This page contains a simple cookbook for setting up Red Hat 6.X as an internet gateway for a home network or small office network. The instructions are very simplified: no special cases will be discussed, and some assumptions will be made about which network addresses are to be used. The most important assumptions are:
root. You know how to install RPM packages
from your Linux CDROM. If you do not meet any of these assumptions, then this document probably isn't for you.
There is nothing special that you have to do during the installation process. Simply choose an installation which makes sense for you and go for it. This document gives directions on installing everything to do with networking from scratch, to avoid making any assumptions about what was installed or configured during installation. To ensure that things work and there is no confusion about what information goes where, all the configuration will be done by directly editing the system configuration files rather than using the GUI configuration tools provided with Red Hat. On the one hand, this might be a little harder than it has to be; on the other hand, your knowledge will be a good deal more transferable to different distributions and situations (like, what if X doesn't work, or you are setting up a headless server).
The latest version of this document should always be available at http://www.coastnet.com/~pramsey/linux/homenet.html for the HTML version and http://www.coastnet.com/~pramsey/linux/homenet.sgml for the SGML version.
Copyright © 2000, Paul Ramsey.
This manual may be reproduced in whole or in part, without fee, subject to the following restrictions:
Exceptions to these rules may be granted for academic purposes: Write to the author and ask. These restrictions are here to protect us as authors, not to restrict you as learners and educators.
Depending on whether you are using a hub or not, your network topolgy will differ slightly. I am only covering networking with RJ45 cabling (the stuff that looks like phone cables on steroids) and not covering thin coax. With thin coax you can network multiple machines without requiring a hub, but have to be more careful about terminating connections and so on. If you know networking already, these instructions will be largely redundant.
If you have a hub, your network will look like this.
Connect the eth0 card on the Linux box to the cable modem or ADSL box using
the cable supplied by the service provider during their install (or one you
know works in that configuration. This is important because sometimes cable
modems like to be connected with a crossover and sometimes they like a straight-through
through cable: the one the company gives you is the one you want to use.
Connect the eth1 card on the Linux box to the hub with a straight-through
cable. Connect all your other computers to the hub with straight-through cables.
If you do not have a hub, you can still connect one computer to your Linux box, using a crossover cable. Your topology will look like this.
Connect the eth0 card on the Linux box to the cable modem or ADSL box using
the cable supplied by the service provider. Connect the eth1 card on the Linux
box to the other computer with a crossover cable.
This is not a recommended configuration (in this configuration your internal and external networks are on the same physical address, and are therefor theoretically more susceptible to cracking; in reality, the risk is probably very low), but it can be done. Your mileage may vary.
The Linux kernel includes support for "IP aliasing", which allows an ethernet
card to service two different IP addresses simultaneously. The stock kernels
shipped with Red Hat and Mandrake do include support for IP aliasing by default.
To set up your gateway with only one ethernet card, in all the subsequent code
examples, simply replace eth1 with eth0:0.
In a single-card situation, running a DHCP server is not recommended.
Plug all your machines and your cable model or ADSL box into the hub. Cross your fingers and continue.
OK, by now you have installed Linux on your gateway computer. You may have even configured one of your networking cards, and set up connectivity to the Internet. However, we are going to start from scratch and pretend that nothing is configured at all.
Log in as root. All the instructions given in this document assume you
are logged in as root.
The Linux kernel refers to your two ethernet cards as eth0 and eth1, so
that is how I'll be referring to them from now on too. The trouble is, which
one is which? Here's a "simple" way of figuring out, guaranteed to
work at least 50% of the time: lay your computer on the desk with the
motherboard horizontal and the back panel facing you (as you would if you were
going to open it and do some work on it). The leftmost card is eth0 -- you
might want to label it with some masking tape. Now, write down on a piece of
paper the make and model of both eth0 and eth1.
OK, let's see if eth0 and eth1 are recognized automatically by the kernel.
