid includes Clone, a tool that copies the partitioning scheme from a single drive to
as many other drives as needed. The Clone option is useful for making multiple copies of a RAID partition/drive
when you are creating a large RAID array of identical partitions or identically partitioned drives. Click the RAID
button to access the Clone tool, which is active only when at least one unallocated RAID partition exists.
Disk Partitioning Setup . During installation, the Disk Partitioning Setup screen gives you the choice between
automatically and manually partitioning the hard disk. Choose automatic partitioning if you do not want to make
any decisions about how to split up the hard disk.
Automatic Partitioning .
The Automatic Partitioning screen gives you the following choices:
•
Remove all Linux partitions Removes all Linux partitions, deleting the data on those partitions and creating
one or more chunks of free space (page 972) on the disk. You can create new partitions using the free space. If
there is only a Linux system on the disk, this choice is the same as the next one.
•
Remove all partitions Deletes all the data on the disk and gives you a free space the size of the disk to work
with, as though you were working with a new drive.
•
Keep all partitions and use existing free space Forces you to install Red Hat Linux in the free space on the
disk. Does not work if there is not enough, free space.
Tip: The Disk Is Not Partitioned Until Later
Disk Druid does not write to the hard disk; it creates a table that specifies how you want the hard disk
partitioned. The disk is partitioned and formatted when you click Next from the About to Install
screen.
Choosing automatic partitioning causes anaconda to split the free space, if possible, into three partitions: /boot
(about 100 megabytes), swap (up to a few thousand megabytes), and / (root), which gets the bulk of the disk space.
The Automatic Partitioning screen has a frame that allows you to choose the hard disk you want to install Red Hat
Linux on. This frame has meaning only if you have more than one hard disk. At the bottom of the screen is a check
box, Review (and modify if needed) the partitions created. Put a check mark in this box to display the initial
Disk Druid window (Figure 3-6) when you click Next.
28
Draft
Using Disk Druid to Partition the Disk
Draft
Figure 3-6. Disk Druid: Main screen with Help hidden
Disk Setup . When you choose to partition the hard disk manually, Anaconda displays the Disk Setup screen,
which is the Disk Druid main screen, shown in Figure 3-6 with the Help column hidden. This screen has three
sections, from the top: a graphical representation of the disk drive(s) showing how each is partitioned, a row of
command buttons, and a graphical table listing one partition per line. If you are starting with a new disk, no partitions
are listed on the screen.
The row of command buttons has the following buttons:
•
New Adds a new partition to the disk (page 55).
•
Edit Edits the highlighted partition (page 56).
•
Delete Deletes the highlighted partition.
•
Reset Cancels the changes you have made and causes the Disk Druid table to revert so it matches the layout
of the disk.
•
RAID Enables you to create software RAID partitions and to join two or more RAID partitions into a RAID
device (page 31).
•
LVM Enables you to create LVM physical volumes, which you can then use to create LVM logical volumes
(page 32).
The Disk Druid table has the following columns:
•
Device The name of the device in the /dev directory (for example, /dev/hda1).
•
Mount Point/RAID/Volume Specified where the partition will be mounted when the system is brought up (for
example, /usr). Also used to specify the RAID device or LVM volume the partition is part of.
29
Draft
Using Disk Druid to Partition the Disk
Draft
•
Type The type of partition, such as ext3, swap, or LVM.
•
Format A check mark in this column indicates the partition will be formatted as part of the installation procedure.
All data on the partition will be lost.
•
Size The size of the partition in megabytes.
•
Start The number of the block the partition starts on.
•
End The number of the block the partition ends on.
At the bottom of the screen is a check box that allows you to hide RAID device and LVM volume group members.
Do not check this box if you want to see all the information about the disk drives.
Add a new partition . In order to add a new partition to a hard drive, there must be enough free space on the
hard drive to accommodate the partition. Click the New button to add a partition; Disk Druid displays the Add
Partition window (Figure 3-7). Specify the mount point (the name of the directory that the partition will be
mounted over [page 442]) and the filesystem type; use the arrow buttons at the right end of these text boxes to
display drop-down menus of choices. If there is more than one drive, mark the check box next to the drive you
want the partition to be created on in the Allowable Drives frame. Specify the desired size of the partition and, in
the Additional Size Options frame, mark Fixed size to create the partition close to the size you specify. Because of
block-size constraints, partitions are not usually exactly the size you specify. Mark Fill all space up to (MB) and
fill in the maximum size you want the partition to be to create a partition that takes up the existing free space, up
to the maximum size you specify. In other words, Disk Druid does not complain if it cannot create the partition as
large as you would like. Mark the third choice, Fill to maximum allowable size, to cause the partition to occupy
all the remaining free space on the disk, regardless of size. Mark the Force to be a primary partition check box
to create a primary partition. Click OK, and Disk Druid adds the partition to its table (but does not write to the
hard disk).
Figure 3-7. Disk Druid: Add Partition window
30
Draft
Using the Kickstart Configurator
Draft
Edit an existing partition . To modify an existing partition, highlight the partition in the Disk Druid table or
the graphical representation of the disk drive and click the Edit button; Disk Druid displays the Edit Partition
window (Figure 3-8). From this window, you can change the mount point of a partition or format the partition as
another type (ext3, vfat, swap, and so on). You cannot change the size of a partition from this window; instead,
you must delete the partition and create a new partition of the desired size.
Figure 3-8. Disk Druid: Edit Partition window
Using the Kickstart Configurator
Kickstart is Red Hat’s program that completely or partially automates the same installation and postinstallation
configuration on one or more machines. You create a single file that answers all the questions that are normally
asked during an installation. Then the installation script refers to this file instead of asking you the questions. Using
Kickstart, you can automate language selection, network configuration, keyboard selection, boot loader installation,
disk partitioning, mouse selection, X Window System configuration, and more.
The system-config-kickstart (FEDORA) and redhat-config-kickstart (RHEL) utilities run the Kickstart Configurator
(Figure 3-9), which creates a Kickstart installation script.
31
Draft
fdisk: Reports On and Partitions a
Hard Disk
Draft
Figure 3-9. Kickstart Configurator
Figure 3-9 shows the first window the Kickstart Configurator displays. The first text box, Language, is the language
that will be used for installation. The Language Support box, toward the bottom of the window, is the language
that the new system will use after installation.
To generate a Kickstart file (ks.cfg by default), go through each section of this window (along the left side) and fill
in the answers and mark the appropriate boxes. Click the Help button for instructions on completing these tasks.
When you are finished and click Save File, the Kickstart Configurator gives you a chance to review the generated
script before it saves the file.
fdisk: Reports On and Partitions a Hard Disk
The fdisk utility reports on and manipulates hard disk partitions. You can display the size (in 1024-byte blocks) of
a hard disk by using the fdisk with the –s option followed by the device name of the hard drive:
# fdisk -s /dev/hda
12714912
Run fdisk with the –l option to display information about partitions on the drives you specify:
32
Draft
fdisk: Reports On and Partitions a
Hard Disk
Draft
# /sbin/fdisk -l /dev/hda
Disk /dev/hda: 123.5 GB, 123522416640 bytes
255 heads, 63 sectors/track, 15017 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device Boot
/dev/hda1
*
/dev/hda2
/dev/hda3
/dev/hda4
/dev/hda5
/dev/hda6
/dev/hda7
/dev/hda8
/dev/hda9
/dev/hda10
/dev/hda11
/dev/hda12
/dev/hda13
/dev/hda14
Start
1
14
2564
5114
5114
7664
10214
11489
12126
12253
12380
12507
12634
12888
End
13
2563
5113
15017
7663
10213
11488
12125
12252
12379
12506
12633
12887
15017
Blocks
104391
20482875
20482875
79553880
20482843+
20482843+
10241406
5116671
1020096
1020096
1020096
1020096
2040223+
17109193+
Id
83
83
83
f
83
83
83
83
83
83
83
83
82
8e
System
Linux
Linux
Linux
Win95 Ext'd (LBA)
Linux
Linux
Linux
Linux
Linux
Linux
Linux
Linux
Linux swap
Linux LVM
In the preceding output, /dev/hda4 is an extended partition (page 28), /dev/hda13 is a swap partition (page 435),
and /dev/hda14 specifies a part of the disk that is managed by LVM (page 32).
In addition to reporting on the layout and size of a disk drive, you can use fdisk interactively to modify the layout.
Be extremely careful when using fdisk in this manner, and always back up the system before starting. Changing the
partition information (the partition table) on a disk destroys the information on the disk. Read the fdisk man page
and the Linux Partition mini-HOWTO before modifying a partition table.
Caution: fdisk Can Destroy Everything
Be as careful with fdisk as you would be with a utility that formats a hard drive. Changes you make
with fdisk can easily result in the loss of large amounts of data. If you are using fdisk and have any
question about what you are doing, back out with q (quit without saving changes)—your changes do
not take effect until you exit from fdisk.
To partition a disk, start fdisk without any options and give an m command, which displays the following help
message. You can safely ignore a warning message about the number of cylinders being too large.
# /sbin/fdisk /dev/hda
Command (m for help): m
Command action
a
toggle a bootable flag
b
edit bsd disklabel
c
toggle the dos compatibility flag
d
delete a partition
l
list known partition types
m
print this menu
n
add a new partition
o
create a new empty DOS partition table
33
Draft
fdisk: Reports On and Partitions a
Hard Disk
p
q
s
t
u
v
w
x
Draft
print the partition table
quit without saving changes
create a new empty Sun disklabel
change a partition's system id
change display/entry units
verify the partition table
write table to disk and exit
extra functionality (experts only)
When you choose p (print), fdisk displays the current partitions on the disk:
Command (m for help): p
Disk /dev/hda: 123.5 GB, 123522416640 bytes
255 heads, 63 sectors/track, 15017 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device Boot
Start
End
Blocks
Id
/dev/hda1
*
1
13
104391
83
/dev/hda2
14
2563 20482875
83
/dev/hda3
2564
5113 20482875
83
/dev/hda4
5114
15017 79553880
f
/dev/hda5
5114
7663 20482843+ 83
/dev/hda6
7664
10213 20482843+ 83
/dev/hda7
10214
11488 10241406
83
/dev/hda8
11489
12125
5116671
83
/dev/hda9
12126
12252
1020096
83
/dev/hda10
12253
12379
1020096
83
/dev/hda11
12380
12506
1020096
83
/dev/hda12
12507
12633
1020096
83
/dev/hda13
12634
12887
2040223+ 82
/dev/hda14
12888
15017 17109193+ 8e
System
Linux
Linux
Linux
Win95 Ext'd (LBA)
Linux
Linux
Linux
Linux
Linux
Linux
Linux
Linux
Linux swap
Linux LVM
Each disk entry includes the filename within the /dev directory, the cylinders the partition starts and ends on (use
u to change the units to sectors), the number of 1024 (1 kilobyte) blocks, the ID number of the partition type (l
[ell] lists the partition types), and the name of the partition type.
In the preceding example partition 4 defines an extended partition that includes almost the entire disk. You cannot
make changes to this partition without affecting all the partitions within it. Following are guidelines to remember
when defining a partition table for a disk. For more information, refer to “Partitioning a Disk” on page 29.
•
Do not modify the partition that defines the entire disk. This partition is called the backup, or overlap, partition.
•
Do not overlap partitions that contain data. If the first partition ends at cylinder 187, the second partition should
begin at cylinder 188. If you overlap partitions, the filesystem will become corrupt as data for overlapping files
is written to the disk. Because the overlap partition contains no data or filesystem, it is safe to have other partitions
overlap it.
•
Never put a raw partition on cylinder 0. A few sectors on cylinder 0 are reserved for such things as the disk label,
bad blocks, and partition tables. A filesystem will preserve this information, but when you use cylinder 0 as
part of a raw partition, this information is deleted, and the disk may become unusable. An example of a raw
partition is a swap partition (page 435) or raw database partition, as used for a Sybase or Oracle database. Start
all raw partitions at cylinder 1 or greater.
•
It is a good idea to put /boot at the beginning of the drive (partition 1) so that there is no issue of Linux having
to boot from a partition too far into the drive. When you can afford the disk space, it is desirable to put each
34
Draft
Setting Up a Dual-Boot System
Draft
major filesystem on a separate partition. Many people choose to combine root, /var, and /usr into a single
partition, which generally results in less wasted space but can, on rare occasions, cause problems. You can also
put the contents of the boot directory in the root filesystem.
•
When using megabytes to specify the size of a partition, remember to check how many cylinders have been allocated so you know where to begin the next partition.
•
Use tune2fs (page 447) to make all partitions, except swap and /boot, type ext3, unless you have a reason to
do otherwise.
The following sequence of commands defines a 300 megabyte, bootable, Linux partition as partition 1 on a clean
disk:
# /sbin/fdisk /dev/hda
Command (m for help): n
(create new partition)
Command action
l
logical (5 or over)
p
primary partition (1-4)
p
(select primary partition)
Partition number (1-4): 1
(select partition number 1)
First cylinder (1-1582, default 1):
(allow first cylinder to default to 1)
Using default value 1
Last cylinder or +size or +sizeM or +sizeK (1-2, default 2): +300M (300 MB partition)
Command (m for help): t
(set partition type)
Partition number (1-12): 1
(specify which partition)
Hex code (type L to list codes): 83
(83 is Linux, 82 is Linux swap, press L for a list)
Changed system type of partition 1 to 83
Command (m for help): a
Partition number (1-12): 1
(specify a bootable partition)
(specify partition 1 as bootable)
After defining a partition using k or m to specify kilobytes or megabytes, run p to check for the ending cylinder.
Do this before defining the next contiguous partition so that you do not waste space or have any overlap. After
setting up all the partitions and exiting from fdisk with a w command, make a filesystem (mkfs, page 397) on each
partition that is to hold a filesystem (not swap). Use mkswap (page 435) to create a swap partition. You can use
e2label (page 396) to label partitions.
Setting Up a Dual-Boot System
Caution: Windows and Dual Boot
Installing Fedora Core 2 in dual boot configurations with Windows XP or 2K can result in the Windows
partition becoming inaccessible. The problem is caused when Anaconda modifies the partition table
in the master boot record (MBR). THIS PROBLEM CAN OCCUR EVEN WHEN YOU DO NOT
MODIFY THE PARTITION TABLE.
Please check the Fedora site (fedora.redhat.com [http://www.fedora.redhat.com]) and the author’s
Web page (www.sobell.com [http://www.sobell.com]) for the latest information.
A dual-boot system is one that can boot one of two operating systems. Dual-boot in this section refers to a system
that can boot Windows or Linux. The biggest problem in setting up a dual-boot system, assuming you want to add
35
Draft
The X Window System
Draft
Linux to a Windows system, is finding disk space for Linux. The Linux+WindowsNT mini-HOWTO covers installing
Linux first and Windows NT second or the other way around. The next section discusses several ways to create
the needed space.
Creating Free Space on a Windows System
Typically, you install Red Hat Linux in free space on a hard disk. In order to add Red Hat Linux to a Windows
system, you must provide enough free space (refer to “Disk space” on page 25) on a hard disk that already contains
Windows. There are several ways to provide/create free space. Following are some ways, from easiest to most difficult:
Use existing free space . If there is sufficient free space on the Windows disk, you can install Linux there. This
technique is best, but there is rarely enough free space on an installed hard disk.
Add a new disk drive . Add another disk drive to the system and install Linux on the new disk, which contains
only free space. This technique is very easy and clean but requires a new disk drive.
Remove a Windows partition . If you can delete a big enough Windows partition, you can install Linux in its
place. In order to delete a Windows partition, you must have multiple partitions under Windows and be willing
to lose any data in the partition you delete. In many cases, you can move the data from the partition you delete to
another Windows partition.
Once you are sure a partition contains no useful information, you can use Disk Druid to delete it when you install
Linux: From the Disk Partition screen (page 49), choose to partition the disk with Disk Druid manually, highlight
the partition you want to delete, and click the Delete button. After deleting the partition, you can install Red Hat
Linux in the free space left by the partition you removed.
Installing Red Hat Linux as the Second Operating System
After creating enough free space on a Windows system (previous section), start installing Red Hat Linux. When
you get to the Disk Partitioning Setup screen (page 49), you must choose manual partitioning if you need to delete
a Windows partition. You can choose either automatic or manual partitioning if you already have free space to install
Red Hat Linux in. If you partition the disk automatically, choose to keep all partitions and install Red Hat Linux
in free space (page 27). If you partition the disk manually, use Disk Druid (page 53) to delete the appropriate
Windows partition if necessary and create the Red Hat Linux partitions in the free space. When you boot, you will
be able to choose which operating system you want to run.
The X Window System
With the introduction of Fedora Core 2, Red Hat replaced the XFree86 X Window System with the X.org Foundation
X11R6.7.0 X Window System (X.org and freedesktop.org). The X.org X server is functionally equivalent to the one
distributed by XFree86 because most of the code is the same. See “XFree86 and X.org” on page 218 for more information on the transition. For more information on the X.org release of X, go to
freedesktop.org/~xorg/X11R6.7.0/doc/RELNOTES.html
[http://freedesktop.org/~xorg/X11R6.7.0/doc/RELNOTES.html]. Red Hat Enterprise Linux and Fedora Core 1
use XFree86.
Tip: The XF86Config and xorg.conf Files
The X Window System configuration files are kept in the /etc/X11 directory. Under XFree86 (Fedora
Core 1 and Red Hat Enterprise Linux), the primary configuration file is named XF86Config; under
X.org (Fedora Core 2 and later), the file is named xorg.conf. They are basically the same file with
two names. If X.org does not find xorg.conf, it uses XF86Config.
36
Draft
system-config-display: Configuring the
Display
Draft
If you have an XF86Config file you want to move to Fedora Core 2, copy it and rename it xorg.conf.
If you specified any kind of graphical desktop, such as GNOME or KDE, you installed X.org or XFree86 when you
installed Linux. X.org and XFree86 each comprise almost twenty rpm packages; the easiest way to install X.org or
XFree86 on an already installed Linux system is to use [system|redhat]-config-packages (page 453).
Most of the X software is installed under the /usr/X11R6 directory. The /usr/X11R6/bin directory contains the
utilities that are part of X and makes interesting browsing. This directory contains most of the utilities whose names
begin with x, such as xterm, xmag, and xeyes. The configuration files are kept in /etc/X11; the file used to guide
the initial setup is /etc/X11/xorg.conf (FEDORA) or /etc/X11/XF86Config (RHEL).
system-config-display: Configuring the Display
The easiest way to configure X.org or XFree86 is to run system-config-display (FEDORA) or redhat-config-xfree86
(RHEL), both of which display the Display settings window. The two utilities display different windows, with
redhat-config-xfree86 offering a subset of the options offered by system-config-display. This section explains how
to use system-config-display. The redhat-config-xfree86 Display and Advanced tabs display the same information
as the system-config-display Settings and Hardware tabs respectively.
Figure 3-10 shows the Settings tab of the Display settings window where you can specify the resolution and color
depth for the monitor. Normally, the system probes the monitor and fills in these values. If not, check the specifications for the monitor and select the appropriate values from these combo boxes. It is all right to specify a lower
resolution than a monitor is capable of, but you can damage an older monitor by specifying a resolution higher
than the monitor is capable of. A color depth of 8 bits equates to 256 colors, 16 bits to thousands of colors, and
24 or 32 bits to millions of colors.
37
Draft
system-config-display: Configuring the
Display
Draft
Figure 3-10. Display Settings Window, Settings tab
Next, click the Hardware tab. Again, the system normally probes for the monitor type and brand and model of
video card; these values appear next to the words Monitor Type and Video Card. You can manually select a
monitor or video card. Figure 3-11 shows the Monitor selection window superimposed on the Hardware tab of the
Display settings window.
38
Draft
system-config-display: Configuring the
Display
Draft
Figure 3-11. Display Settings window, Hardware tab
Specify a monitor . To specify a monitor, click Configure across from the words Monitor Type; system-configdisplay displays the Monitor window. Scroll down until you see the manufacturer of monitor you are using and
click the triangle to the left of the name of the manufacturer; system-config-display opens a list of models made by
that manufacturer. Scroll through the list of models. Click to highlight the model you are using; click OK. If an
appropriate model is not listed, scroll to the top of the list and click the triangle next to Generic CRT Display or
Generic LCD Display, depending on the type of display you are setting up. From one of these lists, select the
maximum resolution your monitor is capable of. Click OK.
Specify a video card . To specify a video card, click Configure adjacent to the words Video Card; systemconfig-display displays the Video Card window. Scroll down until you see the manufacturer and model of the video
card in your system. Click OK.
Specify two monitors . The Dual head tab allows you to specify a second video card that can drive a second
monitor. Specify the monitor type, video card, resolution, and color depth as you did earlier. You can choose to
have each monitor display a desktop or to have the two monitors display a single desktop (spanning desktops).
Click OK to close the Display settings window.
The system-config-display utility generates an xorg.conf (FEDORA) or XF86Config(RHEL) file (next section)
with the information you entered.
39
Draft
The xorg.conf and XF86Config Files
Draft
The xorg.conf and XF86Config Files
The xorg.conf (FEDORA) and XF86Config (RHEL) files comprise sections that can appear in any order. The
format of a section is
Section "name"
entry
...
EndSection
where name is the name of the section. A typical entry occupies multiple physical lines but is actually one logical
line, consisting of a keyword followed by zero or more integer, real, or string arguments. Keywords in these files
are not case sensitive and underscores (_) within keywords are ignored. Most strings are not case sensitive and
SPACEs and underscores in most strings are ignored. All strings must appear within double quotation marks.
The Option keyword provides free-form data to server components and is followed by the name of the option and
optionally a value. All Option values must be enclosed within double quotation marks.
Boolean Options take a value of TRUE (1, on, true, yes) or FALSE (0, off, false, no); no value is the same as
TRUE. You can prepend No to the name of a Boolean Option to reverse the sense of the Option.
The sections that can appear in an xorg.conf or XF86Config file are
ServerFlags
Global Options (optional)
ServerLayout
Binds Screen(s) and InputDevice(s)
Files
Locations of configuration files
Module
Modules to be loaded (optional)
InputDevice
Keyboard(s) and pointer(s)
Monitor
Monitor(s)
Device
Video card(s)
Screen
Binds device(s) and monitor(s)
VideoAdaptor
Configures the Xv extension (optional)
Modes
Video modes (optional)
DRI
Direct Rendering Infrastructure (optional)
Vendor
Vendor-specific information (optional)
This chapter covers the sections you most likely need to work with: ServerLayout, InputDevice, Monitor, Device,
and Screen. The excerpts from the xorg.conf file used in this chapter come from a file generated by system-configdisplay.
ServerLayout
The ServerLayout section appears first in some xorg.conf and XF86Config files because it summarizes the other
sections that are used to specify the server. The following ServerLayout section names the server single head
configuration and specifies that the server comprises the sections named Screen0, Mouse0, Keyboard0, and
DevInputMice.
The term core in this file means primary; there must be exactly one CoreKeyboard and one CorePointer. The AlwaysCore argument indicates that the device reports core events and is used here to allow a non-USB and a USB
40
Draft
The xorg.conf and XF86Config Files
Draft
mouse to work at the same time. The result is that you can use either type of mouse interchangeably without
modifying the xorg.conf or XF86Config file:
Section "ServerLayout"
Identifier "single head configuration"
Screen
0 "Screen0" 0 0
InputDevice "Mouse0" "CorePointer"
InputDevice "Keyboard0" "CoreKeyboard"
InputDevice "DevInputMice" "AlwaysCore"
EndSection
Refer to the following sections for explanations of the sections specified in Server-Layout.
InputDevice
There must be at least two InputDevice sections: one specifying the keyboard and one specifying the pointer (usually
a mouse). The format of an InputDevice section is
Section "InputDevice"
Identifier "id_name"
Driver "drv_name"
options
...
EndSection
where id_name is a unique name for the device and drv_name is the driver to use for the device, typically keyboard
or mouse. The system-config-display and redhat-config-xfree86 utilities typically create three InputDevice sections.
The following section defines a keyboard device named Keyboard0 that uses the keyboard driver. The keyboard
model is a 105-key PC keyboard. You can change pc105 to microsoft if you are using a US Microsoft Natural
keyboard, although the differences are minimal. See www.xfree86.org/current/XKB-Config.html
[http://www.xfree86.org/current/XKB-Config.html] for a complete list of keyboard (XKB) options.
Section "InputDevice"
Identifier "Keyboard0"
Driver
"keyboard"
Option
"XkbModel" "pc105"
Option
"XkbLayout" "us"
EndSection
To change the language the keyboard supports, change the argument to the XkbLayout Option to, for example,
fr for French.
The next InputDevice section defines a mouse named Mouse0 that uses the mouse driver. The Device Option
specifies a PS2 device. The ZAxisMapping Option maps the Z axis, the mouse wheel, to virtual mouse buttons 4
and 5 that are used to scroll a window. For more information, refer to “Remapping Mouse Buttons” on page 223.
When set to YES, the Emulate3Buttons Option enables the user of a two-button mouse to emulate a 3-button
mouse by pressing the two buttons simultaneously. See www.xfree86.org/current/mouse.html
[http://www.xfree86.org/current/mouse.html] for a complete list of mouse options.
Section "InputDevice"
Identifier "Mouse0"
41
Draft
The xorg.conf and XF86Config Files
Driver
Option
Option
Option
Option
EndSection
Draft
"mouse"
"Protocol" "IMPS/2"
"Device" "/dev/psaux"
"ZAxisMapping" "4 5"
"Emulate3Buttons" "no"
The next InputDevice section is similar to the previous one except the Device Option specifies a USB mouse. See
“ServerLayout” on page 66 for a discussion.
Section "InputDevice"
# If the normal CorePointer mouse is not a USB mouse then
# this input device can be used in AlwaysCore mode to let you
# also use USB mice at the same time.
Identifier "DevInputMice"
Driver
"mouse"
Option
"Protocol" "IMPS/2"
Option
"Device" "/dev/input/mice"
Option
"ZAxisMapping" "4 5"
Option
"Emulate3Buttons" "no"
EndSection
Monitor
The xorg.conf and XF86Config files must have at least one Monitor section. The easiest way to set up this section
is to use the system-config-display or redhat-config-xfree86 utility, which either determines the type of monitor
automatically by probing or allows you to select from a list of monitors.
Caution: Do Not Guess at Values for HorizSync or VertRefresh
If you configure the Monitor section manually, do not guess at the scan rates (HorizSync and VertRefresh); on older monitors, you can destroy the hardware by choosing scan rates that are too high.
The following section defines a monitor named Monitor0. The VendorName and ModelName are for reference
only and do not affect the way the system works. The optional DisplaySize specifies the height and width of the
screen in millimeters, allowing X to calculate the DPI of the monitor. HorizSync and VertRefresh specify ranges of
vertical refresh frequencies and horizontal sync frequencies for the monitor. These values are available from the
manufacturer. The dpms Option specifies the monitor is DPMS (page 969) compliant (has built-in energy saving
features).
Section "Monitor"
Identifier
VendorName
ModelName
DisplaySize
HorizSync
VertRefresh
Option
EndSection
42
"Monitor0"
"Monitor Vendor"
"Dell D1028L"
360 290
31.0 - 70.0
50.0 - 120.0
"dpms"
Draft
The xorg.conf and XF86Config Files
Draft
Your Monitor section may mention DDC (Display Data Channel); DDC can be used by a monitor to inform a video
card about its properties.
Device
The xorg.conf and XF86Config files must have at least one Device section to specify the type of video card in the
system. The VendorName and BoardName are for reference only and do not affect the way the system works. The
easiest way to set up this section is to use the system-config-display or redhat-config-xfree86 utility, both of which
usually determine the type of video card by probing. The following Device section specifies that Videocard0 uses
the tdfx driver:
Section "Device"
Identifier
Driver
VendorName
BoardName
EndSection
"Videocard0"
"tdfx"
"Videocard vendor"
"Voodoo3 (generic)"
Screen
The xorg.conf and XF86Config files must have at least one Screen section. This section binds a video card specified
in the Device section with a display specified in the Monitor section. The following Screen section specifies that
Screen0 comprises Videocard0 and Monitor0, both defined elsewhere in the file. The DefaultDepth entry specifies
the default color depth (page 963), which can be overridden in the Display subsection (next).
Each Screen section must have at least one Display subsection. The following subsection specifies a color Depth
and three Modes. The modes specify screen resolutions in dots-per-inch (DPI). The first mode is the default; you
can switch between modes while X is running by pressing CONTROL-ALT-KEYPAD+ or CONTROL-ALTKEYPAD–. You must use the plus or minus on the numeric keypad when giving these commands. X ignores invalid
modes.
Section "Screen"
Identifier "Screen0"
Device
"Videocard0"
Monitor
"Monitor0"
DefaultDepth
24
SubSection "Display"
Depth
24
Modes
"1024x768" "800x600" "640x480"
EndSubSection
EndSection
Multiple Monitors
X has supported multiple screens for a long time. X.org and XFree86 support multimonitor configurations using
either two graphics cards or a dual-head card. Both of these setups are usually configured the same way because
the drivers for dual-head cards provide a secondary virtual device.
Traditionally, each screen in X is treated as a single entity. Each window must be on one screen or another. More
recently, the Xinerama extension allows windows to be split across two or more displays. This extension is supported
by X.org and XFree86 and works with most video drivers. When using Xinerama, you must set all screens to the
same color depth.
43
Draft
gdm: Displays a Graphical Login
Draft
For each screen, you must define a Device, Monitor, and Screen section in the xorg.conf or XF86Config file.
These sections are exactly the same as for a single screen configuration; each screen must have a unique identifier.
If you are using a dual-head card, the Device section for the second head is likely to require a BusID value to enable
the driver to determine that you are not referring to the primary display. The following section identifies the two
heads on an ATi Radeon 8500 card. For other dual-head cards, consult the documentation provided with the driver
(for example, give the command man mga to display information on the mga driver):
Section "Device"
Identifier
Driver
VendorName
BoardName
EndSection
Section "Device"
Identifier
Driver
VendorName
BoardName
BusID
EndSection
"Videocard0"
"radeon"
"ATi"
"Radeon 8500"
"Videocard1"
"radeon"
"ATi"
"Radeon 8500"
"PCI:1:5:0"
Once you have defined the screens, use the ServerLayout section to tell X where they are in relation to each other.
Each screen is defined in the following form:
Screen ScreenNumber “Identifier” Position
The ScreenNumber is optional. If omitted, X numbers screens in the order they are specified, starting with 0. The
Identifier is the same Identifier used in the Screen sections. The Position can be either absolute or relative. The
easiest way to define screen positions is to give one screen an absolute position, usually with the coordinates of the
origin, and then use the LeftOf, RightOf, Above, and Below keywords to indicate the positions of the other screens:
Section "ServerLayout"
Identifier
Screen
0
Screen
1
InputDevice
InputDevice
InputDevice
Option
Option
EndSection
"Multihead layout"
"Screen0" LeftOf "Screen1"
"Screen1" 0 0
"Mouse0" "CorePointer"
"Keyboard0" "CoreKeyboard"
"DevInputMice" "AlwaysCore"
"Xinerama" "on"
"Clone" "off"
Two options can control the behavior of a multimonitor layout: Xinerama causes the screens to act as if they were
a single screen and Clone causes each of the screens to display the same thing.
gdm: Displays a Graphical Login
Traditionally, users were expected to log in on a text-based terminal and then start the X server. Today, most
desktop systems and workstations provide a graphical login. Red Hat Linux uses the GNOME display manager
(gdm) to provide this functionality, even if you are bringing up a KDE desktop.
44
Draft
More Information
Draft
Configuring gdm
The gdmsetup utility configures the login presented by gdm by editing the heavily commented
/etc/X11/gdm/gdm.conf file. By default, root can log in both locally and remotely. It is usually a good idea to
disable remote root logins because, when a user logs in remotely using gdm, the password is sent in cleartext across
the network.
Using kdm
The kdm utility is the KDE equivalent of gdm. There is no benefit in using kdm in place of gdm: Both perform the
same function. Using gdm does not force you to use GNOME.
The configuration file for kdm, /etc/X11/xdm/kdmrc, is heavily commented. You can edit the kdm configuration
using the KDE control panel, but doing so removes the comments from the file.
More Information
Web .
XFree86 .
DRI .
xfree86.org, xfree86.org/current has README files on many topics.
www.xfree86.org/current/DRI.html [http://www.xfree86.org/current/DRI.html]
Mouse Configuration .
www.xfree86.org/current/mouse.html [http://www.xfree86.org/current/mouse.html]
Keyboard Configuration . www.xfree86.org/current/XKB-Config.html
[http://www.xfree86.org/current/XKB-Config.html]
X.org .
X.org, freedesktop.org [http://www.freedesktop.org]
X.org release notes . freedesktop.org/~xorg/X11R6.7.0/doc/RELNOTES.html
[http://freedesktop.org/~xorg/X11R6.7.0/doc/RELNOTES.html]
Chapter Summary
Most installations of Red Hat Linux begin by booting from the first installation CD. When the system boots from
the CD, it displays a boot: prompt. You can respond to this prompt with different commands, by pressing RETURN
without entering a command, or not at all. In all cases, the system boots Red Hat Linux from the CD. If you are
installing from files on the local hard disk or over a network, give the command linux askmethod in response to
the boot: prompt:
The program that installs Red Hat Linux is named Anaconda. Anaconda identifies the hardware, builds the
filesystems, and installs or upgrades the Red Hat Linux operating system. Anaconda can run in textual or graphical
(default) interactive mode or in batch mode (Kickstart). Anaconda does not write to the hard disk until it displays
the About to Install screen. Until you see this screen, you can press CONTROL-ALT-DEL to abort the installation
without making any changes to the hard disk.
The Disk Druid graphical disk-partitioning program can add, delete, and modify partitions on a hard drive during
installation. The fdisk utility reports on and manipulates hard disk partitions before or after installation.
A dual-boot system is one that can boot one of two operating systems, frequently Windows and Linux. The biggest
problem in setting up a dual-boot system, assuming you want to add Linux to a Windows system, is finding disk
space for Linux.
45
Draft
Exercises
Draft
With the introduction of Fedora Core 2, Red Hat replaced XFree86 with X.org Foundation’s X11R6.7.0 X Window
System. The X.org X server is functionally equivalent to the one distributed by XFree86 because most of the code
is the same. Under X.org, the primary configuration file is named /etc/X11/xorg.conf.
Red Hat Linux uses the GNOME display manager (gdm) to provide a graphical login, even if you are using a KDE
desktop. The gdmsetup utility configures the login presented by gdm by editing the /etc/X11/gdm/gdm.conf file.
Exercises
1. What is the difference between Xinerama and traditional multimonitor X11?
2. What command would you give in response to the boot: prompt to begin an FTP installation?
3. Describe the Anaconda installer.
4. Where on the disk should you put your /boot partition or the root (/) partition if you do not use a /boot partition?
5. If the graphical installer does not work, what three things should you try?
6. When should you specify an ext2 filesystem in place of ext3?
7. Describe Disk Druid.
8. When does a Red Hat Linux system start X by default?
Advanced Exercises
9. If you do not install GRUB on the master boot record of the hard disk, how can you boot Linux?
10. Why would you place /var at the start of the disk?
11. Assume you have four screens, screen0 through screen3, configured. How would you instruct X.org that your
screen layout was a T shape with the first screen at the bottom and the others in a row above it?
46
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 11 - System Administration: Core Concepts
System Administration: Core Concepts
IN THIS CHAPTER
System Administrator and Superuser
Rescue Mode
SELinux (FEDORA)
Red Hat Configuration Tools
rpcinfo: Displays Information About portmap
The xinetd Super Server
TCP Wrappers: Client/Server Security (hosts.allow and
hosts.deny)
Setting Up a chroot Jail
DHCP
nsswitch.conf: Which Service to Look at First
PAM
The job of a system administrator is to keep one or more systems useful and convenient for users. On a Linux system,
the administrator and user may both be you, with you and a single computer only a few feet apart. Or the system
administrator may be halfway around the world, supporting a network of systems, with you simply one of thousands
of users. A system administrator can be one person who works part time taking care of a system and perhaps is
also a user of the system. Or the administrator can be several people, all working full-time to keep many systems
running.
A well-maintained system
•
Runs quickly enough so users do not get too frustrated waiting for the system to respond or complete a task.
•
Has enough storage to accommodate users’ reasonable needs.
•
Provides a working environment appropriate to each user’s abilities and requirements.
•
Is secure from malicious and accidental acts altering its performance or compromising the security of the data
it holds and exchanges with other systems.
•
Is backed up regularly with recently backed-up files readily available to users.
•
Has recent copies of the software that users need to get their jobs done.
•
Is easier to administer than a poorly maintained system.
47
Draft
System Administrator and Superuser
Draft
In addition, a system administrator should be available to help users with all types of system-related problems,
from logging in to obtaining and installing software updates to tracking down and fixing obscure network issues.
Part V of this book breaks system administration into the following 7 chapters:
•
Chapter 11 (this chapter) covers the core concepts of system administration, including Superuser, system operation, the Red Hat configuration tools and other useful utilities, general information about setting up and securing a server (including a section on DHCP), and PAM.
•
Chapter 12 (page 425) covers files, directories, and filesystems from an administrator’s point of view.
•
Chapter 13 (page 453) covers installing software on the system, including how to use Red Hat Network (RHN),
up2date, Apt, yum, BitTorrent, and wget.
•
Chapter 14 (page 485) discusses how to set up local and remote printers that use the CUPS printing system.
•
Chapter 15 (page 507) explains how to rebuild the Linux kernel.
•
Chapter 16 (page 521) covers additional system administrator tasks and tools, including setting up users and
groups, backing up files, scheduling tasks, printing system reports, and general problem solving.
•
Chapter 17 (page 551) goes into detail about how to set up a LAN, including setting up and configuring the
hardware and configuring the software.
Because Linux is configurable and runs on various platforms (Sun SPARC, DEC/Compaq Alpha, Intel x86, AMD,
PowerPC, and more), this chapter cannot discuss every system configuration or every action you will have to take
as a system administrator. This chapter familiarizes you with the concepts you need to understand and the tools
you need to use to maintain a Red Hat Enterprise Linux or Fedora Core system. Where it is not possible to go into
depth about a subject, the chapter provides references to other sources.
This chapter assumes that you are familiar with the following terms. Refer to the glossary (page 955) for definitions.
block (device)
environment
mount(a device)
spawn
daemon
filesystem
process
system console
device
fork
root filesystem
X server
device filename
kernel
runlevel
disk partition
login shell
signal
System Administrator and Superuser
Much of what a system administrator does is work that ordinary users do not have permission to do. When doing
one of these tasks, the system administrator logs in as root (or uses another method; see the list starting on page
372) in order to have systemwide powers that are beyond those of ordinary users: A user with root privileges is
referred to as Superuser. The username is root by default. Superuser has the following powers and more:
•
Some commands, such as those that add new users, partition hard drives, and change system configuration,
can be executed only by root. Superuser can use certain tools, such as sudo, which is covered shortly, to give
specific users permission to perform tasks that are normally reserved for Superuser.
•
Read, write, and execute file access and directory access permissions do not affect root: Superuser can read
from, write to, and execute all files, as well as examine and work in all directories.
48
Draft
•
System Administrator and Superuser
Draft
Some restrictions and safeguards that are built into some commands do not apply to root. For example, root
can change any user’s password without knowing the old password.
When you are running with root (Superuser) privileges, the shell by convention displays a special prompt to remind
you of your status. By default, this prompt is or ends with a pound sign (#).
To lessen the chance that a user other than Superuser will try to use them by mistake, many of the commands that
Superuser runs are kept in the /sbin and /usr/sbin directories, rather than in /bin and /usr/bin. (Many of these
commands can be run by ordinary users.) You can execute these commands by giving their full pathnames on the
command line (for example, /sbin/runlevel). When you log in as root, these directories are in your PATH (page
283) by default.
Caution: Least Privilege
When you are working on the computer, especially when you are working as the system administrator,
perform any task by using the least privilege possible. When you can perform a task logged in as an
ordinary user, do so. When you must be logged in as Superuser, do as much as you can as an ordinary
user, log in or use su so that you have root privileges, do as much of the task that has to be done as
Superuser, and revert to being an ordinary user as soon as you can. Because you are more likely to
make a mistake when you are rushing, this concept becomes more important when you have less time
to apply it.
You can gain or grant Superuser privileges in a number of ways:
1.
When you bring the system up in single-user mode (page 387), you are Superuser.
2.
Once the system is up and running in multiuser mode (page 388), you can log in as root: When you supply
the proper password, you will be Superuser.
3.
You can give an su (substitute user) command while you are logged in as yourself, and, with the proper password, you will have Superuser privileges. For more information, refer to “ su: Gives You Another User’s Privileges” on page 373.
4.
You can use sudo selectively to give users Superuser privileges for a limited amount of time on a per-user and
per-command basis. The sudo utility is controlled by the /etc/sudoers file, which must be set up by root.
Refer to the sudo man page for more information.
Security: root-Owned Setuid Programs Are Extremely Dangerous
Because a root-owned setuid program allows someone who does not know the root password
to have the powers of Superuser, it is a tempting target for a malicious user. Your site should
have as few of these programs as necessary. You can disable setuid programs at the filesystem
level by mounting a filesystem with the nosuid option (page 443).
5.
Any user can create a setuid (Set User ID) file (page 175). Setuid programs run on behalf of the owner of the
file and have all the access privileges that the owner has. While you are running as Superuser, you can change
the permissions of a file owned by root to setuid. When an ordinary user executes a file that is owned by root
and has setuid permissions, the program has full root privileges. In other words, the program can do anything
that root can do and that the program does or allows the user to do. The user’s privileges do not change.
When the program finishes running, all user privileges are back to the way they were before the program was
started. Setuid programs that are owned by root are extremely powerful and are also extremely dangerous to
system security, which is why very few of them are on the system. Examples of setuid programs that are owned
by root include passwd, at, and crontab. The following example shows two ways for Superuser to give a program
setuid privileges:
49
Draft
System Administration Tools
# ls -l my*
–rwxr–xr–x 1
root
–rwxr–xr–x 1
root
# chmod 4755 myprog
# chmod u+s myprog2
# ls -l my*
–rwsr–xr–x 1
root
–rwsr–xr–x 1
root
Draft
other 24152 Apr 29 16:30 myprog
other 24152 Apr 29 16:31 myprog2
other 24152 Apr 29 16:30 myprog
other 24152 Apr 29 16:31 myprog2
The s in the owner execute position of the ls –l output (page 173) indicates that the file has setuid permission.
6.
Some programs ask you for a password (either your password or the root password, depending on the particular command and the configuration of the system) when they start. When you provide the root password,
the program runs with Superuser privileges.
When a program requests the root password when it starts, you stop running as the privileged user when you
quit using the program. This setup helps keep you from remaining logged in as Superuser when you do not
need/intend to be. Refer to “ consolehelper: Runs Programs as Root” on page 375.
Some techniques limit the number of ways to become Superuser. PAM (page 416) controls the who, when, and
how of logging in. The /etc/securetty file controls which terminals (ttys) a user can log in on as root. The
/etc/security/access.conf file adds another dimension to login control (see the file for details).
Security: Do Not Allow root Access over the Internet
Prohibiting root logins using login over a network is the default policy of Red Hat Enterprise Linux
and Fedora and is implemented by the PAM securetty module. The /etc/security/access.conf file
must contain the names of all the users and terminals/workstations that you want a user to be able
to log in on as root. Initially, every line in access.conf is commented out.
If you need root access to a system over a network, use ssh (page 563). As shipped by Red Hat, ssh
does not follow the instructions in securetty or access.conf. Also, in /etc/ssh/sshd_config, Red Hat
sets PermitRootLogin to YES to permit root to log in using ssh (page 578).
System Administration Tools
Many tools can help you be an efficient and thorough system administrator. A few of these tools/utilities are described
in this section, another group of administration utilities is described starting on page 393, and many others are
scattered throughout the system administration chapters.
su: Gives You Another User’s Privileges
The su (substitute user) utility can create a shell or execute a program with the identity and permissions of a specified
user. Follow su on the command line with the name of a user; if you are root or if you know that user’s password,
you take on the identity of that user. When you give an su command without an argument, su defaults to Superuser
so that you take on the identity of root (you have to know the root password).
To be sure that you are using the system’s official version of su (and not one planted on your system by a malicious
user), specify su’s absolute pathname (/bin/su) when you use it. (Of course, if someone has compromised your
system enough to have you run a fake su command, you are in serious trouble anyway, but using an absolute
pathname for su is still a good idea.)
50
Draft
System Administration Tools
Draft
When you give an su command to become Superuser, you spawn a new shell, which displays the # prompt. You
return to your normal status (and your former shell and prompt) by terminating this shell: Press CONTROL-D, or
give an exit command. Giving an su command by itself changes your user and group IDs but makes minimal changes
to your environment. You still have the same PATH you did when you logged in as yourself. When you run a
utility that is normally run by root (those in /sbin and /usr/sbin), you need to specify an absolute pathname for
the utility (as in /sbin/service). When you give the command su – (you can use –l or --login in place of the dash),
you get a root login shell: It is as though you logged in as root. Not only are your user and group IDs those of root,
but your entire environment is that of root. The login shell executes the appropriate start-up scripts before giving
you a prompt, and your PATH is set to what it would be if you had logged in as root, typically including /sbin
and /usr/sbin.
Use the id utility to display the changes in your user and group IDs and in the groups you are associated with:
$ id
uid=500(alex) gid=500(alex) groups=500(alex)
$ su
Password:
# id
uid=0(root) gid=0(root) groups=0(root),1(bin),2(daemon),3(sys), ...
You can use su with the –c option to run a single command with root privileges, returning to your original shell
when the command finishes executing. The following example first shows that a user is not permitted to kill a
process. With the use of su –c and the root password, the user is permitted to kill (page 375) the process. The
quotation marks are necessary because su –c takes its command as a single argument.
$ kill -15 4982
-bash: kill: (4982) - Operation not permitted
$ su -c "kill -15 4982"
Password:
$
Security: Superuser, PATH, and Security
The fewer directories you keep in your PATH when you are root, the less likely you will be to execute
an untrusted program as root. If possible, keep only the default directories, along with /sbin and
/usr/sbin, in root’s PATH. Never include the working directory (as . or : : anywhere in PATH, or : as
the last element of PATH) . For more information, refer to “PATH: Where the Shell Looks for Programs”
on page 283.
consolehelper: Runs Programs as Root
The consolehelper utility can make it easier for someone who is logged in on the system console but not logged in
as root to run system programs that normally can be run only by root. PAM (page 416) authenticates users and
can be set to trust all console users, to require user passwords (not the root password), or to require root passwords
before granting trust. As shipped, Red Hat sets up PAM to require a root password for consolehelper. The concept
behind consolehelper is that you may want to consider as trustworthy anyone who has access to the console. You
can turn this feature on, so, for example, Alex could log in on the console as himself and run halt without knowing
the root password. For more information refer to the discussion of consolehelper on page 391 and to the consolehelper man page.
51
Draft
System Administration Tools
Draft
kill: Sends a Signal to a Process
The kill builtin sends a signal to a process. This signal may or may not terminate (kill) the process, depending on
the signal sent and how the process is designed. Refer to “trap: Catches a Signal” on page 889 for a discussion of
the various signals and how a process receives them. Running kill is not the first method a user or system administrator should try when a process needs to be aborted.
Caution: kill: Method of Last Resort
Because of its inherent dangers, using kill is a method of last resort, especially when you are running
as Superuser. One kill command issued by root can bring the system down without warning.
When you do need to use kill, send the termination signal (kill –TERM or kill –15) first. Only when
that does not work should you attempt to use the kill signal (kill –KILL or kill – 9).
Usually a user can kill a process from another window or by logging in on another terminal. Sometimes you may
have to log in as root (or use su) to kill a process for a user. To kill a process, you need to know the PID of the
process. The ps utility can give you this information once you know the name of the program the user is running
and/or the username of the user. The top utility (page 534) can also be helpful in finding and killing (see top’s k
command) a runaway process.
In the following example, Alex complains that xmms is stuck and that he cannot do anything from the xmms window,
not even close it. A more experienced user could open another window and kill the process, but in this case, you
kill it for Alex. First, use ps with the –u option, followed by the name of the user and the –f (full/wide) option to
view all the processes associated with that user.
$ ps -u alex -f
UID
PID
alex
2841
alex
2896
alex
2903
alex
2908
alex
2974
alex
2977
alex
2980
alex
3002
alex
3008
alex
3009
alex
3010
alex
3012
alex
3014
alex
3019
alex
3024
alex
3028
alex
3029
alex
3030
alex
3040
alex
3041
alex
3069
alex
3074
alex
3101
alex
3121
52
PPID
2840
2841
2896
2903
1
1
1
1
1
1
2908
3009
1
1
1
1
1
3028
3009
3040
1
3041
3074
3014
C
0
0
0
0
STIME TTY
19:39 tty1
19:40 tty1
19:40 tty1
19:40 tty1
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 tty1
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 ?
0 19:41 pts/2
0 19:41 ?
0 19:41 pts/3
0 19:41 ?
0 19:42 pts/3
0 19:42 pts/3
33 19:49 ?
TIME
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:03
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:01
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:00
00:00:03
00:00:00
00:00:41
CMD
-bash
/bin/sh /usr/X11R6/bin/startx
xinit /etc/X11/xinit/xinitrc -/bin/bash /usr/bin/startkde
kdeinit:dcopserver --nosid
kdeinit:klauncher
kdeinit:kded
/usr/bin/artsd -F 10 -S 4096 -s 60
kdeinit:knotify
kdeinit:Running...
ksmserver --restore
kdeinit:kwin
kdeinit:kdesktop
kdeinit:kicker
kdeinit:klipper -icon klipper
kdeinit:kwrited
alarmd
/bin/cat
kdeinit:konsole -icon konsole
/bin/bash
kdeinit:kcontrol -caption Control
xmms
(dns helper)
kscience.kss -window-id 4194310
Draft
Rescue Mode
Draft
This list is fairly short, and the process running xmms is easy to find. Another way to go about searching is to use
ps to produce a long list of all the processes and use grep to find all the processes running xmms.
$ ps -ef | grep xmms
alex
3074
1 1 10:22 tty1
alex
3157 2573 0 10:25 pts/1
00:00:01 xmms
00:00:00 grep xmms
Many people may be running xmms, and you may need to look in the left column to find the name of the user so
you can kill the right process. You can combine the two commands as ps –u alex –f | grep xmms.
Now that you know the PID of Alex’s process running xmms is 3074, you can use kill to terminate it. The safest
way to do this is to log in as Alex (perhaps allow him to log in for you or su to alex [su alex] if you are logged in
as root) and give the command
$ kill –TERM 3074
Only if this command fails should you send the kill signal as
$ kill –KILL 3074
The –KILL option instructs kill to send a SIGKILL signal, which the process cannot ignore. You can give the same
command while you are logged in as root, but a typing mistake can have much more far-reaching consequences
than when you make the mistake while you are logged in as an ordinary user. A user can kill only her or his own
processes, whereas Superuser can kill any process, including system processes.
As a compromise between speed and safety, you can combine the su and kill utilities by using the –c option to su.
The following command runs the part of the command line following the –c with the identity of Alex:
# su alex -c "kill -TERM 3074"
Two useful utilities related to kill are killall and pidof. The first is very similar to kill but uses a command name in
place of a PID number. To kill all your processes that are running xmms or vi, you can give the command
$ killall xmms vi
When root gives this command, all processes that are running xmms or vi on the system are killed.
The pidof utility displays the PID number of each of the processes running the command you specify. Because this
utility resides in /sbin, you must give the absolute pathname if you are not running as root:
$ /sbin/pidof httpd
567 566 565 564 563 562 561 560 553
Refer to the man pages for each of these utilities for more information, including lists of available options.
Rescue Mode
Rescue mode is an environment you can use to fix a system that does not boot normally. To bring a system up in
rescue mode, boot the system from the first installation CD or the rescue CD. From the boot: prompt, give the
53
Draft
Avoiding a Trojan Horse
Draft
command linux rescue from the first installation CD or press RETURN without entering a command from the
rescue CD. The system comes up in rescue mode.
In rescue mode, you can change or replace configuration files, check and repair partitions using fsck, rewrite boot
information, and more. The rescue setup first asks if you want to set up the network interface. You may want to
copy files from other systems on the LAN or download files from the Internet. When you choose to set up the
network interface, you need to choose whether to use DHCP to automatically configure the network connection
or supply the IP address and netmask of the interface, as well as the IP addresses of the gateway and up to three
DNS addresses.
If the rescue setup finds an existing Linux installation, you can choose to mount it under /mnt/sysimage, optionally
in readonly mode. With the existing installation mounted, once the system displays a shell prompt (similar to
2.05b#), you can give the command chroot /mnt/sysimage to mount the existing installation as it would be if
you booted normally, with the existing installation’s root mounted at / (root). See page 406 for more information
on chroot. If you choose not to mount the existing installation, you are running a rescue system with standard tools
mounted in standard locations (/bin, /usr/bin and so on). Partitions from your local installation are available for
fixing or mounting. When you exit from the rescue shell, the system reboots. Remove the CD if you want to boot
from the hard drive.
Avoiding a Trojan Horse
A Trojan horse is a program that does something destructive or disruptive to your system while appearing to be
benign. As an example, you could store the following script in an executable file named mkfs:
while true
do
echo 'Good Morning Mr. Jones. How are you? Ha Ha Ha.' > /dev/console
done
If you are running as Superuser when you run this command, it would continuously write a message to the console.
If the programmer were malicious, it could do worse. The only thing missing in this plot is access permissions.
A malicious user could implement this Trojan horse by changing Superuser’s PATH variable to include a publicly
writable directory at the start of the PATH string. (The catch is that you need to be able to write to /etc/profile—where the PATH variable is set for root—and only root can do that.) Then you would need to put the bogus
mkfs program file in that directory. Because the fraudulent version appears in a directory mentioned earlier than
the real one in PATH, the shell runs it. The next time Superuser tries to run mkfs, the fraudulent version would
run.
Trojan horses that wait for and take advantage of the misspellings that most people make are one of the most insidious types. For example, you might type sl instead of ls. Because you do not regularly execute a utility named sl
and you may not remember typing the command sl, it is more difficult to track down this type of Trojan horse
than one that takes the name of a utility you are familiar with.
A good way to prevent executing a Trojan horse is to make sure that your PATH variable does not contain a single
colon (:) at the beginning or end of the PATH string or a period (.) or double colon (::) anywhere in the PATH
string. A common way to check for a Trojan horse is to examine the filesystem periodically for files with setuid
(refer to item 5 on page 372). The following command lists these files:
# find / -perm -4000 -exec ls -lh {} \; 2> /dev/null
-rwsr-xr-x 1 root root 54K Apr 5 16:17 /usr/bin/lppasswd
-rwsr-xr-x 1 root root 111K Feb 17 10:26 /usr/bin/crontab
54
Draft
-rwsr-xr-x
-rws--x--x
-rws--x--x
-rwsr-xr-x
-rwsr-xr-x
-r-s--x--x
-rwsr-xr-x
---s--x--x
-rwsr-xr-x
-rwsr-xr-x
-rwsrwxr-x
-rwsr-xr-x
-rwsr-xr-x
-rws--x--x
...
Getting Help
1
1
1
1
1
1
1
1
1
1
1
1
1
1
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
Draft
7.0K Apr 13 08:26 /usr/bin/kpac_dhcp_helper
18K Mar 23 12:05 /usr/bin/chfn
6.8K Mar 23 12:05 /usr/bin/newgrp
39K Apr 15 12:26 /usr/bin/at
42K Mar 30 13:45 /usr/bin/gpasswd
91K Feb 15 10:04 /usr/bin/passwd
17K Feb 17 09:12 /usr/bin/rlogin
96K Apr 1 08:19 /usr/bin/sudo
6.8K Apr 13 08:26 /usr/bin/kgrantpty
18K Feb 17 09:12 /usr/bin/rcp
364K Mar 16 13:32 /usr/bin/tvtime
40K Mar 30 13:45 /usr/bin/chage
9.6K Feb 17 09:12 /usr/bin/rsh
19K Mar 23 12:05 /usr/bin/chsh
This command uses find to locate all the files that have their setuid bits set (mode 4000). The hyphen preceding
the mode causes find to report on any file that has this bit set, regardless of how the other bits are set. The output
sent to standard error is redirected to /dev/null so that it does not clutter the screen.
You can also set up a program, such as AIDE (Advanced Intrusion Detection Environment), that will take a snapshot
of your system and check it periodically as you specify. See www.cs.tut.fi/~rammer/aide.html
[http://www.cs.tut.fi/~rammer/aide.html] for more information.
Getting Help
Your distribution comes with extensive documentation (page 94). Red Hat maintains a page that points toward
many useful support documents: www.redhat.com/apps/support [http://www.redhat.com/apps/support]. You
can also find help on the System Administrator’s Guild site (www.sage.org [http://www.sage.org]). In addition,
the Internet is a rich source of information on managing a Linux system; refer to Appendix B (page 913) and to the
author’s home page (www.sobell.com [http://www.sobell.com]) for pointers to useful sites.
You do not need to act as a Red Hat system administrator in isolation; a large community of Linux/Red Hat experts
is willing to assist you in getting the most out of your system, although you will get better help if you have already
tried to solve a problem yourself by reading the available documentation. If you are unable to solve a problem
through the documentation, a well thought-out question to the appropriate newsgroup, such as comp.os.linux.misc,
or mailing list can often provide useful information. Be sure you describe the problem and identify your system
carefully. Include information about your version of Red Hat Enterprise Linux or Fedora Core and any software
packages and hardware that you think relate to the problem. The newsgroup comp.os.linux.answers contains
postings of solutions to common problems and periodic postings of the most up-to-date versions of FAQ and
HOWTO documents. See www.catb.org/~esr/faqs/smart-questions.html
[http://www.catb.org/~esr/faqs/smart-questions.html] for a good paper by Eric S. Raymond and Rick Moen titled
“How to Ask Questions the Smart Way.”
SELinux (FEDORA)
Traditional Linux security, called Discretionary Access Control (DAC), is based on users and groups. A process run
by a user has access to anything the user has access to, making fine-grained access control difficult. Fine-grained
access control is particularly important on servers, which often require root privileges to run.
SELinux (Security Enhanced Linux) implements Mandatory Access Control (MAC), wherein you can define a security
policy that controls all objects, such as files and devices, and all subjects, such as processes and users. Using SELinux,
you can grant a program only the permissions it needs.
55
Draft
System Operation
Draft
Under SELinux, users are assigned roles that determine the primary function of the user. SELinux roles are conceptually similar to groups, except that under SELinux, a user must actively switch between permitted roles. You can
use system-config-users (page 521) to assign roles to users.
SELinux can be in one of three states:
•
Enforcing/Active The SELinux security policy is enforced. You will not be able to do anything not permitted
by the security policy.
•
Permissive/Warn SELinux sends warning messages to a log but does not enforce. You can use the log to build
a security policy that matches your requirements.
•
Disabled SELinux does not enforce because no policy is loaded.
Running SELinux in permissive or enforcing mode degrades system performance between five and ten percent.
SELinux is usually of no benefit on a single-user system. You may want to consider SELinux for a server that connects
to the Internet. If you are unsure whether to use SELinux or not, selecting permissive mode allows you to change
to disabled or enforcing mode easily at a later date.
To disable SELinux, change the SELINUX= line in /etc/sysconfig/selinux to SELINUX=disabled. The following
listing shows SELinux disabled:
$ cat /etc/sysconfig/selinux
# This file controls the state of SELinux on the system.
# SELINUX= can take one of these three values:
#
enforcinfg - SELinux security policy is enforced.
#
permissive - SELinux prints warnings instead of enforcing.
#
disabled - No SELinux policy is loaded.
SELINUX=disabled
If you are going to use SELinux in the future but not now, turn it on when you install Linux, and run it in permissive
mode. Permissive mode writes the required extended information to inodes, but it does not stop you from doing
anything on the system. Because SELinux sets extended attributes on files, it is an involved process to turn on
SELinux after you install Linux. If you decide you are never going to use SELinux, you can disable it. For more information, see the Fedora SELinux FAQ at people.redhat.com/kwade/fedora-docs/selinux-faq-en.
system-config-securitylevel . Under Fedora Core 2 and later, you can use system-config-securitylevel to set the
state of SELinux. From the Security Level Configuration window, click the SELinux tab and choose Active, Warm,
or Disabled from the combo box. Click OK. See page 741 for information on the Firewall Options tab.
System Operation
This section covers the basics of how the system functions and how you can make intelligent decisions as a system
administrator. This section does not cover every aspect of system administration in the depth necessary to set up
or modify all system functions. It provides a guide to bringing a system up and keeping it running from day to day.
Booting the System
Booting a system is the process of reading the Linux kernel (page 979) into system memory and starting it running.
Refer to “Boot Loader” on page 514 for more information on the initial steps of bringing a system up.
56
Draft
Init Scripts: Start and Stop System
Services
Draft
As the last step of the boot procedure, Linux runs the init program as PID number 1. The init program is the first
genuine process to run after booting and is the parent of all system processes. (That is why when you run as root
and kill process 1, the system dies.)
initdefault . The initdefault entry in the /etc/inittab file (page 429) tells init what runlevel to bring the system
to (Table 11-1). Set initdefault to 3 to cause the system to present a text login message when it boots; set initdefault
to 5 to present a graphical login screen.
Table 11-1. Runlevels
Number
Name
0
Halt
1 (not S or s)
Single user
2
Undefined
3
Multiuser
4
Undefined
5
Multiuser with X
6
Reboot
Login
Network
Filesystems
Text
Down
Not mounted
Text
Up
Mounted
Graphics
Up
Mounted
Init Scripts: Start and Stop System Services
The first script that init runs is /etc/rc.d/rc.sysinit, which runs initlog to set up command logging for the duration
of the boot process. The rc.sysinit script also performs basic system configuration, including setting the system
clock, hostname, and keyboard mapping; setting up swap partitions; checking the filesystems for errors; and
turning on quota management (page 545).
Next, the /etc/rc.d/rc init script runs the scripts that handle the services that need to be started when you first
bring the system up and that need to be started or stopped when the system goes from single-user to multiuser
mode and back down again.
Tip: List the Kernel Boot Messages
To save a list of kernel boot messages, give the following command immediately after booting the
system and logging in:
$ dmesg > dmesg.boot
This command saves the kernel messages in the dmesg.boot file. This list can be educational. It can
also be useful when you are having a problem with the boot process. For more information, refer to
“dmesg: Display Kernel Messages” on page 517.
The init (initialization) scripts, also called rc (run command) scripts, are shell scripts located in the /etc/rc.d/init.d
directory and run via symbolic links in the /etc/rc.d/rcn.d directory, where n is the runlevel the system is entering.
The /etc/rc.d/rcn.d directories contain scripts whose names begin with K (K15httpd, K72autofs, K30sendmail,
and so on) and scripts whose names begin with S (S05kudzu, S10network, S13portmap, and so on). When entering a new run-level, each of the K (kill) scripts is executed with an argument of stop, and then each of the S
(start) scripts is executed with an argument of start. Each of the K files is run in numerical order. The S files are
run similarly. This setup allows the person who sets up these files to control which services are stopped and which
are started and in what order, whenever the system enters a given runlevel. Using scripts with start and stop argu-
57
Draft
Init Scripts: Start and Stop System
Services
Draft
ments is flexible because it allows one script to both start and kill a process, depending on the argument it is called
with.
To customize system initialization, you can add shell scripts to the /etc/rc.d/init.d directory and add links to the
files in init.d from the /etc/rc.d/rcn.d directories. The following example shows several links to the cups script.
These links are called to run the cups init script to start or stop the cupsd daemon at various runlevels:
$ pwd
/etc/rc.d
$ ls -l */*cups
-rwxr-xr-x 1 root
lrwxrwxrwx 1 root
lrwxrwxrwx 1 root
lrwxrwxrwx 1 root
lrwxrwxrwx 1 root
lrwxrwxrwx 1 root
lrwxrwxrwx 1 root
lrwxrwxrwx 1 root
root 2312 Apr 5 16:17 init.d/cups
root
14 Apr 28 21:40 rc0.d/K10cups -> ../init.d/cups
root
14 Apr 28 21:40 rc1.d/K10cups -> ../init.d/cups
root
14 May 2 22:44 rc2.d/S55cups -> ../init.d/cups
root
14 May 2 22:44 rc3.d/S55cups -> ../init.d/cups
root
14 May 2 22:44 rc4.d/S55cups -> ../init.d/cups
root
14 May 2 22:44 rc5.d/S55cups -> ../init.d/cups
root
14 Apr 28 21:40 rc6.d/K10cups -> ../init.d/cups
Each link in /etc/rc.d/rcn.d should point to a file in /etc/rc.d/init.d. The file /etc/rc.d/rc1.d/K10cups is a link
to the file named cups in /etc/rc.d/init.d. (The numbers that are part of the filenames of the links in the
/etc/rc.d/rcn.d directories may change from one OS release to the next, but the scripts in /etc/rc.d/init.d always
have the same names.) The names of files in the /etc/rc.d/init.d directory are functional. When you want to turn
NFS services on or off, use the nfs script. When you want to turn basic network services on or off, run the network
script. The cups script controls the printer daemon. Each of the scripts takes an argument of stop or start, depending
on what you want to do. Some of the scripts also take arguments of restart, reload, status, and some others. Run
a script without an argument to display a usage message telling you which arguments it accepts.
Following are three examples of calls to init scripts. You may find it easier to use service (next) in place of the
pathnames in these calls:
# /etc/rc.d/init.d/nfs stop
# /etc/rc.d/init.d/network start
# /etc/rc.d/init.d/network restart
The first example stops all processes related to serving filesystems over the network, using NFS. The second example
starts all processes related to basic network services. The third example stops and then starts these same processes.
Tip: Maintain the Links in the /etc/rc*.d Hierarchy
Refer to page 385 for information about using chkconfig to maintain the symbolic links in the /etc/rc*.d
hierarchy.
The /etc/rc.d/rc.local file is executed after the other init scripts. Put commands that customize the system in
rc.local. You can add any commands you like to rc.local; however, it is best to run them in the background so
that if they hang, they do not stop the boot process.
service: Configures Services I
Red Hat provides service, a handy script that reports on or changes the status of any of the system services in
/etc/rc.d/init.d. In place of the commands toward the end of the previous section, you can give the following
commands from any directory:
58
Draft
Init Scripts: Start and Stop System
Services
Draft
# /sbin/service nfs stop
# /sbin/service network start
# /sbin/service network restart
The command /sbin/service --status–all displays the status of all system services.
system-config-services: Configures Services II
The system-config-services (FEDORA) and redhat-config-services (RHEL) graphical utilities (Figure 11-1) each have
two distinct functions: They turn system services on and off, and they control which services are stopped and
started when the system enters and leaves runlevels 3, 4 (not used), and 5. The line below the toolbar gives you
two pieces of information: the current runlevel of the system and the runlevel that you will edit. The [system|redhat]-config-services utilities control two groups of services: independent services listed in /etc/rc.d/init.d and those
controlled by xinetd (page 403) and listed in /etc/xinetd.d (or as specified in /etc/xinetd.conf). Scroll to and
highlight the service you are interested in; a short description appears in the text box at the upper-right of the
window. When the description includes xinetd is required for this service, the highlighted service is dependent
on xinetd; otherwise, it is an independent service. The lower text box has status information.
Figure 11-1. The Service Configuration window
The xinetd superserver is an independent service. You must make sure that xinetd is turned on for the runlevels
at which you want to run services that are dependent on it. It is usually on for runlevels 3, 4, and 5.
The [system|redhat]-config-services utilities do the following:
59
Draft
Init Scripts: Start and Stop System
Services
Draft
•
Turn independent services on and off. When you highlight an independent service, you can click the toolbar
or make a selection from Actions in the menubar to stop, start, or restart (stop and start) the service. The system
turns on/off the service immediately; the change does not affect whether the service will run the next time you
bring up the system, enter another runlevel, or reenter the current runlevel.
•
Turn xinetd-controlled services on and off. When you highlight a service that is controlled by xinetd, the
Start, Stop, and Restart buttons on the toolbar are grayed out. Until you click the box to the left of the service,
the Save and Cancel buttons are grayed out too. The only things you can do are to turn this service off or on.
Click the small box, and click the Save button to change the state of the service. You are changing the yes/no
parameter of the disable line discussed on page 404. When you click Save, the system restarts xinetd with the
service status change you requested. This change affects all runlevels and will stay in effect through changes in
runlevels and reboots unless you change it again.
•
Controls future execution of independent services in runlevels 3, 4, and 5. Select the runlevel you want
to affect, using the Edit Runlevel selection from the menubar. Highlight an independent service, and click the
box next to it to indicate whether you want the service on or off at the specified runlevel. Click the Save button.
When you enter that runlevel in the future, the service will be on or off as you specified. The current state of
the service is not changed. See the first item in this list when you want to change the state of the service.
chkconfig: Configures Services III
The chkconfig character-based utility duplicates much of what [system|redhat]-config-services does: It makes it
easier for a system administrator to maintain the /etc/rc.d directory hierarchy. This utility can add, remove, list
start-up information, and check the state of system services; it changes the configuration only, it does not change
the current state of any service. To see a list of all services, give the following command:
# chkconfig --list
dhcrelay
0:off
dhcpd
0:off
syslog
0:off
atd
0:off
gpm
0:off
...
xinetd based services:
chargen-udp:
rsync: off
chargen:
daytime-udp:
daytime:
...
1:off
1:off
1:off
1:off
1:off
2:off
2:off
2:on
2:off
2:on
3:off
3:off
3:on
3:on
3:on
4:off
4:off
4:on
4:on
4:on
5:off
5:off
5:on
5:on
5:on
6:off
6:off
6:off
6:off
6:off
off
off
off
off
All the services that run their own daemons are listed, one to a line, followed by their configured state for each
runlevel. Following that list, chkconfig displays each of the xinetd-based services and their current status. You can
check on how a specific daemon is configured by adding its name to the previous command:
# chkconfig --list sshd
sshd
0:off
1:off
2:on
3:on
4:on
5:on
6:off
In the next example, chkconfig configures the /etc/rc.d directory hierarchy so that sshd will be off in runlevels 2,
3, 4, and 5 and then confirms the change:
60
Draft
Emergency Mode
# chkconfig --level 2345 sshd off
# chkconfig --list sshd
sshd
0:off
1:off
2:off
3:off
4:off
Draft
5:off
6:off
For convenience, you can omit the –level 2345. When you specify an init script and on or off, chkconfig defaults
to runlevels 2, 3, 4, and 5. The following command is equivalent to the first of the preceding commands:
# chkconfig sshd off
Following, both ps and service confirm that even though chkconfig set things up so that sshd would be off in all
runlevels, it is still running. The chkconfig utility did not shut down sshd. In the following example, the second
command line shows that when you give a service command followed by the name of an init script, you get the
usage message from the script:
# ps -ef | grep sshd
root
697
1 0 Oct01 ?
00:00:00 /usr/sbin/sshd
root
17185 21650 0 15:15 pts/4
00:00:00 grep sshd
# /sbin/service sshd
Usage: /etc/init.d/sshd {start|stop|restart|reload|condrestart|status}
# /sbin/service sshd status
sshd (pid 697) is running...
When you reboot the system, sshd will not start, but you can stop it more easily using service:
# /sbin/service sshd stop
Stopping sshd:
# ps -ef | grep sshd
root
17209 21650 0 15:16 pts/4
# /sbin/service sshd status
sshd is stopped
[ OK ]
00:00:00 grep sshd
Emergency Mode
When you use lilo or grub to boot, you can set the emergency kernel flag that gets passed to init and causes init to
run sulogin (single-user login). Also see “Rescue Mode” on page 377 for booting the system when it will not boot
normally. To bring the system up in emergency mode when you are booting using lilo, enter linux emergency,
and press RETURN in response to the Boot: prompt.
When you are using grub, you need to press e as soon as the grub screen appears so you can edit the boot commands.
If the grub screen shows multiple bootable kernels, move the highlight to the one you want to boot before pressing
e. After pressing e, use the up/down ARROW keys to move the highlight to the line that starts with kernel. Press
e to bring up the grub editor; the cursor will be at the end of the kernel command line. Type SPACE emergency,
press RETURN to return to the previous grub screen, and then press b to boot the system.
With either grub or lilo, when the boot process hands control to init, init runs sulogin, which displays the following
message:
Give root password for system maintenance
(or type Control-D for normal startup):
61
Draft
Single-User Mode
Draft
Enter the root password, and you will see the root shell prompt. At this point the only mounted filesystem is root
(/), and possibly /proc (page 434); root is mounted as a readonly filesystem (despite what the mount command
says). If /proc is not mounted, give the command
# mount /proc
The ps, top, fsck, and other programs require the /proc filesystem. The –L and –U mount options, as well as the
LABEL= and UUID= specifications in the /etc/fstab file (page 445), also require /proc.
Now you can safely use fsck to check the integrity of the root filesystem:
# fsck /
With most types of filesystems, you can use the –f option with fsck to force a check when the filesystem appears
to be clean.
If you need to write any files to the root filesystem, you need to remount it in read-write mode:
# mount -n -o remount,rw /
The –n argument keeps mount from trying to write to the /etc/mtab file first. If it were to attempt to write to a
readonly filesystem, mount would fail, and the filesystem would remain mounted in readonly mode. If you remount
the root filesystem as shown, you may need to remount it in readonly mode before you can continue booting the
system:
# mount -o remount,rw /
When you are ready to bring the system up to multiuser mode, exit from the shell. Linux will either continue
booting or reboot to the default runlevel (as specified by initdefault in /etc/inittab [page 381]).
Single-User Mode
When the system is in single-user mode, only the system console is enabled. You can run programs from the console
in single-user mode as you would from any terminal in multiuser mode. The differences are that not all filesystems
may be mounted, so you may not be able to access some files, and few of the system daemons will be running. The
root filesystem is always mounted in single-user mode. The scripts in /etc/rc.d/rc1.d are run as part of single-user
initialization.
With the system in single-user mode, you can perform system maintenance that requires filesystems unmounted
or just a quiet system—no one except you using it, so that no user programs interfere with disk maintenance and
backup programs. The classical UNIX term for this state is quiescent. See “Backing Up Files” on page 524 for a
discussion of one of the most important and often neglected areas of system administration.
Going Multiuser
After you have determined that all is well with all filesystems, you can bring the operating system up to multiuser
mode. When you exit from the single-user shell, init brings the system to the default runlevel—usually 3 or 5 (page
381). Or, the following command in response to the Superuser prompt brings the system to (textual) multiuser
mode (use 5 to go to graphics [multiuser] mode):
62
Draft
Multiuser/Graphics Mode
Draft
# telinit 3
The telinit utility tells init what runlevel to enter. The telinit executable is a symbolic link to the init executable,
but, by convention, running telinit is preferred to running init directly.
When it goes from single- to (textual) multiuser mode, the system executes the K (kill or stop) scripts and then the
S (start) scripts in /etc/rc.d/rc3.d. For more information, refer to “Init Scripts: Start and Stop System Services” on
page 381. Use chkconfig (page 385) to stop one of these scripts from running.
Runlevel 2 is referred to as multiuser mode, and runlevel 3 is extended multiuser mode. But because runlevel 2 is
rarely used, this chapter uses the term multiuser to refer to runlevel 3. Runlevel 4 is not used, and runlevel 5 is
graphics or X11 mode.
Multiuser/Graphics Mode
Multiuser/graphics mode is the normal state for a Linux system. All appropriate filesystems are mounted, and users
can log in from all connected terminals, dial-in lines, and network connections. All support services and daemons
are enabled and running. Once the system is in multiuser/graphics mode, you will see a login screen or prompt.
Most systems are set up to boot directly to multiuser/graphics mode without stopping at single-user mode.
Logging In
With a text login, the system uses init, mingetty, and login to allow a user to log in; login uses PAM modules (page
416) to authenticate users. Once the system is in multiuser mode, init is responsible for spawning a mingetty process
on each of the lines that a user can log in on.
When you enter your username, mingetty establishes the characteristics of your terminal and then overlays itself
with a login process and passes to the login process whatever you entered in response to the login: prompt. The
login program consults the /etc/passwd file to see whether a username matches the username you entered. This
program then consults the /etc/shadow file to see whether a password is associated with the username. If there
is, login prompts you for a password; if not, it continues without requiring a password. When your username requires
a password, login verifies the password you enter by checking the /etc/shadow file again. If either your username
or password is not correct, login displays Login incorrect and prompts you to log in again.
All passwords in the /etc/shadow file are encrypted or hashed using MD5 (page 982). It is not feasible to recover
an encrypted password. When you log in, the login process encrypts/hashes the password you type at the prompt
and compares it to the encrypted/hashed password in /etc/shadow. If it matches, you are authenticated.
With a graphical login, the init process spawns gdm (the GNOME display manager) on the first free virtual terminal,
providing features similar to mingetty and login. The gdm utility starts an X server and presents a login window.
The gdm display manager then uses PAM to authenticate the user and runs the scripts in the
/etc/X11/gdm/PreSession directory. The scripts inspect the user’s ~/.dmrc file, which stores the user’s default
session and language, and launch the user’s session. GNOME and KDE desktop environments store the state of
the last login and attempt to recreate it when the user logs back in.
With NIS, login compares your username and password with the information in the appropriate naming service
instead of (or in addition to) the passwd and shadow files. If the system is configured to use both of these methods
(/etc/passwd and NIS), it checks the /etc/nsswitch.conf file (page 413) to see which order to consult them in.
PAM (page 416), the Pluggable Authentication Module facility, allows you greater control over user logins than
the /etc/passwd and /etc/shadow files do. Using PAM, you can specify multiple levels of authentication, mutually
exclusive authentication methods, or parallel methods that are each in themselves sufficient to grant access to the
system. For example, you can have a different authentication method for console logins and for TELNET logins.
63
Draft
Running a Program and Logging Out
Draft
And you can require that modem users authenticate themselves via two or more methods (such as a smartcard or
badge reader and a password). PAM modules also provide security technology vendors with a convenient way to
interface their hardware or software products with a system.
When the username and password are correct, login or the scripts in PreSession consult the appropriate services
to initialize your user and group IDs, establish your home directory, and determine which shell or desktop manager
you will be working with.
The login utility/PreSession scripts assign values to the HOME, PATH, LOGNAME, SHELL, TERM, and MAIL
variables. They look in the /etc/group file (page 428) to identify the groups the user belongs to. When login has
finished its work, it overlays itself with the login shell, which inherits the variables login has set. In a graphical environment, the PreSession scripts start the desktop manager.
During a textual login, the login shell assigns values to additional shell variables and executes the commands in
the system start-up shell script(s) /etc/profile and /etc/bashrc. Some systems have additional system start-up shell
scripts. Exactly what these scripts do is system dependent, but they usually display the contents of the/etc/motd
(message of the day) and /etc/issue files, let you know that you have mail, and set umask (page 398), the file-creation
mask.
After executing the system start-up commands, the shell executes the commands from the personal start-up shell
scripts in your home directory. For a list of these scripts, refer to page 272. Because the shell executes these scripts
after the system script, a sophisticated user can override any variables or conventions that were established by the
system, whereas a new user can remain uninvolved in these complications.
Running a Program and Logging Out
When you see a shell prompt, you can execute a program or exit from the shell. If you exit from a shell, the process
running the shell dies and the parent process wakes up. When the shell is a child of another shell, the parent shell
wakes up and displays a prompt. Exiting from a login shell causes the operating system to send init a signal that
one of its children has died. Upon receiving this signal, init takes action based on the contents of the /etc/inittab
file. In the case of a process controlling a line for a terminal, init informs mingetty that the line is free for another
user.
When you are at runlevel 5 and exit from a GUI, the GNOME display manager, gdm, takes care of initiating a new
login display.
Bringing the System Down
The shutdown and halt utilities perform all the tasks needed to bring the system down safely. These utilities can
restart the system, prepare the system to be turned off, put the system in single-user mode, and on some hardware,
power down the system. The poweroff and reboot utilities are linked to halt. When you call halt when the system
is not shutting down (runlevel 0) or rebooting (runlevel 6), halt calls shutdown. (When you are running as other
than Superuser, the link is through console-helper [page 375].)
You must tell shutdown when you would like to bring the system down. This can be expressed as an absolute time
of day, as in 19:15, which causes the shutdown to occur at 7:15 P.M. Alternatively, you can give the number of
minutes from the present time, as in +15, which means fifteen minutes from now. To bring the system down immediately (recommended for emergency shutdowns only or when you are the only user logged in), you can give
the argument +0, or its synonym: now. For shutdown times longer than 5 minutes, all nonroot logins are disabled
for the last 5 minutes before shutdown.
Calling shutdown with the –r option causes the system to reboot (same as the reboot command except that reboot
implies now). Adding the –f option forces a fast reboot, where filesystem checking is disabled (see the shutdown
man page for details). Using –h instead of –r forces the system to halt (same as the halt command except that halt
64
Draft
Bringing the System Down
Draft
implies now). A message appears once the system has been safely halted: System halted, although most ATX
systems turn off automatically after shutdown, so you will not see the message
Because Linux is a multiuser system, shutdown warns all users before taking any action. This gives users a chance
to prepare for the shutdown, perhaps by writing out editor files or exiting from networking applications. You can
replace the default shutdown message with one of your own by following the time specification on the command
line with a message:
# /sbin/shutdown -h 09:30 Going down 9:30 to install disk, up by 10am.
Caution: Do Not Turn the Power Off Before Bringing the System
Down
Avoid rebooting your Linux system without first bringing it down as described here. Linux, like UNIX
system, speeds up disk access by keeping an in-memory collection of disk buffers that are written to
the disk periodically or when system use is momentarily low. When you turn off or reset the computer
without writing the contents of these disk buffers to the disk, you lose any information in the buffers.
Running shutdown forces these buffers to be written. You can force the buffers to be written at any
time by issuing a sync command. However, sync does not unmount filesystems, nor does it bring the
system down.
CONTROL-ALT-DEL: Reboots the System
By default, the /etc/inittab file on an Intel-based computer has the entry
ca::ctrlaltdel:/sbin/shutdown -t3 -r now
This entry allows any user1 to reboot the computer safely by pressing the key sequence CONTROL-ALT-DEL (also
referred to as the three-finger salute, or the Vulcan death grip) from a textual login on the console. (Although it is
not recommended, you can press CONTROL-ALT-BACKSPACE from a graphical session to kill the X server; once
the X server is killed, you can use CONTROL-ALT-DEL.) Because of its hooks into the keyboard driver, this key
sequence sends a SIGINT signal to the init process, which in response runs shutdown. Because it runs as root, init
causes shutdown to run as root also, even if the key sequence is initiated by an ordinary user. You can disable
CONTROL-ALT-DEL by deleting the preceding line from /etc/inittab (or putting a # at the left end of the line)
and then sending init a HUP signal (kill –HUP 1), which causes it to reread the /etc/inittab file.
consolehelper: Allows an Ordinary User to Run a Privileged Command
As shown following, there are two executable halt files:
$ file /sbin/halt /usr/bin/halt
/sbin/halt:
ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV)...
/usr/bin/halt: symbolic link to 'consolehelper'
The one in /sbin runs the halt utility, whereas the one in /usr/bin is a link to console-helper. In root’s PATH
variable, /sbin normally precedes /usr/bin so that when someone running as root gives a halt command, the shell
1
When you include the –a option in the shutdown command in /etc/inittab and the /etc/shutdown.allow file exists, one of the users whose
name appears in this file (or root) must be logged in on one of the virtual consoles in order for a non root user to run shutdown from a virtual
console.
65
Draft
Crash
Draft
executes /sbin/halt (the halt utility). Normally, /sbin does not appear in an ordinary user’s PATH; when an ordinary
user gives a halt command, the shell follows the link from /usr/bin/halt and executes /usr/bin/consolehelper.
What consolehelper does depends on how PAM is set up (see /etc/pam.d/halt for the modules it calls and
/usr/share/doc/pam-*/txts/* for descriptions of the modules). Refer to “PAM” on page 416 for more information.
As shipped by Red Hat, consolehelper prompts for the root password; when you supply it, consolehelper proceeds
to shut the system down.
Going Single-user
Because going from multiuser to single-user mode can affect other users, you must be Superuser to make this
change. Make sure that you give other users enough warning before going to single-user mode; otherwise, they
may lose whatever they were working on.
Following is a method of manually bringing the system down to single-user mode— the point where it is safe to
turn the power off. You must be running as Superuser to perform these tasks.
1.
Use wall (write all) to warn everyone who is using the system to log out.
2.
If you are sharing files via NFS, use exportfs –ua to disable network access to the shared filesystems. (Use
exportfs without an argument to see what filesystems are being shared.)
3.
Use umount –a to unmount all mounted devices. (Use mount without an argument to see what devices are
mounted.)
4.
Give the command telinit 1 to bring the system down to single-user mode.
Turning the Power Off
Once the system is in single-user mode, shutting it down is quite straightforward. Give the command telinit 0
(preferred) or halt to bring the system down. You can build a kernel with apm so it turns the machine off at the
appropriate time. If your machine is not set up this way, turn the power off when the appropriate prompt appears
or when the system starts rebooting.
Crash
A crash occurs when the system suddenly stops/fails when you do not intend it to. A crash may result from software
or hardware problems or a loss of power. As a running system loses power, nothing is regular or predictable. In a
fraction of a second, some components are supplied with enough voltage; others are not. Buffers are not flushed,
corrupt data may be written to the hard disk, and so on. IDE drives do not behave as predictably as SCSI drives
under these circumstances. After a crash, you must bring the operating system up carefully to minimize possible
damage to the filesystems. Frequently there will be little or no damage.
Repairing a Filesystem
Although the filesystems are checked automatically during the boot process if needed, you will have to check them
manually if a problem cannot be repaired automatically. To check the filesystems manually after a crash, boot the
system up to emergency mode (page 386). Do not mount any devices other than root, which Linux mounts automatically. Run fsck on all the local filesystems that were mounted at the time of the crash, repairing them as needed.
Depending on how your system is set up, when fsck cannot repair a filesystem automatically, the system enters
emergency mode so you can run fsck manually. Make note of any ordinary files or directories that you repair (and
can identify), and inform their owners that they may not be complete or correct. Look in the lost+found directory
in each file-system for missing files. After successfully running fsck, type exit to exit from the single-user shell and
resume booting.
66
Draft
Useful Utilities
Draft
If files are not correct or are missing altogether, you may have to recreate them from a backup copy of the filesystem.
For more information, refer to “Backing Up Files” on page 524.
When the System Does Not Boot
When you cannot boot the computer from the hard drive, boot the system into rescue mode. For more information,
refer to “Rescue Mode” on page 377. If that works, run fsck on the root filesystem and try rebooting from the hard
drive again.
When all else fails, go through the install procedure, and perform an “upgrade” to your current version of Linux.
Red Hat Linux/Fedora Core can perform a nondestructive upgrade and can fix quite a bit in the process. For more
information, refer to “Upgrade Examine” on page 48.
Useful Utilities
This section briefly describes a few of the many utilities that can help you perform system administration tasks.
Read the man/info pages for the utilities described in “Linux Utilities” on page 396 to learn more about using them.
Some of these utilities are incorporated as part of the Main menu, and some are useful to users other than the system
administrator.
Red Hat Configuration Tools
The Red Hat configuration tools, most of which are named redhat-config- * in all older releases of Red Hat Linux,
including Red Hat Enterprise Linux and Fedora Core 1, were renamed system-config- * in Fedora Core 2. Many of
these tools bring up a graphical display when called from a GUI and a textual display when called from a non-GUI
command line. In general, these tools, which are listed Table 11-2, are simple to use and require little explanation
beyond what the tool presents. Some have Help selections on their toolbar; most do not have man pages.
