Setting up a Linux system as a firewall for a small network with DMZ is a fairly straight-forward task if you understand the basics and follow the documentation.

This guide doesn't attempt to acquaint you with all of the features of Shorewall. It rather focuses on what is required to configure Shorewall in one of its more popular configurations:

Here is a schematic of a typical installation.

Figure 1. schematic of a typical installation

[root@gateway root]# which ip
/sbin/ip
[root@gateway root]#

If you have an ADSL Modem and you use PPTP to communicate with a server in that modem, you must make the changes recommended here in addition to those detailed below. ADSL with PPTP is most commonly found in Europe, notably in Austria.

The configuration files for Shorewall are contained in the directory /etc/shorewall -- for simple setups, you will only need to deal with a few of these as described in this guide.

After you have installed Shorewall, locate the three-interface Sample configuration:

As each file is introduced, I suggest that you look at the actual file on your system and that you look at the man page for that file. For example, to look at the man page for the /etc/shorewall/zones file, type man shorewall-zones at a shell prompt.

Note: Beginning with Shorewall 4.4.20.1, there are versions of the sample files that are annotated with the corresponding manpage contents. These files have names ending in '.annotated'. You might choose to look at those files instead.

Shorewall views the network where it is running as being composed of a set of zones. In the three-interface sample configuration, the following zone names are used:

#ZONE   TYPE   OPTIONS                 IN_OPTIONS              OUT_OPTIONS
fw      firewall
net     ipv4
loc     ipv4
dmz     ipv4

Zone names are defined in /etc/shorewall/zones.

Note that Shorewall recognizes the firewall system as its own zone. When the /etc/shorewall/zones file is processed, he name of the firewall zone is stored in the shell variable $FW which may be used throughout the Shorewall configuration to refer to the firewall zone.

Rules about what traffic to allow and what traffic to deny are expressed in terms of zones.

For each connection request entering the firewall, the request is first checked against the /etc/shorewall/rules file. If no rule in that file matches the connection request then the first policy in /etc/shorewall/policy that matches the request is applied. If there is a common action defined for the policy in /etc/shorewall/actions or /usr/share/shorewall/actions.std then that action is performed before the action is applied. The purpose of the common action is two-fold:

The /etc/shorewall/policy file included with the three-interface sample has the following policies:

#SOURCE    DEST        POLICY      LOGLEVEL    LIMIT
loc        net         ACCEPT
net        all         DROP        info
all        all         REJECT      info

#SOURCE    DEST        POLICY      LOGLEVEL    LIMIT
$FW        net         ACCEPT

The above policy will:

The word info in the LOG LEVEL column for the DROP and REJECT policies indicates that packets dropped or rejected under those policies should be logged at that level.

Some people want to consider their firewall to be part of their local network from a security perspective. If you want to do this, add these two policies:

#SOURCE    DEST        POLICY      LOGLEVEL    LIMIT
loc        $FW         ACCEPT
$FW        loc         ACCEPT

It is important to note that Shorewall policies (and rules) refer to connections and not packet flow. With the policies defined in the /etc/shorewall/policy file shown above, connections are allowed from the loc zone to the net zone even though connections are not allowed from the loc zone to the firewall itself.

At this point, edit your /etc/shorewall/policy file and make any changes that you wish.

Figure 2. DMZ

The firewall has three network interfaces. Where Internet connectivity is through a cable or DSL “Modem”, the External Interface will be the Ethernet adapter that is connected to that “Modem” (e.g., eth0) unless you connect via Point-to-Point Protocol over Ethernet (PPPoE) or Point-to-Point Tunneling Protocol (PPTP) in which case the External Interface will be a ppp interface (e.g., ppp0). If you connect via a regular modem, your External Interface will also be ppp0. If you connect using ISDN, you external interface will be ippp0.

root@lists:~# ip route ls
192.168.1.1 dev eth0  scope link 
192.168.2.2 dev tun0  proto kernel  scope link  src 192.168.2.1 
192.168.3.0/24 dev br0  proto kernel  scope link  src 192.168.3.254 
10.13.10.0/24 dev tun1  scope link 
192.168.2.0/24 via 192.168.2.2 dev tun0 
192.168.1.0/24 dev br0  proto kernel  scope link  src 192.168.1.254 
206.124.146.0/24 dev eth0  proto kernel  scope link  src 206.124.146.176 
10.10.10.0/24 dev tun1  scope link 
default via 206.124.146.254 dev eth0 
root@lists:~# 

If your external interface is ppp0 or ippp0 then you will want to set CLAMPMSS=yes in /etc/shorewall/shorewall.conf.