Type ifconfig eth0 and ifconfig eth1. In both cases, if the kernel is recognizing
your card, you should see something like this (bearing in mind that the numbers
and whatnot will be different):
eth0 Link encap: Ethernet HWaddr 00:60:67:4A:02:0A
inet addr:0.0.0.0 Bcast:0.0.0.0 Mask:255.255.255.255
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:466 errors:0 dropped:0 overruns:0 frame:0
TX packets:448 errors:0 dropped:0 overruns:0 carrier:0
collisions:85 txqueuelen:100
Interrupt:10 Base address:0xe400
If the kernel is not recognizing your network card you will see something like this:
eth0: error fetching interface information: Device not found.
If both of your cards were found, skip to the next section. Otherwise, read this section.
OK, so one or both of your cards are not recognized by the kernel. This is not a problem, really. What we're going to have to do is tell the kernel more explicitly how to find your cards. There are lots of twists and turns here, and I'm not going to cover all of them. Remember, when the going gets tough, the tough turn to the Ethernet HOWTO. Here's some summary advice:
/proc/pci and noting down makes and models. Now, since you know what the make and model of eth0 and eth1 are you can
go to the
compatibility page of the
Ethernet HOWTO and look up your card. Take note of the recommended driver,
and any information about special options your card may require. Write it down.
It's time to edit a configuration file! The file we will be editing is
/etc/conf.modules. Open this file up in the text editor of your choice. Because
there are so many possibilities and combinations of things which can go in
this file, I'm going to give my own gateway as an example. I have a PCI 10/100Mb
card based on the VIA Rhine chip, and a plain-jane 10Mb NE2000 ISA clone. I
use the 100Mb card for the internal network and the 10Mb card for the external
connection. My /etc/conf.modules file looks like this:
alias parport_lowlevel parport_pc alias eth0 ne options ne io=0x300 irq=10 alias eth1 via-rhine
My conf.modules file is laid out as follows:
alias eth0 ne) tells the kernel to use the ne driver for
the eth0 device. options ne io=0x300 irq=10) tells the ne driver at which
io address and irq interrupt it will find the ISA card at. If you have ISA
cards you will probably have to use this kind of directive, just replace the
driver, io and irq directives with the correct information for your card. alias eth1 via-rhine) tells the kernel to use the via-rhine
driver for eth1. Because my eth1 card is a PCI card, I do not need to provide
io or irq information: the PCI subsystem configures the device automatically.You will want to ensure that you have alias entries in conf.modules for
both your cards, and correct options lines for all your ISA cards. You may
already have lines in conf.modules for any ethernet cards you configured during
installation.
When you have finished editing conf.modules, try ifconfig eth0 and ifconfig
eth1 again. You may have to apply some trial and error if you are messing with
IO addresses and IRQs without a manufacturers manual.
So, you were really really smart, bought two identical network cards for
your Linux gateway, and now you cannot get them to work together? Do not worry,
getting them to coexist is just a matter of using the correct syntax in /etc/conf.modules.
For this example, the addresses and IRQ numbers are made up, and I will assume
that you have bought a matched pair of NE2000 clones (a common choice). Your
/etc/conf.modules file should look like this:
alias eth0 ne alias eth1 ne options ne io=0x330,0x360 irq=7,9
The addressing options are all given on the same line, and the first number
for each addressing type is for eth0, the second number for eth1.
The "inside network" is the network which all your home/office machines will talk on. The "outside network" is the big scary internet on the other side of the Linux box. By and large, the inside network will be completely insulated from the outside network by the Linux box, which will operate as a medium strength firewall.
Now that your drivers are working and you can see both eth0 and eth1 in
ifconfig it is time to set up the internal home network. I am assuming that
you are going to put your internal network on eth1 and your external device
on eth0.
Your internal network is going to be a private network and will therefor
be on a special network reserved for internal networking: 192.168.1.0. This
is a "private Class C network", in case you want to impress your
friends.
First we need to make sure networking is turned on. Edit the file /etc/sysconfig/network
and make sure the following lines exist:
NETWORKING=yes FORWARD_IPV4=yes
The first line tells Linux that we want the network devices brought up at boot time. The second line tells Linux to enable IP forwarding. This is required when we start configuring masquerading in Section 4.