67
Draft
Red Hat Configuration Tools
Draft
Table 11-2. Red Hat Configuration Tools
system-config-bind (FEDORA) redhat-config-bind (RHEL) Displays the Domain Name Service window. For more
information, refer to “JumpStart II: system-config-bind:
Setting Up a Domain” on page 712.
system-config-boot (FEDORA)
Displays the Boot Configuration window, which allows
you to specify which boot entry in /etc/grub.conf (page
514) you want the system to boot from.
system-config-date (FEDORA) redhat-config-date (RHEL) Displays the Date/Time Properties window with two
tabs: Date & Time and Time Zone. You can set the date
and time or enable NTP (network time protocol) from
the first tab. The Time Zone tab allows you to specify
the time zone of the system clock or whether you want
the system clock set to UTC (page 1004).
system-config-display (FEDORA) redhat-config-xfree86
(RHEL)
Brings up the Display settings window with three tabs:
Settings, Hardware, and Dual head. For more information, refer to “ system-config-display: Configuring the
Display” on page 63.
system-config-httpd (FEDORA) redhat-config-httpd
(RHEL)
Displays the HTTP window with four tabs: Main, Virtual
Hosts, Server, and Performance Tuning. For more information, refer to “JumpStart II: Setting Up Apache with
system-config-httpd ” on page 764.
system-config-keyboard (FEDORA) redhat-config-keyboard (RHEL)
Displays the Keyboard window, which allows you to select the type of keyboard attached to the system. This is
the utility you use to select the keyboard when you install
the system.
system-config-kickstart (FEDORA) redhat-config-kickstart Displays the Kickstart Configurator window, which allows you to create a Kickstart script. For more informa(RHEL)
tion, refer to “Using the Kickstart Configurator” on page
57.
system-config-language (FEDORA) redhat-config-language Displays the Language Selection window, which allows
you to specify the default system language from among
(RHEL)
those that are installed. This is the utility you use to select
the system language when you install the system.
redhat-config-mouse (RHEL)
Displays the Mouse Configuration window, which allows
you to specify the type of mouse that is attached to the
system. This is the utility you use to select the system
mouse when you install the system.
redhat-config-netboot (RHEL)
Displays the Network Installation and Diskless Environment window, which allows you to configure network
installation or a diskless environment. The first time you
run this utility, it displays the First Time Druid window.
system-config-network (FEDORA) redhat-config-network Displays the Network Configuration window. For more
information, refer to “ system-config-network: Configuring
(RHEL)
the Hardware” on page 555.
system-config-network-cmd (FEDORA) redhat-confignetwork-cmd (RHEL)
68
Displays the parameters that system-config-network uses.
Draft
Red Hat Configuration Tools
system-config-nfs (FEDORA) redhat-config-nfs (RHEL)
Draft
Displays the NFS Server Configuration window. For more
information, refer to “JumpStart: system-config-nfs:
Configures an NFS Server” on page 664.
system-config-packages (FEDORA) redhat-config-packages Displays the Package Management window. This is the
utility you use to customize the list of packages you install
(RHEL)
when you install the system. For more information, refer
to “ system-config-packages: Adds and Removes Software
Packages” on page 453.
system-config-printer (FEDORA) redhat-config-printer
(RHEL)
Displays the Printer configuration window, which allows
you to set up printers and edit printer configurations.
For more information, refer to “JumpStart I: Configuring
a Local Printer Using system-config-printer ” on page 487.
system-config-proc (FEDORA) redhat-config-proc (RHEL) Displays the Kernel Tuning window, which allows you
to tune the kernel by working with the /proc filesystem
(page 434). Be sure you know what you are doing if you
modify /proc, as you can cause the system to fail by
changing kernel parameters.
system-config-rootpassword (FEDORA) redhat-configrootpassword (RHEL)
Displays the Root Password window, which allows you
to change the root password. While logged in as root,
you can also use passwd from a command line to change
the root password.
system-config-samba (FEDORA) redhat-config-samba
(RHEL)
Displays the Samba Server Configuration window, which
can help you configure Samba. For more information,
refer to “JumpStart: system-config-samba: Configuring
a Samba Server” on page 679.
system-config-securitylevel (FEDORA) redhat-config-se- Displays the Security Level Configuration window with
two tabs: Firewall Options and SELinux (FEDORA).
curitylevel (RHEL)
Refer to “JumpStart: Using system-config-securitylevel to
Build a Firewall” on page 741 for more information about
configuring a firewall. See page 379 for information about
SELinux.
system-config-services (FEDORA) redhat-config-services Displays the Service Configuration window, which allows
you to specify which daemons (services) run at each of
(RHEL)
the runlevels. For more information, refer to “ systemconfig-services: Configures Services II” on page 383.
system-config-soundcard (FEDORA) redhat-configsoundcard (RHEL)
Displays the Audio Devices window, which tells you
which audio device the system detected and gives you
the option of playing a sound.
system-config-users (FEDORA) redhat-config-users
(RHEL)
Displays the User Manager window, which allows you
to work with users and groups. For more information,
refer to “ system-config-users: Manages User Accounts”
on page 521.
system-logviewer (FEDORA) redhat-logviewer (RHEL)
Displays the System Logs window, which can display
any of various system logs.
system-switch-mail (FEDORA) redhat-switch-mail (RHEL) Displays the system-switch-mail window, which allows
you to choose between the sendmail (page 609) and
Postfix (page 633) MTAs.
69
Draft
Linux Utilities
Draft
Linux Utilities
This section lists and explains a few command line system administration tools you may find useful.
chsh . Changes the login shell for a user. When you call chsh without an argument, you change your own login
shell. Superuser can change the shell for any user by calling chsh with that user’s username as an argument. When
changing a login shell with chsh, you must specify an installed shell that is listed in the file /etc/shells; other entries
are rejected. Also, you must give the pathname to the shell exactly as it appears in /etc/shells. In the following
example, Superuser changes Alex’s shell to tcsh:
# chsh alex
Changing the login shell for alex
Enter the new value, or press return for the default
Login Shell [/bin/bash]: /bin/tcsh
clear . Clears the screen. You can also use CONTROL-L from the bash shell to clear the screen. The value of
the environment variable TERM is used to determine how to clear the screen.
dmesg .
Displays recent system log messages (page 517).
e2label . Displays or creates a volume label on an ext2 or ext3 disk partition. The format of an e2label command
is shown following:
e2label device [newlabel]
where device is the name of the device (/dev/hda2, /dev/sdb1, /dev/fd0, and so on) you want to work with. When
you include the optional newlabel parameter, e2label changes the label on device to newlabel. Without this parameter,
e2label displays the label. You can also create a volume label with the –L option of tune2fs (page 447).
kudzu . The kudzu utility finds new and changed hardware and configures it. This utility determines which
hardware is new by probing all devices on internal and external buses and comparing the results to the /etc/sysconfig/hwconf database. In the default configuration, the /etc/rc.d/init.d/kudzu script runs and calls kudzu as the
machine enters runlevels 3 and 5. When it finds new or changed hardware, kudzu gives you a chance to configure
it and permits you to deconfigure any hardware that you have removed.
mkfs . Makes a New Filesystem. This utility is a front end for many utilities, each of which builds a a different
type of filesystem. By default, mkfs builds an ext2 filesystem and works on either a hard disk partition or a floppy
diskette. Although it can take many options and arguments, you can use mkfs simply as
# mkfs device
where device is the name of the device (/dev/hda2, /dev/sdb1, /dev/fd0, and so on) you want to make a
filesystem on. Use the –t option to specify a type of filesystem. The following command creates an ext3 filesystem
on device :
# mkfs -t ext3 device
ping . Sends packets to a remote system. This utility determines if you can reach a remote system through the
network and the time it takes to exchange messages with the remote system. Refer to “ping: Tests a Network
Connection” on page 346.
70
Draft
Linux Utilities
Draft
reset (link to tset) . Resets terminal characteristics. The value of the environment variable TERM (page 920)
is used to determine how to reset the screen. The screen is cleared, the kill and interrupt characters are set to their
default values, and character echo is turned on. From a graphical terminal emulator, this command also changes
the size of the window to its default. The reset utility is useful to restore your screen to a sane state after it has been
corrupted. Similar to an stty sane command.
setserial . Gets and sets serial port information; it is used by Superuser to configure a serial port. The following
command sets the input address of /dev/ttys0 to 0x100, the interrupt (IRQ) to 5, and the baud rate to 115,000
baud:
# setserial /dev/ttys0 port 0x100 irq 5 spd_vhi
You can also check the configuration of a serial port with setserial:
# setserial /dev/ttys0
/dev/ttyS0, UART: 16550A, Port: 0x0100, IRQ: 5, Flags: spd_vhi
Normally, setserial is called while the system is being booted if any of the serial ports needs to be set up specially.
umask . A shell builtin that specifies a mask the system uses to set up access permissions when you create a
file. The format of a umask command is shown following:
umask [
mask
]
where mask is a three-digit octal number or a symbolic value such as you would use with chmod (page 174). The
mask specifies the permissions that are not allowed. When mask is an octal number, the digits correspond to the
permissions for the owner of the file, members of the group the file is associated with, and everyone else. Because
the mask specifies the permissions that are not allowed, the system subtracts each of these from 7 when you create
a file. The result is three octal numbers that specify the access permissions for the file (the numbers you would use
with chmod). A mask that you specify as a symbolic value also specifies the permissions that are not allowed.
Most utilities and applications do not attempt to create files with execute permissions, regardless of the value of
mask; they assume you do not want an executable file. The effective result is that when a utility or application,
such as touch, creates a file, the system subtracts each of the digits in mask from 6. An exception is mkdir, which
assumes that you want the execute (access in the case of a directory) bit set.
The following commands set the file-creation permissions mask and display the mask and its effect when you create
a file and a directory. The mask of 022, when subtracted from 777, gives permissions of 644 (rw–r—r—) for a file
and 755 (rwxr–xr–x) for a directory.
$ umask 022
$ umask
0022
$ touch afile
$ mkdir adirectory
$ ls -ld afile adirectory
drwxr-xr-x 2 sam sam 4096 May 2 23:57 adirectory
-rw-r--r-- 1 sam sam
0 May 2 23:57 afile
71
Draft
Setting Up a Server
Draft
The next example sets the same mask value symbolically.
$ umask g=rx,o=rx
$ umask
0022
uname . Displays information about the system. When you run uname without any arguments, it displays the
name of the operating system (Linux). Giving uname a –a (all) option causes it to display the operating system
name, the hostname, the version number and release date of the operating system, and the type of hardware you
are using:
# uname –a
Linux pb 2.6.3-2.1.253.2.1 #1 Fri Mar 12 14:01:55 EST 2004 i686 athlon i386 GNU/Linux
Setting Up a Server
This section discusses issues that are common to setting up most servers: rules for writing configuration files; how
to specify hosts and subnets; how to use portmap, rpcinfo, xinetd, TCP wrappers (hosts.allow and hosts.deny);
and how to set up a chroot jail. Setting up specific servers is covered in Chapters 14 and 18–26. Setting up a LAN
is covered in Chapter 17 (page 551).
Standard Rules in Configuration Files
Most configuration files, typically named *.conf, rely on the following conventions:
•
Blank lines are ignored.
•
A # anywhere on a line starts a comment which continues to the end of the line. Comments are ignored.
•
When a name contains a SPACE, you must quote the SPACE by preceding it with a backslash (\) or by enclosing
the entire name within single or double quotation marks.
•
To make long lines easier to read and edit, you can break them into several shorter lines. Break a line by inserting
a backslash (\) immediately followed by a NEWLINE (press RETURN in a text editor). When you insert the
NEWLINE before or after a SPACE, you can indent the following line to make it easier to read. Do not break
lines in this manner while editing on a Windows machine, as the NEWLINEs may not be properly escaped
(Windows uses RETURN-LINEFEEDs to end lines).
Configuration files that do not follow these conventions are noted in the text.
Specifying Clients
Table 11-3 shows some of the common ways to specify a host or a subnet (next section). Most of the time you can
specify multiple hosts or subnets by separating the host or subnet specifications with SPACEs.
72
Draft
Standard Rules in Configuration Files
Draft
Table 11-3. Specifying a Client
Client Name Pattern
Matches
n.n.n.n
One IP address
name
One hostname, either local or remote
name that starts with .
Matches a host name that ends with the specified string.
For example, .tcorp.com matches the systems
kudos.tcorp.com and speedy.tcorp.com, among others.
IP address that ends with .
Matches a host address that starts with the specified
numbers. For example, 192.168.0. matches
192.168.0.0 –192.168.0.255. If you omit the trailing
period, this format does not work.
starts with @
Specifies a netgroup.
n.n.n.n/m.m.m.m or n.n.n.n/mm
An IP address and subnet mask specify a subnet (next
section).
starts with /
An absolute pathname of a file containing one or more
names or addresses as specified in this table.
Wildcards
* and ?
Matches one (?) or more (*) characters in a simple hostname or IP address. These wildcards do not match periods in a domain name.
ALL
Always matches.
LOCAL
Matches any hostname that does not contain a period.
Operator
EXCEPT
Examples .
Matches anything in the preceding list that is not in the
following list. For example, a b c d EXCEPT c matches
a, b, and d. So you could use 192.168. EXCEPT
192.168.0.1 to match all IP addresses that start with
192.168. except 192.168.0.1.
Each of the following examples specifies one or more systems:
10.10.
Matches all systems with IP addresses that start with
10.10..
.redhat.com
Matches all named hosts on the Red Hat network.
localhost
Matches the local system.
127.0.0.1
The loopback address; always resolves to the local host.
192.168.*.1
Could match all routers on a network of /24 subnets.
Specifying a Subnet
When you set up a server, you frequently need to specify the clients that are allowed to connect to the server.
Sometimes, it is convenient to specify a range of IP addresses, a subnet. The discussion on page 337 explains what
a subnet is and how to use a subnet mask to specify a subnet. Usually, you can specify a subnet as
n.n.n.n/m.m.m.m
73
Draft
rpcinfo: Displays Information About
portmap
Draft
or
n.n.n.n/mbits
where n.n.n.n is the base IP address and the subnet is represented by m.m.m.m (the subnet mask) or mbits (the
number of bits used for the subnet mask). For example, 192.168.0.1/255.255.255.0 represents the same subnet
as 192.168.0.1/24. In binary, decimal 255.255.255.0 is represented by 24 ones followed by eight zeros. The /24
is shorthand for a subnet mask with 24 ones. Each line in Table 11-4 presents two notations for the same subnet
followed by the range of IP addresses that the subnet includes.
Table 11-4. Different Ways to Represent a Subnet
Bits
Mask
Range
10.0.0.0/8
10.0.0.0/255.0.0.0
10.0.0.0 - 10.255.255.255
172.16.0.0/12
172.16.0.0/255.240.0.0
172.16.0.0 - 72.31.255.255
192.168.0.0/16
192.168.0.0/255.255.0.0
192.168.0.0 - 192.168.255.255
rpcinfo: Displays Information About portmap
The rpcinfo utility can display information about programs registered with portmap and can make RPC calls to
programs to see if they are alive. For more information on portmap, refer to “RPC Network Services” on page 358.
The rpcinfo utility takes the following options and arguments:
rpcinfo –p [
host]
rpcinfo [–n port
] –u | –t host program [
version]
rpcinfo –b | –d program version
–p .
probe .
Lists all RPC programs registered with portmap on host or on the local system if host is not specified.
–n .
(port) number .
With –t or –u, uses port number in place of the port number specified by portmap.
–u .
UDP .
Makes a UDP RCP call to version (if specified) of program on host and reports whether there was a response.
–t .
TCP .
Makes a TCP RCP call to version (if specified) of program on host and reports whether there was a response.
–b .
broadcast .
–d .
74
Makes an RPC broadcast to version of program and lists hosts that respond.
Draft
delete .
rpcinfo: Displays Information About
portmap
Draft
Removes local RPC registration for version of program. Superuser only.
Give the following command to see which RPC programs are registered with the portmap daemon on the system
named peach:
$ /usr/sbin/rpcinfo -p
program vers proto
100000
2
tcp
100000
2
udp
100024
1
udp
100024
1
tcp
100021
1
udp
100021
3
udp
...
peach
port
111
111
32768
32768
32769
32769
portmapper
portmapper
status
status
nlockmgr
nlockmgr
Use the –u option to display a list of versions of a daemon, such as ypserv, registered on a remote system (peach):
$ /usr/sbin/rpcinfo -u peach ypserv
program 100004 version 1 ready and waiting
program 100004 version 2 ready and waiting
Specify localhost to display a list of versions of a daemon registered on the local system:
$ /usr/sbin/rpcinfo -u localhost nfs
program 100003 version 2 ready and waiting
program 100003 version 3 ready and waiting
Locking down portmap . As the portmap daemon holds information about which servers are running on the
local system and which port each is running on, it is important that only trusted systems have access to this information. One way to ensure that only selected systems have access to portmap is to lock it down in the
/etc/hosts.allow and /etc/hosts.deny files (page 404). Put the following line in hosts.deny to prevent all systems
from using portmap on the local (server) system:
portmap: ALL
Test this setup from a remote system with the following command:
$ rpcinfo -p hostname
No remote programs registered.
Replace hostname with the name of the remote system that you changed the hosts.deny file on. The change is
immediate; you do not need to kill/restart a daemon.
Next, add the following line to the hosts.allow file on the server system:
portmap: host-IP
where host-IP is the IP address of the trusted, remote system that you gave the preceding rpcinfo command from.
Use only IP addresses with portmap in hosts.allow; do not use system names that portmap could get stuck trying
75
Draft
The xinetd Super Server
Draft
to resolve. Give the same rpcinfo command, and you should see a list of the servers that RPC knows about, including
portmap. See page 642 for more examples.
Caution: Set the Clocks
The portmap daemon relies on the client’s and server’s clocks being synchronized. A simple DoS attack
(page 968) can be initiated by setting the server’s clock to the wrong time.
The xinetd Super Server
The xinetd daemon is a more secure replacement for the inetd super server that originally shipped with 4.3BSD.
The Internet super server listens for network connections and, when one is made, launches a specified server daemon
and forwards the data from the socket (page 439) to the daemon’s standard input. The xinetd super server performs
these tasks and provides additional features to control access.
The version of xinetd distributed with Red Hat Linux is linked against libwrap.a, so it is capable of using the
/etc/hosts.allow and /etc/hosts.deny files for access control (see “TCP Wrappers,” page 404 for more information).
Using TCP Wrappers can simplify configuration but hides some of the more advanced features of xinetd.
The base configuration for xinetd is stored in the /etc/xinetd.conf file. The file supplied by Red Hat is shown
following:
$ cat /etc/xinetd.conf
# Simple configuration file for xinetd
defaults
{
instances
= 60
log_type
= SYSLOG authpriv
log_on_success
= HOST PID
log_on_failure
= HOST
cps
= 25 30
}
includedir /etc/xinetd.d
The defaults section specifies the default configuration of xinetd; the files in the included directory, /etc/xinetd.d,
specify server-specific configurations. Defaults can be overridden by server-specific configuration files.
In the preceding file, the instances directive specifies that no daemon may run more than 60 copies of itself at one
time. The log_type directive specifies that xinetd send messages to the system log daemon (syslogd, page 546)
using the authpriv facility. The next two lines specify what to log on success and on failure. The cps (connections
per second) directive specifies that no more than 25 connections to a specific service should be made per second
and that the service should be disabled for 30 seconds if this limit is exceeded.
The following xinetd configuration file allows telnet connections from the local system and any system with an IP
address that starts with 192.168.. This configuration file does not rely on TCP wrappers and so does not rely on
the hosts.allow and hosts.deny files.
$ cat /etc/xinetd.d/telnet
service telnet
{
socket_type
= stream
wait
= no
76
Draft
Securing a Server
user
server
only_from
disable
=
=
=
=
Draft
root
/usr/sbin/in.telnetd
192.168.0.0/16 127.0.0.1
no
}
The socket_type indicates whether the socket uses TCP or UDP. TCP-based protocols establish a connection
between the client and the server and are identified by the type stream. UDP-based protocols rely on the transmission
of individual datagrams and are identified by the type dgram.
When wait is set to no, xinetd handles multiple, concurrent connections to this service. Setting wait to yes causes
xinetd to wait for the server process to complete before handling the next request for that service. In general, UDP
services should be set to yes and TCP services to no. If you were to set wait to yes for a service such as telnet, only
one person would be able to use the service at a time.
The user specifies the user that the server runs as. If the user is a member of multiple groups, you can also specify
the group on a separate line using the keyword group. The user directive is ignored if xinetd is run as other than
root. The server provides the pathname of the server program that xinetd runs for this service.
The only_from specifies which systems xinetd allows to use the service. Use IP addresses only because using
hostnames can make the service unavailable if DNS fails. Zeros at the right of an IP address are treated as wildcards:
192.168.0.0 allows access from any system in the 192.168 subnet.
The disable line can disable a service without removing the configuration file. As shipped by Red Hat, a number
of services include an xinetd configuration file with disable set to yes. To run one of these services, you must
change disable to no in the appropriate file in xinetd.d and restart xinetd:
# /sbin/service xinetd restart
Stopping xinetd:
Starting xinetd:
[ OK ]
[ OK ]
Securing a Server
This section discusses how to secure a server using TCP wrappers or by setting up a chroot jail.
TCP Wrappers: Client/Server Security (hosts.allow and hosts.deny)
When you open a local system to access from remote systems, you must ensure that you
•
Open the local system only to systems you want to allow to access it.
•
Allow each remote system to access only the data you want it to access.
•
Allow each remote system to access data only in the manner you want it to (read only, read/write, write only).
As part of the client/server model, TCP Wrappers, which can be used for any daemon that is linked against libwrap.a,
uses the /etc/hosts.allow and /etc/hosts.deny files to form the basis of a simple access control language. This
access control language defines rules that selectively allow clients to access server daemons on a local system based
on the client’s address and the daemon the client tries to access.
Each line in the hosts.allow and hosts.deny files has the following format:
77
Draft
Securing a Server
Draft
daemon_list : client_list [: command
]
where daemon_list is a comma-separated list of one or more server daemons (such as portmap, vsftpd, sshd),
client_list is a comma-separated list of one or more clients (see Table 11-3, “Specifying a Client,” on page 399), and
the optional command is the command that is executed when a client from client_list tries to access a server daemon
from daemon_list.
When a client requests a connection with a local server, the hosts.allow and hosts.deny files are consulted as
follows until a match is found. The first match determines whether the client is allowed to access the server.
1.
If the daemon/client pair matches a line in hosts.allow, access is granted.
2.
If the daemon/client pair matches a line in hosts.deny, access is denied.
3.
If there is no match in either the hosts.allow or the hosts.deny files, access is granted.
When either the hosts.allow or the hosts.deny file does not exist, it is as though that file were empty. Although
it is not recommended, you can allow access to all daemons for all clients by removing both files.
Examples .
For a more secure system, put the following line in hosts.deny to block all access:
$ cat /etc/hosts.deny
...
ALL : ALL : echo '%c tried to connect to %d and was blocked' >> /var/log/tcpwrappers.log
This line blocks any client attempting to connect to any service, unless specifically permitted in hosts.allow. When
this rule is matched, it adds a line to the /var/log/tcpwrappers.log file. The %c expands to client information
and the %d expands to the name of the daemon the client attempted to connect to.
With the preceding hosts.deny file in place, you can put lines in hosts.allow explicitly to allow access to certain
services and systems. For example, the following hosts.allow file allows anyone to connect to the OpenSSH daemon
(ssh, scp, sftp) but allows telnet connections only from the same network as the local system and users on the
192.168. subnet:
$ cat /etc/hosts.allow
sshd : ALL
in.telnet : LOCAL
in.telnet : 192.168.? 127.0.0.1
...
The first line in the preceding file allows connection from any system (ALL) to sshd. The second line allows connection from any system in the same domain as the server (LOCAL). The third line matches any system whose IP
address starts 192.168. and the local system.
Setting Up a chroot Jail
On early UNIX systems, the root directory was a fixed point in the file system. On modern UNIX variants, including
Linux, you can define the root directory on a per-process basis. The chroot utility allows you to run a process with
a root directory other than /.
78
Draft
Securing a Server
Draft
The root directory is the top of the directory hierarchy and has no parents: A process cannot access any files above
the root directory (because they do not exist). If, for example, you run a program (process) and specify its root
directory as /home/sam/jail, the program would have no concept of any files in /home/sam or above: jail is the
program’s root directory and is labeled / (not jail).
By creating an artificial root directory, frequently called a (chroot) jail, you prevent a program from being able to
access and modify, possibly maliciously, files outside the directory hierarchy starting at its root. You must set up
a chroot jail properly in order to increase security: If you do not set up a chroot jail correctly, you can make it easier
for a malicious user to gain access to a system than if there were no chroot jail.
Using chroot
Creating a chroot jail is simple: Working as root, give the command /usr/sbin/chroot directory. The directory
becomes the root directory and the process attempts to run the default shell. Working as root from the /home/sam
directory, the following example attempts to set up a chroot jail in the (existing) /home/sam/jail directory:
# /usr/sbin/chroot jail
/usr/sbin/chroot: /bin/bash: No such file or directory
In the preceding example, chroot sets up the chroot jail, but when it attempts to run a the bash shell, it fails. Once
the jail is set up, the directory that was named jail takes on the name of the root directory: / and chroot cannot
find the file identified by the pathname /bin/bash. The chroot jail is working perfectly but is not very useful.
Getting a chroot jail to work the way you want is more complicated. To get the preceding example to run bash in
a chroot jail, you need to create a bin directory in jail (/home/sam/jail/bin) and copy /bin/bash to this directory.
Because the bash binary is dynamically linked to shared libraries (page 813), you need to copy these libraries into
jail too. The libraries go in /lib. The following example creates the necessary directories, copies bash, uses ldd to
display the shared library dependencies of bash, and copies the necessary libraries into lib:
$ pwd
/home/sam/jail $ mkdir bin lib
$ cp /bin/bash bin
$ ldd bin/bash
libtermcap.so.2 => /lib/libtermcap.so.2 (0x4002a000)
libdl.so.2 => /lib/libdl.so.2 (0x4002f000)
libc.so.6 => /lib/tls/libc.so.6 (0x42000000)
/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
$ cp /lib/libtermcap.so.2 lib
$ cp /lib/libdl.so.2 lib
$ cp /lib/tls/libc.so.6 lib
$ cp /lib/ld-linux.so.2 lib
Now that everything is set up, you can try starting the chroot jail again. All the setup can be done by an ordinary
user; you have to run chroot as Superuser:
$ su
Password:
# /usr/sbin/chroot .
bash-2.05b# pwd
/
bash-2.05b# ls
79
Draft
Securing a Server
Draft
bash: ls: command not found
bash-2.05b#
This time the chroot finds and starts bash, which displays its default prompt (bash-2.05b#). The pwd command
works because it is a shell builtin (page 211). However, bash cannot find the ls utility (it is not in the chroot jail).
You can copy /bin/ls and its libraries into the jail if you want users in the jail to be able to use ls.
To set up a useful chroot jail, you need to determine which utilities the users of the chroot jail are going to need.
Then, you can copy the binaries and their libraries into the jail, or you can build static copies of the binaries and
put them in the jail without the need for separate libraries. (The statically linked binaries are considerably larger
than their dynamic counterparts. The base system with bash and the core utilities is over 50 megabytes.) You can
find the source code for most of the common utilities you need in the bash and coreutils SRPMS (source rpm)
packages.
Whichever technique you choose, you must put a copy of su in the jail. The su command is required to run programs
as a user other than root. Because root can break out of a chroot jail, it is imperative that you run a program in the
chroot jail as a user other than root.
The dynamic version of su that Red Hat distributes requires PAM and will not work within a jail. You need to build
a copy of su from the source to use in a jail. By default, a copy of su you build does not require PAM. Refer to “GNU
Configure and Build System” on page 459 for instructions on how to build packages such as coreutils (which includes
su).
In order to be able to use su, you need to copy the relevant lines from the /etc/passwd and /etc/shadow files into
files with the same names in the etc directory inside the jail.
Tip: Keeping Multiple chroot Jails
If you are going to deploy multiple chroot jails, it is a good idea to keep a clean copy of the bin and
lib files somewhere other than in one of the active jails.
Running a Service in a chroot Jail
Running a shell inside a jail has limited use. You are more likely to need to run a specific service inside the jail. To
run a service inside a jail, you need make sure all the files needed by that service are inside the jail. The format of
a command to start a service in a chroot jail is
# /usr/sbin/chroot jailpath /bin/su user daemonname &
Where jailpath is the pathname of the jail directory, user is the username that runs the daemon, and daemonname
is the path (inside the jail) of the daemon that provides the service.
Several servers are set up to take advantage of chroot jails. The system-config-bind utility (page 712) automatically
sets up DNS so named runs in a jail and the vsftpd FTP server can automatically start chroot jails for clients (page
598).
Security Considerations
Some services need to be run as root, but they release their root privilege once started (Procmail and vsftpd are
examples). If you are running such a service, you do not need to put su inside the jail.
It is possible for a process run as root to escape from a chroot jail. For this reason, you should always su to another
user before starting a program running inside the jail. Also, be careful about what setuid (page 175) binaries you
80
Draft
DHCP
Draft
allow inside a jail because a security hole in one of these can compromise the security of the jail. Also, make sure
the user cannot access executable files that she uploads.
DHCP
Instead of having network configuration information stored in local files on each system, DHCP (Dynamic Host
Configuration Protocol) enables client systems to retrieve network configuration information each time they connect
to the network. A DHCP server assigns temporary IP addresses from a pool of addresses to clients as needed.
This technique has several advantages over storing network configuration information in local files:
•
DHCP makes it possible for a new user to set up an Internet connection without having to work with IP addresses,
netmasks, DNS addresses, and other technical details. DHCP also makes it faster for an experienced user to set
up a connection.
•
DHCP facilitates assignment and management of IP addresses and related network information by centralizing
the process on a server. A system administrator can configure new systems, including laptops that connect to
the network from different locations, to use DHCP, which assigns IP addresses only when each system connects
to the network. The pool of IP addresses is managed as a group on the DHCP server.
•
DHCP enables IP addresses to be used by more than one system, reducing the number of IP addresses needed
overall. This conservation of addresses is important as the Internet is quickly running out of IPv4 addresses.
Although one IP address can be used by only one system at a time, many end-user systems require addresses
only occasionally, when they connect to the Internet. By reusing IP addresses, DHCP lengthens the life of the
IPv4 protocol. DHCP applies to IPv4 only, as IPv6 forces systems to configure their IP addresses automatically
(called autoconfiguration) when they connect to a network (page 340).
DHCP is particularly useful for administrators responsible for a large number of systems because it removes the
requirement for individual systems to store unique configuration information. DHCP allows an administrator to
set up a master system and deploy new systems with a copy of the master’s hard disk. In educational establishments
and other open access facilities, it is common to store the hard disk image on a shared drive and have each workstation automatically restore itself to pristine condition at the end of each day.
More Information
Web .
www.dhcp.org [http://www.dhcp.org]
FAQ .
www.dhcp-handbook.com/dhcp_faq.html [http://www.dhcp-handbook.com/dhcp_faq.html]
HOWTO .
DHCP Mini HOWTO
How DHCP Works
The client daemon, dhclient, contacts the server daemon, dhcpd, to obtain the IP address, netmask, broadcast
address, nameserver address, and other networking parameters. The server provides a lease on the IP address to
the client. The client can request specific terms of the lease, including its duration, and the server can limit these
terms. While connected to the network, a client typically requests extensions of its lease as necessary so its IP address
remains the same. The lease can expire once the client is disconnected from the network; the server can give the
client a new IP address when it requests a new lease. You can also set up a DHCP server to provide static IP addresses
for specific clients (refer to “Static versus Dynamic IP Addresses” on page 334).
DHCP is broadcast based, so the client and server must be on the same subnet (page 337).
81
Draft
DHCP
Draft
DHCP Client
A DHCP client requests network configuration parameters from the DHCP server and uses those parameters to
configure its network interface.
Prerequisites
Install the following package:
•
dhclient
dhclient: The DHCP Client
When a DHCP client system connects to the network, dhclient requests a lease from the DHCP server and configures
the client’s network interface(s). Once a DHCP client has requested and established a lease, it stores information
about the lease in /etc/dhclient.leases. This information is used to reestablish a lease when either the server or
client needs to reboot. The DHCP client configuration file, /etc/dhclient.conf, is required only for custom configurations. The following dhclient.conf file specifies a single interface, eth0:
$ cat /etc/dhclient.conf
interface "eth0"
{
send dhcp-client-identifier 1:xx:xx:xx:xx:xx:xx;
send dhcp-lease-time 86400;
}
In the preceding file, the 1 in the dhcp-client-identifier specifies an Ethernet network and xx:xx:xx:xx:xx:xx is
the MAC address (page 981) of the device controlling that interface. See page 412 for instructions on how to display
a MAC address. The dhcp-lease-time is the duration, in seconds, of the lease on the IP address. While the client
is connected to the network, dhclient automatically renews the lease each time half of the lease is up. The choice
of 86400 seconds, or one day, is a reasonable choice for a workstation.
DHCP Server
The DHCP server maintains a list of IP addresses and other configuration parameters. When requested, the DHCP
server provides configuration parameters to a client.
Prerequisites
Install the following package:
•
dhcpd
Run chkconfig to cause dhcpd to start when the system goes multiuser:
# /sbin/chkconfig dhcpd on
Start dhcpd:
# /sbin/service dhcpd start
82
Draft
DHCP
Draft
dhcpd: The DHCP Daemon
A simple DCHP server allows you to add clients to a network without maintaining a list of assigned IP addresses.
A simple network, such as a home LAN sharing an Internet connection, can use DHCP to assign a dynamic IP address
to almost all nodes. The exceptions are servers and routers, which must be at known network locations in order
to be able to receive connections. If servers and routers are configured without DHCP, you can specify a simple
DHCP server configuration in /etc/dhcpd.conf:
$ cat /etc/dhcpd.conf
default-lease-time 600;
max-lease-time 86400;
option subnet-mask 255.255.255.0;
option broadcast-address 192.168.1.255;
option routers 192.168.1.1;
option domain-name-servers 192.168.1.1;
subnet 192.168.1.0 netmask 255.255.255.0 {
range 192.168.1.2 192.168.1.200;
}
The preceding configuration file specifies a LAN where the router and DNS are both located on 192.168.1.1. The
default-lease-time specifies the number of seconds a dynamic IP lease will be valid if the client does not specify
a duration. The max-lease-time is the maximum time allowed for a lease.
The information in the option lines is sent to each client when it connects. The names following the word option
specify what the following argument represents. For example, the option broadcast-address line specifies the
broadcast address of the network. The routers and domain-name-servers options allow multiple values separated
by commas.
The subnet section includes a range line that specifies the range of IP addresses that the DHCP server can assign.
If you define multiple subnets, you can define options, such as subnet-mask, inside the subnet section. Options
defined outside all subnet sections are global and apply to all subnets.
The preceding configuration file assigns addresses in the range between 192.168.1.2 and 192.168.1.200. The DHCP
server starts at the bottom of the range and attempts to assign a new IP address to each new client. Once the DHCP
server reaches the top of the range, it starts reassigning IP addresses that have been used, but are not currently in
use. If you have fewer systems than IP addresses, the IP address of each system should remain fairly constant. You
cannot use the same IP address for more than one system at a time.
Once you have configured a DHCP server, you can start or restart it using the dhcpd init script:
# /sbin/service dhcpd restart
Once the server is running, clients configured to obtain an IP address from the server using DHCP should be able
to do so.
Static IP Addresses
As mentioned earlier, routers and servers typically require static IP addresses. While you can manually configure
IP addresses for these systems, it may be more convenient to have the DHCP server provide them with static IP
addresses.
When a system that requires a specific static IP address connects to the network and contacts the DHCP server,
the server needs a way to identify the system so it can assign the proper IP address. The DHCP server uses the MAC
83
Draft
nsswitch.conf: Which Service to Look
at First
Draft
address (page 981) of the system’s Ethernet card (NIC) to identify the system. When you set up the server, you
must know the MAC addresses of each of the systems that requires a static IP address.
Displaying a MAC address . You can use ifconfig to display the MAC addresses of the Ethernet boards (NICs)
in a system. The MAC address is the colon-separated series of hexadecimal number pairs following HWaddr:
$ /sbin/ifconfig | grep -i hwaddr
eth0
Link encap:Ethernet HWaddr BA:DF:00:DF:C0:FF
eth1
Link encap:Ethernet HWaddr 00:02:B3:41:35:98
Run ifconfig on each system that requires a static IP address. Once you have determined the MAC address of each
system that requires a static IP address, you can add a host section to the /etc/dhcpd.conf file for each of these
systems, instructing the DHCP server to assign a specific address to each system. The following host section assigns
the address 192.168.1.1 to the system with the MAC address of BA:DF:00:DF:C0:FF:
$ cat /etc/dhcpd.conf
...
host router {
hardware ethernet BA:DF:00:DF:C0:FF;
fixed-address 192.168.1.1;
option host-name router;
}
The name following host is used internally by dhcpd, while the name specified after option host-name is passed
to the client and can be a hostname or a FQDN.
After making changes to dhcpd.conf, restart dhcpd using service and the dhcpd init script (page 411).
nsswitch.conf: Which Service to Look at First
With the advent of NIS and DNS, it was no longer a simple matter of searching a local file for user and system information. Where you used to look in /etc/passwd to get user information and look in /etc/hosts to find system
address information, you can now use several methods to find this type of information. The /etc/nsswitch.conf
(name service switch configuration) file specifies which methods to use and the order to use them in when looking
for a certain type of information. You can also specify what action the system takes based on whether a method
works or fails.
Format . Each line in nsswitch.conf specifies how to search for a piece of information, such as a user’s password.
The format of a line in nsswitch.conf is
info:
[action]
] [
method [
[action]
]...]
method [
where info specifies the type of information that the line describes (discussed following), method is the method
used to find the information (described following), and action is the response to the return status of the preceding
method. The action is enclosed within square brackets.
84
Draft
How nsswitch.conf Works
Draft
How nsswitch.conf Works
When called upon to supply information that nisswitch.conf describes, the system examines the line with the
appropriate info field. The system uses the methods specified on the line starting with the method on the left. By
default, when the information is found, the system stops searching. Without an action specification, when a
method fails to return a result, the next action is tried. It is possible for the search to end without finding the requested
information.
Information
The nisswitch.conf file commonly controls searches for users (in passwd), passwords (in shadow), host IP address,
and group information. The following list describes most of the types of information (info in the format discussed
earlier) that nsswitch.conf controls searches for.
automount
Automount (/etc/auto.master and /etc/auto.misc, page
671)
bootparams
Diskless and other booting options. See the bootparam
man page.
ethers
MAC address (page 981)
group
Groups of users (/etc/group, page 428)
hosts
System information (/etc/hosts, page 428)
netgroup
Netgroup information (/etc/netgroup, page 430)
networks
Network information (/etc/networks)
passwd
User information (/etc/passwd, page 430)
protocols
Protocol information (/etc/protocols, page 432)
publickey
Used for NFS running in secure mode
rpc
RPC names and numbers (/etc/rpc, page 433)
services
Services information (/etc/services, page 433)
shadow
Shadow password information (/etc/shadow, page 433)
Methods
Following is a list of the types of information that nsswitch.conf controls searches for (method in the format on
page 413). For each type of information, you can specify one or more of the following methods:2
files
Searches local files such as /etc/passwd and /etc/hosts.
nis
Searches the NIS database; yp is an alias for nis.
dns
Queries the DNS (hosts queries only).
compat
± syntax in passwd, group, and shadow files (page
416).
2
There are other, less commonly used methods. See the default /etc/nisswitch.conf file and the nisswitch.conf man page for more information.
Although NIS+ belongs in this list, it is not implemented for Linux and is not discussed in this book.
85
Draft
How nsswitch.conf Works
Draft
Search Order
The information provided by two or more methods may overlap: For example, files and nis can each provide
password information for the same user. With overlapping information, you need to consider which method you
want to be authoritative (take precedence), and put that method at the left of the list of methods.
The default nsswitch.conf file lists methods without actions, assuming no overlap (which is normal). In this case,
the order is not critical: When one method fails, the system goes to the next one; all that is lost is a bit of time for
a failure. Order becomes critical when you use actions between methods, or when you have overlapping entries
that differ.
The first of the following lines from nisswitch.conf causes the system to search for password information in
/etc/passwd and, if that fails, to use NIS to find the information. If the user you are looking for is listed in both
places, the information in the local file would be used and therefore would be authoritative. The second line uses
NIS and, if that fails, searches /etc/hosts and, if that fails, checks with DNS to find host information.
passwd
files nis
hosts
nis files dns
Action Items
Each method can optionally be followed by an action item that specifies what to do if the method succeeds or fails
for one of a number of reasons. The format of an action item is
[[!]
STATUS
=
action
]
where the opening and closing square brackets are part of the format and do not indicate that the contents are
optional; STATUS (by convention uppercase although it is not case sensitive) is the status being tested for; and
action is the action to be taken if STATUS matches the status returned by the preceding method. The leading exclamation point (!) is optional and negates the status.
Values for STATUS are
NOTFOUND .
The method worked, but the value being searched for was not found. Default action is continue.
SUCCESS . The method worked, and the value being searched for was found; no error was returned. Default
action is return.
UNAVAIL . The method failed because it is permanently unavailable. For example, the required file may not
be accessible or the required server may be down. Default action is continue.
TRYAGAIN . The method failed because it is temporarily unavailable. For example, a file may be locked or a
server overloaded. Default action is continue.
Values for action are
return .
continue .
method.
86
Returns to the calling routine with or without a value.
Continues with the next method. Any returned value is overwritten by a value found by the next
Draft
PAM
Draft
For example, the following line from nsswitch.conf causes the system first to use DNS to search for the IP address
of a given host. The action item following the DNS method tests to see if the status returned by the method is not
(!) UNAVAIL.
hosts
dns [!UNAVAIL=return] files
The system takes the action associated with the STATUS (return) if the DNS method did not return UNAVAIL
(!UNAVAIL), that is, if DNS returned SUCCESS, NOTFOUND, or TRYAGAIN. The result is that the following
method (files) is used only when the DNS server is unavailable: If the DNS server is not unavailable (read the two
negatives as “is available,”), the search returns the domain name or reports the domain name was not found. The
search uses the files method (check the local /etc/hosts file) only if the server is not available.
compat Method: ± in passwd, group, and shadow files
You can put special codes in the /etc/passwd, /etc/group, and /etc/shadow files that cause the system, when you
specify the compat method in nisswitch.conf, to combine and modify entries in the local files and the NIS maps.
A plus sign (+) at the beginning of a line in one of these files adds NIS information; a minus sign (–) removes information. For example, to use these codes in the passwd file, specify passwd: compat in nisswitch.conf. The
system then goes through the passwd file in order, adding or removing the appropriate NIS entries when it gets
to each line that starts with a + or –.
Although you can put a plus sign at the end of the passwd file and specify passwd: compat in nsswitch.conf to
search the local passwd file and then go through the NIS map, it is more efficient to put passwd: file nis in nsswitch.conf and not modify the passwd file.
PAM
PAM (actually Linux-PAM, or Linux Pluggable Authentication Modules) allows a system administrator to determine
how various applications use authentication (page 957) to verify the identity of a user. PAM provides shared libraries
(page 813) of modules (located in /lib/security) that, when called by an application, authenticate a user. Pluggable
refers to the ease with which you can add and remove modules from the authentication stack. The configuration
files kept in the /etc/pam.d directory determine the method of authentication and contain a list, or stack, of calls
to the modules. PAM may also use other files, such as /etc/passwd, when necessary.
Instead of building the authentication code into each application, PAM provides shared libraries to keep the authentication code separate from the application code. The techniques of authenticating users stay the same from
application to application. PAM enables a system administrator to change the authentication mechanism for a
given application without touching the application.
PAM provides authentication for a variety of system-entry services (login, ftp, and so on). You can take advantage
of PAM’s ability to stack authentication modules to integrate system-entry services with different authentication
mechanisms, such as RSA, DCE, Kerberos, and smart cards.
From login through using su to shutting the system down, whenever you are asked for a password (or not asked
for a password because the system trusts that you are who you say you are), PAM makes it possible for systems
administrators to configure the authentication process and makes the configuration process essentially the same
for all applications that use PAM to do their authentication.
The configuration files stored in /etc/pam.d describe the authentication procedure for each application. These
files usually have names that are the same as or similar to the name of the application that they configure. For example, authentication for the login utility is configured in /etc/pam.d/login. The name of the file is the name of
87
Draft
More Information
Draft
the PAM service3 that the file configures. Occasionally, one file may serve two programs. PAM accepts only
lowercase letters in the names of files in the /etc/pam.d directory.
Tip: Do Not Lock Yourself Out of the System
Editing PAM configuration files correctly takes care and attention. It is easy to lock yourself out of
the computer with a single mistake. To avoid this type of problem, always keep backup copies of the
PAM configuration files you edit, test every change thoroughly, and make sure you can still log in
once the change is installed. Keep a Superuser session open until you are done testing. When a change
fails and you cannot log in, use the Superuser session to replace the newly edited files with their backup
copies.
PAM warns you about any errors it encounters, logging them to the /var/log/messages or /var/log/secure files.
Look in these files if you are trying to figure out why a changed PAM file is not working properly. In order to prevent
possibly giving unnecessary information to a malicious user, PAM sends error messages to a file rather than to the
screen.
More Information
Local .
/usr/share/doc/pam-*/html/index.html
Web . www.kernel.org/pub/linux/libs/pam/Linux-PAM-html/pam.html
[http://www.kernel.org/pub/linux/libs/pam/Linux-PAM-html/pam.html] (Linux-PAM System Administrators’
Guide)
HOWTO .
User Authentication HOWTO
Configuration File, Module Type, and Control Flag
Following is an example of a PAM configuration file. Comment lines begin with a pound sign (#):
Login module .
$ cat /etc/pam.d/login
#%PAM-1.0
auth
required
auth
required
auth
required
account
required
password
required
session
required
session
required
session
optional
pam_securetty.so
pam_stack.so service=system-auth
pam_nologin.so
pam_stack.so service=system-auth
pam_stack.so service=system-auth
pam_selinux.so multiple
pam_stack.so service=system-auth
pam_console.so
The first line is a special comment; it will become significant only if another PAM format is released. Do not use
#% other than its use in the first line of the preceding example.
The rest of the lines tell PAM to do something as part of the authentication process. The first word on each line is
a module type indicator: account, auth, password, or session (Table 11-5). The second is a control flag (Table
3
There is no relationship between PAM services and the /etc/services file. The name of the PAM service is an arbitrary string that each application gives to PAM; PAM then looks up the configuration file with that name and uses it to control how it does authentication. There is no
central registry of PAM service names.
88
Draft
Configuration File, Module Type, and
Control Flag
Draft
11-6), which indicates what type of action to take if authentication fails. The rest of the line contains the pathname
of a PAM module and any arguments for that module. The PAM library itself uses the /etc/pam.d files to determine
which modules to delegate work to.
Table 11-5. Module Type Indicators
Module Type
Description
Controls
account
Account management
Determining whether an already authenticated user is allowed to use the
service he/she is trying to use. (That
is, has the account expired? Is the user
allowed to use this service at this time
of day?)
auth
Authentication
Proving that the user is authorized to
use the service. This may be done using passwords or another mechanism.
password
Password changing
Updating authentication mechanisms
such as user passwords.
session
Session management
Setting things up when the service is
started (for example, when the user
logs in) and breaking them down when
the service is terminated (for example,
when the user logs out).
89
Draft
Configuration File, Module Type, and
Control Flag
Draft
Table 11-6. Control Flag
required
Success is required for authentication to succeed. Control
and a failure result are returned after all the modules in
the stack have been executed. The technique of delaying
the report to the calling program until all modules have
been executed may keep attackers from knowing what
caused their authentication attempts to fail and tell them
less about the system, making it more difficult for them
to break in.
requisite
Success is required for authentication to succeed. Further
module processing is aborted, and control is returned
immediately after a module fails. This technique may
expose information about the system to an attacker. On
the other hand, if it prevents a user from giving a password over an insecure connection, it might keep information out of the hands of an attacker.
sufficient
Success indicates that this module type has succeeded,
and no subsequent required modules of this type are executed. Failure is not fatal to the stack of this module
type. This technique is generally used when one form or
another of authentication is good enough: If one fails,
PAM tries the other. For example, when you use rsh to
connect to another computer, pam_rhosts first checks
to see whether your connection can be trusted without
a password. If the connection can be trusted, the
pam_rhosts module reports success, and PAM immediately reports success to the rsh daemon that called it. You
will not be asked for a password. If your connection is
not considered trustworthy, PAM starts the authentication over, asking for a password. If this second authentication succeeds, PAM ignores the fact that the
pam_rhosts module reported failure. If both modules
fail, you will not be able to log in.
optional
Result is generally ignored. An optional module is relevant only when it is the only module on the stack for a
particular service.
You can use one of the Control Flag keywords listed in Table 11-6 to set the control flags.
PAM uses each of the module types as requested by the application. That is, the application will ask PAM separately
to authenticate, check account status, manage sessions, and change the password. PAM will use one or more
modules from the /lib/security directory to accomplish each of these tasks.
The configuration files in /etc/pam.d list the set of modules to be used for each application to do each task. Each
such set of the same module types is called a stack. PAM calls the modules one at a time in order, from the top of
the stack (the first module listed in the configuration file) to the bottom. The modules report success or failure
back to PAM. When all the stacks of modules (there are exceptions) within a configuration file have been called,
the PAM library reports success or failure back to the application.
90
Draft
Example
Draft
Example
Part of the login service’s authentication stack follows as an example:
$ cat /etc/pam.d/login
#%PAM-1.0
auth
required
auth
required
auth
required
...
/lib/security/pam_securetty.so
/lib/security/pam_stack.so service=system-auth
/lib/security/pam_nologin.so
The login program first asks for a user name and then asks PAM to run this stack to authenticate the user. Refer to
the Table 11-5 on page 418 and Table 11-6 on page 418.
1.
PAM first calls the pam_securetty (secure tty) module to make sure that the root user logs in only from an
allowed terminal (by default, root is not allowed to run login over the network; this policy helps prevent security breaches). The pam_securetty module is required to succeed in order for the authentication stack to
succeed. The pam_securetty module reports failure only if someone is trying to log in as root from an unauthorized terminal. Otherwise (if the user name being authenticated is not root or if the user name is root and
the login attempt is being made from a secure terminal), the pam_securetty module reports success.
Success and failure within PAM are opaque concepts that apply only to PAM. They do not equate to true and
false as used elsewhere in the operating system.
2.
The pam_stack.so module diverts the PAM stack to another module, returning the success or failure as the
other module returns to it. The service=system-auth gives the name of the other module, in this case systemauth in /etc/pam.d. This module checks that the user who is logging in is authorized to do so, including
checks on the username and password.
3.
Next, the pam_nologin module makes sure that if the /etc/nologin.txt file exists, only the root user is allowed
to log in. (That is, the pam_nologin module reports success only if /etc/nologin.txt does not exist or if the
root user is logging in.) Thus when a shutdown has been scheduled for some time in the near future, the system
administrator can keep users from logging in on the system only to experience a shutdown moments later.
The account module type works like the auth module type but is called after the user has been authenticated; it
is an additional security check or requirement for a user to gain access to the system. For example, account modules
can enforce a requirement that a user can log in only during business hours.
The session module type sets up and tears down the session (perhaps mounting and unmounting the user’s home
directory). One common session module on a Red Hat system is the pam_console module, which sets the system
up especially for users who log in at the physical console, not for those who log in remotely. A local user is able to
access the floppy and CD drives, the sound card, and sometimes other devices as defined by the system administrator.
The password module type is a bit unusual: All the modules in the stack are called once and told to get all the information they need to store the password to persistent memory, such as a disk, but not actually to store it. If it
determines that it cannot or should not store the password, a module reports failure. If all the password modules
in the stack report success, they are called a second time and told to store to persistent memory the password they
obtained on the first pass. The password module is responsible for updating the authentication information (that
is, changing the user’s password).
Any one module can act as more than one module type; many modules can act as all four module types.
91
Draft
Modifying the PAM Configuration
Draft
Caution: Brackets ([]) in the Control Flags Field
You can set the control flags in a more complex way. When you see brackets ([]) in the control flags
position in a PAM configuration file, the newer, more complex method is in use. Each comma-delimited
argument is a value=action pair. When the return from the function matches value, action is evaluated. Refer to the PAM System Administrator’s Guide (/usr/share/doc/pam-*/txts/pam.txt) for more
information.
Modifying the PAM Configuration
Some UNIX systems require that a user be a member of the wheel group in order to use the su command. Although
Red Hat is not configured this way by default, PAM allows you to change the default by editing the /etc/pam.d/su
file:
$ cat /etc/pam.d/su
#%PAM-1.0
auth
sufficient
/lib/security/$ISA/pam_rootok.so
# Uncomment the following line to implicitly trust users in the "wheel" group.
#auth
sufficient
/lib/security/$ISA/pam_wheel.so trust use_uid
# Uncomment the following line to require a user to be in the "wheel" group.
#auth
required
/lib/security/$ISA/pam_wheel.so use_uid
auth
required
/lib/security/$ISA/pam_stack.so service=system-auth
account
required
/lib/security/$ISA/pam_stack.so service=system-auth
password
required
/lib/security/$ISA/pam_stack.so service=system-auth
session
required
/lib/security/$ISA/pam_stack.so service=system-auth
session
optional
/lib/security/$ISA/pam_selinux.so multiple
session
optional
/lib/security/$ISA/pam_xauth.so
The third through sixth lines of the su module contain comments that include the lines necessary to permit members
of the wheel group to run su without supplying a password (sufficient) and to permit only users who are in the
wheel group to use su (required). Uncomment one of these lines when you want your system to follow one of these
rules.
Caution: Do Not Create /etc/pam.conf
You may have encountered PAM on other systems where all configuration is arranged in a single file
(/etc/pam.conf). This file does not exist on Red Hat systems. Instead, the /etc/pam.d directory
contains individual configuration files, one per application that uses PAM. This setup makes it easy
to install and uninstall applications that use PAM without having to modify the /etc/pam.conf file
each time. If you create a /etc/pam.conf file on a system that does not use this file, your PAM configuration may become confused. Do not use PAM documentation from a different system. Also, the
requisite control flag is unavailable on some systems that support PAM.
Chapter Summary
A system administrator is someone who keeps the system useful and convenient for its users. This chapter describes
many of the files and programs you will work with to maintain a Red Hat system. Much of the work you do as the
system administrator requires you to log in as root. The root user, called Superuser, has extensive systemwide
powers that normal users do not have. Superuser can read from and write to any file and can execute programs
that ordinary users are not permitted to execute.
92
Draft
Exercises
Draft
When you bring up the system, it is sometimes in single-user mode. In this mode, only the system console is functional, and not all the filesystems are mounted. When the system is in single-user mode, you can back up files and
use fsck to check the integrity of filesystems before you mount them. The telinit utility brings the system to its
normal multiuser state. With the system running in multiuser mode, you can still perform many administration
tasks, such as adding users and printers.
The chapter describes system operation: booting up, running init scripts, single-user mode, emergency mode,
multiuser mode, bringing the system down, and what to do when the system crashes.
The chapter covers the Red Hat configuration tools (redhat-config- * under Red Hat Enterprise Linux and systemconfig- * under Fedora) and other tools that are useful to the system administrator. The section on setting up a
server discusses the xinetd superserver which starts daemons as needed and can be used to help secure a system
by controlling who can use which services. You can also use TCP wrappers to control who can use which system
services by editing the hosts.allow and hosts.deny files in the /etc directory. By limiting the portion of the
filesystem a user sees, setting up a chroot jail can help control the damage a malicious user can do.
The section on DHCP describes how to set up a DHCP server so you do not have to configure each system on a
network manually. It details setting up both static and dynamic IP addresses using DHCP. Whether a system uses
as the source of certain information NIS, DNS, local files, or a combination, and in what order, is determined by
/etc/nsswitch.conf. The final section of this chapter discusses Linux-PAM, which allows you to maintain finegrain control over who can access the system, how they can access it, and what they can do.
Exercises
1. How does single-user mode differ from multiuser mode?
2. How would you communicate each of the following messages?
a.
The system is coming down tomorrow at 6:00 in the evening for periodic maintenance.
b.
The system is coming down in 5 minutes.
c.
Jenny’s jobs are slowing the system down drastically, and she should postpone them.
d.
Alex’s wife just had a baby girl.
3. What do the letters of the su command stand for? (Hint: It is not Superuser.) What can you do with su besides
give yourself Superuser privileges? How would you log in as Alex if you did not know his password but knew
the root password? How would you establish the same environment that Alex has when he first logs on?
4. How would you allow a user to execute privileged commands without giving the user the Superuser password?
5. Assume you are working as Superuser. How do you kill process 1648? How do you kill all processes running
kmail?
6. How can you disable SELinux?
Put the following line in /etc/sysconfig/selinux:
SELINUX=disabled
7. Develop a strategy for coming up with a password that an intruder would not be likely to guess but that you will
be able to remember.
93
Draft
Advanced Exercises
Draft
Advanced Exercises
8. Give the command
$ /sbin/fuser -uv /
What is this a list of? Why is it so long? Give the same command as root (or ask the system administrator to
do so and mail you the results). How does this list differ from the first? Why is it different?
9. When it puts files in a lost+found directory, fsck has lost the directory information for the files and thus has
lost the names of the files. Each file is given a new name, which is the same as the inode number for the file:
$ ls –lg lost+found
–rw–r--r-- 1 alex pubs 110 Jun 10 10:55 51262
What can you do to identify these files and restore them?
10. Take a look at /usr/bin/lesspipe.sh, and explain what it does and six ways it works.
11. Why are setuid shell scripts inherently unsafe?
12. When a user logs in, you would like the system to check for a username in the local /etc/passwd file first and
then check NIS. How do you implement this strategy?
13. Some older kernels contain a vulnerability that allows a local user to gain root privileges. Explain how this kind
of vulnerability negates the value of a chroot jail.
94
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 16 - Administration Tasks
Administration Tasks
IN THIS CHAPTER
Configuring User and Group Accounts
Backing Up Files
System Reports
Keeping Users Informed
Solving Problems
Speeding Up the System
Keeping the System Secure
logrotate: Manages Log Files
Disk Quota System
syslogd: Logs System Messages
The system administrator has many responsibilities. This chapter discusses tasks not covered in Chapter 11, including
configuring user and group accounts, backing up files, scheduling tasks, general problem solving, and using the
system log daemon, syslogd.
Configuring User and Group Accounts
More than a login name is required for a user to be able to log in and use the system. A user should have the necessary
files, directories, permissions, and usually a password in order to log in. Minimally, a user must have an entry in
the /etc/passwd and /etc/shadow files and a home directory. The following sections describe several ways you
can work with user accounts. Refer to page 354 and the NIS-HOWTO when you want to run NIS to manage the
passwd database.
system-config-users Manages User Accounts
The system-config-users (FEDORA) and redhat-config-users (RHEL) utilities display the User Manager window and
enable you to add, delete, and modify system users and groups. The User Manager Window has two tabs: Users
and Groups, each tab displaying information appropriate to its name. Figure 16-1 shows the Users tab.
95
Draft
system-config-users Manages User Accounts
Draft
Figure 16-1. The User Manager window
Search filter . The Search filter, located just below the toolbar, selects users or groups whose names match the
string, which can include wildcards that you enter in the Search filter text box. The string matches the beginning
of a name. For example, *nob matches nobody and nfsnobody, whereas nob matches only nobody. After you
enter the string, click Apply filter or press RETURN. If you have only a few users, you will not need to use the
Search filter.
Adding a user . To create a new user, click the Add User button on the toolbar. The User Manager displays
the Create New User window, which holds the same information as the User Data tab of the User Properties window
(Figure 16-2). Enter the information for the new user and click OK. Refer to page 379 for information on SELinux.
Once you create a user, you can modify the user to add/change/remove information.
96
Draft
useradd: Adds a User Account
Draft
Figure 16-2. The User Properties window, User Data tab
Modifying a user . To modify a user, highlight the user in the User Manager window and click Properties on
the toolbar; the utility displays the User Properties window (Figure 16-2). The User Properties window has four
tabs: User Data, Account Info, Password Info, and Groups. The User Data tab holds basic user information such
as name and password. The Account Info tab allows you to specify an expiration date for the account and to lock
the account so the user cannot log in. The Password Info tab allows you to turn on password expiration and specify
various related parameters. In the Groups tab, you can specify the groups that the user is a member of.
Working with groups . Click the Groups tab in the User Manager window to work with groups. To create a
group, click Add Group on the toolbar and specify the name of the group. To change the name of a group or add
or remove users from a group, highlight the group and click Properties on the toolbar. Click the appropriate tab,
make the changes you want, and click OK.
Help .
The User Manager has extensive help: Click Help on the toolbar.
When you are done working with users and groups, close the window.
useradd: Adds a User Account
The useradd utility (and the link to it, named adduser) adds new user accounts to the system. By default, useradd
assigns the next highest unused user ID to a new account and specifies the bash as the user’s login shell. The following example creates the user’s home directory (in /home), specifies the user’s group ID, and puts the user’s full
name in the comment field:
# useradd -g 500 -c "Alex Watson" alex
Based on the /etc/login.defs file, the system creates a home directory for the new user. When useradd creates a
home directory, it copies the contents of /etc/skel, which contains bash and other startup files, to that directory.
For more information on adding and modifying user information, see the useradd and usermod man pages. Once
you have added a user, use passwd to give the user a password.
97
Draft
userdel: Removes a User Account
Draft
userdel: Removes a User Account
If appropriate, make a backup copy of the files belonging to the user before deleting them. The userdel utility deletes
user accounts. The following command removes alex’s account, his home directory, and all his files:
# userdel -r alex
To turn off a user’s account temporarily, you can use usermod to change the expiration date for the account. Because
it specifies that his account expired in the past (December 31, 2003), the following command line prevents alex
from logging in:
# usermod -e "12/31/03" alex
groupadd: Adds a Group
Just as useradd adds a new user to the system, groupadd adds a new group by adding an entry to /etc/group. The
following example creates a new group, named rtfm:
# groupadd -g 1024 rtfm
Unless you use the –g option to assign a group ID, the system picks the next available sequential number greater
than 500. The –o option allows the group ID to be nonunique if you want to have multiple names for the same
group ID.
The analogue of userdel for groups is groupdel, which takes a group name as an argument. You can also use
groupmod to change the name or group ID of a group, as in the following examples:
# groupmod -g 1025 rtfm
# groupmod -n manuals rtfm
The first example gives the previously created rtfm group a new group ID number. The second example renames
the rtfm group manuals.
Caution: Group ID Cautions
The groupmod utility does not change group numbers in /etc/passwd when you renumber a group.
You must edit /etc/passwd and change the entries yourself. If you change the number of a group,
files that belonged to the group will no longer belong to the group. They may belong to no group or
to another group with the old group ID number.
Backing Up Files
One of the most neglected tasks of system administration is making backup copies of files on a regular basis. The
backup copies are vital in three instances: when the system malfunctions and files are lost, when a catastrophic
disaster (fire, earthquake, and so on) occurs, and when a user or the system administrator deletes or corrupts a file
by accident. Even when you set up RAID (page 31), you still need to back up files. Although RAID is useful for
fault tolerance (disk failure), it does not help in a catastrophic disaster or when a file is accidentally removed or
corrupted. It is a good idea to have a written backup policy and to keep copies of backups offsite (in another
building, at home, or at a completely different facility or campus) in a fireproof vault or safe.
98
Draft
Choosing a Backup Medium
Draft
The time to start thinking about backups is when you partition the disk. Refer to “Partitioning a Disk” on page 29.
Make sure the capacity of the backup device and your partition sizes are in line. Although you can back up a partition
onto multiple volumes, it is easier not to and much easier to restore from a single volume.
You must back up filesystems on a regular basis. Backup files are usually kept on magnetic tape or other removable
media. Exactly how often you should back up which files depends on your system and needs. The criterion is: If
the system crashes, how much work are you willing to lose? Ideally, you would back up all the files on the system
every few minutes so you would never lose more than a few minutes of work.
The tradeoff is: How often are you willing to back up the files? The backup procedure typically slows down the
system for other users, takes a certain amount of your time, and requires that you have and store the media (tape
or disk) you keep the backup on. Avoid backing up an active filesystem; the results may be inconsistent, and
restoring from the backup may be impossible. This requirement is a function of the backup program and the
filesystem you are backing up.
Another question is when to run the backup. Unless you kick the users off and bring the system down to singleuser mode (not a very user-friendly practice), you want to do it when the machine is at its quietest. Depending on
the use of the system, sometime in the middle of the night can work well. Then the backup is least likely to impact
the users, and the files are not likely to change as they are being read for backup.
A full backup makes copies of all files, regardless of when they were created or accessed. An incremental backup
makes copies of the files that have been created or modified since the last (usually full) backup.
The more people using the machine, the more often you should back up the filesystems. A common schedule might
have you perform an incremental backup one or two times a day and a full backup one or two times a week.
Choosing a Backup Medium
Traditionally, personal computers used floppy diskettes for performing backups. However, the large, hard disks
now available for computers makes this impractical. If you have a ninety gigabyte disk on your system, you would
need more than 72,000 floppy diskettes to do a full backup. Even if files are compressed as you back them up, the
number of diskettes required would be unmanageable. If your computer is connected to a network, you can write
your backups to a tape drive on another system. This is often done with networked computers to avoid the cost of
having a tape drive on each computer in the network and to simplify management of doing backups for many
computers in a network. Most likely you want to use a tape system for backing up your computer. Because tape
drives to hold many gigabytes of data are available, using tape simplifies the task of backing up the system, making
it more likely that you regularly do this important task. Other options for holding backups are writable CDs, DVDs,
and removable hard disks. These devices, although not as cost-effective or able to store as much information as
tape systems, offer convenience and improved performance over using tapes.
Backup Utilities
A number of utilities help you back up the system, and most work with any media. Most Linux backup utilities
are based on one of the archive programs—tar or cpio—and augment these basic programs with bookkeeping
support for managing backups conveniently.
You can use any one of the tar, cpio, or dump/restore utilities to construct full or partial backups of the system.
Each utility constructs a large file that contains, or archives, other files. In addition to file contents, an archive includes
header information for each file it holds. This header information can be used when extracting files from the archive
to restore file permissions and modification dates. An archive file can be saved to disk, written directly to tape, or
shipped across the network while it is being created.
99
Draft
Backup Utilities
Draft
In addition to helping you back up your system, these programs are convenient for bundling files for distribution
to other sites. The tar program is often used for this purpose, and some software packages available on the Internet
are bundled as tar archive files.
The amanda (Advanced Maryland Automatic Network Disk Archiver—www.amanda.org [http://www.amanda.org])
utility, one of the more popular backup systems, uses dump or tar and takes advantage of Samba to back up Windows
systems. The amanda utility backs up a LAN of heterogeneous hosts to a single tape drive. It is available under Red
Hat; refer to the amanda man page for details.
tar: Archives Files
The tar (tape archive) utility stores and retrieves files from an archive and can compress the archive to conserve
space. You can specify an archive device with the –f option. If you do not specify an archive device, tar uses
/dev/rmt0 (which may not exist on the local system). With the –f option, tar uses the argument to –f as the name
of the archive device. You can use this option to refer to a device on another system on the network. Although a
lot of options are available with tar, you need only a few in most situations. The following command displays a
complete list of options:
# tar --help | less
Most options for tar can be given either in a short form (a single letter) or as a descriptive word. Descriptive-word
options are preceded by two dashes, as in --help. Single-letter options can be combined into a single command
line argument and do not need to be preceded by a dash (for consistency with other utilities, it is good practice to
use the dash anyway).
Although the following two commands look quite different, they specify the same tar options in the same order.
The first version combines single-letter options into a single command line argument; the second version uses descriptive words for the same options:
# tar –ztvf /dev/st0
# tar --gzip --list --verbose --file /dev/st0
Both commands tell tar to generate a (v, verbose) table of contents (t, list) from the tape on /dev/st0 (f, file), using
gzip (z, gzip) to decompress the files. Unlike the original UNIX tar, the GNU version strips the leading / from absolute pathnames.
The options in Table 16-1 tell the tar program what you want it to do. You must include exactly one of these options
whenever you use tar.
100
Draft
Performing a Simple Backup
Draft
Table 16-1. The tar Utility
Option
Effect
--append (–r)
Appends files to an archive.
--catenate (–A)
Adds one or more archives to the end of an existing
archive.
--create (–c)
Creates a new archive.
--delete
Deletes files in an archive (not on tapes).
-- dereference (–h)
Follows symbolic links.
--diff (–d)
Compares files in an archive with disk files.
--extract (–x)
Extracts files from an archive.
--help
Displays a help list of tar options.
--list (–t)
Lists the files in an archive.
--update (–u)
Like the –r option, but the file is not appended if a newer
version is already in the archive.
The –c, –t, and –x options are used most frequently. You can use many other options to change how tar operates.
The –j option compresses/decompresses the file by filtering it through bzip2 (page 133).
cpio: Archives Files
The cpio (copy in/out) program is similar to tar but can use archive files in a variety of formats, including the one
used by tar. Normally, cpio reads the names of the files to insert into the archive from standard input and produces
the archive file as standard output. When extracting files from an archive, cpio reads the archive as standard input.
As with tar, some options can be given in both a short, single-letter form and a more descriptive word form.
However, unlike tar, the syntax of the two forms differs when the option must be followed by additional information.
In the short form, you must, use a SPACE between the option and the additional information; with the word form,
you must separate the two with an equal sign and no SPACEs.
Running cpio with --help displays a full list of options, although not as nice a list as tar.
Performing a Simple Backup
When you prepare to make a major change to a system, such as replacing a disk drive or updating the Linux kernel,
it is a good idea to archive some or all of the files so you can restore any that are damaged if something goes wrong.
For this type of backup, tar or cpio works well. For example, if you have a SCSI tape drive as device /dev/st0 that
is capable of holding all your files on a single tape, you can use the following commands to construct a backup
tape of the entire system:
# cd /
# tar –cf /dev/st0 .
This command creates an archive (c) on the device /dev/st0 (f). All the commands in this section start by using cd
to change to the root directory so you are sure to back up the entire system. If you would like to compress the
archive, replace the preceding tar command with the following command, which uses j to call bzip2:
# tar –cjf /dev/st0 .
101
Draft
dump, restore: Back Up and Restore
Filesystems
Draft
You can back up your system with a combination of find and cpio. The following commands create an output file
and set the I/O block size to 5120 bytes (the default is 512 bytes):
# cd /
# find . –depth | cpio –oB > /dev/st0
The next command restores all the files in the /home directory from the preceding backup. The options extract
files from an archive (–i) in verbose mode, keeping the modification times and creating directories as needed.
# cd /
# cpio –ivmd /home/\* < /dev/st0
Tip: Exclude Some Directories from a Backup
In practice, you exclude some directories from the backup process. For example, not backing up /tmp
or /var/tmp (or its link, /usr/tmp) can save room in the archive. Also, do not back up the files in
/proc. Because the /proc filesystem is not a disk filesystem but rather a way for the Linux kernel to
provide you with information about the operating system and system memory, you need not back up
/proc; you cannot restore it later. You do not need to back up filesystems that are mounted from disks
on other computers in your network. Do not back up FIFOs; the results are unpredictable. If you plan
on using a simple method, similar to those just shown, create a file naming the directories to exclude
from the backup, and use the appropriate option with the archive program to read the file.
Although any of the archive programs works well for such simple backups, only amanda provides a sophisticated
backup and restore system. For example, to determine whether a file is in an archive requires you to read the entire
archive. If the archive is split across several tapes, this is particularly tiresome. More sophisticated utilities, including
amanda, assist you in several ways, including keeping a table of contents of the files in a backup.
dump, restore: Back Up and Restore Filesystems
The dump utility, first seen in UNIX version 6, backs up an entire filesystem, or only those files that have changed
since the last dump. The restore utility restores an entire filesystem, an individual file, or a directory hierarchy. You
will get the best results if you perform a backup on a quiescent system so that the files are not changing as you
make the backup.
The following command performs a complete backup of all files (including directories and special files) on the root
(/) partition onto SCSI tape 0. Frequently, there is a link to the active tape drive, named /dev/tape, which you can
use in place of the actual entry in the /dev directory.
# dump -0uf /dev/st0 /
The option specifies that the whole filesystem is to be backed up (a full backup). There are ten dump levels: 0–9.
Zero is the highest (most complete) level and always backs up the entire filesystem. Each additional level is incremental with respect to the level above it. For example, 1 is incremental to 0 and backs up only files that have
changed since the last level 0 dump. Level 2 is incremental to 1 and backs up only files that have changed since the
last level 1 dump, and so on. You can construct a very flexible schedule by using this scheme. Also, you do not
need to use sequential numbers for backup levels. You can perform a level 0 dump, followed by level 2 and 5 dumps.
The u option updates the /etc/dumpdates file (page 427) with filesystem, date, and dump level information for
use by the next incremental dump. The f option and its argument (/dev/st0) write the backup to the file named
/dev/st0.
102
Draft
dump, restore: Back Up and Restore
Filesystems
Draft
The following command makes a partial backup containing all the files that have changed since the last level 0
dump. The first argument is a 1, specifying a level 1 dump.
# dump -1uf /dev/st0 /
To restore an entire filesystem from a tape, first restore the most recent complete (level 0) backup. Do this carefully
because restore can overwrite the existing filesystem. When you are logged in as Superuser, cd to the directory the
filesystem is mounted on, and give the following command:
# restore -if /dev/st0
The i option invokes an interactive mode that allows you to choose which files and directories you would like to
restore. As with dump, the f option specifies the name of the device that the backup tape is mounted on. When
restore finishes, load the next lower-level (higher number) dump tape and issue the same restore command. If you
have multiple incremental dumps at a particular level, always restore with the most recent one. You do not need
to invoke restore with any special arguments to restore an incremental dump; it will restore whatever is on the tape.
You can also use restore to extract individual files from a tape using the x option and specifying the filenames on
the command line. Whenever you restore a file, the restored file will be in your working directory. Before restoring
files, make sure you are working in the right directory. The following commands restore the etc/xinetd.conf file
from the tape in /dev/st0. The filename of the dumped file does not begin with a / because all dumped pathnames
are relative to the filesystem that you dumped—in this case /. Because the restore command is given from the /
directory, the file will be restored to its original location: /etc/xinetd.conf:
# cd /
# restore -xf /dev/st0 etc/xinetd.conf
If you use the x option without specifying a file or directory name to extract, the entire dumped filesystem is extracted.
Use the r option to restore an entire filesystem without using the interactive interface. The following command
restores the filesystem from the tape on /dev/st0 into the working directory without interaction:
# restore -rf /dev/st0
You can also use dump and restore to access a tape drive on another system. Specify the file/directory as host:file,
where host is the hostname of the system the tape drive is on and file is the file/directory you want to dump/restore.
Occasionally, restore may prompt you with
You have not read any volumes yet.
Unless you know which volume your file(s) are on you should start
with the last volume and work towards the first.
Specify next volume #:
Enter 1 (one) in response to this prompt. If the filesystem spans more than one tape or disk, this prompt allows
you to switch tapes.
At the end of the dump, you will receive another prompt:
set owner/mode for '.'? [yn]
103
Draft
Scheduling Tasks
Draft
Answer y to this prompt when restoring entire filesystems or files that have been accidentally removed. Doing so
will restore the appropriate permissions to the files and directories being restored. Answer n if you are restoring a
dump to a directory other than the one it was dumped from; the working directory permissions and owner will be
set to those of the person doing the restore (typically root).
Various device names can access the /dev/st0 device. Each name accesses a different minor device number that
controls some aspect of how the tape drive is used. After you complete a dump when you use /dev/st0, the tape
drive automatically rewinds the tape to the beginning. Use the nonrewinding SCSI tape device (/dev/nst0) to keep
the tape from rewinding on completion. This feature allows you to back up multiple filesystems to one volume.
Following is an example of backing up a system where the /home, /usr, and /var directories are on different
filesystems:
# dump -0uf /dev/nst0 /home
# dump -0uf /dev/nst0 /usr
# dump -0uf /dev/st0 /var
The preceding example uses the nonrewinding device for the first two dumps. If you use the rewinding device, the
tape rewinds after each dump, and you are left with only the last dump on the tape. For more information, refer
to the mt (magnetic tape) man page.
You can use mt to manipulate files on a multivolume dump tape. The following mt command positions the tape
(fsf 2 instructs mt to skip forward past two files, leaving the tape at the start of the third file). The restore command
restores the /var file-system from the previous example:
# mt -f /dev/st0 fsf 2
# restore rf /dev/st0
Scheduling Tasks
It is a good practice to schedule certain routine tasks to run automatically. For example, you may want to remove
old core files once a week, summarize accounting data daily, and rotate system log files monthly.
cron and crontab: Schedule Routine Tasks
Using crontab, you can submit a list of commands in a format that can be read and executed by cron. As Superuser,
you can put commands in one of the /etc/cron.* directories to be run at intervals specified by the directory name,
such as cron.daily.
Tip: cron Stops for No One; Try anacron
The cron utility assumes the system it is running on is always running. A similar utility, anacron, does
not make that assumption and is well suited to portable and home computers that are frequently
turned off. The anacron utility takes its instructions from the /etc/anacrontab file unless you specify
otherwise. Refer to the anacron and anacrontab man pages for more information.
at: Runs Occasional Tasks
Like the cron utility, at allows you to run a job sometime in the future. Unlike cron, at runs a job only once. For
instance, you can schedule an at job that will reboot the system at 3 A.M. (when all users are logged off):
104
Draft
kcron: Schedules Tasks
Draft
# at 3am
at> reboot
at> CONTROL-D <EOT>
job 1 at 2004-02-01 03:00
It is also possible to run an at job from within an at job. For instance, you could have an at job that would check
for new patches every 18 days, something that would be more difficult with cron.
kcron: Schedules Tasks
The kcron utility provides an easy-to-use GUI to cron, allowing you to create and modify crontab files. Scheduling
tasks with kcron is a matter of clicking buttons (Figure 16-3).
Figure 16-3. The kcron Task Scheduler
105
Draft
System Reports
Draft
Run kcron when you are logged in as yourself to view and modify your personal crontab file. When you run kcron
as root, you can modify any crontab file on the system. To start, kcron displays a window that lists Users (when
you are running as root), Tasks, and Variables. The Description column of this window is very wide and does not
fit in the window. Use the right-left scroll bar to view its contents. If you are running as root, you need to doubleclick a user to display the Tasks folder. To create a new crontab entry, highlight Tasks, and select New from
Edit on the menubar (or from the right-click menu). To modify an entry, highlight the entry, and select Modify
from Edit on the menubar. From the resulting window, enter the name of the program you want to run in the
Program text box, and depress buttons or place check marks corresponding to the dates and times you want to run
the program. Unless you redirect it, output from the program that kcron runs is mailed to you.
System Reports
Many utilities report on one thing or another. The who, finger, ls, ps, and other utilities generate simple end user
reports. In some cases, these reports can help you with system administration. This section describes utilities that
generate more in-depth reports that can usually be of more help with system administration tasks. Linux has many
other report utilities, including sar (system activity report), iostat (input/output and CPU statistics), netstat (network
report), mpstat (processor statistics), and nfsstat (NFS statistics).
vmstat: Reports Virtual Memory Statistics
The vmstat utility generates virtual memory information along with (limited) disk and CPU activity data. The following example shows virtual memory statistics in 3-second intervals for seven iterations (from the arguments 3
7). The first line covers the time since the system was last booted; the rest of the lines cover the period since the
previous line:
$ vmstat 3 7
procs -----------memory---------- ---swap-- -----io---r b swpd
free
buff
cache
si
so
bi
bo
0 2
0 684328 33924 219916
0
0 430
105
0 2
0 654632 34160 248840
0
0 4897
7683
0 3
0 623528 34224 279080
0
0 5056
8237
0 2
0 603176 34576 298936
0
0 3416
141
0 2
0 575912 34792 325616
0
0 4516
7267
1 2
0 549032 35164 351464
0
0 4429
77
0 2
0 523432 35448 376376
0
0 4173
6577
--system-- ----cpu---in
cs us sy id wa
1052
134 2 4 86 8
1142
237 0 5 0 95
1094
178 0 4 0 95
1161
255 0 4 0 96
1147
231 0 4 0 96
1120
210 0 4 0 96
1135
234 0 4 0 95
The following list explains the column heads displayed by vmstat.
•
•
106
procs process information
•
r
number of waiting, runnable processes
•
b
number of blocked processes (in uninterruptable sleep)
memory memory information in kilobytes
•
swpd
used virtual memory
•
free
idle memory
•
buff
memory used as buffers
•
cache
memory used as cache
Draft
•
•
•
•
top: Lists Processes Using the Most
Resources
Draft
swap system paging activity in kilobytes per second
•
si
memory swapped in from disk
•
so
memory swapped out to disk
io system I/O activity in blocks per second
•
bi
blocks received from a block device
•
bo
blocks sent to a block device
system values are per second
•
in
interrupts (including the clock)
•
cs
context switches
cpu percentage of total CPU time spent in each of these states
•
us
user (nonkernel)
•
sy
system (kernel)
•
id
idle (RHEL includes I/O wait time)
•
wa
waiting for I/O (RHEL shows zeroes)
top: Lists Processes Using the Most Resources
The top utility is a useful supplement to ps. At its simplest, top displays system information at the top and the most
CPU-intensive processes below the system information. The top utility updates itself periodically; type q to quit.
Although you can use command line options, the interactive commands are often more useful. Refer to Table 162 and to the top man page for more information.
$ top
top - 21:30:26 up 18 min, 2 users, load average: 0.95, 0.30, 0.14
Tasks: 63 total,
4 running, 58 sleeping,
1 stopped,
0 zombie
Cpu(s): 30.9% us, 22.9% sy, 0.0% ni, 0.0% id, 45.2% wa, 1.0% hi, 0.0%si
Mem:
1036820k total, 1032276k used,
4544k free,
40908k buffers
Swap: 2048276k total,
0k used, 2048276k free,
846744k cached
PID
1285
1276
7
6
8
300
1064
1224
1275
USER
root
root
root
root
root
root
mgs2
root
mgs2
PR
25
18
15
15
15
15
16
16
16
NI
0
0
0
0
0
0
0
0
0
VIRT
9272
3048
0
0
0
0
8144
4964
2840
RES
6892
860
0
0
0
0
2276
1360
936
SHR
1312
1372
0
0
0
0
6808
3944
1784
S
R
R
S
S
S
S
S
S
R
%CPU
29.3
3.7
0.7
0.3
0.3
0.3
0.3
0.3
0.3
%MEM
0.7
0.1
0.0
0.0
0.0
0.0
0.2
0.1
0.1
TIME+
0:00.88
0:05.25
0:00.27
0:00.11
0:00.06
0:00.24
0:00.69
0:00.03
0:00.15
COMMAND
bzip2
cp
pdflush
pdflush
kswapd0
kjournald
sshd
bash
top
107
Draft
Keeping Users Informed
1284
1
root
root
15
16
0
0
2736
2624
668
520
1416
1312
S
S
0.3
0.0
Draft
0.1
0.1
0:00.01
0:06.51
tar
init
Table 16-2. top: Interactive Commands
A
Sorts processes by age (newest first).
h or ?
Displays a help screen.
k
Prompts for a PID number and type of signal and sends
the process that signal. Defaults to signal 15 (SIGTERM);
specify 9 (SIGKILL) only when 15 does not work.
M
Sorts processes by memory usage.
P
Sorts processes by CPU usage (default).
q
Quits.
s
Prompts for time between updates in seconds. Use 0 for
continuous updates.
SPACE
Updates display immediately.
T
Sorts tasks by time.
W
Writes a startup file named ~/.toprc so that next time
you start top, it uses the same parameters it is currently
using.
Keeping Users Informed
One of your primary responsibilities as system administrator is communicating with the system users. You need
to make announcements, such as when the system will be down for maintenance, when a class on some new software
will be held, and how users can access the new system printer. You can even start to fill the role of a small local
newspaper, letting users know about new employees, RIFs, births, the company picnic, and so on.
Different communications have different priorities. Information about the company picnic in two months is not
as time sensitive as the fact that you are bringing the system down in 5 minutes. To meet these differing needs,
Linux provides different ways of communicating. The most common methods are described and contrasted in the
following list. All these methods are generally available to everyone, except for the message of the day, which is
typically reserved for Superuser.
write . Use write to communicate with a user who is logged in on the local system. You might use it to ask a
user to stop running a program that is bogging down the system. The user might reply that he will be done in 3
minutes. Users can also use write to ask the system administrator to mount a tape or restore a file.
talk . The talk utility performs the same function as write but is more advanced. Although talk uses a characterbased interface, it has a graphical appearance, showing what each user is typing as it is being typed. Unlike write,
you can use talk to have a discussion with someone on another machine on the network.
wall . The wall (write all) utility effectively communicates immediately with all users who are logged in. It works
similarly to write, except users cannot use wall to write back to only you. Use wall when you are about to bring the
system down or are in another crisis situation. Users who are not logged in do not get the message.
Use wall while you are Superuser only in crisis situations; it interrupts anything anyone is doing.
108
Draft
Creating Problems
Draft
email . Email is useful for communicating less urgent information to one or more system and/or remote users.
When you send mail, you have to be willing to wait for each user to read it. The email utilities are useful for reminding
users that they are forgetting to log out, bills are past due, or they are using too much disk space.
Users can easily make permanent records of messages they receive via email, as opposed to messages received via
write or talk, so they can keep track of important details. It would be appropriate to use email to inform users about
a new, complex procedure, so each user could keep a copy of the information for reference.
message of the day . Users see the message of the day each time they log in in a textual environment. You can
edit the /etc/motd file to change the message. The message of the day can alert users to upcoming periodic maintenance, new system features, or a change in procedures.
Creating Problems
Even experienced system administrators make mistakes; new system administrators make more mistakes. Even
though you can improve your odds by carefully reading and following the documentation provided with your
software, many things can still go wrong. A comprehensive list is not possible, no matter how long, as new and
exciting ways to create problems are discovered every day. A few of the more common techniques are described
here.
Failing to Perform Regular Backups
Few feelings are more painful to a system administrator than realizing that important information is lost forever.
If your system supports multiple users, having a recent backup may be your only protection from a public lynching.
If it is a single-user system, having a recent backup certainly keeps you happier when you lose a hard disk.
Not Reading and Following Instructions
Software developers provide documentation for a reason. Even when you have installed a software package before,
you should carefully read the instructions again. They may have changed, or you may simply remember them incorrectly. Software changes more quickly than books are revised, so no book should be taken as offering foolproof
advice; look for the latest documentation online.
Failing to Ask for Help When Instructions Are Not Clear
If something does not seem to make sense, try to find out what does make sense; do not guess. Refer to “Help” on
page 913.
Deleting or Mistyping a Critical File
One sure way to give yourself nightmares is to execute the command
# rm –rf /etc
ü do not do this
Perhaps no other command renders a Linux system useless so quickly. The only recourse is to reboot into rescue
mode (page 377) using the first installation CD and restore the missing files from a recent backup. Although this
example is extreme, many files are critical to proper operation of a system. Deleting one of these files or mistyping
information in one of them is almost certain to cause problems. If you directly edit /etc/passwd, for example, entering the wrong information in a field can make it impossible for one or more users to log in. Do not use rm –rf
109
Draft
Solving Problems
Draft
with an argument that includes wildcard characters; do pause after typing the command, and read it before you
press RETURN. Check everything you do carefully, and make a copy of a critical file before you edit it.
Solving Problems
As the system administrator, it is your responsibility to keep the system secure and running smoothly. When a user
is having a problem, it usually falls to the administrator to help the user get back on track. This section suggests
ways to keep users happy and the system functioning at its peak.
Helping When a User Cannot Log In
When a user has trouble logging in on the system, the problem may be a user error or a problem with the system
software or hardware. The following steps can help you determine where the problem is:
•
Determine if only that one user or only that one user’s terminal/ workstation has a problem or if the problem
is more widespread.
•
Check that the user’s Caps Lock key is not on.
•
Make sure the user’s home directory exists and corresponds to that user’s entry in the /etc/passwd file. Verify
that the user owns his or her home directory and startup files and that they are readable (and, in the case of the
home directory, executable). Confirm that the entry for the user’s login shell in the /etc/passwd file is valid
(that is, that the entry is accurate and that the shell exists as specified).
•
Change the user’s password if there is a chance that he or she has forgotten the correct password.
•
Check the user’s startup files (.profile, .login, .bashrc, and so on). The user may have edited one of these files
and introduced a syntax error that prevents login.
•
Check the terminal or monitor data cable from where it plugs into the terminal to where it plugs into the computer
(or as far as you can follow it). Finally, try turning the terminal or monitor off and then turning it back on.
•
When the problem appears to be widespread, check if you can log in from the system console. If you can, make
sure that the system is in multiuser mode. If you cannot log in, the system may have crashed; reboot it and
perform any necessary recovery steps (the system usually does quite a bit automatically).
•
Check the /etc/inittab file to see that it is starting the appropriate login service (usually some form of getty,
such as mingetty).
