Basic Network Configurations

Networking software is a standard part of the Unix kernel. Although
configuration specifics vary from vendor to vendor there is a level of
continuity across platforms. In addition many vendors provide graphic tools
for configuring most network software. On HP systems SAM can be used. The
Solaris 2.5 CDE (Common Desktop Environment) provides a tool that looks very
much like SAM. SGI provides network setup tools under the System Manager
menu. Even with these tools it is important to understand which files are
involved so they may be edited by hand if the need arises.

In order to configure network software the system administrator must have
the following information:

The full name of the machine
The IP address of the machine
The subnet mask if the network uses subnets
The broadcast address
The default router for the machine
The loopback port address
The IP address of at least one name server if BIND is used

The machine name is a matter of choice. The only real stipulation is that
it be unique on whatever subnet and/or network it is on. At IU, IP addresses
are distributed by the Network Operations Center (NOC). The NOC can be
contacted at 855-3699, [email protected]. An IP address consists of four
octets separated by periods, such as 129.79.123.156. Each octete designates
either a network address or a host address on that network. Which octets
are used to designate the network and host address depends on the network
class. There are three network classes; A, B, and C. IU is a class C network
that uses class B type addresses. The first two octets, 129.79, designate
the IU network. They are common to almost all machines on the IU campus. At
IU third octet is used to designate the subnet. The fourth octet is the host
number and it is unique to each machine. The subnetmask defines which
bytes of
the IP address represent the subnet and the host. If no number is specified a
default value is set according to the class of network that the machine is
on. The subnetmask is conventionally given in hexadecimal. The IU subnetmask
is 0xffffff00, or 255.255.255.0. The first two characters are not part of the
number. They indicate that what follows is a hexadecimal number.

For further recommended information on network addressing see
Article No. 001.

An Ethernet network is type of broadcast network. In a broadcast network
any system can send information and all systems receive every message, although
they discard messages that are not addressed to them. Broadcasting is
accomplished via the broadcast address. This is the address to use for
reaching all other addresses on a network. Any address with the host octet
set to all 1’s, or 255, is by default interpreted as a broadcast address.
So the broadcast address is the address of the subnet, plus 255. If a hosts
IP address is 129.79.149.145, it’s subnet address is
129.79.149 and it’s broadcast address would be
129.79.149.255.

A router is a dedicated computer with at least two Ethernet interfaces. It
decodes and passes network layer packets between different networks. Routing
can be done statically, dynamically or with a combination of the two.
Generally only very small networks use completely static routing. Dynamic
routing is done via the routed daemon which communicates with other
routed programs to learn how to reach another machines on the network.
The default router IP address at IU is the address of the subnet plus 254, or
occasionally 174. If a machines IP address is 129.79.124.35 then the default
router is 129.79.124.254.

Frequently the terms ‘router’ and ‘gateway’ are used interchangeably.
This can be deceiving as they are not necessarily the same thing. In a broad
sense, both have to do with how packets of data find their way between
networks, or between a subnet and a larger network. A gateway is a computer
with at least two Ethernet interfaces, each on a different network. The
gateway acts as a bridge between the two networks. Incoming data packets are
simply passed on to the next network. A router also has multiple Ethernet
interfaces. It looks at incoming data a packets and forwards them to the
appropriate places.

The loopback port is a reserved network interface that a machines uses to
facilitate interprocess communication. It allows the machine to send packet to
itself for testing purposes. The ifconfig command uses this address
for configuration and testing. Every machine that uses TCP/IP has 127.0.0.1
as it’s loopback address.

BIND, the Berkeley Internet Name Domain, is a method of implementing Domain
Name Service (DNS). DNS maps host names to IP addresses. Without it any
command which uses the network and all host table lookup routines must get
this mapping information from a local /etc/hosts file. This requires
maintaining a master file that includes every machine on the network.
While this is a functional solution for small networks it doesn’t work well
for large ones. The use of name servers and BIND eliminates the need for a
single master file.

IU has several name severs. The current IP addresses for these machines
are 198.88.18.1, 198.88.19.1, and 129.79.1.9. This information changes
frequently so it is best to check the
UCS Knowledge Base for the most
up to date information. When BIND is used the /etc/hosts file
is used to list the loopback address as well as the name and IP address of
the machine. The contents of /etc/hosts look like:

127.0.0.1 localhost 129.79.123.143 peanut.ucs.indiana.edu

This file should be owned by root and have permissions set to 0444 or -r–r–r–.

If a workstation is going to be a client of BIND an /etc/resolv.conf
file must be created. This file defines the default domain search list used by
gethostbyname and
gethostbyaddr library routines.
Please note, these man pages came from a Sun machine and may differ slightly
from those found on other workstations.