All the network interface settings for Red Hat and Red Hat derivatives
are contained in files in the /etc/sysconfig/network-scripts directory. Enter
that directory, and create a new file ifcfg-eth1. Put the following into the
ifcfg-eth1 file:
DEVICE=eth1 IPADDR=192.168.1.1 ONBOOT=yes
This code tells the networking scripts to configure eth1 at boot time and
to give it a particular IP address. Activate your network with the new settings
with the following command: /etc/rc.d/init.d/network restart
A DHCP server will automatically configure devices on your internal home network with IP addresses. This is very useful for people with laptops: they can simply plug their machines in and be immediately properly configured. If you do not want a DHCP server on your internal network, just skip to the next section.
First you need to be sure you have the DHCP server installed. Mount your
Linux CD and install the dhcp RPM. Now edit the /etc/dhcpd.conf file and put
the following (and only the following) in it:
subnet 192.168.1.0 netmask 255.255.255.0 {
range 192.168.1.2 192.168.1.60;
default-lease-time 86400;
max-lease-time 86400;
option routers 192.168.1.1;
option ip-forwarding off;
option broadcast-address 192.168.1.255;
option subnet-mask 255.255.255.0;
}
If you are going to set up your Linux box as a caching domain name server, insert the following option:
option domain-name-servers 192.168.1.1;
If you know your outside DNS addresses and you are not going to use the Linux box for DNS, insert the following option, where x.x.x.x and y.y.y.y are IP numbers of the DNS servers:
option domain-name-servers x.x.x.x, y.y.y.y;
If you are going to run Samba file sharing on the Linux box for your Windows computers, add the following options to use the Linux box as the default WINS and browsing server:
option netbios-name-servers 192.168.1.1; option netbios-dd-server 192.168.1.1; option netbios-node-type 8; option netbios-scope "";
Configuring Samba and WINS is well beyond the scope of this document. If you need some pointers, start with the SMB HOWTO and go on from there.
There are still a few more steps. Next, edit the /etc/rc.d/rc.local file
and add the following lines at the end:
# Set up the broadcast device to support DHCP /sbin/route add -host 255.255.255.255 dev eth1
Windows DHCP clients require a particular broadcast address in DHCP responses, and this command forces the Linux TCP/IP stack to produce it.
The next step is to alter the /etc/rc.d/init.d/dhcpd file to use eth1 as
the default device. Replace the line:
daemon /usr/sbin/dhcpd
With:
daemon /usr/sbin/dhcpd eth1
OK, now we are ready to start up DHCP. First start the DHCP server with
the command: /etc/rc.d/init.d/dhcpd start. Now add the special broadcast address
by invoking /sbin/route add -host 255.255.255.255 dev eth1.
Finally, we have to make sure that the DHCP server will start at re-boot
time. Some RPM packages of the DHCP server do not include directives to ensure
the server starts every time, so we'll make sure it gets started by invoking
the command chkconfig dhcpd on.
This command causes RedHat to add the dhcp startup script to the various
runlevel directories under /etc/rc.d. In runlevels 3 and 5 (multiuser console
and multiuser X) the DHCP server is started. In runlevels 0, 1 and 6 (shutdown,
single user and reboot) the DHCP server is stopped.
If you have set DHCP up, configuring your client computers is very easy: just enable DHCP configuration. For Windows computers, this involves opening the "Control Panel" and then the "Networking" option. Find the "TCP/IP" protocol and opt to "Configure" it. Check the box that says to "Configure TCP/IP address automatically", apply your changes, and reboot.
Before you reboot, you might want to type the following command: tail -f
/var/log/messages. This will watch the Linux system log continuously. If all
goes well, when you reboot your Windows computer, you will see it request an
IP address and see the DHCP server respond. Control-C exits the tail -f command.
If you have not set up DHCP, configuration is still fairly easy. Again, open the "Networking" option from the "Control Panel", and choose to configure the TCP/IP protocol. You can assign your client computers any address in the 192.168.1.0 network except 192.168.1.0 (the network address), 192.168.1.255 (the broadcast address) or 192.168.1.1 (your Linux server). Never give two computers the same IP address. Set the "Gateway" address to 192.168.1.1, so that outgoing traffic is routed through your Linux gateway.