•
Check the /var/log/messages file. This file accumulates system errors, messages from daemon processes, and
other important information. It may indicate the cause or more symptoms of a problem. Also, check the system
console. Occasionally messages about system problems that do not get written to /var/log/messages (for instance, if the disk is full) get displayed on the console.
•
If the user is logging in over a network connection, use “system-config-services: Configures Services II” on page
383 to make sure that the service the user is trying to use (such as telnet or ssh) is enabled.
•
Use df to check for full filesystems. Sometimes, if the /tmp filesystem or the user’s home directory is full, login
fails in unexpected ways. In some cases you may be able to log in to a textual environment but not a graphical
one. When applications that start when the user logs in cannot create temporary files or cannot update files in
the user’s home directory, the login process itself may terminate.
110
Draft
Speeding Up the System
Draft
Speeding Up the System
When the system is running slowly for no apparent reason, perhaps a process did not exit when a user logged out.
Symptoms include poor response time and a system load, as shown by w or uptime, that is greater than 1.0. Use
ps –ef to list all processes. The top utility is excellent for quickly finding rogue processes. One thing to look for in
ps –ef output is a large number in the TIME field. For example, if you find a Netscape process that has a TIME
field over 100.0, this process has likely run amok. However, if the user is doing a lot of Java work and has not
logged out for a long time, this value may be normal. Look at the STIME field to see when the process was started.
If the process has been running for longer than the user has been logged in, it is a good candidate to be killed.
When a user gets stuck and leaves his or her terminal unattended without notifying anyone, it is convenient to kill
(page 375) all processes owned by that user. If the user is running a window system, such as GNOME or KDE on
the console, kill the window manager process. Manager processes to look for include startkde, gnome-session,
or another process name that ends in wm. Usually the window manager is either the first or the last thing to be
run, and exiting from the window manager logs the user out. If killing the window manager does not work, try
killing the X server process itself. This process is typically listed as /etc/X11/X. If that fails, you can kill all processes
owned by a user by running kill –1 –1, or equivalently kill –TERM –1 as the user. Using –1 (one) in place of the
process ID tells kill that it should send the signal to all processes that are owned by that user. For example, as root
you could type
# su jenny -c 'kill -TERM -1'
If this does not kill all processes (sometimes TERM does not kill a process), you can use the KILL signal. The following line will definitely kill all processes owned by Jenny and will not be friendly about it:
# su jenny -c 'kill -KILL -1'
(If you do not use su jenny –c, the same command brings the system down.)
lsof: Finds Open Files
The name lsof is short for ls open files; this utility locates open files. Its options let you look only at certain processes,
look only at certain file descriptors of a process, or show certain network connections (network connections use
file descriptors just as normal files do and lsof can show those as well). Once you have identified a suspect process
using ps –ef, run the following command:
# lsof -sp pid
Replace pid with the process ID of the suspect process; lsof displays a list of all file descriptors that process pid has
open. The –s option displays the size of all open files. The size information may be helpful in determining whether
the process has a very large file open. If it does, contact the owner of the process or, if necessary, kill the process.
The –rn option redisplays the output of lsof every n seconds.
Keeping a Machine Log
A machine log that includes the information shown in Table 16-3 can help you find and fix system problems. Note
the time and date for each entry in the log. Avoid the temptation to keep the log only on the computer because it
will be most useful to you at times when the machine is down. Another good idea is to keep a record of all email
about user problems. One way to do this is to save all this mail to a separate file or folder as you read it. Another
way is to set up a special mail alias that users send mail to when they have problems. This alias can then forward
111
Draft
Keeping the System Secure
Draft
mail to you and also store a copy in an archive file. Following is an example of an entry in the /etc/aliases file (page
614) that sets up this type of alias:
trouble: admin,/var/mail/admin.archive
Table 16-3. Machine Log
Hardware modifications
Keep track of the system hardware configuration: which
devices hold which partitions, the model of the new NIC
you added, and so on.
System software modifications
Keep track of the options used when building Linux.
Print such files as /usr/src/linux*/.config (Linux kernel
configuration), /etc/modules.conf (RHEL), and the X11
configuration file /etc/X11/XF86Config (RHEL) or
/etc/X11/xorg.conf (FEDORA). The file hierarchy under
/etc/sysconfig contains valuable information about
network configuration and so on.
Hardware malfunctions
Keep as accurate a list as possible of any problems with
the system. Make note of any error messages or numbers
that the system displays on the system console and what
users were doing when the problem occurred.
User complaints
Make a list of all reasonable complaints made by knowledgeable users (for example, “machine is abnormally
slow”).
Email sent to the trouble alias will be forwarded to the admin user and also stored in the file /var/mail/admin.archive.
Keeping the System Secure
No system with dial-in lines or public access to terminals is absolutely secure. You can make a system as secure as
possible by changing the Superuser password frequently and choosing passwords that are difficult to guess. Do
not tell anyone who does not absolutely need to know the Superuser password. You can also encourage system
users to choose difficult passwords and to change them periodically.
By default, passwords on Red Hat Linux use MD5 (page 982) hashing, which makes them more difficult to break
than DES (page 924) encrypted passwords. It makes little difference how well encrypted your password is if you
make it easy for someone to find out or guess what it is.
A password that is difficult to guess is one that someone else would not be likely to think you would have chosen.
Do not use words from the dictionary (spelled forward or backward); names of relatives, pets, or friends; or words
from a foreign language. A good strategy is to choose a couple of short words, include some punctuation (for example,
put a ^ between them), mix the case, and replace a couple of the letters in the words with numbers. If it were not
printed in this book, an example of a good password would be C&yGram5 (candygrams). Ideally you would use
a random combination of ASCII characters, but that would be difficult to remember.
You can use one of several excellent password-cracking programs to find users who have chosen poor passwords.
These programs work by repeatedly encrypting words from dictionaries, phrases, names, and other sources. If the
encrypted password matches the output of the program, then the program has found the password of the user.
Two programs that crack passwords are crack and cops. These and many other security tips and programs are
available from CERT (www.cert.org [http://www.cert.org]), which was originally called the computer emergency
response team. Specifically look at www.cert.org/tech_tips [http://www.cert.org/tech_tips].
112
Draft
Log Files and Mail for root
Draft
Make sure that no one except Superuser can write to files containing programs that are owned by root and run in
setuid mode (for example, mail and su). Also make sure that users do not transfer programs that run in setuid mode
and are owned by root onto the system by means of mounting tapes or disks. These programs can be used to circumvent system security. One technique that prevents users from having setuid files is to use the –nosuid flag to
mount, which you can set in the flags section in the fstab file. Refer to “fstab: Keeps Track of Filesystems” on page
445.
The BIOS in many machines gives you some degree of protection from an unauthorized person modifying the BIOS
or rebooting the system. When you set up the BIOS, look for a section named Security. You can probably set up a
BIOS password. If you depend on the BIOS password, lock the computer case. It is usually a simple matter to reset
the BIOS password with access to a jumper on the motherboard.
Log Files and Mail for root
Users frequently email root and postmaster to communicate with the system administrator. If you do not forward
root’s mail to yourself (refer to “/etc/aliases” on page 614), remember to check root’s mail periodically. You will
not receive reminders about mail that arrives for root when you use su to perform system administration tasks.
However, after using su to become root, you can give the command mail –u root to look at root’s mail.
Look at the system log files regularly for evidence of problems. Two important files are /var/log/messages, where
the operating system and some applications record errors, and /var/log/maillog, which contains errors from the
mail system. You can use system-logviewer (FEDORA) or redhat-logviewer (RHEL) to view many of the system
logs.
The logwatch utility (/usr/sbin/logwatch points to the /etc/log.d/scripts/logwatch.pl Perl script) is a report
writer that sends email reports on log files. By default, the script is run daily (/etc/cron.daily/00-logwatch also
points to /etc/log.d/scripts/logwatch.pl) and emails its output to root. Refer to the logwatch man page and to
the script itself for more information.
Monitoring Disk Usage
Sooner or later, you will probably start to run out of disk space. Do not fill up a disk; Linux can write to files significantly faster if at least 5 to 30 percent of the disk space in a given filesystem is free. The result is that using more
than the maximum optimal disk space in a filesystem can degrade system performance.
Fragmentation . When the filesystem becomes full, it can become fragmented. This is similar to the DOS
concept of fragmentation but is not nearly as pronounced and is typically rare on modern Linux filesystems; by
design Linux filesystems are resistant to fragmentation. Keep filesystems from running near full capacity, and you
may never need to worry about fragmentation. If there is no space on a filesystem, you cannot write to it at all.
To check on fragmentation, you can unmount the filesystem and run fsck on it. As part of fsck execution, fragmentation is computed and displayed. You can defragment a filesystem by backing it up, using mkfs (page 397) to make
a clean, empty image, and then restoring the filesystem. The utility that you use to do your backup and restore is
irrelevant and completely up to you. You can use dump/restore, tar, cpio, or a third-party backup program.
Reports . Linux provides several programs that report on who is using how much disk space on what filesystems.
Refer to the du, quot, and df man pages and the –size option in the find utility man page. In addition to these utilities, you can use the disk quota system to manage disk space.
The main ways to increase the amount of free space on a filesystem are to compress files, delete files, grow filesystems,
and condense directories. This section contains some ideas on ways to maintain a filesystem so that it does not get
overloaded.
113
Draft
logrotate: Manages Log Files
Draft
Files that grow quickly . Some files, such as log files and temporary files, grow over time. Core dump files take
up space and are rarely needed. Also, users occasionally run programs that accidentally generate huge files. As the
system administrator, you must review these files periodically so that they do not get out of hand.
If a filesystem is running out of space quickly (that is, over a period of an hour rather than weeks or months), first
figure out why it is running out of space. Use a ps –ef command to determine whether a user has created a runaway
process that is creating a huge file. In evaluating the output of ps, look for a process that has used a large amount
of CPU time. If such a process is running and creating a large file, the file will continue to grow as you free up space.
If you remove the huge file, the space it occupied will not be freed until the process terminates, so you need to kill
the process. Try to contact the user running the process, and ask the user to kill it. If you cannot contact the user,
log in as root and kill the process. Refer to kill on page 375 for more information.
You can also truncate a large log file rather than removing it, although you can better deal with this recurring
situation with logrotate. For example, if the /var/log/messages file has become very large because a system daemon
is misconfigured, you can use /dev/null to truncate it:
# cp /dev/null /var/log/messages
or
# cat /dev/null > /var/log/messages
or, without spawning a new process,
# : > /var/log/messages
If you remove /var/log/messages, you have to restart the syslogd daemon. Without restarting syslogd, the space
on the filesystem is not released.
When no single process is consuming the disk space but it has instead been used up gradually, locate unneeded
files and delete them. You can archive them by using cpio, dump, or tar before you delete them. You can safely remove
most files named core that have not been accessed for several days. The following command line performs this
function without removing necessary files named core (such as /dev/core):
# find / -type f -name core | xargs file | grep 'B core file' | sed 's/:ELF.*//g' | xargs
rm -f
The find command lists all ordinary files named core and sends its output to xargs, which runs file on each of the
files in the list. The file utility displays a string that includes B core file for files created as the result of a core dump.
These files need to be removed. The grep command filters out from file lines that do not contain this string. Finally,
sed removes everything following the colon so that all that is left on the line is the pathname of the core file; xargs
removes the file.
Look through the /tmp and /var/tmp directories for old temporary files and remove them. Keep track of disk usage
in /var/mail, /var/spool, and /var/log.
logrotate: Manages Log Files
Rather than deleting or truncating log files, you may want to keep the contents around for a while in case you need
to refer to them. The logrotate utility helps you manage system log (and other) files automatically by rotating (page
994), compressing, mailing, and removing each as you specify. The logrotate utility is controlled by the
/etc/logrotate.conf file, which sets default values and can optionally specify files to be rotated. Typically, logrotate.conf has an include statement that points to utility-specific specification files in /etc/logrotate.d. Following
is the default logrotate.conf file:
114
Draft
logrotate: Manages Log Files
Draft
$ cat /etc/logrotate.conf
# see "man logrotate" for details
# rotate log files weekly
weekly
# keep 4 weeks worth of backlogs
rotate 4
# create new (empty) log files after rotating old ones
create
# uncomment this if you want your log files compressed
#compress
# RPM packages drop log rotation information into this directory
include /etc/logrotate.d
# no packages own wtmp -- we'll rotate them here
/var/log/wtmp {
monthly
create 0664 root utmp
rotate 1
}
# system-specific logs may be also be configured here.
The logrotate.conf file sets default values for common parameters. Whenever logrotate runs into another value
for one of these parameters, it resets the default value. You have a choice of rotating files daily, weekly, or monthly.
The number following the rotate keyword specifies the number of rotated log files that you want to keep. The
create keyword causes logrotate to create a new log file with the same name and attributes as the newly rotated
log file. The compress keyword (commented out in the default file) causes log files to be compressed using gzip.
The include keyword specifies the standard /etc/logrotate.d directory for program-specific logrotate specification
files. When you install a program using rpm (page 455), rpm puts the logrotate specification file (if it is part of the
package) in this directory.
The last set of instructions in logrotate.conf takes care of the /var/log/wtmp log file (wtmp holds login records;
you can view this file with the command who /var/log/wtmp). The keyword monthly overrides the default value
of weekly for this utility only (because the value is within brackets). The create keyword is followed by the arguments
establishing the permissions, owner, and group for the new file. Finally, rotate establishes that one rotated log file
should be kept.
The /etc/logrotate.d/samba file is an example of a utility-specific logrotate specification file:
$ cat /etc/logrotate.d/samba
/var/log/samba/*.log {
notifempty
missingok
sharedscripts
copytruncate
postrotate
/bin/kill -HUP 'cat /var/run/smbd.pid /var/run/nmbd.pid /var/run/winbindd.pid
2> /dev/null' 2> /dev/null || true
115
Draft
Removing Unused Space from Directories
Draft
endscript
}
This file, which is incorporated in /etc/logrotate.d because of the include statement in logrotate.conf, works
with each of the files in /var/log/samba that has a filename extension of log (*.log). The notifempty keyword
causes logrotate not to rotate the log file if it is empty, overriding the default action of rotating empty log files. The
missingok keyword means that no error will be issued when the file is missing. The sharedscripts keyword causes
logrotate to execute the command(s) in the prerotate and postrotate sections one time only, not one time for each
log that is rotated. The copytruncate keyword causes logrotate to truncate the original log file immediately after
it copies it. This keyword is useful for programs that cannot be instructed to close and reopen their log files because
they might continue writing to the original file even after it has been moved. The commands between postrotate
and endscript are executed after the rotation is complete. Similarly, commands between prerotate and endscript
are executed before the rotation is started.
The logrotate utility has many keywords, and many of these take arguments and have side effects. Refer to the
logrotate man page for details.
Removing Unused Space from Directories
A directory with too many filenames in it is inefficient. The point at which a directory on an ext2 or ext3 filesystem
becomes inefficient varies, depending partly on the length of the filenames it contains. Keep your directories relatively
small. Having fewer than a few hundred files (or directories) in a directory is generally a good idea, and having
more than a few thousand is generally a bad idea. Additionally, Linux uses a caching mechanism for frequently
accessed files to speed the process of locating an inode from a filename. This caching mechanism works only on
filenames of up to 30 characters in length, so avoid extremely long filenames for frequently accessed files.
When you find a directory that is too large, you can usually break it into several smaller directories by moving its
contents into new directories. Make sure that you remove the original directory once you have moved its contents.
Because Linux directories do not shrink automatically, removing a file from a directory does not shrink the directory,
even though it makes more space on the disk. To remove unused space and make a directory smaller, you must
copy or move all the files into a new directory and remove the original directory.
The following procedure removes unused directory space. First, remove all unneeded files from the large directory.
Then create a new, empty directory. Next, move or copy all the remaining files from the old large directory to the
new empty directory. Remember to copy hidden files. Finally, delete the old directory and rename the new directory:
#
#
#
#
116
mkdir /home/alex/new
mv /home/alex/large/* /home/alex/large/.[A-z]* /home/alex/new
rmdir /home/alex/large
mv /home/alex/new /home/alex/large
Draft
syslogd: Logs System Messages
Draft
Disk Quota System
The disk quota system limits the disk space and number of files owned by individual users. You can choose to
limit each user’s disk space, the number of files each user can own, or both. Each resource that is limited has two
limits. The lower limit, or quota, can be exceeded by the user, although a warning is presented each time the user
logs in when he or she is above the quota. After a certain number of warnings (set by the system administrator),
the system will behave as if the user had reached the upper limit. Once the upper limit is reached or the user has
received the specified number of warnings, the user will not be allowed to create any more files or use any more
disk space. The user’s only recourse at that point is to remove some files.
Users can review their usage and limits with the quota command. Superuser can use quota to obtain information
about any user.
First, you must decide which filesystems to limit and how to allocate space among users. Typically, only filesystems
that contain users’ home directories, such as /home, are limited. Use the edquota command to set the quotas, and
then use quotaon to start the quota system. You will probably want to put the quotaon command into the appropriate init script so that the quota system will be enabled when you bring up the system (page 381). Unmounting
a filesystem automatically disables the quota system for that filesystem.
syslogd: Logs System Messages
Traditionally, UNIX programs sent log messages to standard error. If a more permanent log was required, the
output was redirected to a file. Because of the limitations of this approach, 4.3BSD introduced the system log daemon
(syslogd) that Linux uses. This daemon listens for log messages and stores them in the /var/log hierarchy. In addition to providing logging facilities, syslogd allows a single machine to serve as a log repository for a network and
allows arbitrary programs to process specific log messages.
syslog.conf . The /etc/syslog.conf file stores configuration information for syslogd. Each line in this file contains
a selector and an action, separated by whitespace. The selector defines the origin and type of the message and the
action specifies how syslogd is to process the message. Sample lines from syslog.conf follow (a # indicates a
comment):
# Log all kernel messages to the console.
kern.*
# Log all the mail messages in one place.
mail.*
# Log cron stuff
cron.*
# Everybody gets emergency messages
*.emerg
# Save boot messages also to boot.log
local7.*
/dev/console
/var/log/maillog
/var/log/cron
*
/var/log/boot.log
Selectors . A selector is split into two parts, a facility and a priority, separated by a period. The facility indicates
the origin of the message; for example, kern messages come from the kernel and mail messages come from the
mail subsystem. Following is a list of facility names used by syslogd and the systems that generate these messages:
auth
Authorization and security systems including login
authpriv
Same as auth, but should be logged to a secure location
cron
cron
117
Draft
syslogd: Logs System Messages
Draft
daemon
System and network daemons without their own categories
kern
Kernel
lpr
Printing subsystem
mail
Mail subsystem
news
Network news subsystem
user
Default facility; all user programs use this
uucp
The UNIX to UNIX copy protocol subsystem
local0 to local7
Reserved for local use
The priority indicates the severity of the message. The following list of the priority names and the conditions they
represent is in priority order:
debug
Debugging information
info
Information not requiring intervention
notice
Conditions that may require intervention
warning
Warnings
err
Errors
crit
Critical conditions such as hardware failures
alert
Conditions requiring immediate attention
emerg
Emergency conditions
A selector consisting of a single facility and priority, such as kern.info, causes the corresponding action to be applied
to every message from that facility with that priority or higher (more urgent). Use .= to specify a single priority
(kern.=info applies the action to kernel messages of info priority). An exclamation point specifies that a priority
is not matched, so kern.!info matches kernel messages with a priority lower than info and kern.!=info matches
kernel messages with a priority other than info.
A line with multiple selectors, separated by semicolons, applies the action if any of the selectors is matched. Each
of the selectors on a line with multiple selectors constrains the match, with subsequent selectors frequently tightening the constraints. For example, the selectors mail.info;mail.!err match mail subsystem messages with info,
notice, or warning priorities.
You can replace either part of the selector with an asterisk to match anything. Similarly, the keyword none in either
part of the selector indicates no match is possible. The selector *.crit;kern.none matches all critical or higher
messages, except those from the kernel.
Actions . The action specifies how syslogd processes a message that matches the selector. The simplest actions
are regular files, specified by an absolute pathname; syslogd appends messages to these files. Because a terminal
is represented by a device file, you can specify /dev/console to have messages sent to the system console. If you
want a hard copy record of messages you can specify a device file that represents a dedicated printer.
You can write important messages to a specific user’s terminal by specifying a user-name, such as root, or a commaseparated list of usernames. Very important messages can be written to every logged-in terminal by using an asterisk.
To forward messages to syslogd on a remote system, specify the name of the system preceded by @. It is a good
idea to forward critical messages from the kernel to another system because these messages often precede a system
crash and may not be saved to the local disk. The following line from syslog.conf sends critical kernel messages
to grape:
118
Draft
Chapter Summary
kern.crit
Draft
@grape
Chapter Summary
This chapter starts by describing how to use system-config-users (FEDORA) and redhat-config-users (RHEL) to add
new users and groups to the system and how to modify existing users’ accounts. It also explains how to use the
equivalent command line tools to work with user accounts. The section on making backups explains how to use
the tar, cpio, dump, and restore utilities to back up and restore files.
The section on scheduling tasks discusses cron, the system scheduling daemon, and how to schedule tasks using
crontab, at, and KDE’s kcron. The system reports section covers vmstat, which details virtual memory, I/O, and
CPU statistics, and top, a useful tool to see how the system is performing from moment to moment and help you
figure out what might be slowing it down. The final sections cover general problem solving, discussing several tools
that can help track down system problems. One of the most important of these tools is syslogd, the system log
daemon. Using /etc/syslogd.conf, you can control which error messages appear on the console, which are sent
as email, and which go to one of several log files.
Exercises
1. How would you list all the processes running vi?
2. How would you use kill to cause a server process to reread its configuration files?
3. From the command line, how would you create a user named John Doe who has the username jd and who belongs
to group 65535?
4. How would you notify the users of the system that you are going to reboot the system in ten minutes?
5. Give a command that will create a level 0 dump of the /usr filesystem on the first tape device on the system.
What command would you use to take advantage of a drive that supports compression? What command would
place a level 3 dump of the /var filesystem immediately after the level 0 dump on the tape?
Advanced Exercises
6. If your system is not as responsive as it usually is, what is a good first step in figuring out where the problem is?
7. A process stores its PID in a file named process.pid. What command line will terminate the process?
8. Working as root, you are planning to delete some files but want to make sure that the wildcard expression you
will use is correct. Suggest two ways you could make sure that you deleted the correct files.
9. Create a cron file that will regularly perform the following backups:
a.
Performs a level 0 backup once per month.
b.
Performs a level 2 dump one day per week.
c.
Performs a level 5 dump every day that neither a level 0 nor a level 2 dump is performed.
In the worst case, how many restores would you have to perform to recover a file that was dumped using the
preceding schedule?
119
120
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 17 - Configuring a LAN
Configuring a LAN
IN THIS CHAPTER
Setting Up the Hardware
Gateways and Routers
Network Interface Card (NIC)
Configuring the Systems
system-config-network: Configuring the Hardware
iwconfig: Configuring a Wireless NIC
Setting Up Servers
Networks allow computers to communicate and share resources. A local area network (LAN) connects together
computers at one site, such as an office, home, or library, and can allow the connected computers to share an Internet connection and a printer. Of course, one of the most important reasons to set up a LAN is to allow systems
to communicate while users play multiplayer games.
This chapter covers the two aspects of configuring a LAN: setting up the hardware and configuring the software.
This chapter is not necessarily organized in the order you will perform the tasks involved in setting up a particular
LAN: Read the chapter through, figure out how you are going to set up the LAN, then read the parts of the chapter
in the order appropriate to your setup.
Setting Up the Hardware
Each system, or node, on a LAN must have a network interface card (NIC). NICs can be connected to the network
with cables or radio waves (wireless); in either case, there must be a hub that each of the systems connects to. If
the network is connected to another network, such as the Internet, it must also have a router, or gateway. The
router can be one of the systems on the LAN or a dedicated piece of hardware.
Connecting the Computers
A modern Ethernet-based LAN has a connection between each computer and a central hub. There are two kinds
of hubs: passive (sometimes just called a hub) and switching (called a switch). A passive hub simply connects all
systems together and shares the network bandwidth among the systems. A switching hub puts each system on its
own network with the switch and routes packets between those networks, providing each system with the full
network bandwidth.
In the simple network shown in Figure 17-1, four computers are connected to a single hub. Assuming the hub is
passive, when computers 1 and 2 are communicating at the same time as computers 3 and 4, each conversation is
121
Draft
Gateways and Routers
Draft
limited to a maximum of half the network bandwidth. If the hub were a switch, each conversation could use the
full network bandwidth.
Figure 17-1. A simple network
Usually, hubs are less expensive than switches. If you are planning on using a network for sharing an Internet
connection and light file sharing, a hub is likely to be fast enough. If systems on the network are going to exchange
files regularly, a switch may be more appropriate. Refer to “Ethernet” on page 327 for a discussion of switches,
hubs, and cables.
Each computer on a LAN must be connected to the hub. If you are using use more than one hub, connect the port
labeled uplink on one to a normal port on another.
Wireless access point (WAP) . A wireless access point (WAP) connects a wireless network to a wired one.
Typically, a WAP acts as a transparent bridge, forwarding packets between the two networks as if they were one.
If you connect multiple WAPs in different locations to the same wired network, it is possible for wireless clients to
roam transparently between the WAPs.
Wireless networks do not require a hub, although a WAP can optionally fill a similar role. In a wireless network,
the bandwidth is shared among all the nodes within range of each other; the maximum speed is limited by the
slowest node.
Gateways and Routers
If the LAN you are setting up is connected to another network, such as the Internet, you need a router, sometimes
called a gateway. A router can perform several functions, the most common of which is allowing several systems
to share a single Internet connection and IP address (masquerading). When a router masquerades packets, the
packets from each system on the LAN appear to come from a single IP address; return packets are passed back to
the correct system.
There are several choices for routers:
•
A simple hardware router is relatively cheap and does most things required by a small network.
•
You can set up a Red Hat Linux system as a router. The Linux kernel can use iptables to route packets between
network adapters.
•
You can use a Linux distribution tailored for use as a router. For example, SmoothWall (www.smoothwall.org
[http://www.smoothwall.org]) provides a browser-based configuration in the style of a hardware router.
122
Draft
Network Interface Card (NIC)
Draft
Network Interface Card (NIC)
Each system’s NIC may be a separate Ethernet card (wired or wireless) or it may be built into the motherboard.
Supported NICs . Linux supports most wired Ethernet NICs. Fewer wireless NICs are supported. See “More
Information” on page 559 for references.
Unsupported wireless NICs . If a wireless network card is not supported under Linux directly, you may be
able to get it to work with NdisWrapper (ndiswrapper.sourceforge.net), which uses Win32 drivers. NdisWrapper
is a kernel module that provides a subset of the Windows network driver API.
Wireless bridge . An alternative to a wireless NIC is a wireless bridge. A wireless bridge forwards packets
between wired and wireless interfaces, eliminating the need for wireless drivers. It is a simple device with an ethernet
port that plugs into a NIC and an 802.11 (wireless) controller. While carrying a bridge around is usually not possible
for mobile users, it is an easy way to migrate a desktop computer to a wireless configuration.
Mode . Wireless networks operate in either ad hoc or infrastructure mode. In ad hoc mode, individual nodes
in the network communicate directly with each other. In infrastructure mode, nodes communicate via a WAP (page
552). Infrastructure mode is generally more reliable if the wireless LAN communicates with a wired LAN.
If you do not want to use a WAP, it may be possible to set up a WLAN card so it acts as a WAP; consult the
NIC/driver documentation.
Configuring the Systems
kudzu . Once the hardware is in place, you need to configure each system so that it knows about the NIC that
connects it to the network. Normally, kudzu, the Red Hat utility that detects and configures new hardware, gives
the system the information it needs about the NIC. The kudzu utility probes the NIC when you install Red Hat
Linux or the first time you boot after you install a NIC.
You can use system-config-network (next) to augment the information kudzu collects and to activate the NIC. When
it prompts you for information, kudzu allows you to specify only one nameserver. It is a good idea to specify at two
or three nameservers; you can use system-config-network to add additional nameservers.
System information .
formation:
In addition to information about the NIC, each system must also have the following in-
•
The system’s IP address
•
The netmask (subnet mask) for the system’s address (pages 337 and 400)
•
The IP address of the gateway
•
The IP addresses of the nameservers (DNS addresses)
•
The system’s hostname (set when you install Red Hat Linux)
If you set up a DHCP server (page 408) to distribute network configuration information to systems on the LAN,
you do not need to specify the preceding information on each system; you just specify that the system is using
DHCP to obtain this information. You need to specify this information when you set up the DHCP server.
Private address space . When you set up a LAN, the IP addresses of the systems on the LAN are generally not
made public on the Internet. There are special IP addresses, part of the private address space defined by IANA (page
123
Draft
system-config-network: Configuring
the Hardware
Draft
976), that are reserved for private use and are appropriate to use on a LAN (Table 17-1). Unless you have been assigned IP addresses for the systems on the LAN, choose addresses from the private address space.
Table 17-1. Private IP Ranges
Range of IP Addresses
From IP Address
To IP Address
10.0.0.0/8
10.0.0.0
10.255.255.255
172.16.0.0/12
172.16.0.0
172.31.255.255
192.168.0.0/16
192.168.0.0
192.168.255.255
system-config-network: Configuring the Hardware
The system-config-network (FEDORA) and redhat-config-network (RHEL) utilities display the Network Configuration
window (Figure 17-2) with tabs to specify hosts (/etc/hosts, page 428) and DNS servers (/etc/resolv.conf, page
432), as well as to configure network hardware and logical devices associated with the hardware.
Figure 17-2. The Network Configuration window, Devices tab
Adding a device . Normally, kudzu identifies and adds new hardware to the system, after which you can then
use [system|redhat]-config-network to edit the configuration information. If you do need to add a NIC to the system,
click the Devices tab; then click New on the toolbar. The utility displays the Select Device Type window (Figure
17-3).
124
Draft
system-config-network: Configuring
the Hardware
Draft
Figure 17-3. Select Device Type window
The Select Device Type window can set up seven types of connections (most of which do not pertain to setting up
a LAN): CIPE (page 963), Ethernet (page 327), ISDN (page 978), modem, token ring (page 1002), wireless, and
xDSL (page 1006). CIPE, ISDN, modem, wireless, and xDSL are PPP (Point-to-Point Protocol) connections. PPP
is a serial line protocol that establishes a connection between two systems, putting them on the same network. PPP
is capable of handling several protocols, the most common of which is TCP/IP, which provides compression for
increased efficiency. The two systems can then run ssh, X, or any other network application between them. Ethernet
and token ring are used to connect to LANs.
Choose the type of connection you want to establish and click Forward. Some selections probe for information.
You can accept entries in the text boxes that are filled in in the following window. Fill in blank text boxes as appropriate. When you have finished setting up the device, click Apply. The Select Device Type window closes, leaving
the Network Configuration window displaying the device you just added. Follow the instructions in the next
paragraph to edit the configuration information or click the Devices tab, highlight the new device, click Menubar:
File Save, and click Activate to bring the new device on line.
Editing a device . The Network Configuration window (Figure 17-2) has four tabs; two pertain to hardware
devices and two to the system. The Hosts tab modifies the /etc/hosts file (page 428) and the DNS tab modifies the
system’s hostname and /etc/resolv.conf file (page 432). Make changes in these tabs as necessary.
To modify the configuration of network hardware, such as a NIC, click the Hardware tab, highlight the description
of the hardware, and click Edit on the toolbar. The utility displays the Network Adapters Configuration window.
In this window, you can change the name of the device (eth0, eth1, and so on) and the resources it uses. Typically,
you will change only the name. Click OK to accept the changes and close the window.
To modify the device represented by a piece of hardware, click the Devices tab, highlight the device, and click Edit
on the toolbar. The utility displays a window appropriate to the device you are editing. For example, if you are
working with an Ethernet NIC, [system|redhat]-config-network displays the Ethernet Device window (Figure 174).
125
Draft
iwconfig: Configuring a Wireless NIC
Draft
Figure 17-4. The Ethernet Device window
From this window, you can set up the device to use DHCP or manually specify the necessary IP addresses. The
Hardware Device tab allows you to associate the device with a piece of hardware and specify a MAC address (page
981). When you are finished making changes, click OK, click the Devices tab, highlight the new device, and click
Menubar: File Save. Activate the device if necessary.
iwconfig: Configuring a Wireless NIC
You can configure a wireless NIC using system-config-network (page 555) or iwconfig. The iwconfig utility is based
on ifconfig and configures elements of a wireless NIC not supported by ifconfig, such as setting up Master mode
and binding a card to a WAP.
The most common parameters you will change with iwconfig are the encryption key, the mode, and the name of
the network. Most devices support a minimum of 40-bit Wired Equivalent Privacy (WEP) encryption. The encryption
key is defined by a string of 10 hexadecimal digits. The contents of the string are arbitrary, but must be the same
on all nodes.
# iwconfig eth1 key 19FEB47A5B
126
Draft
Setting Up Servers
Draft
The algorithm used by WEP is known to be flawed; using it does not give much protection. If you require privacy,
use an encrypted protocol, such as SSH or HTTPS. If you are have difficulty connecting, disable encryption on all
nodes:
# iwconfig eth1 key off
The mode defines whether you are connecting to an ad hoc or an infrastructure network. Normally, you can set
mode to Auto, which selects the correct mode automatically:
# iwconfig eth1 mode Auto
The exception is if you want to use the NIC as a WAP, in which case you need to set mode to Master. Not all
wireless NICs are capable of acting as masters.
# iwconfig eth1 mode Master
The network name is defined by the ESSID (Extended Service Set ID), an arbitrary string. With the ESSID set (it
must be the same on every node, including the WAP), you should be able to roam between any set of nodes with
the same network name:
# iwconfig eth1 essid "My Wireless Network"
See the iwconfig man page for more information.
Setting Up Servers
Setting up local clients and servers can make a LAN easier to use and more useful. The following list briefly describes
some of these tools and references the pages that describe them in detail.
•
NIS . NIS can provide a uniform login regardless of which system you log in on. The NIS authentication
server is covered on page 645 and the client on page 640. NIS is often combined with home directories mounted
using NFS.
•
NFS . NFS allows you to share directory hierarchies. Sharing directories using NFS requires that the server
export the directory hierarchy (page 665) and the clients mount the hierarchy (page 658).
Using NFS, you can store all home directories on one system and mount them from other systems as needed.
This configuration works well with NIS login authentication. With this setup, it can be convenient to create a
world-writable directory, for example /home/shared, which users can use to exchange files. If you set the sticky
bit (page 998) on this directory (chmod 1777 /home/shared), users are able to delete only files they created.
If you do not set the sticky bit, any user can delete any file.
•
OpenSSH . OpenSSH tools include ssh (logs in on a remote system, page 570) and scp (copies files to/from
a remote system, page 572). You can also set up automatic logins with OpenSSH: If you set up a shared home
directory with NFS, each user’s ~/.ssh directory (page 565) is the same on each system; a user who sets up a
personal authentication key (page 575) will be able to use OpenSSH tools between systems without entering a
password. See page 575 to set up an OpenSSH server. You can just use the ssh and scp clients; you do not have
to set them up.
•
DHCP . DHCP enables a client system to retrieve network configuration information from a server each
time it connects to a network. See page 408 for more information.
127
Draft
•
More Information
Draft
Samba . Samba allows Linux systems to participate in a Windows network, sharing directories and printers,
and accessing those shared by Windows systems. Samba includes a special share for accessing users’ home
directories. For more information, refer to “The [homes] Share: Sharing Users’ Home Directories” on page 691.
You can also use Samba to set up a shared directory similar to the one described under “NFS.” To share a Linux
directory with Windows computers, place the following code in /etc/smb.conf (page 685). Any Windows user
can access this share; it can be used to exchange files between users and between Linux and Windows systems:
[public]
comment = Public file space
path = /home/shared
read only = no
public = yes
browseable = yes
More Information
Web .
SmoothWall Linux distribution www.smoothwall.org [http://www.smoothwall.org]
NdisWrapper ndiswrapper.sourceforge.net [http://ndiswrapper.sourceforge.net]
Hardware compatibility list hardware.redhat.com [http://hardware.redhat.com]
HOWTOs . Linux Wireless Lan HOWTO www.hpl.hp.com/personal/Jean_Tourrilhes/Linux
[http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux] Wireless HOWTO
Summary
A local area network (LAN) connects together computers at one site and can allow the connected computers to
share an Internet connection and a printer. Each system, or node, on a LAN must have a network interface card
(NIC). NICs can be connected to the network with cables or radio waves (wireless).
An Ethernet-based LAN has a connection between each computer and a central hub. There are two kinds of hubs:
passive (sometimes just called a hub) and switching (faster, called a switch). A wireless access point (WAP) connects
a wireless network to a wired one. If the LAN you are setting up is connected to another network, such as the Internet, you need a router (gateway). A router can perform several functions, the most common of which is allowing
several systems to share a single Internet connection and IP address, called masquerading.
You can set up the LAN to use NIS as a login server so that you do not have to set up accounts on each system.
You can use NFS, which allows you to mount remote directory hierarchies, to set up a universal home directory.
Samba is an important part of many LANs: It allows Linux systems to participate in a Windows network, sharing
directories and printers, and accessing those shared by Windows systems.
Exercises
1. What advantage does a switch have over a passive hub?
2. What server would you set up to allow users to log in with the same username and password on all computers
on a LAN?
3. Name two servers that allow you to share directories between systems.
128
Draft
Advanced Exercises
Draft
4. What is a WAP and what does it do?
5. What is a common function of a router? What is this function called?
6. What does a wireless bridge do?
7. What is kudzu? What does it do when you install a new NIC?
8. What is the private address space? When would you use a private address?
Advanced Exercises
9. If you set a system’s subnet mask to 255.255.255.0, how many computers can you put on the network without
using a router?
10. Which file stores information about which DNS servers the system uses?
129
130
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 25 IN THIS CHAPTER
How iptables Works
Rules, Matches, Targets, and Chains
Network Packet
Jumps and Targets
JumpStart: Using system-config-securitylevel to Build a
Firewall
Anatomy of an iptables Command
Building a Set of Rules
A Rule Set Generated by system-config-securitylevel
Sharing an Internet Connection Using NAT
The iptables utility builds and manipulates network packet filtering rules in the Linux kernel. You can use iptables
to set up a firewall to protect your system from malicious users and to set up NAT (Network Address Translation,
page 984), which can allow multiple systems to share a single Internet connection. The iptables utility is flexible
and extensible, allowing you to set up simple or complex network packet filtering solutions. It provides connection
tracking (stateful packet filtering), allowing you to handle packets based on the state of their connection. For example,
you can set up rules that reject inbound packets trying to open a new connection and accept inbound packets that
are responses to locally initiated connections. Features not included in the base iptables package are available as
patches via the patch-o-matic program. See page 740 for links to more information.
Some of the concepts involved in fully understanding iptables are beyond the scope of this book. You can use iptables
at different levels and this chapter attempts to present the fundamentals. There are, however, some sections of this
chapter that delve into areas that may require additional understanding or explanation. If a concept is not clear,
refer to one of the resources in “More Information” on page 740.
How iptables Works
What is frequently referred to as iptables is actually composed of two components: netfilter and iptables. Running
in kernelspace (page 979), the netfilter component is a set of tables that hold rules that the kernel uses to control
packet filtering. Running in userspace (page 1003), the iptables utility sets up, maintains, and displays the rules
stored by netfilter.
Rules, matches, targets, and chains . A rule comprises one or more criteria (matches or classifiers) and a single
action (a target). If, when a rule is applied to a packet, the packet matches all the criteria, the action is applied to
the packet. Rules are stored in chains. Each of the rules in a chain is applied, in order, to a packet, until there is a
match. If there is no match, the chain’s policy, or default action, is applied to the packet (page 745).
In the kernel, iptables replaces the earlier ipchains as a method of filtering network packets and provides multiple
chains for increased filtration flexibility. The iptables utility also provides stateful packet inspection (page 740).
131
Draft
How iptables Works
Draft
As an example of how rules work, assume that a chain has two rules (Figure 25-1). The first rule tests to see if a
packet’s destination is port 23 (FTP) and drops the packet if it is. The second rule tests to see if a packet was received
from the IP address 192.1681.1 and alters the packet’s destination if it was. When a packet is processed by the example chain, the kernel first applies the first rule in the chain to see if the packet arrived on port 23. If yes, the
packet is dropped and that is the end of processing for that packet. If no, the kernel applies the second rule in the
chain to see if the packet is from the specified IP address. If yes, the destination in the packet’s header is changed
and the modified packet is sent on its way. If no, the packet is sent on without being changed.
Figure 25-1. Example of how rules work
Chains are collected in three tables: Filter, NAT, and Mangle. Each of the tables has builtin chains as described
following. You can create additional, user-defined chains in Filter, the default table.
The default table. This table is mostly used to DROP or ACCEPT packets based on their content and does not alter
packets. Builtin chains are INPUT, FORWARD, and OUTPUT. All user-defined chains go in this table.
Note
NAT
The Network Address Translation table. Packets that create new connections are routed through this table, which
is used exclusively to translate the source or destination field of the packet. Builtin chains are PREROUTING,
OUTPUT, and POSTROUTING. Use with DNAT, SNAT, and MASQUERADE targets only.
•
DNAT (Destination NAT) alters the destination IP address of the first inbound packet in a connection so it is
rerouted to another host. Subsequent packets in the connection are automatically DNATed. Useful for redirecting
packets from the Internet that are bound for a firewall or a NATed server (page 755).
•
SNAT (Source NAT) alters the source IP address of the first outbound packet in a connection so that it appears
to come from a fixed IP address, for example, a firewall or router. Subsequent packets in the connection are
automatically SNATed. Replies to SNATed packets are automatically De-SNATed so they go back to the original sender. SNAT is useful for hiding LAN addresses from systems outside the LAN and using a single IP
address to serve multiple, local hosts. See also MASQUERADE, following.
•
MASQUERADE differs from SNAT only in that it checks for an IP address to apply to each outbound packet,
making it suitable for use with dynamic IP addresses such as those provided by DHCP (page 408). MASQUERADE is slightly slower than SNAT.
Used exclusively to alter TOS (Type of Service), TTL (Time To Live), and MARK fields in a packet. Builtin chains
are PREROUTING and OUTPUT.
Network Packet . When a packet from the network enters the kernel’s network protocol stack, it is given some
basic sanity tests, including checksum verification. After passing these tests, the packet goes through the
PREROUTING chain where it can have its destination address changed (Figure 25-2).
132
Draft
About iptables
Draft
Figure 25-2. Filtering a packet in the kernel
Next, the packet is routed, based on its destination address. If it is bound for the local system, it first goes through
the INPUT chain where it can be filtered (accepted, dropped, or sent to another chain) or altered. If the packet is
not addressed to the local system (the local system is forwarding the packet), it goes through the FORWARD and
POSTROUTING chains where it can again be filtered or altered.
Packets that are created locally pass through the OUTPUT and POSTROUTING chains where they can be filtered
or altered before being sent on the network.
The connection tracking machine or, as it is sometimes called, the state machine, gives you information on the
state of a packet, allowing you to define rules that match on the state of the connection the packet is part of. For
example, when a connection is opened, the first packet is part of a NEW connection, while subsequent packets
are part of an ESTABLISHED connection. Connection tracking is handled by the conntrack module.
The OUTPUT chain handles connection tracking for locally generated packets. The PREROUTING chain handles
connection tracking for all other packets. For more information, refer to “State” on page 748.
Before the advent of connection tracking, it was sometimes necessary to open many or all nonprivileged ports to
make sure that you accepted all RETURN and RELATED traffic. Because connection tracking allows you to
identify these kinds of traffic, you can keep many more ports closed to general traffic, making a system more secure.
Jumps and Targets . A jump or target specifies the action the kernel takes if a packet matches all the match
criteria for the rule being processed (page 748).
About iptables
This section contains information about iptables: resources, prerequisites for running iptables, and notes.
More Information
Documentation, HOWTOs, FAQs, patch-o-matic, security information .
Tutorial .
netfilter.org
www.faqs.org/docs/iptables [http://www.faqs.org/docs/iptables]
Scripts and more .
www.linuxguruz.com/iptables [http://www.linuxguruz.com/iptables]
KernelAnalysis-HOWTO
IP Masquerade HOWTO (contains useful scripts)
133
Draft
Prerequisites
Draft
Netfilter Extensions HOWTO at netfilter.org and www.iptables.org/documentation/HOWTO/netfilter-extensionsHOWTO.html [http://www.iptables.org/documentation/HOWTO/netfilter-extensions-HOWTO.html]
TCP Illustrated by W. Richard Stevens, Addison-Wesley, 1993.
Prerequisites
Install the following package:
•
iptables
Run chkconfig to cause iptables to start when the system comes up:
# /sbin/chkconfig iptables on
To ensure maximum protection, the init script (page 381) starts packet filtering by running iptables very soon after
entering runlevels 2–5 while runlevels 0, 1, and 6 do not stop packet filtering almost until the system leaves the
runlevel.
Notes
The iptables utility differs from most other Linux utilities in its setup and use. Where other Linux utilities such as
Apache, vsftpd, and sshd read the data that controls their operation from a configuration file, iptables requires
that you give a series of iptables commands to build a set of packet filtering rules that are kept in the kernel.
There are two ways to set up reliably the same set of rules each time you bring the system up. First, you can put
iptables commands in a script and run that script each time the system comes up. You can call this script from
/etc/rc.d/rc.local.
Second, you can put the arguments to the iptables commands you want to execute in /etc/sysconfig/iptables. The
system-config-securitylevel utility (page 751), as well as the Anaconda installer (page 46), uses this technique,
building sets of rules and storing the corresponding iptables command arguments in /etc/sysconfig/iptables.
For information on copying packet filtering rules to and from the kernel, refer to “Copying Rules to and from the
Kernel” on page 750. You can run iptables with the –L option or you can run service iptables status to display
the packet filtering rules the kernel is using.
If you have a problem with firewall rules, you can return packet processing rules in the kernel to their default state
without rebooting by giving the following commands:
# iptables --flush && iptables --delete-chain
These commands flush all chains and delete any user-defined chains, leaving the system without a firewall. In an
emergency you can give the following command to unload all iptables modules from the kernel and set a policy of
DROP for all tables:
# /sbin/service iptables panic
134
Draft
JumpStart: Using system-config-securitylevel to Build a Firewall
Draft
JumpStart: Using system-config-securitylevel to Build a Firewall
The system-config-securitylevel (FEDORA) and redhat-config-securitylevel (RHEL) utilities (Figure 25-3) build an
extremely simple firewall. It cannot deal with complex setups; it cannot even deal with NFS. The system-configsecuritylevel utility has two tabs; the SELinux tab is discussed on (page 380). The Firewall Options tab displays a
few more choices than redhat-config-securitylevel and is discussed here.
Figure 25-3. The Security Level Configuration window
From the Security level combo box, select Enable firewall. The firewall automatically allows packets that originate
locally through to the outside (generally the Internet) and allows responses to those packets back in.
Click the check boxes of the services you want to provide; these boxes set up a firewall that allows the local system
to be one or more of the following types of servers: HTTP (Web), FTP, SSH, TELNET, and SMTP (mail).
Click the check box next to any devices you trust. The sit0 device is a virtual device that can be used for tunneling
IPv6 over IPv4; you can ignore this device if you have not configured IPv6. Do not specify a device that connects
directly or indirectly to the Internet as a trusted device. You can enter the numbers of other ports you want to open
in the Other ports text box.
Click OK and [system|redhat]-config-securitylevel sets up and turns on the firewall. For more information, refer to
“A Rule Set Generated by system-config-securitylevel” on page 751.
135
Draft
Anatomy of an iptables Command
Draft
Anatomy of an iptables Command
This section lists the components of an iptables command line that follow the name of the utility, iptables. Except
as noted, the iptables utility is not sensitive to the position of arguments on the command line. The examples in
this chapter reflect a generally accepted syntax that allows commands to be easily read, understood, and maintained.
Not all commands have all components.
Many tokens on an iptables command line have two forms: a short form, consisting of a single letter preceded by
a single hyphen, and a long form, consisting of a word preceded by two hyphens. Most scripts use the short forms
for brevity; lines using the long forms can get unwieldy. The following iptables command lines are equivalent and
are used as examples in this section:
# iptables --append FORWARD --in-interface eth1 --out-interface eth0 --jump ACCEPT
# iptables -A FORWARD -i eth1 -o eth0 -j ACCEPT
Specifies the name of the table the command operates on: FILTER, NAT, or MANGLE. You can specify a table
name in any iptables command. When you do not specify a table name, the command operates on the FILTER
table. Most of the examples in this chapter do not specify table names and, therefore, work on the FILTER table.
Specify a table as –t tablename or --table tablename .
Tells iptables what to do with the rest of the command line: for example, add or delete a rule, display rules, or add
a chain. The example commands, –A or --append, append the rule specified by the command line to the specified
table and chain. See (page 744) for a list of commands.
Specifies the name of the chain that this rule belongs to or that this command works on. The chain is INPUT,
OUTPUT, FORWARD, PREROUTING, POSTROUTING, or the name of a user-defined chain. Specify a chain
by putting the name of the chain on the command line without preceding hyphens. The examples at the beginning
of this section work with the FORWARD table.
There are two kinds of match criteria: packet match criteria that match a network packet and rule match criteria
that match an existing rule.
Packet match criteria identify network packets and implement rules that take action on packets that match the
criteria. Packet match criteria and an action together are called a rule specification. Rule specifications form the
basis for packet filtering. The first example at the beginning of this section uses the --in-interface eth1 --out-interface eth0 rule match criteria; the second example uses the short form of the same criteria: –i eth1 –o eth0.
Both of these rules forward packets that come in on device eth1 to device eth0.
Rule match criteria identify existing rules. An iptables command can modify, remove, or position a new rule adjacent
to a rule specified by a rule match criterion. There are two ways to identify an existing rule. You can use the same
rule specification that was used to create the rule or you can use the rule’s ordinal number, called a rule number.
Rule numbers begin with 1, the first rule in a chain, and can be displayed with iptables –L --line-numbers. The
first of the following commands deletes the rule listed at the beginning of this section; the second replaces rule
number 3 in the INPUT chain with a rule that rejects all packets from IP address 192.168.0.10.
# iptables --delete -A FORWARD -i eth1 -o eth0 -j ACCEPT
# iptables -R INPUT 3 -source 192.168.0.10 -jump REJECT
A jump or target specifies what action the kernel takes on packets that match all the match criteria for this rule.
Specify a jump or target as –j target or --jump target . The examples at the beginning of this section specify the
ACCEPT target using the following commands: --jump ACCEPT and –j ACCEPT.
136
Draft
Building a Set of Rules
Draft
A jump transfers control to a different chain within the same table. The following command adds (--append) a
rule to the INPUT chain that transfers packets that use the TCP protocol (--protocol tcp) to a user-defined chain
named tcp_rules (--jump tcp_rules).
# iptables --append INPUT --protocol tcp --jump tcp_rules
When the packet finishes traversing the tcp_rules chain, assuming it has not been dropped or rejected, it continues
traversing the INPUT chain from the rule following the one it jumped from.
A target specifies an action the kernel takes on the packet; the simplest actions are ACCEPT and DROP or REJECT.
The following command adds a rule to the FORWARD chain that rejects packets coming from the FTP port
(/etc/services, the file iptables consults to determine which port to use, shows that FTP uses port 22).
# iptables --append FORWARD --sport ftp --jump REJECT
Some targets, such as LOG, are non-terminating: Control passes to the next rule after the target is executed. See
page 748 for information on how to use targets.
Building a Set of Rules
To specify a table, it is common practice to put the table declaration on the command line immediately following
iptables. For example, the following command flushes (deletes all the rules from) the NAT table:
# iptables -t NAT -F
Commands
Following is a list of iptables commands:
Note
--append
–A . Adds rule(s) specified by rule-specifications to the end of chain. When a packet matches rule-specifications,
target processes it.
iptables –A chain rule-specification –jump target
Note
--delete
–D .
Removes one or more rules, as specified by rule-numbers or rule-specifications.
iptables –D chain rule-number | rule-specification
Note
--insert
137
Draft
Commands
Draft
–I . Adds rule(s) specified by rule-specifications and target to the location in chain specified by rule-number. If
you do not specify rule-number, it defaults to 1, the head of the chain.
iptables –I chain rule-number rule-specification –jump target
Note
--replace
–R . Replaces rule number rule-number with rule-specification and target. Command fails if rule-specification
resolves to more than one address.
iptables –R chain rule-number | rule-specification –jump target
Note
--list
–L . Displays rules in chain. Omit chain to display rules for all chains. Use --line-numbers to display rule
numbers.
iptables –L [chain] display-options
Note
--flush
–F .
Deletes all rules from chain. Omit chain to delete all rules from all chains.
iptables –F [chain]
Note
--zero
–Z . Change to 0 the value of all packet and byte counters in chain or in all chains when you do not specify
chain. Use with –L to display the counters before clearing them.
iptables –Z [–L] [chain]
Note
--delete-chain
–X . Removes the user-defined chain named chain. If you do not specify chain, removes all user-defined chains.
You cannot delete a chain that a targets point to.
iptables –X chain
Note
--policy
–P . Sets the default target or policy for a builtin chain. This policy is applied to packets that do not match any
rule in the chain. If a chain does not have a policy, unmatched packets are ACCEPTed.
138
Draft
Packet Match Criteria
Draft
iptables –P builtin-chain builtin-target
Note
--rename-chain
–E .
Changes the name of chain old to new.
iptables –E old new
Note
--help
–h .
Displays a summary of iptables command syntax.
iptables –h
Follow a match extension protocol with –h to display options you can use with that protocol. For more information,
refer to “Help with extensions” on page 746.
Packet Match Criteria
The following criteria match network packets. When you precede a criterion with an exclamation point (!), the
rule matches packets that do not match the criterion.
Note
--protocol [!] proto–p
Matches on protocol. This is a match extension (page 746).
Note
--source [!] address[/mask]
–s or --src . Matches if the packet came from address. The address can be a name or IP address. See (page 400)
for formats of the optional mask that you can use only with an IP address.
Note
--destination [!] address[/mask]
–d or --dst . Matches if the packet is going to address. The address can be a name or IP address. See (page 400)
for formats of the optional mask, which you can use only with an IP address.
Note
--in-interface [!] iface[+]
–i . For the INPUT, FORWARD, and PREROUTING chains, matches if iface is the name of the interface the
packet was received from. Append a plus sign (+) to iface to match any interface whose name begins with iface.
When you do not specify in-interface, the rule matches packets coming from any interface.
139
Draft
Display Criteria
Draft
Note
--out-interface [!] iface[+]
–o . For the FORWARD, OUTPUT, and POSTROUTING chains, matches if iface is the interface the packet
will be sent to. Append a plus sign (+) to iface to match any interface whose name begins with iface. When you do
not specify out-interface, the rule matches a packet going to any interface.
Note
[!] –fragment
–f . Matches the second and subsequent fragments of fragmented packets. Because these packets do not contain
source or destination information, they do not match any other rules.
Display Criteria
The following criteria display information. All packets match these criteria.
Note
--verbose
–v .
Displays additional output.
Note
--numeric
–n .
Displays IP addresses and port numbers as numbers, not names.
Note
--exact
–x .
Use with –L to display exact packet and byte counts instead of rounded values.
Note
--line-numbers
Display line numbers when listing rules. The line numbers are also the rule numbers that you can use in rule match
criteria (page 743).
Match Extensions
Rule specification (packet match criteria) extensions, called match extensions, add matches based on protocols and
state to matches described previously. Each of the protocol extensions is kept in a module that must be loaded
before the match extension can be used. The command that loads the module must appear in the same rule specification as, and to the left of, the command that uses the module. There are two types of match extensions: implicit and explicit.
140
Draft
Match Extensions
Draft
Implicit Match Extensions
Implicit extensions are loaded (somewhat) automatically when you use a --protocol command (following). Each
protocol has its own extensions. Follow the protocol with –h to display extensions you can use with that protocol.
For example, iptables –p tcp –h displays TCP extensions at the end of the help output.
Note
--protocol [!] proto
–p . Loads the proto module and matches if the packet uses proto. The proto can be a name or number from
/etc/protocols, including tcp, udp, and icmp (page 976).
Specifying all or 0 (zero) matches any of tcp, udp, or icmp and is the same as not including this match in a rule.
The following criteria load the TCP module and match TCP protocol packets coming from port 22 (ssh packets):
--protocol tcp --source-port 22
The following command expands the preceding match to cause iptables to drop all incoming ssh packets. This
command uses ssh, which iptables looks up in /etc/services, in place of 22.
# iptables --protocol tcp --source-port ssh -jump DROP
This section does not describe all extensions; use –h, described preceding, for a complete list.
TCP
Note
--destination-port [!] [port][:port]]
--dport . Matches a destination port number or service name (see /etc/services). You can also specify a range
of port numbers, :port to specify ports 0 through port, or port: to specify ports port through 65535.
Note
--source-port [!] [port][:port]]
--sport . Matches a source port number or service name (see /etc/services). You can also specify a range of port
numbers, :port to specify ports 0 through port, or port: to specify ports port through 65535.
Note
[!] --syn
Matches packets with the SYN bit set and ACK and FIN bits cleared. This match extension is shorthand for --tcpflags SYN,RST,ACK SYN.
Note
--tcp-flags [!] mask comp
141
Draft
Match Extensions
Draft
Defines TCP flag settings that constitute a match. Valid flags are SYN, ACK, FIN, RST, URG, PSH, ALL, NONE.
The mask is a comma-separated list of flags to be examined and comp is a comma-separated subset of mask that
specifies the flags that must be set for there to be a match. Flags not specified in mask must be unset.
Note
--tcp-option [!] n
Matches TCP options based on their decimal value (n).
UDP
You can specify a source and/or destination port in the same manner as described under “TCP,” preceding.
ICMP
ICMP (page 976) packets carry messages only.
Note
--icmp-type [!] name
Matches when the packet is an ICMP packet of type name. The name can be a numeric ICMP type or one of the
names returned by
# iptables -p icmp -h
Explicit Match Extensions
Explicit match extensions differ from implicit match extensions in that you must use a –m or --match option to
specify a module before you can use the extension. There are many explicit match extension modules; this chapter
covers state, one of the most important.
State
This match extension matches on the state of the connection the packet is part of (page 740).
Note
--state state
Matches a packet whose state is defined by state, a comma-separated list of states from the following list:
ESTABLISHED . Any packet, within a specific connection, following the exchange of packets in both directions
for that connection.
INVALID .
NEW .
A stateless or unidentifiable packet.
The first packet within a specific connection, typically a SYN packet.
RELATED . Any packets exchanged in a connection spawned from an ESTABLISHED connection. For example,
an FTP data connection might be related to the FTP control connection. (You need the ip_conntrack_ftp module
for FTP connection tracking.)
142
Draft
Targets
Draft
The following command establishes a rule that matches and drops invalid packets as well as packets from new
connections:
# iptables --match state --state INVALID,NEW -jump DROP
Targets
All targets are built in; there are no user-defined targets. This section lists some of the targets available with iptables
as distributed by Red Hat. Applicable target options are listed following each target.
Note
ACCEPT
Continues processing the packet.
Note
DNAT
Destination Network Address Translation .
Rewrites the destination address of the packet.
--to-destination ip[-ip][:port-port]
Same as SNAT with to-source, except changes the destination addresses of packets to the specified address(es)
and port(s) and is valid only in the PREROUTING or OUTPUT chains of the NAT table and any user-defined
chains called from those chains. The following command adds to the PREROUTING chain of the NAT table a rule
that changes the destination in the headers of TCP packets with a destination of 66.187.232.50 to 192.168.0.10.
# iptables -t NAT -A PREROUTING -p tcp -d 66.187.232.50 -j DNAT --to-destination 192.168.0.10
Note
DROP
Ends the packet’s life without notice.
Note
LOG
Turns on logging for the packet being processed. The kernel uses syslogd (page 546) to process output generated
by this target. LOG is a nonterminating target; processing continues with the next rule. Use two rules to LOG
packets that you REJECT, one each with the targets LOG and REJECT, with the same matching criteria.
--og-level n
Specifies a logging level as per syslog.conf (page 546).
--log-prefix string
Prefixes log entries with string, which can be up to 14 characters.
143
Draft
Targets
Draft
--log-tcp-options
Logs options from the TCP packet header.
--log-ip-options
Logs options from the IP packet header.
Note
MASQUERADE
Similar to SNAT with --to-source, except the IP information is grabbed from the interface on the specified port.
For use on systems with dynamically assigned IP addresses, such as those that use DHCP, including most dial-up
lines. Valid only in rules in the POSTROUTING chain of the NAT table.
--to-ports port[-port]
Specifies the port for the interface you want to masquerade. Forgets connections when the interface goes down, as
is appropriate for dial-up lines. You must specify a TCP or UDP protocol (--protocol tcp or udp) with this target.
Note
REJECT
Similar to DROP, except it notifies the sending system that the packet was blocked.
--reject-with type
Returns the error type to the originating system. The type can be one of the following, which returns the appropriate
ICMP (page 976) error: icmp-net-unreachable, icmp-host-unreachable, icmp-port-unreachable, icmp-protounreachable, icmp-net-prohibited, or icmp-host-prohibited. You can specify type as echo-reply from rules
that specify an ICMP ping packet to return a ping reply. You can specify tcp-reset from rules in or called from the
INPUT chain to return a TCP RST packet. Valid in the INPUT, FORWARD, and OUTPUT chains and user-defined
chains called from these chains.
Note
RETURN
Stops traversing this chain and returns the packet to the calling chain.
Note
SNAT
Source Network Address Translation .
a LAN that share an Internet connection.
Rewrites the source address of the packet. Appropriate for hosts on
--to-source ip[-ip][:port-port]
Alters the source IP address of an outbound packet, and all future packets in this connection, to ip. Skips additional
rules, if any. Returning packets are automatically de-SNATed so they return to the originating host. Valid only in
the POSTROUTING chain of the NAT table.
144
Draft
Copying Rules to and from the Kernel
Draft
When you specify a range of IP addresses (ip-ip), or use multiple to-source targets, iptables assigns the addresses
in a round-robin fashion, cycling through the addresses, one for each new connection.
When the rule specifies the TCP or UDP protocol (–p tcp or –p udp), you can specify a range of ports. When you
do not specify a range of ports, the rule matches all ports. Every connection on a NATed subnet must have a unique
IP address and port combination. If two computers on a NATed subnet try to use the same port, the kernel maps
one of the ports to another (unused) one. Ports less than 512 are mapped to other ports less than 512, ports from
512 to 1024 are mapped to other ports from 512 to 1024, and ports above 1024 are mapped to other ports above
1024.
Copying Rules to and from the Kernel
The iptables-save utility copies packet filtering rules from the kernel to standard output so you can save them in a
file. The iptables-restore utility copies rules from standard input, as written by iptables-save, to the kernel. Sample
output from iptables-save follows:
# iptables-save
# Generated by iptables-save v1.2.8 on Sun Mar 14 20:41:37 2004
*filter
:INPUT ACCEPT [0:0]
:FORWARD ACCEPT [0:0]
:OUTPUT ACCEPT [0:0]
:RH-Firewall-1-INPUT - [0:0]
-A INPUT -j RH-Firewall-1-INPUT
-A FORWARD -j RH-Firewall-1-INPUT
-A RH-Firewall-1-INPUT -i lo -j ACCEPT
...
COMMIT
Most of the lines that iptables-save writes are iptables command lines without the iptables at the beginning. Lines
that begin with a pound sign (#) are comments. Lines that start with an asterisk are names of tables that the following
commands work on; all the preceding commands work on the Filter table. The COMMIT line must appear at the
end of all the commands for a table; it executes the preceding commands. Lines that begin with colons specify
chains in the following format:
:chain policy [packets:bytes]
Where chain is the name of the chain, policy is the policy (default target) for the chain, and packets and bytes are
the packet and byte counters. The square brackets must appear in the line; they do not indicate optional parameters.
Refer to www.faqs.org/docs/iptables/iptables-save.html [http://www.faqs.org/docs/iptables/iptables-save.html]
for more information.
A Rule Set Generated by system-config-securitylevel
This section describes a set of rules generated by system-config-securitylevel (page 741) when you ask it to create a
default firewall. The redhat-config-securitylevel works the same way. The system-config-securitylevel utility, which
replaces gnome-lokkit, writes the rules in the format used by iptables-save (see the preceding section) to /etc/sysconfig/iptables, so the firewall is implemented each time the system boots.
In the following listing, *filter indicates that the following commands work on the FILTER table and the first line
that begins with a colon specifies that the policy for the INPUT chain in the FILTER table is ACCEPT. FORWARD
145
Draft
A Rule Set Generated by system-configsecuritylevel
Draft
and OUTPUT chains are specified similarly. Because the counters for all the chains are zero, the counters will be
reset to zero each time the system boots and starts iptables from this file.
The system-config-securitylevel utility creates a user-defined chain that is named RH-Firewall-1-INPUT. No policy
is specified because user-defined chains cannot have policies.
# cat /etc/sysconfig/iptables
# Firewall configuration written by system-config-securitylevel
# Manual customization of this file is not recommended.
*filter
:INPUT ACCEPT [0:0]
:FORWARD ACCEPT [0:0]
:OUTPUT ACCEPT [0:0]
:RH-Firewall-1-INPUT - [0:0]
-A INPUT -j RH-Firewall-1-INPUT
-A FORWARD -j RH-Firewall-1-INPUT
-A RH-Firewall-1-INPUT -i lo -j ACCEPT
-A RH-Firewall-1-INPUT -p icmp --icmp-type any -j ACCEPT
-A RH-Firewall-1-INPUT -p 50 -j ACCEPT
-A RH-Firewall-1-INPUT -p 51 -j ACCEPT
-A RH-Firewall-1-INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
-A RH-Firewall-1-INPUT -j REJECT --reject-with icmp-host-prohibited
COMMIT
The first two lines that begin with –A add rules to the INPUT and FORWARD chains that cause control to transfer
to the RH-Firewall-1-INPUT chain. The subsequent lines append rules to the RH-Firewall-1-INPUT chain.
COMMIT executes the preceding commands. Following is a description of what the rest of the lines do:
-A RH-Firewall-1-INPUT -i lo -j ACCEPT
The preceding line accepts packets from the local interface.
-A RH-Firewall-1-INPUT -p icmp --icmp-type any -j ACCEPT
The preceding line accepts all ICMP packets.
-A RH-Firewall-1-INPUT -p 50 -j ACCEPT
-A RH-Firewall-1-INPUT -p 51 -j ACCEPT
The preceding lines accept packets that match protocol 50 and 51, which /etc/protocols lists as IPv6-Crypt, an
encryption header for IPv6, and IPv6-Auth, an encryption header for IPv6.
-A RH-Firewall-1-INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
The preceding line uses –m to specify the state module and accepts ESTABLISHED and RELATED packets.
-A RH-Firewall-1-INPUT -j REJECT --reject-with icmp-host-prohibited
The preceding line rejects all packets that have not been accepted and returns ICMP error icmp-host-prohibited
to the system that sent the packet.
146
Draft
Sharing an Internet Connection Using
NAT
Draft
With the preceding rules loaded, you can use iptables to list the rules and see the defaults that iptables puts in place.
# iptables -L
Chain INPUT (policy ACCEPT)
target
prot opt source
RH-Firewall-1-INPUT all --
anywhere
destination
anywhere
Chain FORWARD (policy ACCEPT)
target
prot opt source
RH-Firewall-1-INPUT all -- anywhere
destination
anywhere
Chain OUTPUT (policy ACCEPT)
target
prot opt source
destination
Chain RH-Firewall-1-INPUT (2 references)
target
prot opt source
destination
ACCEPT
all -anywhere
anywhere
ACCEPT
icmp -- anywhere
anywhere
ACCEPT
ipv6-crypt-- anywhere
anywhere
ACCEPT
ipv6-auth-- anywhere
anywhere
ACCEPT
all -anywhere
anywhere
REJECT
all -anywhere
anywhere
icmp any
state RELATED,ESTABLISHED
reject-with icmp-host-prohibited
Sharing an Internet Connection Using NAT
On the Internet there are a lot of scripts available that set up Internet connection sharing using iptables. Each of
the scripts boils down to the same few basic iptables commands with minor differences. This section discusses
those few statements to explains how a connection can be shared. You can use the statements presented in this
section or refer to the Linux IP Masquerade HOWTO for complete scripts, the simplest of which is at
tldp.org/HOWTO/IP-Masquerade-HOWTO/firewall-examples.html.
There are two ways you can share a single connection to the Internet (one IP address). Both involve setting up
NAT to alter addresses in and forward packets. The first allows clients (browsers, mail readers, and so on) on
several systems on a LAN to share a single IP address to connect to servers on the Internet. The second allows
servers (mail, web, FTP, and so on) on different systems on a LAN to provide their services over a single connection
to the Internet. You can use iptables to set up one or both of these configurations. In both cases, you need to set
up a system that is a router: It must have two network connections—one connected to the Internet and the other
to the LAN.
For optimum security, use a dedicated system as a router. Because data transmission over a connection to the Internet, even a broadband connection, is relatively slow, using an slower, older system as a router does not generally
slow down a LAN. This setup gives you some defense against intrusion from the Internet. A workstation on the
LAN can also function as a router, but this setup means you have data on a system that is directly connected to the
Internet. The following sections discuss the security of each setup.
The examples in this section assume that the device named eth0 connects to the Internet on 10.255.255.255 and
that eth1 connects to the LAN on 192.168.0.1. Substitute the devices and IP addresses that your systems use. If
you use a modem to connect to the Internet, you need to substitute ppp0 (or another device) for eth0 in the examples.
In order for the examples in this section to work, you must turn on IP forwarding. First, give the following command
and make sure everything is working:
147
Draft
Connecting Several Clients to a Single
Internet Connection
Draft
# /sbin/sysctl -w net.ipv4.ip_forward=1
Once you know that iptables is working the way you want, change the 0 to a 1 in the following line in /etc/sysctl.conf
to make the kernel always perform IP forwarding:
net.ipv4.ip_forward = 0
Connecting Several Clients to a Single Internet Connection
Configuring the kernel of the router system to allow clients on multiple, local systems on the LAN to connect to
the Internet requires you to set up IP masquerading, or SNAT (source NAT). IP masquerading translates the source
and destination addresses in the headers of network packets that originate on local systems and the packets that
remote servers send in response to those packets. These packets are part of connections that originate on a local
system. The example in this section does nothing to packets that are part of connections that originate on the remote
systems (on the Internet): These packets cannot get past the router system, providing a degree of security.
The point of rewriting the packet headers is to allow systems with different local IP addresses to share a single IP
address on the Internet. The router system translates the source or origin address of packets from local systems to
that of the Internet connection, so that all packets passing from the router to the Internet appear to come from a
single system, 10.255.255.255 in the example. All packets sent in response by remote systems on the Internet to
the router system have the address of the Internet connection, 10.255.255.255 in the example, as their destination
address. The router system remembers each connection and alters the destination address on each response packet
to that of the local, originating system.
The router system is established by four iptables commands, one of which sets up a log of masqueraded connections.
The first command puts the first rule in the FORWARD chain of the FILTER (default) table (–A FORWARD):
# iptables -A FORWARD -i eth0 -o eth1 -m state --state ESTABLISHED,RELATED -j ACCEPT
To match this rule, a packet must be
1.
Received on eth0 (coming in from the Internet): –i eth0.
2.
Going to be sent out on eth1 (going out to the LAN): –o eth1.
3.
Part of an established connection or a connection that is related to an established connection: --state ESTABLISHED,RELATED.
The kernel accepts (–j ACCEPT) packets that meet these three criteria. Accepted packets pass to the next appropriate chain/table. Packets that are not accepted pass to the next rule in the FORWARD chain. Packets from the
Internet that attempt to create a new connection are not accepted by this rule.
The second command puts the second rule in the FORWARD chain of the FILTER table:
# iptables -A FORWARD -i eth1 -o eth0 -j ACCEPT
To match this rule, a packet must be
1.
Received on eth1 (coming in from the LAN): –i eth1.
2.
Going to be sent out on eth0 (going out to the Internet): –o eth0.
148
Draft
Connecting Several Clients to a Single
Internet Connection
Draft
The kernel accepts packets that meet these two criteria, which means that all packets that originate locally and are
going to the Internet are accepted. Accepted packets pass to the next appropriate chain/table. Packets that are not
accepted pass to the next rule in the FORWARD chain.
The third command puts the third rule in the FORWARD chain of the FILTER table:
# iptables -A FORWARD -j LOG
This rule has no match criteria so it acts on all packets it processes. This rule’s action is to log packets, which means
it logs packets from the Internet that attempt to create a new connection.
Packets that get to the end of the FORWARD chain of the FILTER table are done with the rules set up by iptables
and are handled by the local tcp stack. Packets from the Internet that attempt to create a new connection on the
router system are accepted or returned, depending on whether the service they are trying to connect to is available
on the router system.
The fourth command puts the first rule in the POSTROUTING chain of the NAT table. Only packets that are establishing a new connection are passed to the NAT table. Once a connection has been set up for SNAT or MASQUERADE, the headers on all subsequent ESTABLISHED and RELATED packets are altered the same way as the
first packet. Packets that are sent in response to these packets automatically have their headers adjusted so that
they return to the originating local system.
# iptables -t NAT -A POSTROUTING -o eth0 -j MASQUERADE
To match this rule, a packet must be
1.
Establishing a new connection (otherwise it would not have come to the NAT table).
2.
Going to be sent out on eth0 (going out to the Internet): –o eth0.
The kernel MASQUERADEs all packets that meet these criteria, which means that all locally originating packets
that are establishing new connections have their source address changed to the address that is associated with eth0
(10.255.255.255 in the example).
Following are the four commands together:
#
#
#
#
iptables
iptables
iptables
iptables
-A
-A
-A
-t
FORWARD -i eth0 -o eth1 -m state --state ESTABLISHED,RELATED -j ACCEPT
FORWARD -i eth1 -o eth0 -j ACCEPT
FORWARD -j LOG
NAT -A POSTROUTING -o eth0 -j MASQUERADE
You can put these commands in /etc/rc.local or in a script called by this file on the router system to have them
executed each time the system boots. Or you can put them in /etc/sysconfig/iptables, leaving off the iptables
command at the start of each line. When you put the commands in the iptables file, they are be executed by the
iptables init script each time it is called. For more information, refer to “Copying Rules to and from the Kernel”
on page 750.
To limit the local systems that can connect to the Internet, you can add a –s (source) match criterion to the last
command as shown following:
# iptables -t NAT -A POSTROUTING -o eth0 -s 192.168.0.0-192.168.0.32 -j MASQUERADE
149
Draft
Connecting Several Servers to a Single
Internet Connection
Draft
In the preceding command, –s 192.168.0.0-192.168.0.32 causes only packets from an IP address in the specified
range to be MASQUERADEd.
Connecting Several Servers to a Single Internet Connection
DNAT (destination NAT) can set up rules to allow clients from the Internet to send packets to servers on the LAN.
This example sets up an SMTP mail server on 192.168.1.33 and an HTTP (Web) server on 192.168.1.34. Both
protocols use TCP; SMTP uses port 25 and HTTP uses port 80, so the rules match TCP packets with destination
ports of 25 and 80. The example assumes the mail server does not make outgoing connections and uses another
server on the LAN for DNS and mail relaying. Both commands put rules in the PREROUTING chain of the NAT
table (–A PREROUTING –t NAT):
# iptables -A PREROUTING -t NAT -p tcp --dport 25 --to-source 192.168.0.33:25 -j DNAT
# iptables -A PREROUTING -t NAT -p tcp --dport 80 --to 192.168.0.34:80 -j DNAT
To match these rules, the packet must use the TCP protocol (–p tcp) and have a destination port of 25 (first rule,
--dport 25) or 80 (second rule, --dport 80).
The --to-source is a target specific to the PREROUTING and OUTPUT chains of the NAT table; it alters the destination address and port of matched packets as specified. As with MASQUERADE and SNAT, subsequent packets
in the same and related connections are appropriately altered.
The fact that the servers cannot originate connections means that neither server can be exploited to participate in
a DDoS attack (page 966) on systems on the Internet and cannot send private data from the local system back to
a malicious user’s system.
Chapter Summary
The iptables utility is used to set up firewalls that help to prevent unauthorized access to a system or network. An
iptables command sets up or maintains in the kernel rules that control the flow of network packets; rules are stored
in chains. Each rule has a criteria part and an action part, called a target. When the criteria part matches a network
packet, the kernel applies the action from the rule to the packet.
There are three tables that hold chains: Filter, NAT, and Mangle. Filter, the default table, DROPs or ACCEPTs
packets based on their content. NAT, the Network Address Translation table, translates the source or destination
field of packets. Mangle is used exclusively to alter TOS (Type of Service), TTL (Time To Live), and MARK fields
in a packet. The connection tracking machine, handled by the conntrack module, defines rules that match on the
state of the connection a packet is part of.
In an emergency you can give the following command to unload all iptables modules from the kernel and set a
policy of DROP for all tables.
# /sbin/service iptables panic
Exercises
1. How would you remove all iptables rules and chains?
2. How would you list all current iptables rules?
3. How is configuring iptables different from configuring most Linux services?
150
Draft
Advanced Exercises
Draft
4. Define an iptables rule that will reject incoming connections on the TELNET port.
5. What does NAT stand for? What does the NAT table do?
Advanced Exercises
6. What does the conntrack module do?
7. What do rule match criteria do? What are they used for?
8. What do packet match criteria do? What are they used for?
9. What utilities copy packet filtering rules to and from the kernel? How do they work?
10. Define a rule that will silently block incoming SMTP connections from spmr.com.
151
152
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 13 - Downloading and Installing Software
Downloading and Installing Software
IN THIS CHAPTER
system-config-packages: Adds and Removes Software
Packages
rpm: Red Hat Package Manager
Keeping Software Up-to-Date
Keeping the System Up-to-Date
Red Hat Network (RHEL)
yum: Updates and Installs Packages
Apt: An Alternative to yum
BitTorrent
wget: Download Files Noninteractively
A software package is the collection of scripts, programs, files, and directories required to run a software application,
including system software. Using packages makes it easier to transfer, install, and uninstall applications. A package
contains either executable files or source code files that you need to compile and install. Executable files are precompiled for a specific processor and operating system, whereas source files need to be compiled but will run on
a wide range of machines and operating systems.
Software for your system can come in different kinds of packages, such as rpm (page 455), the GNU Configure and
Build System (page 459), tar, compressed tar, and others. The most popular package is rpm. Other packages (such
as tar), which were popular before the introduction of rpm, are used less now because they require more work on
the part of the installer (you) and do not provide the depth of prerequisite and compatibility checking that rpm
offers. Newer programs such as yum (page 469) and Apt (page 472) not only check for compatibility but obtain
over the Internet additional software required to install a given software package.
system-config-packages: Adds and Removes Software Packages
Red Hat has made the process of adding and removing software packages that they supply much easier with the
system-config-packages (FEDORA) and redhat-config-packages (RHEL) package management utilities (Figure 131). These are the same tools you use during installation when you choose to select packages manually.
153
Draft
system-config-packages: Adds and Removes Software Packages
Draft
Figure 13-1. Package Management window
These utilities divide software packages into five categories: Desktops, Applications, Servers, Development, and
System. Within each category are package groups. Figure 13-1 shows the Desktops category and a little bit of the
Applications category. The Desktops category has three package groups. The pair of numbers in brackets to the
left of each of the Details buttons tells you how many packages are (to be) installed and how many are available.
Before you make any changes, the first of each of these pairs of numbers tells you how many packages are installed;
after you make changes, it tells you how many packages will be installed after you click Update.
Boxes to the left of the names of the package groups, such as the one next to Editors in the Application category,
indicate whether a group is selected (with a check mark) or not (no check mark). Click a box to add or remove a
package group.
The Desktop package groups cannot be removed (there are no boxes to the left of the group names), although they
can be added if they were not installed when the system was first configured. All other package groups can be removed.
Once a package group is selected, you can click Details to display the Package Details window. Figure 13-2 shows
the Details window for the KDE package group. Within a package group are two sets of packages: those that are
standard when you select the group (Standard Packages) and those that are optional (Extra Packages). Click the
triangle to the left of the name of the package set to hide or display the list of packages in that set. Put check marks
in the boxes next to the names of the Extra Packages you want on the system. Remove check marks next to Extra
Packages you do not want. Click Close to return to the Package Management window when you finish selecting
packages. Click Update to install/remove the packages you specified. The package manager asks you for the installation CDs it needs.
154
Draft
rpm: Red Hat Package Manager
Draft
Figure 13-2. Package Management, Details window
rpm: Red Hat Package Manager
The rpm (Red Hat Package Manager) utility works only with software packages that have been built for processing
by rpm; it installs, uninstalls, upgrades, queries, and verifies rpm packages. Because Red Hat released rpm under
the GPL (page 4), rpm is used by several distributions. Together with information contained within software
packages assembled for use with rpm (named *.rpm), the rpm utility keeps track of where software packages should
be installed, the versions of the packages that you have installed, and the dependencies between the packages.
Source rpm packages are frequently found in a directory named SRPMS (source rpms), whereas binary rpm packages
are in RPMS. When you download binary packages, make sure that they are relevant to your operating system
(both distribution and release—for example, Fedora Core 2) and were compiled on the appropriate architecture
(i386 covers all Intel- and most AMD-based systems; i586 covers Pentium-class processors and above; i686 refers
to Pentium II or better, and includes MMX extensions; S390 is for IBM System/390; ia64 is for the 64-bit Intel
processor; alpha is for the DEC/Compaq Alpha chip; athlon denotes the AMD Athlon family; x86-64, x86_64,
and AMD64 for AMD64 technology; ppc is for the Power PC chip; and sparc covers the Sun Sparc processor).1
The name of the rpm file contains almost all the necessary information. Each of the following lines from a search
for sendmail on www.rpmfind.net [http://www.rpmfind.net] gives you the information you need:
1
Many rpm packages run on releases and even distributions other than the ones they were compiled on/for. However, installing packages intended
for other distributions can create problems. In particular, Mandrake packages are rarely compatible with other systems.
155
Draft
rpm: Red Hat Package Manager
Draft
sendmail-8.12.10-1mdk.src.html ...
Mandrake Cooker
sendmail-8.12
.10-1mdk.src.rpm
sendmail-8.12.10-1.src.html ...
PLD Linux Distribution
sendmail-8.12
.10-1.src.rpm
sendmail-8.12.9-7mdk.src.html ...
Mandrake
sendmail-8.12
.9-7mdk.src.rpm
sendmail-8.12.9-7mdk.i586.html ...
Mandrake
sendmail-8.12
.9-7mdk.i586.rpm
sendmail-8.12.8-9.90asp.i386.html...
ASPLinux
sendmail-8.12.8-9
.90asp.i386.rpm
...
RedHat-9 Updates for i386
sendmail-8.12.8-9
...
Red Hat Linux 9 Updates Sources
sendmail-8.12.8-9
...
RedHat-8.0 Updates for i386
sendmail-8.12.8-9
.90.i386.rpm
.90.src.rpm
.80.i386.rpm
sendmail-8.12.8-5.90asp.src.html...
ASPLinux
.90asp.src.rpm
sendmail-8.12.8-5.90.src.html...
Netwinder.Org
sendmail-8.12.8-5
...
SuSE Linux 8.2 Updates for i386
sendmail-8.12
...
SuSE Linux 8.1 Updates for i386
sendmail-8.12
...
RedHat-8.0 for i386
sendmail-8.12.5-7
...
Yellow Dog PPC
sendmail-8.11
sendmail-8.12.8-5
.90.src.rpm
.7-77.i586.rpm
.6-159.src.rpm
.i386.rpm
sendmail-8.11.6-3a.src.html
.6-3a.src.rp
Click the html filename at the left to display the information about the file. Not all packages have an HTML description file. Click the rpm filename at the right to download the file. Both of these names tell you the name of
the program, its version number, and its format (source or compiled for i386, alpha, ia64, and so on). The column
to the left of the rpm filename tells you which distribution the file is from/for.
Packages marked noarch, short for no architecture, contain resources, such as images or scripts, which are run by
an interpreter. You can install and run noarch packages on any architecture. Another site to search for rpm (and
non-rpm) files is www.rpmseek.com [http://www.rpmseek.com].
156
Draft
Querying Packages and Files
Draft
Querying Packages and Files
The rpm utility can be run from a command line. Use rpm –qa to get a list of one-line summaries of all packages
installed on the system (any user can run this utility). Use rpm –q, followed by the name of the package, to get
more information about a particular package. For instance, rpm –q nis tells you whether NIS is installed and if so,
which version. Use the –ql options to get a list of files in a package:
$ rpm -q nis
package nis is not installed
$ rpm -ql logrotate
/etc/cron.daily/logrotate
/etc/logrotate.conf
/etc/logrotate.d
/usr/sbin/logrotate
/usr/share/doc/logrotate-3.7
/usr/share/doc/logrotate-3.7/CHANGES
/usr/share/man/man8/logrotate.8.gz
/var/lib/logrotate.status
When you run rpm with the –qi options, you get quite a bit of information about a package:
$ rpm -qi logrotate
Name
: logrotate
Relocations: (not relocatable)
Version
: 3.7
Vendor: Red Hat, Inc.
Release
: 4.1
Build Date: Sun Feb 15 05:58:51 2004
Install Date : Tue Mar 30 18:34:53 2004
Build Host: tweety.devel.redhat.com
Group
: System Environment/Base
Source RPM: logrotate-3.7-4.1.src.rpm
Size
: 46526
License: GPL
Signature
: DSA/SHA1, Wed Mar 17 11:20:06 2004, Key ID da84cbd430c9ecf8
Packager
: Red Hat, Inc. <http://bugzilla.redhat.com/bugzilla>
Summary
: Rotates, compresses, removes and mails system log files.
Description :
The logrotate utility is designed to simplify the administration of
log files on a system which generates a lot of log files. Logrotate
allows for the automatic rotation compression, removal and mailing of
log files. Logrotate can be set to handle a log file daily, weekly,
monthly or when the log file gets to a certain size. Normally,
logrotate runs as a daily cron job.
Install the logrotate package if you need a utility to deal with the
log files on your system.
Installing, Upgrading, and Removing Packages
You can use rpm to install or upgrade a package. Log in as, or su to, root. (Although you can run rpm as a nonprivileged user, you will not have permission to write to the necessary directories during an install or uninstall, and the
install/uninstall will fail. During a query, you do not need this permission, so you can and should work as a nonprivileged user.) Use the –U option, followed by the name of the file that contains the rpm version of the package
you want to install. The –U option upgrades existing packages and installs new packages (as though you had used
the –i option). For kernels, use –i, not –U in order to leave the old kernel intact when you install a new kernel. Add
the –v (verbose) option to get more information about what is going on and the –h (or --hash) option to see hash
157
Draft
Installing a Linux Kernel Binary
Draft
marks as the package is unpacked and installed. For example, while logged in as root, use the following command
to add samba to your system:
# rpm -Uvh samba-3.0.0-15.i386.rpm
warning: samba-3.0.0-15.i386.rpm: V3 DSA signature: NOKEY, key ID 4f2a6fd2
Preparing...
############################################# [100%]
1:samba
############################################# [100%]
When you install a package, your working directory must be the directory that contains the rpm file, or you must
use a pathname that points to the file.
To remove the same package, give the following command from any directory:
# rpm -e samba
The rpm utility queries its database to find out the information it needs to uninstall the package and removes links,
unloads device drivers, and stops daemons as necessary. Refer to the rpm man page for more rpm options.
Installing a Linux Kernel Binary
The following steps install a new Linux kernel binary. Refer to Chapter 15 when you want to configure and rebuild
a kernel, not to install a new, prebuilt kernel binary. Rebuilding a kernel is more involved than installing a new
one.
1.
Run rpm with the –i option to install the new kernel. Do not use the –U option. You are installing a new kernel
that has a name different from the old kernel: You are not upgrading the existing kernel.
2.
Make sure the new kernel works before you remove the old kernel. To verify that the new kernel works, reboot
the system using the new kernel.
3.
Remove the old kernel by removing the files that contain the release number (and EXTRAVERSION number
[page 511], if applicable) in their filenames from /boot or / (root). Remove information about the old kernel
from lilo.conf or grub.conf. Remember to run lilo if you modify lilo.conf (RHEL only). You may want to
wait a while before removing the old kernel to make sure that there are no problems with the new one. Instead
of removing the old kernel manually, you may be able to remove it using the tool you used to install it (rpm,
yum, Apt, or other).