An /etc/resolv.conf file is set up as follows:

domain                  physics.indiana.edu
search			physics.indiana.edu ucs.indiana.edu indiana.edu .
nameserver		198.88.18.1    #ns.indiana.edu
nameserver		198.88.19.1    #ns2.indiana.edu
nameserver 		129.79.1.9     #iugate.ucs.indiana.edU

The search line is used to tell gethostbyname and
gethostbyaddr what order host resolution services are to be used
in. Host resolution should begin in whatever subnet the machine is on
and move outward to the domain as a whole. This line is not needed for
SunOS machines. The ‘.’ at the end of the search line is optional and
pertains only to machines at IU. It allows non-fully qualified domain
names to be resolved and is leftover from an older DNS system.

By default gethostbyname and gethostbyaddr are configured to
access name information in the following order; NIS, BIND, local.
local refers to the local /etc/hosts file.

On SGIs an additional line is added to /etc/resolv.conf between
the domain and search lines:

hostresorder     local bind

This line is used to change the order in which name information is accessed. There are three possible options; local for local files, bind,and dns.

On HP and SunOS 5.0 (Solaris) information on name service switch is also
kept in another file, /etc/nsswitch.conf
. This file is used to specify which name service to use and in what
order. More than one option can be listed. Examples of /etc/nsswitch.conf
files can be found below.

For an HP using DNS:

hosts: dns

For Solaris 2.5 using primarily local files:

passwd:     files
group:      files
hosts:      files dns
networks:   files
protocols:  files
rpc:        files
ethers:     files
netmasks:   files
bootparams: files
publickey:  files

A default /etc/nsswitch file is created during the installation of
the operating system. On HPs the default name service order is dns, nis,
local. SunOS 5.0 creates a default nsswitch.conf as well as
nsswitch.files, nsswitch.nis and nsswitch.nisplus. The
idea is that the contents of nsswitch.files, nsswitch.nis or
nsswitch.nisplus can be copied into nsswitch.conf according to
what the primary name service is.

The nameserver lines are used to specify the IP address of the
local DNS name servers. Up to three may be listed. Listing more than one can
be helpful if the primary name server goes down.

Unix workstations have at least one Ethernet device, also referred to as an
interface device or just an interface. In terms of networking the IP address
is actually the address of the interface. If a single machine uses more than
one interface each interface is given its own IP address and name.
Workstations all come with pre designated primary Ethernet interface. On SGI
this interface is ec0, on HP it is lan0 and on Suns it is le0. Unless the
workstation is intended for some special purpose, such as routing, only the
primary interface will need to be accounted for when configuring network
software.

On an SGI, HP or Sun system running Solaris most of the network
configuration is done during the installation of the operating system. Linux
varies from distribution to distribution in terms of when networking is
configured. Even if an initial configuration is set up during the installation
the /etc/resolv.conf file needs to be created if using name servers or
BIND. On HP and Solaris systems the /etc/nsswitch.conf file also needs
to be edited. Both these task must be done manually.

Unix uses the ifconfig command to
configure network interfaces. ifconfig is used to disable or enable
network interfaces, as well as setting the IP address, subnet mask. and
broadcast address. It is conventionally run at boot time, but can be run
from the command line to make changes. The syntax for ifconfig is
as follows:

ifconfig interface [family] address up options

interface refers to the Ethernet interface to be
configured. family specifies what protocol layer is to be
configured. An interface can be configured at the hardware level, the
IP level, or some other protocol level. Configuring at any level other
than the IP level is rare, so family should be set to
inet. In some versions of
ifconfig, inet is the default setting that is taken if
the family argument is left out entirely. HP-UX requires that the
family argument be explicitly stated.

address is the IP address of the Ethernet interface.
up turns the interface on and down turns it
off. options conventionally include the subnet mask and broadcast
address.

For essential information on network configurations for Suns please see
Sun-network.
This document is from another university so names of subnets,
netmask and such are not applicable to configuring machines at IU. However,
the steps are the same.

For essential information on networking under Linux see
Configuring TCP/IP on your system.

For essential information on setting up networking under IRIX see
Setting up the Network on System
V and SGI Networking.
For further recommended reading see the
SGI Technical Publications Library. Click on “Bookshelves”, then on
“IRIX 5.3 Bookshelf”. Choose “Administrator Collection”, them move down the
list to “IRIX Advanced Site and Server Administration Guide” and on to
“Chapter 17 Setting Up a Network”.

For essential information on network configurations for HP-UX please read
HP Networking.

Terms used: IP address,
subnet, gateway,
router, BIND,
DNS, Ethernet, datagram,
ARP, fully qualified domain name.