The IP Masquerading HOWTO has very detailed information on client configuration in the Configuration Section.
Setting up your Linux box as a caching DNS server will (slightly) improve your netsurfing speed, because commonly used DNS addresses will get cached inside your network and not have to be retrieved from the outside.
If you are interesting in doing full blown DNS, there is a great deal of complexity to be learned. There is a DNS HOWTO available, and the book DNS and BIND is a good (and very comprehensive) paper reference.
In order for your client machines to take advantage of the caching server, they must be configured to use the Linux gateway as their primary DNS server. The DHCP directives given in section 3.2.2 are one way to accomplish this. If you are configuring your client computers by hand, you can change the DNS configurations in the same control tabs you used to set the IP address of the machine.
To install the DNS server, first install the bind RPM, then install the
caching-nameserver RPM. At this point, you are almost ready.
As installed, the caching server will work fine, but if you know the IP
addresses of the internet providers DNS servers you can improve performance
slightly by editing the /etc/named.conf file and adding the following line
after the directory line (where x.x.x.x and y.y.y.y are the primary and secondary
DNS servers):
forwarders { x.x.x.x; y.y.y.y; };
This change makes your DNS server first query the ISPs DNS servers before traversing the internet in search of a given address. The ISPs servers often have a rich cache of DNS information and can provide a much faster answer than your server could.
OK, now you can start your DNS server: /etc/rc.d/init.d/named start
Until we configure the outside network, the DNS service will not work (since
it has to communicate with other DNS servers on the internet), but we can test
out the basic internal connectivity with the ping program.
On one of your client computers, open up a terminal (MSDOS) window, and
type ping 192.168.1.1. This will send out packets to your Linux computer at
regular intervals, and your Linux computer will reflect the packets back. If
things are working right, you should see a set of packet return times.
Now we're ready to configure the outside network. Sometimes this will be difficult, depending on how well your internet provider supports Linux. If you have difficulty, there is an ADSL mini-HOWTO which covers ADSL issues in some detail. If I can find a Cable Modem HOWTO, I will link to it also.
The main problem with most outside connections is getting an IP address.
Some internet providers hand out static IP addresses to cable or ADSL subscribers,
and in that case configuration is easy. However, most providers have now moved
to dynamic configuration via (you guessed it) DHCP. This means that your Linux
computer will likely be a DHCP server on your eth1 interface, and a DHCP client
on your eth0 interface.
Additionally, many providers have taken to providing their services in specialized non-standard ways which assume their customers will be using Windows. Some of those cases will be discussed at the end of section 3.3.2.
If your internet provider has assigned you a static IP address, you are
sitting pretty. First, create a new interface configuration file, /etc/sysconfig/network-scripts/ifcfg-eth0
and put the following in it:
DEVICE=eth0 IPADDR=x.x.x.x NETMASK=y.y.y.y ONBOOT=yes
Just fill in x.x.x.x and y.y.y.y with the values given by your internet
provider. Now edit the /etc/resolv.conf file and enter the following information:
search provider_domain_here nameserver n.n.n.n nameserver m.m.m.m
The provider_domain should be supplied by your internet provider. Also
enter the primary and secondary DNS servers in the n.n.n.n and m.m.m.m lines.
If you have set up the Linux box as a DNS server, you can add a line before
the other nameserver entries: nameserver 127.0.0.1. This will make your Linux
server use the caching server before asking the outside servers for DNS information.
If your internet provider uses DHCP configuration, you need to create a
new interface configuration file, /etc/sysconfig/network-scripts/ifcfg-eth0and put the following in it:
DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes
Now make sure that the dhcpcd client daemon is installed on your system.
Go to your Linux CD and install the dhcpcd RPM package.
It's time to test your new network configuration. Just use the command
/etc/rc.d/init.d/network restart. Now test your outside connection with ping.
Ping a computer on the internet, like www.yahoo.com and see if anything comes
back.