Installing Non-rpm Software
Most software that does not come in rpm format comes with detailed instructions on how to configure, build (if
necessary), and install it. Some binary distributions (those containing prebuilt executables that run on Red Hat)
require you to unpack the software from the root directory.
The /opt and /usr/local Directories
Some newer application packages include scripts to install themselves automatically into a directory hierarchy under
/opt, with files in a /opt subdirectory that is named after the package and executables in /opt/bin or /opt/ package
/bin. These scripts are relatively new additions to Red Hat but are familiar to Sun Solaris users.
Other software packages allow you to choose where you unpack them. Because the software available for Linux is
developed by many different people, there is no consistent method for doing installations. As you acquire local
158
Draft
GNU Configure and Build System
Draft
software, you should install it on your system in as consistent and predictable a manner as possible. The standard
Linux file structure has a directory hierarchy under /usr/local for binaries (/usr/local/bin), manual pages
(/usr/local/man), and so forth. To prevent confusion later and to avoid overwriting or losing the software when
you install standard software upgrades in the future, avoid installing nonstandard software in standard system
directories (such as /usr/bin). On a multiuser system, make sure that users know where to find the local software,
and make an announcement whenever you install, change, or remove local tools.
GNU Configure and Build System
The GNU Configure and Build System makes it easy to build a program that is distributed as source code (see
autoconf at developer.gnome.org/tools/build.html). This two-step process does not require special tools other than
a shell, make, and gcc (the GNU C compiler). You do not need to work with root privileges for either of these steps.
The following example assumes you have downloaded the GNU chess program (www.gnu.org/software/chess/chess.html [http://www.gnu.org/software/chess/chess.html]) to the working directory. First, unpack
and decompress the file and cd to the new directory:
$ tar -xvzf gnuchess*
gnuchess-5.03/
gnuchess-5.03/book/
gnuchess-5.03/book/README
...
$ cd gnuchess*
After reading the README and INSTALL files, run the configure script, which finds out about your system and
generates the Makefile file:
$ ./configure
checking for a BSD compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for mawk... mawk
checking whether make sets ${MAKE}... yes
checking for gcc... gcc
checking for C compiler default output... a.out
checking whether the C compiler works... yes
...
checking for memset... yes
configure: creating ./config.status
config.status: creating Makefile
config.status: creating src/Makefile
config.status: creating src/config.h
Refer to the configure info page, specifically the --prefix option, which causes the install phase to place the software
in a directory other than /usr/local. The second step is to run make.
$ make
Making all in src
make[1]: Entering directory '/hdd4/gnuchess-5.03/src'
cd .. \
&& CONFIG_FILES= CONFIG_HEADERS=src/config.h \
/bin/sh ./config.status
159
Draft
Keeping Software Up-to-Date
Draft
config.status: creating src/config.h
config.status: src/config.h is unchanged
make all-am
make[2]: Entering directory '/hdd4/gnuchess-5.03/src'
source='atak.c' object='atak.o' libtool=no \
depfile='.deps/atak.Po' tmpdepfile='.deps/atak.TPo' \
depmode=gcc3 /bin/sh ../depcomp \
gcc -DHAVE_CONFIG_H -I. -I. -I.
-g -O2 -c 'test -f atak.c || echo './''atak.c
.
.
gcc -g -O2 -o gnuchess atak.o book.o cmd.o epd.o eval.o genmove.o hash.o hung.o init.o
iterate.o main.o move.o null.o output.o players.o pgn.o quiesce.o random.o repeat.o
search.o solve.o sort.o swap.o test.o ttable.o util.o version.o -lreadline -lncurses -lm
make[2]: Leaving directory '/hdd4/gnuchess-5.03/src'
make[1]: Leaving directory '/hdd4/gnuchess-5.03/src'
make[1]: Entering directory '/hdd4/gnuchess-5.03'
make[1]: Nothing to be done for 'all-am'.
make[1]: Leaving directory '/hdd4/gnuchess-5.03'
$ ls src/gnuchess
src/gnuchess
After make finishes, the gnuchess executable is in the src directory. If you want to install it, give the following
command while running with root privileges:
# make install
Making install in src
make[1]: Entering directory '/hdd4/gnuchess-5.03/src'
make[2]: Entering directory '/hdd4/gnuchess-5.03/src'
/bin/sh ../mkinstalldirs /usr/local/bin
/usr/bin/install -c gnuchess /usr/local/bin/gnuchess
make[2]: Nothing to be done for 'install-data-am'.
...
You can run the two steps and install the software with this command line:
# ./configure && make && make install
The Boolean operator && (AND) allows the execution of the next step only if the previous step returned a successful
exit status.
Keeping Software Up-to-Date
Of the many reasons to keep the software on your system up-to-date, one of the most important is security. Although
you hear about software-based security breaches, you do not hear about the fixes that were available but never installed before the breach. Timely installation of software updates is critical to system security. Linux Open Source
software is the ideal environment to find and fix bugs and make repaired software available quickly. When you
keep your system and application software up-to-date, you keep abreast of bug fixes, new features, support for
new hardware, speed enhancements, and more.
160
Draft
Bugs
Draft
Bugs
A bug is an unwanted and unintended program property, especially one that causes the program to malfunction
(definition courtesy www.foldoc.org [http://www.foldoc.org]). Bugs have been around forever, in many types of
systems, machinery, thinking, and so on. All sophisticated software contains bugs. Bugs in system software or application packages can crash the system or cause programs not to run correctly. Security holes (a type of bug) can
compromise the security of the system, allowing malicious users to read and write files, send mail to your contacts
in your name, or destroy all the data on the system, rendering the system useless. If the engineers fixed all the bugs,
there would still be feature requests as long as anyone used the software. Bugs, feature requests, and security holes
are here to stay; they must be properly tracked if developers are to fix the most dangerous/important bugs first,
users are to research and report bugs in a logical manner, and administrators are to be able to apply the developer’s
fixes quickly and easily.
Early on, Netscape used an internal bug-tracking system named BugSplat. Later, after Netscape created Mozilla
(mozilla.org) as an Open Source browser project, the Mozilla team decided that it needed its own bug-tracking
system. Netscape’s IS department wrote a very short-lived version of Bugzilla. Then Terry Weissman, who had
been maintaining BugSplat, wrote a new, Open Source version of Bugzilla in Tcl, rewriting it in Perl a couple of
months later.
Bugzilla belongs to a class of programs formally known as Defect Tracking Systems, of which Bugzilla is now
preeminent. It is the tool that almost all Linux developers use to track problems and enhancement requests for
their software. Red Hat uses Bugzilla to track bugs and bug fixes for its Linux distributions; Red Hat Network takes
advantage of Bugzilla to notify users of and distribute these fixes.
Errata
For both Red Hat Enterprise Linux and Fedora, Red Hat processes security, bugfix, and new feature (enhancement)
updates. The easiest way to learn about new updates and to obtain and install them is to use up2date (page 462).
As the Linux community, including Red Hat, finds and fixes operating system and software package bugs, including
security holes, Red Hat generates rpm files (page 455) that contain the code that fixes the problems. Installing a
new version of a package is almost the same as installing the package in the first place. When running rpm, use the
–U (upgrade) option, not –i. The upgrade option is the same as the install option except that it removes the earlier
version of the package. When you upgrade a system software package, rpm renames modified configuration files
with a .rpmsave extension. You must manually merge the changes you made to the original files into the new files.
RHEL . Lists of errata for all recent releases of Red Hat Enterprise Linux are available at www.redhat.com/security [http://www.redhat.com/security]. However, if you are running RHEL, you probably have a subscription
to RHN (page 467) and can use this service to find and download updates.
FEDORA . For information on Fedora Core updates, point a browser at fedora.redhat.com and click Download
Updates. Information about updates is posted to the Fedora Announce List (www.redhat.com/mailman/listinfo/fedora-announce-list [http://www.redhat.com/mailman/listinfo/fedora-announce-list]). You can use up2date, yum,
or Apt to find and download updates.
Fedora, Red Hat Enterprise Linux, Red Hat Network, and up2date
Before Red Hat introduced Fedora, RHN (Red Hat Network), for a fee, provided updates for software on Red Hat
systems (page 467). The tool used to download and install the updates was up2date (page 462), which used yum/Aptlike tools to download rpm files. The round button on the panel that changed colors to let you know when updates
were available was called the RHN Alert Notification Tool (page 466).
161
Draft
Fedora, Red Hat Enterprise Linux, Red
Hat Network, and up2date
Draft
When Red Hat split its product line, RHN, a Red Hat profit center, went with Red Hat Enterprise Linux. The
up2date utility went with both Red Hat Enterprise Linux and Fedora. Although up2date remained the same, its
configuration files changed: Under Fedora, up2date uses yum (and optionally Apt) to download updated rpm files
from the Fedora site. Also under Fedora, the round button that lets you know when updates are available has the
same name and works the same way, except when it calls up2date (up2date does not use RHN). The letters rhn
appear in Fedora filenames (for example, /etc/sysconfig/rhn), even though Fedora does not use RHN. See page
469 for information on yum and page 472 for Apt.
up2date: Updates Packages
The up2date utility downloads and optionally installs rpm packages. It works with many files and directories, in
graphical and character-based modes, and has many options.
RHEL .
FEDORA .
positories.
The up2date utility works with the RHN server.
The up2date utility uses yum (and optionally Apt) to download files from the Fedora and other re-
The --configure option generates /etc/sysconfig/rhn/up2date, up2date’s system profile file. The up2date-config
utility (next section) is a link to up2date with the --configure option. You do not normally use this option because
up2date configures itself (creates the up2date system profile) when necessary. The --nox option (also up2date-nox)
runs up2date in character-based mode. Refer to the up2date man page for more information.
In addition to updating packages on the system, up2date can download and install Red Hat packages that are not
on the system. In the following example, a Fedora user calls links, the character-based browser program, finds it
is not on the system, and confirms that finding with whereis. Then, up2date, with the --whatprovides option,
queries the Fedora repository to find that the elinks package provides links. Finally, up2date, with an argument of
the rpm package to be installed, downloads the elinks package. In this case, up2date installs the package because
that is what the up2date profile is set up to do. The same commands work under Red Hat Enterprise Linux when
you are subscribed to RHN. You must run up2date as Superuser to install or upgrade a package.
# links
bash: links: command not found
# whereis links
links:
# up2date --whatprovides links
http://fedora.redhat.com/download/up2date-mirrors/fedora-core-2
using mirror: http://ftp.dulug.duke.edu/pub/fedora/linux/core/development/i386/
elinks-0.9.1-1
# up2date elinks
http://fedora.redhat.com/download/up2date-mirrors/fedora-core-2
using mirror: http://ftp.dulug.duke.edu/pub/fedora/linux/core/development/i386/
Fetching Obsoletes list for channel: fedora-core-2...
Fetching rpm headers...
########################################
Name
Version
Rel
---------------------------------------------------------elinks
0.9.1
1
Testing package set / solving RPM inter-dependencies...
162
i386
Draft
Fedora, Red Hat Enterprise Linux, Red
Hat Network, and up2date
Draft
########################################
Preparing
########################################### [100%]
Installing...
1:elinks
########################################### [100%]
When you give it a command, up2date determines where to look for the file you request by looking at the /etc/sysconfig/rhn/sources configuration file. In the preceding example, up2date says it is looking in fedora.redhat.com/download/up2date-mirrors/fedora-core-2, a list of Fedora repository mirrors. Then up2date displays
the URL of the repository it is using following the words using mirror.
up2date-config: Configures up2date
The up2date-config utility sets parameters in /etc/sysconfig/rhn/up2date, the up2date profile file. You can run
up2date-config from a command line, but it is not usually required because up2date configures itself as necessary
the first time you run it. In a graphical environment, this tool displays a window with three tabs: General, Retrieval/Installation (Figure 13-3), and Package Exceptions. See Table 13-1.
Figure 13-3. Configuring up2date, Retrieval/Installation tab
163
Draft
Fedora, Red Hat Enterprise Linux, Red
Hat Network, and up2date
Draft
Table 13-1. Configuring up2date
General/Network Settings
Select a Red Hat Network Server to use
This text box is already filled in. Do not change it unless
you have reason to do so. Works for RHN (RHEL), Fedora, and other repositories.
Enable HTTP Proxy
If you need to use a proxy server, enter the HTTP proxy
server in the required format.
Use Authentication
Select Use Authentication and fill in the Username and
Password text boxes when the proxy server requires authentication. These spaces are for the proxy server, not
for the RHN user-name and password.
Retrieval/Installation
Package Retrieval Options
Do not install packages after retrieval
Download, but do not install packages. You will need to
install the new packages manually.
Do not upgrade packages when local configuration file Do not download or install packages that have been
has been modified
customized. This option is not necessary unless you are
using other than the standard Red Hat packages.
Retrieve source RPM along with binary package
Download the source code (*.src.rpm) file in addition
to the binary file (*.arch.rpm) that is to be installed. The
up2date utility does nothing with the source file except
download it.
Package Verification Options
Use GPG to verify package integrity
Uses Red Hat’s GPG signature to verify the authenticity
of the files you are downloading. If the Red Hat signature
is not on the local system, up2date asks if you want the
system to download it for you. This is a critical security
link; it is a good idea to select this option.
Package Installation Options
After installation, keep binary packages on disk
Normally, binary rpm files are removed once the files
they contain have been installed. Select this option if you
want them left on the system in the Package storage directory (following).
Enable RPM rollbacks (allows “undo” but requires addi- By using extra disk space, up2date can store information
tional storage space)
so it can uninstall a package it has installed and reinstall
the version that was installed previously.
Override version stored in System Profile
Downloads and installs packages for a version of Red
Hat that you specify in the text box, overriding the version number that is stored in the system profile.
Package storage directory
Specifies a directory to store the downloaded files in. By
default, they are stored in /var/spool/up2date.
Package Exceptions
Specifies packages and files that you do not want to
download. These names can include wildcard characters.
164
Draft
Fedora, Red Hat Enterprise Linux, Red
Hat Network, and up2date
Draft
General/Network Settings
Package Names to Skip
By default kernel* appears in this list box, meaning that
no rpm packages whose names begin with the letters
kernel will be downloaded. Installing a new kernel is an
important event, and Red Hat assumes you do not want
this to happen without your knowledge. Use the Add,
Edit, and Remove buttons to adjust the list box to meet
your requirements. Normally, you do not have to make
any changes here.
File Names to Skip
Similar to Package Names to Skip except you can specify
filenames you want to skip here.
Tip: (FEDORA) When You Do Not Want up2date to Run RHN
When the following line appears uncommented anywhere in /etc/sysconfig/rhn/sources, up2date
tries to contact RHN and fails; up2date ignores other repositories listed in the file:
up2date default
Registering with RHN can put this line in sources, but up2date still will not work. Remove this line
or comment it out by putting a # at the beginning of the line to cause up2date to contact Fedora or
other repositories.
Red Hat Network Alert Notification Tool
RHEL+FEDORA . The Red Hat Network Alert Notification Tool can take care of everything you need to do
from your system to set up and run up2date to keep a system up to date. Although its title includes the words Red
Hat Network, you can use it on a Fedora system to control up2date without using RHN. The Red Hat Network
Alert Notification Tool is represented by a round button on both the GNOME and KDE Main panel. It shows one
of four icons:
•
Blue with a check mark indicates that all is well: There are no pending downloads.
•
Red with an exclamation point indicates that files need to be downloaded. Click the button to display the
window shown in Figure 13-4.
165
Draft
Red Hat Network (RHEL)
Draft
Figure 13-4. RHN Alert Notification Tool, Available Updates tab
•
Green with half arrows pointing left and right indicates that the system is communicating with the server.
•
Grey with a question mark indicates that there has been an error. Click the icon to display the error message.
If the button is not on the Main panel, run rhn-applet-gui from Run Application on the GNOME Main menu or
Run Command on the KDE Main menu to display it. Table 13-2 describes the selections in the Red Hat Network
Alert Notification Tool panel icon menu (right click).
Table 13-2. Red Hat Network Alert Notification Tool Icon Menu
Check for updates
Runs up2date (page 462) in the background to check for
updates. The green icon with arrows on the Red Hat
Network Alert Notification button shows that the system
is communicating with the server.
Launch up2date
Runs up2date (page 462) in the foreground, opening a
series of windows that does not give you many options.
Configuration
Opens a series of windows that displays the Terms of
Service, allows you to configure a proxy, and checks for
updates.
RHN Website
Opens a Mozilla window displaying the RHN Website.
Red Hat Network (RHEL)
RHN (Red Hat Network, rhn.redhat.com), a service provided by Red Hat, is an Internet-based system that can
keep the software on one or more of your Red Hat Enterprise Linux systems up-to-date with minimal work on your
part. You must subscribe to this service in order to use it. Red Hat uses the term entitle to indicate that a system is
subscribed to RHN: Your system must be entitled before you can use RHN. To date, Red Hat allows each user to
register one system for free. You can choose to make RHN more or less automated, giving you various degrees of
control over the update process.
The systems that are entitled are the clients; Red Hat maintains the RHN server. The RHN Server is much more
than a single server; it involves many systems and databases that are replicated and located in different areas. For
166
Draft
Red Hat Network (RHEL)
Draft
the purpose of understanding how to use the client tools on your system, picture the RHN Server as a single server.
For additional information, refer to the Red Hat Network manuals at www.redhat.com/docs/manuals/RHNetwork
[http://www.redhat.com/docs/manuals/RHNetwork].
Red Hat built RHN with security as a priority. Any time you allow a remote system to put a program on your system
and run it, the setup must be very close to the theoretical ideal of absolutely secure. Toward this end, RHN never
initiates communication with your system. Once a program running on your system sends a message to the RHN
server, the server can respond and your system can trust the response.
Subscribing to Red Hat Network (RHEL)
Perform the following tasks to subscribe to and start using RHN.
•
•
Give the command rhn_register to open the RHN registration window.
•
Read the Welcome page and click Forward.
•
Step 1: Review the Red Hat Privacy Statement window opens. If you are comfortable with the information
in this window, click Forward.
•
Step 2: Login window opens. Choose whether you want to Create New Account or Use Existing Account.
Fill in the requested information. Click Forward.
•
Step 3: Register a System Profile - Hardware window opens. This window establishes a profile name for
your computer. The default name is your node (system) name. You can put any information to help you
identify this system in this text box. The window also confirms that you want to include the hardware and
network information in the profile information that you will send to RHN at the end of this process. When
you click Forward, the program compiles a list of the rpm packages installed on your system.
•
Step 3: Register a System Profile - Packages window opens (there are two Step 3 windows). This window
confirms that you want to include a list of installed rpm packages as part of the system profile and gives you
the option of removing packages from the profile so they do not get updated. Click Forward.
•
Click Forward to send your profile to RHN; or click Cancel, and nothing will be sent.
Entitle the system. The error: Service not enabled for server profile: “pro-filename” means that your system
is not entitled. If you get this message, go to rhn.redhat.com, and log in with the username and password you
set up in Step 2.
Click the Systems tab at the top of the page, and then click the System Entitlements box at the left. The system
you just registered should be listed. Change the entry in the combo box (in the Entitlement column) on the line
with the proper system name to Update. You can click Buy them now at the top of the page to buy an entitlement, if necessary.
•
Check for updates. Run up2date, or choose Red Hat Network Alert Notification Tool Icon menu: Check
for updates (page 466) to see if the RHN server downloads files to or exchanges information with the local
system. Alternatively, give the command up2date --list to see whether any packages are available for the system,
thereby testing your connection with the RHN server.
You can start the flow of updates from the system or from the Web site. From the system, run up2date. From
the Web site, log in, click the Systems tab, click the name of the system in the table, click update now if applicable, and click Apply Errata. In either case, the next time the rhnsd daemon (next) on the local system contacts
the RHN server, you will get updates per the up2date profile (installed or not, left on the system or not, source
code or not, and so on).
167
Draft
Keeping the System Up-to-Date
Draft
rhnsd: RHN Daemon
The RHN daemon (rhnsd) is a background service that periodically queries the RHN server to find out whether
any new packages are available to be downloaded. This daemon is one of the keys to RHN security; it is the component that initiates contact with the RHN server so the server never has to initiate contact with the local system.
Refer to “ service: Configures Services I” on page 383 to start, stop, or display the status of rhnsd immediately;
refer to “ system-config-services: Configures Services II” on page 383 or to “ chkconfig: Configures Services III” on
page 385 to start or stop rhnsd at specified runlevels.
Keeping the System Up-to-Date
Apt and yum both fill the same role: They install and update software packages. Apt is slightly faster, especially on
slow connections, and supports a few more features, such as undoing upgrades. The yum utility is installed by default
and is easier to configure and use than Apt. If you are familiar with Debian systems or find yum lacks some features
you need, try using Apt; otherwise use yum.
yum: Updates and Installs Packages
Early releases of RedHat Linux did not include a tool for managing updates. The RPM tool could install or upgrade
individual software packages, but it was up to the user to locate the packages and the packages they were dependent
on. When Terra Soft produced their RedHat-based Linux distribution for the PowerPC, they created the Yellow
Dog Updater to fill this gap. This program has since been ported to other architectures and distributions. The result,
Yellow Dog Updater, Modified (yum), is included with Fedora Core.
Configuring yum
The yum utility is designed to be easy to use. The configuration file, /etc/yum.conf, has two parts: The [main]
section contains general settings and the rest of the file holds a list of servers.
The [main] section must be present for yum to function. The cachedir specifies the directory yum should store
downloaded packages in and logfile specifies where yum keeps its log. The amount of information logged is specified
by debuglevel, with a value of 10 producing the most information.
$ cat /etc/yum.conf
[main]
cachedir=/var/cache/yum
debuglevel=2
logfile=/var/log/yum.log
pkgpolicy=newest
distroverpkg=fedora-release
tolerant=1
exactarch=1
...
The pkgpolicy defines which version of a software package yum installs; always set to newest to install the newest
version of a package. You can also configure yum to try to install from a specific server, falling back to other servers
on failure, ignoring package versions. The distroverpkg specifies the distribution the system is running. You should
not need to change this setting.
With tolerant set to 1, yum automatically corrects simple command line errors, such as attempting to install a
package already on the system. Setting tolerant to 0 tuns this feature off. Setting exactarch to 1 causes yum to
168
Draft
yum: Updates and Installs Packages
Draft
update packages only with packages of the same architecture, preventing an i686 package from replacing an i386
one, for example.
The last three sections contain lists of servers holding updates. The first, [base], contains the packages present on
the installation CDs. The second, [updates-released], contains updated versions of packages considered to be stable,
and the last, [updates-testing], contains updates that are not ready for release. The last section is commented out;
do not uncomment it unless you are testing unstable packages. Never uncomment on production systems.
$ cat /etc/yum.conf
...
[base]
name=Fedora Core $releasever - $basearch - Base
baseurl=http://fedora.redhat.com/releases/fedora-core-$releasever
[updates-released]
name=Fedora Core $releasever - $basearch - Released Updates
baseurl=http://fedora.redhat.com/updates/released/fedora-core-$releasever
#[updates-testing]
#name=Fedora Core $releasever - $basearch - Unreleased Updates
#baseurl=http://fedora.redhat.com/updates/testing/fedora-core-$releasever
Each server section contains a name and a baseurl. The name provides a friendly name for the server and is displayed
by up2date when listing possible servers. The baseurl indicates the location of the server. These definitions use
two variables: yum sets $basearch to the architecture of the system and $releasever to the version of the release.
Refer to the yum.conf man page for more options.
Using yum
Working as root, you can run yum from a command line. The behavior of yum depends on the options you specify.
The update option updates all installed packages:
It downloads package headers for installed packages, prompts you to proceed, and downloads and installs the
updated packages.
# yum update
Gathering header information file(s) from server(s)
Server: Fedora Core 2 - i386 - Base
Server: Fedora Core 2 - i386 - Released Updates
Finding updated packages
Downloading needed headers
getting /var/cache/yum/updates-released/headers/pango-0-1.2.54.i386.hdr
pango-0-1.2.5-4.i386.hdr 100% |=========================| 6.5 kB 00:00
...
[update: rhn-applet 2.1.4-3.i386]
Is this ok [y/N]: y
Getting pango-1.2.5-4.i386.rpm
pango-1.2.5-4.i386.rpm
100% |=========================| 341 kB 00:06
...
You can update individual packages by specifying the names of the packages on the command line following the
word update.
169
Draft
Apt: An Alternative to yum
Draft
To install a new package together with the packages it is dependent on, give the command yum install, followed
by the name of the package:
# yum install tcsh
Gathering header information file(s) from server(s)
Server: Fedora Core 1 - i386 - Base
Server: Fedora Core 1 - i386 - Released Updates
Finding updated packages
Downloading needed headers
getting /var/cache/yum/base/headers/tcsh-0-6.12-5.i386.hdr
tcsh-0-6.12-5.i386.hdr
100% |=======================| 3.8 kB
Resolving dependencies
Dependencies resolved
I will do the following:
[install: tcsh 6.12-5.i386]
Is this ok [y/N]: y
Getting tcsh-6.12-5.i386.rpm
tcsh-6.12-5.i386.rpm
100% |=======================| 443 kB
Running test transaction:
Test transaction complete, Success!
tcsh 100 % done 1/1
Installed: tcsh 6.12-5.i386
Transaction(s) Complete
00:00
00:10
You can also use yum to remove packages, using a similar syntax.
# yum remove tcsh
Gathering header information file(s) from server(s)
Server: Fedora Core 1 - i386 - Base
Server: Fedora Core 1 - i386 - Released Updates
Finding updated packages
Downloading needed headers
Resolving dependencies
Dependencies resolved
I will do the following:
[erase: tcsh 6.12-5.i386]
Is this ok [y/N]: y
Running test transaction:
Test transaction complete, Success!
Erasing: tcsh 1/1
Erased: tcsh 6.12-5.i386
Transaction(s) Complete
Apt: An Alternative to yum
The Apt (Advanced Package Tool) utility can help with the issue of dependencies: Apt tries to resolve package dependencies automatically by looking for the packages that the package you are installing is dependent on. Starting
life as part of the Debian Linux distribution using Debian’s .deb package format, Apt has been ported to rpm-based
distributions, including Red Hat; download it from ayo.freshrpms.net.
The Apt utility uses repositories of rpm files as the basis for its actions. It can use public repositories, such as those
maintained by freshrpms.com, or you can create a private repository as explained on page 478. To make things
170
Draft
Apt: An Alternative to yum
Draft
quicker, Apt keeps locally a list of packages that are on each of the repositories it uses. Any software you want to
install or update must reside in a repository.
The only connection between Apt repositories and Red Hat is that the repositories include rpm files produced by
Red Hat.
When you give Apt a command to install a package, Apt looks for the package in its local package list. If the
package is in the list, Apt fetches the package and the packages that the package you are installing is dependent on
and calls rpm to install the packages. Because Apt calls rpm, it maintains the rpm database.
Using Apt on Your System
This section shows how to configure Apt to use an external repository and how to create a local repository.
Installing and Setting Up Apt
Once you have downloaded the apt * .rpm file, install it as shown following (your Apt version number will be
different):
# rpm -Uvh apt-0.5.15cnc6-1.1.fc2.fr.i386.rpm
Preparing...
###########################################[100%]
1:apt
###########################################[100%]
Update the local package list . The primary Apt command is apt-get; its arguments determine what the command
does. After you install Apt, you need to give the command apt-get update to update the local package list:
# apt-get update
Get:1 http://ayo.freshrpms.net fedora/linux/2/i386 release [1991B]
Fetched 1991B in 0s (4922B/s)
Get:1 http://ayo.freshrpms.net fedora/linux/2/i386/core pkglist [1445kB]
Get:2 http://ayo.freshrpms.net fedora/linux/2/i386/core release [151B]
Get:3 http://ayo.freshrpms.net fedora/linux/2/i386/updates pkglist [251kB]
Get:4 http://ayo.freshrpms.net fedora/linux/2/i386/updates release [157B]
Get:5 http://ayo.freshrpms.net fedora/linux/2/i386/freshrpms pkglist [98kB]
Get:6 http://ayo.freshrpms.net fedora/linux/2/i386/freshrpms release [161B]
Fetched 1847kB in 28s (64.7kB/s)
Reading Package Lists... Done
Building Dependency Tree... Done
Because the available packages change frequently, it is a good idea to create a cron job (page 531) to update the
local package list automatically. Create the following file to perform this task daily:
$ cat /etc/cron.daily/apt-update
apt-get update
Check the dependency tree . The Apt utility does not tolerate a broken rpm dependency tree. To check the
status of your dependency tree, run apt-get check:
# apt-get check
Reading Package Lists... Done
Building Dependency Tree... Done
171
Draft
Apt: An Alternative to yum
Draft
The easiest way to fix errors that apt-get reveals is to erase the offending packages and then reinstall them using
Apt.
At the time of writing, Apt was incompatible with the Ximian Desktop.
Update the system . There are two arguments to apt-get that upgrade all packages on the system: upgrade,
which upgrades all packages on the system that do not require new packages to be installed, and dist-upgrade,
which upgrades all packages on the system, installing new packages as needed.
The following command updates all rpm-based packages on the system that depend only on packages that are
already installed:
# apt-get upgrade
Reading Package Lists... Done
Building Dependency Tree... Done
The following packages will be upgraded
bash binutils dia ethereal foomatic gaim gdm ghostscript gimp-print
...
rhn-applet rsync sed slocate strace vnc-server yum
The following packages have been kept back
gstreamer-plugins gthumb rhythmbox
57 upgraded, 0 newly installed, 0 removed and 3 not upgraded.
Need to get 59.7MB/87.9MB of archives.
After unpacking 11.8MB of additional disk space will be used.
Do you want to continue? [Y/n]
Enter Y to upgrade the listed packages; otherwise enter N. Packages that are not upgraded because they depend
on packages that are not already installed are listed as kept back.
Use dist-upgrade to upgrade all packages, including packages that are dependent on packages that are not installed.
Also installs dependencies:
# apt-get dist-upgrade
Reading Package Lists... Done
Building Dependency Tree... Done
Calculating Upgrade... Done
The following packages will be upgraded
gstreamer-plugins gthumb rhythmbox
The following NEW packages will be installed:
Hermes flac libexif libid3tag
3 upgraded, 4 newly installed, 0 removed and 0 not upgraded.
Need to get 4510kB of archives.
After unpacking 6527kB of additional disk space will be used.
Do you want to continue? [Y/n]
Adding and Removing Individual Packages
The format of a command to install a specific software package and the packages it is dependent on is shown following:
apt-get install package
172
Draft
Apt: An Alternative to yum
Draft
where package is the name of the package, such as zsh, not the name of the rpm, which usually includes version
and architecture information (for example, zsh-1.2.i386.rpm).
# apt-get install zsh
Reading Package Lists... Done
Building Dependency Tree... Done
The following NEW packages will be installed:
zsh
0 upgraded, 1 newly installed, 0 removed and 0 not upgraded.
Need to get 1435kB of archives.
After unpacking 2831kB of additional disk space will be used.
Get:1 http://ayo.freshrpms.net fedora/linux/1/i386/core zsh 4.0.7-1.1 [1435kB]
Fetched 1435kB in 21s (66.0kB/s)
Committing changes...
Preparing...
########################################### [100%]
1:zsh
########################################### [100%]
Done.
Remove a package the same way you install a package, substituting remove for install:
# apt-get remove zsh
Reading Package Lists... Done
Building Dependency Tree... Done
The following packages will be REMOVED:
zsh
0 upgraded, 0 newly installed, 1 removed and 0 not upgraded.
Need to get 0B of archives.
After unpacking 2831kB disk space will be freed.
Do you want to continue? [Y/n] y
Committing changes...
Preparing...
########################################### [100%]
Done.
So that you can later reinstall a package with the same configuration, the apt-get remove command does not remove
configuration files from the /etc directory hierarchy. Although it is not recommended, you can use the --purge
option to remove all files, including configuration files. Alternatively, you can move these files to an archive so you
can restore them if you want to.
apt.conf: Configuring Apt
The /etc/apt/apt.conf file contains Apt configuration information and is split into three sections: Apt, which
contains global settings for the Apt tools; Acquire, which describes settings related to the package-fetching mechanism; and RPM, which contains rpm specific settings. In this file, semicolons (;) separate statements, and double
forward slashes (//) introduce comments.
APT section .
The APT section is shown following:
$ cat /etc/apt/apt.conf
APT {
Clean-Installed "false";
Get {
Assume-Yes "false";
173
Draft
Apt: An Alternative to yum
Draft
Download-Only "false";
Show-Upgraded "true";
Fix-Broken "false";
Ignore-Missing "false";
Compile "false";
};
};
...
When you set Clean-Installed to TRUE, Apt removes packages that are no longer in the repository. This option
is useful if Apt is keeping a number of systems in the same state using a private repository: You can uninstall a
package from all the systems by deleting the package from the repository.
The options in the Get subsection, listed following, apply to the apt-get utility. (The apt-get utility has command
line arguments with the same names.)
Assume-Yes . TRUE runs apt-get in batch mode, automatically answering YES whenever it would otherwise
prompt you for input.
Download-Only . TRUE retrieves packages from the repository but does not install them. FALSE retrieves and
installs the packages.
Show-Upgraded .
TRUE displays a list of upgraded packages.
Fix-Broken . TRUE attempts to fix dependency tree problems with varying degrees of success. FALSE quits if
it finds a dependency tree problem.
Ignore-Missing . TRUE holds back missing or corrupt packages and continues to install other packages. FALSE
aborts the entire install or upgrade upon finding a missing or corrupt package.
Compile . TRUE compiles and installs source rpm (SRPM) packages that you ask apt-get to retrieve. FALSE
downloads these files without compiling or installing them.
Acquire section .
The Acquire section controls options related to fetching packages:
$ cat /etc/apt/apt.conf
...
Acquire {
Retries "0";
Http {
Proxy ""; // http://user:[email protected]:port/
}
};
...
The Retries option specifies the number of times Apt attempts to fetch a package when an attempt failed. The
Http Proxy setting specifies the proxy to use when fetching packages using HTTP. The argument to this option
is blank by default, indicating that Apt is not to use a proxy. An example proxy is shown as a comment.
RPM section .
Following is the RPM section of apt.conf:
$ cat /etc/apt/apt.conf
...
RPM {
174
Draft
Apt: An Alternative to yum
Draft
Ignore { };
Hold { };
Allow-Duplicated { "^kernel$"; "^kernel-"; "^kmodule-"; "^gpg-pukey$"
};
Options { };
Install-Options "";
Erase-Options "";
Source {
Build-Command "rpmbuild --rebuild";
};
};
The Ignore and Hold options perform similar functions and contain lists of packages that Apt ignores or holds
(does not upgrade) respectively. These are usually blank.
The Allow-Duplicated section lists packages that can have more than one version on the system at one time. In
general, you do not want multiple versions of the same package. The kernel is an exception because it is good
practice to leave the old kernel installed when you install a new kernel in case you are unable to boot the new one.
The Options section contains options that are passed to rpm. The Install-Options and Erase-Options sections
contain options passed to rpm whenever it is used to install or erase a package.
Finally, the Source Build-Command option specifies the command that Apt uses to build a source rpm file.
synaptic: The Apt GUI
The GNU project has produced for Apt a graphical front-end named synaptic (Figure 13-5), which you can install
using apt-get:
# apt-get install synaptic
...
Committing changes...
Preparing...
###########################################[100%]
1:synaptic
###########################################[100%]
Done.
175
Draft
Apt: An Alternative to yum
Draft
Figure 13-5. Synaptic, a graphical front end for Apt
The synaptic utility supports most of the features of the command line apt-get utility. Once installed, synaptic can
be launched from the Main menu or the command line:
# synaptic &
Each time you run synaptic, it pauses for up to several minutes as it creates a dependency tree.
The synaptic utility lists packages hierarchically, providing a nice way of browsing the packages available in the
repositories. From the hierarchy, you can select packages to upgrade, install, and remove.
Creating a Repository
If you are maintaining systems on a network, you may want to set up a private Apt repository, which has a number
of advantages over a public repository:
•
Updates do not require each system to connect to the Internet.
•
You can test packages before deploying them.
•
You can add custom packages to the repository.
An Apt repository is a directory hierarchy with a specific layout. The repository can be local, mounted using NFS
or Samba, or shared using FTP or HTTP.
Refer to freshrpms.net/apt for more information.
176
Draft
BitTorrent
Draft
BitTorrent
BitTorrent is the name of a protocol that implements a hybrid client-server and P2P (page 987) file transfer mechanism. BitTorrent efficiently distributes large amounts of static data, such as the Fedora installation ISO images
(page 35). BitTorrent can replace protocols such as anonymous FTP, where client authentication is not required.
Each BitTorrent client that downloads a file provides additional bandwidth for uploading the file, reducing the
load on the initial source. In general, BitTorrent downloads proceed faster than FTP downloads.
Unlike protocols such as FTP, BitTorrent groups multiple files into a single package called a torrent. For example,
you can download the Fedora Core ISO images, together with the release notes and MD5 values, as a single torrent.
BitTorrent, like other P2P systems, does not use a dedicated server. The functions of a server are performed by the
tracker, peers, and seeds. The tracker allows clients to communicate with each other. A client, called a peer when
it has downloaded part of the torrent and a seed once it has downloaded the entire torrent, acts as an additional
source for the torrent. As with a P2P network, each peer and seed that downloads a torrent uploads to other clients
the sections of the torrent it already has. There is nothing special about a seed: It can be removed at any time once
the torrent is available for download from other seeds.
The first step in downloading a torrent using BitTorrent is to locate or acquire a .torrent file. A .torrent file contains
the information about the torrent, such as its size and the location of the tracker. You can use a .torrent file using
its URI (page 1003), or you can acquire it via the Web, an email attachment, or other means. The next step is for
your BitTorrent client to connect to the tracker to learn the locations of other clients that it can download the torrent
from.
Once you have downloaded a torrent, it is good manners to allow BitTorrent to continue to run so another clients
can upload at least as much information as you downloaded.
Prerequisites
BitTorrent is not distributed with Red Hat Linux; you can download the rpm files from torrent.dulug.duke.edu/btrpms. If there is no rpm file for your version of Linux, use an rpm file for a similar version. Because
BitTorrent is written in Python and runs on any platform with a Python interpreter, it is not dependent on system
architecture. The noarch in the name of the RPM file stands for no architecture.
In order to run, BitTorrent requires Python, which is installed as /usr/bin/python on most Red Hat systems. Python
is available in the python rpm package.
How BitTorrent Works
The official BitTorrent distribution includes three client applications. You can use any of these applications to
download BitTorrent files:
•
btdownloadheadless.py . A text-based client that writes the status to standard output. Good for unattended
downloads where the output is redirected to a file.
•
btdownloadcurses.py . A text-based client that provides a pseudographical interface. Good for attended
downloads to machines not running a GUI.
•
btdownloadgui.py .
A graphical client (Figure 13-6).
177
Draft
BitTorrent
Draft
Figure 13-6. The graphical BitTorrent client
In addition to the official clients, there are other clients that provide additional features. Some of these clients are
available on sourceforge.net.
Using BitTorrent
First locate the .torrent file for the torrent you want to download. You can copy the .torrent file to the working
directory (first format following) or specify it with a --url option (second format). The simplest BitTorrent command
lines have the following formats:
$ btdownloadheadless.py --responsefile tfile.torrent --saveas savefile
or
$ btdownloadheadless.py --url http://domain/tfile.torrent --saveas savefile
Where tfile.torrent is the name of, or http://domain/tfile.torrent is the URI for, the .torrent file, and savefile is the
location to save the torrent in. In the case of torrents containing a single file, the file is saved as savefile. For torrents
containing multiple files, the files are saved in a directory named savefile. If you omit the --saveas argument, the
files are saved in the name specified in the .torrent file. Because each of the btdownload*.py applications takes
the same arguments, the preceding formats work for all three applications.
The following example shows how to download Fedora Core 2 ISO images. These are large files and take a while
to download. See page 35 for information about burning installation CDs from ISO images. Go to fedora.redhat.co/download to find the location of the latest .torrent file. To start the download, give the following command.
Because the command line is long, it is broken by a backslash (\). Make sure no character follows the backslash,
or the backslash will not quote the following RETURN, and the command will fail. (The shell supplies the > on
the second line.)
$ btdownloadheadless.py --max_upload_rate 8 \
> --url http://torrent.dulug.duke.edu/tettnang-binary-i386-iso.torrent
The preceding command uses a URI to specify a .torrent file and saves the downloaded files in a directory named
tettnang (the name of the Fedora release) as specified by the .torrent file.
178
Draft
wget: Download Files Noninteractively
Draft
The --max_upload_rate 8 option prevents BitTorrent from using more than 8 kilobytes per second of upstream
bandwidth. BitTorrent usually gives higher download rates to clients that upload more, so feel free to increase this
value if you have spare bandwidth. You need to leave enough free upstream bandwidth for the acknowledgement
packets from your download to get through, or your download will be very slow. By default, the client uploads to
a maximum of 7 other clients at once. You can change this value by specifying the --max_uploads argument, followed by the number of concurrent uploads you wish to permit. The default value of 7 is about right for most
consumer broadband connections.
After you give the preceding command, the screen quickly fills with output that looks similar to the following:
saving: tettnang-binary-i386-iso (2179.7 MB)
percent done: 0.0
time left: 9 hour 27 min 28 sec
download to: tettnang-binary-i386-iso
download rate: 34.08 kB/s
upload rate: 4.48 kB/s
download total: 0.8 MiB
upload total: 0.1 MiB
The file size is that of all the files you are downloading: four ISO images and several smaller files.
You can abort the download by pressing CONTROL-C. The download will automatically resume from where it
left off when you download the same torrent to the same location again.
Use the following command to perform the same download as in the previous example, throttling the rate and
number of uploads to values sensible for modem users. (The shell supplies the > on the second line, you do not
enter it.)
$ btdownloadcurses.py --max_upload_rate 3 --max_uploads 2 \
> --url http://torrent.dulug.duke.edu/tettnang-binary-i386-iso.torrent
The preceding command displays output similar to the following:
-----------------------------------------------------------file: tettnang-binary-i386-iso
|
size: 2,285,617,943 (2.1 GB)
|
dest: /home/mark/tettnang-binary-i386-iso
|
progress: ________________________________________________ |
status: finishing in 1:17:56 (11.7%)
|
speed: 450.3 KB/s down 3.1 KB/s up
|
totals: 256.0 MB
down 2.0 MB
up
|
error(s):
|
|
------------------------------------------------------------
|
|
|
|
|
|
|
|
|
wget: Download Files Noninteractively
The wget utility is a noninteractive, command line utility that can retrieve files from the Web using HTTP, HTTPS,
and FTP.
The following simple example of wget downloads Red Hat’s home page, named index.html, to a file with the same
name.
179
Draft
Chapter Summary
$ wget http://www.redhat.com
--18:03:34-- http://www.redhat.com/
=> 'index.html'
Resolving www.redhat.com... done.
Connecting to www.redhat.com[66.187.232.50]:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 29,537 [text/html]
100%[=================================>] 29,537
54.32K/s
18:03:35 (54.32 KB/s) - 'index.html' saved [29537/29537]
Draft
ETA 00:00
Use the –b option to run wget in the background and to redirect its standard error to a file named wget-log:
$ wget -b http://example.com/big_file.tar.gz
Continuing in background, pid 10752.
Output will be written to 'wget-log'.
If you download a file that would overwrite a local file, wget appends a dot followed by a number to the filename.
In the same way, subsequent background downloads are logged to wget-log.1, wget-log.2, and so on.
The –c option continues an interrupted download. The next command continues the download from the previous
example:
$ wget -b -c http://example.com/big_file.tar.gz
Chapter Summary
As a system administrator, you need to keep application and system software current. Of the many reasons to keep
the software on a system up-to-date, one of the most important is system security. Red Hat has made the process
of adding and removing the software packages they supply much easier with the system-config-packages (FEDORA)
and redhat-config-packages (RHEL) package management utilities. In addition, you can use the rpm utility to install,
uninstall, upgrade, query, and verify rpm packages. For packages distributed as source code, the GNU Configure
and Build System makes it easy to build executable files.
For both Red Hat Enterprise Linux and Fedora, Red Hat processes security, bugfix, and new feature (enhancement)
updates. The easiest way to learn about new updates is to use up2date. The up2date utility downloads and optionally
installs rpm packages. It works with many files and directories, in graphical and character-based modes, and has
many options.
Red Hat Network (RHN), a service provided by Red Hat, is an Internet-based system that can keep the software
on one or more Red Hat Enterprise Linux systems up-to-date.
For Fedora systems, yum and Apt both fill the same role of installing and updating software packages. Apt is slightly
faster, especially on slow connections, and supports a few more features, such as undoing upgrades. The yum
utility is installed by default and is easier to configure and use than Apt. The GNU Project distributes for Apt a
graphical front-end named synaptic, which is easy to use.
BitTorrent is a good tool for downloading large static data files such as the Fedora installation ISO images. BitTorrent
can replace protocols such as anonymous FTP, where client authentication is not required.
180
Draft
Exercises
Draft
Exercises
1. Why would you use HTTP or FTP instead of BitTorrent for downloading large files?
2. What command would you give to perform a complete upgrade using
a.
up2date?
b.
yum?
c.
Apt?
3. Why would you build a package from source when a (binary) rpm file is available?
4. Suggest two advantages rpm files have over source distributions.
5. Why would you choose to use RHN instead of yum or Apt? (RHEL only)
Advanced Exercises
6. What are some steps you should take before performing an upgrade on a mission-critical server?
7. When should you use rpm –i instead of rpm –U?
8. The system-config-packages utility does not allow you to remove KDE. If you decide that you no longer wish to
have KDE installed, how would you remove it?
9. When you compile a package yourself, not from an rpm file, which directory hierarchy should you put it in?
181
182
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 26 - Apache (httpd): Setting Up a Web Server
Apache (httpd): Setting Up a Web
Server
IN THIS CHAPTER
JumpStart I: Getting Apache Up and Running
JumpStart II: Setting Up Apache with system-config-httpd
Filesystem Layout
Directives I: Directives You May Want to Modify As You
Get Started
Contexts and Containers
The Red Hat httpd.conf File
Redirects
Multiviews
VirtualHosts
Troubleshooting
The World Wide Web, WWW or Web for short, is a collection of servers that hold material, called content, that
Web browsers, or just browsers, can display. Each of the servers on the Web is connected to the Internet, a network
of networks (an internetwork). Much of the content on the Web is coded in HTML (Hypertext Markup Language,
page 975). Hypertext, the code behind the links that you click on a Web page, allows browsers to display and react
to links that point to other Web pages on the Internet.
Apache is the most popular Web server on the Internet today. It is robust and extensible. The ease with which you
can install, configure, and run it in the Linux environment makes it an obvious choice for publishing content on
the World Wide Web. The Apache server and related projects are developed and maintained by the Apache Software
Foundation (ASF), a not-for-profit corporation formed in June of 1999. The ASF grew out of The Apache Group,
which was formed in 1995 to develop the Apache server.
This chapter starts with introductory information about Apache, followed by the first JumpStart section, which
describes the minimum you need to do to get Apache up and running. The second JumpStart section covers the
use of the Red Hat system-config-httpd configuration script. Following these sections is “Filesystem Layout,” which
tells you where the various Apache files are located.
Configuration directives, a key part of Apache, are discussed starting on page 767. This section includes coverage
of contexts and containers, two features/concepts that are critical to understanding Apache. The next section explains
the main Apache configuration file, httpd.conf, as modified by Red Hat. The final pages of the chapter cover virtual hosts, troubleshooting, and modules you can use with Apache, including CGI and SSL.
183
Draft
Introduction
Draft
Introduction
Apache is a server that responds to requests from Web browsers, or clients, such as Mozilla, Netscape, lynx, and
Internet Explorer. When you enter the address of a Web page (a URI [page 1003]) in a Web browser’s location
bar, the browser sends a request over the Internet to the (Apache) server at that address. In response, the server
sends the requested content back to the browser. The browser then displays or plays the content, which might be
a song, picture, video clip, or other information.
Aside from add-on modules that can interact with the content, Apache is oblivious to the content itself. Server administration and content creation are two different aspects of bringing up a Web site. This chapter concentrates
on setting up and running an Apache server and spends little time discussing content creation.
Apache, like the Linux kernel, uses external modules to increase load-time flexibility and allow parts of its code to
be recompiled without recompiling the whole program. Not part of the Apache binary, modules are stored as separate files that can be loaded when Apache is started.
Apache uses external modules, called Dynamic Shared Objects or DSOs, for basic and advanced functions; there
is not much to Apache without these modules. Apache also uses modules to extend its functionality: There are
modules that can process scripts written in Perl, PHP, Python, and other languages, use several different methods
to authenticate users, facilitate publishing content, and process nontextual content, such as audio. The list of
modules written by the Apache group and third-party developers is always growing. For more information, refer
to “Modules” on page 793.
About Apache
This section describes the packages you need to install and provides references for the programs this chapter covers.
The “Notes” section on page 762 covers terminology and other topics that will help you make better sense of this
chapter. The next section, “JumpStart I” on page 762, gets Apache up and running as quickly as possible.
Prerequisites
Install the following packages:
•
httpd
•
apr (Apache portable runtime, FEDORA)
•
apr-util (library, FEDORA)
Run chkconfig to cause httpd to start when the system goes multiuser:
# /sbin/chkconfig httpd on
After you configure Apache, use service to start httpd:
# /sbin/service httpd start
After changing the Apache configuration, restart httpd with the following command, which will not disturb clients
connected to the server:
184
Draft
More Information
Draft
# /sbin/service httpd graceful
You can install the following optional packages:
•
httpd-manual The Apache manual (FEDORA)
•
webalizer Web server log analyzer (page 799)
•
mod_perl Embedded Perl scripting language
•
mod_python Embedded Python scripting language
•
mod_ssl Secure Sockets Layer extension (page 795)
•
php Embedded PHP scripting language, including IMAP & LDAP support
•
mrtg MRTG traffic monitor (page 799).
•
net-snmp and net-snmp-utils SNMP, required for MRTG (page 799).
More Information
The Apache Reference Manual and Users’ Guide /var/www/manual Point a browser at http://localhost/manual
or /var/www/manual/index.html.en (en is for English, substitute for en: fr [French], ja.jis [Japanese], or ko.euckr [Korean] as needed). On Fedora, present only if you installed the httpd-manual package.
Apache Documentation httpd.apache.org/docs-2.0
Apache Directives List httpd.apache.org/docs-2.0/mod/directives.html
Apache Software Foundation (newsletters, mailing lists, projects, module registry, and more) www.apache.org
[http://www.apache.org]
mod_perl perl.apache.org
mod_python www.modpython.org [http://www.modpython.org]
mod_php www.php.net [http://www.php.net]
mod_ssl www.modssl.org [http://www.modssl.org] (page 795)
MRTG mrtg.hdl.com/mrtg.html
SNMP net-snmp.sourceforge.net
SSI httpd.apache.org/docs-2.0/howto/ssi.html [http://www.httpd.apache.org/docs-2.0/howto/ssi.html]
webalizer www.mrunix.net/webalizer [http://www.mrunix.net/webalizer].
Notes
Apache is the name of a server that serves HTTP content, among others. The Apache daemon is named httpd as
it is an HTTP server daemon. This chapter uses the terms Apache and httpd interchangeably.
185
Draft
JumpStart I: Getting Apache Up and
Running
Draft
An Apache server is the same thing as an Apache process: an Apache child process exists to handle incoming client
requests, hence it is referred to as a server.
Because Apache serves content on privileged ports, you must start it as root. For security, the processes that Apache
spawns run as the user and group apache.
The root of the directory hierarchy that Apache serves content from is called the document root. As shipped by Red
Hat, the document root is /var/www/html. You can use the DocumentRoot directive (page 770) to change the
location of the document root.
As shipped, only root can add or modify content in /var/www/html. To avoid having people work as root when
working with content, create a group (webwork, for example), put people who need to work with Web content
in this group, and make the directory hierarchy starting at /var/www/html (or other document root) writable by
that group. In addition, if you make the directory hierarchy setgid (chmod g+s filename ), all new files created in
the hierarchy will belong to the group, making sharing files easy. See page 523 for information about working with
groups.
JumpStart I: Getting Apache Up and Running
To get Apache up and running, you should modify the /etc/httpd/conf/httpd.conf configuration file. “Directives
I: Directives You May Want to Modify As You Get Started” on page 768 explains more about this file and explores
other changes you may want to make to it.
Modifying the httpd.conf Configuration File
Apache runs as installed, but it is best if you add the two lines described in this section to the /etc/httpd/conf/httpd.conf configuration file before starting Apache. If you do not add these lines, Apache assigns values that may
not work for you.
The ServerName line establishes a name for the server; add one of the following lines to httpd.conf to set the name
of the server to the domain name of the server or, if you do not have a domain name, to the IP address of the server:
ServerName example.com
or
ServerName IP_address
where example.com is the domain name of the server and IP_address is the IP address of the server. If you are not
connected to a network, you can use the localhost address, 127.0.0.1, so you can start the server and experiment
with it.
When a client has trouble getting information from a server, the server frequently displays an error page that says
what the problem is. For example, when Apache cannot find a requested page, it displays a page that says Error
404: Not Found. Each error page has a link that the user can click on to send mail to the administrator of the
server. ServerAdmin specifies the email address that the server displays on error pages. Add a ServerAdmin line to
httpd.conf that follows this format:
ServerAdmin email_address
where email_address is the email address of the person who needs to know if people are having trouble using the
server. Make sure that someone checks this email account frequently.
After making the changes to httpd.conf, start or restart httpd as explained on page 761.
186
Draft
Testing Apache
Draft
Testing Apache
Once you start the httpd daemon, you can check that Apache is working by pointing a browser on the local system
to http://localhost/. From a remote system, point a browser to http:// followed by the ServerName you specified
in httpd.conf. For example, you might use either of these URIs: http://192.168.0.16 or http://example.org .
The browser should display the Red Hat/Apache test page. This test page is actually an error page that says there
is no content. For more information, refer to “Red Hat test page” on page 789.
If you are having problems getting Apache to work, see “Troubleshooting” on page 792.
Putting Your Content in Place
Place the content you want Apache to serve in /var/www/html. Apache automatically displays the file named index.html in this directory. Working as root (or as a member of the group [webwork] you set up for this purpose),
give the following command to create such a page:
# cat > /var/www/html/index.html
<html><body><p>This is my test page.</p></body></html>
CONTROL-D
After creating this file, refresh the browser, if it is still running from testing, or start it again and point it at the
server. The browser should display the page you just created.
JumpStart II: Setting Up Apache with system-config-httpd
You can use the system-config-httpd (FEDORA) and redhat-config-httpd (RHEL) utilities to display the HTTP
window, which can help you set up Apache. The HTTP window has four tabs: Main, Virtual Hosts, Server, and
Performance Tuning. The Virtual Hosts tab has buttons that open other, extensive windows. Each of the fields in
these tabs/windows corresponds to a directive in the /etc/httpd/conf/httpd.conf file. This section discusses some
of the basic directives you can change with [system|redhat]-config-httpd. For more information, click Help at the
bottom of the HTTP window.
The Main tab (Figure 26-1) allows you to establish an FQDN (page 972) as the name of the server (ServerName,
page 769), an email address for the server administrator (ServerAdmin, page 769), and the ports and addresses that
Apache listens for requests on (Listen, page 768). Highlight an entry in the Available Addresses subwindow, and
click Edit to edit that entry or Add to add a new entry. Both actions bring up a window that allows you to specify
a port and select whether you want to listen to all IP addresses on that port or listen to a specific address.
187
Draft
JumpStart II: Setting Up Apache with
system-config-httpd
Draft
Figure 26-1. HTTP window, Main tab
The Virtual Hosts tab can establish default settings for Apache and set up virtual hosts (page 791). Click the Virtual
Hosts tab, and then click Edit Default Settings to configure default values for the Apache server (Figure 26-2).
The entries in the frame at the left of the window work like tabs: Click an entry, such as Logging, to display the
corresponding selections on the right. You do not have to change most of the values in this window. When you
are done making changes, click OK to close the window.
Figure 26-2. Establishing default values for Apache
188
Draft
Filesystem Layout
Draft
Click the Virtual Hosts tab, highlight an entry in the Virtual Hosts subwindow, and click either Edit to edit the
entry or Add to add a new virtual host. Both actions bring up the Virtual Hosts Properties window for a specific
virtual host (Figure 26-3). This window is similar to the one you used to establish default settings, except it pertains
to a specific virtual host and has more entries in the frame at the left. You do not have to change most of the values
in this window. Click OK when you are done making changes in this window.
Figure 26-3. Virtual Hosts Properties window for a specific virtual host
Usually, you do not need to change the values in this tab. You can specify the pathname of the lock file (LockFile
directive), the PID file (PidFile directive), and the directory that Apache stores core dumps in (CoreDumpDirectory).
The lower portion of the tab allows you to specify a user (User, page 785) and group (Group, page 783) that Apache
runs as.
The selections in this tab control the maximum number of connections that Apache allows (MaxClients, page 776),
the number of seconds after which a connection will disconnect (Timeout, page 778), the maximum number of
requests Apache allows per connection (MaxRequestsPerChild, page 776), and whether to allow persistent connections (KeepAlive directive). Initially, the values in this tab do not need to be changed. Click OK when you are done
making changes and restart httpd as discussed on page 761.
Filesystem Layout
This section tells you where you can find many of the files you may need to work with as you set up and modify
an Apache server.
The Apache server and related binary files are kept in several directories:
/usr/bin/httpd .
The Apache server (daemon)
/usr/sbin/apachectl .
/usr/bin/htpasswd .
797).
Starts and stops Apache, the httpd init script that service executes calls apachectl.
Creates and maintains password files used by the Apache authentication module (page
189
Draft
Configuration Directives
Draft
/usr/bin/rotatelogs . Rotates Apache log files so the files do not get too large. See logrotate (page 543) for more
information about rotating log files.
/etc/httpd/modules . Module binaries, including mod_perl.so and mod_python.so are kept in this directory
(a symbolic link to /usr/lib/httpd/modules, page 793).
Apache configuration files are kept in the /etc/httpd/conf directory and in the /etc/httpd/conf.d directory hierarchy:
/etc/httpd/conf/httpd.conf . The main Apache configuration file; contains configuration directives. The discussion of configuration directives starts on page 767. Refer to “The Red Hat httpd.conf File” on page 788 for a
description of the httpd.conf file.
/etc/httpd/conf/magic .
/etc/httpd/conf/ssl.* .
/etc/httpd/conf.d .
Mime file type identification; generally not changed.
Files and directories used by mod_ssl (page 795).
Configuration files for modules including php and mod_perl.
Logs are kept in /var/log/httpd (there is a symbolic link at /etc/httpd/logs):
/var/log/httpd/access_log .
Requests made to the server.
/var/log/httpd/error_log .
Request and run-time server errors.
/var/log/httpd/ssl_*_log .
Logs for mod_ssl.
Web documents, including the Web pages displayed by client browsers, custom error messages, and CGI scripts
are kept in /var/www by default:
/var/www/cgi-bin .
CGI scripts (page 794).
/var/www/error . Multilanguage default error documents. You can modify these documents to conform to the
style of your Web site: See ErrorDocument (page 780).
/var/www/icons .
Icons used to display directory entries.
/var/www/manual . FEDORA The Apache Reference Manual and Users’ Guide. On Fedora, present only if you
installed the httpd-manual package.
By default, the document root (page 762) is /var/www/html. You can change this location with the DocumentRoot
directive (page 770). In addition to content for the Web pages Apache serves, this directory can house the usage
directory, which holds webalizer (page 799) output.
A .htaccess file contains configuration directives and can appear in any directory in the document root hierarchy.
The location of a .htaccess file is critical: The directives in a .htaccess file apply to all files in the hierarchy rooted
at the directory that holds the .htaccess file. You must use the AllowOverride directive (page 786) to cause Apache
to examine .htaccess files. Clients cannot read .htaccess files; based on the Red Hat httpd.conf file, Apache does
not answer requests for files whose names start with .ht.
Configuration Directives
Configuration directives, or simply directives, are lines in a configuration file that control some aspect of how Apache
functions. A configuration directive is composed of a keyword followed by one or more arguments (values) separated
by SPACEs. For example, the following configuration directive sets Timeout to 300 (seconds):
190
Draft
Directives I: Directives You May Want
to Modify As You Get Started
Draft
Timeout 300
You must enclose arguments that contain SPACEs within double quotation marks. Keywords are not case-sensitive,
but arguments (pathnames, filenames, and so on) often are.
The most important file that holds Apache configuration directives is, by default, /etc/httpd/conf/httpd.conf.
This file holds global directives that affect all content served by Apache. An Include directive (page 783) within
httpd.conf can incorporate the contents of another file as though it was part of httpd.conf.
Local directives can appear in .htaccess files (page 767). A .htaccess file can appear in any directory within the
document root hierarchy and affects files in the directory hierarchy rooted at the directory the .htaccess file appears
in.
When you specify an absolute pathname in a configuration directive, the directive uses that pathname without
modifying it. When you specify a relative pathname, including a simple filename or the name of a directory, Apache
prepends to the name the value specified by the ServerRoot (page 782) directive (/etc/httpd by default).
Directives I: Directives You May Want to Modify As You Get Started
When it starts, Apache reads the /etc/httpd/conf/httpd.conf configuration file (by default) for instructions governing every aspect of how Apache runs and delivers content. The httpd.conf file shipped by Red Hat is over a
thousand lines long. This section details some of the lines in this file you may want to change as you are getting
started with Apache. You can use each of the following directives in httpd.conf; the Context line in each explanation
shows which other files the directives can appear in. Context is explained on page 772. The section titled “Directives
II” on page 776 describes more directives.
Note
Listen
Specifies the port(s) that Apache listens for requests on.
Listen [ IP-address :] portnumber
where IP-address is the IP address that Apache listens on, and portnumber is the number of the port that Apache
listens on for the given IP-address. When IP-address is absent, or set to 0.0.0.0, Apache listens on all network interfaces.
There must be at least one Listen directive in httpd.conf or Apache will not work. The following minimal directive
from the Fedora httpd.conf file listens for requests on all interfaces on port 80:
Listen 80
Red Hat Enterprise Linux uses the following Listen directive, which is equivalent to the preceding one, in its httpd.conf file:
Listen 0.0.0.0:80
The next directive changes the port from the default value of 80 to 8080:
Listen 8080
191
Draft
Directives I: Directives You May Want
to Modify As You Get Started
Draft
When you specify a port other than 80, each request to the server must include a port number (as in www.example.org:8080), or the kernel will return a Connection Refused message. Use multiple Listen directives to cause
Apache to listen on multiple IP addresses and ports. For example,
Listen 80
Listen 192.168.1.1:8080
Listen 192.168.1.2:443
accepts connections on all network interfaces on port 80, on 192.168.1.1 on port 8080, and on 192.168.1.2 on
port 443.
Context: server config
Default: none (Apache will not start without this directive)
Red Hat: Listen 0.0.0.0:80
Note
ServerAdmin
Sets the email address displayed on error pages.
ServerAdmin email-address
where email-address is the email address of the person responsible for Web content. This address appears on
Apache-generated error pages, except in Fedora Core 2, where Red Hat has removed email addresses from error
pages. Make sure email-address points to an email account that someone checks frequently. Users can use this address
to get help with the Web site or to inform the administrator of errors. There is no default value for ServerAdmin;
if you do not use this directive, the value is undefined and no email address appears on error pages.
Red Hat provides the following ServerAdmin template in its httpd.conf file:
#
ServerAdmin [email protected]
Copy this line, remove the #, and substitute the email address of the person responsible for Web content in place
of [email protected] Because webmaster is a common name, you can use webmaster
at your domain and use the /etc/aliases file (page 614) to forward mail that is sent to webmaster to the person
responsible for maintaining the Web site.
Contexts: server config, virtual host
Default: none
Red Hat: none
Note
ServerName
Specifies the server’s name and the port it listens on.
ServerName FQDN [:port]
192
Draft
Directives I: Directives You May Want
to Modify As You Get Started
Draft
where FQDN is the fully qualified domain name or IP address of the server and port is the optional port number
Apache listens on. The domain name of the server must be able to be resolved by DNS and may be different from
the hostname of the system running the server. If you do not specify a ServerName, Apache performs a DNS reverse
name resolution (page 708) on the IP address of the system and assigns that value to ServerName. If the reverse
lookup fails, Apache assigns the IP address of the system to ServerName.
Red Hat provides the following ServerName template in its httpd.conf file:
# ServerName dummy-host.example.com
Copy this line, remove the #, and substitute the FQDN or IP address of the server in place of dummy-host.example.com. Append to this name or number a colon and the port number Apache listens on if you are using other
than port 80.
The ports specified by ServerName and Listen (page 768) must be the same if you want the FQDN specified by
ServerName tied to the IP address specified by Listen.
Apache uses ServerName to construct a URI when it redirects a client (page 790).
Context: server config, virtual host
Default: none
Red Hat: none
Note
DocumentRoot
Points to the root of the directory hierarchy that holds the server’s content.
DocumentRoot dirname
where dirname is the absolute pathname of the directory at the root of the directory hierarchy that holds the content
Apache serves. Do not use a trailing slash. You can put the document root wherever you like, as long as the user
apache has read access to the ordinary files and execute access to the directory files in the directory hierarchy. The
FHS (page 167) specifies /srv as the top-level directory for this purpose. The following directive puts the document
root at /home/www:
DocumentRoot /home/www
Context: server config, virtual host
Default:/usr/local/apache/htdocs
Red Hat: /var/www/html
Note
UserDir
Allows users to publish content from their home directories
UserDir dirname | disabled | enabled user-list
193
Draft
Directives I: Directives You May Want
to Modify As You Get Started
Draft
where dirname is the name of a directory which, if it appears in a local user’s home directory, Apache publishes to
the Web. The disabled keyword prevents content from being published from users’ home directories, and enabled
causes content to be published from the home directories of users specified in the SPACE separated user-list. When
you do not specify a directory, Apache publishes to ~/public_html.
Apache combines the effects of multiple UserDir directives. Assume the following directives:
UserDir disabled
UserDir enabled user1 user2 user3
UserDir web
The first of the preceding directives turns off user publishing for all users. The second directive enables user publishing for three users. The third directive makes web the name of the directory which, if it appears in one of the
specified users’ home directories, Apache publishes to the Web.
To cause a browser to display the content published by a user, specify in the Location bar the name of the Web
site followed by a /~ and the user’s username. For example, if Sam published content in the public_html directory
in his home directory and the URI of the Web site was www.example.com, you would enter http://www.example.com/~sam to display his Web page. To display a user’s Web page, Apache must have execute permission
(as user apache) to the user’s home directory and the directory holding the content, and read permission to the
content files.
Red Hat provides the following ServerName directive and template in its httpd.conf file:
UserDir disable
#UserDir public_html
Put a pound sign (#) in front of the first line and remove the pound sign from the second line to allow users to
publish content from directories named public_html in their home directories.
Context: server config, virtual host
Default: public_html
Red Hat: disabled
Note
DirectoryIndex
Specifies which file to display when a user asks for a directory.
DirectoryIndex filename [filename ...]
where filename is the name of the file that Apache serves.
This directive specifies a list of filenames. When a client requests a directory, Apache attempts to find a file in the
specified directory whose name matches that of a file in the list. When Apache finds a match, it returns that file.
When this directive is absent or when none of the files specified by this directive exists in the specified directory,
Apache displays a directory listing as specified by the IndexOptions directive (page 780).
Red Hat provides the following DirectoryIndex directive in its httpd.conf file:
194
Draft
Contexts and Containers
Draft
DirectoryIndex index.html index.html.var
This directive causes Apache to return from the specified directory the file named index.html and, if that file does
not exist, the file named index.html.var.
The index.html is the name of the standard, default HTML document. If you supply PHP or CGI documents, you
may want to add the index.php or index.cgi values to this directive. The name index is standard but arbitrary.
The .var filename extension denotes a content-negotiated document that allows Apache to serve the Apache
manual and other documents in one of several languages as specified by the client. If you are not using the local
copy of the Apache manual and do not provide content in different languages, you can remove index.html.var
from this directive.
Context: server config, virtual host
Default: index.html
Red Hat: index.html index.html.var
Contexts and Containers
To make it flexible and easy to customize, Apache uses configuration directives, contexts, and containers. Configuration directives are covered starting on page 767. This section discusses contexts and containers, which are critical to managing an Apache server.
Contexts
There are four locations, called contexts, that define where a configuration directive can appear. This chapter marks
each configuration directive to indicate which context(s) it can appear in. Table 26-1 lists the contexts with a description of each.
Table 26-1. Contexts
Context
Location(s) Directives Can Appear In
server config
Directive can appear in the httpd.conf file only, but not
inside <VirtualHost> or <Directory> containers (next
section) unless so marked.
virtual host
Directive can appear inside <VirtualHost> containers in
the httpd.conf file only.
directory
Directive can appear inside <Directory>, <Location>,
and <Files> containers in the httpd.conf file only.
.htaccess
Directive can appear in .htaccess files (page 767) only.
Directives in files incorporated by means of the Include directive (page 783) are part of the context they are included
in and must be allowed in that context.
Putting a directive in the wrong context generates a configuration error and can cause Apache to serve content incorrectly or not start.
195
Draft
Contexts and Containers
Draft
Containers
Containers, or special directives, are directives that group other directives. Containers are delimited by XML-style
tags. Three examples are shown following:
<Directory> ... </Directory>
<Location> ... </Location>
<VirtualHost> ... </VirtualHost>
Look in httpd.conf for examples of containers. Containers are limited, just as other directives, to use within specified
contexts. This section describes some of the more frequently used containers.
Note
<Directory>
Applies directives to directories within specified directory hierarchies.
<Directory directory> ... </Directory>
where directory is an absolute pathname specifying the root of the directory hierarchy that holds the directories
the directives in the container apply to. The directory can include wildcards; a * does not match a /.
A <Directory> container provides the same functionality as a .htaccess file. While an administrator can use a
<Directory> container in the httpd.conf file, regular users cannot. Regular users can use .htaccess files to control
access to their own directories.
The directives in the following <Directory> container apply to the /var/www/html/corp directory hierarchy: The
Deny directive denies access to all clients, the Allow directive grants clients from the 192.168.10. subnet access,
and the AllowOverride directive (page 786) enables the use of .htaccess files in the hierarchy:
<Directory /var/www/html/corp>
Deny from all
Allow from 192.168.10.
AllowOverride All
</Directory>
Contexts: server config, virtual host
Note
<Files>
Applies directives to specified ordinary files.
<Files directory> ... </Files>
where directory is an absolute pathname specifying the root of the directory hierarchy that holds the ordinary files
the directives in the container apply to. The directory can include wildcards; a * does not match a /. This container
is similar to <directory> but applies to ordinary files and not to directories.
The following directive, from the Red Hat httpd.conf file, denies access to all files whose filenames start with .ht.
The tilde (~) changes how Apache interprets the following string. Without a tilde, the string is a simple shell match
196
Draft
Contexts and Containers
Draft
that interprets shell special characters (page 207). With a tilde, Apache interprets the string as a regular expression
(page 903):
<Files ~ "^\.ht">
Order allow,deny
Deny from all
</Files>
Contexts: server config, virtual host, directory, .htaccess
Note
<IfModule>
Applies directives if a specified module is loaded.
<IfModule [!]module-name> ... </IfModule>
where module-name is the name of the module (page 793) that is tested for. Apache executes the directives in this
container if module-name is loaded or, with !, if module-name is not loaded.
Apache will not start if you specify a configuration directive that is specific to a module that is not loaded.
The following <IfModule> container from the Red Hat httpd.conf file depends on the mod_mime_magic.c
module being loaded. If the module is loaded, Apache runs the MIMEMagicFile directive, which tells the
mod_mime_magic.c module where its hints file is located.
<IfModule mod_mime_magic.c>
MIMEMagicFile conf/magic
</IfModule>
See page 788 for another example of the <IfModule> container.
Contexts: server config, virtual host, directory, .htaccess
Note
<Limit>
Limits access control directives to specified HTTP methods.
<Limit method [method] ... > ... </Limit>
where method is an HTTP method. An HTTP method specifies what action is to be performed on a URI. The most
frequently used methods are GET, POST, and PUT and are case-sensitive. GET is the default method; it sends any
data indicated by the URI. PUT stores data from the body section of the communication at the specified URI. POST
creates a new document containing the body of the request at the specified URI.
This container binds a group of access-control directives to specified HTTP methods: Only methods named by the
<Limit> container are affected by this group of directives.
The following example disables HTTP uploads (PUTs) from systems not in a subdomain of example.com:
197
Draft
Contexts and Containers
Draft
<Limit PUT>
order deny,allow
deny from all
allow from .example.com
</Limit>
Caution: Use <LimitExcept> in Place of <Limit>
It is safer to use the <LimitExcept> container in place of the <Limit> container, as it protects against
arbitrary methods. When you use <Limit>, you must be careful to name explicitly all possible methods
that the group of directives could affect.
It is safer still not to put access control directives in any container.
Contexts: server config, virtual host, directory, .htaccess
Note
<LimitExcept>
Limits access control directives to all but specified HTTP methods.
<LimitExcept method [method] ... > ... </LimitExcept>
where method is an HTTP method. See <Limit> for a discussion of methods.
This container causes a group of access control directives not to be bound to specified HTTP methods: Methods
not named in <LimitExcept> are affected by this group of directives.
The access control directives within the following <LimitExcept> container affect HTTP methods other than GET
and POST. You could put this container in a <Directory> container to limit its scope:
<LimitExcept GET POST>
Order deny,allow
Deny from all
</LimitExcept>
Contexts: server config, virtual host, directory, .htaccess
Note
<Location>
Applies directives to specified URIs
<Location URI | URI> ... </Location>
where URI is a URI that points to content and specifies a file or the root of the directory hierarchy that the directives
in the container apply to. While the <Directory> container points within the local filesystem, <Location> points
outside the local file system. The URI can include wildcards; a * does not match a /.
The following <Location> container limits access to http://server/pop , where server is the FQDN of the server,
to clients from the example.net domain:
198
Draft
Contexts and Containers
Draft
<Location /pop>
Order deny,allow
Deny from all
Allow from .example.net
</Location>
Contexts: server config, virtual host
Caution: Use <Location> with Care
Use this powerful container with care. Do not use it in place of the <Directory> container: When
several URIs point to the same location in a filesystem, a client may be able to circumvent the desired
access control by using a URI not specified by this container.
Note
<LocationMatch>
Applies directives to matched URIs.
<LocationMatch regexp > ... </LocationMatch>
where regexp is a regular expression that matches one or more URIs. This container works the same way as <Location>, except it applies to any URIs that regexp matches:
# Disable autoindex for the root directory and present a
# default welcome page if no other index page is present.
#
<LocationMatch "^/$">
Options -Indexes
ErrorDocument 403 /error/noindex.html
</LocationMatch>
Contexts: server config, virtual host
Note
<VirtualHost>
Applies directives to a specified virtual host.
<VirtualHost addr [:port] [addr[:port]] ... > ... </VirtualHost>
where addr is an FQDN or IP address of the virtual host and port is the port that Apache listens on for the virtual
host. This container holds commands that Apache applies to a virtual host. For an example and more information,
refer to “VirtualHosts” on page 791.
Context: server config
199
Draft
Directives II
Draft
Directives II
This section discusses configuration directives that you may want to use after you have gained some experience
with Apache.
Directives That Control Processes
Note
MaxClients
Specifies the maximum number of child processes.
MaxClients num
where num is the maximum number of child processes (servers) Apache runs at one time, including idle processes
and those serving requests. When Apache is running num processes and there are no idle processes, Apache issues
server too busy errors to new connections; it does not start new child processes. A value of 150 is usually sufficient,
even for moderately busy sites.
Context: server config
Default: 256
Red Hat: 150
Note
MaxRequestsPerChild
Specifies the maximum number of requests a child process can serve.
MaxRequestsPerChild num
where num is the maximum number of requests a child process (server) can serve during its lifetime. After a child
process serves num requests, it does not process any more requests but dies after it finishes processing its current
requests. At this point additional requests are processed by other processes from the server pool.
Set num to 0 so as not to set a limit on the number of requests a child can process, except for the effects of MinSpareServers. By limiting the life of processes, this directive can prevent memory leaks from consuming too much
system memory. However, setting MaxRequestsPerChild to a small value can hurt performance by causing Apache
to create new child servers constantly.
Context: server config
Default: 10000
Red Hat: 1000
Note
MaxSpareServers
Specifies the maximum number of idle processes.
200
Draft
Directives II
Draft
MaxSpareServers num
where num is the maximum number of idle processes (servers) Apache keeps running to serve requests as they
come in. Do not set this number too high as each process consumes system resources.
Context: server config
Default: 10
Red Hat: 20
Note
MinSpareServers
Specifies the minimum number of idle processes.
MinSpareServers num
where num is the minimum number of idle processes (servers) Apache keeps running to serve requests as they come
in. More idle processes occupy more computer resources; increase this value for busy sites only.
Context: server config
Default: 5
Red Hat: 5
Note
StartServers
Specifies the number of child processes that Apache starts with.
StartServers num
where num is the number of child processes, or servers, that Apache starts when it is brought up. This value is significant only when Apache starts; MinSpareServers and MaxSpareServers control the number of idle processes once
Apache is up and running. Starting Apache with multiple servers ensures a pool of servers waiting to serve requests
immediately.
Context: server config
Default: 5
Red Hat: 8
Networking Directives
Note
HostnameLookups
Specifies whether Apache puts a client’s hostname or IP address in its logs.
HostnameLookups On | Off | Double
201
Draft
Directives II
Draft
where
On: Performs DNS reverse name resolution (page 708) to determine the hostname of each client for logging.
Off: Logs each client’s IP address.
Double: For security, performs DNS reverse name resolution (page 708) to determine the hostname of each client,
performs a forward DNS lookup to verify the original IP address, and logs the hostname.
Contexts: server config, virtual host, directory
Default: Off
Red Hat: Off
Tip: Lookups Can Consume a Lot of System Resources
Use the On and Double options with caution as they can consume a lot of resources on a busy system.
You can use a program such as logresolve to perform reverse name resolution offline for statistical
purposes.
If you perform host name resolution offline, you run the risk that the name has changed; you usually
want the name that was current at the time of the request. To minimize this problem, perform the
host name resolution as soon as possible after writing the log.
Note
Timeout
Specifies the time Apache waits for network operations to complete.
Timeout num
where num is the number of seconds that Apache waits for network operations to complete. You can usually set
this directive to a lower value; five minutes is a long time to wait on a busy server. The Apache documentation says
that the default is not lower, “...because there may still be odd places in the code where the timer is not reset when
a packet is sent.”
Context: server config
Default: 300
Red Hat: 300
Note
UseCannonicalName
Specifies the method the server uses to identify itself
UseCannonicalName On | Off | DNS
On: .
Apache uses the value of the ServerName directive (page 769) as its identity.
Off: .
Apache uses the name and port from the incoming request as its identity.
202
Draft
Directives II
Draft
DNS: . Apache performs a DNS reverse name resolution (page 708) on the IP address from the incoming request
and uses the result as its identity. Rarely used.
This directive is important when a server has more than one name and needs to perform a redirect. Red Hat sets
this directive to OFF because the ServerName directive (page 769) is commented out. Once you set ServerName,
change UseCannonicalName to ON. See page 790 for a discussion of redirects and this directive.
Contexts: server config, virtual host, directory
Default: ON
Red Hat: OFF
Logging Directives
Note
ErrorLog
Specifies where Apache sends error messages.
ErrorLog filename | syslog[:facility]
where
filename: .
to.
syslog: .
facility: .
Specifies the name of the file, relative to ServerRoot (page 782), that Apache sends error messages
Specifies that Apache send errors to syslogd (page 546).
Specifies which syslogd facility to use. The default facility is local7.
Contexts: server config, virtual host
Default logs/error_log
Red Hat logs/error_log
Note
LogLevel
Specifies the level of error messages that Apache logs.
LogLevel level
where level specifies that Apache logs errors of that level and higher (more urgent). Choose level from the following
list, which is in order of decreasing urgency and increasing verbosity:
emerg System unusable messages
alert Need for immediate action messages
crit Critical condition messages
error Error condition messages
203
Draft
Directives II
Draft
warn Nonfatal warning messages
notice Normal but significant messages
info Operational messages and recommendations
debug Messages for finding and solving problems
Contexts: server config, virtual host
Default: warn
Red Hat: warn
Directives That Control Content
Note
AddHandler
Creates a mapping between filename extensions and a builtin Apache handler.
AddHandler handlerextension [extension] ...
where handler is the name of a builtin handler and extension is a filename extension that maps to the handler.
Handlers are actions that are built into Apache and are directly related to loaded modules. Apache uses a handler
when a client requests a file with a specified filename extension.
For example, the following AddHandler directive causes Apache to process files that have a filename extension of
.cgi with the cgi-script handler:
AddHandler cgi-script .cgi
Contexts: Server config, virtual host, directory, htaccess
Default: none
Red Hat: imap-file map, type-map var
Note
Alias
Maps a URI to a directory or file.
Alias alias pathname
where alias must match part of the URI that the client requested in order to invoke the alias and pathname is the
absolute pathname of the target of the alias, usually a directory.
For example, the following alias in httpd.conf causes Apache to serve /usr/local/pix/milk.jpg when a client requests
http://www.example.com/pix/milk.jpg:
Alias /pix /usr/local/pix
204
Draft
Directives II
Draft
In some cases, you need to use a <Directory> container (page 772) to grant access to aliased content.
Contexts: server config, virtual host
Default: None
Red Hat (provides two aliases, one for /icons/ and one for /error/)
Note
ErrorDocument
Specifies the action Apache takes when the specified error occurs.
ErrorDocument code action
where code is the error code (page 799) that this directive defines a response for and action is one of the following:
string: .
Defines the message that Apache returns to the client.
absolute pathname: .
URI: .
Points to a local script or other content that Apache redirects the client to.
Points to an external script or other content that Apache redirects the client to.
When you do not specify this directive for a given error code, Apache returns a hard-coded error message when
that error occurs. See (page 789) for an explanation of how an ErrorDocument directive returns the Red Hat test
page when the system is first installed.
Some examples of ErrorDocument directives follow:
ErrorDocument 403 "Sorry, access is forbidden."
ErrorDocument 403 /cgi-bin/uh-uh.pl
ErrorDocument 403 http://errors.example.com/not_allowed.html
Contexts: server config, virtual host, directory, .htaccess
Default: none: Apache returns hard-coded error messages.
Red Hat: 403 /error/noindex.html; refer to “Red Hat test page” on page 789.
Note
IndexOptions
Specifies how Apache displays directory listings.
IndexOptions [±]option [[±] option] ...
where option can be any combination of the following:
DescriptionWidth= n: Sets the width of the description column to n characters. Use * in place of n to accommodate
the widest description.
FancyIndexing: . In directory listings, displays column headers that are links. When you click one of these
links, Apache sorts the display based on the content of the column. Clicking a second time reverses the order.
205
Draft
Directives II
FoldersFirst: .
Draft
Sorts the listing so that directories come before plain files. Use only with FancyIndexing.
IconsAreLinks: .
Makes the icons clickable. Use only with FancyIndexing.
IconHeight= n: .
Sets the height of icons to n pixels. Use only with IconWidth.
IconWidth: .
Sets the width of icons to n pixels. Use only with IconHeight.
IgnoreCase: .
Ignores case when sorting names.
IgnoreClient: .
Ignores options the client supplied in the URI.
NameWidth= n: . Sets the width of the filename column to n characters. Use * in place of n to accommodate
the widest filename.
ScanHTMLTitles: . Extracts and displays titles from HTML documents. Use only with FancyIndexing. Not
normally used because it is CPU and disk intensive.
SuppressColumnSorting: .
only with FancyIndexing.
SuppressDescription: .
Suppresses file descriptions. Use only with FancyIndexing.
SuppressHTMLPreamble: .
if that file exists.
SuppressIcon: .
SuppressSize: .
Suppresses the contents of the file specified by the HeaderName directive, even
Suppresses icons. Use only with FancyIndexing.
SuppressLastModified: .
SuppressRules: .
Suppresses clickable column headings that can be used for sorting columns. Use
Suppresses the modification date. Use only with Fancy-Indexing.
Suppresses horizontal lines. Use only with FancyIndexing.
Suppresses file sizes. Use only with FancyIndexing.
VersionSort: . Sorts version numbers (in filenames) in a natural way; character strings, except for substrings
of digits, are not affected.
When a client requests a URI that points to a directory (such as http://www.exam-ple.com/support/ ) and none
of the files specified by the DirectoryIndex directive (page 771) is present in that directory and, if the directory
hierarchy is controlled by a .htaccess file, AllowOverride (page 786) has been set to allow indexing, then Apache
displays a directory listing according to the options specified by this directive.
When this directive appears more than once within a directory, Apache merges the options from the directives.
Use + and – to merge options with options from higher-level directories. (Unless you use + or – with all options,
Apache discards any options set in higher-level directories.) For example, the following directives and containers
set the options for /custsup/download to VersionSort; Apache discards FancyIndexing and IgnoreCase in the
download directory because there is no + or – before VersionSort in the second <Directory> container.
<Directory /custsup>
IndexOptions FancyIndexing
IndexOptions IgnoreCase
</Directory
<Directory /custsup/download>
206
Draft
Directives II
Draft
IndexOptions VersionSort
</Directory>
Because of the + before VersionSort, the next directives and containers set the options for /custsup/download to
FancyIndexing, IgnoreCase, and VersionSort.
<Directory /custsup>
IndexOptions FancyIndexing
IndexOptions IgnoreCase
</Directory
<Directory /custsup/download>
IndexOptions +VersionSort
</Directory>
Contexts: server config, virtual host, directory, .htaccess
Default: None. Lists only filenames.
Red Hat: FancyIndexing VersionSort NameWidth=*
Note
ServerRoot
Specifies the root directory for server files (not content).
ServerRoot directory
where directory specifies the pathname of the root directory for files that make up the server; Apache prepends
directory to relative pathnames in httpd.conf. This directive does not specify the location of the content that Apache
serves; the DocumentRoot directive (page 770) performs that function. Do not change this value unless you move
the server files.
Context: server config
Default: /usr/local/apache
Red Hat: /etc/httpd
Note
ServerTokens
Specifies the server information that Apache returns to a client.
ServerTokens Prod | Major | Minor | Min | OS | Full
Prod: .
Returns the product name (Apache). Also ProductOnly.
Major: .
Returns the major release number of the server (Apache/2).
Minor: .
Returns the major and minor release numbers of the server (Apache/2.0).
Minimal: .
Returns complete version (Apache/2.0.40). Also Min.
207
Draft
Directives II
Draft
OS: . Returns the name of the operating system and the complete version (Apache/2.0.40 (Red Hat Linux)).
Provides less information that might help a malicious user then Full does.
Full: . Same as OS, plus sends the names and versions of non-Apache group modules (Apache/2.0.40 (Red
Hat Linux) PHP/4.2.2).
Unless there is a reason you want to let clients know the details of the software you are running, set ServerTokens
to reveal as little as possible.
Contexts: server config
Default: Full
Red Hat: OS
Note
ServerSignature
Adds a line to server generated pages.
ServerSignature On | Off | EMail
On: Turns the signature line on. The signature line contains the server version as specified by the ServerTokens
directive (page 782) and the name specified by the <VirtualHost> container (page 776).
Off: .
Turns the signature line off.
EMail: . To the signature line, adds a mailto: link to the server email address. This option produces output
that can attract spam.
Contexts: server config, virtual host, directory, .htaccess
Default: Off
Red Hat: On
Configuration Directives
Note
Group
Sets the GID of the processes that run the servers.
Group # grouprid | groupname
where groupid is a GID value, preceded by a #, and groupname is the name of a group. The processes (servers) that
Apache spawns are run as the group specified by this directive. See the User directive (page 785) for more information.
Context: server config
Red Hat: apache
Default: #–1
208
Draft
Directives II
Draft
Note
Include
Loads directives from files.
Include filename | directory
where filename is the relative pathname of a file that contains directives. Apache prepends ServerRoot (page 782)
to filename. The directives in filename are included in the file holding this directive at the location of the directive.
Because filename can include wildcards, it can specify more than one file.
And where directory is the relative pathname that specifies the root of a directory hierarchy that holds files containing
directives. Apache prepends ServerRoot to directory. The directives in ordinary files in this hierarchy are included
in the file holding this directive at the location of the directive. The directory can include wildcards.