Your Local Interface will be an Ethernet adapter (eth0, eth1 or eth2) and will be connected to a hub or switch. Your local computers will be connected to the same switch (note: If you have only a single local system, you can connect the firewall directly to the computer using a cross-over cable).

Your DMZ Interface will also be an Ethernet adapter (eth0, eth1 or eth2) and will be connected to a hub or switch. Your DMZ computers will be connected to the same switch (note: If you have only a single DMZ system, you can connect the firewall directly to the computer using a cross-over cable).

The Shorewall three-interface sample configuration assumes that the external interface is eth0, the local interface is eth1 and the DMZ interface is eth2. If your configuration is different, you will have to modify the sample /etc/shorewall/interfaces file accordingly. While you are there, you may wish to review the list of options that are specified for the interfaces. Some hints:

Prior to Shorewall 5.1.9, it is also required to change the snat and stoppedrules file, to replace eth0 with the name of your external interface and eth1 with the name of your local interface.

Before going further, we should say a few words about Internet Protocol (IP) addresses. Normally, your ISP will assign you a single Public IP address. This address may be assigned via the Dynamic Host Configuration Protocol (DHCP) or as part of establishing your connection when you dial in (standard modem) or establish your PPP connection. In rare cases, your ISP may assign you a static IP address; that means that you configure your firewall's external interface to use that address permanently. Regardless of how the address is assigned, it will be shared by all of your systems when you access the Internet. You will have to assign your own addresses for your internal network (the local and DMZ Interfaces on your firewall plus your other computers). RFC 1918 reserves several Private IP address ranges for this purpose:

10.0.0.0    - 10.255.255.255
172.16.0.0  - 172.31.255.255
192.168.0.0 - 192.168.255.255

You will want to assign your local addresses from one sub-network or subnet and your DMZ addresses from another subnet. For our purposes, we can consider a subnet to consists of a range of addresses x.y.z.0 - x.y.z.255. Such a subnet will have a Subnet Mask of 255.255.255.0. The address x.y.z.0 is reserved as the Subnet Address and x.y.z.255 is reserved as the Subnet Broadcast Address. In Shorewall, a subnet is described using Classless InterDomain Routing (CIDR) notation with consists of the subnet address followed by /24. The 24 refers to the number of consecutive “1” bits from the left of the subnet mask.

It is conventional to assign the internal interface either the first usable address in the subnet (10.10.10.1 in the above example) or the last usable address (10.10.10.254).

One of the purposes of subnetting is to allow all computers in the subnet to understand which other computers can be communicated with directly. To communicate with systems outside of the subnetwork, systems send packets through a gateway (router).

Your local computers (Local Computers 1 & 2) should be configured with their default gateway set to the IP address of the firewall's internal interface and your DMZ computers (DMZ Computers 1 & 2) should be configured with their default gateway set to the IP address of the firewall's DMZ interface.

The foregoing short discussion barely scratches the surface regarding subnetting and routing. If you are interested in learning more about IP addressing and routing, I highly recommend “IP Fundamentals: What Everyone Needs to Know about Addressing & Routing”, Thomas A. Maufer, Prentice-Hall, 1999, ISBN 0-13-975483-0.

The remainder of this guide will assume that you have configured your network as shown here:

Figure 3. DMZ

The default gateway for the DMZ computers would be 10.10.11.254 and the default gateway for the Local computers would be 10.10.10.254.

Warning

Your ISP might assign your external interface an RFC 1918 address. If that address is in the 10.10.10.0/24 subnet then you will need to select a DIFFERENT RFC 1918 subnet for your local network and if it is in the 10.10.11.0/24 subnet then you will need to select a different RFC 1918 subnet for your DMZ.

The addresses reserved by RFC 1918 are sometimes referred to as non-routable because the Internet backbone routers don't forward packets which have an RFC-1918 destination address. When one of your local systems (let's assume local computer 1) sends a connection request to an Internet host, the firewall must perform Network Address Translation (NAT). The firewall rewrites the source address in the packet to be the address of the firewall's external interface; in other words, the firewall makes it look as if the firewall itself is initiating the connection. This is necessary so that the destination host will be able to route return packets back to the firewall (remember that packets whose destination address is reserved by RFC 1918 can't be routed across the Internet). When the firewall receives a return packet, it rewrites the destination address back to 10.10.10.1 and forwards the packet on to local computer 1.

On Linux systems, the above process is often referred to as IP Masquerading and you will also see the term Source Network Address Translation (SNAT) used. Shorewall follows the convention used with Netfilter:

In Shorewall, both Masquerading and SNAT are configured with entries in the /etc/shorewall/masq file (/etc/shorewall/snat when running Shorewall 5.0.14 or later).