Your situation may differ from the very simple situations described above. Here are some short remarks on the various difficulties and links to more authoritative resources and addressing them. Thanks to John Mellor for supplying the links and impetus for adding this section.
Several ADSL providers (Bell Atlantic, for example) are now insisting that their new customers connect to the service using the "PPP over Ethernet" protocol (PPPoE). To this end, they provide a Windows client program: not very useful for Linux users. Fortunately, PPPoE is a simple protocol and several efforts are underway to support it under Linux, although all are in the early stages.
One of the favorite tricks network providers play is to tie your service to a unique hostname, or even a unique network interface card. This is presumably to keep you from plugging multiple computers into your ethernet port using a hub (of course, by using Linux and Masquerading you're getting the same effect with better security and the cable company has no way of knowing!).
If the provider has given you a hostname and insisted that you set your Windows box with that name in order you use their service, then you'll have to make sure that your Linux box sends in that hostname when requesting an address from the DHCP server.
The Red Hat DHCP client is called when you set the BOOTPROTO to dhcp in
the interface configuration file, but it is called without reference to a hostname.
To call the program with a hostname, in Red Hat 6.1, edit the /etc/sysconfig/network
file, and change the line:
HOSTNAME=
To read this:
HOSTNAME=your_isp_assigned_name
This may not work in some of the Red Hat variants. If it does not work,
check the /sbin/ifup script and see if the call to dhcpcd and pump include
a -h $HOSTNAME switch. If they do not, add them, so the calls look like
/sbin/dhcpcd -i $DEVICE -h $HOSTNAME and /sbin/pump -i $DEVICE
-h $HOSTNAME.
The Road Runner cable service has a special login process which much be run before the server can be used. Fortunately, a detailed Linux Road Runner HOWTO is available.
Now you can admire your work. Type ifconfig to see all your configured
devices. On my gateway computer, it looks like this:
eth0 Link encap:Ethernet HWaddr 00:60:67:4A:02:0A
inet addr:24.65.182.43 Bcast:24.65.182.255 Mask:255.255.255.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:487167 errors:0 dropped:0 overruns:0 frame:0
TX packets:467064 errors:0 dropped:0 overruns:0 carrier:0
collisions:89 txqueuelen:100
Interrupt:10 Base address:0xe400
eth1 Link encap:Ethernet HWaddr 00:80:C8:D3:30:2C
inet addr:192.168.1.1 Bcast:192.168.1.255 Mask:255.255.255.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:284112 errors:0 dropped:0 overruns:0 frame:1
TX packets:311533 errors:0 dropped:0 overruns:0 carrier:0
collisions:37938 txqueuelen:100
Interrupt:5 Base address:0xe800
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
UP LOOPBACK RUNNING MTU:3924 Metric:1
RX packets:12598 errors:0 dropped:0 overruns:0 frame:0
TX packets:12598 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
Note that the eth0 interface has a fancy outside IP address, and the eth1
address has a private internal address.
You can look at the network routes by typing the route command. On my gateway
computer it looks like this:
Kernel IP routing table Destination Gateway Genmask Flags Metric Ref Use Iface 255.255.255.255 * 255.255.255.255 UH 0 0 0 eth1 192.168.1.0 * 255.255.255.0 U 0 0 0 eth1 24.65.182.0 * 255.255.255.0 U 0 0 0 eth0 127.0.0.0 * 255.0.0.0 U 0 0 0 lo default 24.65.182.1 0.0.0.0 UG 0 0 0 eth0
Here we can see the outside network is set up, the inside network is set up, the local device is set up, the special 255.255.255.255 broadcast address is set up, and the default route is set up to point to the internet providers gateway. Perfect!
Now you have the outside, and the inside. All the remains is to open the door between the two. First though, we have to make sure no monsters can get in from the outside.
One of the drawbacks of being permanently connected to the internet via ADSL or cable is that your computer is exposed to potential security threats 24 hours a day, 7 days a week. Using Linux as a gateway reduces the risks, because it hides all your other computers: as far as the rest of the internet is concerned, only your Linux box is available for connections. This means that your network is only as secure as your Linux box, so at this point I'll give a few basic tips to make your box more secure.
First, you need to shut out all the bad guys. To do this, edit the file
/etc/hosts.deny and make sure it looks just like this:
# # hosts.deny This file describes the names of the hosts which are # *not* allowed to use the local INET services, as decided # by the '/usr/sbin/tcpd' server. # # The portmap line is redundant, but it is left to remind you that # the new secure portmap uses hosts.deny and hosts.allow. In particular # you should know that NFS uses portmap! ALL: ALL
This tells the "TCP wrappers" -- which control 95% of
incoming connections -- to deny all connections from all hosts. That's a pretty
good rule! But, it will also keep you from connecting to your Linux box from
inside your home network, which is annoying, so we will make one exception.
Edit the file /etc/hosts.allow and make sure it looks just like this:
# # hosts.allow This file describes the names of the hosts which are # allowed to use the local INET services, as decided # by the '/usr/sbin/tcpd' server. # ALL: 127.0.0.1 ALL: 192.168.1.
This tells the "TCP wrappers" that they can allow connections to all services from the local device (127.0.0.1) and from your home network (192.168.1.).
You have now locked the monsters outside, with a strong padlock. If you want to put up bars and alarm systems, you will have to be alot more sophisticated. The Security HOWTO is a good place to start if you want to learn more about securing your Linux box.
All right! The preliminaries are over, this is where the magic begins. IP masquerading is one of the truly magical services Linux provides. There are commercial products for Windows which do the same thing, but not nearly as efficiently: an ancient 386 can merrily provide IP masquerading services to a whole medium sized office, but cannot even run Windows 95, let alone the add on masquerading package.
Linux has an extremely versatile firewalling capability, and we are going
to be using it in the simplest and crudest possible manner. If you want to
learn how to do firewalling like an expert, you should read both the
Firewalling HOWTO for an
understanding of the theory and the
IPChains HOWTO for instructions on the new ipchains firewalling
tool which ships with the Linux 2.2.X kernel (and by extension Red Hat 6.X).
There is also now a very good
IP Masquerading HOWTO available which has more details on masquerading
tweaks.
Configuring simple masquerading is very very easy once your internal and
external networking is operational. Edit the /etc/rc.d/rc.local file and add
the following lines to the bottom:
# 1) Flush the rule tables. /sbin/ipchains -F input /sbin/ipchains -F forward /sbin/ipchains -F output # 2) Set the MASQ timings and allow packets in for DHCP configuration. /sbin/ipchains -M -S 7200 10 60 /sbin/ipchains -A input -j ACCEPT -i eth0 -s 0/0 68 -d 0/0 67 -p udp # 3) Deny all forwarding packets except those from local network. # Masquerage those. /sbin/ipchains -P forward DENY /sbin/ipchains -A forward -s 192.168.1.0/24 -j MASQ # 4) Load forwarding modules for special services. /sbin/modprobe ip_masq_ftp /sbin/modprobe ip_masq_raudio
The last two lines insert kernel modules which allow FTP and RealAudio to work for computers on the inside network. There are other modules for special services which you can tack on if you need them:
/sbin/modprobe ip_masq_cuseeme) /sbin/modprobe ip_masq_irc) /sbin/modprobe ip_masq_quake) /sbin/modprobe ip_masq_vdolive) Now you're ready to try masquerading! Run the rc.local script with the
command /etc/rc.d/rc.local and you are ready to go! Sit down at one of your
other computers and try some web surfing. With any luck, everything is now
hunky dory.
There are lots and lots of things which can go wrong using a simple document like this, because there are plenty of special cases. The majority of possible problems adhere to the configuration of the internal and external network devices. I will try and respond to people with problems, figure out what went wrong and add links down here so that people with special case problems can track down help. Feel free to contact me at pramsey@refractions.net.
Some portions of ICQ work fine over masquerading. Other portions do not
work well at all. There is a
beta quality ICQ module under development, however, which addresses some
(but not all) of the deficiencies of running ICQ over masquerading. The README
file in the source code distribution describes how to compile the module. Once
you have it compiled and installed, invoke /sbin/modprobe ip_masq_icq.