When you install Apache and its modules, rpm puts configuration files, which have a filename extension of conf,
in the conf.d directory within the ServerRoot directory. The Include directive in the Red Hat httpd.conf file incorporates module configuration files for whichever modules are installed.
Contexts: server config, virtual host, directory
Default: None.
Red Hat: conf.d/*.conf
Note
LoadModule
Loads a module.
LoadModule module filename
where module is the name of an external DSO module and filename is the relative pathname of the named module.
Apache prepends ServerRoot (page 782) to filename. Apache loads the external module specified by this directive.
For more information, refer to “Modules” on page 793.
Context: server config
Default: None. Nothing is loaded by default if omitted.
Red Hat: Loads over 40 modules, refer to httpd.conf for the list.
Note
Options
Controls server features by directory.
Options [±]option [[±]option ...]
This directive controls which server features are enabled for a directory hierarchy. The directory hierarchy is specified
by the container this directive appears in. A + or the absence of a – turns an option on and a – turns it off.
where option is one of the following:
209
Draft
Directives II
Draft
None None of the features this directive can control are enabled.
All All of the features this directive can control are enabled, except for Multi-Views, which you must explicitly
enable.
ExecCGI Apache can execute CGI scripts (page 794).
FollowSymLinks Apache follows symbolic links.
Includes Permits SSIs (server side includes). SSIs are containers embedded in HTML pages and evaluated on the
server before the content is passed to the client.
IncludesNOEXEC The same as Includes but disables the #exec and #exec cgi commands that are part of SSIs.
Does not prevent the #include command from referencing CGI scripts.
Indexes Generates a directory listing if DirectoryIndex (page 771) is not set.
MultiViews Allows multiviews (page 791).
SymLinksIfOwnerMatch The same as FollowSymLinks but follows the link only if the file or directory being
pointed to has the same owner as the link.
The following Options directive from the Red Hat httpd.conf file sets the Indexes and FollowSymLinks options
and, because the <Directory> container specifies the /var/www/html directory hierarchy (the document root),
affects all content:
<Directory "/var/www/html">
Options Indexes FollowSymLinks
...
<Directory>
Contexts: directory
Default: All
Red Hat: None
Note
ScriptAlias
Maps a URI to a directory or file and declares the target a server (CGI) script.
ScriptAlias alias pathname
where alias must match part of the URI the client requested in order to invoke the ScriptAlias and pathname is the
absolute pathname of the target of the alias, usually a directory. Similar to the Alias directive, this directive specifies
the target is a CGI script (page 794).
The following ScriptAlias directive from the Red Hat httpd.conf file maps client requests that include /cgi-bin/
to the /var/www/cgi-bin directory (and indicates that these requests will be treated as CGI requests):
ScriptAlias /cgi-bin/ "/var/www/cgi-bin/"
210
Draft
Directives II
Draft
Contexts: server config, virtual host
Default: None
Red Hat: /cgi-bin/ “/var/www/cgi-bin”
Note
User
Sets the UID of the processes that run the servers.
User # userid | username
where userid is a UID value, preceded by a #, and username is the name of a local user. The processes (servers) that
Apache spawns are run as the user specified by this directive.
Apache must start as root in order to listen on a privileged port. For reasons of security, Apache’s child processes
(servers) run as nonprivileged users. The default UID of –1 does not map to a user under Red Hat. Instead, Red
Hat’s httpd rpm creates a user named apache during installation and sets User to that user.
Security: Do Not Set User to root or 0
For a more secure system, do not set User to root or 0 (zero) and do not allow the Apache user to
have write access to the DocumentRoot directory hierarchy (except as needed for storing data), especially not to configuration files.
Context: server config
Red Hat: apache
Default: #–1
Security Directives
Note
Allow
Specifies which clients can access specified content.
Allow from All | host [host ...] | env=var [env=var ...]
This directive, which must be written as Allow from, grants to specified clients access to a directory hierarchy.
The directory hierarchy is specified by the container or .htaccess file this directive appears in.
All: .
Serves content to any client.
host: . Serves content to the client(s) specified by host , which can take several forms: host can be an FQDN, a
partial domain name (such as example.com), an IP address, a partial IP address, or a network/netmask pair.
var: . Serves content when the environment variable named var is set. You can set a variable with the SetEnvIf
directive. See the Order directive for an example.
Contexts: directory, .htaccess
211
Draft
Directives II
Draft
Default: none; default behavior depends on the Order directive
Red Hat: All
Note
AllowOverride
Specifies the classes of directives that are allowed in .htaccess files.
AllowOverride All | None | directive-class [directive-class ...]
This directive specifies whether Apache reads .htaccess files in the directory hierarchy specified by its container.
If Apache does reads .htaccess files, this directive specifies which kinds of directives are valid within .htaccess
files.
None: .
All: .
Ignores .htaccess files.
Allows all classes of directives in .htaccess files.
where directive-class is one of the following directive class identifiers:
AuthConfig: . Class of directives that control authorization (AuthName, AuthType, Require, and so on). This
class is used mostly in .htaccess files when requiring a username and password to access the content. For more
information, refer to “Authentication Modules and .htaccess” on page 797.
FileInfo: .
on).
Class of directives that controls document types (DefaultType, ErrorDocument, SetHandler, and so
Indexes: .
so on).
Class of directives relating to directory indexing (DirectoryIndex, FancyIndexing, IndexOptions, and
Limit: .
Class of client access directives (Allow, Deny, and Order).
Options: .
Class of directives controlling directory features.
Context: directory
Default: All
Red Hat: None
Note
Deny
Specifies which clients are not allowed to access specified content.
Deny from All | host [host ...] | env=var [env=var ...]
This directive, which must be written as Deny from, denies to specified clients access to a directory hierarchy. The
directory hierarchy is specified by the container or .htaccess file this directive appears in. See the Order directive
for an example.
All: .
212
Denies content to all clients.
Draft
Directives II
Draft
host: . Denies content to the client(s) specified by host , which can take several forms: host can be an FQDN, a
partial domain name (such as example.com), an IP address, a partial IP address, or a network/netmask pair.
var: . Denies content when the environment variable named var is set. You can set a variable with the SetEnvIf
directive.
Contexts: directory, .htaccess
Default: none
Red Hat: none
Note
Order
Specifies default access and the order Allow and Deny directives are evaluated in.
Order Deny,Allow | Allow,Deny
Deny,Allow: . Allows access by default: Denies access only to clients specified in Deny directives. (First evaluates
Deny directives, then evaluates Allow directives.)
Allow,Deny: . Denies access by default: Allows access only to clients specified in a Allow directives. (First
evaluates Allow directives, then evaluates Deny directives.)
Access granted or denied by this directive applies to the directory hierarchy specified by the container or .htaccess
file this directive appears in.
There must not be SPACEs on either side of the comma. Although Red Hat has a default of Allow,Deny, which
denies access to all clients not specified by Allow directives, the next directive in httpd.conf, Allow from all,
grants access to all clients:
Order allow,deny
Allow from all
You can restrict access by specifying Deny,Allow to deny all access and then specifying the clients you want to
grant access to in an Allow directive. The following directives grant access to clients from the example.net domain
only and would typically appear within a <Directory> container (page 772):
Order deny,allow
Deny from all
Allow from .example.net
Contexts: directory, .htaccess
Default: Deny,Allow
Red Hat: Allow,Deny
213
Draft
The Red Hat httpd.conf File
Draft
The Red Hat httpd.conf File
This section highlights some of the important features of the Red Hat httpd.conf file, which is based on the httpd.conf file distributed by Apache. This heavily commented file is broken into the following parts (as is this section):
1.
Global Environment: .
Controls the overall functioning of the Apache server.
2.
Main Server Configuration: . Configures the default server (as opposed to virtual hosts) and provides
default configuration information for virtual hosts.
3.
Virtual Hosts: .
Configures virtual hosts. For more information, refer to “VirtualHosts” on page 791.
Section 1: Global Environment
The ServerTokens directive (page 782) is set to OS, which causes Apache, when queried, to return the name of the
operating system and the complete version number of Apache:
ServerTokens OS
The ServerRoot directive (page 782) is set to /etc/httpd, which is the pathname that Apache prepends to relative
pathnames in httpd.conf.
ServerRoot "/etc/httpd"
Multiprocessing Modules (MPMs) allow you to change the way Apache works by changing modules it uses. The
<IfModule> containers (page 773) allow you to use the same httpd.conf file with different modules: The directives
in an <IfModule> container are executed only if the specified module is loaded.
The section of httpd.conf that starts with the comment
## Server-Pool Size Regulation (MPM specific)
holds two <IfModule> containers (page 773) that configure Apache, depending on which module, prefork or
worker, is loaded. Red Hat ships Apache with the prefork module loaded; this section does not discuss the <IfModule> container for the worker module.
The prefork <IfModule> container, shown following, holds directives that control the functioning of Apache when
it starts and as it runs:
<IfModule prefork.c>
StartServers
8
MinSpareServers
5
MaxSpareServers
20
MaxClients
150
MaxRequestsPerChild
</IfModule>
1000
The Listen directive (page 768) does not specify an IP address (the 0.0.0.0 is the same as no IP address) and so
Apache listens on all network interfaces:
214
Draft
Section 2: Main Server Configuration
Draft
Listen 80 FEDORA
Listen 0.0.0.0:80 RHEL
There are quite a few LoadModule directives (page 784); these directives load the Apache DSO modules (page
793).
The Include directive (page 783) includes the files that match *.conf in the /etc/httpd/conf.d directory, as though
they were part of httpd.conf:
Include conf.d/*.conf
When you first install Apache, there is no index.html file in /var/www/html; when you point a browser at the
local Web server, Apache generates an error 403, which returns the Red Hat test page. The mechanism by which
this page is returned is convoluted: The Red Hat httpd.conf file holds an Include directive that includes all files
in the conf.d directory that is in the ServerRoot directory (page 782). The welcome.conf file in this directory has
an ErrorDocument 403 directive (page 780) that redirects users who receive this error to error/noindex.html in
the DocumentRoot directory (page 770). The noindex.html is the Red Hat test page that confirms the server is
working but there is no content to display.
Section 2: Main Server Configuration
Note
ServerAdmin
ServerName
As shipped, the ServerAdmin and ServerName directives are commented out. Change them to useful values as
suggested in ServerAdmin (page 769) and Server-Name (page 769).
The DocumentRoot directive (page 770) appears as follows:
DocumentRoot "/var/www/html"
You need to modify this directive only if you want to put your content somewhere other than /var/www/html.
The following <Directory> container (page 772) sets up restrictive environment for the entire local filesystem
(specified by /):
<Directory />
Options FollowSymLinks
AllowOverride None
</Directory>
The preceding Options directive (page 784) allows Apache to follow symbolic links but disallows many options.
The AllowOverride directive (page 786) causes Apache to ignore .htaccess files. You must explicitly enable less
restrictive options if you want them, but doing so can expose the root filesystem and compromise the system.
Next, another <Directory> container sets up less restrictive options for the DocumentRoot (/var/www/html).
The code in httpd.conf is interspersed with many comments. Without the comments it looks like this:
215
Draft
Section 3: Virtual Hosts
Draft
<Directory "/var/www/html">
Options Indexes FollowSymLinks
AllowOverride None
Order allow,deny
Allow from all
</Directory>
The Indexes option in the Options directive allows Apache to display directory listings. The Order (page 787) and
Allow (page 786) directives combine to allow requests from all clients. This container is slightly less restrictive than
the preceding one, although it still does not allow Apache to follow directives in .htaccess files.
As explained on page 771, the DirectoryIndex directive causes Apache to return the file named index.html from
a requested directory and, if that file does not exist, to return index.html.var. Because Options Indexes is specified
in the preceding <Directory> container, if neither of these files exists in a queried directory, Apache returns a directory listing:
DirectoryIndex index.html index.html.var
There are many more directives in this part of the httpd.conf file. The comments in the file provide a guide to what
they do. There is nothing here you need to change as you get started using Apache.
Section 3: Virtual Hosts
All lines in this section are comments or commented out directives. If you want to set up virtual hosts, see page
791.
Redirects
Apache can respond to a request for a URI by asking the client to request a different URI. This response is called
a redirect. A redirect works because redirection is part of the HTTP implementation: Apache sends the appropriate
response code and the new URI, and a compliant browser requests the new location.
The Redirect directive can establish an explicit redirect that sends a client to a different page when a Web site is
moved. Or, when a user enters the URI of a directory in a browser but leaves off the trailing slash, Apache automatically redirects the client to the same URI terminated with a slash.
The ServerName directive (page 769), which establishes the name of the server, and UseCannonicalName directive
(page 778) are important when a server has more than one name and needs to perform an automatic redirect. For
example, assume the server with the name zach.example.com and an alias of www.example.com has ServerName
set to www.example.com. When a client specifies a URI of a directory but leaves off the trailing slash (zach.example.com/dir), Apache has to perform a redirect to determine the URI of the requested directory. When UseCannonicalName is set to ON, Apache uses the value of ServerName and returns www.example.com/dir/. With
UseCannonicalName set to OFF, Apache uses the name from the incoming request and returns zach.example.com/dir/.
Multiviews
Multiviews is a way to represent a page in different ways, most commonly in different languages. Using request
headers, a browser can request a specific language from a server. Servers that cannot handle these requests ignore
them.
216
Draft
Server Generated Directory Listings
(Indexing)
Draft
Multiviews is demonstrated by the Apache manual, which can be installed locally in /var/www/manual. When
you point a browser to http://server/manual/index.html , the browser displays the page in the browser’s default
language. If you change the browser’s default language setting and reload the page, the browser displays the page
in the new language. The browser can display the pages in different languages because the server has a copy of the
page for each language. For example, the files index.html.en and index.html.fr both exist in the
/var/www/manual directory.
Server Generated Directory Listings (Indexing)
When a client requests a directory, Apache configuration determines what it returns to the client. Apache can return
a file as specified by the DirectoryIndex directive (page 771), a directory listing if there is no file matching DirectoryIndex and the Options Indexes directive (page 784) is set, or an error message if there is no file matching DirectoryIndex and Options/Indexes is not set.
VirtualHosts
Apache supports virtual hosts, which means one instance of Apache can respond to requests directed to multiple
IP addresses or hostnames as though it were multiple servers. Each IP address or hostname can provide different
content and be configured differently.
There are two types of virtual hosts: host-by-name and host-by-IP. Host-by-name relies on the FQDN the client
uses in its request to Apache; for example, www.example.com versus www2.example.com. Host-by-IP examines
the IP address the host resolves as and responds according to that match.
Host-by-name is useful if there is only one IP address, but Apache must support multiple FQDNs. You can use
host-by-IP if a given Web server has aliases, but Apache should serve the same content regardless of which name
is used.
Virtual hosts inherit their configuration from httpd.conf Section 1 (page 788) and Section 2 (page 789). Then, in
Section 3, <VirtualHost> containers create the virtual hosts and specify directives that override inherited and default
values. You can specify many virtual hosts for a single instance of Apache.
The following <VirtualHost> container sets up a Host-by-name for the site named intranet.example.com. This
virtual host handles requests that Apache receives directed to intranet.example.com.
<VirtualHost intranet.example.com>
ServerName intranet.example.com
DocumentRoot /usr/local/www
ErrorLog /var/log/httpd/intra.error_log
CustomLog /var/log/httpd/intra.server_log
<Directory /usr/local/www>
Order deny,allow
Deny from all
Allow from 192.168. # allow from private subnet only
</Directory>
</VirtualHost>
Troubleshooting
You can use service and the httpd init script to check the syntax of the Apache configuration files:
217
Draft
Modules
Draft
# service httpd configtest
Syntax OK
Once you start the httpd daemon, you can check that Apache is working by pointing a browser on the local system
at http://localhost/. From a remote system, use http://server/, substituting the hostname of the server for server.
Apache displays the Red Hat test page.
If the browser does not display the test page, it will display one of two errors: Connection refused or an error
page. If you get a connection refused error, make sure port 80 is not blocked by iptables and check that the server
is running:
# /sbin/service httpd status
httpd (pid 21406 21405 21404 21403 21402 21401 13622) is running...
If the server is running, check that you did not specify a port other than 80 in a Listen directive. If you did, the URI
you specify in the browser must reflect this port number. Otherwise, check the error log (/var/log/httpd/error_log)
for information on what is not working.
To make sure the browser is not at fault, use telnet to try to connect to port 80 of the server:
$ telnet www.example.com 80
Trying 192.0.34.166...
Connected to www.example.com.
Escape character is '^]'.
CONTROL-]
telnet> quit
Connection closed.
If you see Connection refused, you have verified that you cannot get through to the server.
Modules
Apache is a skeletal program that uses external modules, called Dynamic Shared Objects (DSOs), for most of its
functionality. This section lists these modules and discusses some of the more important ones. In addition to the
modules included with Red Hat, there are many other modules. See httpd.apache.org/modules for more information.
Module List
The following Apache modules are included with Red Hat:
access (mod_access.so) Controls access based on client characteristics.
actions (mod_actions.so) Allows execution of CGI scripts based on request method.
alias (mod_alias.so) Allows outside directories to be mapped to DocumentRoot.
asis (mod_asis.so) Allows sending files that contain their own headers.
auth (mod_auth.so) Provides user authentication using .htaccess.
auth_anon (mod_auth_anon.so) Provides anonymous user access to restricted areas.
218
Draft
Module List
Draft
auth_dbm (mod_auth_dbm.so) Uses DBM files for authentication.
auth_digest (mod_auth_digest.so) Uses MD5 digest for authentication.
autoindex (mod_autoindex.so) Allows directory indexes to be generated.
cern_meta (mod_cern_meta.so) Allows the use of CERN httpd metafile semantics.
cgi (mod_cgi.so) Allows the execution of CGI scripts.
dav (mod_dav.so) Allows Distributed Authoring and Versioning.
dav_fs (mod_dav_fs.so) Provides a filesystem for mod_dav.
dir (mod_dir.so) Allows directory redirects and listings as index files.
env (mod_env.so) Allows CGI scripts to access environment variables.
expires (mod_expires.so) Allows generation of Expires HTTP headers.
headers (mod_headers.so) Allows customization of request and response headers.
imap (mod_imap.so) Allows image maps to be processed on the server side.
include (mod_include.so) Provides server side includes (SSIs).
info (mod_info.so) Allows the server configuration to be viewed.
log_config (mod_log_config.so) Allows logging of requests made to the server.
mime (mod_mime.so) Allows association of file extensions with content.
mime_magic (mod_mime_magic.so) Determines MIME types of files.
negotiation (mod_negotiation.so) Allows content negotiation.
proxy (mod_proxy.so) Allows Apache to act as a proxy server.
proxy_connect (mod_proxy_connect.so) Allows connect request handling.
proxy_ftp (mod_proxy_ftp.so) Provides an FTP extension proxy.
proxy_http (mod_proxy_http.so) Provides an HTTP extension proxy.
rewrite (mod_rewrite.so) Allows on-the-fly URI rewriting based on rules.
setenvif (mod_setenvif.so) Sets environment variables based on a request.
speling (mod_speling.so) Auto-corrects spelling if requested URI has incorrect capitalization and one spelling
mistake.
status (mod_status.so) Allows the server status to be queried and viewed.
unique_id (mod_unique_id.so) Generates a unique ID for each request.
userdir (mod_userdir.so) Allows users to have content directories (public_html).
usertrack (mod_usertrack.so) Allows tracking of user activity on a site.
219
Draft
mod_cgi and CGI Scripts
Draft
vhost_alias (mod_vhost_alias.so) Allows the configuration of virtual hosting.
mod_cgi and CGI Scripts
The CGI (Common Gateway Interface) allows external application programs to interface with Web servers. Any
program can be a CGI program if it runs in real time (at the time of the request) and relays its output to the requesting
client. Various kinds of scripts, including shell, Perl, Python, and PHP, are the most common CGI programs because
a script can call a program and reformat its output in HTML for a client.
Apache can handle requests for CGI programs in different ways. The most common method is to put a CGI program
in the cgi-bin directory and enable its execution from that directory only. The location of the cgi-bin directory,
as specified by the ScriptAlias directive (page 785), is /var/www/cgi-bin. Alternatively, an AddHandler directive
(page 779) can identify filename extensions of scripts, such as .cgi or .pl, within the regular content (for example,
AddHandler cgi-script .cgi). If you use an AddHandler, you must also specify the ExecCGI option within an
Options directive within the appropriate <Directory> container. The mod_cgi module must be loaded in order to
access and execute CGI scripts.
The following Perl CGI script displays the Apache environment. This script should be used for debugging only as
it presents a security risk if outside clients can access it.
#!/usr/bin/perl
##
## printenv -- demo CGI program which just prints its environment
##
print "Content-type: text/plain\n\n";
foreach $var (sort(keys(%ENV))) {
$val = $ENV{$var};
$val =~ s|\n|\\n|g;
$val =~ s|"|\\"|g;
print "${var}=\"${val}\"\n";
}
mod_ssl
SSL (Secure Socket Layer), which is implemented by the mod_ssl module, has two functions: It allows a client to
verify the identity of a server and it enables secure two-way communication between a client and server. SSL is used
on Web pages with forms that require passwords, credit card numbers, or other sensitive data.
Apache uses the HTTPS protocol, not HTTP, for SSL communication. When Apache uses SSL, it listens on a second
port (443 by default) for a connection and performs a handshaking sequence before sending the requested content
to the client.
Server verification is critical for financial transactions; you do not want to give your credit card number to a
fraudulent Web site posing as a known company. SSL uses a certificate to positively identify a server. Over a public
network such as the Internet, the identification is reliable only if the certificate has a digital signature from an authoritative source such as VeriSign or Thawte. SSL Web pages are denoted by a URI beginning with https://.
Data encryption prevents malicious users from evesdropping on Internet connections and copying personal information. To encrypt communication, SSL sits between the network and an application and encrypts communication
between the server and client.
220
Draft
mod_ssl
Draft
Setting Up mod_ssl
The /etc/httpd/conf.d/ssl.conf file configures mod_ssl. The first few directives in this file load the mod_ssl
module, instruct Apache to listen on port 443, and set various parameters for SSL operation. About a third of the
way through the file is a section labeled SSL Virtual Host Context that sets up virtual hosts (page 791).
A <VirtualHost> container in ssl.conf is similar to one in httpd.conf. As with any <VirtualHost> container, it
holds directives such as ServerName and ServerAdmin that need to be configured. In addition, it holds some SSLrelated directives.
Using a Self-Signed Certificate for Encryption
If you require SSL for encryption and not verification, that is, if the client already trusts the server, it is possible to
generate and use a self-signed certificate and bypass the time and expense of obtaining a digitally signed certificate.
Self-signed certificates generate a warning when you connect to the server: Most browsers display a dialog box that
allows you to examine and accept the certificate. The sendmail daemon also uses certificates (page 632).
To generate the private key that the encryption relies on, cd to /etc/httpd/conf and give the following command:
# cd /etc/httpd/conf
# make server.key
umask 77 ; \
/usr/bin/openssl genrsa -des3 1024 > server.key
Generating RSA private key, 1024 bit long modulus
...................................++++++
.......++++++
e is 65537 (0x10001)
Enter pass phrase:
Verifying - Enter pass phrase:
The preceding command generates a file named server.key that is protected by the pass phrase you entered; you
will need this pass phrase to start the server. Keep the server.key file secret.
The next command generates the certificate. This process uses the private key you just created; you need to supply
the same pass phrase you entered when you created the private key.
# make server.crt
umask 77 ; \
/usr/bin/openssl req -new -key server.key -x509 -days 365 -out server.crt
Enter pass phrase for server.key:
You are about to be asked to enter information that will be incorporated
into your certificate request.
What you are about to enter is what is called a Distinguished Name or a DN.
There are quite a few fields but you can leave some blank
For some fields there will be a default value,
If you enter '.', the field will be left blank.
----Country Name (2 letter code) [GB]:US
State or Province Name (full name) [Berkshire]:California
Locality Name (eg, city) [Newbury]:San Francisco
Organization Name (eg, company) [My Company Ltd]:Sobell Associates Inc.
Organizational Unit Name (eg, section) []:Executive
221
Draft
Authentication Modules and .htaccess
Draft
Common Name (eg, your name or your server's hostname) []:www.sobell.com
Email Address []:[email protected]
The answers to the first five questions are arbitrary: They can help clients identify you when they examine the certificate. The answer to the sixth question (Common Name) is critical. Because certificates are tied to the name of
the server, it is important you enter your server’s FQDN accurately. If you mistype this information, the server
name and that of the certificate will not match and the browser will generate a warning message each time a connection is made.
The preceding commands created two files in the working directory; move server.key into the ssl.key directory,
and server.crt into the ssl.crt directory. After you restart Apache, the new certificate will be in use.
Notes on Certificates
•
Although the server name is part of the certificate, the SSL connection is tied to the IP address of the server:
You can have only one certificate per IP address. In order for multiple virtual hosts to have separate certificates,
you must specify host-by-IP rather than host-by-name virtual hosts (page 791).
•
As long as the server is identified by the name for which the certificate was issued, you can use the certificate
on another server and/or IP address.
•
A root certificate (root CA) is the certificate that signs the server certificate. Every browser contains a database
of the public keys for the root certificates of the major signing authorities including VeriSign and Thawte.
•
It is possible to generate a root certificate (root CA) and sign all your server certificates with this root CA. Regular clients can import the public key of the root CA so that they recognize every certificate signed by that root
CA. This setup is convenient for a server with multiple SSL-enabled virtual hosts and no commercial certificates.
For more information see www.modssl.org/docs/2.8/ssl_faq.html#ToC29
[http://www.modssl.org/docs/2.8/ssl_faq.html#ToC29].
•
You cannot use a CA you generate if clients need to verify the identity of the server.
Authentication Modules and .htaccess
To restrict access to a Web page, Apache and third parties provide authentication modules and methods that can
verify a user’s credentials, such as a username and password. Some modules enable authentication against various
databases including LDAP (page 980) and NIS (page 637).
Commonly, user authentication directives are placed in a .htaccess file. A basic .htaccess file that uses the Apache
default authentication module (mod_auth) follows. Substitute appropriate values for the local server.
# cat .htaccess
AuthUserFile /var/www/.htpasswd
AuthGroupFile /dev/null
AuthName "Browser dialog box query"
AuthType Basic
require valid-user
The /var/www/.htpasswd is a typical absolute pathname of a .htpasswd file and Browser dialog box query is
the string that the user will see as part of the dialog box that requests a username and password.
The second line of the preceding .htaccess file turns off the group function. The fourth line specifies the user authentication type Basic, which is implemented by the default mod_auth module. The last line tells Apache which
users can access the protected directory. The entry valid-user grants access to the directory to any user who is in
222
Draft
Scripting Modules
Draft
the Apache password file and who enters the correct password. You can also specify Apache usernames separated
by SPACEs.
You can put the Apache password file anywhere on the system, as long as Apache can read it. It is safe to put this
file in the same directory as the .htaccess file as, by default, Apache will not answer any requests for files whose
names start with .ht.
The following command creates a .htpasswd file for Sam:
$ htpasswd -c .htpasswd sam
New password:
Re-type new password:
Adding password for user sam
Omit the –c option to add a user or to change a password in an existing .htpasswd file. Remember to use an AllowOverride directive (page 786) to permit Apache to read the .htaccess file.
Scripting Modules
Apache can process content before serving it to a client. In earlier versions of Apache, only CGI scripts could process
content; in the current version, scripting modules can work with scripts that are embedded in HTML documents.
Scripting modules manipulate content before Apache serves it to a client. Because they are built into Apache, they
are fast. Scripting modules are especially efficient at working with external data sources such as relational databases.
Clients can pass data to a scripting module that modifies the information that Apache serves.
Contrast scripting modules with CGI scripts that are run external to Apache: CGI scripts do not allow client interaction and are slow because of the external calls.
RedHat provides packages that allow you to embed Perl, Python, and PHP code in HTML content. Perl and Python,
general-purpose scripting languages, are encapsulated for use directly in Apache and are implemented in the
mod_perl and mod_python modules.
PHP, developed for manipulating Web content, outputs HTML by default. PHP, implemented in the mod_php
module, is easy to set up, has a syntax similar to Perl and C, and comes with a large number of Web-related functions.
webalizer: Analyzing Web Traffic
The webalizer package, which is typically installed as part of Apache, creates a directory at /var/www/usage and
a cron file (page 531) at /etc/cron.daily/00webalizer. Once a day, the cron file generates usage data and puts it
in the usage directory; you can view the data by pointing a browser at http://server/usage/ where server is the
hostname of the server.
The /etc/webalizer.conf file controls the behavior of the webalizer utility. If you change the location of the DocumentRoot or log files, you must edit this file to reflect those changes. For more information on webalizer, refer to
the webalizer man page and the sites listed under “More Information” on page 761.
MRTG: Monitoring Traffic Loads
Multi Router Traffic Grapher (MRTG) is an open source application that graphs statistics available through SNMP
(Simple Network Management Protocol). SNMP information is available on all high-end routers and switches, as
well as on some other networked equipment, such as printers and wireless access points. You can use the net-snmp
223
Draft
Error Codes
Draft
and net-snmp-utils rpm packages supplied by Red Hat to install SNMP on a system. You also need to install the
mrtg package.
Once MRTG and SNMP are installed and running, you can view the reports at http://server/mrtg where server
is the FQDN of your server. For more information, see the mrtg man page and the sites listed under “More Information” on page 761.
Error Codes
Following is a list of Apache error codes:
•
100 Continue
•
101 Switching Protocols
•
200 OK
•
201 Created
•
202 Accepted
•
203 Non-Authoritative Information
•
204 No Content
•
205 Reset Content
•
206 Partial Content
•
300 Multiple Choices
•
301 Moved Permanently
•
302 Moved Temporarily
•
303 See Other
•
304 Not Modified
•
305 Use Proxy
•
400 Bad Request
•
401 Unauthorized
•
402 Payment Required
•
403 Forbidden
•
404 Not Found
•
405 Method Not Allowed
•
406 Not Acceptable
•
407 Proxy Authentication Required
224
Draft
•
408 Request Time-out
•
409 Conflict
•
410 Gone
•
411 Length Required
•
412 Precondition Failed
•
413 Request Entity Too Large
•
414 Request-URI Too Large
•
415 Unsupported Media Type
•
500 Internal Server Error
•
501 Not Implemented
•
502 Bad Gateway
•
503 Service Unavailable
•
504 Gateway Time-out
•
505 HTTP Version not supported
Chapter Summary
Draft
Chapter Summary
Apache is the most popular Web server on the Internet today. It is robust and extensible. The /etc/httpd/conf/httpd.conf configuration file controls many aspects of how Apache runs. The Red Hat httpd.conf file, based on the
httpd.conf file distributed by Apache, is heavily commented and broken into three parts: Global Environment,
Main Server Configuration, and Virtual Hosts. You can use the system-config-httpd (FEDORA) and redhat-confighttpd (RHEL) utilities to modify httpd.conf.
Content to be served must be placed in /var/www/html, called the document root. Apache automatically displays
the file named index.html in this directory.
Configuration directives, or simply directives, are lines in a configuration file that control some aspect of how
Apache functions. There are four locations, called contexts, that define where a configuration directive can appear:
server config, virtual host, directory, and .htaccess. Containers, or special directives, are directives that group other
directives.
To restrict access to a Web page, Apache and third parties provide authentication modules and methods that can
verify a user’s credentials, such as a username and password. Some modules enable authentication against various
databases including LDAP and NIS.
Apache can respond to a request for a URI by asking the client to request a different URI. This response is called
a redirect. Apache can also process content before serving it to a client using scripting modules that work with
scripts embedded in HTML documents.
Apache supports virtual hosts, which means that one instance of Apache can respond to requests directed to multiple
IP addresses or hostnames as though it were multiple servers. Each IP address or hostname can provide different
content and be configured differently.
225
Draft
Exercises
Draft
The CGI (Common Gateway Interface) allows external application programs to interface with Web servers. Any
program can be a CGI program if it runs in real time and relays its output to the requesting client.
SSL (Secure Socket Layer) has two functions: It allows a client to verify the identity of a server and it enables secure
two-way communication between a client and server.
Exercises
1. How would you tell Apache that your content is in /usr/local/www?
2. How would you instruct an Apache server to listen on port 81 instead of port 80?
3. How would you enable Sam to publish Web pages from his ~/website directory but not allow anyone else to
publish to the Web?
4. Apache must be started as root. Why? Why does this not present a security risk?
Advanced Exercises
5. If you are running Apache on a firewall system, perhaps to display a Web front end for firewall configuration,
how would you make sure that it is accessible only from inside the local network?
6. Why is it more efficient to run scripts using mod_php or mod_perl than through CGI?
7. What two things does SSL provide and how does this differ if the certificate is self-signed?
8. Some Web sites generate content by retrieving data from a database and inserting it into a template using PHP
or CGI each time the site is accessed. Why is this often a poor practice?
9. Assume you want to provide Webmail access for employees on the same server that hosts the corporate Web
site. The Web site address is example.com, you want to use mail.example.com for Webmail, and the Webmail
application is located in /var/www/webmail. Describe two ways you can set this up.
10. Part of a Web site is a private intranet and is accessed as http://example.com/intranet. Describe how you would
prevent people outside the company accessing this site. Assume the company uses the 192.168.0.0/16 subnet
internally.
226
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 18 - OpenSSH: Secure Network Communication
OpenSSH: Secure Network Communication
IN THIS CHAPTER
About OpenSSH
OpenSSH Clients
JumpStart: Using ssh and scp
sshd: OpenSSH Server
JumpStart: Starting the sshd Daemon
Authorized Keys: Automatic Login
Troubleshooting
Tunneling/Port Forwarding
OpenSSH is a suite of secure network connectivity tools that replaces telnet, rcp, rsh/rshd, rlogin/rlogind, and
ftp/ftpd. Unlike the tools it replaces, OpenSSH tools encrypt all traffic, including passwords, thwarting malicious
users who would eavesdrop, hijack connections, and steal passwords.
This chapter covers the following OpenSSH tools:
scp .
Copies files to/from another system
sftp .
Copies files to/from other systems (a secure replacement for ftp)
ssh .
Runs a command on or logs in on another system
sshd .
The OpenSSH daemon (runs on the server)
ssh-keygen .
Creates RSA or DSA host/user authentication keys
Introduction
Using public-key encryption (page 923), OpenSSH provides two levels of authentication: server and client/user.
First, the client verifies that it is connected to the correct server. Then, OpenSSH encrypts communication between
the systems. Second, once a secure, encrypted connection has been established, OpenSSH makes sure that the user
is authorized to log in on or copy files from/to the server. Once the system and user have been verified, OpenSSH
allows different services to be passed through the connection. These services include interactive shell sessions (ssh),
remote command execution (ssh and scp), X11 client/server connections, and TCP/IP port tunneling.
227
Draft
About OpenSSH
Draft
SSH1 versus SSH2 . SSH protocol version 2 (SSH2) is a complete rewrite of the SSH protocol version 1 (SSH1)
with improved security, performance, and portability. The two protocols are not compatible. Because SSH1 is being
rapidly supplanted by SSH2 and because SSH1 is subject to a person in the middle attack (footnote 3 on page 926),
this chapter does not discuss SSH1. Because version 2 is floating-point intensive, version 1 does have a place on
systems without FPUs (floating-point units or accelerators), such as old 486SX systems. As initially installed, the
OpenSSH tools supplied with Red Hat support both protocols; you need run only one server to communicate with
systems using either protocol.
ssh . The ssh utility allows you to log in on a remote system over a network. You might choose to use a remote
system to access a special-purpose application or to use a device that is available only on that system, or you might
use a remote system because you know that it is faster or not as busy as the local computer. While traveling, many
people use ssh on a laptop to log in on a system at headquarters. From a GUI you can use several systems simultaneously by logging in on each from a different terminal emulator window.
X11 forwarding . With X11 forwarding turned on, as it is when you install Red Hat Linux, it is a simple matter
to run an X11 program over an ssh connection: Run ssh from a terminal emulator running in a GUI and give an
X11 command such as xclock; the graphical output appears on the local display. For more information, refer to
“Forwarding X11” on page 580.
About OpenSSH
This section discusses configuration files that OpenSSH clients and servers use, describes how OpenSSH works,
and provides additional OpenSSH resources.
Files
OpenSSH clients and servers rely on many files. Global files are kept in /etc/ssh and user files in ~/.ssh. In the
description of each file, the first word indicates whether the client or the server uses the file.
Caution: rhost Authentication Is a Security Risk
Although OpenSSH can get authentication information from /etc/hosts.equiv, /etc/shosts.equiv,
~/.rhosts, and ~/.shosts, this chapter does not cover the use of these files because they are security
risks and the default settings in the /etc/ssh/sshd_config configuration file prevent their use.
/etc/ssh: Global Files
Global files listed in this section effect all users but can be overridden by files in a user’s ~/.ssh directory.
moduli . client and server Contains key exchange information that is used to establish a secure connection. Do
not modify this file.
ssh_config . client The global OpenSSH configuration file (page 573). Entries here can be overridden by entries
in a user’s ~/.ssh/config file.
sshd_config .
server The configuration file for sshd (page 577).
ssh_host_dsa_key, ssh_host_dsa_key.pub . server SSH protocol version 2 DSA host keys. Both files should
be owned by root. The ssh_host_dsa_key.pub public file should be readable by anyone but writable only by its
owner (644 permissions). The ssh_host_dsa_key private file should not be readable or writable by anyone except
its owner (600 permissions).
228
Draft
How OpenSSH Works
Draft
ssh_host_rsa_key, ssh_host_rsa_key.pub . server SSH protocol version 2 RSA host keys. Both files should be
owned by root. The ssh_host_rsa_key.pub public file should be readable by anyone but writable only by its
owner (644 permissions). The ssh_host_rsa_key private file should not be readable or writable by anyone except
its owner (600 permissions).
ssh_known_hosts . client Contains public RSA (by default) keys of hosts that users on the local system can
connect to. This file contains information similar to ~/.ssh/known_hosts, except it is set up by the administrator
and is available to all users. This file should be owned by root and should be readable by anyone but writable only
by its owner (644 permissions).
sshrc . server Contains initialization routines. If ~/.ssh/rc is not present, this script is run after ~/.ssh/environment and before the user’s shell is started.
~/.ssh: User Files
OpenSSH creates the ~/.ssh directory and the known_hosts file therein automatically when you connect to a remote
system.
authorized_keys . server Enables you to log in on or copy files from/to another system without supplying a
password (page 575). No one except the owner should be able to write to this file.
config . client A user’s private OpenSSH configuration file (page 573). Entries here override those in
/etc/ssh/ssh_config.
environment . server Contains commands that are executed when a user logs in with ssh. Similar in function
to ~/.bashrc for a local bash shell.
id_dsa, id_dsa.pub . client User authentication DSA keys generated by ssh-keygen (page 576). Both files should
be owned by the user in whose home directory they appear. The id_dsa_key.pub public file should be readable
by anyone but writable only by its owner (644 permissions). The id_dsa_key private file should not be readable
or writable by anyone except its owner (600 permissions).
id_rsa, id_rsa.pub . client User authentication RSA keys generated by ssh-keygen (page 576). Both files should
be owned by the user in whose home directory they appear. The id_rsa_key.pub public file should be readable
by anyone but writable only by its owner (644 permissions). The id_rsa_key private file should not be readable
or writable by anyone except its owner (600 permissions).
known_hosts . client Contains public RSA keys (default) of hosts that the user has connected to. OpenSSH
automatically adds entries each time the user connects to a new server (page 568). Refer to “HostKeyAlgorithms”
(page 574) for information on using DSA keys.
rc . server Contains initialization routines. This script is run after environment and before the user’s shell is
started. If this file is not present, OpenSSH runs /etc/ssh/sshrc; if that file does not exist, OpenSSH runs xauth.
How OpenSSH Works
When OpenSSH starts, it first establishes an encrypted connection and then authenticates the user. Once these
two tasks are taken care of, OpenSSH allows the two systems to send information back and forth.
OpenSSH uses two key pairs to negotiate an encrypted session: a host key pair and a session key pair. The host key
pair is a set of public/private keys that is established the first time the server system runs sshd (page 575). The
session key pair is a set of public/private keys that changes hourly.
The first time an OpenSSH client connects with an OpenSSH server, you are asked to verify that it is connected to
the correct server (see “JumpStart: Using ssh and scp ” on page 567). After verification, the client makes a copy of
229
Draft
More Information
Draft
the server’s public host key. On subsequent connections, the client compares the key provided by the server with
the key it stored. Although this test is not foolproof, the next one is quite secure.
Next, the client generates a random key, which it encrypts with both the server’s public host key and the session
key. The client then sends this encrypted key to the server. The server uses its private keys to decrypt the key. This
process creates a key that is known only to the client and server and is used to encrypt the rest of the session.
More Information
Local .
Web .
man pages ssh scp ssh-keygen ssh_config sshd sshd_config
OpenSSH home page www.openssh.com [http://www.openssh.com]
Search tldp.org for ssh for various HOWTOs and other documents.
Books .
2003)
Implementing SSH: Strategies for Optimizing the Secure Shell by Dwivedi; John Wiley & Sons (October
SSH, The Secure Shell: The Definitive Guide by Barrett & Silverman; O’Reilly & Associates, 1st edition (February
15, 2001)
OpenSSH Clients
This section covers setting up and using the ssh, scp, and sftp clients.
Prerequisites
Install the following packages:
•
openssh
•
openssh-clients
There are no startup commands for OpenSSH clients.
JumpStart: Using ssh and scp
The ssh and scp clients do not require setup beyond installing the requisite packages, although you can create/edit
files that facilitate their use. In order to run a secure shell on, or securely copy a file to/from a remote system, the
remote system must be running the OpenSSH daemon (sshd); you must have an account on the remote system;
and the server must positively identify itself to the client. The following example shows a user logging in on grape
as zach and then giving an exit command to return to the shell on the local system:
$ ssh [email protected]
[email protected]'s password:
[[email protected] zach]$ exit
Connection to grape closed.
$
You can omit [email protected] ([email protected] in the preceding example) from the command line if you have the same username
on both systems. The first time you connect to a remote OpenSSH server, ssh or scp asks you to make sure that
you are connected to the right system. Refer to “Message on initial connection to a server” on page 568.
230
Draft
Setup
Draft
Following is an example of copying a file to a remote system using scp:
$ scp ty1 [email protected]:
[email protected]'s password:
ty1
100% |*****************************| 1311
00:00
Setup
This section describes how to set up OpenSSH on the client side.
Recommended Settings
X11 forwarding . The configuration files provided by Red Hat provide a mostly secure system and may or may
not meet your needs. The important OpenSSH default value that the Red Hat configuration files override is ForwardX11, which is set to yes in the Red Hat /etc/ssh/ssh_config configuration file (page 578). See page 580 for
more information on X11 forwarding.
Server Authentication/Known Hosts
known_hosts, ssh_known_hosts . There are two files that list the hosts the local system has connected to and
positively identified: ~/.ssh/known_hosts (user) and /etc/ssh/ssh_known_hosts (global). No one except the
owner (root in the case of the second file) should be able to write to either of these files. No one except the owner
should have any access to a ~/.ssh directory.
When you connect to an OpenSSH server for the first time, the OpenSSH client prompts you, following, to confirm
that you are connected to the right system. This checking can help prevent a person-in-the-middle attack (footnote
3 on page 926).
Message on initial connection to a server .
The authenticity of host 'grape (192.168.0.3)' can't be established.
RSA key fingerprint is c9:03:c1:9d:c2:91:55:50:e8:19:2b:f4:36:ef:73:78.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'grape,192.168.0.3' (RSA) to the list of
known hosts.
Before you respond to the preceding query, make sure you are logging in on the correct system and not on an imposter. If you are not sure, a telephone call to someone who logs in on that system locally can help verify that you
are on the intended system. When you answer yes (you must spell it out), the client appends the server’s public
host key (the single line in the /etc/ssh/ssh_host_rsa_key.pub or /etc/ssh/ssh_host_dsa_key.pub file on the
server) to the user’s ~/ssh/known_hosts file on the local client, creating the ~/.ssh directory if necessary. So that
it can keep track of which line in known_hosts applies to which server, OpenSSH prepends the name of the server
and the server’s IP address (by default) to the line.
Subsequently, when you use OpenSSH to connect to that server, the client verifies that it is connected to the correct
server by comparing this key to the one the server supplies.
The known_hosts file uses one very long line to identify each host it keeps track of. Each line starts with the
hostname and IP address of the system the line corresponds to, followed by the type of encryption being used and
the server’s public host key. The following line (it is one logical line wrapped on to three physical lines) from
known_hosts is used to connect to grape at 192.168.0.3 using RSA (page 994) encryption.
231
Draft
ssh: Connect to and Execute Commands on a Remote System
Draft
known_hosts file .
$ cat ~/.ssh/known_hosts
grape,192.168.0.3 ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAIEArinPGsaLUtnSL4V7b
T51ksF7KoScsIk7wqm+2sJEC43rxVNS5+MO/O64UXp5qQOHBmeLCCFCsIJg8xseuVkg9iwO
BKKOdlZdBNVqFS7tnJdBQTFf+ofPIDDip8w6ftHOdM8hZ/diQq5gXqMH+Mpac31pQXAxXgY
SP8NYIgb3X18=
OpenSSH automatically stores keys from servers it has connected to in user-private files (~/.ssh/known_hosts).
These files work only for the user whose directory they appear in. Working as root and using a text editor, you
can copy lines from a user’s private list of known hosts to the public list in /etc/ssh/ssh_known_hosts to make
a server known globally on the local system.
If, after you have a remote system’s private key stored in one of the known hosts files, the remote system supplies
a different fingerprint when you attempt to connect, OpenSSH displays the following message and does not complete
the connection:
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@
WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED!
@
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
IT IS POSSIBLE THAT SOMEONE IS DOING SOMETHING NASTY!
Someone could be eavesdropping on you right now (man-in-the-middle attack)!
It is also possible that the RSA host key has just been changed.
The fingerprint for the RSA key sent by the remote host is
f1:6f:ea:87:bb:1b:df:cd:e3:45:24:60:d3:25:b1:0a.
Please contact your system administrator.
Add correct host key in /home/sam/.ssh/known_hosts to get rid of this message.
Offending key in /home/sam/.ssh/known_hosts: 1
RSA host key for grape has changed and you have requested strict checking.
Host key verification failed.
When you see this message, it is possible that you are the subject of a person-in-the-middle attack. It is more likely
that something on the remote system changed, causing it to supply a new fingerprint. Check with the remote system’s
administrator. If all is well, remove the offending key from the specified file (the third line from the bottom in the
preceding example points to the line you need to remove) and try connecting again. You will see the “Message on
initial connection to a server” (page 568) again as OpenSSH verifies that you are connecting to the correct system.
Follow the same steps as when you initially connected to the remote host.
ssh: Connect to and Execute Commands on a Remote System
The format of an ssh command line is
ssh [[email protected]]host [command]
where host, the name of the OpenSSH server you want to connect to, is the only required argument. The host can
be a local system name, an FQDN of a system on the Internet, or an IP address. Give the command ssh host to log
in on the remote system host with the same username that you are using on the local system. Include user @ when
you want to log in with a username other than the one you are using on the local system. Depending on how things
are set up, you may need to supply your password.
232
Draft
ssh: Connect to and Execute Commands on a Remote System
Draft
Opening a remote shell . When you do not include command, ssh logs you in on host. You get a shell prompt
and can run commands on host. Give the command exit to close the connection to host and return to the local
system’s prompt.
[bravo]$ ssh speedy
[email protected]'s password:
Last login: Sat Sep 14 06:51:59 from bravo
Have a lot of fun...
You have new mail.
[speedy]$
...
[speedy]$ exit
Connection to speedy closed.
[bravo]$
Running a remote command . When you include command, ssh logs in on host, executes command, closes the
connection to host, and returns control to the local system; you never get the remote system’s prompt.
The following example runs ls in the memos directory on the remote system speedy. The example assumes that
the user running the command (Alex) has a login on speedy and that the memos directory is in Alex’s home directory on speedy:
[bravo]$ ssh speedy ls memos
[email protected]'s password:
memo.0921
memo.draft
[bravo]$
For the next example, assume a file named memo.new is in the working directory on the local system (bravo) and
you cannot remember whether it contains certain changes or you made these changes to the file named memo.draft
on speedy. You could copy memo.draft to the local system and run diff (page 126) on the two files, but then you
would have three similar copies of the file spread across two systems. If you are not careful about removing the old
copies when you are done, you may be confused again in a few days. Instead of copying the file, you can use ssh:
[bravo]$ ssh speedy cat memos/memo.draft | diff memos.new –
When you run ssh, standard output of the command run on the remote system is passed to the local shell as though
the command had been run in place on the local system. As with all shell commands, you must quote special
characters that you do not want interpreted by the local system. In the preceding example, the output of the cat
command on speedy is sent through a pipe on bravo to diff (running on bravo), which compares the local file
memos.new to standard input (–). The following command line has the same effect but causes diff to run on the
remote system:
[bravo]$ cat memos.new | ssh speedy diff – memos/memo.draft
Standard output from diff on the remote system is sent to the local shell, which displays it on the screen (because
it is not redirected).
Options
This section lists some of the options you can use with ssh.
233
Draft
scp: Copying a File from/to a Remote
System
Draft
–C .
compression .
compression.)
Enables compression. (In the commercial version of ssh, –C disables compression and +C enables
–f .
not foreground . Sends ssh to the background after asking for a password and before executing the command.
Useful when you want to run the command in the background but must supply a password. Implies –n.
–L . Forwards a port on the local client to a remote system. For more information, refer to “Tunneling/Port
Forwarding” on page 579.
–l user .
login .
(ell) Attempts to log in as user.
–n .
null .
Redirects standard input to ssh to come from /dev/null. Required when running ssh in the background.
–o option .
option .
Specifies option in the format used in configuration files (page 573).
–p .
port . Specifies the port on the remote host that the connection is made to. Using the host declaration (page
573) in the configuration file, you can specify a different port for each system you connect to.
–R . Forwards a port on the remote system to the local client. For more information, refer to “Tunneling/Port
Forwarding” on page 579.
–v .
verbose .
pected.
Displays debugging messages about the connection and transfer. Useful if things are not going as ex-
–X .
X11 . Turns on X11 forwarding. You do not need this option if you turn on X11 forwarding in the configuration
file (page 568).
–x .
X11 .
Turns off X11 forwarding.
scp: Copying a File from/to a Remote System
The scp (secure copy) utility copies a file from one computer to another on a network, using the same authentication
mechanism as ssh and therefore providing the same security. As does ssh, scp asks you for a password when it is
needed. The format of an scp command is
scp [[user
@]
234
Draft
scp: Copying a File from/to a Remote
System
Draft
fromhost:]source-file [[[email protected]]tohost:]destination-file
where source-file is the file you want to copy and destination-file is the resulting copy. A simple or relative filename
is assumed to be relative to your home directory on the remote system and relative to your working directory on
the local system. An absolute pathname describes a path from the root directory on either system. Make sure that
you have read permission to the file you are copying and write permission for the directory you are copying it into.
The fromhost: and tohost: are the names of the systems that (will) hold the files. When you do not specify fromhost
or tohost, each defaults to the local system. The fromhost and tohost can be local system names, FQDNs (page 972)
of a systems on the Internet, or IP addresses. You can copy from or to the local system or between two remote
systems.
Sam has an alternate username, sls, on grape. In the following example, Sam uses scp to copy memo.txt from the
home directory of his sls account on grape to the allmemos directory in the working directory on the local system.
If allmemos were not the name of a directory, memo.txt would be copied to a file named allmemos in the working
directory.
$ scp [email protected]:memo.txt allmemos
[email protected]'s password:
memo.txt
100% |*****************************| 14664
00:00
As the transfer progresses, the percent and number of bytes transferred increase and the time remaining decreases.
The asterisks provide a visual representation of the progress of the transfer.
The next example shows Sam, working from peach, copying the same file as in the previous example to the directory
named old in Sam’s home directory on speedy. For this example to work, Sam must be able to use ssh to log in
on speedy from grape without using a password. For more information, refer to “Authorized Keys: Automatic
Login” on page 575.
$ [[email protected]] scp [email protected]:memo.txt speedy:old
[email protected]'s password:
Options
This section lists some of the options you can use with scp.
–C .
compression .
Enables compression.
–o option .
option .
Specifies option in the format used in configuration files (next section).
–P port .
port .
Connects to port port on the remote host.
–p .
preserve .
Preserves the modification and access times as well as the modes of the original file.
–q .
235
Draft
quiet .
sftp: A Secure FTP Client
Draft
Does not display the progress meter.
–r .
recursive .
Recursively copies a directory hierarchy.
–v .
verbose .
pected.
Displays debugging messages about the connection and transfer. Useful if things are not going as ex-
sftp: A Secure FTP Client
As part of OpenSSH, Red Hat provides sftp, a secure alternative to ftp (page 583). Functionally the same as ftp, sftp
maps ftp commands into OpenSSH commands. You can replace ftp with sftp when you are logging into a server
that is running the OpenSSH daemon, sshd. Once you are connected to a system with sftp, give the command ?
to display a list of commands. For secure communication, use sftp or scp for all file transfers requiring authentication.
Refer to the sftp man page for more information.
~/.ssh/config and /etc/ssh/ssh_config Configuration Files
It is usually not necessary to modify OpenSSH client configuration files. For a given user there are two configuration
files: ~/.ssh/config (user) and /etc/ssh/ssh_config (global). These files are read in this order and, for a given
parameter, the first one found is the one that is used. A user can override a global parameter setting by setting the
same parameter in his user configuration file. Parameters given on the ssh or scp command line take precedence
over parameters set in either of these files.
Lines in the configuration files contain declarations that start with a keyword, which is not case sensitive, followed
by whitespace, and end with case-sensitive arguments.
You can use the Host keyword to cause declarations to apply to a specific system. A Host declaration applies to
all the lines that follow it until the next Host declaration. You can use * and ? wildcards within a hostname.
Host hostnames . Specifies that the following declarations, until the next Host declaration, apply to hostnames
only. The hostnames should be in the same form you would use on a command line and can contain ? and * wildcards.
Use a single * to specify all hosts.
CheckHostIP yes | no . Uses an IP address in addition to a hostname to identify a system in the known_hosts
file when set to yes (default). Set to no to use a hostname only.
ForwardX11 yes | no . Automatically forwards X11 connections over a secure channel and sets the DISPLAY
shell variable when set to yes. (Alternatively, you can use –X on the command to redirect X11 connections.) The
default is no, but the Red Hat configuration files set ForwardX11 to yes. In order for X11 forwarding to work,
X11Forwarding must also be set to yes in the /etc/sshd_config file on the server (page 578). For more information,
refer to “Forwarding X11” on page 580.
HostbasedAuthentication yes | no .
set to no (default).
Tries rhosts authentication when set to yes. For a more secure system,
HostKeyAlgorithms algorithms . The algorithms is a comma-separated list of algorithms that the client wants
to use in order of preference. Choose algorithms from: ssh-rsa, ssh-dss (default is ssh-rsa, ssh-dss).
KeepAlive yes | no . Periodically checks to see if a connection is alive when set to yes (default). Checking causes
the ssh or scp connection to be dropped when the server crashes or the connection dies for another reason, even if
it is only temporary. Setting this parameter to no causes the client not to check if the connection is alive.
236
Draft
sshd: OpenSSH Server
Draft
This declaration uses the TCP keepalive option, which is not encrypted and is susceptible to IP spoofing (page
978). Refer to ClientAliveInterval on page 577 for a server-based nonspoofable alternative.
StrictHostKeyChecking yes | no | ask . Determines whether and how OpenSSH adds host keys to a user’s
known_hosts file. Set to ask (default) to ask whether to add a host key when connecting to a new system, no to
add a host key automatically, and yes to require that host keys be manually added. The yes and ask arguments
cause OpenSSH to refuse to connect to a system whose host key has changed. For a more secure system, set to yes
or ask.
User name . Specifies a username to use when logging in on a system. Specify systems with the Host declaration.
Avoids having to enter a username on the command line when you are using a username that differs from your
username on the local system.
sshd: OpenSSH Server
This section discusses how to set up an OpenSSH server.
Prerequisites
Install the following packages:
•
openssh
•
openssh-server
Run chkconfig to cause sshd to start when the system goes multiuser:
# /sbin/chkconfig sshd on
See “Starting sshd for the First Time,” following, for information on starting the server for the first time.
JumpStart: Starting the sshd Daemon
Install the requisite packages and start the sshd daemon as described following. Look in /var/log/messages to
make sure everything is working properly.
Recommended Settings
The configuration files provided by Red Hat establish a mostly secure system and may or may not meet your needs.
The Red Hat /etc/ssh/sshd_config file turns on X11 forwarding (page 580). For a more secure system, you can
set PermitRootLogin to no, thereby removing a known-name, privileged account that is exposed to the outside
world with only password protection.
Starting sshd for the First Time
When you start the sshd OpenSSH daemon for the first time, generally when you boot the system, it creates host
key files (page 565) in /etc/ssh:
# /sbin/service sshd start
Generating SSH1 RSA host key:
Generating SSH2 RSA host key:
[
[
OK
OK
]
]
237
Draft
Authorized Keys: Automatic Login
Generating SSH2 DSA host key:
Starting sshd:
[
[
OK
OK
Draft
]
]
OpenSSH uses the files it creates to identify the server.
Authorized Keys: Automatic Login
You can configure OpenSSH so you do not have to enter a password each time you connect to a remote system.
To set things up, you need to generate a personal authentication key, place the public part of the key on the remote
server, and keep the private part of the key on the local client. When you connect, the remote system issues a
challenge based on the public part of the key. The private part of the key is required to respond properly to the
challenge. When the local system provides the proper response, the remote system logs you in.
The first step in setting up an automatic login is to generate your personal authentication keys. Check to see if authentication keys are already in place: Look in ~/.ssh for either id_dsa and id_dsa.pub or id_rsa and id_rsa.pub.
If one of these pairs of files is present, skip the next step (do not create a new key).
The ssh-keygen utility creates the public and private parts of an RSA key:
ssh-keygen .
$ ssh-keygen -t rsa
Generating public/private rsa key pair.
Enter file in which to save the key (/home/sam/.ssh/id_rsa): RETURN
Enter passphrase (empty for no passphrase):RETURN
Enter same passphrase again:RETURN
Your identification has been saved in /home/sam/.ssh/id_rsa.
Your public key has been saved in /home/sam/.ssh/id_rsa.pub.
The key fingerprint is:
f2:eb:c8:fe:ed:fd:32:98:e8:24:5a:76:1d:0e:fd:1d [email protected]
Replace rsa with dsa to generate DSA keys. In this example, the user pressed RETURN in response to each of the
queries. You have the option of specifying a passphrase (10–30 characters is a good length) to encrypt the private
part of the key. There is no way to recover a lost passphrase. See the following security tip for more information
about the passphrase. The ssh-keygen utility generates two keys: A private key or identification in ~/.ssh/id_rsa
and a public key in ~/.ssh/id_rsa.pub. No one except the owner should be able to write to either of these files.
Only the owner should be able to read from the private key file.
authorized_keys . To enable you to log in on or copy files from/to another system without supplying a password,
first create a ~/.ssh directory with permissions set to 700 on the remote system. Automatic login will fail if anyone
except the owner has permission to read from or write to the ~/.ssh directory on either system. Next, copy
~/.ssh/id_rsa.pub on the local system to a file named ~/.ssh/authorized_keys on the remote system. No one
except the owner should be able to read from or write to this file. Now when you run ssh or scp to access the remote
system, you do not have to supply a password.
Security: Encrypt Your Personal Key
The private part of the key is kept in a file that only you can read. If a malicious user compromises
your account or the root account on the local system, that user then has access to your account on
the remote system because she can read the private part of your personal key.
Encrypting the private part of your personal key protects the key and, therefore, access to the remote
system should someone compromise your local account. However, if you encrypt your personal key,
238
Draft
Command Line Options
Draft
you have to supply the passphrase you used to encrypt the key each time you use the key, negating
the benefit of not having to type a password in order to log in on the remote system. Also, most
passphrases that you can remember can be cracked quite quickly by a powerful computer.
A better idea is to store the private keys on a removable medium, such as a USB flash drive, and have
your ~/.ssh directory as the mountpoint for the filesystem stored on this drive.
Command Line Options
Command Line options override declarations in the configuration files (following). Following is a list of some of
the more useful sshd options.
–d .
debug . Sets debug mode wherein sshd sends debugging messages to the system log and the server stays in the
foreground. You can specify this option up to three times to increase the verbosity of the output. See also –e. (The
ssh client uses –v for debugging, page 571.)
–e .
error .
Sends output to standard error, not to the system log. Useful with –d.
–f file .
file .
Specifies the file with the pathname file as the default configuration file in place of /etc/ssh/sshd_config.
–t .
test .
Checks the configuration file syntax and the sanity of the key files.
–D .
noDetach .
Keeps sshd in the foreground. Useful for debugging, implied by –d.
/etc/ssh/sshd_config Configuration File
Lines in the /etc/ssh/sshd_config configuration file contains declarations that start with a keyword, which is not
case sensitive, followed by whitespace, and end with case-sensitive arguments.
AllowUsers userlist . The userlist is a SPACE-separated list of usernames that specifies users who are allowed
to log in using sshd. The list can include * and ? wildcards. You can specify a user as user or [email protected] If you use
the second format, make sure that you specify the host as returned by hostname. Without this declaration, any user
who can log in locally can log in using an OpenSSH client.
ClientAliveInterval n . Sends a message through the encrypted channel after n seconds of not receiving a
message from the client. See ClientAliveCountMax. Default is 0, meaning that no messages are sent.
This declaration passes messages over the encrypted channel and is not susceptible to IP spoofing (page 978). This
declaration differs from KeepAlive, which uses the TCP keepalive option and is susceptible to IP spoofing.
ClientAliveCountMax n . The n specifies the number of client alive messages that can be sent without receiving
a response before sshd disconnects from the client. See ClientAliveInterval. Default is 3.
HostbasedAuthentication yes | no .
set to no (default).
Tries rhosts authentication when set to yes. For a more secure system,
239
Draft
Troubleshooting
Draft
IgnoreRhosts yes | no . Ignores .rhosts and .shosts files for authentication. Does not affect the use of
/etc/hosts.equiv and /etc/ssh/shosts.equiv files for authentication. For security, set to no (default).
KeepAlive yes | no . Periodically checks to see if a connection is alive when set to yes (default). Checking causes
the ssh or scp connection to be dropped when the client crashes or the connection dies for another reason, even if
it is only temporary. Setting this parameter to no causes the server not to check if the connection is alive.
This declaration uses the TCP keepalive option, which is not encrypted and is susceptible to IP spoofing (page
978). Refer to ClientAliveInterval (page 577) for a nonspoofable alternative.
LoginGraceTime n . Waits n seconds for a user to log in on the server before disconnecting. A value of 0 means
there is no time limit. The default is 120.
LogLevel val . Specifies how detailed the log messages are. Choose val from QUIET, FATAL, ERROR, INFO,
and VERBOSE. The default is INFO.
PermitEmptyPasswords .
no.
PermitRootLogin .
default is yes.
Permits a user to log in to an account that has an empty password. The default is
Permits root to log in using an OpenSSH client. For a more secure system, set to no. The
StrictModes yes | no . Checks modes and ownership of user’s home directory and files. Login fails if directories
and/or files can be written to by anyone. For security, set to yes (default).
VerifyReverseMapping yes | no . Checks to see that the client’s IP address reverse maps (page 708) to the host
name the user is logging in from. The default is no.
X11Forwarding yes | no . Allows X11 forwarding when set to yes. The default is no, but the Red Hat configuration files set X11Forwarding to yes. In order for X11 forwarding to work, ForwardX11 must also be set to
yes in either the ~/.ssh/config or the /etc/ssh/ssh_config client configuration file (page 574). For security, do
not allow untrusted users to open X11 connections.
Troubleshooting
Log files . There are several places to look for clues when you have problems connecting with ssh or scp. First,
look for sshd entries in /var/log/secure and /var/log/messages on the server. Following are messages you may
see when you are using an AllowUsers declaration but have not included the user who is trying to log in (page
577).
# grep sshd /var/log/secure
grape sshd[16]: User sam not allowed because not listed in AllowUsers
grape sshd[16]: Failed password for illegal user sam from 192.168.0.6 port 59276 ssh2
The next messages originate with PAM (page 416) and indicate that the user is not known to the system.
# grep sshd /var/log/messages
grape sshd(pam_unix)[2817]: check pass; user unknown
grape sshd(pam_unix)[2817]: authentication failure; logname= uid=0
euid=0 tty=NODEVssh ruser= rhost=peach.sobell.com
240
Draft
Tunneling/Port Forwarding
Draft
Debug the client . If entries in these files do not help solve the problem, try connecting with the –v option (ssh
or scp, the results should be the same). OpenSSH displays a lot of messages and one of them may help you figure
out what the problem is.
$ ssh -v grape
OpenSSH_3.1p1, SSH protocols 1.5/2.0, OpenSSL 0x0090602f
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: Applying options for *
debug1: Rhosts Authentication disabled, originating port will not be trusted.
debug1: restore_uid
debug1: ssh_connect: getuid 500 geteuid 0 anon 1
debug1: Connecting to grape [192.168.0.3] port 22.
debug1: temporarily_use_uid: 500/500 (e=0)
...
debug1: Host 'grape' is known and matches the RSA host key.
debug1: Found key in /home/sam/.ssh/known_hosts:1
debug1: bits set: 1617/3191
debug1: ssh_rsa_verify: signature correct
...
debug1: authentications that can continue: publickey,password,keyboard-interactive
debug1: next auth method to try is publickey
debug1: try privkey: /home/sam/.ssh/identity
debug1: try privkey: /home/sam/.ssh/id_rsa
debug1: try privkey: /home/sam/.ssh/id_dsa
debug1: next auth method to try is keyboard-interactive
debug1: authentications that can continue: publickey,password,keyboard-interactive
debug1: next auth method to try is password
[email protected]'s password:
Debug the server . You can debug from the server side by running sshd with the –de options. The server will
run in the foreground and its display may help you solve the problem.
Tunneling/Port Forwarding
The ssh utility allows you to forward a port (port forwarding, page 989) through the encrypted connection it establishes. Because the data sent across the forwarded port uses the encrypted ssh connection as its data link layer (page
332), the term tunneling (page 1002) is applied to this type of connection: “The connection is tunneled through
ssh.” You can secure protocols including POP, X, IMAP, and WWW by tunneling them through ssh.
Forwarding X11 . The ssh utility makes it easy to tunnel the X11 protocol. In order for X11 tunneling to work,
you must enable it in both the server and the client. Red Hat enables X11 tunneling in its default setup. If necessary,
you can enable X11 tunneling on the server by setting the X11Forwarding declaration to yes in the
/etc/ssh/sshd_config file (page 578). On the client, you can use the –X option to ssh on the command line or set
the ForwardX11 declaration to yes in the ~/.ssh/config or /etc/ssh/ssh_config file (page 574) to enable X11
tunneling.
With X11 forwarding turned on, ssh tunnels the X11 protocol, setting the $DISPLAY environment variable on the
system it connects to and forwarding the required port. You must be running from a GUI, which usually means
that you are using ssh on a terminal emulator to connect to a remote system. When you give an X11 command
from an ssh prompt, OpenSSH creates a new secure channel that carries the X11 data. The graphical output from
the X11 program appears on your screen.
241
Draft
Tunneling/Port Forwarding
Draft
[peach] $ ssh -X speedy
[speedy] $ echo $DISPLAY
localhost:10.0
By default, ssh uses X Window System display numbers 10 and higher (port numbers 6010 and higher) for forwarded
X sessions. Once you connect to a remote system using ssh –X, you can give a command to run an X application
and it will run on the remote system with its display on the local system: It appears to run locally.
Port forwarding . You can forward arbitrary ports using the –L and –R options. The –L option forwards a
local port to a remote system, so that a program that tries to connect to the forwarded port on the local system
transparently connects to the remote system. The –R option does the reverse, forwarding remote ports to the local
system. The –N option, which prevents ssh from executing remote commands, is generally used with –L and –R.
When you specify –N, ssh works only as a private network to forward ports. The format of an ssh command line
using one of these options follows:
$ ssh –N –L | –R local-port:remote-host:remote-port target
where
local-port is the number of the local port that is being forwarded to or from remote-host.
remote-host is the name or IP address of the system that local-port gets forwarded to or from.
remote-port is the number of the port on remote-host that is being forwarded from or to the local system.
target is the name or IP address of the system ssh connects to.
As an example, assume that there is a POP mail client on the local system and that the POP server is on a remote
network, on a system named pophost. POP is not a secure protocol; passwords are sent in cleartext each time the
client connects to the server. You can make it more secure by tunneling POP through ssh. (POP-3 connects on port
110; port 1550 is an arbitrary port on the local system.)
$ ssh -N -L 1550:pophost:110 pophost
After giving the preceding command, you can point the POP client at localhost:1550, and the connection between
the client and the server will be encrypted. (When you set up an account on the POP client, specify the location of
the server as localhost, port 1550; details vary with different mail clients.) In this example, remote-host and target
are the same system.
Firewalls . The system specified for port forwarding (remote-host) does not have to be the same as the destination
of the ssh connection (target). For example, assume the POP server is behind a firewall and you cannot connect to
it via ssh. If you can connect to the firewall via the Internet using ssh, you can encrypt the part of the connection
over the Internet.
$ ssh -N -L 1550:pophost:110 firewall
In the preceding example, remote-host, the system receiving the port forwarding, is pophost, and target, the system
that ssh connects to, is firewall.
You can also use ssh when you are behind a firewall (that is running sshd) and want to forward a port into your
system without modifying the firewall settings.
242
Draft
Chapter Summary
Draft
$ ssh -R 1678:localhost:80 firewall
The preceding command forwards connections from the outside to port 1678 on your firewall back to the local
Web server. Forwarding connections in this manner allows you to use a Web browser to connect to port 1678 on
your firewall in order to connect to the Web server on your machine. This setup would be useful if you ran a
Webmail program (page 625) on your system because it would allow you to check your mail from anywhere using
an Internet connection.
Compression . Compression, enabled with the –C option, can speed up communication over a low-bandwidth
connection. This option is commonly used with port forwarding. Compression can increase latency to an extent
that may not be desirable for an X session forwarded over a high-bandwidth connection.
Chapter Summary
OpenSSH is a suite of secure network connectivity tools that encrypts all traffic, including passwords, thwarting
malicious users who would eavesdrop, hijack connections, and steal passwords. The components discussed in this
chapter are sshd (the server daemon), ssh (runs a command on or logs in on another system), scp (copies files
to/from another system), sftp (securely replaces ftp) and ssh-keygen (creates authentication keys).
In order to ensure secure communications, when an OpenSSH client opens a connection, it first verifies that it is
connected to the correct server. Then OpenSSH encrypts communication between the systems. Finally, OpenSSH
makes sure that the user is authorized to log in on or copy files from/to the server.
OpenSSH also enables secure X11 forwarding. With this feature, you can run securely a graphical program on a
remote system and have the display appear on the local system.
Exercises
1. What is the difference between the scp and sftp utilities?
2. How can you use ssh to find out who is logged in on a remote system?
3. How would you use scp to copy your ~/.bashrc file from bravo to the local system?
4. How would you use ssh to run xterm on bravo and show the display on the local system?
5. What problem can enabling compression present when using ssh to run remote X11 applications on a local display?
6. When you try to connect to another system using an OpenSSH client and you see a message warning you that
the remote host identification has changed, what has happened? What should you do?
Advanced Exercises
7. What scp command would you use to copy your home directory from bravo to the local system?
8. What single command could you give to log in as root on the remote system named bravo, if bravo has remote
root logins disabled?
9. How could you use ssh to compare the contents of the ~/memos directories on bravo and the local system?
243
Draft
Advanced Exercises
$ ls ~/memos > filelist
$ ssh bravo ls ~/memos | diff filelist -
244
Draft
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 15 - Rebuilding the Linux Kernel
Rebuilding the Linux Kernel
IN THIS CHAPTER
Locating the Source Code
Installing the Source Code
Configuring and Compiling the Linux Kernel
Installing the Kernel and Associated Files
Boot Loader
dmesg: Display Kernel Messages
Once you have installed Red Hat Enterprise Linux or Fedora Core, you may want to reconfigure and rebuild the
Linux kernel. Red Hat comes with a prebuilt kernel that simplifies the installation process. This kernel may not be
properly configured for all your system’s features. By reconfiguring and rebuilding the kernel, you can build one
that is customized for your system and needs.
Tip: Maybe You Just Need to Install a New Linux Kernel Binary?
Refer to “Installing a Linux Kernel Binary” on page 458 when you want to install a Linux kernel binary
that you do not need to configure or build.
Because recent releases of the Linux kernel are modular, you do not usually need to rebuild them. You can dynamically change many things that used to require rebuilding the kernel. Two ways to make these changes are using
boot options or modifying /etc/sysctl.conf, which is used by sysctl when the system is booted.
grub . If you are using grub, append string to the kernel line in /boot/grub/grub.conf or to its symbolic link,
/etc/grub.conf. For example, norelocate prevents the substitution of CPU-specific optimizations and selinux=0
turns off SELinux.
lilo . The append kernel configuration parameter in /etc/lilo.conf appends a string to the parameter line that
is passed to the kernel. You can use this parameter to specify parameters so you do not have to rebuild the kernel,
parameters that cannot be detected automatically, and parameters you do not want to probe for. The following
example turns off the Advanced Power Management BIOS:
append="apm=off"
See the BootPrompt HOWTO for more information.
sysctl . The sysctl utility modifies kernel parameters while the system is running. This utility uses the facilities
of /proc/sys, which defines the parameters that sysctl can modify.
245
Draft
Preparing the Source Code
Draft
You can get a complete list of sysctl parameters with a sysctl –a command. An example of displaying and changing
the domainname kernel parameter follows. The quotation marks are not required in this example but will quote
any characters that would otherwise be interpreted by the shell.
# /sbin/sysctl kernel.domainname
kernel.domainname = tcorp.com
# /sbin/sysctl -w kernel.domainname="testing.com"
kernel.domainname = testing.com
Caution: Have the First Installation CD at Hand When You Rebuild
the Kernel
When you rebuild the Linux kernel to install a new version or to change the configuration of the existing version, make sure that you have the First Installation CD handy. This disk allows you to reboot
your computer, even when you have destroyed your system software completely. (You can also use
the rescue CD. See page 39.) Always follow these instructions; having this CD can make the difference
between momentary panic and a full-scale nervous breakdown. Refer to “Rescue Mode” on page 377
for instructions on bringing the system up in rescue mode.
Rebuilding a Linux kernel is fairly straightforward. When you do need to rebuild the kernel, perform the steps described in this chapter.
Preparing the Source Code
Before you can start, you must locate/install and clean the source code. If you want code that has not been customized
(patched) by Red Hat, go to kernel.org.
Locating the Source Code
When you have the kernel source on your system, the /usr/src directory will look something like the following:
$ ls -l /usr/src
total 8
drwxr-xr-x 19 root root 4096 Apr 28 21:49 linux-2.6.5-1.327
drwxr-xr-x
7 root root 4096 Apr 28 21:47 redhat
In the preceding example, the name linux-2.6.5-1.327 means that the directory contains version 2.6 of the Linux
kernel, release 5-1.327. If the source code is present on your system, skip to “Cleaning the Source Tree” on page
509.
Installing the Source Code
When the source is not present on your system, you need to install it. The easiest way to install the kernel source
is by giving the command up2date kernel-source. See page 462 for more information on up2date.
Alternatively, you can load the kernel source from the installation CD set or the Red Hat/Fedora Web site. You
need kernel-source*.rpm. The kernel used in the examples comes from kernel-source-linux-2.6.51.327.i386.rpm. You need to work as root to use rpm to install the kernel package. You can also use yum (page
469) or Apt (page 472) to download and install the kernel source.
246
Draft
Read the Documentation
Draft
Once you have located the kernel-source rpm package, give an rpm command to install it. The following command
installs the kernel-source rpm package from the directory that contains the package (refer to “ rpm: Red Hat
Package Manager” on page 455 for more information on using rpm):
$ rpm -ivh kernel-source*.rpm
Do not use the –U option as it overwrites the source for the existing kernel. You generally want to keep old kernel
source code around for a while after you install a new kernel in case you need to go back to the old kernel.
Tip: Now the Working Directory Is /usr/src/linux*
All commands in this section on building a kernel are given relative to the top-level directory that
holds the kernel source. Make sure that this directory is your working directory before proceeding.
The working directory for the examples in this chapter is /usr/src/linux-2.6.5-1.327.
Read the Documentation
The kernel package includes the latest documentation, some of which may not be available in other documents.
Review the README file and the relevant files in the Documentation directory. Read the Linux Kernel-HOWTO
for an excellent, detailed generic guide to installing and configuring the Linux kernel.
Configuring and Compiling the Linux Kernel
This section describes how to configure the kernel to meet your needs and how to compile it.
Cleaning the Source Tree
If you want to save the existing configuration file (/usr/src/linux*/.configure), copy it to another directory (such
as your home directory) before you proceed, because the following command removes it. Purge the source tree (all
the subdirectories and files within /usr/src/linux*) of all configuration and potentially stale *.o files by giving the
following command:
$ make mrproper
Configuring the Linux Kernel
Before you can compile the code and create the Linux kernel, you must decide and specify what features you want
the kernel to support. A kernel can support most features in two ways: by building the feature into the kernel or
by specifying the feature as a loadable kernel module (page 512), which is loaded into the kernel only as needed.
Trade off the size of the kernel against the time it takes to load a module. Make the kernel as small as possible while
minimizing how often modules have to be loaded. Do not make the SCSI driver modular unless you have reason
to do so.
The configs directory has sample configuration files for various processors, multiple processors, and configurations.
You may want to look at these before you get started or even use one of these as your starting point. To use one of
these files, copy it from the configs directory to the linux* directory and rename it .config.
The three standard commands to configure the Linux kernel are
247
Draft
Configuring the Linux Kernel
Draft
$ make config
$ LANG=C make menuconfig
$ make xconfig
FEDORA . The LANG=C is required before the make menuconfig command because the default encoding
in Fedora is UTF8, which the kernel’s build routines do not understand. The make xconfig command uses Qt
(www.trolltech.com [http://www.trolltech.com]), which is normally installed with KDE or GNOME under Fedora.
If you prefer to use GTK+ (www.gtk.org [http://www.gtk.org]) and it is installed on the system, give the command
make gconfig.
RHEL .
Under Red Hat Enterprise Linux, the make xconfig command uses Tk.
Each command asks the same questions and produces the same result, given the same responses. The first and
second commands work in character-based environments; the second and third commands work in graphical environments. For most administrators in most situations, the third (graphical) method is the easiest to use (Figures
15-1, and 15-2). The figures show the windows displayed by Fedora; the Red Hat Enterprise Linux windows look
different but perform the same function.
Figure 15-1. The qconf window as displayed by make xconfig on Fedora
248
Draft
Compiling the Linux Kernel
Draft
Figure 15-2. The qconf Processor Type submenu
The make xconfig command displays the qconf window. You can view the qconf window in three configurations:
single, split, or full view. Choose a view by clicking one of the three icons to the right of the floppy diskette on the
toolbar. Figure 15-1 shows the default split view. In the split view, the left frame shows the options and the topright view lists the features for each option. The bottom-right view describes the highlighted option or feature.
Figure 15-2 shows the full view.
In any view, click the boxes and circles next to the choices and subchoices: An empty box/circle indicates the feature
is disabled, a check mark indicates it is to be included in the kernel, and a dot means it is to be compiled as a
module. Select Menubar: Options Show All Options to display all options and features.
Go through the options and mark the features as you would like them to be configured in the new kernel. At any
time during the configuration process, you can store the currently defined configuration to a file, load a configuration
from a file, or exit with or without saving. See the selections in File on the Menubar. When you are done, select
Menubar: File Save and close the window.
EXTRAVERSION Number
To prevent overwriting existing kernel files and to identify various compilations of the kernel, you can use the
EXTRAVERSION variable in Makefile. This variable is initially set to a dash and the release number, followed
by the word custom (for example –2.6.5-1.327custom). Whatever value you assign to this variable is placed at
the end of the kernel name. It is a good idea to keep the original release number, but you can append the date or
time to ensure a unique name for each kernel you build. You can also make note of patches you applied to the
kernel in this string to help people track down problems later on.
Compiling the Linux Kernel
Before compiling the kernel, make sure, once again, that no files are in the source tree from previous work:
$ make clean
Then give the following command to compile the kernel:
249
Draft
Using Loadable Kernel Modules
Draft
$ make bzImage
CHK
include/linux/version.h
UPD
include/linux/version.h
SYMLINK include/asm -> include/asm-i386
SPLIT
include/linux/autoconf.h -> include/config/*
HOSTCC scripts/basic/fixdep
HOSTCC scripts/basic/split-include
HOSTCC scripts/basic/docproc
...
Using Loadable Kernel Modules
A loadable kernel module (page 980) (sometimes called a module or loadable module) is an object file, part of the
kernel, that is linked into the kernel at runtime. Modules are compiled separately from the kernel and can be inserted
into and removed from a running kernel at almost any time except when the module is being used. This ability
gives the kernel the flexibility to be as small as possible at any given time. Modules are a good way to code some
kernel features, including drivers that are not used all the time (such as a tape driver).
Tip: Module Filename Extensions Have Changed
Filenames of modules in the 2.4 and earlier kernels ended in .o. Starting with the 2.6 kernel (Fedora
Core 2 and greater), module filenames end in .ko.
If you are not using loadable kernel modules, skip to “Installing the Kernel and Associated Files” on page 513. If
you do not know for a fact that you are not using loadable modules, you probably are, so continue with this section.
When you configure the kernel to support loadable modules, you need to build and install the modules. Give the
following command to compile the modules that you specified when you configured the kernel:
$ make modules
The next command installs the modules in the /lib/modules/ kernel-versionEXTRAVERSION directory. Run this
command as root even when you did not build any modules:
# make modules_install
Table 15-1 lists some of the tools that can help you work with modules. Refer to the corresponding man pages for
options and more information.
Table 15-1. Tools for Working with Modules
depmod
Works with dependencies for modules.
insmod
Loads modules in a running kernel.
lsmod
Lists information about all loaded modules.
modinfo
Lists information about a module.
modprobe
Loads, unloads, and reports on modules. When it loads
a module, it also loads dependencies.
rmmod
Unloads modules from a running kernel.
250
Draft
Installing the Kernel and Associated
Files
Draft
Installing the Kernel and Associated Files
Next, copy the compiled kernel and associated files to the appropriate directory, usually either root (/) or /boot.
When you have a boot partition, the files are kept in the root of this partition (/boot). Without a boot partition,
the files are kept in the root directory. Run the following command as root to install the new kernel files in the
proper directory:
# make install
Changing lilo.conf (RHEL)
If you are using lilo (Fedora Core does not support lilo), edit /etc/lilo.conf, and add a section giving the absolute
pathname of the new kernel image and, when you use modules, the pathname of the initrd file. Run lilo to make
your changes to lilo.conf take effect. Refer to “ lilo: The Linux Loader (RHEL)” on page 516 for details.
grub.conf
When using grub, you do not have to make any changes to /etc/grub.conf.
Rebooting
Reboot the computer by logging out and selecting Reboot from the login screen (Figure 4-1, page 77). If you are
working at the console, you can press CONTROL-ALT-DEL or give a reboot command.
Boot Loader
A boot loader is a very small program that takes its place in the bootstrap (page 960) process, which brings a computer
from off or reset to a fully functional state. It frequently resides on the starting sectors of a hard disk called the MBR
(Master Boot Record).
The BIOS (page 959), stored in an EEPROM (page 969) on the system’s motherboard, gains control of a system
when you turn it on or reset it. After testing the hardware, the BIOS transfers control to the MBR, which usually
passes control to the partition boot record. This transfer of control starts the boot loader, which is responsible for
locating the operating system kernel (kept in the / or /boot directory), loading that kernel into memory and starting
it running. Refer to “Booting the System” on page 381 for more information on what happens from this point forward.
You can place the /boot directory on a very small filesystem that is near the beginning of the hard drive where the
BIOS can access it. With this setup, the root (/) file-system can be anywhere on any hard drive that Linux can access
and that perhaps the BIOS cannot.
grub: The Linux Loader
The term grub (see the grub info page and www.gnu.org/software/grub [http://www.gnu.org/software/grub]) stands
for Grand Unified Boot Loader. The grub loader is a product of the GNU project and conforms to the multiboot
specification (page 984), which allows it to load many free operating systems directly, as well as chain loading (page
962) proprietary operating systems. In many ways, grub is more flexible than lilo. The grub loader can recognize
various types of filesystems and kernel executable formats, allowing it to load an arbitrary operating system: You
must specify the kernel’s filename and location (drive and partition). You can pass this information to grub by using
either the command line or the menu interface. When you boot the system, grub displays a menu of choices that
251
Draft
grub: The Linux Loader
Draft
is generated by the /boot/grub/grub.conf file (or see its symbolic link, /etc/grub.conf). At this point you can
modify the menu, choose which operating system to boot, or do nothing and allow grub to boot the default system.
When you install a Red Hat Enterprise Linux, you have the choice of using grub or lilo. Fedora runs grub only.
When you install grub at the time you install Linux on a system, the installation program configures grub, and you
do not have to.
The /boot/grub/grub.conf file is the default grub configuration file and is similar in function to /etc/lilo.conf.
The grub.conf file following is from a system that had its kernel replaced (there are two versions of vmlinuz and
initrd). The system has a separate boot partition so that all kernel and initrd (for systems using loadable modules,
page 512) image paths are relative to /boot/ (see the NOTICE in the file). Without a separate boot partition, the
boot files reside in the root partition (/) so that kernel and initrd paths are relative to /. (Thus you would specify
the kernel as kernel /vmlinuz- version, replacing version with the version number of the kernel.)
The file starts with comments that Anaconda, the graphical installer, puts there, followed by three assignments.
The default is the section number of the default boot specification. The numbering starts with 0. The following
example includes two boot specifications. The first, numbered 0, is for the 2.6.5-1.327custom kernel, and the
second, numbered 1, is for the 2.6.5-1.327 kernel. The timeout is the number of seconds that grub waits after it
has prompted you for a boot specification before it boots the system with the default boot specification. The
splashimage is the grub menu interface background that you see when you boot the system.
$ cat /etc/grub.conf
# grub.conf generated by anaconda
#
# Note that you do not have to rerun grub after making changes to this file
# NOTICE: You have a /boot partition. This means that
#
all kernel and initrd paths are relative to /boot/, eg.
#
root (hd0,0)
#
kernel /vmlinuz-version ro root=/dev/hda10
#
initrd /initrd-version.img
default=1
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title Fedora Core (2.6.5-1.327custom)
root (hd0,0)
kernel /vmlinuz-2.6.5-1.327custom ro root=LABEL=/ rhgb quiet
initrd /initrd-2.6.5-1.327custom.img
title Fedora Core (2.6.5-1.327)
root (hd0,0)
kernel /vmlinuz-2.6.5-1.327 ro root=LABEL=/ rhgb quiet
initrd /initrd-2.6.5-1.327.img
Following the splashimage assignment in the preceding example are two boot specifications, differentiated by the
title lines as explained previously. The three lines following the title line in each specification specify the location
of the root (drive 0, partition 0), kernel, and initrd images. In this case, because there is a /boot partition, the
pathnames are relative to /boot. For the default boot specification (the second one, numbered 1), the absolute
pathname of the kernel is /boot/vmlinuz-2.6.5-1.327custom, which is specified with the options ro root=LABEL=/
rhgb quiet.
These options tell grub that it is to be mounted readonly and that root (/) is mounted on the device labeled / in
/etc/fstab (page 445). The rhgb (Red Hat graphical boot) is the software that generates a graphical display that
tells you what is going on as the system boots. The quiet option produces less debugging output so it is easier to
tell what is happening. You specify the initrd (initialize RAM disk, page 991) image in a manner similar to the
252
Draft
lilo: The Linux Loader (RHEL)
Draft
kernel. Substitute your kernel and initrd names and version numbers for the ones in the example. Make sure that
when you install a new kernel manually, its title line is different from the others present in grub.conf.
lilo: The Linux Loader (RHEL)
The term lilo (Linux loader) applies to both the Linux loader utility and the boot loader that lilo writes to the beginning of the active partition on the hard drive or to a floppy. Although written to support Linux, lilo is a generalpurpose boot loader that can start many operating systems, including DOS, OS/2, Windows, and versions of BSD.