If your external firewall interface is eth0 then you do not need to modify the file provided with the sample. Otherwise, edit /etc/shorewall/masq or /etc/shorewall/snat and change it to match your configuration.

If, in spite of all advice to the contrary, you are using this guide and want to use one-to-one NAT or Proxy ARP for your DMZ, you will need to modify the SOURCE column to list just your local interface (10.10.10.0/24 in the above example).

If your external IP is static then, if you are running Shorewall 5.0.13 or earlier, you can enter our static IP in the third column in the /etc/shorewall/masq entry if you like although your firewall will work fine if you leave that column empty (Masquerade). Entering your static IP in column 3 (SNAT) makes the processing of outgoing packets a little more efficient.

When running Shorewall 5.0.14 or later, the rule in /etc/shorewall/snat must be change from a MASQUERADE rule to an SNAT rule.

#ACTION                      SOURCE                DEST         PROTO      PORT
SNAT(static-ip)              ...

If you are using the Debian package, please check your shorewall.conf file to ensure that the following is set correctly; if it is not, change it appropriately:

Shorewall does not maintain a log itself but rather relies on your system's logging configuration. The following commands rely on knowing where Netfilter messages are logged:

It is important that these commands work properly because when you encounter connection problems when Shorewall is running, the first thing that you should do is to look at the Netfilter log; with the help of Shorewall FAQ 17, you can usually resolve the problem quickly.

The Netfilter log location is distribution-dependent:

If you are running a distribution that logs netfilter messages to a log other than /var/log/messages, then modify the LOGFILE setting in /etc/shorewall/shorewall.conf to specify the name of your log.

Beginning in Shorewall 4.4.7, /etc/shorewall/shorewall.conf contains a LOAD_HELPERS_ONLY option which is set to Yes in the samples. This causes Shorewall to attempt to load the modules listed in /usr/share/shorewall/helpers. In addition, it sets sip_direct_media=0 when loading the nf_conntrack_sip module. That setting is somewhat less secure than sip_direct_media=1, but it generally makes VOIP through the firewall work much better.

The modules in /usr/share/shorewall/helpers are those that are not autoloaded. If your kernel does not support module autoloading and you want Shorewall to attempt to load all netfilter modules that it might require, then set LOAD_HELPERS_ONLY=No. That will cause Shorewall to try to load the modules listed in /usr/share/shorewall/modules. That file does not set sip_direct_media=0.

If you need to modify either /usr/share/shorewall/helpers or /usr/share/shorewall/modules then copy the file to /etc/shorewall and modify the copy.

Modify the setting of LOAD_HELPER_ONLY as necessary.

One of your goals will be to run one or more servers on your DMZ computers. Because these computers have RFC-1918 addresses, it is not possible for clients on the Internet to connect directly to them. It is rather necessary for those clients to address their connection requests to your firewall who rewrites the destination address to the address of your server and forwards the packet to that server. When your server responds, the firewall automatically performs SNAT to rewrite the source address in the response.

The above process is called Port Forwarding or Destination Network Address Translation (DNAT). You configure port forwarding using DNAT rules in the /etc/shorewall/rules file.

The general form of a simple port forwarding rule in /etc/shorewall/rules is:

#ACTION   SOURCE    DEST                                          PROTO      DPORT
DNAT      net       dmz:<server local IP address>[:<server port>] <protocol> <port>

If you don't specify the <server port>, it is assumed to be the same as <port>.

#ACTION     SOURCE    DEST                PROTO      DPORT
Web(DNAT)   net    dmz:10.10.11.2  
Web(ACCEPT)  loc    dmz:10.10.11.2

At this point, add the DNAT and ACCEPT rules for your servers.

Normally, when you connect to your ISP, as part of getting an IP address your firewall's Domain Name Service (DNS) resolver will be automatically configured (e.g., the /etc/resolv.conf file will be written). Alternatively, your ISP may have given you the IP address of a pair of DNS name servers for you to manually configure as your primary and secondary name servers. It is your responsibility to configure the resolver in your internal systems. You can take one of two approaches:

If you run the name server on the firewall:

#ACTION     SOURCE    DEST                PROTO      DPORT                     
DNS(ACCEPT) loc       $FW
DNS(ACCEPT) dmz       $FW            

Run name server on DMZ computer 1:

#ACTION     SOURCE    DEST                PROTO      DPORT                      
DNS(ACCEPT) loc       dmz:10.10.11.1
DNS(ACCEPT) $FW       dmz:10.10.11.1             

In the rules shown above, “DNS”(ACCEPT)is an example of a defined macro. Shorewall includes a number of defined macros and you can add your own. To see the list of macros included with your version of Shorewall, run the command shorewall show macros.