You can configure lilo to select from various operating systems and versions of the Linux kernel each time you start
your system (called dual booting).
When you run lilo without any options, it reads /etc/lilo.conf to determine which operating systems are to be
made available at boot time. Then it writes this information to the MBR. You must run lilo to reinstall the boot
loader whenever you change /etc/lilo.conf, as when you rebuild the Linux kernel.
The /etc/lilo.conf file gives you a great deal of control over lilo:
$ cat /etc/lilo.conf
# lilo configuration file
boot=/dev/hda
map=/boot/map
install=/boot/boot.b
prompt
timeout=50
default=linux
image=/boot/vmlinuz-2.4.18-14custom
label=linux
initrd=/boot/initrd-2.4.18-14custom.img
read-only
root=/dev/hda12
Comments in the lilo.conf file start with a pound sign (#) and run through the end of the line, just as they do in
a shell script. In the example, the first line that is not a comment identifies the disk that holds the Master Boot
Record: /dev/hda. The next line gives the location of the map file. The install line tells lilo where to find the file
it is to install as the boot loader. The prompt presents the user with a boot: prompt, whereas timeout gives the
time (in tenths of a second) that the system will wait before booting automatically. When you set prompt and do
not set timeout, the system cannot boot automatically. Press TAB in response to the boot: prompt to see the
choices.
Security: Protect lilo.conf Too
When you protect the images specified in lilo.conf with passwords, also change permissions on
lilo.conf to 600, as anyone can read it with its default permissions.
The lilo boot loader allows you to specify more than one boot image. The default line specifies the label of the
image to boot if none is specified, such as when you press RETURN in response to the boot: prompt. In this example,
the label of the default (and only) image is linux (see the label line). The image line specifies the boot image of
the Linux kernel. The initrd line initializes the boot loader RAM disk (page 991), which is used in the first phase
of the system boot when you use loadable modules (page 512). Specify the read-only line for every image on a
Linux system so it is safe to run fsck automatically as you bring the system up.
253
Draft
LOADLIN: A DOS-Based Linux
Loader
Draft
You can add lines to lilo.conf to pass arguments to the Linux kernel, adjust for strange disk geometries, require
passwords to start specific kernels, and add many other options. See the lilo.conf man page for the technical details
and the /usr/share/doc/lilo*/doc directory for the User and Technical Guides. View the *.ps files with gs (ghostscript).
Before you modify an existing lilo.conf file, save the existing configuration as lilo.conf.old or lilo.conf.1. Then
edit lilo.conf and create a new section that looks similar to the preceding example. Use the kernel version number
and EXTRAVERSION (page 511) in place of 2.4.18-14custom.
If you use modules, set up the appropriate initrd line in lilo.conf:
initrd=/boot/initrd-2.4.18-14custom.img
When you are finished editing lilo.conf, run lilo to write the new information to the MBR.
LOADLIN: A DOS-Based Linux Loader
The LOADLIN loader, a DOS utility that loads Linux from DOS and some versions of Windows, can load big
kernels (bzImage) and RAM disk images (initrd). Refer to elserv.ffm.fgan.de/~lermen, where you can find the
LOADLIN Users Guide and other information. See also the Loadlin+Win95/98/ME mini-HOWTO.
dmesg: Display Kernel Messages
The dmesg utility displays the kernel-ring buffer, where the kernel stores messages. When the system boots, the
kernel fills this buffer with messages regarding hardware and module initialization. Messages in the kernel-ring
buffer are often useful for diagnosing system problems. Run dmesg to view the messages:
$ dmesg
...
VFS: Disk quotas vdquot_6.5.1
Detected PS/2 Mouse Port.
pty: 2048 Unix98 ptys configured
Serial driver version 5.05c (2001-07-08) with MANY_PORTS MULTIPORT
SHARE_IRQ SER
IAL_PCI ISAPNP enabled
ttyS0 at 0x03f8 (irq = 4) is a 16550A
ttyS1 at 0x02f8 (irq = 3) is a 16550A
Real Time Clock Driver v1.10e
NET4: Frame Diverter 0.46
RAMDISK driver initialized: 16 RAM disks of 8192K size 1024 blocksize
...
The dmesg utility, which is frequently used with grep, is useful if you are having hardware-related problems. If you
find that your hard disks are performing poorly, you can use dmesg to check that they are running in DMA mode:
$ dmesg | grep DMA
...
ide0: BM-DMA at 0xa400-0xa407, BIOS settings: hda:pio, hdb:DMA
ide1: BM-DMA at 0xa408-0xa40f, BIOS settings: hdc:DMA, hdd:DMA
...
254
Draft
Chapter Summary
Draft
The preceding lines tell you which mode each IDE device is operating in.
If you are having problems with the Ethernet connection, search the dmesg log for eth:
$ dmesg | grep eth
eth0: RealTek RTL8139 Fast Ethernet at 0xe4d05000, 00:d0:70:01:a0:64, IRQ 5
eth0: Identified 8139 chip type 'RTL-8139C'
eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
If everything is working properly, dmesg displays the hardware configuration information for each network card.
If you have configured a system service incorrectly, the dmesg log quickly fills up with errors; it is a good place to
start when diagnosing faults.
Chapter Summary
This chapter describes how to rebuild the Linux kernel. Sometimes you do not need to rebuild the kernel; you can
dynamically change many things by using boot options in /etc/lilo.conf or /etc/grub.conf, or by modifying
/etc/sysctl.conf.
Before you can rebuild the kernel, you must have the kernel source files on the system. These files are located in
/usr/src/linux*. The chapter discusses the steps involved in compiling and installing both the kernel and loadable
modules. It concludes with a discussion of the dmesg utility, which displays the kernel-ring buffer, where the kernel
stores messages.
Exercises
1. What is the purpose of the kernel?
2. How would you display a list of all loaded modules in the current kernel?
3. What command would you give to upgrade the kernel from an rpm file, and how is this different from upgrading
other packages?
4. How would you display information from the kernel about the hard disk on the first IDE channel?
5. The noreplacement kernel argument tells the kernel not to use CPU-specific sections of code. How would you
use this argument?
6. What is a boot loader?
Advanced Exercises
7. What is the EXTRAVERSION variable? Where is it used and what is it used for?
8. You have just installed an Adaptec SCSI card. How can you find out if it has been recognized and which entry
in /dev represents it?
9. When you install an experimental kernel for testing, how do you instruct grub not to load it by default?
10. How would you obtain a list of all network-related kernel parameters?
255
256
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 14 - Printing with CUPS
Printing with CUPS
IN THIS CHAPTER
JumpStart I: Configuring a Local Printer Using systemconfig-printer
JumpStart II: Configuring a Remote Printer Using CUPS
Traditional UNIX Printing
The CUPS Web Interface
CUPS on the Command Line
Integration with Windows
A printing system handles the tasks involved in getting a print job from an application (or the command line) through
the appropriate filters (page 971) into a queue for a suitable printer and getting it printed. While handling a job, a
printing system can keep track of billing information so that the proper accounts can be charged for printer use.
When a printer fails, the printing system can redirect jobs bound for that printer to other, similar printers.
Introduction
LPD and LPR . Traditionally, UNIX had two printing systems: the BSD Line Printer Daemon (LPD) and the
System V Line Printer system (LPR). Linux adopted those systems at first, and both UNIX and Linux have seen
modifications to and replacements for these systems. Today, CUPS is the default printing system under Red Hat
Linux.
CUPS . CUPS (Common UNIX Printing System) is a cross-platform print server built around IPP (Internet
Printing Protocol), which is based on HTTP. CUPS provides a number of printer drivers and can print different
types of files including PostScript. Because it is built on IPP and written to be portable, CUPS runs under many
operating systems, including Linux and Windows. Other UNIX variants, including MacOS X, use CUPS, and recent
versions of Windows include the ability to print to IPP printers, making CUPS an ideal solution for printing in a
heterogeneous environment. CUPS provides System V and BSD command line interfaces and, in addition to IPP,
supports LPD/LPR, HTTP, SMB, and JetDirect (socket) protocols, among others.
IPP . The IPP project (www.ietf.org/html.charters/ipp-charter.html
[http://www.ietf.org/html.charters/ipp-charter.html]) was started in 1996 when Novell and several other companies
decided to design a protocol for printing over the Internet. The IPP enables users to
•
Determine the capabilities of a printer
•
Submit jobs to a printer
•
Determine the status of a printer
257
Draft
Prerequisites
•
Determine the status of a print job
•
Cancel a print job
Draft
IPP is a client/server protocol in which the server side can be a print server or a network capable stand-alone
printer.
Printers and queues . On a modern computing system, when you “send a job to the printer,” you actually add
the job to the list of jobs waiting their turn to be printed on a printer. The list is called a print queue or simply a
queue. The phrase configuring (or setting up) a printer is often used to mean configuring a (print) queue. This chapter
uses the phrases interchangeably.
Prerequisites
Install the following packages:
•
cups
•
system-config-printer (FEDORA) or redhat-config-printer (RHEL) (optional)
•
kdebase (optional, provides kprinter)
•
printman (optional, provides gnome-print-manager)
Run chkconfig to cause CUPS (the cupsd daemon) to start when the system goes multiuser:
# /sbin/chkconfig cups on
Start CUPS:
# /etc/rc.d/init.d/cups start
To use the Web interface to CUPS, you need an X server and a Web browser.
More Information
Local . CUPS Documentation With the CUPS Web interface up (page 493), point a local browser at localhost:631/documentation.html.
Web . www.linuxprinting.org [http://www.linuxprinting.org] information on printers and printing under Linux;
hosts a support database with details about many printers, including notes and driver information, also forums,
articles, and a HOWTO on printing
CUPS home page .
www.cups.org [http://www.cups.org]
IPP information . www.pwg.org/ipp [http://www.pwg.org/ipp]
www.ietf.org/html.charters/ipp-charter.html [http://www.ietf.org/html.charters/ipp-charter.html]
HOWTO .
258
SMB HOWTO has a section named “Sharing a Windows Printer with Linux Machines.”
Draft
JumpStart I: Configuring a Local
Printer Using system-config-printer
Draft
JumpStart I: Configuring a Local Printer Using
system-config-printer
This JumpStart configures a printer that is connected directly to the local system. The fastest way to add a new
printer is to use system-config-printer (FEDORA) or redhat-config-printer (RHEL), a simple wrapper script that
launches either the textual or the graphical version of printconf-gui, depending on whether it detects an X server.
You can run [system|redhat]-config-printer from a command line or from the Gnome or K (Red Hat) menus: System
Setting->Printing.
From the [system|redhat]-config-printer Printer configuration window (Figure 14-1), click the New button to
display the Add a new print queue wizard. Click Forward to display Queue name window (Figure 14-2). Enter
a name for, and a short description of, the printer. The name is a short identifier that starts with a letter and does
not contain any SPACEs. The description can be a short sentence. Click Forward.
Figure 14-1. The main Printer configuration window
259
Draft
JumpStart I: Configuring a Local
Printer Using system-config-printer
Draft
Figure 14-2. The Queue name window
The next window, Queue type, asks you to specify the printer connection (Figure 14-3). By default, the Locallyconnected queue type is selected in the combo box at the top of the window.
Figure 14-3. The Queue type window
Most printers connect to a parallel or USB port, although older printers may connect to a serial port. The default
list in the box in the middle of the window lists the system’s parallel ports. Under Linux, a parallel port is identified
as /dev/lp n, where n is the number that identifies the port. The first parallel port (LPT1 under Windows) is
/dev/lp0. Unless you have several parallel ports on the local system, the printer is probably connected to /dev/lp0.
260
Draft
JumpStart I: Configuring a Local
Printer Using system-config-printer
Draft
USB devices appear in the /dev/usb directory. USB printers appear as standard parallel printers within the usb
directory hierarchy. The first USB printer port is /dev/usb/lp0, with subsequent printers appearing as lp1, lp2,
and so on, exactly as parallel ports do. The first serial port (COM1 under DOS or Windows) is /dev/tty0.
If your device is not listed, click Custom device and enter the pathname of the device as /dev/ xxx .
Click to highlight the line that names the port the printer you are installing is connected to and click Forward.
The wizard displays the Printer model window (Figure 14-4). Specifying the printer model is a two-step process:
First, click the bar that has the words Generic (click to select manufacturer) on it; the wizard displays a long
pop-up menu of manufacturers. Move the mouse pointer over the menu until the manufacturer of your printer is
highlighted; then click. The wizard replaces the words in the bar with the name of the manufacturer and displays
that manufacturer’s known printer models in the box below the bar. Scroll through the models and click the
printer that is attached to the system. Click Notes to display notes on the selected printer from the Linux Printing
Database.
Figure 14-4. The Printer model window
If your printer is not listed, check to see if it can emulate another printer (if it has an emulation mode). If it can,
check to see if the printer it can emulate is listed and set it up that way. If all else fails, reselect Generic in the bar
and choose a generic printer from the list in the box. Choose PostScript Printer if your printer is PostScript capable.
If there is no match, select Text Only Printer; you will not be able to print graphics, but you should be able to
print text.
FEDORA .
If you are using a winprinter, select GCI Printer from the list of Generic printers.
Click Forward to display the Finish, and create the new print queue window. The wizard has not saved any
information at this point; you can click Cancel to abort the process or Apply to create the new print queue.
Next, the wizard asks you if you want to print a test page; do so to ensure that the configuration was successful.
Printing the test page automatically commits the configuration changes. If you do not print a test page, select Apply
from the Action drop down menu to commit the changes.
If you have more than one print queue and want to make the new print queue the default one, highlight the print
queue and select Set as default from the Action dropdown menu.
261
Draft
JumpStart II: Configuring a Remote
Printer Using CUPS
Draft
JumpStart II: Configuring a Remote Printer Using CUPS
This JumpStart uses the Web interface to CUPS to configure either a printer that is connected to a different
UNIX/Linux system that provides IPP support or an LPD/LPR queue, or a printer connected directly to the network.
If the printer you are configuring is on an older Linux system or another UNIX-like operating system that does not
run CUPS, the system is probably running LPD/LPR. Newer versions of Linux and UNIX variants that support
CUPS (including Mac OS X) support IPP. Most devices that connect printers directly to a network support LPR/LPD
and may support IPP.
Printers connected directly to a network are functionally equivalent to printers connected to a system running a
print server: They listen on the same ports as systems running print servers and queue jobs.
Connect to the Web interface to CUPS by pointing a Web browser at localhost:631 on the system you are configuring the printer on (Figure 14-5).
Figure 14-5. CUPS Web interface: main page
Click Printers on the navigation bar at the top of the page; then click Add Printer to display the Admin->Add
New Printer page (Figure 14-6). If you are prompted for a username and password, enter root and the root
password. The Name field holds the system name for the printer; it must start with a letter and not contain any
SPACEs. Fill in the Location and Description fields with text that will help users identify the printer and click
Continue.
262
Draft
JumpStart II: Configuring a Remote
Printer Using CUPS
Draft
Figure 14-6. CUPS Web interface: Admin->Add New Printer
The next page asks you to select the device that the printer is attached to. Click the down arrow at the right of the
Device combo box to display the list of printer devices. Select Internet Printing Protocol (ipp) or an LPD/LPR
Host or Printer, depending on what the printer is attached to. Select AppSocket/HP JetDirect for an HP JetDirect
compatible network printer. Click Continue.
The next page asks for the URI (location on the network) of the printer. For an LPD printer, use the form lpd://
hostname / queue ; for an IPP printer, use ipp:// hostname /ipp; for an HP JetDirect compatible network printer,
use socket:// hostname. Replace hostname with the name of the host that the printer is attached to or the name of
printer for a network printer. You can specify an IP address in place of hostname. Replace queue with the name of
the queue on the remote system. Enter the URI of the printer and click Continue.
Next is the first of two Model/Driver pages (Figure 14-7). Highlight the brand of printer and click Continue. If
the printer is PostScript capable and it is not listed, select Postscript. If the printer is not PostScript capable and
is not listed, check to see if the printer supports PCL; if it does, select another, similar PCL printer. If all else fails,
determine which listed printer is most similar to the one you are configuring and specify that printer. You can also
try configuring the printer using [system|redhat]-config-printer (page 487) which offers a different choice of models.
263
Draft
Traditional UNIX Printing
Draft
Figure 14-7. CUPS Web interface: Admin->Model/Driver page
The final page in this process is the second Model/Driver page. Select the model of the printer from the scrollable
list and click Continue.
If the printer was configured properly, when you select a printer and click Continue, you will get a message saying
that the printer was added successfully. Click the name of the printer on this page or click Printers at the top of
the page to display the Printer page (Figure 14-8). Once you have set up the printer, it is a good idea to print a test
page.
Figure 14-8. CUPS Web interface: Printer page
Traditional UNIX Printing
Before the advent of GUIs and wysiwyg (page 1006) word processors, UNIX users would create documents using
an editor such as vi and a typesetting markup language such as TeX or nroff/troff, convert the resulting files to
PostScript using an interpreter, and send the PostScript files to the printer using lp (System V) or lpr (BSD). Red
Hat Linux implements both BSD and System V command line printing utilities for compatibility: These utilities
are now wrappers around the equivalent functionality in CUPS rather than core components of the printing system.
The corresponding utilities are functionally equivalent; use whichever you prefer (Table 14-1).
264
Draft
Configuring Printers Using CUPS
Draft
Table 14-1. BSD and System V Command Line Utilities
BSD/SysV
Purpose
lpr/lp
Sends job(s) to the printer.
lpq/lpstat
Displays the status of the print queue.
lprm/cancel
Removes job(s) from the print queue.
From the command line, you can print a text or PostScript file using lp:
$ lp memo.txt
request id is MainPrinter-25 (1 file(s))
The preceding command adds memo.txt to the print queue of the default printer as job 25. When this printer is
available, it prints the file.
You can specify a printer using the –d option:
$ lp -d colorprinter graph.ps
request id is colorprinter-26 (1 file(s))
The –P option to lpr is equivalent to the –d option to lp.
Without an argument, lp (and lpr) sends its standard input to the printer:
$ cat memo2.txt | lp
request id is MainPrinter-27 (1 file(s))
The lpq and lpstat commands display information about the print queue:
$ lpstat
MainPrinter-25
ColorPrinter-26
MainPrinter-27
mark
mark
mark
13312
75776
8192
Sun Feb 22 18:28:38 2004
Sun Feb 22 18:28:48 2004
Sun Feb 22 18:28:57 2004
Use cancel or lprm to remove jobs from the print queue. Only the owner of a print job or root can remove a job.
$ cancel 27
$ lpstat
MainPrinter-25
ColorPrinter-26
mark
mark
13312
75776
Sun Feb 22 18:28:38 2004
Sun Feb 22 18:28:48 2004
Use cancel –a or lprm – to remove all your jobs from the print queue.
Configuring Printers Using CUPS
You can use the Web interface or the command line interface to CUPS to manage printers and queues.
265
Draft
The CUPS Web Interface
Draft
The CUPS Web Interface
CUPS, designed for Internet printing, provides a Web interface to configure printers. To connect to this interface,
point a Web browser running on the local system at localhost:631.
Setting Up and Modifying a Printer
“JumpStart II: Configuring a Remote Printer Using CUPS” (page 489) discusses how to set up a remote printer
using CUPS. The procedure for setting up a local printer is similar. The major difference is the second step: specifying
the device that the printer is connected to.
A local printer is generally connected to USB Port #1 or Parallel Port #1. After specifying one of these devices,
the Web interface displays the page on which you specify the brand of the printer; you do not specify a URI for a
local printer. If you are setting up a serial printer, you will need to specify characteristics of the printer, including
its baud rate. After these steps, the procedure is the same as explained in JumpStart II.
To modify a printer, click Printers from the Web interface navigation bar and click the Modify Printer button
adjacent to the printer you want to modify. The Web interface takes you through the same steps as setting up a
new printer.
Click the red Stop Printer button to pause the printer. Click the Reject Jobs button to prevent jobs from being
added to the printer’s queue.
Jobs
The Jobs page (Figure 14-9) lists jobs in the print queues. From this page you can hold (pause), release (unpause),
and cancel print jobs. Click Show Complete Jobs to display a list of recently completed jobs; in some cases, you
can restart completed jobs from this page.
Figure 14-9. CUPS Web interface: Jobs page
Classes
CUPS allows you to group similar printers; this group is called a class. A print job sent to a class will be printed on
the first available printer in the class. For example, you may be able to divide your print jobs into black and white
and color. If you have more than one printer that can fulfil each role, you can allow users to select a printer
manually, or you can define two printer classes (black and white and color) and have users send jobs to a class of
printers.
266
Draft
CUPS on the Command Line
Draft
TIP: Plan for the Future
If you expect to add printers to the network, you may want to configure classes containing the existing
printers when you set up the network. You can then add printers later without having to change
printer configuration on client systems.
Adding printers to a class is a two-step process. First, you need to define a class. Second, you need to add existing
printers to the class. To define a class, click Classes from the navigation bar at the top of the page and then click
Add Class. To clients, a class of printers appears as a single printer; for each class, you need to specify a name,
location, and description. Once you have defined a class, you can add printers to the class. Repeat this process for
the classes you want to define. A printer can belong to more than one class.
CUPS on the Command Line
In addition to using the Web interface, you can control CUPS and manage print queues from the command line.
This section details utilities that enable you to manage printers and print queues and establish printing quotas.
lpinfo: Display Available Drivers
The lpinfo utility provides information about the printer drivers and interfaces available to CUPS. The –m option
displays the list of available PostScript Printer Definition (PPD) files/drivers.
PPD files .
$ /usr/sbin/lpinfo -m
raw Raw Queue
dymo.ppd.gz DYMO Label Printer CUPS v1.1
epson9.ppd.gz EPSON 9-Pin Series CUPS v1.1
epson24.ppd.gz EPSON 24-Pin Series CUPS v1.1
stcolor2.ppd.gz EPSON New Stylus Color Series CUPS v1.1
stphoto2.ppd.gz EPSON New Stylus Photo Series CUPS v1.1
stcolor.ppd.gz EPSON Stylus Color Series CUPS v1.1
stphoto.ppd.gz EPSON Stylus Photo Series CUPS v1.1
deskjet.ppd.gz HP DeskJet Series CUPS v1.1
laserjet.ppd.gz HP LaserJet Series CUPS v1.1
deskjet2.ppd.gz HP New DeskJet Series CUPS v1.1
okidata9.ppd.gz OKIDATA 9-Pin Series CUPS v1.1
okidat24.ppd.gz OKIDATA 24-Pin Series CUPS v1.1
...
postscript.ppd.gz Generic postscript printer
CUPS uses URIs (page 1003) to identify printer ports by location and type, just as a Web browser identifies documents
by location and protocol. A parallel port has a URI with the format: parallel:/dev/lp 0 ; a remote LPD printer uses
this format: lpd:// 192.168.0.101. With the –v option, lpinfo provides a list of available connections.
$ lpinfo -v
network socket
network http
network ipp
network lpd
direct parallel:/dev/lp0
direct scsi serial serial:/dev/ttyS0?baud=115200
267
Draft
CUPS on the Command Line
Draft
...
serial serial:/dev/ttyS31?baud=115200
direct usb:/dev/usb/lp0
...
direct usb:/dev/usb/lp15
network smb
The the –v option to lpinfo does not display every possible network address for the socket, HTTP, IPP, LPD, and
SMB protocols; there are more than four billion of these addresses in the IPv4 address space.
lpadmin: Configure Printers
The lpadmin utility can add and remove printers from the system, modify printer configurations, and manage
printer classes. It has three major options: –d (set the default printer), –x (remove a printer), and –p (add or
modify a printer). The first two options are simple; examples are shown after the next section. Each of the options
takes an argument of the name of a printer. The name of the printer must start with a letter and cannot contain
SPACEs.
Adding or Modifying a Printer
Add a printer or modify an existing printer by giving the following command:
# /usr/sbin/lpadmin -p printer options
where printer is the name of the printer and options is a combination of options from the following list:
–c class .
Adds the printer to the class class, creating the class if necessary.
–D info . The info is a string that describes the printer for users. This string has no meaning to the system. Enclose
info within quotation marks if it contains SPACEs.
–E .
enable .
Enables the printer and instructs it to accept jobs.
–L loc . The loc is a string that physically locates the printer for users (office, building, floor, and so on). This
string has no meaning to the system. Enclose loc within quotation marks if it contains SPACEs.
–m model . The model is the name of the PPD file (page 495) that describes the printer. Use lpinfo –m to display
a list of all of the installed PPD files. If you have a manufacturer-provided PPD file, copy it to
/usr/share/cups/model. Use the –P option to specify the pathname of the file. Specifying –m postscript.ppd.gz
is the same as specifying –P /usr/share/cups/model/postscript.ppd.gz.
–P file . The file is the absolute pathname of the PPD file (page 495) that holds the printer driver. See –m for
an alternative way to specify a PPD file.
–r class . Removes the printer from the class class . This option removes the class if after removing the printer
the class would be left empty.
–v URI .
268
The URI is the device to which the printer is attached. Use lpinfo –v to list possible devices.
Draft
CUPS on the Command Line
Draft
Example lpadmin commands
At a minimum, you need to provide a device and a model when you add a printer to the system. The following
command adds an Epson Stylus Color printer to the system and enables it for use. The printer is connected locally
to the first parallel port and is named ColorPrinter.
# lpadmin -p ColorPrinter -E -v parallel:/dev/lp0 -m stcolor.ppd.gz
The printer information generated by the preceding command is stored in the /etc/cups/printers.conf file.
# cat /etc/cups/printers.conf
# Printer configuration file for CUPS v1.1.17
# Written by cupsd on Mon 23 Feb 2004 03:08:58 AM GMT
<Printer ColorPrinter>
Info ColorPrinter
DeviceURI parallel:/dev/lp0
State Idle
Accepting Yes
JobSheets none none
QuotaPeriod 0
PageLimit 0
KLimit 0
</Printer>
The printer driver information from /usr/share/cups/model/stcolor.ppd.gz is uncompressed and copied to
/etc/cups/ppd. The resulting file is given the printer’s name: /etc/cups/ppd/ColorPrinter.ppd.
You can modify a printer configuration with lpadmin using the same options that you used to add it: When you
specify the name of a printer that exists, lpadmin modifies the printer rather than creating a new one.
The next command configures an HP LaserJet compatible printer with a JetDirect interface that is connected directly
to the LAN at 192.168.1.103 and names it HPLJ. Specifying socket in the protocol part of the URI instructs CUPS
to use the JetDirect protocol, a proprietary protocol developed by HP for printers connected directly to a network.
# lpadmin -p HPLJ -E -v socket://192.168.1.103 -m laserjet.ppd.gz
The lpstat utility with the –d option displays the name of the default printer:
$ lpstat -d
system default destination: MainPrinter
CUPS automatically makes the first printer you defined the default printer. The following command makes HPLJ
the default printer:
# lpadmin -d HPLJ
The following command removes the configuration for the printer named Color-Printer:
# lpadmin -x ColorPrinter
269
Draft
CUPS on the Command Line
Draft
Printing Quotas
CUPS provides rudimentary printing quotas. You can define two forms of quotas: page count and file size. File size
quotas are almost meaningless because a small PostScript file can take a long time to interpret and can use a lot
more ink than a large one. Page quotas are more useful, although their implementation is flawed. To determine
the number of pages in a document, CUPS examines the PostScript input. If a job is submitted in the printer’s
native language, such as PCL, CUPS bypasses this accounting mechanism. Also, if mpage is used to create a PostScript
file with multiple pages printed on each sheet, CUPS counts each page in the original document, rather than each
sheet of paper it prints on.
Use the job-quota-period and either job-page-limit or job-k-limit to establish a quota for each user on a given
printer. The job-quota-period option specifies the number of seconds that the quota is valid. The following
command establishes a quota of 20 pages per day per user for the printer named HPLJ:
$ lpadmin -p HPLJ -o job-quota-period=86400 -o job-page-limit=20
The job-k-limit option works similarly but defines a file size limit in kilobytes. The limit is the total number of
kilobytes that each user can print over the quota period. Once a user has exceeded her quota, she will not be allowed
to print until the next quota period.
Managing Print Queues
When a printer is operating normally, it accepts jobs into its print queue and prints them in the order received.
Two factors determine how a printer handles a job: if the printer is accepting jobs and if it is enabled. Table 14-2
shows what happens in each of the four combinations of the two factors.
Table 14-2. Printer Status
Accepting Jobs
Rejecting Jobs
Enabled
Disabled
Accepts new jobs into the queue.
Accepts new jobs into the queue.
Prints jobs from the queue.
Does not print jobs from the queue
until the printer is enabled.
Rejects new jobs.
Rejects new jobs.
Prints jobs from the queue.
Does not print jobs from the queue
until the printer is enabled.
The utilities that change these factors are disable, enable, reject, and accept. Each of these utilities takes the name
of a printer as an argument. The following commands disable and then enable the printer named HPLJ:
# /usr/bin/disable HPLJ
# /usr/bin/enable HPLJ
The next commands cause HPLJ to reject and then accept jobs:
# /usr/sbin/reject HPLJ
# /usr/sbin/accept HPLJ
The enable and disable utilities are located in /usr/bin, while reject and accept are located in /usr/sbin. Depending
on how the PATH environment variable (page 283) is set, you may need to specify absolute pathnames for disable,
270
Draft
Sharing CUPS Printers
Draft
reject, and accept. Because enable is a bash builtin (page 211), you always need to specify the absolute pathname
of this utility. You may want to create easier-to-use aliases (page 305) for these commands.
Sharing CUPS Printers
IPP is designed for remote printing. By default, CUPS binds to localhost and accepts connections from the local
system only. To allow other systems to connect to CUPS on the local system, you need to instruct CUPS to bind
to an IP address that the other computers can reach. The Listen directive in the CUPS configuration file,
/etc/cups/cupsd.conf, specifies which IP address CUPS binds to and accepts requests on. The format of the Listen
directive follows:
Listen IP : port
where IP is the IP address that CUPS accepts connections on and port is the port number that CUPS listens on for
connections on IP . CUPS typically uses port 631. For example, the following directive causes CUPS to listen on
IP address 192.168.0.10, port 631:
Listen 192.168.0.10:631
After you change cupsd.conf, you need to restart the CUPS daemon:
# /sbin/service cups restart
Stopping cups:
Starting cups:
[ OK ]
[ OK ]
Once you restart the CUPS daemon, remote systems can print on the local system’s printers using the IP address
and port number specified with Listen directive. Make sure the system’s firewall allows LAN users to connect to
port 631 on the local system and does not allow systems outside the LAN to connect.
Alternatively, you can use CUPS’s access control list to permit only selected machines to connect to local printers.
An access control list is defined inside a <Location> container. The following example allows only the system at
IP 192.168.1.101 and the local system to print to the specified printer:
<Location /printers>
Order Deny,Allow
Allow from 192.168.1.101
Allow from @LOCAL
</Location>
The /printers indicates that this container refers to all local printers. Alternatively, you can control access on a
per-printer basis by specifying /printers/ name , where name is the printer name, or by specifying /printers/ path
.ppd, where path .ppd is the full pathname of the PPD file (page 495) that the printer uses.
The Order Deny,Allow line denies print requests by default and allows requests only from explicitly specified addresses. Specifying Order Allow,Deny allows print requests by default and denies requests from explicitly specified
addresses.
Allow from specifies the IP addresses that CUPS accepts connections from. Use Deny from with Order Allow,Deny
to specify IP addresses that CUPS will not accept connections from.
The @LOCAL macro specifies the local machine: It accept jobs from any address that resolves to the local machine.
Specifying 127.0.0.1 in place of @LOCAL would work as long as no application tried to print to the print server
271
Draft
The GNOME Print Manager
Draft
using its external IP address. Do not use the machine’s external IP address. Most processes use the loop-back device
(127.0.0.1) to connect to the printer, and the loopback device does not allow connections to any IP other than itself.
You can also use domain names, including wildcards, and IP ranges with either wildcards or netmasks in Allow
from and Deny from directives.
The GNOME Print Manager
The GNOME Print Manager (Figure 14-10) is similar to the Windows printer control panel; by default, it displays
an icon for each printer. You can run the GNOME Print Manager from the GNOME (Red Hat) menu: System
Tools -> Print Manager or from a virtual terminal.
Figure 14-10. The GNOME Print Manager
From the GNOME Print Manager window, double-click an icon to display the print queue for the printer represented by the icon. From this list you can cancel a job by highlighting it, right clicking, and clicking Cancel Documents.
$ gnome-print-manager &
When you ask the GNOME Print Manager to configure or add a printer and when you start it when no printers
are configured, it calls [system|redhat]-config-printer (page 487).
The KDE Printing Manager
KDE includes a printing abstraction layer (kprinter) that provides the print dialog box for KDE applications and
a utility for managing printers (Figure 14-11). Give the command kcmshell printmgr to display the Printing
Manager window; this is the same window that Control Center: Peripherals -> Printers displays. You can use
the related kprinter utility to print files.
272
Draft
The KDE Printing Manager
Draft
Figure 14-11. The KDE Printing Manager
The kprinter abstraction layer is not a stand-alone program and does not replace CUPS. Rather, kprinter is an
interface between an application or user submitting a print job and the printing system. KDE’s Printing Manager
provides much of the same functionality as CUPS and can use CUPS as a printing mechanism. With proper permissions, the Printing Manager
•
Starts print jobs . You can start a print job with the Printing Manager, as well as from a command line or
within an application.
•
Controls print jobs . The Printing Manager can display information on each of your print jobs. From these
windows, you can cancel print jobs (even when they have started printing), hold and release print jobs, and
move print jobs to different queues (as long as they have not started printing). You can also send a print job as
a fax or save it as a PDF or PostScript file.
•
Works with printers .
•
Works with multiple printing systems .
and other printing systems.
•
Works with filters .
install new ones.
You can add, remove, and modify printers and their properties.
The Printing Manager works with CUPS, LPRng, RLPR, PDQ,
The Printing Manager allows you to import existing printing filters (page 971) or to
273
Draft
Integration with Windows
Draft
The benefits of the Printing Manager are such that you may wish to use it from non-KDE applications. For example,
the Mozilla Web browser, as shipped by Red Hat, prints using lpr. You can change Mozilla so that it uses kprinter
by following these steps:
1.
Start Mozilla and select File -> Print from the Menubar.
2.
Click Properties in the Printer frame to display the Mozilla Printer Properties window.
3.
Change the command that starts with lpr in the Print Command text box to kprinter – –stdin. This command
runs kprinter and causes it to take its input from standard input.
4.
Click OK.
With these changes, when you print from Mozilla, kprinter opens a window that allows you to control where the
be printed output goes. Select the printer you want to use at the top of the window, make any other changes in the
window as appropriate, and click Print.
Refer to the KDEPrint Handbook (click “Help” from the kprinter window) for more information.
Integration with Windows
This section explains how to use Linux printers from Windows computers and how to use Windows printers from
Linux systems.
Printing from Windows
This section assumes that Samba (page 675) is installed and working on the Linux system that controls the printer
you want to use from Windows. Samba must be set up so that the Windows user who will be printing is mapped
to a Linux user (including mapping the Windows guest user to the Linux user nobody). Make sure that these users
have Samba passwords. Refer to “Samba Users, User Maps, and Passwords” on page 677.
Windows supports printer sharing via SMB, allowing a printer to be shared transparently between Windows systems
using the same mechanism as file sharing. Samba allows Windows users to use printers connected to Linux systems
just as they use any other shared printers. Because all Linux printers traditionally appear to be PostScript printers,
the Linux print server appears to share a PostScript printer. Windows does not include a generic PostScript printer
driver. Instead, Windows users must select a printer driver for a PostScript printer. The Apple Color LaserWriter
driver is a good choice.
When you use rpm to install Samba, it creates a directory named /var/spool/samba that is owned by root and
that anyone can read from and write to. The sticky bit (page 998) is set for this directory, allowing a Windows user
who starts a print job as a Linux user to be able to delete that job, but not allowing users to delete print jobs of
other users. Make sure this directory is in place and has the proper ownership and permissions:
$ ls -ld /var/spool/samba
drwxrwxrwt
2 root
root
4096 Feb 24 12:29 /var/spool/samba
Put the following two lines in the [global] section of the /etc/smb.conf file:
[global]
...
printing = cups
printcap name = cups
274
Draft
Printing to Windows
Draft
The printer’s share is listed in the [printers] section in smb.conf. Following, the path is the path Samba uses as a
spool directory and is not a normal share path. The following settings allow anyone, including guest, to use the
printer. Setting use client driver to YES causes Windows systems to use their own drivers. Not setting this option,
or setting it to NO, can cause printing from Windows not to work. Make sure the [printers] section in smb.conf
has the following entries:
[printers]
comment = All Printers
path = /var/spool/samba
printer admin = root
guest ok = Yes
printable = Yes
use client driver = Yes
browseable = No
Ideally, each user who is going to print should have an account. When multiple users share the same account (for
example, the nobody account), these users can delete each other’s print jobs.
Modern versions of Windows (2000 and later) support IPP and, as a result, can communicate directly with CUPS.
To use this feature, you must have CUPS configured on the Linux print server to allow remote IPP printing and
you need to create a new printer on the Windows system that points to the IP address of the Linux print server.
The details involved in configuring a Windows machine are beyond the scope of this book. You can use testparm
(page 694) and testprns to check the syntax of the Samba setup.
Printing to Windows
CUPS views a printer on a Windows computer exactly the same way it views any other printer. The only difference
is the URI you need to specify when connecting it. To configure a printer connected to a Windows machine, go to
the Printers page in the CUPS Web interface. From here, select Add Printer, as you would for a local printer.
When you are asked to select the device, choose Windows Printer via SAMBA. Next, enter the URI of the
printer using the following format: smb:// windows_system / printer_name. Once you have added the printer, you
can use it as you would any other printer.
Chapter Summary
A printing system such as CUPS sets up printers and moves a print job from an application or the command line
through the appropriate filters into a queue for a suitable printer and prints it.
CUPS is a cross-platform print server built around the IPP printing protocol. CUPS handles setting up and sending
jobs through print queues. The easiest way to work with CUPS is to use the Web interface which you can connect
to by pointing a Web browser at localhost:631 on the printer system connected to the printer. From the Web interface, you can configure print queues and modify print jobs in the queues.
You can use the traditional UNIX commands from a command line to send jobs to a printer (lpr/lp), display a print
queue (lpq/lpstat), and remove jobs from a print queue (lprm/cancel). In addition, CUPS provides the lpinfo and
lpadmin utilities to configure printers from the command line.
Samba enables you to print on a Linux printer from a Windows system and vice versa.
275
Draft
Exercises
Draft
Exercises
1. What commands can you use from a command line to send a file to the default printer?
2. What command would you give to cancel all your print jobs?
3. What commands list your outstanding print jobs?
4. What is the purpose of sharing a Linux printer using Samba?
5. Name three printing protocols that CUPS supports. Which is the CUPS native protocol?
Advanced Exercises
6. What command lists the installed printer drivers available to CUPS?
7. How would you send a text file to a printer connected to the first parallel port without using a print queue?
Why is doing this not a good idea?
8. Assume you have a USB printer with a manufacturer-supplied PostScript printer definition file named newprinter.ppd. What command would you use to add this printer to the system on the first USB port with the name
USBPrinter?
9. How would you define a quota that would allow each user to print up to 50 pages per week to the printer named
LaserJet?
10. Define a set of access control rules for a <Location> container inside /etc/cups/cupsd.conf that would allow
anyone to print to all printers as long as they were either on the local machine or in the mydomain.com domain.
276
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 20 - sendmail: Setting Up Mail Clients, Servers, and More
sendmail: Setting Up Mail Clients,
Servers, and More
IN THIS CHAPTER
JumpStart I: Configuring sendmail on a Client
JumpStart II: Configuring sendmail on a Server
How sendmail Works
Configuring sendmail
SpamAssassin
Webmail
Mailing Lists
627
Setting Up an IMAP or POP3 Server
628
Setting Up KMail
Authenticated Relaying
Sending and receiving email require three pieces of software. At each end, there is a client, called an MUA (Mail
User Agent), which is a bridge between a user and the mail system. Common MUAs are mutt, Kmail, Mozilla Mail,
and Outlook. When you send an email, the MUA hands it to an MTA (a Mail Transfer Agent such as sendmail),
which transfers it to the destination server. At the destination, an MDA (a Mail Delivery Agent such as procmail)
puts the mail in the recipient’s mailbox file. On Linux systems, the MUA on the receiving system either reads the
mailbox file or retrieves mail from a remote MUA or MTA, such as an ISP’s SMTP (mail) server, using POP (Post
Office Protocol) or IMAP (Internet Message Access Protocol).
Most Linux MUAs expect a local copy of sendmail will deliver outgoing email. On some systems, including those
with a dialup connection to the Internet, sendmail relays email to an ISP’s mail server. Because sendmail uses
SMTP (Simple Mail Transfer Protocol) to deliver email, sendmail is often referred to as an SMTP server.
In the default Red Hat setup, the sendmail MTA uses procmail as the local MDA. By default, procmail writes
email to the end of the recipient’s mailbox file. You can also use procmail to sort email according to a set of rules,
either on a per-user basis or globally. The global filtering function is useful for systemwide filtering to detect spam
and for other tasks, but the per-user feature is largely superfluous on a modern system. Traditional UNIX MUAs
were simple programs that could not filter mail and delegated this function to MDAs such as procmail. Modern
MUAs incorporate this functionality.
277
Draft
Introduction
Draft
Caution: You Do Not Need to Set Up sendmail to Send and Receive
Email
Most MUAs can use POP or IMAP for receiving email. These protocols do not require an MTA such
as sendmail. Thus you do not need to install or configure sendmail (or other MTA) to receive email.
You still need SMTP to send email. However, the SMTP server can be at a remote location, such as
your ISP, so you do not need to concern yourself with it.
Introduction
When the network that was to evolve into the Internet was first set up, it connected a few computers, each serving
a large number of users and running several services. Each computer was capable of sending and receiving email
and had a unique host name, which was used as a destination for email.
Today, the Internet has a large number of transient clients. Because these clients do not have fixed IP addresses or
hostnames, they cannot receive email directly. Users on these systems usually have an account on an email server
run by their employer or an ISP, and they collect email from this account using POP or IMAP. Unless you own a
domain that you want to receive email at, you will not need to set up sendmail as an incoming SMTP server.
You can set up sendmail on a client system so all it does is relay outbound mail to an SMTP server. This configuration is required by organizations that use firewalls to prevent email from being sent out on the Internet from any
system other than the company’s official mail servers. As a partial defense against spreading viruses, some ISPs
block outbound port 25 to prevent their customers from sending email directly to a remote computer. This configuration is required by these ISPs.
You can also set up sendmail as an outbound server that does not use an ISP as a relay. In this configuration,
sendmail connects directly to the SMTP servers for the domains receiving the email. An ISP set up as a relay is
configured this way.
You can set up sendmail to accept email for a registered domain name as specified in the domain’s DNS MX record
(page 706). However, most mail clients (MUAs) do not interact directly with sendmail to receive email. Instead,
they use POP or IMAP, protocols that include features for managing mail folders, leaving messages on the server,
and reading only the subject of an email without downloading the entire message. If you want to collect your email
from a system other than the one running the incoming mail server, you may need to set up a POP or IMAP server,
as discussed on page 628.
Prerequisites
Install the following packages:
•
sendmail (required)
•
sendmail-cf (required to configure sendmail)
•
squirrelmail (optional, provides Webmail, page 625)
•
spamassassin (optional, provides spam filtering, page 622)
•
mailman (optional, provides mailing list support, page 627)
•
imap (optional, provides IMAP and POP incoming mail server daemons)
278
Draft
More Information
Draft
Run chkconfig to cause sendmail to start when the system goes multiuser (by default, sendmail does not run in
single user mode):
# /sbin/chkconfig sendmail on
Start sendmail. Because sendmail is normally running, you need to restart it to cause sendmail to reread its configuration files. The following restart command works even when sendmail is not running; it just fails to shut
down sendmail:
# /sbin/service sendmail restart
Shutting down sendmail:
Shutting down sm-client:
Starting sendmail:
Starting sm-client:
[
[
[
[
OK
OK
OK
OK
]
]
]
]
Run chkconfig to cause the spamassassin daemon, spamd, to start when the system goes multiuser (spamassassin
is normally installed in this configuration):
# /sbin/chkconfig spamassassin on
As with sendmail, spamassassin is normally running; restart it to cause spamd to reread its configuration files:
# /sbin/service spamassassin restart
Starting spamd:
[ OK ]
The IMAP and POP protocols are implemented as several different daemons that are controlled by xinetd. See
page 628 for information on these daemons and how to start them.
More Information
Web .
sendmail .
IMAP .
www.sendmail.org [http://www.sendmail.org]
www.imap.org [http://www.imap.org]
SquirrelMail .
www.squirrelmail.org [http://www.squirrelmail.org]
Postfix .
www.postfix.org/docs.html [http://www.postfix.org/docs.html]
Qmail .
www.qmail.org [http://www.qmail.org]
Mailman .
www.list.org [http://www.list.org]
procmail .
www.procmail.org [http://www.procmail.org]
SpamAssassin .
Spam database .
spamassassin.org
razor.sourceforge.net
279
Draft
JumpStart I: Configuring sendmail on
a Client
Draft
JumpStart I: Configuring sendmail on a Client
This JumpStart configures an outbound sendmail server. This server
•
Uses a remote SMTP server, typically an ISP, to relay email to its destination
•
Sends to the SMTP server email originating from the local system only; it does not forward email originating
from other systems
•
Does not handle inbound email; as is frequently the case, you need to use POP or IMAP to receive email
To set up this server, you must edit /etc/mail/sendmail.mc and restart sendmail.
Change sendmail.mc .
The dnl at the start of the following line in sendmail.mc says that the line is a comment:
dnl define(`SMART_HOST', `smtp.your.provider')
To specify a remote SMTP server, you must open sendmail.mc in an editor and change the preceding line, deleting
dnl from the beginning of the line and replacing smtp.your.provider with the FQDN of your ISP’s SMTP server
(obtain this name from your ISP). Be careful not to alter the back tick (′) preceding or the single quotation mark
(’) following the FQDN. If your ISP’s SMTP server is at smtp.myisp.com, you would change the line:
dnl define(`SMART_HOST', `smtp.your.provider')
Restart sendmail .
you edited:
When you restart it, sendmail regenerates the sendmail.cf file from the sendmail.mc file
# /sbin/service sendmail restart
Test .
Test sendmail with the following command:
$ echo "my sendmail test" | /usr/sbin/sendmail [email protected]
Replace [email protected] with an email address on another system where you receive email. You need to send
email to a remote system to make sure that sendmail is relaying your email.
JumpStart II: Configuring sendmail on a Server
If you want to receive inbound email sent to a registered domain that you own, you need to set up sendmail as an
incoming mail server; this JumpStart describes how to set up such a server. This server
•
Accepts outbound email from the local system only
•
Delivers outbound email directly to the recipient’s system, without using a relay
•
Accepts inbound email from any system
This server does not relay outbound email originating on other systems. Refer to “access: Setting Up a Relay Host”
on page 620 if you want the local system to act as a relay. For this configuration to work, you must be able to make
outbound connections from, and receive inbound connections to, port 25.
280
Draft
How sendmail Works
Draft
The line in sendmail.mc that limits sendmail to accept inbound email from the local system only is
DAEMON_OPTIONS(`Port=smtp, Addr=127.0.0.1, Name=MTA')dnl
To allow sendmail to accept inbound email from other systems, remove the parameter Addr=127.0.0.1, from
the preceding line, leaving the following line:
DAEMON_OPTIONS(`Port=smtp, Name=MTA')dnl
By default sendmail does not use a remote SMTP server to relay email, so there is nothing to change to cause
sendmail to send email directly to recipients’ systems. (JumpStart I set up a SMART_HOST to relay email.)
Once you have restarted sendmail, it will accept mail addressed to the local system, as long as there is a DNS MX
record (page 706) pointing at the local system. If you are not running a DNS server, you must ask your ISP to set
up an MX record.
How sendmail Works
Outbound email . When you send email, the MUA passes the email to sendmail, which creates in the
/var/spool/mqueue directory two files that hold the message while sendmail processes it. In order to generate
unique filenames for a particular piece of email, sendmail generates a random string for each piece of email and
uses the string in filenames pertaining to the email. The sendmail daemon stores the body of the message in a file
named df (data file) followed by the generated string. It stores the headers and other information in a file named
qf (queue file) followed by the generated string.
If a delivery error occurs, sendmail creates a temporary copy of the message that it stores in a file whose name starts
with tf and logs errors in a file whose name starts xf. Once an email has been sent successfully, sendmail removes
all files pertaining to that email from /var/spool/mqueue.
Incoming email . By default, the MDA stores incoming messages in users’ files in the mail spool directory,
/var/spool/mail, in mbox format (next paragraph). Within this directory, each user has a mail file named with
the user’s username. Mail remains in these files until it is collected, typically by an MUA. Once an MUA collects
the mail from the mail spool, the MUA stores the mail as directed by the user, usually in the user’s home directory
hierarchy.
mbox versus maildir . The mbox format, which sendmail uses, stores all messages for a user in a single file.
To prevent corruption, the file must be locked while a process is adding messages to or deleting messages from the
file; you cannot delete a message at the same time the MTA is adding messages. A competing format, maildir,
stores each message in a separate file. This format does not use locks, allowing an MUA to read and delete messages
at the same time as new mail is delivered. In addition, the maildir format is better able to handle larger mailboxes.
The downside is that the maildir format adds overhead when using a protocol such as IMAP to check messages.
Qmail (page 633) uses maildir format mailboxes.
Mail logs
The sendmail daemon stores log messages in /var/log/maillog. Other mail servers, such as the imapd and ipop3d
daemons may log information to this file. Following is a sample log entry:
/var/log/maillog .
# cat /var/log/maillog
281
Draft
Aliases and Forwarding
Draft
...
Mar 3 16:25:33 MACHINENAME sendmail[7225]: i23GPXvm007224:
to=<[email protected]>, ctladdr=<[email protected]>
(0/0), delay=00:00:00, xdelay=00:00:00, mailer=local, pri=30514,
dsn=2.0.0, stat=Sent
Each log entry starts with a timestamp, the name of the system sending the email, the name of the mail server
(sendmail), and a unique identification number. The address of the recipient follows the to= label and the address
of the sender follows ctladdr=. Additional fields provide the name of the mailer and the time it took to send the
message. If a message is sent correctly, the stat= label is followed by Sent.
A message is marked Sent when sendmail sends it; Sent does not indicate that the message has been delivered. If
a message is not delivered due to an error down the line, the sender usually receives an email saying that it was not
delivered, giving a reason why.
If you get a lot of email, the maillog file can grow quite large; the syslog logrotate (page 543) entry is set up to
archive and rotate the maillog files regularly.
Aliases and Forwarding
There are three files that can forward email: .forward (page 615), aliases (next), and virtusertable (page 621).
See page 621 for a table comparing the three files.
/etc/aliases . Most of the time when you send email, it goes to a specific person; the recipient, [email protected],
maps to a specific, real user on the specified system. Sometimes you may want email to go to a class of user and
not to a specific recipient. Examples of classes of users are postmaster, webmaster, root, tech_support, and so
on. Different users may receive this email at different times or the email may be answered by a group of users. You
can use the /etc/aliases file to map inbound addresses to local users, files, commands, and remote addresses.
Each line in /etc/aliases contains the name of a local pseudouser, followed by a colon, whitespace, and a commaseparated list of destinations. The default installation includes a number of aliases that redirect messages for certain
pseudousers to root. These have the form
system:
root
Sending messages to the root account is a good way of making them easy to review, but, because it is rare that
anyone checks root’s email, you may want to send copies to a real user. The following line forwards mail sent to
abuse on the local system to root and alex:
abuse:
root, alex
You can create simple mailing lists with this type of alias. For example, the following alias sends copies of all email
sent to admin on the local system to several users, including Zach, who is on a different system:
admin:
sam, helen, mark, [email protected]
You can direct email to a file by specifying an absolute pathname in place of a destination address. The following
alias, which is quite popular among less conscientious system administrators, redirects email sent to complaints
to /dev/null (page 426) where they disappear:
complaints:
282
/dev/null
Draft
Related Programs
Draft
You can also send email to standard input of a command by preceding the command with a pipe character (|). This
technique is commonly used with mailing list software such as Mailman (page 627). For each list it maintains,
Mailman has entries, such as the following entry for mylist, in the aliases file:
mylist: "|/var/mailman/mail/mailman post mylist"
newaliases . After you edit /etc/aliases, you must either run newaliases as root or restart sendmail to recreate
the aliases.db file that sendmail reads.
praliases .
You can use praliases to list aliases currently loaded by sendmail:
# /usr/sbin/praliases | head -5
postmaster:root
daemon:root
adm:root
lp:root
shutdown:root
~/.forward . Systemwide aliases are useful in many cases, but nonroot users cannot make or change them.
Sometimes you may want to forward your own mail: Maybe you want to have mail from several systems go to one
address or perhaps you just want to forward your mail while you are at another office for a week. The ~/.forward
file allows ordinary users to forward their email.
Lines in a .forward file are the same as the right column of the aliases file explained previously: Destinations are
listed one per line and can be a local user, a remote email address, a filename, or a command preceded by a pipe
character (|).
Mail that you forward does not go to your local mailbox. If you want to forward mail and keep a copy in your
local mailbox, you must specify your local username preceded by a backslash to prevent an infinite loop. The following example sends Sam’s email to himself on the local system and on the system at tcorp.com:
$cat ~sam/.forward
[email protected]
\sam
Related Programs
sendmail . The sendmail distribution includes several programs. The primary program, sendmail, reads from
standard input and sends an email to the recipient specified by its argument. You can use sendmail from the
command line to check that the mail delivery system is working and to email the output of scripts:
$ echo "sample email" | /usr/sbin/sendmail [email protected]
mailq . The mailq utility displays the status of the outgoing mail queue and normally reports there are no
messages in the queue. Messages in the queue usually indicate a problem with the local or remote sendmail configuration or a network problem.
# /usr/bin/mailq
/var/spool/mqueue is empty
Total requests: 0
283
Draft
Configuring sendmail
Draft
mailstats . The mailstats utility reports on the number and sizes of messages sendmail has sent and received
since the date it displays on the first line:
# /usr/sbin/mailstats
Statistics from Sat Dec 22 16:02:34 2001
M
msgsfr bytes_from
msgsto
bytes_to
msgsrej msgsdis
0
0
0K
17181
103904K
0
0
4
368386
4216614K
136456
1568314K
20616
0
9
226151
26101362K
479025
12776528K
4590
0
==============================================================
T
594537
30317976K
632662
14448746K
25206
0
C
694638
499700
146185
Mailer
prog
esmtp
local
In the preceding output, each mailer is identified by the first column, which displays the mailer number, and by
the last column, which displays the name of the mailer. The second through fifth columns display the number and
total sizes of messages sent and received by the mailer. The sixth and seventh columns display the number of
messages rejected and discarded respectively. The row that starts with T lists the column totals and the row that
starts with C lists the number of TCP connections.
makemap . The makemap utility processes text configuration files in /etc/mail into the database format that
sendmail reads (*.db files). You do not need to run makemap manually; it is invoked by the sendmail init script
when you start or restart sendmail.
Configuring sendmail
The sendmail configuration files reside in /etc/mail where the primary configuration file is sendmail.cf. This
directory contains other text configuration files, such as access, mailertable, and virtusertable. The sendmail
daemon does not read these files but reads the corresponding *.db files in the same directory.
You can use makemap or give the command make from the /etc/mail directory to generate all the *.db files, although
this step is not usually necessary. The sendmail init script automatically generates these files when you restart
sendmail:
# /sbin/service sendmail restart
The sendmail.mc and sendmail.cf Files
This sendmail.cf file is not intended to be edited by hand and contains a large warning to this effect:
$ cat /etc/mail/sendmail.cf
...
###########################################################
#####
#####
DO NOT EDIT THIS FILE! Only edit the source .mc file.
#####
###########################################################
...
284
Draft
The sendmail.mc and sendmail.cf Files
Draft
Editing sendmail.mc and Generating sendmail.cf
The sendmail.cf file is generated from sendmail.mc using the m4 macro processor. It can be helpful to use a text
editor that supports syntax highlighting, such as vim, to edit sendmail.mc.
dnl . Many of the lines in sendmail.mc start with dnl, which stands for delete to new line, and instructs m4
to delete from the dnl to the end of the line (the next NEWLINE character). Because m4 ignores anything on a line
after a dnl instruction, you can use dnl to introduce comments; it works the same way as a # does in a shell script.
Many of the lines in sendmail.mc end with dnl. Because NEWLINEs immediately follow these dnls, these dnls
are superfluous; you can remove them if you like.
After editing sendmail.mc, you need to regenerate sendmail.cf and restart sendmail to make your changes take
effect. You can give the command make from the /etc/mail directory to regenerate sendmail.cf, although this
step is not usually necessary. The sendmail init script automatically regenerates sendmail.cf when you restart
sendmail.
About sendmail.mc
Lines near the beginning of sendmail.mc provide basic configuration information.
divert(-1)dnl
include(`/usr/share/sendmail-cf/m4/cf.m4')dnl
VERSIONID(`setup for Red Hat Linux')dnl
OSTYPE(`linux')dnl
The line that starts with divert tells m4 to discard extraneous output it may generate when processing this file.
The include statement, which tells m4 where to find the macro definition file that it will use to process the rest of
this file, points to the file named cf.m4. The cf.m4 file contains other include statements that include parts of the
sendmail configuration rule sets.
The VERSIONID statement defines a string that indicates the version of this configuration. You can change this
string to include a brief comment about the changes you make to this file or other information. The value of this
string is not significant to sendmail.
Do not change the OSTYPE statement unless you are migrating a sendmail.mc file from another operating system.
Other statements you may want to change are explained in the following sections and in the sendmail documentation.
Tip: Quoting m4 Strings
The m4 macro processor, which converts sendmail.mc to sendmail.cf, requires strings to be preceded
by a back tick (`) and closed with a single quotation mark (’).
Masquerading
Typically, you want your email to appear to come from the user and the domain where you receive email; sometimes
the outbound server is in a different domain than the inbound server. You can cause sendmail to alter outbound
messages so that they appear to come from a user and/or domain other than the one they are sent from: You masquerade (page 982) the message.
285
Draft
The sendmail.mc and sendmail.cf Files
Draft
There are several lines in sendmail.mc that pertain to this type of masquerading; each is commented out in the
file that Red Hat distributes:
dnl MASQUERADE_AS(`mydomain.com')dnl
dnl MASQUERADE_DOMAIN(localhost)dnl
dnl FEATURE(masquerade_entire_domain)dnl
The MASQUERADE_AS statement causes email that you send from the local system to appear to come from the
domain specified within the single quotation marks (mydomain.com in the commented out line in the distributed
file). Remove the leading dnl and change mydomain.com to the domain name that you want mail to appear to
come from.
The MASQUERADE_DOMAIN statement causes email from the specified system or domain to be masqueraded,
just as local email is. That is, email from the system specified in this statement is treated as though it came from
the local system: It is changed to appear to come from the domain specified in the MASQUERADE_AS statement.
Remove the leading dnl and change localhost to the name of the system or domain that sends the email that you
want to masquerade. If the name you specify has a leading period, it specifies a domain; if there is no leading period,
it specifies a system or host. You can have as many MASQUERADE_DOMAIN statements as necessary.
The masquerade_entire_domain feature statement causes sendmail also to masquerade subdomains of the domain
specified in the MASQUERADE_DOMAIN statement. Remove the leading dnl to masquerade entire domains.
Accepting Email from Unknown Hosts
As shipped by Red Hat, sendmail is configured to accept email from domains that it cannot resolve (and that may
not exist). To turn this feature off and cut down the amount of spam you receive, add dnl to the beginning of the
following line:
FEATURE(`accept_unresolvable_domains')dnl
When this feature is off, sendmail uses DNS to look up the domains of all email it receives; if it cannot resolve the
domain, it rejects the email.
Setting Up a Backup Server
You can set up a backup mail server to hold email when the primary mail server experiences problems. For maximum
coverage, the backup server should be on a different connection to the Internet from the primary server.
Setting up a backup server is easy: Remove the leading dnl from the following line in the backup mail server’s
sendmail.mc file:
dnl FEATURE(`relay_based_on_MX')dnl
DNS MX records (page 706) specify where email for a domain should be sent. You can have multiple MX records
for a domain, each pointing to a different mail server. When a domain has multiple MX records, each record usually
has a different priority; priority is specified by a two-digit number with lower numbers specifying higher priorities.
When attempting to deliver email, an MTA first tries to deliver email to the highest priority server. Failing that
delivery, it tries to deliver to a lower-priority server. If you activate the relay_based_on_MX feature and point a
low-priority MX record at a secondary mail server, the mail server will accept email for the domain. The mail
server will then forward email to the server identified by the highest priority MX record for the domain when that
server becomes available.
286
Draft
Other Files in /etc/mail
Draft
Other Files in /etc/mail
The /etc/mail directory holds most of the files that control sendmail. This section discusses three of those files:
mailertable, access, and virtusertable.
mailertable: Forwarding Email from One Domain to Another
When you run a mail server, you may want to send mail destined for one domain to a different location. The
sendmail daemon uses the /etc/mail/mailertable file for this purpose. Each line in mailertable holds the name
of a domain and a destination mailer separated by whitespace; when sendmail receives email for the specified domain, it forwards it to the mailer specified on the same line. Red Hat enables this feature by default: Put an entry
in the mailertable file and restart sendmail to use it.
The following line in mailertable forwards email sent to tcorp.com to the mailer at bravo.com:
$ cat /etc/mail/mailertable
tcorp.com
smtp:[bravo.com]
The square brackets in the example instruct sendmail not to use MX records but to send email directly to the
SMTP server. Without the brackets, email could be put in an infinite loop.
A period in front of a domain name acts as a wildcard and causes the name to match any domain that ends in the
specified name. For example, .tcorp.com matches sales.tcorp.com, mktg.tcrop.com, and so on.
The sendmail init script regenerates mailertable.db from mailertable each time you run it, as when you restart
sendmail.
access: Setting Up a Relay Host
On a LAN, you may want to set up a single server to process outbound mail, keeping local mail inside the network.
A system that processes outbound mail for other systems is called a relay host. The /etc/mail/access file specifies
which systems the local server relays email for. As distributed by Red Hat, this file lists only the local system:
$ cat /etc/mail/access
...
# by default we allow relaying from localhost...
localhost.localdomain
RELAY
localhost
RELAY
127.0.0.1
RELAY
You can add systems to the list in access by adding an IP address followed by whitespace and the word RELAY.
The following line adds the 192.168. subnet to the list of hosts that the local system relays mail for:
192.168.
RELAY
The sendmail init script regenerates access.db from access each time you run it, as when you restart sendmail.
287
Draft
Other Files in /etc/mail
Draft
virtusertable: Serving Email to Multiple Domains
When the DNS MX records are set up properly, a single system can serve email to multiple domains. On a system
that serves mail to many domains, you need a way to sort the incoming mail so that it goes to the right places. The
virtusertable file can forward inbound email addressed to different domains (aliases cannot do this).
As sendmail is shipped by Red Hat, virtusertable is enabled; you need to put forwarding instructions in the
/etc/mail/virtusertable file and restart sendmail to get it to work. The virtusertable file is similar to the aliases
file (page 614), except the left column contains full email addresses, not just local ones. Each line in virtusertable
has the address that the email was sent to, followed by whitespace and the address sendmail will forward the email
to. As with aliases, the destination can be a local user, an email address, a file, or a pipe symbol (|), followed by a
command.
The following line from virtusertable forwards mail addressed to [email protected] to zcs, a local user:
[email protected]
zcs
You can also forward email for a user to a remote email address:
[email protected]
[email protected]
You can forward all email destined for a domain to another domain without specifying each user individually. To
forward email for every user at bravo.com to tcorp.com, specify @bravo.com as the first address on the line.
When sendmail forwards email, it replaces %1 in the destination address with the name of the recipient. The next
rule forwards all email addressed to bravo.com to tcorp.com, keeping the original recipients’ names:
@bravo.com
%[email protected]
Finally, you can specify that email intended for a specific user should be rejected by using the error: namespace in
the destination. The next example bounces email addressed to [email protected] with the message 5.7.0:550
Invalid address:
[email protected]
error:5.7.0:550
Invalid address
.forward, aliases, and virtusertable . The .forward (page 615), aliases (page 614), and virtusertable files
do the same thing: They forward email addressed to one user to another user. They can also redirect email to a file
or as input to a program. The difference between them is scope and ownership; see Table 20-1.
Table 20-1. Comparison of Forwarding Techniques
Controlled by
.forward
aliases
virtusertable
nonroot user
root
root
Forwards email addressed nonroot user
to
Any real or virtual user on Any real or virtual user on
the local system
any domain recognized by
sendmail
Order of precedence
Second
288
Third
First
Draft
In Addition to sendmail
Draft
In Addition to sendmail
This section covers SpamAssassin, Webmail, and mailing lists. In addition, it discusses how to set up IMAP and
POP3 servers and a KMail client.
SpamAssassin
Spam, or more correctly, UCE (Unsolicited Commercial Email), accounts for about half of all email. SpamAssassin
evaluates each piece of incoming email and assigns it a number that indicates how likely it is that the email is spam.
The higher the number, the more likely it is that the email is spam. You can filter the email based on its rating.
SpamAssassin is effective as installed or you can modify its configuration files to make it better meet your needs.
How SpamAssassin works . You can set up SpamAssassin on a mail server so that it rates all inbound email
before it is sent to users. Or individual users can run it from their mail clients. Either way, you run the Spamassassin
spamd daemon and filter email through it using the spamc client.
SpamAssassin uses several techniques to identify spam:
•
Header analysis Checks for tricks that people who send spam use to make you think email is legitimate
•
Text analysis Checks the body of an email for characteristics of spam
•
Blacklists Checks various lists to see if the sender is known for sending spam
•
Database Checks the signature of the message against Vipul’s Razor (razor.sourceforge.net), a spam-tracking
database
With spamd running, you can see how spamc works by sending it a simple string:
$ echo "hi there" | spamc
X-Mail-Format-Warning: Bad RFC2822 header formatting in hi there
X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on
speedy.tcorp.com
X-Spam-Level: **
X-Spam-Status: No, hits=2.9 required=5.0
tests=DATE_MISSING,FROM_NO_LOWER
autolearn=no version=2.63
Of course, SpamAssassin complains because the string you gave it did not have standard email headers. The logical
line that starts with X-Spam-Status contains the heart of the report on the string hi there. First, it says No (it does
not consider the message to be spam). SpamAssassin uses a rating system that assigns a number of hits to a piece
of email. If the email receives over the required number of hits (5.0 by default), SpamAssassin marks it as spam.
The tests the string failed, because it did not have standard email headers, are DATE_MISSING and
FROM_NO_LOWER (the From line in the header had no lowercase characters [because there was no From line]).
The following listing is from a real piece of spam processed by SpamAssassin. It received 24.5 hits, indicating that
it is almost certainly spam.
X-Spam-Status: Yes, hits=24.5 required=5.0
tests=DATE_IN_FUTURE_06_12,INVALID_DATE_TZ_ABSURD,
MSGID_OE_SPAM_4ZERO,MSGID_OUTLOOK_TIME,
MSGID_SPAMSIGN_ZEROES,RCVD_IN_DSBL,RCVD_IN_NJABL,
RCVD_IN_UNCONFIRMED_DSBL,REMOVE_PAGE,VACATION_SCAM,
289
Draft
SpamAssassin
Draft
X_NJABL_OPEN_PROXY
version=2.55
X-Spam-Level: ************************
X-Spam-Checker-Version: SpamAssassin 2.55 (1.174.2.19-2003-05-19-exp)
X-Spam-Report:
This mail is probably spam. The original message has been attached
along with this report, so you can recognize or block similar unwanted
mail in future. See http://spamassassin.org/tag/ for more details.
Content preview: Paradise SEX Island Awaits! Tropical 1 week vacations
where anything goes! We have lots of WOMEN, SEX, ALCOHOL, ETC! Every
man's dream awaits on this island of pleasure. [...]
Content analysis details: (24.50 points, 5 required)
MSGID_SPAMSIGN_ZEROES (4.3 points) Message-Id generated by spam tool (zeroes variant)
INVALID_DATE_TZ_ABSURD (4.3 points) Invalid Date: header (timezone does not exist)
MSGID_OE_SPAM_4ZERO (3.5 points) Message-Id generated by spam tool (4-zeroes variant)
VACATION_SCAM
(1.9 points) BODY: Vacation Offers
REMOVE_PAGE
(0.3 points) URI: URL of page called "remove"
MSGID_OUTLOOK_TIME (4.4 points) Message-Id is fake (in Outlook Express format)
DATE_IN_FUTURE_06_12 (1.3 points) Date: is 6 to 12 hours after Received: date
RCVD_IN_NJABL
(0.9 points) RBL: Received via a relay in dnsbl.njabl.org
[RBL check: found 94.99.190.200.dnsbl.njabl.org.]
RCVD_IN_UNCONFIRMED_DSBL (0.5 points) RBL: Received via a relay in unconfirmed.dsbl.org
[RBL check: found 94.99.190.200.unconfirmed.dsbl.org.]
X_NJABL_OPEN_PROXY (0.5 points) RBL: NJABL: sender is proxy/relay/formmail/spam-source
RCVD_IN_DSBL
(2.6 points) RBL: Received via a relay in list.dsbl.org
[RBL check: found 211.157.63.200.list.dsbl.org.]
X-Spam-Flag: YES
Subject: [SPAM] re: statement
Because SpamAssassin considered the preceding email to be spam, it modified the Subject line by adding [SPAM]
at the beginning of the line.
Configuration . You can configure SpamAssassin globally by editing /etc/mail/spamassassin/local.cf. Users
can override the global options and add their own options in ~/.spamassassin/user_prefs. You can put the options
discussed in this section in either of these files. Use perldoc to display the configuration document that lists all the
options:
Documentation .
$ perldoc Mail::SpamAssassin::Conf
As shown in the preceding example, SpamAssassin rewrites the Subject line of email that it rates as spam. The rewrite_subject keyword in the configuration files controls this behavior. A 1 following this keyword indicates that
SpamAssassin will rewrite Subject lines. Change the 1 to a 0 (zero) to turn off this behavior:
rewrite_subject 0
The required_hits keyword specifies the minimum number of hits a piece of email must receive before SpamAssassin considers it to be spam. The default is 5.0. With a higher number, SpamAssassin marks fewer pieces of email
as spam.
required_hits 5.00
290
Draft
SpamAssassin
Draft
Sometimes, mail from addresses that should be marked as spam is not, or mail from addresses that should not be
marked as spam is. Use the whitelist_from keyword to specify addresses that should never be marked as spam
and blacklist_from to specify addresses that should always be marked as spam:
whitelist_from [email protected]
blacklist_from spammer.net
You can specify multiple addresses, separated by SPACEs, on the whitelist_from and blacklist_from lines and
each address can include wildcards. You can also use multiple lines.
Using SpamAssassin with a mail server . To add these headers to every email that arrives on the system, you
need to configure your MDA to pipe email through the spamc. The first step is to make sure you have procmail
configured as your MDA. The first of the following lines in sendmail.mc specifies the procmail command, its
path and flags. The MAILER line defines procmail as the mailer. You should not have to change either of these
lines.
FEATURE(local_procmail,`',`procmail -t -Y -a $h -d $u')dnl
MAILER(procmail)dnl
The procmail configuration file, /etc/procmailrc, may not exist on the server. If the file does not exist, create it
so that it is owned by root and has 644 permissions and the following contents. If it does exist, append the last
two lines from the following file to it:
$ cat /etc/procmailrc
DROPPRIVS=yes
:0 fw
| /usr/bin/spamc
The first line of this file ensures that procmail runs with the least possible privileges. The next two lines implement
a rule that pipes each user’s incoming email through spamc. The :0 tells procmail that a rule follows. The f flag
indicates a filter and the w causes procmail to wait for the filter to complete and check the exit code. The last line
specifies the /usr/bin/spamc file as the filter.
With the preceding changes in place, all email that comes into the system passes through SpamAssassin, which
rates it according to the options in the global configuration file. For users who have home directories on the server
system, SpamAssassin allows users’ configuration files to override the global file.
When you run SpamAssassin on a server, you typically want to rate the email more conservatively so fewer pieces
of good email are marked as spam. Setting required_hits in the range of 6–10 is generally appropriate. Also, you
do not want to remove any email automatically as there is a chance that you could prevent a user from getting a
piece of nonspam email. When the server marks email that is potentially spam, users can manually or automatically
filter the spam and decide what to do with it.
Using SpamAssassin with a mail client . With the SpamAssassin (spamd) daemon running and the configuration files set up, you are ready to have SpamAssassin filter your email. You need to set up two rules in your mail
client: The first passes each piece of email through SpamAssassin using spamc (page 622) and the second filters
email based on whether the X-SPAM-Flag line has a YES or NO on it.
In general, you do not want to pass very large pieces of email through SpamAssassin. This example, which uses
Kmail, passes messages smaller than 200,000 bytes through SpamAssassin. The first filter rule (Figure 20-1) processes
all messages where the size is less than 200000 (bytes). The action the rule takes is to pipe the message through
291
Draft
Webmail
Draft
/usr/bin/spamc. This rule uses nice to conserve system resources. Use nice or not as your system requires. The
rule is applied to incoming messages and filtering does not stop if the message matches this rule.
Figure 20-1. The first rule passes messages through SpamAssassin
The first rule adds an X-Spam-Flag line to each piece of email it processes. The second rule checks the value of the
flag on this line. If the X-Spam-Flag line contains YES, the second rule moves the email to a folder named spam
(Figure 20-2). Because the If this filter matches, stop processing here box is checked, Kmail does not further
process messages marked as spam. Messages not marked as spam can be processed by other rules.
Figure 20-2. The second rule checks the X-Spam-Flag
Webmail
Traditionally, you read email using a dedicated email client such as Kmail. Recently it has become more common
to use a Web application to read email. If you have an email account with a commercial provider such as HotMail
or Yahoo! Mail, you use a Web browser to read email. Email read in this manner is called Webmail. Unlike email
292
Draft
Webmail
Draft
you read on a dedicated client, you can read Webmail from anywhere you can open a browser on the Internet: You
can check your email from an Internet cafe or a friend’s computer.
SquirrelMail, distributed with Red Hat, provides Webmail services; the SquirrelMail files reside in
/usr/share/squirrelmail. If you want to run SquirrelMail, you must run IMAP (page 628) because SquirrelMail
uses IMAP to receive and authenticate email. You must also run Apache, Chapter 26, so a user can use a browser
to connect to SquirrelMail.
SquirrelMail is modular: You can easily add functionality using plugins. There are plugins that allow you to share
a calendar and plugins that give you the ability to change passwords using the Webmail interface. See the plugins
section of the SquirrelMail Web site for more information.
Create the following link to make SquirrelMail accessible from the Web:
# ln -s /usr/share/squirrelmail /var/www/html/mail
With this link in place, you can point a Web browser at http://localhost/mail to display the SquirrelMail login
page (Figure 20-3).
Figure 20-3. SquirrelMail login page
Next, use the conf.pl script in /usr/share/squirrelmail/config to configure SquirrelMail:
# cd /usr/share/squirrelmail/config
# ./conf.pl
SquirrelMail Configuration : Read: config.php (1.4.0)
--------------------------------------------------------Main Menu -1. Organization Preferences
2. Server Settings
293
Draft
3.
4.
5.
6.
7.
8.
9.
Mailing Lists
Draft
Folder Defaults
General Options
Themes
Address Books (LDAP)
Message of the Day (MOTD)
Plugins
Database
D. Set pre-defined settings for specific IMAP servers
C. Turn color off
S Save data
Q Quit
Command >>
The only item that you must set to get SquirrelMail to work is the server’s domain name (from the Server Settings
page). SquirrelMail provides several themes; if you do not like the way SquirrelMail looks, choose another theme
from the Themes page.
Mailing Lists
A mailing list can be an asset if you regularly send email to the same, large group of people. A mailing list provides
advantages over listing numerous recipients in the To or Cc field of an email or sending the same email individually
to many people:
•
Anonymity .
None of the recipients of the email can see the addresses of the other recipients.
•
Archiving . Email sent to the list is stored in a central location where list members or the public, as specified
by the list administrator, can browse through it.
•
Access control .
•
Consistency . When you send mail to a group of people using To or Cc, it is easy to leave people who want
to be on the list off and leave people who want to be off the list on.
•
Efficiency .
server.
You can easily specify who can send email to the list.
A mailing list application spreads sending email over time so it does not overload the mail
Mailman, included with Red Hat, provides mailing list support. Mailman resides in /var/mailman; the configuration
file is /var/mailman/Mailman/mm_cfg.py. Before you can use Mailman, you need to change the two following
lines in mm_cfg.py so that they point to your domain:
DEFAULT_URL_HOST
= 'mm_cfg_has_not_been_edited_to_set_host_domains'
DEFAULT_EMAIL_HOST = 'mm_cfg_has_not_been_edited_to_set_host_domains'
After making these changes, create a new mailing list with the newlist script:
# /var/mailman/bin/newlist
Enter the name of the list: painting_class
Enter the email of the person running the list: [email protected]
Initial list password:
To finish creating your mailing list, you must edit your /etc/aliases (or equivalent)
294
Draft
Setting Up an IMAP or POP3 Server
Draft
file by adding the following lines, and possibly running the 'newaliases' utility:
## list mailing list
list:
"|/var/mailman/mail/mailman post list"
list-admin:
"|/var/mailman/mail/mailman admin list"
list-bounces:
"|/var/mailman/mail/mailman bounces list"
list-confirm:
"|/var/mailman/mail/mailman confirm list"
list-join:
"|/var/mailman/mail/mailman join list"
list-leave:
"|/var/mailman/mail/mailman leave list"
list-owner:
"|/var/mailman/mail/mailman owner list"
list-request:
"|/var/mailman/mail/mailman request list"
list-subscribe:
"|/var/mailman/mail/mailman subscribe list"
list-unsubscribe: "|/var/mailman/mail/mailman unsubscribe list"
Hit enter to notify painting_class owner...
Before the list can receive email, you need to copy the lines generated by newlist to the end of /etc/aliases (page
614) and run newaliases.
Mailman includes a Web configuration interface that you can enable by configuring a Web server to run the scripts
in /var/mailman/cgi-bin. An entry such as the following in /etc/httpd/conf/httpd.conf (page 767) sets up this
interface (pipermail is the archive manager that Mailman uses):
# Mailman configs
ScriptAlias /mailman/ /var/mailman/cgi-bin/
Alias /pipermail/ /var/mailman/archives/public/
<Directory /var/mailman/archives>
Options +FollowSymLinks
</Directory>
Setting Up an IMAP or POP3 Server
There are two protocols that allow users to retrieve email remotely: IMAP (Internet Message Access Protocol) and
POP (Post Office Protocol). The imap package that Red Hat ships includes the imapd and ipop3d daemons that
implement these protocols.
The imapd and ipop3d daemons run through the xinetd super server (page 403). The imap package includes the
imap, imaps, ipop3, and ipop3s xinetd control files in the /etc/xinetd.d directory. The files whose names end
in s control the SSL versions of the protocols. By default, each of these protocols is disabled in its xinetd control
file.
To start one of these daemons, change the yes following disable = to no in the control file that corresponds to the
service you want to enable and restart xinetd.
If you enable an insecure protocol, you may want to restrict access to the local system by adding the following line
to the control file:
only_from = 127.0.0.1.
To use one of these protocols from a mail client, tell the mail client which protocol you are using and whether the
server is running SSL.
295
Draft
Setting Up KMail
Draft
Setting Up KMail
KMail is the graphical email client for KDE and is compatible with the MIME, SMTP, POP3, and IMAP standards.
To start KMail, give the command kmail from a terminal emulator window or from a Run Command window
(press ALT-F2 to open this window). Or you can choose Internet KMail from the KDE main menu. You can
run KMail from any desktop environment, including GNOME. Figure 20-4 shows the initial KMail window.
Figure 20-4. The initial KMail window
Selecting Configure KMail from the Settings menu on the menubar displays the Configure KMail window (Figure
20-5). This window has buttons along the left side; click the buttons to display different configuration pages on
the right.
296
Draft
Setting Up KMail
Draft
Figure 20-5. The Configure KMail window
Identity . KMail sets up a minimal identity for you. Click the Identities button to display the Identities page
where you can specify your email address, a reply-to address (if it is different from your email address), a signature
that KMail automatically appends to your outgoing email messages, and more.
Help . KMail provides help in setting up KMail to send and receive email. Click the Help button at the lowerleft of any KMail window to display the appropriate page of the online Configure KMail manual (part of the KDE
Help Center).
Once you have an identity, you need to set up incoming and outgoing accounts. Click the Network button to
display the Network page where you can set up accounts for sending and receiving messages. This page has two
tabs: Sending and Receiving.
Outgoing account . Click the Sending tab on the Network page to display the outgoing accounts. The outgoing
account defaults to sendmail on the local system. If you use the local sendmail, you need to configure it as explained
in “Jumpstart I: Configuring sendmail on a Client” on page 612. If you are using SMTP, you need to remove the
sendmail account and add an SMTP account: Highlight the sendmail account and click Remove; next, click Add
to display the Add Transport window where you can select sendmail or SMTP.
Incoming account . Click the Receiving tab on the Network page to display the incoming accounts; there is
no default incoming account. Click Add to display the Add Account window where you can select a type of account
such as Local mailbox, POP3, or IMAP. If you receive mail locally and from an ISP, you need to set up two accounts.
For a POP3 or IMAP account you need to specify the server (host) and your login name and password on the
server. If you want KMail to check for mail periodically, turn on Enable interval mail checking and specify how
often you want KMail to check for mail.
297
Draft
Authenticated Relaying
Draft
You do not have to change any settings on other pages. Following is a summary of what you will find on each of
the Configure KMail pages:
•
Identities . Specify one or more email identities including name, email addresses, signature, and if you are
using PGP or GnuPG (page 926), you can specify your OpenPGP key in the Advanced tab.
•
Network .
•
Appearance .
•
Compose . Specify what outgoing messages look like and what headers are included when you reply to or
forward a message.
•
Security . Specify various security features including whether you want to receive HTML messages in HTML
or plain text. Receiving HTML messages in HTML can make a system less secure.
•
Misc . Specify various KMail options including which warnings you receive, how messages you read are
marked, and what happens when you exit from KMail.
Specify outgoing and incoming email accounts.
Specify how KMail looks, including fonts, colors, layout, and headers.
KMail has a lot of options and features. Use the Help button to get assistance. It is easy to set up KMail for basic
use. As you become more comfortable using it, you can configure KMail to take care of more tasks for you.
Authenticated Relaying
If you travel with a portable computer such as a laptop, you may connect to the Internet through a different connection at each location you work. Perhaps you travel for work, or maybe you just bring your laptop home at night.
This section does not apply if you always dial into the network through your ISP. In that case, you are always
connected to your ISP’s network and it is as though you never moved your computer.
On a laptop you do not use a local instance of sendmail to send email: you use SMTP to connect to an ISP or to
a company’s SMTP server, which relays the outgoing mail. To avoid relaying email for anyone, including those
who would send spam, SMTP servers restrict whom they relay email for, based on IP address. By implementing
authenticated relaying, you can cause the SMTP server to authenticate, based on user identification. In addition,
SMTP can encrypt communication when you send mail from your email client and use the SMTP server.
An authenticated relay provides these advantages over a plain connection:
•
You can send email from any Internet connection.
•
The secure connection makes it more difficult to intercept email as it traverses the Internet.
•
The outgoing mail server requires authentication, preventing it from being used for spam.
You set up authenticated relaying by creating an SSL certificate (or using an existing one), enabling SSL in sendmail,
and telling your email client to connect to the SMTP server using SSL. If you have an SSL certificate from a company
such as Verisign, you can skip the next step, in which you create a self-signed certificate.
Creating a Self-Signed Certificate
The default location for SSL certificates is /usr/share/ssl/certs. Working as root, you can use the Makefile in this
directory to generate the required certificates. Apache uses a similar procedure for creating a certificate (page 796).
# cd /usr/share/ssl/certs
298
Draft
Enabling SSL in sendmail
Draft
# make sendmail.pem
...
Generating a 1024 bit RSA private key
..................................................++++++
...............++++++
writing new private key to '/tmp/openssl.OK6561'
----You are about to be asked to enter information that will be incorporated
into your certificate request.
What you are about to enter is what is called a Distinguished Name or a DN.
There are quite a few fields but you can leave some blank
For some fields there will be a default value,
If you enter '.', the field will be left blank.
----Country Name (2 letter code) [GB]:US
State or Province Name (full name) [Berkshire]:California
Locality Name (eg, city) [Newbury]:San Francisco
Organization Name (eg, company) [My Company Ltd]:Sobell Associates Inc.
Organizational Unit Name (eg, section) []:
Common Name (eg, your name or your server's hostname) []:sobell.com
Email Address []:[email protected]
You can enter any information you wish, with one exception. TLS/SSL negotiation fails if the hostname you specify
does not match that of the server.
Enabling SSL in sendmail
Once you have a certificate, instruct sendmail to use it. First, add the following lines to sendmail.mc. The first
of these lines tells sendmail to allow authenticated users to relay. The next two lines specify the authentication
mechanisms.
The first option for confAUTH_OPTIONS, A, instructs sendmail to use the AUTH parameter when sending mail
only if authentication succeeded. The second option, p, instructs sendmail, for connections that are not secure,
not to allow authentication methods that could be cracked by a packet sniffer.
define('confAUTH_OPTIONS', 'A p')
TRUST_AUTH_MECH('EXTERNAL DIGEST-MD5 CRAM-MD5 LOGIN PLAIN')
define('confAUTH_MECHANISMS', 'EXTERNAL GSSAPI DIGEST-MD5 CRAM-MD5 LOGIN PLAIN')
Next, add the following lines to sendmail.mc to tell sendmail where the certificate is:
define('CERT_DIR', '/usr/share/ssl/certs')
define('confCACERT_PATH', 'CERT_DIR')
define('confCACERT', 'CERT_DIR/sendmail.pem')
define('confSERVER_CERT', 'CERT_DIR/sendmail.pem')
define('confSERVER_KEY', 'CERT_DIR/sendmail.pem')
define('confCLIENT_CERT', 'CERT_DIR/sendmail.pem')
define('confCLIENT_KEY', 'CERT_DIR/sendmail.pem')
Encrypted connections are made in one of two ways: SSL (simpler) and TLS. SSL requires a dedicated port and has
the client and server negotiate a secure connection and continue the transaction as if the connection were unencrypted. TLS has the client connect to the server using an insecure connection and then issue a STARTTLS command
299
Draft
Enabling SSL in the Mail Client
Draft
to negotiate a secure connection. TLS runs over the same port as an unencrypted connection. Because many clients
support only SSL, it is a good idea to instruct sendmail to listen on the SMTPS port. The final line that you need
to add to sendmail.mc instructs sendmail to listen on the SSL port:
DAEMON_OPTIONS('Port=smtps, Name=TLSMTA, M=s')
Enabling SSL in the Mail Client
Enabling SSL in a mail client is usually quite simple. For example, the Kmail setup provides a Security tab that allows
you to choose the type of encryption you want to use: None, SSL, or TLS.
Alternatives to sendmail
Over its years, sendmail has grown to be enormously complex. Its complexity makes it hard to configure if you
want to set up something more than a simple mail server. And its size and complexity add to its vulnerability. For
optimum security, make sure you run the latest version of sendmail and always keep sendmail up to date. Or,
consider using one of the following alternatives.
Postfix . Postfix, which is included with Red Hat, is an alternative MTA. Postfix attempts to be fast and easy
to administer while, at the same time, being sendmail compatible enough to not upset sendmail users. Postfix
has a good reputation for ease of use and security and is a drop-in replacement for sendmail. Documentation for
Postfix is at www.postfix.org/docs.html [http://www.postfix.org/docs.html].
Qmail . Qmail is a direct competitor of Postfix and has the same objectives. By default, Qmail stores email using
the maildir format as opposed to the mbox format that other MTAs use (page 614). The Qmail Web site is
www.qmail.org [http://www.qmail.org].
Chapter Summary
The sendmail daemon is an MTA (Mail Transfer Agent). When you send a message, sendmail works with other
software to get it to the proper recipients. You can set up sendmail to relay email to an SMTP server that sends it
on to its destination or you can have sendmail send email directly to the SMTP servers for the domains receiving
the email. By default, sendmail stores incoming messages in the mail spool directory, /var/spool/mail.
The file that controls many aspects of how sendmail works is sendmail.cf. You edit sendmail.mc and when you
restart sendmail, the sendmail init script generates sendmail.cf. The system administrator can use the /etc/aliases
file and ordinary users can use ~/.forward files to reroute email to one or more local or remote addresses, to files,
or as input to programs.
You can use a program such as SpamAssassin to grade and mark email as to the likelihood of it being spam. You
can then decide what to do with the marked email: You can either look at each piece of potential spam and decide
where to put it or have your MUA automatically put potential spam in a special mailbox for spam.
Exercises
1. By default, email addressed to system goes to root. How would you also save a copy in /var/logs/systemmail?
2. How would Max store a copy of his email in ~/mbox and send a copy to [email protected]?
3. If your firewall allowed only the machine with the IP address 192.168.1.1 to send email outside the network,
how would you instruct your local copy of sendmail to use this server as a relay?
300
Draft
Advanced Exercises
Draft
4. What does dnl stand for in the m4 macro language; what are dnl commands used for?
5. SpamAssassin is installed on your mail server, with the threshold set to an unusually low value of 3, resulting in
a lot of false positives. What rule could you give to your mail client to allow it to identify spam with a score of
5 or above?
6. Describe the software and protocols used when Max sends an email to Sam on a remote Linux system.
Advanced Exercises
7. Your company’s current mail server runs on a commercial UNIX server, and you are planning on migrating it
to Linux. After copying the configuration files across to the Linux box, you find that it does not work. What
might you have forgotten to change?
8. Assume you have a script that sends its output to standard output. How would you modify the script to send
the output in an email to a user specified by the first argument on the command line? (You may assume that
the data is stored in $RESULT.)
9. Give a simple way of reading your email that does not involve the use of an MUA.
10. If you accidentally delete the /etc/aliases file, how could you easily recreate it (assuming that you had not restarted sendmail)?
301
302
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 19 - FTP: Transferring Files Across a Network
FTP: Transferring Files Across a Network
IN THIS CHAPTER
FTP Client
JumpStart: Downloading Files Using ftp
Anonymous FTP
Automatic Login
Binary versus ASCII Transfer Mode
FTP Server (vsftpd)
JumpStart: Starting a vsftpd Server
vsftpd.conf: Configuring vsftpd
The ftp File Transfer Protocol (FTP) is a method of downloading files from and uploading files to another system
using TCP/IP over a network. File Transfer Protocol is the name of a client/server protocol (FTP) and a client utility
(ftp) that invokes the protocol. In addition to the original ftp utility, there are many line-oriented and graphical
FTP client programs, including most browsers, that run under many different operating systems. There are also
many FTP server programs.
Introduction
First implemented under 4.2BSD, FTP has played an essential role in the propagation of Linux; it is the protocol/program frequently used to distribute free software. The term FTP site refers to an FTP server that is connected to a
network, usually the Internet. FTP sites can be public, allowing anonymous users to log in and download software
and documentation. Private FTP sites require you to log in with a username and password. Some sites allow you
to upload programs.
ftp and vsftpd . Although most FTP clients are similar, the servers differ quite a bit. This chapter describes the
ftp client with references to sftp, a secure FTP client, and it covers the FTP server that Red Hat uses internally and
offers as part of its distribution, vsftpd (very secure FTP).
Security . FTP is not a secure protocol. All usernames and passwords exchanged in setting up an FTP connection
are sent in cleartext, data exchanged over an FTP connection is not encrypted, and the connection is subject to hijacking. FTP is best used for downloading public files. In most cases, the OpenSSH clients, ssh (page 570), scp
(page 572), and sftp (page 573), offer secure alternatives to FTP.
303
Draft
More Information
Draft
Security: Use FTP Only to Download Public Information
You can use scp for almost all FTP functions other than allowing anonymous users to download information. Because scp uses an encrypted connection, user passwords and data cannot be sniffed. See
page 570 for more information on scp.
The vsftpd server does not make usernames, passwords, data, and connections more secure. The vsftpd server is
secure in that it is harder for a malicious user to compromise directly the system running it, even if vsftpd is poorly
implemented. One of the features that makes vsftpd more secure than, for example ftpd, is that it does not run
with root privileges. See also “Security” on page 595.
ftp utility . The ftp utility is a user interface to the standard File Transfer Protocol (FTP), which transfers files
between systems that can communicate over a network.
sftp utility .
Part of the OpenSSH suite, sftp is a secure alternative to ftp. See page 573 for more information.
FTP connections . FTP uses two connections: one for control (you establish this connection when you log in
on an FTP server) and one for data transfer (FTP sets this up when you ask it to transfer a file). An FTP server listens
for incoming connections on port 21 by default and handles user authentication and file exchange.
Passive versus active connections . A client can ask an FTP server to establish either a PASV (passive, default)
or a PORT (active) connection for data transfer. Some servers are limited to one type of connection or the other.
The difference between a passive and an active FTP connection is in whether the client or server initiates the data
connection. In passive mode, the client initiates the connection to the server (on port 20 by default), while in active
mode the server initiates the connection (there is no default port; see “Connection Parameters” on page 603 for
the parameters that determine which ports are used). Neither is inherently more secure. The reasons that passive
connections are more common are that a client behind a NAT (page 738) can connect to a passive server and it is
simpler to program a scalable passive server.
The parameters that control the type of connection that vsftpd server allows are discussed under “Connection
Parameters” on page 603.
More Information
Local . Type help or ? at any ftp> prompt to see a list of commands; follow the ? with a SPACE and an ftp
command for information about the command man pages ftp netrc vsftpd.conf
Web .
vsftpd home page.
HOWTO .
vsftpd.beasts.org
FTP mini-HOWTO
FTP Client
ftp . Red Hat supplies several different FTP clients including ftp (an older version of the BSD ftp utility). This
section discusses ftp as most other FTP clients provide a superset of ftp commands.
sftp .
Part of the OpenSSH suite, sftp is a secure alternative to ftp. See page 573 for more information.
gftp . Red Hat also provides gftp, a graphical client that works with FTP, SSH, and HTTP servers. This client
has many useful features, including the ability to resume an interrupted file transfer. See the gftp man page for more
information.
304
Draft
Prerequisites
Draft
ncftp . FEDORA includes ncftp, a text-based client with many more features than ftp, including filename completion and command line editing. See the ncftp man page for details.
Prerequisites
The ftp and sftp utilities are installed on most Red Hat systems. You can check for their presence by giving either
of these utilities’ names as commands:
$ ftp
ftp> quit
$ sftp
usage: sftp [-vC1] [-b batchfile] [-o option] [-s subsystem|path] ...
[-F config] [-P direct server path] [-S program]
[[email protected]]host[:file [file]]
Install the ftp or openssh-clients (contains sftp, see [page 567] for information on installing OPENssh) rpm
package if needed.
JumpStart: Downloading Files Using ftp
This JumpStart section is broken into two parts: a description of the basic commands and a tutorial session that
shows a user working with ftp.
Basic Commands
Give the command
$ ftp hostname
where hostname is the name of the FTP server you want to connect to. If you have an account on the server, log in
with your username and password. If it is a public system, log in as user anonymous (or ftp) and give your email
address as your password. Use the ls and cd ftp commands on the server as you would use the corresponding
utilities from a shell. The get file command copies file from the server to the local system, put file copies file from
the local system to the server, status displays information about the FTP connection, and help displays a list of
commands.
The preceding instructions, except for status, also work from sftp and ncftp.
Tutorial Session
Following are two ftp sessions wherein Alex transfers files from and to a vsftpd server named bravo. When Alex
gives the command ftp bravo, the local ftp client connects to the server, which asks for a username and password.
Because he is logged in on his local system as alex, ftp suggests that he log in on bravo as alex. To log in as alex,
he could just press RETURN; but his username on bravo is watson, so he types watson in response to the Name
(bravo:alex): prompt. Alex responds to the Password: prompt with his normal system password, and the vsftpd
server greets him and informs him that it is Using binary mode to transfer files. With ftp in binary mode, Alex
can transfer ASCII and binary files (page 589).
Connect and log in .
$ ftp bravo
Connected to bravo.
305
Draft
JumpStart: Downloading Files Using
ftp
Draft
220 (vsFTPd 1.2.0)
530 Please login with USER and PASS.
530 Please login with USER and PASS.
KERBEROS_V4 rejected as an authentication type
Name (bravo:alex): watson
331 Please specify the password.
Password:
230 Login successful.
Remote system type is UNIX.
Using binary mode to transfer files.
ftp>
After logging in, Alex uses the ftp ls command to see what is in his remote working directory, which is his home
directory on bravo. Then he cds to the memos directory and displays the files there.
ls and cd .
ftp> ls
227 Entering Passive Mode (192,168,0,6,79,105)
150 Here comes the directory listing.
drwxr-xr-x
2 500
500
4096 Oct 10 23:52 expenses
drwxr-xr-x
2 500
500
4096 Oct 10 23:59 memos
drwxrwxr-x
22 500
500
4096 Oct 10 23:32 tech
226 Directory send OK.
ftp> cd memos
250 Directory successfully changed.
ftp> ls
227 Entering Passive Mode (192,168,0,6,114,210)
150 Here comes the directory listing.
-rw-r--r-1 500
500
4770 Oct 10
-rw-r--r-1 500
500
7134 Oct 10
-rw-r--r-1 500
500
9453 Oct 10
-rw-r--r-1 500
500
3466 Oct 10
-rw-r--r-1 500
500
1945 Oct 10
226 Directory send OK.
23:58
23:58
23:58
23:59
23:59
memo.0514
memo.0628
memo.0905
memo.0921
memo.1102
Next, Alex uses the ftp get command to copy memo.1102 from the server to the local system. Binary mode ensures
that he will get a good copy of the file regardless of whether it is binary or ASCII. The server gives him confirmation
that the file was copied successfully and tells him the size of the file and how long it took to copy. Alex then copies
the local file memo.1114 to the remote system. The file is copied into his remote working directory, memos.
get and put .
ftp> get memo.1102
local: memo.1102 remote: memo.1102
227 Entering Passive Mode (192,168,0,6,194,214)
150 Opening BINARY mode data connection for memo.1102 (1945 bytes).
226 File send OK.
1945 bytes received in 7.1e-05 secs (2.7e+04 Kbytes/sec)
ftp> put memo.1114
306
Draft
JumpStart: Downloading Files Using
ftp
Draft
local: memo.1114 remote: memo.1114
227 Entering Passive Mode (192,168,0,6,174,97)
150 Ok to send data.
226 File receive OK.
1945 bytes sent in 2.8e-05 secs (6.8e+04 Kbytes/sec)
After a while, Alex decides he wants to copy all the files in the memo directory on bravo to a new directory on his
local system. He gives an ls command to make sure he is going to copy the right files, but ftp has timed out. Instead
of exiting from ftp and giving another ftp command from the shell, he gives ftp an open bravo command to reconnect
to the server. After logging in, he uses the ftp cd command to change directories to memos on the server.
timeout and open .
ftp> ls
421 Timeout.
Passive mode refused.
ftp> open bravo
Connected to bravo (192.168.0.6).
220 (vsFTPd 1.1.3)
...
ftp> cd memos
250 Directory successfully changed.
local cd (lcd) . At this point, Alex realizes he has not created the new directory to hold the files he wants to
download. Giving an ftp mkdir command would create a new directory on the server, but Alex wants a new directory
on his local system. He uses an exclamation point (!) followed by a mkdir memos.hold command to invoke a
shell and run mkdir on his local system, creating a directory named memos.hold in his working directory on the
local system. (You can display the name of your working directory on the local system with !pwd.) Next, because
he wants to copy files from the server to the memos.hold directory on his local system, he has to change his
working directory on the local system. Giving the command !cd memos.hold will not accomplish what Alex wants
to do because the exclamation point spawns a new shell on the local system and the cd command would be effective
only in the new shell, which is not the shell that ftp is running under. For this situation, ftp provides the lcd (local
cd) command, which changes the working directory for ftp and reports on the new local working directory.
ftp> !mkdir memos.hold
ftp> lcd memos.hold
Local directory now /home/alex/memos.hold
Alex uses the ftp mget (multiple get) command followed by the asterisk (*) wildcard to copy all the files from the
remote memos directory to the memos.hold directory on the local system. When ftp prompts him for the first
file, he realizes that he forgot to turn off prompts, responds with n, and presses CONTROL-C to stop copying files
in response to the second prompt. The server checks if he wants to continue with his mget command.
Next, he gives the ftp prompt command, which toggles the prompt action (turns it off if it is on and turns it on if
it is off). Now when he gives a mget * command, ftp copies all the files without prompting him. After getting the
files he wants, Alex gives a quit command to close the connection with the server, exit from ftp, and return to the
local shell prompt.
mget and prompt .
ftp> mget *
mget memo.0514? n
mget memo.0628? CONTROL-C
Continue with mget?
n
307
Draft
Notes
Draft
ftp> prompt
Interactive mode off.
ftp> mget *
local: memo.0514 remote: memo.0514
227 Entering Passive Mode (192,168,0,6,53,55)
150 Opening BINARY mode data connection for memo.0514 (4770 bytes).
226 File send OK.
4770 bytes received in 8.8e-05 secs (5.3e+04 Kbytes/sec)
local: memo.0628 remote: memo.0628
227 Entering Passive Mode (192,168,0,6,65,102)
150 Opening BINARY mode data connection for memo.0628 (7134 bytes).
226 File send OK.
...
150 Opening BINARY mode data connection for memo.1114 (1945 bytes).
226 File send OK.
1945 bytes received in 3.9e-05 secs (4.9e+04 Kbytes/sec)
ftp> quit
221 Goodbye.
Notes
A Linux system running ftp can exchange files with any of the many operating systems that support the FTP protocol.
Many sites offer archives of free information on an FTP server, although for many it is just an alternate to an easierto-access Web site (see for example, ftp://ftp.ibiblio.org/pub/Linux and http://www.ibiblio.org/pub/Linux). Most
browsers can connect to and download files from FTP servers. Many systems that permit anonymous access store
interesting files in the pub directory.
The ftp utility makes no assumptions about filesystem naming or structure because you can use ftp to exchange
files with non-UNIX/Linux systems (whose filenaming conventions may be different).
Anonymous FTP
Many systems, notably those that allow you to download free software, allow you to log in as anonymous. Most
systems that support anonymous logins accept the name ftp as an easier-to-spell and quicker-to-enter synonym
for anonymous. An anonymous user is usually restricted to a portion of a filesystem set aside to hold files that are
to be shared with remote users. When you log in as an anonymous user, the server prompts you to enter a password.
Although any password may be accepted, by convention you are expected to supply your email address.
Automatic Login
You can store server specific FTP username and password information so that you do not have to enter it each time
you visit an FTP site. Each line of ~/.netrc identifies a server. When you connect to an FTP server, ftp reads the
~/.netrc file to determine whether you have an automatic login set up for that server. The format of a line in
~/.netrc is
machine server login username password passwd
where server is the name of the server, username is your username and passwd is your password on server. Replace
machine with default on the last line of the file to specify a username and password for systems not listed in
~/.netrc. The default line is useful for logging in on anonymous servers. A sample ~/.netrc file follows:
308
Draft
Binary versus ASCII Transfer Mode
Draft
$ cat ~/.netrc
machine bravo login alex password mypassword
default login anonymous password [email protected]
Protect the account information in .netrc by making it readable by only the user whose home directory it appears
in. Refer to the netrc man page for more information.
Binary versus ASCII Transfer Mode
The vsftpd FTP server can, but does not always, provide two modes to transfer files. Binary mode transfers always
copy an exact, byte-for-byte image of a file and never changes line endings. Transfer all binary files using binary
mode. Unless you need to convert line endings, use binary mode to transfer ASCII files too.
ASCII files, such as text or program source code, created under Linux with a text editor such as vi, use a single
NEWLINE character (CONTROL-J, written as \n) to mark the end of each line. Other operating systems mark the
ends of lines differently. Windows marks the end of each such line with a RETURN (CONTROL-M, written as \r)
followed by a NEWLINE (two characters). Apple uses a RETURN by itself. These descriptions do not apply to files
created by word processors such as Word or OpenOffice as these programs generate binary files.
The vsftpd FTP server can map Linux line endings to Windows line endings as you upload files and Windows line
endings to Linux line endings as you download files. Although these features should arguably be on the client and
not the server, they are incorporated in vsftpd where the ASCII download feature can be a security risk.
To use ASCII mode on an FTP server that allows it, give an ascii command (page 592) after you log in and set cr
to ON (the default, page 592). If the server does not allow you to change line endings as you transfer a file, you
can use the unix2dos or dos2unix utilities before/after you transfer a file in binary mode.
Security . When run against a very large file, the ftp size command, which displays the size of a file, consumes
a lot of server resources and can be used to initiate a DoS attack (page 968). For security, by default, vsftpd transfers
every file in binary mode, even when it appears to be using ASCII mode. On the server side, you can enable real
ASCII mode transfers by setting the ascii_upload_enable and ascii_download_enable parameters (page 601)
to YES. With the server set to allow ASCII transfers, the client controls whether line endings are mapped by using
the ascii, binary, and cr commands (page 592).
ftp Specifics
This section covers the details of using ftp.
Format
The format of an ftp command line is shown following:
ftp [options] [ftp-server]
where options is one or more options from the list in the next section and ftp-server is the name or network address
of the FTP server that you want to exchange files with. If you do not specify an ftp-server, you will need to use the
ftp open command to connect to a server once ftp is running.
Command Line Options
–i .
interactive .
(page 593).
Turns off prompts during file transfers with mget (page 591) and mput (page 592). See also prompt
309
Draft
ftp Specifics
Draft
–g .
globbing .
Turns off globbing. See glob (page 592).
–v .
verbose . Tells you more about how ftp is working. Responses from the remote computer are displayed, and
ftp reports information on how quickly files are transferred. See also verbose (page 594).
–n .
no automatic login .
Disables automatic logins as described on page 589.
ftp Commands
The ftp utility is interactive: After you start ftp, it prompts you to enter commands to set parameters or transfer
files. You can abbreviate commands as long as the abbreviations are unique. Enter a question mark (?) in response
to the ftp> prompt to display a list of commands. Follow the question mark by a SPACE and a command to display
a brief description of what the command does:
ftp> ? mget
mget
get multiple files
Shell Command
![command] . Escapes to (spawns) a shell on the local system (use CONTROL-D or exit to return to ftp when
you are finished using the local shell). Follow the exclamation point with a command to execute that command
only; ftp returns you to the ftp> prompt when the command completes executing. Because the shell that ftp spawns
with this command is a child of the shell that is running ftp, no changes you make in this shell are preserved when
you return to ftp. Specifically, when you want to copy files to a local directory other than the directory that you
started ftp from, you need to use the ftp lcd command to change your local working directory: Issuing a cd command
in the spawned shell will not make the change you desire. See “local cd (lcd)” on page 587 for an example.
Transfer Files
In the following descriptions, the remote-file and local-file can be pathnames.
append local-file [ remote-file ] .
remote-file if specified.
Appends local-file to the file of the same name on the remote system or to
get remote-file [ local-file ] . Copies remote-file to the local system under the name local-file. Without localfile, ftp uses remote-file as the filename on the local system.
mget remote-file-list .
multiple get . The mget command copies several files to the local system, each maintaining its original filename.
You can name the remote files literally or use wildcards (see glob, page 592). See prompt (page 593) to turn off
prompts during transfers.
mput local-file-list .
multiple put . The mput command copies several files to the server, each maintaining its original filename.
You can name the local files literally or use wildcards (see glob, page 592). See prompt (page 593) to turn off
prompts during transfers.
310
Draft
ftp Specifics
Draft
put local-file [ remote-file ] . Copies local-file to the remote system under the name remote-file. Without remotefile, ftp uses local-file as the filename on the remote system.
newer remote-file [ local-file ] . If the modification time of remote-file is more recent than that of local-file or
if local-file does not exist, copies remote-file to the local system under the name local-file. Without local-file, ftp
uses remote-file as the filename on the local system. Similar to get.
reget remote-file [ local-file ] . If local-file exists and is smaller than remote-file, assumes that a previous get
of local-file was interrupted and continues from where the previous get left off. This command can save you time
when a get of a large file fails part way through the transfer.
Status
ascii .
binary .
bye .
Sets the file transfer type to ASCII. The cr command must be ON for ascii to work (page 589).
Sets the file transfer type to binary (page 589).
Closes the connection to the server and terminates ftp. Same as quit.
case . Toggles and displays case mapping status. Default is OFF. When ON, for get and mget commands,
maps filenames that are all uppercase on the server to all lower-case on the local system.
close .
Closes the connection to the server without exiting from ftp.
cr . Carriage RETURN Toggles and displays (carriage) RETURN stripping status. Effective only when file
transfer type is ascii. Set cr to ON (default) to remove RETURN characters from RETURN/LINEFEED line termination sequences used by Windows, yielding the standard Linux line termination of LINEFEED. Set cr to OFF to
leave line ending unmapped (page 589).
debug [n] . Toggles/sets and displays debugging status/level. The n is the debugging level. OFF or 0 (zero) is
the default. When n > 0, displays each command ftp sends to the server.
glob . Toggles and displays filename expansion (page 207) status for mdelete (page 593), mget (page 591),
and mput (page 592) commands.
hash . Toggles and displays pound sign (#, also called a hash mark) display status. When ON, ftp displays one
pound sign for each 1024-byte data block it transfers.
open .
Interactively specifies the name of the server. Useful when a connection times out or otherwise fails.
passive . Toggles between active (PORT, default) and passive (PASV) transfer modes and displays the transfer
mode. For more information, refer to “Passive versus active connections” on page 584.
prompt . Toggles and displays the prompt status. When ON (default), mdelete (page 593), mget (page 591),
and mput (page 592) ask for verification before transferring each file. Turn OFF to turn off these prompts.
quit .
Closes the connection to the server and terminates ftp. Same as bye.
umask [nnn] . Changes the umask (page 398) applied to files created on the server to nnn. Without an option,
displays the umask.
user [username] [password] . Prompts for or accepts the username and password that enable you to log in on
the server. Unless you call it with the –n option, ftp prompts you for a username and password automatically.
311
Draft
FTP Server (vsftpd)
Draft
Directories
cd remote-directory .
cdup .
Changes the working directory on the server to remote-directory.
Changes the working directory on the server to the parent of the working directory.
lcd [local_directory] .
local change directory . Changes the working directory on the local system to local_directory. Without an argument, this command changes the working directory on the local system to your home directory (just as the cd
shell builtin does without an argument). See “local cd (lcd)” on page 587 for an example.
Files
chmod mode remote-file . Changes access permissions of remote-file on the server to mode. See chmod on
(page 174) for more information on how to specify mode.
delete remote-file .
Removes remote-file from the server.
mdelete remote-file-list .
multiple delete .
Deletes the files specified by remote-file-list from the server.
Display Information
dir [remote-directory] [file] . Displays a listing of remote-directory from the server. Displays the working directory when you do not specify remote-directory. When you specify file, the listing is saved on the local system in
a file named file.
help .
Displays a list of local ftp commands.
ls [remote-directory] [file] .
Similar to dir but produces a more concise listing from some servers.
pwd . Displays the pathname of the working directory on the server. Use !pwd to display the pathname of the
local working directory.
status .
Displays ftp connection and status information.
verbose . Toggles and displays verbose mode, which displays responses from the server and reports on how
quickly files are transferred. Same as specifying the –v option on the command line.
FTP Server (vsftpd)
This section discusses the vsftpd server as supplied and installed under Red Hat Linux.
Prerequisites
Install the following packages:
•
vsftpd
Run chkconfig to cause vsftpd to start when the system goes multiuser.
# /sbin/chkconfig vsftpd on
312
Draft
Notes
Draft
Start vsftpd. If you change the vsftpd.conf configuration file, you need to restart vsftpd.
# /sbin/service vsftpd start
Notes
The vsftpd server can run in normal mode (the xinetd daemon [page 403] calls vsftpd each time a client tries to
make a connection) or it can run in standalone mode (vsftpd runs as a daemon and handles connections directly).
Standalone mode . Although by default vsftpd runs in normal mode, Red Hat sets it up to run in standalone
mode by setting the listen parameter (page 596) to YES in the vsftpd.conf file. Under Red Hat Linux, with vsftpd
running in standalone mode, you start and stop the server using service and the vsftpd init script.
Normal mode . You must install an xinetd control file (page 403) if you want to run vsftpd in normal mode.
There is a sample file at /usr/share/doc/vsftpd*/vsftpd.xinetd. Copy the sample file to /etc/xinetd.d, rename it
vsftpd, and edit the file to change the disable parameter to no. With the listen parameter set to NO, xinetd will
take care of starting vsftpd as needed.
Security . The safest policy is not to allow users to authenticate against FTP: Use FTP for anonymous access
only. If you do allow local users to authenticate and upload files to the server, be sure to put local users in a chroot
jail (page 598). Because FTP sends usernames and passwords in cleartext, a malicious user can easily sniff (page
997) them. With a username and password, the same user can impersonate a local user, upload a Trojan horse
(page 1002), and compromise the system.
JumpStart: Starting a vsftpd Server
As vsftpd is installed from the Red Hat rpm file, local and anonymous users can log in on the server; there is no
guest account. You do not have to configure anything.
Testing the Setup
Make sure vsftpd is working by logging in from the system running the server. You can refer to the server as localhost
or by using its hostname on the command line. Log in as anonymous; use any password.
$ ftp localhost
Connected to localhost.localdomain.
220 (vsFTPd 1.2.0)
530 Please login with USER and PASS.
530 Please login with USER and PASS.
KERBEROS_V4 rejected as an authentication type
Name (bravo:alex): anonymous
331 Please specify the password.
Password:
230 Login successful.
Remote system type is UNIX.
Using binary mode to transfer files.
ftp> quit
221 Goodbye.
If you are not able to connect to the server, check the following:
Make sure the server is running.
313
Draft
vsftpd.conf: Configuring vsftpd
Draft
# /sbin/service vsftpd status
vsftpd (pid 3091) is running...
Check that permissions on /var/ftp, or the home directory of ftp as specified in /etc/passwd, are set to 755. If the
ftp user can write to /var/ftp, connections will fail.
# ls -ld /var/ftp
drwxr-xr-x
4 root
root
4096 Aug 27 23:54 /var/ftp
Once you are able to log in from the local system, log in from another system, either one on your LAN or another
system with access to the server. On the command line, use the hostname from within your LAN or the FQDN
(page 972) from outside your LAN. The dialog should appear the same as in the previous example. If you cannot
log in from a system that is not on your LAN, use ping (page 346) to test the connection and make sure your firewall
is set up to allow FTP access. See “FTP connections” on page 584 for a discussion of active and passive modes and
the ports that each uses.
vsftpd.conf: Configuring vsftpd
The configuration file for vsftpd, /etc/vsftpd/vsftpd.conf, lists Boolean, numeric, and string name-value pairs of
configuration parameters, called directives. Each name-value pair is joined by an equal sign with no SPACEs on
either side. Red Hat provides a well-commented /etc/vsftpd/vsftpd.conf file that changes many of the compiledin defaults. This section covers most of the options, noting their default values and their values as specified in the
vsftpd.conf file supplied by Red Hat.
Set Boolean options to YES or NO and numeric options to a nonnegative integer. Octal numbers, useful for setting
umask options, must have a leading 0 (zero), or the number will be treated as base ten. Following are examples
from vsftpd.conf of setting each type of option:
anonymous_enable=YES
local_umask=022
xferlog_file=/var/log/vsftpd.log
Where Red Hat has overridden the default value by putting a different value in vsftpd.conf, both the default and
Red Hat values are noted. The directives are broken into the following groups:
•
Standalone Mode (page 596)
•
Logging In (page 597)
•
Working Directory and the chroot Jail (page 598)
•
Downloading and Uploading Files (page 600)
•
Messages (page 602)
•
Display (page 602)
•
Logs (page 603)
•
Connection Parameters (page 603)
314
Draft
vsftpd.conf: Configuring vsftpd
Draft
Standalone Mode
Refer to “Notes” on page 588 for a discussion of normal and standalone modes. This section describes the parameters
that affect standalone mode.
listen .
YES runs vsftpd in standalone mode; NO runs it in normal mode.
Default: NO
Red Hat: YES
listen_port .
In standalone mode, specifies the port that vsftpd listens on for incoming connections.
Default: 21
listen_address . In standalone mode, specifies the IP address of the local interface that vsftpd listens on for
incoming connections. When not set, vsftpd uses the default network interface.
Default: none
max_clients .
In standalone mode, specifies the maximum number of clients. Zero (0) indicates unlimited clients.
Default: 0
max_per_ip . In standalone mode, specifies the maximum number of clients from the same IP address. Zero
(0) indicates unlimited clients from the same IP address.
Default: 0
Logging In
There are three classes of users who can log in on a vsftpd server: anonymous, local, and guest. The guest user is
rarely used and is not covered in this chapter. Local users log in with their system username and password, while
anonymous users log in with anonymous or ftp, using their email address as a password. You can control
whether each of these classes of users can log in on the server and what they can do once they log in. You can specify
what a local user can do on a per-user basis; refer to user_config_dir on page 606.
Local Users
userlist_enable . The /etc/vsftpd.user_list file (page 606), or other file specified by userlist_file, contains a
list of zero or more users. YES consults the list and takes action based on userlist_deny, either granting or denying
users in the list permission to log in on the server. To prevent the transmission of cleartext passwords, access is
denied immediately after the user enters his/her username. NO does not consult the list. For a more secure system,
set to NO.
Default: NO
Red Hat: YES
userlist_deny . YES prevents users listed in /etc/vsftpd.user_list (page 606) from logging in on the server. NO
allows only users listed in /etc/vsftpd.user_list to log in on the server. Use userlist_file to change the name of
the file that this parameter consults. This parameter is checked only when userlist_enable is set to YES.
Default: YES
userlist_file .
The name of the file consulted when userlist_enable is set to YES.
315
Draft
vsftpd.conf: Configuring vsftpd
Draft
Default: /etc/vsftpd.user_list
local_enable .
YES permits local users (users listed in /etc/passwd) to log in on the server.
Default: NO
Red Hat: YES
Anonymous Users
anonymous_enable .
YES allows anonymous logins.
Default: YES
no_anon_password .
YES skips asking anonymous users for passwords.
Default: NO
deny_email_enable . YES checks to see if the password (email address) that an anonymous user enters is listed
in /etc/vsftpd.banned_emails or other file specified by banned_email_file. If it is, the user is not allowed to log
in on the system. NO does not perform this check. Using iptables (page 737) to block specific hosts is generally
more productive than using this parameter.
Default: NO
banned_email_file .
The name of the file consulted when deny_email_enable is set to YES.
Default: (/etc/vsftpd.banned_emails)
Working Directory and the chroot Jail
When a user logs in on a vsftpd server, standard filesystem access permissions control which directories and files
the user can access and how the user can access them. There are three basic parameters that control a user who is
logged in on a vsftpd server. They are the user’s
•
User ID (UID)
•
Initial working directory
•
Root directory
By default, the vsftpd server sets the user ID of a local user to that user’s login name and sets the user ID of an
anonymous user to ftp. A local user starts in her home directory and an anonymous user starts in /var/ftp.
By default, anonymous users are placed in a chroot jail for security; local users are not. For example, when an anonymous user logs in on a vsftpd server, his home directory is /var/ftp. But all that user sees is that his home directory is /. The user sees the directory at /var/ftp/upload as /upload. The user cannot see, nor work with, for
example, the /home, /usr/local, or /tmp directories. The user is in a chroot jail. For more information, refer to
“Setting Up a chroot Jail” on page 406.
You can use the chroot_local_user option to put each local user in a chroot jail whose root is the user’s home
directory. You can use chroot_list_enable to put selected local users in chroot jails.
chroot_list_enable . Upon login, YES checks to see whether a local user is listed in /etc/vsftpd.chroot_list
(page 606) or other file specified by chroot_list_file.
316
Draft
vsftpd.conf: Configuring vsftpd
Draft
When a user is in the list and chroot_local_user is set to NO, the user is put in a chroot jail in his home directory.
Only users listed in /etc/vsftpd.chroot_list are put in chroot jails.
When a user is in the list and chroot_local_user is set to YES, that user is not put in a chroot jail. Users not listed
in /etc/vsftpd.chroot_list are put in chroot jails.
Default: NO
chroot_local_user . See chroot_list_enable. Set to NO for a more open system, but remember to add new
users to the chroot_list_file as needed as you add users to the system. Set to YES for a more secure system. New
users are automatically restricted unless you add them to chroot_list_file.
Default: NO
chroot_list_file .
The name of the file consulted when chroot_list_enable is set to YES.
Default: /etc/vsftpd.chroot_list
passwd_chroot_enable . YES enables you to change the location of the chroot jail that the chroot_list_enable
and chroot_local_user settings impose on a local user.
The location of the chroot jail can be moved up the directory structure by including a /./ within the home directory
string for that user in /etc/passwd. This change has no effect on the standard system login, just as cd . has no effect
on your working directory.
For example, changing the home directory field in /etc/passwd (page 430) for Sam from /home/sam to /home/./sam
allows Sam to cd to /home after logging in using vsftpd. Given the proper permissions, this ability can allow Sam
to view files and possibly collaborate with another user.
Default: NO
secure_chroot_dir . The name of an empty directory that is not writable by user ftp. The vsftpd server uses
this directory as a secure chroot jail when the user does not need access to the filesystem.
Default: /usr/share/empty
local_root . After a local user logs in on the server, this directory becomes the user’s working directory. No error
results if the specified directory does not exist.
Default: none
Downloading and Uploading Files
By default, any user, local or anonymous, can download files from the server, assuming proper filesystem access
and permissions. You must change write_enable from NO (default) to YES to permit local users to upload files.
By default, local_umask is set to 022, giving uploaded files 644 permissions (page 173).
Security . Refer to “Security” on page 595 for information on the security hole that is created when you allow
local users to upload files.
The following actions set up vsftpd to allow anonymous users to upload files:
1.
Set write_enable (page 600) to YES.
2.
Create a directory under /var/ftp that an anonymous user can write to but not read from (mode 333). You
do not want a malicious user to be able to see, download, modify, and upload a file that another user originally
317
Draft
vsftpd.conf: Configuring vsftpd
Draft
uploaded. The following commands create a /var/ftp/uploads directory that anyone can write to but no one
can read from:
# mkdir /var/ftp/uploads
# chmod 333 /var/ftp/uploads
Because of the security risk, vsftpd prevents anonymous connections when an anonymous user (ftp) can write
to /var/ftp.
3.
Set anon_upload_enable (page 601) to YES.
4.
See other options in this section.
Download/Upload for Local Users
local_umask .
The umask (page 398) setting for local users.
Default: 077
Red Hat: 022
file_open_mode . Uploaded file permissions for local users. The umask (page 398) is applied to this value.
Change to 0777 to make uploaded files executable.
Default: 0666
write_enable . YES permits users to create and delete files and directories (assuming appropriate filesystem
permissions). NO prevents users from making changes to the filesystem.
Default: NO
Red Hat: YES
Anonymous Users
anon_mkdir_write_enable . YES permits an anonymous user to create new directories when write_enable=YES
and the anonymous user has permission to write to the parent directory.
Default: NO
anon_other_write_enable . YES grants an anonymous user write permission in addition to the permissions
granted by anon_mkdir_write_enable and anon_upload_enable. For example, YES allows an anonymous user
to delete and rename files, assuming permission to write to the parent directory. Not recommended for secure sites.
Default: NO
anon_root . After an anonymous user logs in on the server, this directory becomes the user’s working directory.
No error results if the specified directory does not exist.
Default: none
anon_umask . The umask (page 398) setting for anonymous users. The default setting gives only anonymous
users access to files uploaded by anonymous users; set to 022 to give everyone read access to these files.
Default: 077
318
Draft
vsftpd.conf: Configuring vsftpd
Draft
anon_upload_enable . YES allows anonymous users to upload files when write_enable=YES and the anonymous
user has permission to write to the directory.
Default: NO
anon_world_readable_only . YES limits the files that a user can download to those that are readable by the
owner, group, and others. It may not be desirable to allow one anonymous user to download a file that another
anonymous user uploaded. Setting this parameter to YES can prevent this scenario.
Default: YES
ascii_download_enable . YES allows a user to download files using ASCII mode. Setting this parameter to YES
can create a security risk (page 590).
Default: NO
ascii_upload_enable .
YES allows a user to upload files using ASCII mode (page 589).
Default: NO
chown_uploads . YES causes files uploaded by anonymous users to be owned by root (or other user specified
by chown_username).
Default: NO
chown_username .
See chown_uploads.
Default: root
ftp_username .
The login name of anonymous users.
Default: ftp
nopriv_user . The name of the user with minimal privileges, as used by vsftpd. For security, because other
programs use nobody, replace nobody with the name of a dedicated user such as ftp.
Default: nobody
Messages
You can replace the standard greeting banner that vsftpd displays when a user logs in on the system (banner_file
and ftpd_banner) and you can display a message each time a user enters a directory (dirmessage_enable and
message_file). When you set dirmessage_enable=YES, each time a user enters a directory using cd, vsftpd displays
the contents of the file in that directory named .message (or other file specified by message_file).
dirmessage_enable . YES displays .message or other file specified by message_file as an ftp user enters a new
directory by giving a cd command.
Default: NO
Red Hat: YES
message_file .
See dirmessage_enable.
Default: .message
319
Draft
vsftpd.conf: Configuring vsftpd
banner_file .
ftpd_banner.
Draft
The absolute pathname of the file that is displayed when a user connects to the server. Overrides
Default: none
ftpd_banner .
server.
This string overrides the standard vsftpd greeting banner displayed when a user connects to the
Default: none, uses standard vsftpd banner
Display
This section describes parameters that can improve security and performance by controlling how vsftpd displays
information.
hide_ids .
YES lists all users and groups in directory listings as ftp. NO lists real owners.
Default: NO
setproctitle_enable . NO causes ps to display the process running vsftpd as vsftpd. YES causes ps to display
what vsftpd is currently doing (uploading and so on). Set to NO for a more secure system.
Default: NO
text_userdb_names .
YES displays names.
NO improves performance by displaying numeric UIDs and GIDs in directory listings.
Default: NO
use_localtime . NO causes ls, mls, and modtime FTP commands to display UTC (page 1004); YES causes
these commands to display local times.
Default: NO
ls_recurse_enable . YES permits allows users to give ls –R commands. Setting this parameter to YES may pose
a security risk because giving an ls –R command at the top of a large directory hierarchy can consume a lot of system
resources.
Default: NO
Logs
By default, logging is turned off. However, the vsftpd.conf file that Red Hat supplies turns it on. This section describes parameters that control the detail and location of logs.
log_ftp_protocol .
YES logs FTP requests and responses, provided that xferlog_std_format is set to NO.
Default: NO
xferlog_enable . YES maintains a transfer log in /var/log/vsftpd.log (or other file specified by xferlog_file).
NO does not create a log.
Default: NO
xferlog_std_format . YES causes a transfer log (not covering connections) to be written in standard xferlog
format, as used by wu-ftpd, as long as xferlog_file is explicitly set. The default vsftpd log format is more readable
320
Draft
vsftpd.conf: Configuring vsftpd
Draft
than xferlog format, but it cannot be processed by programs that generate statistical summaries of xferlog files.
Search for xferlog on the Internet for more information.
Default: NO
xferlog_file .
See xferlog_enable and xferlog_std_format.
Default: /var/log/vsftpd.log
Connection Parameters
You can allow clients to establish passive and/or active connections (page 584). Setting timeouts and maximum
transfer rates can improve server security and performance. This section describes parameters that control types
of connections that a client can establish, the length of time vsftpd will wait while establishing a connection, and
the speeds of connections for different types of users.
Passive (PASV) Connections
pasv_enable .
NO prevents the use of PASV connections.
Default: NO
pasv_promiscuous . NO causes PASV to perform a security check that ensures that the data and control connections originate from a single IP address. YES disables this check and is not recommended for a secure system.
Default: NO
pasv_max_port .
up a firewall.
The highest port number that vsftpd will allocate for a PASV data connection, useful in setting
Default: 0 (use any port)
pasv_min_port .
up a firewall.
The lowest port number that vsftpd will allocate for a PASV data connection, useful in setting
Default: 0 (use any port)
pasv_address .
Specifies an IP address other than the one used by the client to contact the server.
Default: none, the address is the one used by the client
Active (PORT) Connections
port_enable .
NO prevents the use of PORT connections.
Default: YES
port_promiscuous . NO causes PORT to perform a security check that ensures that outgoing data connections
connect only to the client. YES disables this check and is not recommended for a secure system.
Default: NO
connect_from_port_20 . YES specifies port 20 (ftp-data, a privileged port) on the server for PORT connections,
as required by some clients. NO allows vsftpd to run with fewer privileges (on a nonprivileged port).
Default: NO
321
Draft
vsftpd.conf: Configuring vsftpd
Draft
Red Hat: YES
ftp_data_port .
tions.
With connect_from_port_20 set to NO, specifies the port that vsftpd uses for PORT connec-
Default: 20
Timeouts
accept_timeout .
The number of seconds the server waits for a client to establish a PASV data connection.
Default: 60
connect_timeout .
The number of seconds the server waits for a client to respond to a PORT data connection.
Default: 60
data_connection_timeout .
disconnecting.
The number of seconds the server waits for a stalled data transfer to resume before
Default: 300
idle_session_timeout .
The number of seconds the server waits between FTP commands before disconnecting.
Default: 300
local_max_rate .
For local users, the maximum data transfer rate in bytes per second. Zero indicates no limit.
Default: 0
anon_max_rate .
no limit.
For anonymous users, the maximum data transfer rate in bytes per second. Zero indicates
Default: 0
one_process_model . YES establishes one process per connection, improving performance but degrading security.
NO allows multiple processes per connection. NO is recommended for a more secure system.
Default: NO
Miscellaneous
This section describes parameters not discussed elsewhere.
pam_service_name .
The name of the PAM service vsftpd uses.
Default: ftp
Red Hat: vsftpd
tcp_wrappers . YES causes incoming connections to use tcp_wrappers (page 404) if vsftpd was compiled
with tcp_wrappers support. When tcp_wrappers sets the environment variable VSFTPD_LOAD_CONF,
vsftpd loads the configuration file specified by this variable, allowing per-IP configuration.
Default: NO
Red Hat: YES
322
Draft
Chapter Summary
Draft
user_config_dir . Specifies a directory that contains files named for local users. Each of these files, which
mimics vsftpd.conf, contains parameters that override, on a per-user basis, default parameters and those specified
in vsftpd.conf. For example, assume that user_config_dir is set to /etc/vsftpd_user_conf. If the default configuration file, /etc/vsftpd/vsftpd.conf, sets idlesession_timeout=300 and Sam’s individual configuration file,
/etc/vsftpd_user_conf/sams, sets idlesession_timeout=1200, all users’ sessions, except for Sam’s, will timeout
after 300 seconds of inactivity. Sam’s sessions will timeout after 1200 seconds.
Default: none
Files
In addition to /etc/vsftpd.conf, the following files control the functioning of vsftpd. The directory hierarchy that
user_config_dir points to is not included in this list as it has no default name.
/etc/vsftpd.ftpusers . Lists users, one per line, who are never allowed to log in on the FTP server, regardless of
how userlist_enable (page 597) is set and regardless of the users listed in vsftpd.user_list. The default file lists
root, bin, daemon, and others.
/etc/vsftpd.user_list . Lists either the only users who can log in on the server or the users who are not allowed
to log in on the server. The userlist_enable (page 597) option must be set to YES for vsftpd to examine the list
of users in this file. Setting userlist_enable to YES and userlist_deny (page 597) to YES (or not setting it) prevents
listed users from logging in on the server. Setting userlist_enable to YES and userlist_deny to NO permits only
the listed users to log in on the server
/etc/vsftpd.chroot_list . Depending on the chroot_list_enable (page 599) and chroot_local_user (page 599)
settings, this file lists either users who are forced into a chroot jail in their home directories or users who are not
placed in a chroot jail.
/var/log/vsftpd.log .
Log file. For more information, refer to “Logs” on page 603.
Chapter Summary
FTP, a protocol for downloading files from and uploading files to another system over a network, is the name of
both a client/server protocol (FTP) and a client utility (ftp) that invokes the protocol. FTP is not a secure protocol
and should be used only to download public information. You can run the vsftpd FTP server in the restricted environment of a chroot jail to make it significantly less likely that a malicious user can compromise the system.
There are many servers and clients that implement the FTP protocol. The ftp utility is the original client implementation; sftp is a secure implementation that uses the facilities of OpenSSH to encrypt the connection. The vsftpd
daemon is a secure FTP server; it is secure in that it better protects the server from malicious users than do other
FTP servers.
Public FTP servers allow you to log in as anonymous or ftp. Convention has you supply your email address as a
password when you log in as an anonymous user. Public servers frequently have interesting files in the pub directory.
FTP provides two modes of transferring files: binary and ASCII. It is safe to use binary mode to transfer all types
of files, including ASCII files. If you transfer a binary file using ASCII mode, the transfer will not be successful.
Exercises
1. What changes does FTP make to an ASCII file when you download it in ASCII mode to a Windows machine
from a Linux server? What changes are made when you download the file to a Mac?
323
Draft
Advanced Exercises
Draft
2. What happens if you transfer an executable program file in ASCII mode?
3. When would ftp be better than sftp?
4. How would you prevent local users logging in on the vsftpd server using their system username and password?
5. What advantage does sftp have over ftp?
6. What is the difference between cd and lcd in ftp?
Advanced Exercises
7. Why might you have problems connecting to an FTP server in PORT mode?
8. Why is it advantageous to run vsftp in a chroot jail?
9. After downloading a file, you find that it does not match the MD5 checksum provided. Downloading it again
gives the same incorrect checksum. What have you done wrong and how would you fix it?
10. How would you configure vsftpd to run through xinetd, and what would be the main advantage of doing this?
324
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 12 - Files, Directories, and Filesystems
Files, Directories, and Filesystems
IN THIS CHAPTER
Important Files and Directories
Ordinary Files, Directories, Links, and Inodes
Special Files
Filesystems
mount: Mounts a Filesystem
fstab: Keeps Track of Filesystems
fsck: Checks Filesystem Integrity
Filesystems hold directories of files. These structures store user data and system data that are the basis of users’
work on the system and the system’s existence. This chapter discusses important files and directories, various types
of files and how to work with them, and the use and maintenance of filesystems.
Important Files and Directories
This section details the most common files used to administer the system. Also refer to “Important Standard Directories and Files” on page 167.
~/.bash_profile . Contains an individual user login shell initialization script. The shell executes the commands
in this file in the same environment as the shell each time a user logs in. The file must be located in a user’s home
directory.
The default Red Hat .bash_profile file executes the commands in ~/.bashrc. You can use .bash_profile to specify
a terminal type (for vi, terminal emulators, and other programs), run stty to establish the terminal characteristics
a user desires, and perform other housekeeping functions when a user logs in.
A typical .bash_profile file specifying a vt100 terminal and CONTROL-H as the erase key follows:
$ cat .bash_profile
export TERM=vt100
stty erase '^h'
For more information, refer to “Startup Files” on page 272.
~/.bashrc . Contains an individual user, interactive, nonlogin shell initialization script. The shell executes the
commands in this file in the same environment as the (new) shell each time a user creates a new interactive shell.
325
Draft
Important Files and Directories
Draft
The .bashrc script differs from .bash_profile in that it is executed each time a new shell is spawned, not only
when a user logs in. For more information, refer to “Startup Files” on page 272.
/dev/null . Output sent to this file disappears; also called a bit bucket. The /dev/null file is a device file and
must be created with mknod. Input that you redirect to come from this file appears as nulls, creating an empty file.
You can create an empty file named nothing by giving the following command:
$ cat /dev/null > nothing
or
$ cp /dev/null nothing
or, without explicitly using /dev/null
$ > nothing
This last command redirects the output of a null command to the file with the same result as the previous commands.
You can use this technique to truncate an existing file to zero length without changing its permissions. You can
also use /dev/null to get rid of output that you do not want:
$ grep portable * 2>/dev/null
The preceding command looks for the word portable in all files in the working directory. Any output to standard
error (page 260), such as permission or directory errors, is discarded, while output to standard output appears on
the screen.
/dev/pts . The /dev/pts pseudofilesystem is a hook into the Linux kernel and is part of the pseudoterminal
support. Pseudoterminals are used by remote login programs, such as ssh, telnet, and xterm and other graphical
terminal emulators. The following sequence of commands demonstrates that the user is logged in on /dev/pts/1.
After using who am i to verify the line the user is logged in on and using ls to show that that line exists, the user
redirects the output of an echo command to /dev/pts/1, whereupon the output appears on the user’s screen:
$ who am i
bravo.example.com!alex
pts/1
$ ls /dev/pts
0 1 2
$ echo Hi there > /dev/pts/1
Hi there
May 10 13:03
/dev/zero . Input you take from this file contains an infinite string of zeros (numerical zeros, not ASCII zeros).
You can fill a file (such as a swap file, page 435) or overwrite a file with zeros with a command such as the following.
The od utility shows the contents of the new file:
$ dd if=/dev/zero of=zeros bs=1024 count=10
10+0 records in
10+0 records out
$ ls -l zeros
-rw-rw-r-1 alex
alex
10240 Dec 3 20:26 zeros
$ od -c zeros
326
Draft
Important Files and Directories
0000000
*
0024000
\0
\0
\0
\0
\0
\0
\0
\0
\0
\0
\0
\0
\0
Draft
\0
\0
\0
When you try to do with /dev/zero what you can do with /dev/null, you quickly fill the partition you are working
in:
$ cp /dev/zero bigzero
cp: writing 'bigzero': No space left on device
$ rm bigzero
/etc/aliases . Used by the mail delivery system (typically sendmail) to hold aliases for users. Edit this file to suit
local needs. For more information, refer to “/etc/aliases” on page 614.
/etc/at.allow, /etc/at.deny, /etc/cron.allow, and /etc/cron.deny . By default, users can use the at and cron
utilities. The at.allow file lists the users who are allowed to use at. The cron.allow file works in the same manner
for cron. The at.deny and cron.deny files specify users who are not permitted to use the corresponding utilities.
As Red Hat is configured, an empty at.deny file and the absence of an at.allow file allows anyone to use at; the
absence of cron.allow and cron.deny files allows anyone to use cron. To prevent anyone except Superuser from
using at, remove the at.allow and at.deny files. To prevent anyone except Superuser from using cron, create a
cron.allow file with the single entry root. Refer to “Scheduling Tasks” on page 531.
/etc/dumpdates . Contains information about the last execution of dump. For each filesystem, it stores the time
of the last dump at a given dump level. The dump utility uses this information to determine which files to back up
when executing at a particular dump level. Refer to “Backing Up Files” on page 524 and the dump man page for
more information. Following is a sample /etc/dumpdates file from a system with four file-systems and a backup
schedule that uses three dump levels:
/dev/hda1
/dev/hda8
/dev/hda9
/dev/hda10
/dev/hda1
/dev/hda1
/dev/hda8
/dev/hda9
/dev/hda10
5
2
2
2
2
0
0
0
0
Thu
Sun
Sun
Sun
Sun
Sun
Sun
Sun
Sun
Apr
Apr
Apr
Apr
Apr
Mar
Mar
Mar
Mar
23
19
19
19
19
22
22
22
22
03:53:55
08:25:24
08:57:32
08:58:06
09:02:27
22:08:35
22:33:40
22:35:22
22:43:45
2004
2004
2004
2004
2004
2004
2004
2004
2004
The first column contains the device name of the dumped filesystem. The second column contains the dump level
and the date of the dump.
/etc/fstab .
filesystem (mount) table . Contains a list of all mountable devices as specified by the system administrator.
Programs do not write to this file but only read from it. Refer to “fstab: Keeps Track of Filesystems” on page 445.
/etc/group . Groups allow users to share files or programs without allowing all system users access to them.
This scheme is useful if several users are working with files that are not public. The /etc/group file associates one
or more user names with each group (number).
An entry in the /etc/group file has four fields in the following format:
group-name:password:group-ID:login-name-list
327
Draft
Important Files and Directories
Draft
The group-name is the name of the group. The password is an optional encrypted password. This field is rarely
used and frequently contains an x, indicating that group passwords are not used. The group-ID is a number, with
1–499 reserved for system accounts. The login-name-list is a comma-separated list of users who belong to the group.
If an entry is too long to fit on one line, end the line with a backslash (\), which quotes the following RETURN,
and continue the entry on the next line. A sample entry from a group file follows. The group is named pubs, has
no password, and has a group ID of 503:
pubs:x:503:alex,jenny,scott,hls,barbara
Each user has a primary group, which is the group that user is assigned in the /etc/passwd file. By default, Red
Hat Linux has user private groups: Each user’s primary group has the same name as the user. In addition, a user
can belong to other groups, depending on which login-name-lists the user appears on in the /etc/group file. In effect,
you simultaneously belong to both your primary group and any groups you are assigned to in /etc/group. When
you attempt to access a file you do not own, the operating system checks whether you are a member of the group
that has access to the file. If you are, your access permissions are controlled by the group access permissions for
the file. If you are not a member of the group that has access to the file and you do not own the file, you are subject
to the public access permissions for the file.
When you create a new file, it is assigned to the group associated with the directory the file is being written into,
assuming that you belong to the group. If you do not belong to the group that has access to the directory, the file
is assigned to your primary group.
Refer to page 523 for information on using system-config-users to work with groups.
/etc/hosts . The /etc/hosts file stores the name, IP address, and optional alias of the other systems that your
system knows about. At the very least, the file must have the hostname and IP address that you have chosen for
your local system and a special entry for localhost. This entry is for the loopback service, which allows the local
system to talk to itself (for example, for RPC services). The IP address of the loopback service is always 127.0.0.1.
Following is a simple /etc/hosts file:
$ cat /etc/hosts
127.0.0.1
localhost
192.168.0.1
gateway
192.168.0.2
mp3server
192.168.0.3
workstation
192.168.0.4
windowsbox
...
If you are not using NIS or DNS to look up hostnames (called hostname resolution), you must include in /etc/hosts
all the systems that you want the local system to be able to contact. The order in which hostname resolution services
are checked is controlled by the hosts entry in the /etc/nsswitch.conf file (page 413).
/etc/inittab .
initialization table .
Controls how the init process behaves. Each line in inittab has four colon-separated fields:
id : runlevel:action:process
The id uniquely identifies an entry in the inittab file. The runlevel is the system runlevel(s) at which process is executed. The runlevel (s) are zero or more characters chosen from 0123456S. If more than one runlevel is listed,
the associated process is executed at each of the specified runlevels. When you do not specify a runlevel, init executes
process at all runlevels. When the system changes runlevels, the processes specified by all entries in inittab that do
328
Draft
Important Files and Directories
Draft
not include the new runlevel are sent the SIGTERM signal to allow them to terminate gracefully. After 5 seconds,
these processes are killed with SIGKILL if they are still running. The process is any bash command line.
The action is one of the following keywords: respawn, wait, once, boot, bootwait, ondemand, powerfail,
powerwait, powerokwait, powerfailnow, ctrlaltdel, kbre-quest, off, ondemand, initdefault, or sysinit. This
keyword controls how the process is treated when it is executed. The most commonly used keywords are wait and
respawn.
The wait keyword instructs init to start process and wait for it to terminate. All subsequent scans of inittab ignore
this wait entry. Because a wait entry is started only once (on entering runlevel) and is not executed again while the
system remains at runlevel, it is often used to redirect init output to the console.
The respawn entry tells init to start process if it does not exist but not to wait for it to terminate. If process does
exist, init goes on to the next entry in the inittab. The init utility continues to rescan inittab, looking for processes
that have died. When a process dies, a respawn entry causes init to restart it.
The initdefault entry tells init what runlevel to bring the system to (see Table 11-1 on page 381). Without this information, init prompts for a runlevel on the system console. The value of the initdefault entry is set when you
configure the system or when you edit inittab directly.
Each virtual console (page 103) has in inittab a mingetty entry that includes a unique terminal identifier (such as
tty1, which is short for /dev/tty1). Add or remove mingetty lines to add or remove virtual consoles. Remember
to leave a virtual console for each X window that you want to run. Following is the mingetty entry for /dev/tty2:
Caution: Use Caution when you Edit inittab
Be careful when you edit inittab by hand. Always make a backup copy in the same directory before
you edit the file. If you make a mistake, you may not be able to boot the system. If you cannot boot
the system, refer to “Emergency Mode” on page 386.
2:2345:respawn:/sbin/mingetty tty2
The id on a mingetty line corresponds to the tty number.
All the action s are documented in the inittab man page. For more information, refer to “Booting the System” on
page 381.
/etc/motd . Contains the message of the day, which can be displayed each time someone logs in using a textual
login. This file typically contains site policy and legal information. Keep this file short because users tend to see the
message many times.
/etc/mtab . When you call mount without any arguments, it consults this file and displays a list of mounted
devices. Each time you (or an init script) call mount or umount, these utilities make the necessary changes to mtab.
Although this is an ASCII text file, you should not edit it. See also /etc/fstab.
Tip: Fixing mtab
The operating system maintains its own internal mount table, which may occasionally differ from this
file. Use cat to display the contents of /proc/mounts to see the internal mount table. To bring the
mtab file in line with the operating system’s mount table, you can reboot the system or replace
/etc/mtab with a symbolic link to /proc/mounts (some information will be lost). Refer to “ mount:
Mounts a Filesystem” on page 442 for more information.
329
Draft
Important Files and Directories
Draft
/etc/netgroup . Defines netgroups, which are used for checking permissions when performing remote logins
and remote mounts and when starting remote shells.
/etc/nsswitch.conf . Specifies whether a system uses as the source of certain information NIS, DNS, local files,
or a combination, and in what order (page 413).
/etc/pam.d .
Files in this directory specify the authentication methods used by PAM (page 416) applications.
Caution: Be Cautious Changing PAM Files
Unless you understand how to configure PAM, avoid changing the files in /etc/pam.d. Mistakes in
the configuration of PAM can make the system unusable.
/etc/passwd . Describes users to the system. Do not edit this file directly, but instead, use one of the utilities
discussed in “Configuring User and Group Accounts” on page 521. Each line in passwd has seven colon-separated
fields that describe one user:
login-name:dummy-password:user-ID:group-ID:info:directory:program
The login-name is the user’s login name—the name you enter in response to the login: prompt or GUI login screen.
The value of the dummy-password is the character x. An encrypted/hashed password is stored in /etc/shadow
(page 433). For security, every account should have a password. By convention, disabled accounts have an asterisk
(*) in this field.
The user-ID is a number, with 0 indicating Superuser and 1–499 reserved for system accounts. The group-ID
identifies the user as a member of a group. It is a number, with 0–499 reserved for system accounts; see /etc/group.
You can change these values and set maximum values in /etc/login.defs.
The info is information that various programs, such as accounting programs and email, use to identify the user
further. Normally it contains at least the first and last name of the user. It is referred to as the GECOS (page 972)
field.
The directory is the absolute pathname of the user’s home directory. The program is the program that runs once
the user logs in. If program is not present, /bin/bash is assumed. You can put /bin/tcsh here to log in using the
TC Shell or /bin/zsh to log in using the Z Shell. The chsh utility (page 396) changes this value.
The program is usually a shell, but it can be any program. The following line in the passwd file creates a “user”
whose only purpose is to execute the who utility:
who:x:1000:1000:execute who:/usr:/usr/bin/who
Using who as a login name causes the system to log you in, execute the who utility, and log you out. The output
of who flashes by in a hurry as the new login prompt clears the screen immediately after who finishes running. This
entry in the passwd file does not provide a shell; there is no way for you to stay logged in after who finishes executing.
This technique is useful for providing special accounts that may do only one thing. For instance, sites may create
an FTP (page 583) account in order to enable anonymous FTP access to their systems. Because no one logs in on
this account, set the shell to /bin/false (which returns a false exit status) or to /sbin/nologin (which does not
permit the user to log in). When you put a message in /etc/nologin.txt, nologin displays that message (except it
has the same problem as the output of who: It is removed so quickly that you cannot see it).
330
Draft
Important Files and Directories
Draft
Security: Do Not Replace a Login Shell with a Shell Script
Do not use shell scripts as replacements for shells in /etc/passwd. A user may be able to interrupt a
shell script, giving him or her full shell access when you did not intend to do so. When installing a
dummy shell, use a compiled program, not a shell script.
/etc/printcap .
UNIX.
The printer capability database. This file describes system printers and is derived from 4.3BSD
/etc/profile . Contains a systemwide interactive shell initialization script for environment and start-up programs.
When you log in, the first thing the shell does is to execute the commands in this file in the same environment as
the shell. (For more information on executing a shell script in this manner, refer to the discussion of the . [dot]
command on page 286.) This file allows the system administrator to establish systemwide environment parameters
that individual users can override. Using this file, you can set shell variables, execute utilities, and take care of
other housekeeping tasks. See also “~/.bash_profile” on page 425.
Following is an example of a /etc/profile file that displays the message of the day (the /etc/motd file), sets the filecreation mask (umask), and sets the interrupt character to CONTROL-C:
# cat /etc/profile
cat /etc/motd
umask 022
stty intr '^c'
See the /etc/profile file on your system for a more complex example.
/etc/protocols .
Do not modify.
Provides protocol numbers, aliases, and brief definitions for DARPA Internet TCP/IP protocols.
/etc/rc.d . Holds the system init scripts, also called run command (rc) scripts. The init program executes several
init scripts each time it changes state or runlevel. The /etc/rc.d/init.d directory holds all the scripts. Each runlevel
has a dedicated directory within /etc/rc.d. For example, runlevel 3 has the /etc/rc.d/rc3.d directory. The files in
each of these directories are links to the files in init.d. The init scripts perform such tasks as mounting filesystems
(when the system goes multiuser), removing temporary files after the filesystems are mounted, and unmounting
filesystems when the system is returned to single-user mode or brought down. For more information, refer to “Init
Scripts: Start and Stop System Services” on page 381.
/etc/resolv.conf .
The resolver (page 701) configuration file, used to provide access to DNS.
The following example shows the resolv.conf file for the example.com domain. A resolv.conf file usually has at
least two lines: a domain line and a nameserver line:
# cat /etc/resolv.conf
domain example.com
nameserver 10.0.0.50
nameserver 10.0.0.51
The first line (optional) specifies the domain name. A resolv.conf file may use search in place of domain: In the
simple case, the two perform the same function. In either case, this domain name is appended to all hostnames
that are not fully qualified. See FQDN on page 972.
The domain keyword takes a single domain name as an argument: This name is appended to all DNS queries,
shortening the time to query hosts at your site. When you put domain example.com in resolv.conf, any reference
331
Draft
Important Files and Directories
Draft
to a host within the example.com domain or a subdomain (such as marketing.example.com) can use the abbreviated form of the host: Instead of ping speedy.marketing.example.com, you can use ping speedy.marketing.
This search keyword is similar to domain but can contain multiple domain names. The domains are searched in
order in the process of resolving a hostname. The following line in resolv.conf causes the marketing subdomain
to be searched first, followed by sales, and finally the whole example.com domain:
search marketing.example.com sales.example.com example.com
Put the most frequently used domain names first to try to outguess possible conflicts. If both speedy.marketing.example.com and speedy.example.com exist, the order of the search determines which one you get when you invoke
DNS. Do not overuse this feature. The longer the search path (three or four names is typically enough), the more
network DNS requests are generated, and the slower the response is.
The nameserver line(s) indicate which systems the local system should query to resolve hostnames to IP addresses
and vice versa. These machines are consulted in the order they appear with a 10-second timeout between queries.
The preceding file causes this machine to query 10.0.0.50, followed by 10.0.0.51 when the first machine does not
answer within 10 seconds. The resolv.conf file may be automatically updated when a PPP- (Point to Point Protocol)
or DHCP- (Dynamic Host Configuration Protocol) controlled interface is activated. Refer to the resolv.conf and
resolver man pages for more information.
/etc/rpc . Maps RPC services to RPC numbers. The three columns in this file show the name of the server for
the RPC program, the RPC program number, and aliases.
/etc/services . Lists system services. The three columns in this file show the friendly name of the service, the
port number/protocol the service frequently uses, and aliases for the service. This file does not specify which services
are running on the local machine, nor is it used to map services to port numbers. The services file is used internally
to map port numbers to services for display purposes.
/etc/shadow . Contains encrypted or MD5 (page 982) hashed user passwords. Each entry occupies one line
composed of nine fields, separated by colons:
login-name:password:last-mod:min:max:warn:inactive:expire:flag
The login-name is the user’s login name—the name that the user enters in response to the login: prompt or GUI
login screen. The password is an encrypted or hashed password that passwd puts into this file. When setting up
new user accounts manually, run passwd as Superuser to assign a password to a new user.
The last-mod field indicates when the password was last modified. The min is the minimum number of days that
must elapse before the password can be changed; max is the maximum number of days before the password must
be changed. The warn specifies how much advance warning (in days) to give the user before the password expires.
The account will be closed if the number of days between login sessions exceeds the number of days specified in
the inactive field. The account will also be closed as of the date in the expire field. The last field in an entry, flag, is
reserved for future use. You can use the Password Info tab in system-config-users (“Modifying a user” on page 522)
to modify these fields.
The shadow password file should be owned by root and should not be publicly readable or writable, making it
more difficult for someone to break into your system by identifying accounts without passwords or by using specialized programs that try to match hashed passwords.
A number of conventions exist for making special shadow entries. An entry of *LK* or NP in the password field
indicates locked and no password, respectively. No password is different from an empty password, implying that
this is an administrative account that nobody ever logs in on directly. Occasionally, programs will run with the
332
Draft
Important Files and Directories
Draft
privileges of this account for system maintenance functions. These accounts are set up under the principle of least
privilege.
Entries in the shadow file must appear in the same order as in the passwd file. There must be one and only one
shadow entry for each passwd entry.
/etc/sysconfig . A directory containing a hierarchy of system configuration files. For more information, refer
to the /usr/share/doc/initscripts*/sysconfig.txt file.
/proc . The /proc pseudofilesystem provides a window into the Linux kernel. Through /proc you can obtain
information on any process running on your computer, including its current state, memory usage, CPU usage,
terminal, parent, group, and more. You can also use /proc to modify the kernel as with the system-config-proc
utility (page 395). You can extract information directly from the files in /proc. An example follows:
$ sleep 10000 &
[1] 17924
$ cd /proc/17924
$ ls -l
total 0
-r--r--r-1 alex
lrwx-----1 alex
-r-------1 alex
lrwx-----1 alex
dr-x-----2 alex
pr--r--r-1 alex
-rw------1 alex
lrwx-----1 alex
-r--r--r-1 alex
-r--r--r-1 alex
-r--r--r-1 alex
$ cat status
Name:
sleep
State: S (sleeping)
Pid:
17924
PPid:
17909
TracerPid:
0
Uid:
0
0
0
Gid:
0
0
0
FDSize: 256
Groups: 0 1 2 3 4 6 10
VmSize:
1144 kB
VmLck:
0 kB
VmRSS:
420 kB
VmData:
20 kB
...
alex
alex
alex
alex
alex
alex
alex
alex
alex
alex
alex
0 Feb 2 17:13 cmdline
0 Feb 2 17:13 cwd -> /home/alex
0 Feb 2 17:13 environ
0 Feb 2 17:13 exe -> /bin/sleep
0 Feb 2 17:13 fd
0 Feb 2 17:13 maps
0 Feb 2 17:13 mem
0 Feb 2 17:13 root -> /
0 Feb 2 17:13 stat
0 Feb 2 17:13 statm
0 Feb 2 17:13 status
0
0
In this example, bash creates a background process (PID 17924) for sleep. Next, the user changes directories to the
directory in /proc that has the same name as the PID of the subject background process (cd /proc/17924). This
directory holds information about the process for which it is named. In this case, it holds information about the
sleep process. The ls –l command shows that some of the entries in this directory are links (cwd is a link to the
directory the process was started from, and exe is a link to the executable file that this process is running), and
some appear to be regular files. All appear to be empty. When you cat one of these pseudofiles (status in the example), you get the output shown. Obviously this is not a regular file.
333
Draft
/sbin/shutdown .
Important Files and Directories
Draft
A utility that brings the system down (see page 390).
swap . Even though swap is not a file, swap space can be added and deleted from the system dynamically. Swap
space is used by the virtual memory subsystem. When it runs low on real memory (RAM), the system writes memory
pages from RAM to the swap space on the disk. Which pages are written and when they are written are controlled
by finely tuned algorithms in the Linux kernel. When needed by running programs, these pages are brought back
into RAM. This technique is called paging (page 987). When a system is running very short on memory, an entire
process may be paged out to disk.
Running an application that requires a large amount of virtual memory may result in the need for additional swap
space. If you run out of swap space, you can use mkswap to create a new swap file and swapon to enable it. Normally,
you use a disk partition as swap space, but you can also use a file.
If you are using a file as swap space, first use df to make sure that you have enough space in the partition for the
file. In the following sequence of commands, the administrator first uses dd and /dev/zero (page 426) to create an
empty file (do not use cp as you may create a file with holes, which may not work) in the working directory. Next,
mkswap takes an argument of the name of the file created in the first step to set up the swap space. For security,
change the file so that it cannot be read from or written to by anyone but root. Use swapon with the same argument
to turn the swap file on, and then use swapon –s to confirm that the swap space is available. The final two commands
turn off the swap file and remove it:
# dd if=/dev/zero of=swapfile bs=1024 count=65536
65536+0 records in
65536+0 records out
# mkswap swapfile
Setting up swapspace version 1, size = 67104768 bytes
# chmod 600 swapfile
# swapon swapfile
# swapon -s
Filename
Type
Size
Used
/dev/hda12
partition
265032
38216
/var/swapfile
file
65528
0
...
# swapoff swapfile
# rm swapfile
Priority
-1
-2
/usr/share/magic . Most files begin with a unique identifier called a magic number. This file is a text database
listing all known magic numbers on the system. When you use the file utility, it consults /usr/share/magic to determine the type of a file. Occasionally, you will acquire a new tool that creates a new type of file that is unrecognized
by the file utility: You need to update the /usr/share/magic file; refer to the magic man page for details. See also
“magic number” on page 982.
/var/log . Holds system log files. You can use system-logviewer (FEDORA) or redhat-logviewer (RHEL) to view
log messages.
/var/log/messages . Contains messages from daemons, the Linux kernel, and security programs. For example,
you will find filesystem full warning messages, error messages from system daemons (NFS, syslog, printer daemons),
SCSI and IDE disk error messages, messages from such security-related programs as su, and more in messages.
Check /var/log/messages periodically to keep informed about important system events. Much of the information
displayed on the system console is also sent to messages. If the system has a problem and you do not have access
to the console, check this file for messages about the problem.
334
Draft
File Types
Draft
File Types
Linux supports many types of files. The following sections discuss these types of files:
•
Ordinary Files, Directories, Links, and Inodes (following)
•
Symbolic links (page 437)
•
Special Files (page 437)
•
FIFO Special File (Named Pipe) (page 438)
•
Sockets (page 439)
•
Block and Character Devices (page 439)
•
Raw Devices (page 440)
Ordinary Files, Directories, Links, and Inodes
Ordinary and directory files . An ordinary file stores user data, such as textual information, programs, or an
image, such as a jpeg or tiff file. A directory is a standard-format disk file that stores information, including names,
about ordinary files and other directory files.
Inodes . An inode is a data structure (page 966), stored on disk, that defines a file’s existence and is identified
by an inode number. A directory relates each of the filenames it stores to a specific inode. An inode contains critical
information, such as the name of the owner of the file, where it is physically located on the disk, and how many
hard links point to it. In addition, SELinux (page 379) stores extended information about files in inodes.
When you move (mv) a file within a filesystem, you change the filename portion of the directory entry associated
with the inode that describes the file. You do not create a new inode. If you move a file to another filesystem, mv
first creates a new inode on the destination filesystem and then deletes the original inode. You can also use mv to
move a directory recursively, in which case all the files are copied and deleted.
When you make an additional hard link (ln, page 178) to a file, you create another reference (an additional filename)
to the inode that describes the file. You do not create a new inode.
When you remove (rm) a file, you delete the directory entry that describes the file. When you remove the last hard
link to a file, the operating system puts all the blocks the inode pointed to back in the free list (the list of blocks
that are available for use on the disk) and frees the inode to be used again.
The . and .. directory entries . Every directory has at least two entries (. and ..). The . entry is a link to the
directory itself. The .. entry is a link to the parent directory. In the case of the root directory, there is no parent,
and the .. entry is a link to the root directory itself. It is not possible to create hard links to directories.
Symbolic links . Because each filesystem has a separate set of inodes, you can create hard links to a file only
from within the filesystem that holds that file. To get around this limitation, Linux provides symbolic links, which
are files that point to other files. Files that are linked by a symbolic link do not share an inode: You can create a
symbolic link to a file from any filesystem. You can also create a symbolic link to a directory, device, or other special
file. For more information, refer to “Symbolic Link” on page 180.
335
Draft
Special Files
Draft
Special Files
Special files represent Linux kernel routines that provide access to an operating system feature. FIFO (first in, first
out) special files allow unrelated programs to exchange information. Sockets allow unrelated processes on the same
or different computers to exchange information. One type of socket, the UNIX domain socket, is a special file.
Symbolic links are another type of special file.
Device files . Device files, which include block and character special files, represent device drivers that let you
communicate with peripheral devices, such as terminals, printers, and hard disks. By convention, device files appear
in the /dev directory and its subdirectories. Each device file represents a device: You read from and write to the file
to read from and write to the device it represents.For example, using cat to send an audio file to /dev/dsp plays
the file. The following example shows part of the output an ls –l command produces for the /dev directory:
$ ls -l /dev
crw------crw------crw------crw------brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---brw-rw---...
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
root
sys
root
root
root
floppy
floppy
floppy
floppy
floppy
disk
disk
disk
disk
disk
disk
disk
14,
5,
5,
5,
2,
2,
2,
2,
2,
3,
3,
3,
3,
3,
3,
3,
4
1
64
65
0
12
16
28
12
0
1
2
3
4
5
6
Apr 17 2004 audio
Jan 22 08:31 console
May 5 2004 cua0
May 5 2004 cua1
May 5 2004 fd0
May 5 2004 fd0D360
May 5 2004 fd0D720
May 5 2004 fd0H1440
May 5 2004 fd0H360
May 5 2004 hda
May 5 2004 hda1
May 5 2004 hda2
May 5 2004 hda3
May 5 2004 hda4
May 5 2004 hda5
May 5 2004 hda6
The first character of each line is always –, b, c, d, l, or p, representing ordinary (plain), block, character, directory,
symbolic link, or named pipe (see the following section). The next nine characters represent the permissions for
the file, followed by the number of hard links and the names of the owner and group. Where the number of bytes
in a file would appear for an ordinary or directory file, a device file shows its major and minor device numbers (page
439) separated by a comma. The rest of the line is the same as any other ls –l listing (page 173).
FIFO Special File (Named Pipe)
A FIFO special file, also called a named pipe, represents a pipe: You read from and write to the file to read from and
write to the pipe. The term FIFO stands for first in, first out—the way any pipe works. The first information that
you put in one end is the first information that comes out the other end. When you use a pipe on a command line
to send the output of a program to the printer, the printer prints the information in the same order that the program
produced it and sent it into the pipe.
Unless you are writing sophisticated programs, you will not be working with FIFO special files. However, programs
that you use on Linux use named pipes for interprocess communication. You can create a pipe using mkfifo:
# mkfifo AA
# ls -l AA
prw-rw-r--
336
1
root
root
0 Apr 26 13:11 AA
Draft
Special Files
Draft
The p at the left end of the output of ls –l indicates that the file is a pipe.
The UNIX and Linux systems have had pipes for many generations. Without named pipes, only processes that
were children of the same ancestor could use pipes to exchange information. Using named pipes, any two processes
on a single system can exchange information. One program writes to a FIFO special file. Another program reads
from the same file. The programs do not have to run at the same time or be aware of each other’s activity. The
operating system handles all buffering and information storage. The term asynchronous (async) applies to this type
of communication because programs on the ends of the pipe do not have to be synchronized.
Sockets
Like a FIFO special file, a socket allows asynchronous processes that are not children of the same ancestor to exchange
information. Sockets are the central mechanism of the interprocess communication that is the basis of the networking
facility. When you use networking utilities, pairs of cooperating sockets manage the communication between the
processes on the local computer and the remote computer. Sockets form the basis of such utilities as ssh and scp.
Major and Minor Device Numbers
A major device number represents a class of hardware devices: terminal, printer, tape drive, hard disk, and so on.
In the list of the /dev directory on page 438, all the hard disk partitions have a major device number of 3.
A minor device number represents a particular piece of hardware within a class. Although all the hard disk partitions
are grouped together by their major device number, each has a different minor device number (hda1 is 1, hda2 is
2, and so on). This setup allows one piece of software (the device driver) to service all similar hardware and to be
able to distinguish among different physical units.
Block and Character Devices
This section describes typical device drivers. Because device drivers can be changed to suit a particular purpose,
the descriptions in this section do not pertain to every system.
A block device is an I/O (input/output) device that is characterized by
•
Being able to perform random access reads.
•
Having a specific block size.
•
Handling only single blocks of data at a time.
•
Accepting only transactions that involve whole blocks of data.
•
Being able to have a filesystem mounted on it.
•
Having the Linux kernel buffer its input and output.
•
Appearing to the operating system as a series of blocks numbered from 0 through n – 1, where n is the number
of blocks on the device.
The common block devices on a Linux system are hard disks, floppy diskettes, and CD-ROMs.
A character device is any device that is not a block device. Some examples of character devices are printers, terminals,
tape drives, and modems.
The device driver for a character device determines how a program reads from and writes to the device. For example,
the device driver for a terminal allows a program to read the information you type on the terminal in two ways. A
337
Draft
Filesystems
Draft
program can read single characters from a terminal in raw mode (that is, without the driver doing any interpretation
of the characters). This mode has nothing to do with the raw device described in the following section. Alternatively,
a program can read a line at a time. When a program reads a line at a time, the driver handles the erase and kill
characters so the program never sees corrected typing mistakes. In this case, the program reads everything from
the beginning of a line to the RETURN that ends a line; the number of characters in a line can vary.
Raw Devices
Device driver programs for block devices usually have two entry points so they can be used in two ways: as block
devices or as character devices. The character device form of a block device is called a raw device. A raw device is
characterized by
•
Direct I/O (no buffering through the Linux kernel).
•
A one-to-one correspondence between system calls and hardware requests.
•
Device-dependent restrictions on I/O.
An example of a utility that uses a raw device is fsck. It is more efficient for fsck to operate on the disk as a raw
device, not restricted by the fixed size of blocks in the block device interface. Because it has full knowledge of the
underlying filesystem structure, fsck can operate on the raw device using the largest possible units. When a
filesystem is mounted, processes normally access the disk through the block device interface, which explains why
it is important to allow fsck to modify only an unmounted filesystem. On a mounted filesystem, there is the danger
that, while fsck is rearranging the underlying structure through the raw device, another process would change a
disk block using the block device, resulting in a corrupted filesystem.
Filesystems
Table 12-1 lists some of the types of filesystems available under Linux.
338
Draft
Filesystems
Draft
Table 12-1. Filesystems
adfs
The Acorn Disc Filing System.
affs
The Amiga Fast Filesystem(FFS).
autofs
Automounting filesystem (page 671).
coda
The CODA distributed filesystem (developed at Carnegie
Mellon).
devpts
A pseudofilesystem for pseudoterminals (page 426).
ext2
A standard filesystem for Red Hat systems, usually with
the ext3 extension.
ext3
A journaling (page 978) extension to the ext2 filesystem;
greatly improves recovery time from crashes (it takes a
lot less time to run fsck), promoting increased availability.
As with any filesystem, a journaling filesystem can lose
data during a system crash or hardware failure.
hfs
Hierarchical Filesystem: used by older Macintoshes.
Newer Macs use hfs+.
hpfs
High Performance Filesystem: the native filesystem for
IBM’s OS/2.
iso9660
The standard filesystem for CDs.
minix
Very similar to Linux, the filesystem of a small operating
system that was written for educational purposes by Prof.
Andrew S. Tanenbaum (www.cs.vu.nl/~ast/minix.html
[http://www.cs.vu.nl/~ast/minix.html]).
msdos
The filesystem used by DOS and subsequent Microsoft
operating systems. Do not use msdos for mounting
Windows filesystems; it does not read vfat attributes.
ncpfs
Novell NetWare NCP Protocol Filesystem: used to mount
remote filesystems under NetWare.
nfs
Network Filesystem: Developed by Sun Microsystems,
a protocol that allows a computer to access remote files
over a network as if they were local (page 655).
ntfs
NT Filesystem: the native filesystem of Windows NT.
Still very experimental; use caution.
proc
An interface to several Linux kernel data structures (page
966) that behaves like a filesystem (page 434).
qnx4
The QNX 4 Operating System filesystem.
reiserfs
A journaling (page 978) filesystem, based on balancedtree algorithms. See ext3 for more on journaling filesystems.
romfs
A dumb, readonly filesystem used mainly for RAM disks
(page 991) during installation.
software RAID
RAID implemented in software. Refer to “RAID Filesystem” on page 449.
smbfs
Samba Filesystem (page 675).
sysv
System V UNIX filesystem.
339
Draft
mount: Mounts a Filesystem
Draft
ufs
Default filesystem under Sun’s Solaris Operating System
and other UNIXs.
umsdos
A full-feature UNIX-like filesystem that runs on top of a
DOS FAT filesystem.
vfat
Developed by Microsoft, a standard that allows long filenames on FAT partitions.
xfs
SGI’s journaled filesystem (ported from Irix)
mount: Mounts a Filesystem
The mount utility connects directory hierarchies, typically filesystems, to the Linux directory hierarchy. These directory hierarchies can be on remote and local disks, CDs, and floppy diskettes. Linux also allows you to mount
virtual filesystems that have been built inside regular files, filesystems built for other operating systems, and the
special /proc filesystem (page 434), which maps useful Linux kernel information to a pseudodirectory.
The mount point for the filesystem/directory hierarchy that you are mounting is a directory in the local filesystem.
The directory must exist before mounting; its contents disappear as long as a filesystem is mounted on it and reappear
when you unmount the filesystem.
Without any arguments, mount lists the currently mounted filesystems, showing the physical device holding each
filesystem, the mount point, the type of filesystem, and any options set when each filesystem was mounted:
$ mount
none on /proc type proc (rw)
/dev/hdb1 on / type ext2 (rw)
/dev/hdb4 on /tmp type ext2 (rw)
/dev/hda5 on /usr type ext3 (rw)
/dev/sda1 on /usr/X386 type ext2 (rw)
/dev/sda3 on /usr/local type ext2 (rw)
/dev/hdb3 on /home type ext3 (rw)
/dev/hda1 on /dos type msdos (rw,umask=000)
tuna:/p04 on /p04 type nfs (rw,addr=192.168.0.8)
/dev/scd0 on /mnt/cdrom type iso9660 (ro,noexec,nosuid,nodev)
The mount utility gets this information from the /etc/mtab file (page 430). This section covers mounting local
filesystems; refer to page 655 for information on using NFS to mount remote directory hierarchies.
The first entry in the preceding example shows the /proc pseudofilesystem (page 434). The next six entries show
disk partitions holding standard Linux ext2 and ext3 file-systems. Disk partitions are on three disks: two IDE disks
(hda, hdb) and one SCSI disk (sda). Disk partition /dev/hda1 has a DOS (msdos) filesystem mounted at the directory /dos in the Linux filesystem. You can access the DOS files and directories on this partition as if they were
Linux files and directories, using Linux utilities and applications. The line starting with tuna shows a mounted,
remote NFS filesystem. The last line shows a CD mounted on a SCSI CD drive (/dev/scd0).
On occasion the list of files in /etc/mtab may not be synchronized with the partitions that are mounted. You can
remedy this situation by rebooting the system, or you can refer to the contents of the /proc/mounts file, which
may have slightly different information than mtab but is always correct. You can even replace mtab with a symbolic
link to /proc/mounts:
# rm /etc/mtab
# ln -s /proc/mounts /etc/mtab
340
Draft
mount: Mounts a Filesystem
Draft
Caution: Do Not Mount Anything on Root (/)
Always mount network directory hierarchies and removable devices at least one level below the root
level of the filesystem. The root filesystem is mounted on /; you cannot mount two filesystems in the
same place. If you were to try to mount something on /, all the files, directories, and filesystems that
were under root would no longer be available, and the system would crash.
When you add a line for a filesystem to the /etc/fstab file (page 427), you can mount that filesystem by giving the
associated mount point (or the device) as the argument to mount. For example, the SCSI CD listed earlier was
mounted using the following command:
$ mount /mnt/cdrom
This command worked because /etc/fstab contains the additional information needed to mount the file:
/dev/scd0 /mnt/cdrom iso9660 user,noauto,ro
You can also mount filesystems that do not appear in /etc/fstab. For example, when you insert a floppy diskette
that holds a DOS filesystem into the floppy diskette drive, you can mount that filesystem using the following
command:
# mount –t msdos /dev/fd0 /mnt/floppy
The –t msdos specifies a filesystem type of msdos. You can mount DOS filesystems only if you have configured
your Linux kernel (page 507) to accept DOS filesystems. You do not need to mount a DOS filesystem in order to
read from and write to it, such as when you use mcopy (page 130). You do need to mount a DOS filesystem to use
Linux commands (other than Mtools commands) on files on the diskette.
Mount Options
The mount utility takes many options which you can specify on the command line or in the /etc/fstab file (page
445). For a complete list of mount options for local file-systems, see the mount man page; for remote directory
hierarchies, the nfs man page.
The noauto option causes Linux not to mount the filesystem automatically. The nosuid option forces mounted
setuid executables to run with regular permissions (no effective user ID change) on the local system (the system
that mounted the file-system). Always mount removable devices nosuid so a malicious user does not, for example,
put a setuid copy of bash on a disk and have a root shell.
Unless you specify the user, users, or owner option, only Superuser can mount and unmount a filesystem. The
user option means that any user can mount the filesystem, but it must be unmounted by the same user who
mounted it; users means that any user can mount and unmount the filesystem. These options are frequently used
for CD and floppy drives. The owner option, used only under special circumstances, is similar to the user option
except that the user mounting the device must own the device.
Mounting a Linux Floppy Diskette
Mounting a Linux floppy diskette is similar to mounting a partition of a hard disk. Put an entry similar to the following in /etc/fstab for a diskette in the first floppy drive. Specifying a filesystem type of auto causes the system
to probe the filesystem to determine its type and allows users to mount a variety of diskettes.
341
Draft
/dev/fd0
umount: Unmounts a Filesystem
/mnt/floppy
auto
noauto ,users
Draft
0 0
Create the /mnt/floppy directory if necessary. Insert a diskette and try to mount it. The diskette must be formatted
(use fdformat). In the following examples, the error message following the first command usually indicates there
is no filesystem on the diskette: Use mkfs (page 397) to create a filesystem, but be careful, because mkfs destroys
all data on the diskette:
# mount /dev/fd0
mount: wrong fs type, bad option, bad superblock on /dev/fd0,
or too many mounted file systems
# mkfs /dev/fd0
mke2fs 1.35 (28-Feb-2004)
Filesystem label=
OS type: linux
Block size=1024 (log=0)
Fragment size=1024 (log=0)
184 inodes, 1440 blocks
72 blocks (5.00%) reserved for the super user
First data block=1
1 block group
8192 blocks per group, 8192 fragments per group
184 inodes per group
Writing inode tables: done
Writing superblocks and filesystem accounting information: done
This filesystem will be automatically checked every 28 mounts or
180 days, whichever comes first. Use tune2fs -c or -i to override.
Try the mount command again:
# mount /dev/fd0
# mount
.
.
.
/dev/fd0 on /mnt/floppy type ext2 (rw,nosuid,nodev)
# df -h /dev/fd0
Filesystem
/dev/fd0
Size
1.4M
Used Avail Use% Mounted on
13K 1.3M
1% /mnt/floppy
The mount command without any arguments and df /dev/fd0 show the floppy is mounted and ready for use.
umount: Unmounts a Filesystem
The umount utility unmounts a filesystem as long as it does not house any files or directories that are in use (open).