You don't have to use defined macros when coding a rule in /etc/shorewall/rules. The first example above (name server on the firewall) could also have been coded as follows:

#ACTION   SOURCE    DEST                PROTO      DPORT                      
ACCEPT    loc       $FW                 tcp        53
ACCEPT    loc       $FW                 udp        53
ACCEPT    dmz       $FW                 tcp        53
ACCEPT    dmz       $FW                 udp        53              

In cases where Shorewall doesn't include a defined macro to meet your needs, you can either define the macro yourself or you can simply code the appropriate rules directly. This page can be of help if you don't know the protocol and port involved.

The three-interface sample includes the following rule:

#ACTION     SOURCE    DEST                PROTO      DPORT                      
DNS(ACCEPT) $FW       net       

That rule allow DNS access from your firewall and may be removed if you commented out the line in /etc/shorewall/policy allowing all connections from the firewall to the Internet.

The sample also includes:

#ACTION     SOURCE    DEST                PROTO      DPORT                      
SSH(ACCEPT) loc       $FW
SSH(ACCEPT) loc       dmz        

Those rules allow you to run an SSH server on your firewall and in each of your DMZ systems and to connect to those servers from your local systems.

If you wish to enable other connections between your systems, the general format for using a defined macro is:

#ACTION         SOURCE        DEST                PROTO      DPORT                      
<macro>(ACCEPT) <source zone> <destination zone>

The general format when not using a defined macro is:

#ACTION   SOURCE        DEST                PROTO      DPORT                      
ACCEPT    <source zone> <destination zone>  <protocol> <port> 

#ACTION     SOURCE    DEST                PROTO      DPORT
DNS(ACCEPT) net       $FW

#ACTION   SOURCE    DEST                PROTO      DPORT                      
ACCEPT    net       $FW                 tcp        53
ACCEPT    net       $FW                 udp        53        

If you don't know what port and protocol a particular application uses, look here.

#ACTION     SOURCE    DEST                PROTO      DPORT                      
SSH(ACCEPT) net       $FW

Bering users will want to add the following two rules to be compatible with Jacques's Shorewall configuration:

#ACTION   SOURCE    DEST                PROTO      DPORT                      
ACCEPT    loc       $FW                 udp        53
ACCEPT    net       $FW                 tcp        80       

Now modify /etc/shorewall/rules to add or remove other connections as required.

The installation procedure configures your system to start Shorewall at system boot but startup is disabled so that your system won't try to start Shorewall before configuration is complete. Once you have completed configuration of your firewall, you can enable Shorewall startup by editing /etc/shorewall/shorewall.conf and setting STARTUP_ENABLED=Yes.

While you are editing shorewall.conf, it is a good idea to check the value of the SUBSYSLOCK option. You can find a description of this option by typing 'man shorewall.conf' at a shell prompt and searching for SUBSYSLOCK

The firewall is started using the shorewall start command and stopped using shorewall stop. When the firewall is stopped, routing is enabled on those hosts that have an entry in /etc/shorewall/stoppedrules (/etc/shorewall/routestopped on Shorewall 4.5.7 and earlier). A running firewall may be restarted using the shorewall reload command. If you want to totally remove any trace of Shorewall from your Netfilter configuration, use shorewall clear.

The three-interface sample assumes that you want to enable routing to/from eth1 (your local network) and eth2 (DMZ) when Shorewall is stopped. If these two interfaces don't connect to your local network and DMZ or if you want to enable a different set of hosts, modify /etc/shorewall/routestopped accordingly.

The firewall will start after your network interfaces have been brought up. This leaves a small window between the time that the network interface are working and when the firewall is controlling connections through those interfaces. If this is a concern, you can close that window by installing the Shorewall Init Package.

Before starting Shorewall for the first time, it's a good idea to stop your existing firewall. On older Redhat/CentOS/Fedora:

service iptables stop

On recent Fedora systems that run systemd, the command is:

systemctl stop iptables.service

If you are running SuSE, use Yast or Yast2 to stop SuSEFirewall.

On other systems that use a classic SysV init system:

/etc/init.d/iptables stop

Once you have Shorewall running to your satisfaction, you should totally disable your existing firewall. On older Redhat/CentOS/Fedora:

chkconfig --del iptables

On Debian systems:

update-rc.d iptables disable

On recent Fedora system running systemd:

systemctl disable iptables.service

At this point, disable your existing firewall service.

I highly recommend that you review the Common Configuration File Features page -- it contains helpful tips about Shorewall features than make administering your firewall easier. Also, Operating Shorewall and Shorewall Lite contains a lot of useful operational hints.