For example, a logged-in user’s working directory must not be on the filesystem you want to unmount. The next
command unmounts the CD shown earlier:
342
Draft
fstab: Keeps Track of Filesystems
Draft
$ umount /mnt/cdrom
Unmount a floppy or a remote directory hierarchy the same way you would unmount a partition of a hard drive.
The umount utility consults /etc/fstab to get the necessary information and then unmounts the appropriate
filesystem from its server. When a process has a file open on the filesystem that you are trying to unmount, umount
displays a message similar to the following:
umount: /home: device is busy
Tip: When You Cannot Unmount a Device Because It Is in Use
When a process has a file open on a device you need to unmount, use fuser to determine which process
has the file opened and to kill it. For example, when you want to unmount the floppy, give the command
fuser –ki /mnt/floppy, which, after checking with you, kills the process using the floppy.
Use the –a option to umount in order to unmount all filesystems, except for the one mounted at /, which can never
be unmounted. You can combine –a with the –t option to unmount filesystems of a given type (ext3, nfs, or others).
For example, use the following command to unmount all mounted nfs directory hierarchies that are not being
used:
# umount -at nfs
fstab: Keeps Track of Filesystems
The system administrator maintains the /etc/fstab file, which lists local and remote directory hierarchies, most of
which the system mounts automatically when it boots. The fstab file has six columns; a dash keeps the place of a
column that has no value. The six columns are
1.
Name The name of the block device (page 439) or remote directory hierarchy. A remote directory hierarchy
appears as hostname : pathname , where hostname is the name of the host that houses the filesystem, and
pathname is the absolute pathname of the directory that is to be mounted. You can substitute the volume label
of a local filesystem by using the form LABEL= xx, where xx is the volume label. Refer to e2label on page 396.
2.
Mount point The name of a directory file that the filesystem/directory hierarchy is to be mounted over. If it
does not exist, create this directory with mkdir.
3.
Type The type of filesystem/directory hierarchy that is to be mounted. Local filesystems are generally ext2
or ext3, and remote directory hierarchies are nfs. See Table 12-1 on page 440 for a list of filesystem types.
4.
Mount options A comma-separated list of mount options, such as whether the filesystem is mounted for
reading and writing (rw, the default) or readonly (ro). Refer to the mount and nfs man pages for lists of options.
5.
Dump Used by dump (page 529) to determine when to back up the filesystem.
6.
Fsck Determines which filesystem fsck should check first. Root (/) should have a 1 in this column, other
filesystems that need to be checked should have a 2, and filesystems/directory hierarchies that do not need to
be checked (for example, remotely mounted directory hierarchies or a CD) should have a 0.
The following example shows a typical fstab file:
343
Draft
fsck: Checks Filesystem Integrity
# cat /etc/fstab
LABEL=/
/
LABEL=/boot
/boot
LABEL=/home
/home
/dev/cdrom
/mnt/cdrom
/dev/hda7
/tmp
/dev/hda6
/usr
/dev/hda11
swap
/dev/fd0
/mnt/floppy
none
/proc
none
/dev/pts
bravo:/home
/bravo_home
kudos:/home/alex /kudos_alex
ext3
ext2
ext3
iso9660
ext2
ext3
swap
ext2
proc
devpts
nfs
nfs
defaults
defaults
defaults
noauto,owner,ro
defaults
defaults
defaults
noauto,owner
defaults
gid=5,mode=620
defaults
defaults
Draft
1
1
1
0
1
1
0
0
0
0
0
0
1
2
2
0
2
2
0
0
0
0
0
0
Tip: Exporting Symbolic Links and Device Files
When you export a symbolic link, make sure the object of the link is available on the client (remote)
system. If the object of the link does not exist on a client system, you must export and mount it along
with the exported link; otherwise, it will not point to the file it points to on the server.
A device file refers to a Linux kernel interface. When you export a device file, you export that interface.
If the client system does not have the same type of device, the exported device will not work.
A mounted filesystem with a mount point within an exported filesystem will not be exported with the exported
filesystem. You need to export each filesystem that you want exported, even if it resides within an already exported
filesystem. When you have two filesystems, /opt/apps and /opt/apps/oracle, residing on two partitions to export,
you must export each explicitly, even though oracle is a subdirectory of apps. Most other subdirectories and files
are exported automatically.
fsck: Checks Filesystem Integrity
The fsck (filesystem check) utility verifies the integrity of filesystems and, if possible, repairs any problems it finds.
Because many filesystem repairs can destroy data, particularly on a nonjournaling filesystem (page 978), such as
ext2, fsck asks you for confirmation, by default, before making each repair.
Caution: Do Not Run fsck on a Mounted Filesystem
Do not run fsck on a mounted filesystem (except / ). When you attempt to check a mounted filesystem,
fsck warns you and asks you whether you want to continue. Reply no. You can run fsck with the –N
option on a mounted filesystem as it will not write to the filesystem, so no harm will come of running
it.
The following command checks all the filesystems that are marked to be checked in /etc/fstab (page 427) except
for the root filesystem:
# fsck -AR
The –A option causes fsck to check filesystems listed in fstab, and –R skips the root filesystem. You can check a
specific filesystem with a command similar to one of the following:
# fsck /home
344
Draft
tune2fs: Changes Filesystem Parameters
Draft
or
# fsck /dev/hda6
Crash flag . The /etc/rc.d/rc.sysinit start-up script looks for two flags in the root directory of each partition
to help determine whether fsck needs to be run on that partition before it is mounted. The .autofsck flag (the crash
flag) indicates that the partition should be checked. By default, the person bringing up the system has 5 seconds
to respond to a prompt with a y, or the check is skipped. The other flag, forcefsck, is also in the root directory of
a partition. When this flag is set, the user is given no choice; fsck is automatically run on the partition. These checks
are in addition to those established by tune2fs (next section). The .autofsck flag is present while the system is
running and is removed when the system is properly shut down. When the system crashes, the flag is present when
the system is brought up. The forcefsck flag is placed on the filesystem when a hard error is on the disk and the
disk must be checked.
tune2fs: Changes Filesystem Parameters
The tune2fs utility displays and modifies filesystem parameters on ext2 filesystems and on ext3 filesystems, as they
are modified ext2 filesystems. This utility can also set up journaling on an ext2 filesystem, turning it into an ext3
filesystem. With more reliable hardware and software, the system is rebooted less frequently and it becomes more
important to check filesystems regularly. By default, fsck is run on each partition while the system is brought up,
before the partition is mounted. (The checks scheduled by tune2fs are separate and scheduled differently from the
checks that are done following a system crash or hard disk error [see the previous section].) Depending on the flags,
fsck may do nothing more than display a message saying that the filesystem is clean. The larger the partition, the
more time it takes to check it, assuming a nonjournaling filesystem. These checks are frequently unnecessary. The
tune2fs utility helps you to find a happy medium between checking filesystems each time you reboot the system
and never checking them. It does this by scheduling when fsck checks a filesystem (these checks occur only when
the system is booted).1 You can use two scheduling patterns: time elapsed since the last check and number of
mounts since the last check. The following command causes /dev/hda6 to be checked when fsck runs after it has
been mounted eight times or after 15 days have elapsed since its last check, whichever happens first:
# tune2fs -c 8 -i 15 /dev/hda6
tune2fs 1.35 (28-Feb-2004)
Setting maximal mount count to 8
Setting interval between check 1296000 seconds
The next tune2fs command is similar except that it works on a different partition and sets the current mount count
to 4. When you do not specify a current mount count, as in the previous example, it is assumed to be zero:
# tune2fs -c 8 -i 15 -C 4 /dev/hda10
tune2fs 1.35 (28-Feb-2004)
Setting maximal mount count to 8
Setting current mount count to 4
Setting interval between check 1296000 seconds
The –l option displays a list of information about the partition. You can combine this option with others. Below
the Maximum mount count is –1, which means that the mount count information is ignored by fsck and the
kernel. A mount count of 0 works the same way:
1
For systems whose purpose in life is to run continuously, this kind of scheduling does not work. You must come up with a schedule that is
not based on system reboots but rather on a clock. Each filesystem must be unmounted periodically, checked with fsck (preceding section),
and remounted.
345
Draft
# tune2fs -l /dev/hda5
tune2fs 1.35 (28-Feb-2004)
Filesystem volume name:
Last mounted on:
Filesystem UUID:
Filesystem magic number:
Filesystem revision #:
Filesystem features:
Filesystem state:
Errors behavior:
Filesystem OS type:
Inode count:
Block count:
...
Last mount time:
Last write time:
Mount count:
Maximum mount count:
Last checked:
Check interval:
RAID Filesystem
Draft
/free1
<not available>
f93d69a8-a419-11d4-944c-e77d13cd6039
0xEF53
1 (dynamic)
has_journal filetype needs_recovery sparse_super
clean
Continue
Linux
513024
1024135
Mon Feb 2 01:54:30 2004
Mon Feb 2 01:54:30 2004
6
-1
Mon Feb 2 01:54:30 2004
0 (<none>)
Set the filesystem parameters on your system so that they are appropriate to the way you use your computer. Using
the –C option to stagger the checks ensures that all the checks do not occur at the same time. Always check new
and upgraded filesystems to make sure that they have checks scheduled as you desire.
To change an ext2 filesystem to an ext3 filesystem, you must put a journal (page 978) on the filesystem, and the
kernel must support ext3 filesystems. Use the –j option to set up a journal on an unmounted filesystem:
# tune2fs -j /dev/hdd3
tune2fs 1.35 (28-Feb-2004)
Creating journal inode: done
This filesystem will be automatically checked every -1 mounts or
0 days, whichever comes first. Use tune2fs -c or -i to override.
Before you can use fstab (page 427) to mount the changed filesystem, you must modify its entry in the fstab file
to reflect its new type. Change the third column to ext3.
The following command changes an unmounted ext3 filesystem to an ext2 filesystem:
# tune2fs -O ^has_journal /dev/hdd3
tune2fs 1.35 (28-Feb-2004)
Refer to the tune2fs man page for more details.
RAID Filesystem
RAID (Redundant Arrays of Inexpensive/Independent Disks) spreads information across several disks to combine
several physical disks into one larger virtual device. RAID improves performance and creates redundancy. There
are more than six types of RAID configurations. Using Red Hat tools, you can set up software RAID. Hardware
RAID requires hardware that is designed to implement RAID and is not covered here.
346
Draft
Chapter Summary
Draft
Caution: Do Not Replace Backups with RAID
Do not use RAID as a replacement for regular backups. If your system undergoes a catastrophic failure,
RAID will be useless. Earthquake, fire, theft, and so on may leave your entire system inaccessible (if
your hard drives are destroyed or missing). RAID does not take care of something as simple as replacing
a file when you delete it by accident. In these cases a backup on removable media (that has been removed) is the only way you will be able to restore a filesystem.
RAID can be an effective addition to a backup. Red Hat has RAID software that you can install when you install
your Red Hat system or as an afterthought. The Linux kernel can automatically detect RAID disk partitions at boot
time if the partition ID is set to 0xfd, which fdisk recognizes as Linux raid autodetect.
The kernel disk code implements software RAID so it is much cheaper than a hardware RAID. Not only does it
avoid specialized RAID disk controllers, but it also works with both the less expensive IDE disks as well as SCSI
disks. For more information refer to the Software-RAID HOWTO.
Chapter Summary
This chapter covers the Linux filesystem, which was introduced in Chapter 6, from a system administrator’s point
of view. It lists many important system files and directories, explaining what each does. The section on file types
explains the difference between ordinary and directory files and the inodes that hold each. It also covers the use of
hard and symbolic links.
Special files provide access to operating system features. The section covering these files details major and minor
device numbers, which the kernel uses to identify classes of devices and specific devices within each class, and
character and block devices, which represent I/O devices such as hard disks and printers.
The section on the filesystem discusses how to mount and unmount a filesystem, how to edit /etc/fstab to mount
filesystems automatically when the system boots, and how to verify or repair a filesystem with fsck.
Exercises
1. What is the function of the /etc/hosts file? What services can you use in place of, or to supplement, the hosts
file?
2. What does the /etc/resolv.conf file do? What does the nameserver line in this file do?
3. What is an inode? What happens to the inode when you move a file within a filesystem?
4. What does the .. entry in a directory point to? What does this entry point to in the root (/) directory?
5. What is a device file? Where can you find device files?
6. What is a FIFO? What does FIFO stand for? What is another name for a FIFO? How does a FIFO work?
Advanced Exercises
7. Write a line for /etc/fstab file that would mount the /dev/hdb1 ext3 filesystem on /extra with the following
characteristics: The filesystem will not be mounted automatically when the system boots, and anyone can mount
and unmount the filesystem.
8. Without using rm, how can you delete a file? (Hint: How do you rename a file?)
347
Draft
Advanced Exercises
Draft
9. After burning an ISO image file named image.iso to a CD on /dev/hdc, how can you can verify the copy from
the command line?
10. Why should /var be on a separate partition from /usr?
11. Suggest a way to ensure that deleted files can not be recovered.
12. How would you mount an ISO image so that you could copy files from it without burning it to a CD?
348
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 22 - NFS: Sharing Filesystems
NFS: Sharing Filesystems
IN THIS CHAPTER
NFS Client
JumpStart: Mounting a Remote Directory Hierarchy
Improving Performance
NFS Server
JumpStart: system-config-nfs: Configures an NFS Server
Exporting a Directory Hierarchy
automount: Mounting Directory Hierarchies Automatically
The NFS (Network Filesystem) protocol, a de facto standard originally developed by Sun Microsystems, allows a
server to share selected local directory hierarchies (page 967) with client systems on a heterogeneous network. NFS
runs on UNIX, DOS, Windows, VMS, Linux, and more. Files on the remote computer (the fileserver) appear as if
they are present on the local system (the client). The physical location of a file is irrelevant to an NFS user.
NFS reduces storage needs and system administration. As an example, each system in a company traditionally
holds its own copy of an application program. In order to upgrade the program, the administrator needs to upgrade
it on each of the systems. NFS allows you to store a copy of a program on one system and give other users access
to it over the network. This scenario uses less storage by reducing the number of locations that the same data needs
to be maintained. In addition to efficiency, NFS gives users on the network access to the same data (not just application programs), improving data consistency and reliability. By consolidating data, NFS reduces administrative
overhead and provides a convenience to users.
Introduction
Figure 22-1 shows the flow of data from a client to a server in a typical NFS client/server setup. An NFS directory
hierarchy appears to users and application programs as just another directory hierarchy. By looking at it, you cannot
tell that a given directory holds a remotely mounted NFS directory hierarchy and not a local ext3 filesystem. The
NFS server translates commands from the client into operations on the server’s filesystem.
349
Draft
Introduction
Draft
Figure 22-1. Flow of data in a typical NFS client/server setup
Diskless . In many computer facilities, user files are stored on a central fileserver equipped with many largecapacity disk drives and devices that quickly and easily make backup copies of the data. A diskless system boots
from a fileserver (netboots, next paragraph), a CD, or a floppy diskette and loads system software from a fileserver.
The Linux Terminal Server Project (LTSP.org) Web site says it all: “Linux makes a great platform for deploying
diskless workstations that boot from a network server. The LTSP is all about running thin client computers in a
Linux environment.” Because a diskless workstation does not require a lot of computing power, you can give older,
retired computers a second life by using them as diskless systems.
Netboot/PXE . You can netboot (page 985) systems that are appropriately set up. Red Hat includes the PXE
(Preboot Execution Environment) server package for netbooting Intel systems. Older systems sometimes use tftp
(trivial file transfer protocol) for netbooting. Non-Intel architectures have historically included netboot capabilities
that Red Hat Linux also supports. You can build the Linux kernel so that it mounts root (/) using NFS. Of the
many ways to set up your system, the one you choose depends on what you want to do. See the Remote-boot mini
HOWTO for more information.
Dataless . Another type of Linux system is a dataless system, in which the client has a disk but stores no user
data (only Linux and the applications are kept on the disk). Setting up this type of system is a matter of choosing
which directory hierarchies are mounted remotely.
df shows where directory hierarchies are mounted . The df utility displays a list of the directory hierarchies
available on your system, along with the amount of disk space, free and used, on each. The –h (human) option
makes the output more intelligible. Directory hierarchy names that are prepended with hostname: are available
through NFS.
[bravo]$ cd;pwd
/speedy.home/jenny
[bravo]$ df -h
Filesystem
Size
/dev/hda1
981M
/dev/hda6
20G
350
Used
287M
2.7G
Avail
645M
16G
Use%
31%
15%
Mounted on
/
/usr
Draft
More Information
/dev/hda7
grape:/gc1
grape:/gc5
speedy:/home
9.7G
985M
3.9G
3.9G
384M
92M
3.0G
2.4G
8.8G
844M
738M
1.4G
5%
10%
81%
64%
Draft
/home
/grape.gc1
/grape.gc5
/speedy.home
In the preceding example, Jenny’s home directory, /home/jenny, is on the remote system speedy. Using NFS, the
/home filesystem on speedy is mounted on bravo; to make it easy to recognize, it is mounted as /speedy.home.
The /gc1 and /gc5 filesystems on grape are mounted on bravo as /grape.gc1 and /grape.gc5.
You can use the –T option to df to add a Type column to the display. The following command uses –t nfs to display
NFS filesystems only:
[grape]$ df -ht nfs
Filesystem
Size
grape:/gc1
985M
grape:/gc5
3.9G
speedy:/home
3.9G
Used
92M
3.0G
2.4G
Avail
844M
738M
1.4G
Use%
10%
81%
64%
Mounted on
/grape.gc1
/grape.gc5
/speedy.home
Errors . At times you may lose access to remote files. A network problem or a remote system crash may make
these files temporarily unavailable: When you try to access a remote file, you get an error message, such as NFS
server speedy not responding. When the local system can contact the remote server again, you see a message,
such as NFS server speedy OK. Setting up a stable network and server (or not using NFS) is the best defense
against these kinds of problems.
Security . NFS is based on the trusted-host paradigm (page 343) and therefore has all the security shortcomings
that plague services based on this paradigm. In addition, NFS is not encrypted. Because of these issues, you should
implement NFS on a single LAN segment only, where you can be (reasonably) sure that systems on a LAN segment
are the systems they claim to be. Make sure a firewall blocks NFS traffic from outside the LAN and never use NFS
over the Internet.
To improve security, make sure UIDs and GIDs are the same on server and clients (page 668).
More Information
Web .
nfs.sourceforge.net [http://nfs.sourceforge.net]
HOWTO .
NFS HOWTO
Netboot and PXE Remote-boot mini HOWTO
Book .
NFS Illustrated by Callaghan, Addison-Wesley; 1st edition (December 17, 1999)
NFS Client
This section covers setting up a NFS client, mounting remote directory hierarchies, and improving NFS performance.
Prerequisites
Install the following packages:
•
nfs-utils
351
Draft
•
JumpStart: Mounting a Remote Directory Hierarchy
Draft
system-config-nfs (FEDORA) or redhat-config-nfs (RHEL) (optional)
The portmap utility (refer to “RPC Network Services” on page 358) must be running to enable reliable file locking.
There are no daemons to start for NFS clients.
JumpStart: Mounting a Remote Directory Hierarchy
To set up an NFS client, mount the remote directory hierarchy the same way you mount a local directory hierarchy
(page 442). The following sections detail the process.
mount: Mounts a Remote Directory Hierarchy
The following examples show two ways to mount a remote directory hierarchy, assuming that speedy is on the
same network as the local system and is sharing /home and /export with the local system. The /export directory
on speedy holds two directory hierarchies that you want to mount: /export/progs and /export/oracle. The example
mounts speedy’s /home directory on /speedy.home on the local system, /export/progs on /apps, and /export/oracle on /oracle.
First, use mkdir to create the directories that are the mount points for the remote directory hierarchies:
# mkdir /speedy.home /apps /oracle
You can mount any directory from an exported directory hierarchy: In this example, speedy exports /export and
the local system mounts /export/progs and /export/oracle.
The following commands manually mount the directory hierarchies one time. The error mount: RPC: Program
not registered may mean NFS is not running on the server.
# mount speedy:/home /speedy.home
# mount -o r,nosuid speedy:/export/progs /apps
# mount -o r speedy:/export/oracle /oracle
By default, directory hierarchies are mounted read-write, assuming the NFS server is exporting them with readwrite permissions. The first preceding command mounts the /home directory hierarchy from speedy on the local
directory /speedy.home. The second and third mount lines use the –o r option to force a readonly mount. The
second mount line adds the nosuid option, which forces setuid (page 175) executables in the mounted directory
hierarchy to run with regular permissions on the local system.
nosuid option . Giving a user the ability to run a setuid program can give that user the power of Superuser and
should be limited. Unless you know that a user will need to run a program with setuid permissions from a mounted
directory hierarchy, always mount a directory hierarchy with the nosuid option. An example of when you would
need to mount a directory hierarchy with setuid privileges is a diskless workstation that has its root partition
mounted using NFS.
nodev option . Mounting a device file creates another potential security hole. Although the best policy is not
to mount untrustworthy directory hierarchies, it is not always possible to implement this policy. Unless a user
needs to use a device on a mounted directory hierarchy, mount directory hierarchies with the nodev option, which
prevents character and block special files (page 439) on the mounted directory hierarchy from being used as devices.
fstab . If you mount directory hierarchies frequently, you can add entries for the directory hierarchies to the
/etc/fstab file (page 663). (Alternatively, you can use automount, page 671.) The following /etc/fstab entries
352
Draft
mount: Mounts a Directory Hierarchy
Draft
automatically mount the same directory hierarchies as the previous example at the same time as the system mounts
the local filesystems:
speedy:/home
speedy:/export/progs
speedy:/export/oracle
/speedy.home
/apps
/oracle
nfs
nfs
nfs
r,nosuid
r
0
0
0
0
0
0
A file that is mounted using NFS is always type nfs on the local system, regardless of what type it is on the remote
system. Typically, you never run fsck on or back up an NFS directory hierarchy. The entries in the third, fifth, and
sixth columns of fstab are usually nfs (filesystem type), 0 (do not back up this directory hierarchy with dump, page
529), and 0 (do not run fsck [page 447] on this directory hierarchy). The options for mounting an NFS directory
hierarchy differ from those for mounting an ext3 or other type of filesystem. See the next section for details.
umount: Unmounts a Remote Directory Hierarchy
Use umount to unmount a remote directory hierarchy the same way you would unmount a local one.
mount: Mounts a Directory Hierarchy
The mount utility (page 442) associates a directory hierarchy with a mount point (a directory). You can use mount
to mount an NFS (remote) directory hierarchy. This section describes some mount options and lists default options
first and nondefault options, enclosed in parentheses, second. You can use these options on the command line or
in /etc/fstab (page 663). For a complete list of options, refer to the mount and nfs man pages.
Attribute Caching
File attributes, stored in a file’s inode (page 437), are information about a file, such as file modification time, size,
links, and owner. File attributes do not include the data stored in a file. Typically, file attributes do not change very
often for an ordinary file and change less often for a directory file. Even the size attribute does not change with
every write instruction: When a client is writing to an NFS-mounted file, several write instructions may be given
before the data is actually transferred to the server. Finally, many file accesses, such as that performed by ls, are
read-only and do not change the file’s attributes or its contents. Thus a client can cache attributes and avoid costly
network reads.
The kernel uses the modification time of the file to determine when its cache is out of date. If the time the attribute
cache was saved is later than the modification time of the file itself, the data in the cache is current. The attribute
cache of an NFS-mounted file must be periodically refreshed from the server to see if another process has modified
the file. This period is specified as a minimum and maximum number of seconds for ordinary and directory files.
Following is a list of options that affect attribute caching.
ac (noac) . attribute cache Permits attribute caching (default). The noac option disables attribute caching.
Although noac slows the server, it avoids stale attributes when two NFS clients actively write to a common directory
hierarchy.
acdirmax=n . attribute cache directory file maximum The number of seconds, at a maximum, that NFS
waits before refreshing directory file attributes (default is 60 seconds).
acdirmin=n . attribute cache directory file minimum The number of seconds, at a minimum, that NFS waits
before refreshing directory file attributes (default is 30 seconds).
acregmax=n . attribute cache regular file maximum The number of seconds, at a maximum, that NFS waits
before refreshing regular file attributes (default is 60 seconds).
353
Draft
mount: Mounts a Directory Hierarchy
Draft
acregmin=n . attribute cache regular file minimum The number of seconds, at a minimum, that NFS waits
before refreshing regular file attributes (default is 3 seconds).
actimeo=n . attribute cache timeout Sets acregmin, acregmax, acdirmin, and acdirmax to n (without this
option, each individual option takes on its default value).
Error Handling
The following options control what NFS does when the server does not respond or there is an I/O error. To allow
for a mount point located on a mounted device, a missing mount point is treated as a timeout.
Note
fg (bg)
foreground . Retries failed NFS mount attempts in the foreground (default). The bg (background) option retries
failed NFS mount attempts in the background.
Note
hard (soft)
Displays server not responding on the console on a major timeout and keeps retrying (default). The soft option
reports an I/O error to the calling program on a major timeout. In general, it is not advisable to use soft. The mount
man page says of soft, “Usually it just causes lots of trouble.” For more information, refer to “Improving Performance” on page 662.
Note
nointr (intr)
no interrupt . Does not allow a signal to interrupt a file operation on a hard mounted directory hierarchy when
a major timeout occurs (default). The intr option allows this type of interrupt.
Note
retrans=n
retransmission value . After n minor timeouts, NFS generates a major timeout (default is 3). A major timeout
aborts the operation or displays server not responding on the console, depending on whether hard or soft is set.
Note
retry=n
retry value .
The number of minutes that NFS retries a mount operation before giving up (default is 10,000).
Note
timeo=n
timeout value . The number of tenths of a second that NFS waits before retransmitting following an RPC, or
minor, timeout (default is 7). The value is increased at each timeout to a maximum of sixty seconds or until a major
354
Draft
mount: Mounts a Directory Hierarchy
Draft
timeout occurs (see retrans). On a busy network, a slow server, or when the request passes through multiple
routers/gateways, increasing this value may improve performance.
Miscellaneous Options
Following are additional, useful options:
Note
lock (nolock)
Permits NFS locking (default). The nolock option disables NFS locking (does not start the lockd daemon) and is
useful with older servers that do not support NFS locking.
Note
mounthost=name
The name of the host running mountd, the NFS mount daemon.
Note
mountport=n
The port that mountd uses.
Note
nodev
no device .
Causes mounted device files not to function as devices (page 659).
Note
port=n
The port used to connect to the NFS server (defaults to 2049 if the NFS daemon is not registered with portmap).
When n=0 (the default), NFS queries portmap on the server to determine the port.
Note
rsize=n
read block size . The number of bytes read at one time from an NFS server. The default block size is 4096.
Refer to “Improving Performance” on page 662.
Note
wsize=n
write block size . The number of bytes written at one time to an NFS server. The default block size is 4096.
Refer to “Improving Performance” on page 662.
355
Draft
Improving Performance
Draft
Note
tcp
Use TCP in place of the default UDP protocol for an NFS mount. This option may improve performance on a
congested network.
Note
udp
Use the default UDP protocol for an NFS mount.
Improving Performance
There are several parameters that can affect the performance of NFS, especially over slow connections such as a
line with a lot of traffic or one controlled by a modem. If you have a slow connection, make sure hard (page 661)
is set (this is the default) so that timeouts do not abort program execution.
One of the easiest ways to improve NFS performance is to increase the block size, the number of bytes NFS transfers
at a time. The default of 4096 is low for a fast connection using modern hardware. Try increasing rsize (page 662)
and wsize (page 662) to 8192 or higher. Experiment until you find the optimal block size. Unmount and mount
the directory hierarchy each time you change an option. See the NFS-HOWTO for more information on testing
different block sizes.
NFS waits the amount of time specified by the timeo (timeout, page 661) option for a response to a transmission.
If it does not receive a response in this amount of time, it sends another transmission. The second transmission
uses bandwidth that, over a slow connection, may slow things down further. You may be able to increase performance
by increasing timeo.
The default value of timeo is 7 tenths of a second or 700 milliseconds. After a timeout, NFS doubles the time it
waits to 1400 milliseconds. On each timeout it doubles the amount of time it waits to a maximum of 60 seconds.
You can test the speed of a connection with the size packets you are sending (rsize (page 662) and wsize [page
662]) using ping with the –s (size) option:
$ ping -s 4096 speedy
PING speedy.tcorp.com (192.168.0.1) 4096(4124) bytes of data.
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=0 ttl=64 time=1.43 ms
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=1 ttl=64 time=1.17 ms
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=2 ttl=64 time=1.17 ms
...
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=26 ttl=64 time=1.16 ms
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=27 ttl=64 time=1.16 ms
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=28 ttl=64 time=1.16 ms
4104 bytes from speedy.tcorp.com (192.168.0.1): icmp_seq=29 ttl=64 time=1.26 ms
--- speedy.tcorp.com ping statistics --30 packets transmitted, 30 received, 0% packet loss, time 29281ms
rtt min/avg/max/mdev = 1.154/1.192/1.431/0.067 ms
The preceding example uses Red Hat’s default packet size of 4096 bytes and shows a fast average packet roundtrip time of just over 1 millisecond. Over a modem line, you can expect times of several seconds. If there is other
traffic on the connection, the time will be longer. Run the test during a period of heavy traffic. Try increasing timeo
356
Draft
/etc/fstab: Mounts Directory Hierarchies Automatically
Draft
to three or four times the average round-trip time (to allow for unusually bad network conditions, as when the
connection is made) and see if things improve. Remember that the timeo value is tenths of a second (100 milliseconds
= 1 tenth of a second).
/etc/fstab: Mounts Directory Hierarchies Automatically
The /etc/fstab file (page 445) lists directory hierarchies that the system mounts automatically as it comes up. You
can use the options discussed in the preceding section on the command line or in the fstab file.
The first example line from fstab mounts grape’s /gc1 filesystem on the /grape.gc1 mount point. A mount point
should be an empty, local directory. (Files in a mount point are hidden when a directory hierarchy is mounted on
it.) The type of a filesystem mounted using NFS is always nfs, regardless of its type on the local system. You can
increase the rsize and wsize options to improve performance. Refer to “Improving Performance” on page 662.
grape:/gc1
/grape.gc1
nfs
rsize=8192,wsize=8192
0
0
The next example from fstab mounts a filesystem from speedy. Because the local system connects to speedy over
a slow connection, timeo is increased to 5 seconds (50 tenths of a second). Refer to “Timeouts” on page 662. In
addition, hard is set to make sure that NFS keeps trying to communicate with the server after a major timeout.
Refer to “hard/soft” on page 662.
speedy:/export
/speedy.export
nfs
timeo=50,hard
0
0
The final example from fstab shows a remote-mounted home directory. Because speedy is a local server and is
connected via a reliable, high-speed connection, timeo is decreased and rsize and wsize are increased substantially.
speedy:/export/home
/home
nfs
timeo=4,rsize=16384,wsize=16384
0
0
NFS Server
Prerequisites
Install the following package:
•
nfs-utils
Run chkconfig to cause nfs to start when the system goes multiuser:
# /sbin/chkconfig nfs on
Start nfs:
# /etc/rc.d/init.d/nfs start
The nfs init script starts mountd, nfsd, and rquotad.
The portmap utility (page 358) must be running to enable file locking.
357
Draft
JumpStart: system-config-nfs: Configures an NFS Server
Draft
JumpStart: system-config-nfs: Configures an NFS Server
You can generate an /etc/exports file, which is almost all there is to setting up an NFS server, using system-confignfs (FEDORA) or redhat-config-nfs (RHEL). These utilities display the NFS Server Configuration window (Figure
22-2), which allows you to specify which directory hierarchies are shared and how they are shared using NFS. You
use the same window to add and modify directory hierarchies; in the first case it is named Add NFS Share and in
the second it is named Edit NFS Share.
Figure 22-2. NFS Server Configuration window
The Add/Edit NFS Share window has three tabs: Basic, General Options, and User Access. Figure 22-3 shows the
Basic tab where you can specify the pathname of the root of the directory hierarchy you want to share, the names
or IP addresses of the systems (hosts) you want to share the hierarchy with, and whether you want users from the
specified systems to be able to write to the shared files.
Figure 22-3. Edit NFS Share window
The selections in the other two tabs correspond to options that you can use in the /etc/exports file. Following is
a list of the check box descriptions in these tabs and the options each corresponds to. Refer to the options starting
on page 666 for more information.
Allow connections from ports 1023 and higher insecure
Allow insecure file locking no_auth_nlm or insecure_locks
Disable subtree checking no_subtree_check
358
Draft
Exporting a Directory Hierarchy
Draft
Sync write operations on request sync
Force sync of write operations immediately no_wdelay
Treat remote root user as local root no_root_squash
Treat all client users as anonymous users all_squash
Specify local user ID for anonymous users anonuid
Specify local group ID for anonymous users anongid
After making the changes you want, click OK to close the Add/Edit NFS Share window and click OK again to close
the NFS Server Configuration window. There is no need to restart any daemons.
Exporting a Directory Hierarchy
Exporting a directory hierarchy makes the directory hierarchy available for mounting by designated systems via a
network. Exported does not mean mounted: When a directory hierarchy is exported, it is placed in the list of directory hierarchies that can be mounted by other systems. An exported directory hierarchy may be mounted or not
at any given time. A server holds three lists of exported directory hierarchies:
•
/etc/exports . Access control list for exported directory hierarchies (next section). The system administrator
can modify this file by editing it or by running [system|redhat]-config-nfs.
•
/var/lib/nfs/xtab . Access control list for exported directory hierarchies. Initialized from /etc/exports when
the system is brought up. Read by mountd when a client asks to mount a directory hierarchy. Modified by exportfs (page 670) as directory hierarchies are mounted and unmounted by NFS.
•
Kernel’s export table . List of active exported directory hierarchies. The kernel obtains this information
from /var/lib/nfs/xtab. You can display this table by giving the command cat /proc/fs/nfs/exports.
/etc/exports: List of Exported Directory Hierarchies
The /etc/exports file is the access control list for exported directory hierarchies that NFS clients can mount and
is the only file you need to edit to set up an NFS server. The exports file controls
•
Which clients can access files on the server
•
Which directory hierarchies on the server each client can access
•
How each client can access each directory hierarchy
•
How client usernames are mapped to server usernames
•
Various NFS parameters
The format of each line in the exports file is
export-point client1(options) [client2(options) ... ]
Where export-point is the absolute pathname of the root directory of the directory hierarchy to be exported, client1n is the name of one or more clients or is one or more IP addresses, separated by SPACEs, that are allowed to mount
the export-point. The options, described in the following section, apply to the preceding client.
359
Draft
Exporting a Directory Hierarchy
Draft
Either you can use system-config-nfs (page 664) to make changes to exports or you can edit this file directly. The
following simple exports file gives grape read and write access and gives speedy read-only access to the files in
/home. In each case, access is implicitly granted for all subdirectories. For historical reasons, exportfs complains
when you do not specify either sync or async.
# cat /etc/exports
/home grape(rw,sync)
/home speedy(ro,sync)
You can use IP addresses and include more than one system on a line:
# cat /etc/exports
/home grape(rw,sync)
speedy(ro,sync) 192.168.0.22(rw,sync)
General Options
This section lists default options first and nondefault options, enclosed in parentheses, second.
Note
auth_nlm (no_auth_nlm) or secure_locks (insecure_locks)
Causes the server to require authentication of lock requests (using the NLM [NFS Lock Manager] protocol). Use
no_auth_nlm for older clients when you find that only files that anyone can read can be locked.
When a server exports two directory hierarchies, one of which is mounted on the other, a client has to mount both
directory hierarchies explicitly in order to access both. When the second, or child directory hierarchy is not explicitly
mounted, its mount point appears as an empty directory and the directory hierarchy is hidden. The nohide option
causes the second directory hierarchy to appear without being explicitly mounted, but does not work in all cases.
Caution: Exporting Symbolic Links and Device Files
When you export a directory hierarchy that contains a symbolic link, make sure that the object of the
link is available on the client (remote) system. Either the object of the link must exist on a client system,
or you must export and mount the object along with the exported link, otherwise the symbolic link
will not point to the file it points to on the server.
A device file refers to a Linux kernel interface. When you export a device file, you export that interface.
If the client system does not have the same type of device, the exported device will not work. From a
client, you can use mount’s nodev option (page 659) to prevent device files on mounted directory
hierarchies from being used as devices.
read only .
Permits only read requests on an NFS directory hierarchy. Use rw to permit read and write requests.
Requires that NFS requests originate on a secure port (<1024) so that a program without root permissions cannot
mount a directory hierarchy. This option does not guarantee a secure connection.
Checks subtrees for valid files. Assume that you have an exported directory hierarchy that has its root below the
root of the filesystem that holds it (that is, an exported subdirectory of a filesystem). When the NFS server receives
a request for a file in that directory hierarchy, it performs a subtree check to make sure the file is in the exported
directory hierarchy.
360
Draft
Exporting a Directory Hierarchy
Draft
Subtree checking can cause problems with files that are renamed while opened and, when no_root_squash is
used, files that only root can access. The no_subtree_check option disables subtree checking and can improve
reliability in some cases.
For example, you may need to disable subtree checking for home directories. Home directories are frequently
subtrees (of /home), written to often, and can have files within them frequently renamed. You would probably not
need to disable subtree checking for directory hierarchies that contain files that are mostly read, such as /usr.
synchronize . Specifies that the server is to reply to requests only after it has written to disk changes made by
the request. The async option specifies that the server does not have to wait for information to be written to disk
and can improve performance at the cost of possible data corruption if the server crashes or the connection is interrupted.
Because the default changed with release 1.0.0 of nfs-utils, exportfs displays a warning when you do not specify
either sync or async.
write delay . Causes the server to delay committing write requests when it anticipates that another, related request
follows, improving performance by committing multiple write requests within a single operation. The no_wdelay
option does not delay committing write requests and can improve performance when the server receives multiple,
small, unrelated requests.
User ID Mapping Options
Each user has a UID number and a primary GID number on the local system. The local /etc/passwd and /etc/group
files map these numbers to names. When a user makes a request of an NFS server, the server uses these numbers
to identify the user on the remote system, raising several issues:
•
The user may not have the same ID numbers on both systems and may therefore have owner access to files of
another user (see “NIS and NFS,” following, for a solution).
•
You may not want the root user on the client system to have owner access to root-owned files on the server.
•
There are some important system files that are not owned by root (such as those owned by bin) that you may
not want a remote user to have owner access to.
Security: Critical Files in NFS-Mounted Directories Should Be
Owned by root
Despite the mapping done by the root-squash option, the root user on a client system can use su to
take on the identity of any user on the system and then access that user’s files on the server. Thus,
without resorting to all-squash, you can protect only files owned by root on an NFS server. Make
sure that root, and not bin or another user, owns and is the only user who can modify or delete all
critical files within any NFS-mounted directory hierarchy.
The preceding precaution does not completely protect against an attacker with root privileges, but it
can help protect a system from less experienced malicious users.
Owner access to a file means that the remote user can execute, remove, or, worse, modify the file. NFS gives you
two ways to deal with these cases:
•
You can use the root_squash option to map the ID number of the root user on a client to the nfsnobody user
on the server.
•
You can use the all-squash option to map all NFS users on the client to nfsnobody on the server.
361
Draft
Exporting a Directory Hierarchy
Draft
The /etc/passwd file shows that nfsnobody has a UID and GID of 65534. You can use the anonuid and anongid
options to override these values.
When you use NIS (page 637) for user authorization, users automatically have the same UIDs on both systems. If
you are using NFS on a large network, it is a good idea to use a directory service such as LDAP (page 980) or NIS
for authorization. Without such a service, you will have to synchronize the passwd files on all the systems manually.
Maps requests from root on a remote system to appear to come from the UID for nfsnobody, an unprivileged user
on the local system, or as specified by anonuid. Does not affect other sensitive UIDs such as bin. The
no_root_squash option turns off this mapping so that requests from root appear to come from root.
Does not change the mapping of users making requests of the NFS server. The all_squash option maps requests
from all users, not just root, on remote systems to appear to come from the UID for nfsnobody, an unprivileged
user on the local system, or as specified by anonuid. This option is useful for controlling access to exported public
FTP, news, and other directories.
Note
anonuid=un and anongid=gn
Set the UID or GID of the anonymous account to un or gn, respectively. NFS uses these accounts when it does not
recognize an incoming UID or GID and when instructed to do so by root_squash or all_squash.
showmount: Displays NFS Status Information
Without any options, the showmount utility displays a list of systems that are allowed to mount local directories.
To display information for a remote system, give the name of the remote system as an argument. You typically use
showmount to display a list of directory hierarchies that a server is exporting. The information that showmount
provides may not be complete because it depends on mountd and trusts that remote servers are reporting accurately.
In the following example, bravo and grape can mount local directories, but you do not know which ones:
# /usr/sbin/showmount
Hosts on localhost:
bravo.tcorp.com
grape.tcorp.com
If showmount displays an error such as RPC: Program not registered, NFS is not running on the server. Start NFS
on the server with the nfs init script (page 664).
Note
–a
all .
Tells which directories are exported to which remote systems. This information is stored in /etc/exports.
# /usr/sbin/showmount -a
All mount points on localhost:
bravo.tcorp.com:/home
grape.tcorp.com:/home
362
Draft
exportfs: Maintains the List of Exported Directory Hierarchies
Draft
Note
–e
exports . Displays the export list. Displays the same information as –a from another vantage point: The local
/home directory is exported to bravo and grape:
# /usr/sbin/showmount -e
Export list for localhost:
/home bravo.tcorp.com,grape.tcorp.com
exportfs: Maintains the List of Exported Directory Hierarchies
The exportfs utility maintains the kernel’s list of exported directory hierarchies. Without changing /etc/exports,
exportfs can add or remove from the list of exported directory hierarchies. The format of an exportfs command
follows:
/usr/sbin/exportfs [options] [client:dir ...]
Where options is one or more options as detailed in the next section, client is the name of the system that dir is exported to, and dir is the absolute pathname of the directory at the root of the directory hierarchy being exported.
The system executes the following command when it comes up (it is in the nfs init script); it reexports the entries
in /etc/exports and removes invalid entries from /var/lib/nfs/xtab (page 665) so that /var/lib/nfs/xtab is synchronized with /etc/exports:
# exportfs -r
Replace the –r with –a to export only the entries in /etc/exports. Remove an exported directory hierarchy with
the –u option and remove all exported directory hierarchies with the –ua options.
Options
Note
–a
all . Exports directory hierarchies specified in /etc/exports. This option does not un-export entries you have
removed from exports (that is, it does not remove invalid entries from /var/lib/nfs/xtab); use –r to do this.
Note
–i
ignore .
Ignores /etc/exports; uses what is specified on the command line only.
Note
–o
363
Draft
Testing the Server Setup
Draft
options . Specifies options. You can specify options following –o the same way you do in the exports file. For
example, exportfs –i –o ro speedy:/home/sam exports /home/sam on the local system to speedy for readonly
access.
Note
–r
reexport . Reexports the entries in /etc/exports and removes invalid entries from /var/lib/nfs/xtab so that
/var/lib/nfs/xtab is synchronized with /etc/exports.
Note
–u
unexport . Makes an exported directory hierarchy no longer exported. If a directory hierarchy is mounted when
you unexport it, you see the message Stale NFS file handle when you try to access the directory hierarchy from
the remote system.
Note
–v
verbose .
ation.
Provides more information. Displays export options when you use exportfs to display export inform-
Testing the Server Setup
From the server, run the nfs init script with an argument of status. If all is well, the system displays something
similar to the following:
# /sbin/service nfs status
rpc.mountd (pid 15795) is running...
nfsd (pid 15813 15812 15811 15810 15809 15808 15807 15806) is running...
rpc.rquotad (pid 15784) is running...
Next, from the server, use rpcinfo to make sure NFS is registered with portmap.
$ /usr/sbin/rpcinfo -p localhost | grep nfs
100003
2
udp
2049 nfs
100003
3
udp
2049 nfs
Repeat the preceding command from the client, replacing localhost with the name of the server. The results should
be the same.
Finally, try mounting directory hierarchies from remote systems and verify access.
364
Draft
automount: Mounting Directory Hierarchies Automatically
Draft
automount: Mounting Directory Hierarchies Automatically
With distributed computing, when you log in on any system on the network, all your files, including startup scripts,
are available. A distributed computing environment commonly has all systems able to mount all directory hierarchies
on all servers: Whichever system you log in on, your home directory is waiting for you.
As an example, assume that /home/alex is a remote directory hierarchy that is mounted on demand. When you
issue the command ls /home/alex, autofs goes to work: It looks in the /etc/auto.home map, finds that alex is a
key that says to mount bravo:/export/homes/alex, and mounts the remote directory hierarchy. Once the directory
hierarchy is mounted, ls displays the list of files you want to see. If after this mounting sequence you give the
command ls /home, ls shows that alex is present within the /home directory. The df utility shows that alex is
mounted from bravo.
Prerequisites
Install the following package:
•
autofs
Run chkconfig to cause autofs to start when the system goes multiuser:
# /sbin/chkconfig autofs on
Start autofs:
# /sbin/service nfs start
More Information
man pages autofs, automount, and auto.master
tutorial www.linuxhq.com/lg/issue24/nielsen.html [http://www.linuxhq.com/lg/issue24/nielsen.html]
Automount mini-HOWTO
autofs: Automatically Mounts Directory Hierarchies
An autofs directory hierarchy is like any other directory hierarchy, but remains unmounted until it is needed, at
which time the system mounts it automatically (demand mounting). The system unmounts an autofs directory
hierarchy when it is no longer needed: by default after five minutes of inactivity. Automatically mounted directory
hierarchies are an important part of administrating a large collection of systems in a consistent way. The automount
utility is particularly useful when there is a large number of servers or a large number of directory hierarchies. It
also helps to remove server-server dependencies (following).
When you boot a system that uses traditional fstab-based mounts and an NFS server is down, the system can take
a long time to come up as it waits for the server to time out. Similarly, when you have two servers, each mounting
directory hierarchies from the other and both systems are down, both may hang as they are brought up and each
tries to mount a directory hierarchy from the other. This situation is called a server-server dependency. The automount
facility gets around these issues by mounting a directory hierarchy from another system only when a process tries
to access it.
365
Draft
Chapter Summary
Draft
When a process attempts to access one of the directories within an unmounted autofs directory hierarchy, the
kernel notifies the automount daemon, which mounts the directory hierarchy. You have to give a command, such
as cd /home/alex, that accesses the autofs mount point (in this case /home/alex) in order to create the demand
that causes automount to mount the autofs directory hierarchy so you can see it. Before you issue the cd command,
alex does not appear to be in /home.
The main file that controls the behavior of automount is /etc/auto.master. A simple example follows:
# cat /etc/auto.master
/free1 /etc/auto.misc --timeout 60
/free2 /etc/auto.misc2 --timeout 60
The auto.master file has three columns: The first column names the parent of the autofs mount point—the location
where the autofs directory hierarchy is to be mounted (/free1 and /free2 in the example are not mount points but
will hold the mount points when the directory hierarchies are mounted). The second column names the files, called
map files, that store supplemental configuration information. The optional third column holds mount options for
map entries that do not specify an option.
The map files can have any names, but one is traditionally named auto.misc. Following are the two map files
specified in auto.master:
# cat /etc/auto.misc
sam
-fstype=ext3
# cat /etc/auto.misc2
helen
-fstype=ext3
:/dev/hda8
:/dev/hda9
The first column of a map file holds the relative autofs mount point (sam and helen). This mount point is appended
to the corresponding autofs mount point from column one of the auto.master file to create the absolute autofs
mount point. In this example, sam (from auto.misc) is appended to /free1 (from auto.master) to make /free1/sam.
The second column holds the options, and the third column shows the server and directory hierarchy to be
mounted. This example shows local drives; an NFS-mounted device would have the hostname of the remote system
before the colon (for example, grape:/home/sam).
Before the new setup can work, you have to create directories for the parents of the mount points (/free1 and /free2
in the preceding example) and start or restart the automount daemon using the autofs init script. The following
command displays information about configured and active autofs mount points:
# /sbin/service autofs status
Chapter Summary
NFS allows a server to share selected local directory hierarchies with client systems on a heterogeneous network,
reducing storage needs and administrative overhead. NFS defines a client/server relationship in which a server
provides directory hierarchies that clients can mount.
On the server, the /etc/exports file lists the directory hierarchies that the system exports. Each line in exports lists
the systems that are allowed to mount the hierarchy and specifies the options for each hierarchy (read only, readwrite, and so on). Give an exportfs –r command to cause NFS to reread this file.
From a client, you can give a mount command to mount an exported NFS directory hierarchy or you can put an
entry in /etc/fstab to have the system automatically mount the directory hierarchy when it comes up.
366
Draft
Exercises
Draft
Automatically mounted directory hierarchies help administrate large groups of systems with many servers and
filesystems in a consistent way and can help remove server-server dependencies. The automount daemon automatically mounts autofs directory hierarchies when they are needed and unmounts them when they are no longer
needed.
Exercises
1. List three reasons to use NFS.
2. What command would you give to mount on the local system the /home directory hierarchy that resides on the
file server named bravo? Assume the mounted directory hierarchy will appear as /bravo.home on the local
system. How would you mount the same directory hierarchy if it resided on the fileserver at 192.168.1.1? How
would you unmount /home?
3. How would you list the mount points on the remote system named bravo that the local system named grape
can mount?
4. What command line lists the currently mounted NFS directory hierarchies?
5. What does the /etc/fstab file do?
The fstab file facilitates mounting directory hierarchies. You can use fstab to mount automatically directory
hierarchies when the system comes up and to allow an ordinary user to mount a directory hierarchy.
6. From a server, how would you allow read-only access to /opt for any system in example.com?
Advanced Exercises
7. When is it a good idea to disable attribute caching?
8. Describe the difference between the root_squash and the all_squash options in /etc/exports.
9. Why does the secure option in /etc/exports not really provide any security?
10. Some diskless workstations use NFS as swap space. Why is this useful? What is the downside?
11. NFS maps client users to users on the server. Explain why this mapping is a security risk.
12. What does the mount nosuid option do? Why would you want to do this?
367
368
Draft
Draft
A Practical Guide to Red Hat® Linux®: Fedora™ Core and Red Hat Enterprise Linux,
Second Edition
By Mark Sobell
Publisher: Prentice Hall PTR
Pub Date: July 30, 2004 ISBN: 0-13-147024-8
A custom excerpt from Chapter 23 - Samba: Integrating Linux and Windows
Samba: Integrating Linux and Windows
IN THIS CHAPTER
About Samba
JumpStart: system-config-samba: Configuring a Samba
Server
swat: Configuring a Samba Server
Manually Configuring a Samba Server
Accessing Linux Shares from Windows
Accessing Windows Shares from Linux
Troubleshooting
Samba is a free suite of programs that enables UNIX-like operating systems, including Linux, Solaris, FreeBSD,
and MacOS X, to work with other operating systems, such as OS/2 and Windows, as both a server and a client.
As a server, Samba shares Linux files and printers with Windows systems. As a client, Samba gives Linux users
access to files on Windows systems. Its ability to share files across operating systems makes Samba an ideal tool
in a heterogeneous computing environment.
Refer to “Integration with Windows” on page 502 for information about printing using Samba.
Which Version of Samba?
Note
FEDORA
Fedora Core 2 provides Samba 3.
Note
RHEL
Red Hat Enterprise Linux version 3 provides Samba 2.
Samba changed significantly between versions 2 and 3. Most of the changes are internal; however some may require
you to change configuration parameters.
369
Draft
Introduction
Draft
The primary focus when developing Samba 3 was to provide a migration path for organizations running a Windows
NT 4 primary domain controller. Windows NT 4 is officially unsupported at the end of 2004, and Samba on top
of Linux is now a viable alternative to Windows 2000 as an upgrade path.
If you are migrating from Samba version 2 to version 3, read the Samba HOWTO, Chapter 30, “Upgrading from
Samba-2.x to Samba-3.0.0,” which you can find at www.samba.org/samba/docs/man/upgrading-to-3.0.html
[http://www.samba.org/samba/docs/man/upgrading-to-3.0.html].
Introduction
This chapter starts with a technical discussion of Samba followed by some basic information. The JumpStart section
discusses how to set up a Samba server using system-config-samba, a minimal GUI. The next section covers how
to use swat, a Web-based advanced GUI configuration tool, to set up a Samba server. The final server section discusses
how to set up a Samba server by hand, using text editor to edit manually the files that control Samba. The next
two sections, “Accessing Linux Shares from Windows” (page 691) and “Accessing Windows Shares from Linux”
(page 692), explain how to work with Linux and Windows files and printers. The final section of the chapter,
“Troubleshooting” (page 694), discusses what to do if you have a problem setting up or using Samba.
Table 23-1 lists the utilities and daemons that make up the Samba suite of programs.
Table 23-1. Samba Utilities and Daemons
make_smbcodepage
Makes SMB a codepage definition file. Used in Samba
internationalization features. Maps upper- to lowercase
and vice versa for characters > ASCII 127.
net
FEDORA Samba 3 includes a new utility: net. This utility
has the same syntax as the Windows net command and,
over time, will eventually replace other Samba utilities
such as smbpasswd.
nmbd
The NetBIOS nameserver program, run as a daemon by
default. Provides NetBIOS over IP naming services for
Samba clients. Also provides browsing (as in the Windows Network Neighborhood or My Network Places
view) support.
nmblookup
Makes NetBIOS name queries (page 695).
smbclient
Displays shares on a Samba server such as a Windows
machine (page 692).
smbd
The Samba program, run as a daemon by default.
Provides file and print services for Samba clients.
smbpasswd
Changes Windows NT password hashes on Samba and
Windows NT servers (page 678).
smbstatus
Displays information about current smbd connections.
smbtree
Displays a hierarchical diagram of available shares (page
692).
swat
Samba Web Administration Tool. A graphical editor for
the smb.conf file (page 681).
testparm
Checks syntax of the smb.conf file (page 694).
testprns
Checks printer names in the printcap file.
370
Draft
About Samba
Draft
About Samba
This section covers the packages you need to install to run Samba, where you can get more information, and users
and passwords under Samba.
Prerequisites
Install the following packages:
•
samba
•
samba-client
•
samba-common
•
system-config-samba (FEDORA) or redhat-config-samba (RHEL) (optional)
•
samba-swat (optional, but a good idea)
Run chkconfig to cause smb to start when the system goes multiuser:
# /sbin/chkconfig smb on
Start smb:
# /sbin/service smb start
If you want to use swat, modify /etc/xinetd.d/swat, as explained in “ swat: Configuring a Samba Server” on page
681, and restart xinetd:
# /sbin/service xinetd restart
More Information
Samba/swat home page has links to local Samba documentation (page 681)
Documentation /usr/share/doc/samba-*
Samba .
CIFS .
www.samba.org [http://www.samba.org] (mailing lists, documentation, downloads, and more)
www.samba.org/cifs [http://www.samba.org/cifs]
Unofficial Samba HOWTO .
hr.uoregon.edu/davidrl/samba
Samba HOWTO Collection .
point a browser at
/usr/share/doc/samba-*/docs/htmldocs/index.html
371
Draft
Samba Users, User Maps, and Passwords
Draft
Samba Users, User Maps, and Passwords
In order for a Windows user to gain access to Samba services on a Linux system, you must provide a Windows
username and a Samba password. In some cases, Windows supplies the username and password for you. It is also
possible to authenticate using other methods. Samba can use LDAP (page 980) or PAM (page 416) in place of the
default password file. Refer to the Samba documentation for more information on authentication methods.
The supplied username must be the same as a Linux username or must map to a Linux username. Samba keeps
the username maps in /etc/samba/smbusers. Users with the same username on Linux and Samba do not need to
appear in this file, but they still need a Samba password.
When you install Samba, smbusers has two entries:
$ cat /etc/samba/smbusers
# Unix_name = SMB_name1 SMB_name2 ...
root = administrator admin
nobody = guest pcguest smbguest
The first entry maps the two Windows usernames (administrator and admin) to the Linux username root. The
second entry maps three Windows usernames, including guest, to the Linux username nobody: When a Windows
user attempts to log in on the Samba server as guest, Samba authenticates the Linux user named nobody.
Samba uses Samba passwords, not Linux passwords, to authenticate users. By default, Samba keeps passwords in
/etc/samba/smbpasswd. As Samba is installed, authentication for root or nobody would fail because Samba is
installed without any passwords: The smbpasswd file does not exist.
Each of the configuration techniques described in this chapter allows you to add users to smbusers and passwords
to smbpasswd. You can always use smbpasswd (see “Example,” following) to add and change passwords in smbpasswd.
When you attempt to connect from Windows to a Samba server, Windows presents your Windows username and
password to Samba. If your Windows username is the same as, or maps to, your Linux username and if your
Windows and Samba passwords are the same, you do not have to enter a username or password to connect to the
Samba server.
You can add the following line to smbusers to map the Windows username sam to the Linux username sls:
sls = sam
You can add a password for sls to smbpasswd with the following command:
# smbpasswd -a sls
New SMB password:
Retype new SMB password:
Added user sls.
Now, when Sam logs in on the Samba server as sam, Samba maps sam to sls and looks up sls in smbpasswd.
Assuming Sam provides the correct password, he logs in on the Samba server as sls.
372
Draft
JumpStart: system-config-samba:
Configuring a Samba Server
Draft
JumpStart: system-config-samba: Configuring a Samba Server
The system-config-samba (FEDORA) and redhat-config-samba (RHEL) utilities can set up only basic features of a
Samba server. They are, however, the best tool to use if you are not familiar with Samba and you want to set up a
simple Samba server quickly. The [system|redhat]-config-samba utility has three basic functions: configuring the
server, configuring users, and setting up shares or directories that are exported to the Windows machines.
The [system|redhat]-config-samba utility edits the /etc/samba/smb.conf file; make a copy of this file for safekeeping
before you start. Start this tool by giving the command [system|redhat]-config-samba from a terminal emulator
command line. The utility displays the Samba Server Configuration window (Figure 23-1).
Figure 23-1. Samba Server Configuration window
Select Menubar: Preferences Server Settings to display the Server Settings window Basic tab (Figure 23-2).
Change the workgroup to the one in use on your Windows machines. Change the description of the server if you
like. Click the Security tab and make sure Authentication Mode is set to User; you do not need to specify an Authentication Server. If you are using Windows 98 or newer, set Encrypt Passwords to Yes. When you specify a
username in Guest Account, anyone logging into the Samba server as guest maps to that user’s ID. Typically, the
guest account maps to the UID of the Linux user named nobody. Click OK.
373
Draft
JumpStart: system-config-samba:
Configuring a Samba Server
Draft
Figure 23-2. Server Settings window, Basic tab
Select Menubar: Preferences Samba Users to display the Samba Users window (Figure 23-3). If the user you
want to log in as is not already specified in this window, click Add User. With the proper permissions, the Create
New Samba User window displays a combo box next to Unix Username that allows you to select a Linux user.
Otherwise, your username is displayed as the Unix Username. The Windows Username is the Windows username
that you want to map to the specified Linux (UNIX) username. The Samba Password is the password this user or
Windows enters to gain access to the Samba server.
Figure 23-3. Samba Users window
If Sam has accounts named sam on both the Windows and Linux systems, you would select sam from the Unix
Username combo box, enter sam in the Windows Username text box, and enter Sam’s Windows password in the
two Samba Password text boxes.
374
Draft
JumpStart: system-config-samba:
Configuring a Samba Server
Draft
Tip: Adding a Samba Password for the Linux User nobody
Because the user nobody exists in smbusers when you install Samba, you cannot add the user nobody,
nor can you add a password for nobody from system-config-samba; you must use smbpasswd from
the command line as follows:
# smbpasswd -a nobody
New SMB password:
Retype new SMB password:
Normally, the user nobody does not have a password because it is the guest login. Just press RETURN
(without typing any characters) in response to each of the SMB password prompts to add nobody
to the password file without a password.
Click OK to close the Create New Samba User window and click OK to close the Samba Users window.
Next, you need to add a share, which is the directory you export from the Linux system to the Windows system.
Click Add on the Toolbar to display the Create Samba Share window Basic tab (Figure 23-4). In the Directory text
box, enter the absolute pathname of the directory you want to share (/tmp is an easy directory to practice with).
Enter a description if you like. It can be useful to enter the Linux hostname and the pathname of the directory you
are sharing here. Specify Readonly or Read/Write as suits your purpose. Click the Access tab and specify if you
want to limit access to specified users or if you want to allow anyone to access this share. Click OK. Close the
Samba Server Configuration window.
Figure 23-4. Create Samba Share window, Basic tab
There is no need to restart the Samba server; you should be able to access the share from a Windows machine (page
691).
Caution: Make a Copy of smb.conf
Before you run swat, make a copy of /etc/samba/smb.conf because swat overwrites it.
375
Draft
swat: Configuring a Samba Server
Draft
swat: Configuring a Samba Server
The swat (Samba Web Administration Tool) utility is a browser-based graphical editor for the smb.conf file. For
each of the configurable parameters, it provides help links, default values, and a text box to change the value. The
swat utility is a well-designed tool in that it remains true to the lines in the smb.conf file you edit: You can use and
learn from swat and the transition to using a text editor on smb.conf will be straightforward.
The swat utility is run from xinetd (page 403). Before you can run swat, you need to edit /etc/xinetd.d/swat, as
shown following:
$ cat /etc/xinetd.d/swat
# Default: off
# description: SWAT is the Samba Web Admin Tool. Use swat \
#
to configure your Samba server. To use SWAT, \
#
connect to port 901 with your favorite web browser.
service swat
{
port
= 901
socket_type
= stream
wait
= no
only_from
= 127.0.0.1
user
= root
server
= /usr/sbin/swat
log_on_failure += USERID
disable
= yes
}
First, you must turn swat on by changing the yes that follows disable = to no. If you want to access swat from
other than the local system, add the names or IP addresses of the other systems you want to access swat from on
the line that starts with only_from. Separate the system names or IP addresses with SPACEs. Finally, you need to
restart xinetd so that it rereads its configuration files:
# /etc/rc.d/init.d/xinetd restart
Stopping xinetd:
Starting xinetd:
[ OK ]
[ OK ]
Alternatively, you can give the command killall –HUP xinetd.
After making these changes and restarting xinetd, you should be able to run swat. From the local system, open a
browser, enter http://127.0.0.1:901 or http://localhost:901 in the location bar, and give the username root and
the root password in response to swat’s request for a username and password. From a remote system, replace
127.0.0.1 with the IP address of the server in the preceding command.
Security: Do Not Allow Remote Access to swat
Do not allow access to swat from a remote system. When you do so and log in, your password is sent
in cleartext over whatever connection you are using and can easily be sniffed.
The browser displays the local Samba/swat home page (Figure 23-5). This page has links to local Samba documentation and the following buttons:
376
Draft
swat: Configuring a Samba Server
Draft
Note
HOME
Links to local Samba documentation. When you click the word Samba (not the logo, but the one just before the
word Documentation on the HOME window), swat displays the Samba man page, which defines each Samba
program.
Note
GLOBALS
Edits global variables (parameters) in smb.conf.
Note
SHARES
Edits share information in smb.conf.
Note
PRINTERS
Edits printer information in smb.conf.
Note
WIZARD
Rewrites the smb.conf file, removing all comment lines and lines that specify default values.
Note
STATUS
Shows the active connections, active shares, and open files. Stops and restarts smbd and nmbd.
Note
VIEW
Displays a subset or all of the configuration parameters as determined by default values and settings in smb.conf.
Note
PASSWORD
Manages passwords.
It is quite easy to establish a basic Samba setup so that you can see a Linux directory from a Windows system (any
version of Windows from 3.1 on). More work is required to set up a secure connection or one with special features.
The following example creates a basic setup based on the sample smb.conf file that is included with Red Hat.
377
Draft
swat: Configuring a Samba Server
Draft
Figure 23-5. The local swat home page
Each of the variables/parameters in swat has a link named Help next to it. Click Help and a new browser window
containing an explanation of the parameter appears. Each variable/parameter also has a Set Default button that
you can click to reset the variable/parameter to its default value.
For this example, do not click any of the Set Default buttons. Make sure to click Commit Changes at the top of
each page after you finish making changes on a page and before you click a menu button at the top of the window;
otherwise, swat will not keep your changes.
First, click the GLOBALS button at the top of the Samba/swat home page. Leave everything at its current setting
except workgroup, hosts allow, and hosts deny. Set workgroup to the workgroup used on your Windows systems.
(If you followed the preceding JumpStart, the workgroup is already set.) Scroll to the bottom of the Security Options
and set hosts allow to the names or IP addresses of machines that you want to be able to access the local system’s
shares and printers (including local-host [127.0.0.1]). Set hosts deny to ALL. See page 686 for more information
on various ways you can set hosts allow. Click Commit Changes when you are done with the GLOBALS page.
Tip: If You Can No Longer Use swat...
If you can no longer use swat, you probably changed the hosts allow setting incorrectly. If this is the
case, you need to edit /etc/samba/smb.conf and fix the line with the words hosts allow in it:
378
Draft
swat: Configuring a Samba Server
Draft
# grep hosts smb.conf
hosts allow = 127.0.0.1, 192.168.0.8
hosts deny = ALL
The preceding entries allow access from the local system and from 192.168.0.8 only.
Next, click the SHARES button at the top of the page. Three buttons and two text boxes appear in addition to the
two Change View To buttons (Figure 23-6). In the box adjacent to the Create Share button, enter the name you
want to assign to the share you are setting up. This name can be anything you want; it is the name that a Windows’
user sees and selects when working with the share. Click Create Share.
Figure 23-6. The Share Parameters page
When you want to modify an existing share, bring up the name of the share in the combo box adjacent to Choose
Share, and click Choose Share. Either of these actions displays the complete Share Parameters window.
Leave everything at its default setting except path, which specifies the absolute pathname on the local Linux system
of the share, and optionally comment, which you can use to specify the Linux system and directory that this share
points to. The values for hosts allow and hosts deny are taken from the global variables that you set previously.
Click Commit Changes when you are done with the SHARES page. If you want to see how many parameters there
really are, click the Advanced button near the top of the page.
Tip: You Do Not Need to Restart Samba When You Change smb.conf
Samba rereads its configuration files each time a client connects; you do not need to restart Samba
when you change smb.conf.
Now, from a Windows machine, you should be able to access the share you just created (page 691).
379
Draft
Manually Configuring a Samba Server
Draft
Manually Configuring a Samba Server
The /etc/samba/smb.conf file controls most aspects of how Samba works and is divided into sections, each beginning with a line that starts with an open bracket ([), includes some text, and ends with a close bracket (]). The text
within the brackets identifies the section. Typical sections are
[globals] .
defines global parameters
[printers] .
[homes] .
defines printers
defines shares in the homes directory
[share name] .
defines a share (you can have more than one of these sections)
Installing Samba using the Red Hat rpm files creates a sample configuration file (/etc/samba/smb.conf), which
has extensive comments and commented-out examples. Comment lines in smb.conf can start with a pound sign
(#) or a semicolon (;). The sample file uses pound signs to begin lines that are intended to remain as comments
and semicolons to begin lines that you may want to mimic or use as is by removing the semicolons. In the following
segment of smb.conf, there are two lines of true comments and seven lines beginning with semicolons that you
may want to un-comment and make changes to:
# A private directory, usable only by fred. Note that fred requires
# write access to the directory.
;[fredsdir]
;
comment = Fred's Service
;
path = /usr/somewhere/private
;
valid users = fred
;
public = no
;
writable = yes
;
printable = no
Assuming the global parameters in smb.conf are set properly, you need to add a share to get a Windows system
to be able to access a directory on the local Linux system. Add the following simple share to the end of the smb.conf
file to enable a user on a Windows system to be able to read from/write to the local /tmp directory:
[tmp]
comment = temporary directory
path = /tmp
writeable = yes
guest ok = yes
The name of the share under Windows is tmp; the path under Linux is /tmp. Any Windows user, including guest,
who can log in on Samba, can read from/write to the directory, assuming that that user’s Linux permissions allow
it. The Linux permissions that apply to a Windows user using Samba are the permissions that apply to the Linux
user that the Windows user maps to.
Parameters in the smbd.conf File
The the smb.conf man page and the help feature of swat list all the parameters you can set in smb.conf. The following sections list some of the parameters you are likely to want to change.
380
Draft
Parameters in the smbd.conf File
Draft
Global Parameters
Note
interfaces
A SPACE separated list of the networks that Samba uses. Specify as interface names such as eth0 or IP address/net
mask pairs (page 400).
Default: all active interfaces except 127.0.0.1
Note
printing
The printing system in use on the server. On a Red Hat system, you will normally set to CUPS. Other choices are
BSD, LPRNG, and SYSV. You can set this parameter on a per-printer basis. See page 502 for information about
printing under Samba.
Default: none
Note
server string
The string that is displayed in various places on the Windows machine. Within the string, Samba replaces %v with
the Samba version number and %h with the hostname.
Default: Samba %v
Red Hat: Samba Server
Note
workgroup
The workgroup that the server belongs to. Set to the same workgroup as the Windows clients who use the server.
This parameter controls the domain name that Samba uses when security (page 687) is set to DOMAIN.
Default: WORKGROUP
Red Hat: MYGROUP
Security Parameters
Note
encrypt passwords
YES accepts only encrypted passwords from clients. Windows 98 and Windows NT 4.0 Service Pack 3 and above
use encrypted passwords by default. Uses smbpasswd to authenticate passwords unless you set security to
SERVER or DOMAIN, in which case Samba authenticates using another server.
Default: YES
381
Draft
Parameters in the smbd.conf File
Draft
Note
FEDORA
Samba 3 defaults to storing encrypted passwords in the smbpasswd file if you do not set up passdb (a password
database). Storing passwords in the smbpasswd file is sensible on servers with under 250 users. For high-load
servers, consult the Samba HOWTO collection for information about configuring a database back end.
Note
guest account
The username that is assigned to users logging in as guest or mapped to guest; applicable only when guest ok
(page 690) is set to YES. This username should be present in /etc/passwd but should not be able to log in on the
system. Typically, guest account is assigned a value of nobody because the user nobody can access only files that
any user can access. If you are using the nobody account on the Linux system, set this variable to a name other
than nobody.
Default: nobody
Note
hosts allow
Analogous to the /etc/hosts.allow file (page 404), this parameter specifies hosts that are allowed to connect to
the server. Overrides hosts specified in hosts deny. A good strategy is to specify ALL in hosts deny and specify
the hosts you want to grant access to in this file. Specify hosts in the same manner as hosts.allow.
Default: none (all hosts permitted access)
Note
hosts deny
Analogous to the /etc/hosts.deny file (page 404), this parameter specifies hosts that are not allowed to connect to
the server. Overridden by hosts specified in hosts allow. If you specify ALL in this file, remember to include the
local machine (127.0.0.1) in hosts allow. Specify hosts in the same manner as hosts.deny.
Default: none (no hosts excluded)
Note
map to guest
Defines when a failed login is mapped to the guest account.
Never: Allows guest to login only when the user explicitly provides guest as the username and a blank password.
Bad User: Treats any attempt to log in as a user who does not exist as a guest login. This parameter is a security
risk because it allows a malicious user to retrieve a list of users on the system quickly.
Bad Password: Silently log in as guest any user who incorrectly enters his or her password. This parameter may
confuse a user when she mistypes her password and is unkowingly logged in as guest because she will suddenly
see fewer shares than she is used to.
382
Draft
Parameters in the smbd.conf File
Draft
Default: Never
Note
passwd chat
The chat script that Samba uses to converse with the passwd program. If the script is not followed, Samba does not
change the password. Used only when unix password sync is set to YES.
Default: *new*password* %n\n ... *changed*,
Red Hat: *New*UNIX*password* %n\n ... *successfully*
Note
passwd program
The program Samba uses to set Linux passwords. Samba replaces %u with the user’s username.
Default: /usr/bin/passwd %u
Note
security
Specifies if and how clients transfer user and password information to the server. Choose one of the following:
USER: Causes Samba to require a username and password from Users or Windows when logging in on the Samba
server. With this setting you can use
•
username map (page 688) to map usernames to other names
•
encrypt passwords (page 686) to encrypt passwords (recommended)
•
guest account (page 686) to map users to the guest account
SHARE: Causes Samba not to authenticate clients on a per-user basis. Instead, Samba uses the system found in
Windows 9x where each share can have an individual password for either read or full access. This option is not
compatible with more recent versions of Windows.
SERVER: Causes Samba to use another SMB server to validate usernames and passwords. Failing remote validation,
the local Samba server tries to validate as though security were set to USER.
DOMAIN: Samba passes an encrypted password to a Windows NT Domain Controller for validation.
Note
FEDORA
ADS: Samba 3 adds this parameter, which instructs Samba to use an Active Directory Server for authentication,
allowing a Samba server to participate as a native Active Directory member. (Active Directory is the centralized
information system that Windows 2000 and later uses. It replaces Windows Domains used by Windows NT and
earlier.)
Default: USER
383
Draft
Parameters in the smbd.conf File
Draft
Note
unix password sync
YES causes Samba to change a user’s Linux password when the associated user changes the encrypted Samba
password.
Default: NO
Note
update encrypted
YES allows users to migrate from cleartext passwords to encrypted passwords without logging in on the server and
using smbpasswd. To migrate users, set to YES and set encrypt passwords to NO. As each user logs in on the
server with a cleartext Linux password, smbpasswd encrypts and stores the password in /etc/samba/smbpasswd.
Set to NO and set encrypt passwords to YES after all users have been converted.
Default: NO
Note
username map
The name of the file that maps usernames from a client to those of the server. Each line of the map file starts with
a server username, followed by a SPACE, an equal sign, another SPACE, and one or more SPACE-separated client
usernames. An asterisk (*) on the client side matches any client username. This file frequently maps Windows
usernames to Linux usernames and/or maps multiple Windows usernames to a single Linux username to facilitate
file sharing. A sample map file is shown following:
$ cat /etc/samba/smbusers
# Unix_name = SMB_name1 SMB_name2 ...
root = administrator admin
nobody = guest
sam = sams
Default: no map
Red Hat /etc/samba/smbusers
Logging Parameters
Note
log file
The name of the Samba log file; Samba replaces %m with the name of the client system, allowing you to generate
a separate log file for each client.
Default: /var/log/samba/%m.log
384
Draft
Parameters in the smbd.conf File
Draft
Note
log level
Sets the log level, with 0 (zero) being off and higher numbers being more verbose.
Default: 0 (off)
Note
max log size
An integer specifying the maximum size of the log file in kilobytes. A 0 (zero) specifies no limit. When a file reaches
this size, Samba appends a .old to the filename and starts a new log, deleting any old log file.
Default: 5000, Red Hat 50
Browser Parameters
The domain master browser is the system that is responsible for maintaining the list of machines on a network used
when browsing a Windows Network Neighborhood or My Network Places. SMB uses weighted elections every
11–15 minutes to determine which machine will be the domain master browser.
Whether a Samba server wins this election depends on two parameters: First, setting domain master to YES instructs
the Samba server to enter the election. Second, the os level determines how much weight the Samba server’s vote
receives. Setting os level to 2 should cause the Samba server to win against any Windows 9x machines. NT Server
series domain controllers, including Windows 2000, XP, and 2003, use an os level of 32. The maximum setting
for os level is 255, although setting it to 65 should ensure that the Samba server wins.
Note
domain master
YES causes nmbd to attempt to be the domain master browser. If a domain master browser exists, then local
master browsers will forward copies of their browse lists to it. If there is no domain master browser, then browse
queries may not be able to cross subnet boundaries. A Windows PDC (Primary Domain Controller) will always
try to become the domain master and may behave unexpectedly if it fails. Refer to the preceding discussion.
Default: AUTO
Note
local master
YES causes nmbd to enter elections for the local master browser on a subnet. A local master browser stores a cache
of the NetBIOS names of entities on the local subnet, allowing browsing. Windows computers automatically enter
elections; for browsing to work, you need on a network at least one Windows computer or one Samba server with
local master set to YES. It is poor practice to set local master to NO. If you do not want a computer to act as a
local master, set its os level to a lower number, allowing it to be used as the local master, if all else fails.
Default: YES
385
Draft
Parameters in the smbd.conf File
Draft
Note
os level
An integer that controls how much Samba advertises itself for browser elections and how likely nmbd is to become
the local master browser for its workgroup. A higher number increases the chances of the local server becoming
the local master browser. Refer to the discussion at the beginning of this section.
Default: 20
Note
preferred master
YES forces nmbd to hold an election for local master and enters the local system with a slight advantage. With
domain master set to YES, this parameter helps ensure that the local Samba server becomes the domain master.
Setting this parameter to YES on more than one server causes the servers to keep competing to become master,
generating a lot of network traffic and sometimes causing unpredictable results. A Windows PDC (Primary Domain
Controller) automatically acts as if this parameter is set.
Default: AUTO
Communication Parameters
Note
dns proxy
When acting as a WINS server, YES causes nmbd to use DNS if NetBIOS resolution fails.
Default: YES
Red Hat: NO
Note
socket options
Tunes network parameters used when exchanging data with a client. The Red Hat setting is appropriate in most
cases.
Default: TCP_NODELAY,
Red Hat: TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192
Note
wins server
The IP address of the WINS server that nmbd should register with.
Default: not enabled
386
Draft
Parameters in the smbd.conf File
Draft
Note
wins support
YES specifies that nmbd act as a WINS server.
Default: NO
Share Parameters
Each of the following parameters can appear many times in smb.conf, once in each share definition.
Note
available
YES specifies the share as active. Set to NO to disable the share, but continue logging requests for it.
Note
browseable
Determines whether the share can be browsed, for example, in Windows My Network Places.
Default: YES
Red Hat: NO
Note
comment
A description of the share, shown when browsing the network from Windows.
Default: none
Red Hat: various
Note
guest ok
Allows a user who logs in as guest to access this share.
Default: NO
Note
path
The path of the directory that is being shared.
Default: none
Red Hat: various
387
Draft
The [homes] Share: Sharing Users’
Home Directories
Draft
Note
read only
Does not allow write access.
Default: YES
Red Hat: YES
The [homes] Share: Sharing Users’ Home Directories
Frequently, users want to share their Linux home directories with a Windows machine. To make this task easier,
Samba provides the [homes] share. When you define this share, each user’s home directory is shared with the
specified parameters. In most cases, the following parameters are adequate:
[homes]
comment = Home Directories
read only = No
browseable = No
guest ok = No
These settings prevent users other than the owners from browsing home directories, while allowing logged-in
owners full access.
Accessing Linux Shares from Windows
Browsing Shares
To access a share on a Samba server, open My Computer or Explorer on the Windows system and, in the Address
text box, enter \\ followed by the NetBIOS name (or just the hostname if you have not assigned a different NetBIOS
name) of the Samba server. Windows displays the directories that the Linux system is sharing. To view the shares
on the Linux system named bravo, you would enter \\bravo. From this window, you can view and browse the
shares available on the Linux system. If you set a share so that it is not browseable, you need to enter the path of
the share using the format \\servername\sharename .
Mapping a Share
Another way to access a share on a Samba server is by mapping a share: Open My Computer or Explorer on the
Windows system and click Map Network Drive from one of the drop-down menus on the Menubar (it is on the
Tools menu on Windows XP). Windows displays the Map Network Drive window. Select an unused Windows
drive letter from the Drive combo box and enter the Windows path to the share you just created. (When you use
system-config-samba to create a share, the share has the same name as the name of the directory you are sharing.)
The format of the windows path is \\hostname\sharename . For example, to map /tmp on bravo to Windows
drive J, assuming the share is named tmp on the Linux system, select J in the Drive combo box, enter \\bravo\tmp
in the Folder text box, and click Finish. You should be able to access the /tmp directory from bravo as J (tmp)
on the Windows machine. If you cannot map the drive, see “Troubleshooting” on page 694.
388
Draft
Accessing Windows Shares from Linux
Draft
Accessing Windows Shares from Linux
As a client, Samba enables you to view and work with files on a Windows system from a Linux system. This section
discusses several ways of accessing Windows files from Linux.
smbtree: Displaying Windows Shares
The smbtree utility displays a hierarchical diagram of available shares. When you run smbtree, it prompts you for
a password; do not enter a password if you want to browse shares visible to the guest user. The password allows
you to view restricted shares, such as a user’s home directory in the [homes] share. Following is sample output
from smbtree:
$ smbtree
Password:
MGS
\\PB
\\PB\mark
\\PB\MainPrinter
\\PB\ADMIN$
\\PB\IPC$
\\PB\tmp
pb Samba
Home Directories
MainPrinter
IPC Service (pb Samba)
IPC Service (pb Samba)
mgs temp
In the preceding output, MGS is the name of the workgroup, PB is the name of the Windows machine, mark and
tmp are directory shares, and MainPrinter is a shared printer. Workgroup and machine names are always shown
in capitals. Refer to the smbtree man page for more information.
smbclient: Connecting to Windows Shares
The smbclient utility functions similarly to ftp (page 583) and connects to a Windows share; smbclient uses Linuxstyle forward slashes (/) as path separators rather than Windows-style backslashes (\). The next example connects
to one of the shares displayed in the preceding example:
$ smbclient //PB/mark
Password:
Domain=[PB] OS=[Unix] Server=[Samba 3.0.2-7.FC1]
smb: \> ls
.
D
0 Wed Feb 25
..
D
0 Mon Feb 2
.kde
DH
0 Tue Feb 3
.xemacs
DH
0 Mon Feb 2
.bash_logout
H
24 Tue Oct 28
.bash_profile
H
191 Tue Oct 28
.bashrc
H
124 Tue Oct 28
...
15:10:03
12:40:17
22:24:17
10:12:45
06:15:04
06:15:04
06:15:04
2004
2004
2004
2004
2003
2003
2003
You can use most ftp commands from smbclient. Refer to “Tutorial Session” on page 586 for some examples or
give the command help to display a list of commands.
389
Draft
Browsing Windows Networks
Draft
Browsing Windows Networks
Browsing Windows shares using smbtree and smbclient is quite cumbersome compared with the ease of browsing
a network from Windows; Gnome and KDE provide much more user friendly alternatives. From either Konqueror
or Nautilus (the KDE and Gnome file managers), enter smb:/// in the Location bar to browse the Windows shares
on the network (Figure 23-7).
Figure 23-7. Nautilus displaying a Windows share
Because both Konqueror and Nautilus use virtual filesystem add-ons, which are part of the respective desktop environments and not part of the native Linux system, only native Gnome or KDE applications can open files on remote
shares; normal Linux programs cannot. For example, gedit and kedit will be able to open files on remote shares,
while OpenOffice, mplayer, and xedit will not be able to.
smbmount: Mounting Windows Shares
The smbmount utility mounts Windows shares just as mount mounts a Linux directory hierarchy. When you mount
a Windows share, you can write to the files on the share; you cannot write to files on a share using smbclient.
The syntax of a smbmount command is the same as that of a mount command (page 442):
390
Draft
Troubleshooting
Draft
# smbmount //host/share dir
where host is the name of the system that the share is on, share is the name of the Windows share that you want
to mount, and dir is the absolute pathname of the Linux directory that you are mounting the share on.
The following command, run as root, mounts the share used in the preceding example on the /share directory:
# smbmount //PB/mark /share -o username=mark
Password:
# ls /share
Desktop
mansmbconf
httpd.conf
NVIDIA-Linux-x86-1.0-5336-pkg1.run
smb.conf
smbhold
smbout
x
Troubleshooting
Samba provides three utilities that can help troubleshoot a connection: The smbstatus utility displays a report on
open Samba connections; testparm checks the syntax of /etc/samba/smb.conf and displays its contents; and
testprns checks the validity of the name of a printer.
The following steps can help you narrow down the problem when you cannot get Samba to work.
1.
Restart the smbd and nmbd daemons. Make sure the last two lines of output end with OK.
# /etc/rc.d/init.d smb restart
Shutting down SMB services:
Shutting down NMB services:
Starting SMB services:
Starting NMB services:
2.
[
[
[
[
OK
OK
OK
OK
]
]
]
]
Run testparm to check that your smb.conf file is syntactically correct.
$ testparm
Load smb config files from /etc/samba/smb.conf
Processing section "[homes]"
Processing section "[printers]"
Processing section "[tmp]"
Loaded services file OK.
Press enter to see a dump of your service definitions
...
If you misspell a keyword in smb.conf, you get an error such as the following:
# testparm
Load smb config files from /etc/samba/smb.conf
Unknown parameter encountered: "workgruop"
Ignoring unknown parameter "workgruop"
...
391
Draft
Troubleshooting
Draft
3.
Use ping (page 346) from both sides of the connection to make sure your network is up.
4.
From a Windows command prompt, use net view to display a list of shares available from the server (pb in
this example):
C:>net view \\pb
Shared resources at \\pb
pb Samba
Share name
Type
Used as Comment
-------------------------------------------------------------------MainPrinter Print
MainPrinter
mark
Disk
(UNC)
Home Directories
tmp
Disk
mgs temp
The command completed successfully.
5.
See if you can map the drive from a Windows command prompt. The following command attempts to mount
the share named tmp on pb as drive X:
C:>net use x: \\pb\tmp
The command completed successfully.
6.
From the server, query the nmbd server, using the special name __SAMBA__, for the server’s NetBIOS name.
The –d 2 option turns the debugger on at level 2, which generates a moderate amount of output.
$ nmblookup -d 2 -B pb __SAMBA__
added interface ip=192.168.0.10 bcast=192.168.0.255
nmask=255.255.255.0
querying __SAMBA__ on 192.168.0.10
Got a positive name query response from 192.168.0.10 ( 192.168.0.10 )
192.168.0.10 __SAMBA__<00>
7.
From the server, query the nmbd server for the client’s NetBIOS name. (The machine named jam is the
Windows client.)
$ nmblookup -B jam \*
querying * on 192.168.0.9
192.168.0.9 *<00>
Omit the –B jam option to query for all NetBIOS names.
8.
From the server, use smbclient with the –L option to generate a list of shares offered by the server.
$ smbclient -L pb
Password:
Domain=[PB] OS=[Unix] Server=[Samba 3.0.2-7.FC1]
392
Draft
Chapter Summary
Draft
Sharename
Type
Comment
-----------------tmp
Disk
mgs temp
IPC$
IPC
IPC Service (pb Samba)
ADMIN$
IPC
IPC Service (pb Samba)
MainPrinter
Printer
MainPrinter
mark
Disk
Home Directories
Domain=[PB] OS=[Unix] Server=[Samba 3.0.2-7.FC1]
Server
Comment
--------------Workgroup
--------MGS
9.
Master
------TUNAER
To query for the master browser from the server, run nmblookup with the –M option followed by the name
of the workgroup:
$ nmblookup -M MGS
querying MGS on 192.168.0.255
192.168.0.8 MGS<1d>
Chapter Summary
Samba is a suite of programs that enables Linux to work with Windows to share directories and printers. A directory
or printer that is shared between Linux and Windows systems is called a share. In order to access a share on a Linux
system, a Windows user must supply a username and password. Usernames must correspond to Linux usernames
either directly or as mapped by the /etc/samba/smbusers file. Samba passwords are generated by smbpasswd and
kept in /etc/samba/smbpasswd.
The main Samba configuration file is /etc/samba/smb.conf, which you can edit using a text editor, swat (a Webbased administration utility), or system-config-samba (a minimal configuration GUI). The swat utility is a powerful
configuration tool that provides integrated online documentation and clickable default values to help set up Samba.
From a Windows machine, you can access a share on a Linux Samba server by opening My Computer or Explorer
and, in the Address text box, entering \\ followed by the name of the server. Windows displays the shares on the
server and you can work with them as though they were Windows files.
From a Linux system, you can use any of several Samba tools to access Windows shares. These tools include smbtree
(displays shares), smbclient (similar to ftp), and smbmount (mounts shares). In addition, you can enter smb:/// in
the Location bar of Konqueror or Nautilus and browse the shares.
Samba is a suite of tools that makes heterogeneous computing a reality.
Exercises
1. What two daemons are part of the Samba suite? What does each do?
2. What steps are required for mapping a Windows user to a Linux user?
3. How would you allow access to swat only from machines on the 192.168.1.0/8 subnet?
393
Draft
Advanced Exercises
Draft
4. What is the purpose of the [homes] share?
Advanced Exercises
5. Describe how Samba’s handling of users differs from that of NFS.
6. What configuration changes would you need to apply to your routers if you wanted to allow SMB/CIFS browsing
across multiple subnets without configuring master browsers?
7. How could you use SWAT securely from a remote location?
8. WINS resolution allows hosts to define their own names. Suggest a way that Samba could be used to assign
names from a centralized list.
394
Draft
Draft
Index
395
396
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement