This article is a tutorial for turning a computer into an internet gateway/router. To strengthen its security it should not run any services available to the outside world. Towards the LAN, run only gateway specific services; especially do not run httpd, ftpd, samba, nfsd, etc. as those belong on a server in the LAN since they introduce security risks.
This article does not attempt to show how to set up a shared connection between two machines using cross-over cables. For a simple internet sharing solution, see Internet sharing.
- 1 Hardware Requirements
- 2 Network interface configuration
- 3 DNS and DHCP
- 4 Connection sharing
- 5 IPv6 tips
- 6 Optional additions
- 7 See also
- At least 1 GB of hard drive space. The base install will take up around 500MB of space and if you want to use a caching web proxy, you will need to reserve space for the cache as well.
- At least two physical network interfaces: a gateway connects two networks with each other (actually a router can be made using a single physical interface that underlays two VLAN interfaces and is connected to a VLAN-aware switch, so-called router-on-a-stick configuration, but it is not covered in this article). You will need to be able to connect those networks to the same physical computer. One interface must connect to the external network, while the other connects to the internal network.
- A hub, switch or UTP cable: You need a way to connect the other computers to the gateway
Network interface configuration
Persistent interface naming
Systemd automatically chooses unique interface names for all your interfaces. These are persistent and will not change when you reboot. However you might want to rename your interfaces e.g. in order to highlight their different networks to which they connect. Throughout the following sections of this guide, the convention stated below is used:
- intern0: the network card connected to the LAN. On an actual computer it will probably have the name enp2s0, enp1s1, etc.
- extern0: the network card connected to the external network (or WAN). It will probably have the name enp2s0, enp1s1, etc.
You may change the assigned names of your devices via a configuration file using Systemd-networkd described in Systemd-networkd#Renaming an interface or by a udev-rule following Network configuration#Change interface name. Due to the example-rich nature of this article, you might want to choose the names above.
Now you will need to configure the network interfaces. One way to do so, is using netctl profiles. You will need to create two profiles.
Description='Public Interface.' Interface=extern0 Connection=ethernet IP='dhcp'
Description='Private Interface' Interface=intern0 Connection=ethernet IP='static' Address=('10.0.0.1/24')
/27will give you
10.0.0.30. There are many CIDR calculators, online and offline, for example .
SkipNoCarrier=yesin the LAN profile to make sure the connection is enabled even when the guest on LAN is not yet up.
Next, we set up the interfaces with netctl:
# netctl enable extern0-profile # netctl enable intern0-profile
A straight-forward and simple way to configure network interfaces is via systemd-networkd.
- Apply a DHCP client configuration for the
- Apply a static IP configuration for the
See systemd-networkd#Configuration files for configuration details and an overview of the available options. Run
networkctl reload to apply the configuration changes.
Using rp-pppoe, we can connect an ADSL modem to the
extern0 interface of the firewall and have Arch manage the connection. Make sure to put the modem in bridged mode though (either half-bridge or RFC1483), otherwise, the modem will act as a router too. Install the package.
It should be noted that if you use only PPPoE to connect to the internet (i.e. you do not have another WAN port, except for the one that connects to your modem) you do not need to set up the
extern0-profile as the external pseudo-interface will be ppp0.
You can use netctl to setup the PPPoE connection. To get started, do
# cp /etc/netctl/examples/pppoe /etc/netctl/
and start editing. For the interface configuration, choose the interface that connects to the modem. If you only connect to the internet through PPPoE, this will probably be
extern0. Fill in the rest of the fields with your ISP information. See the PPPoE section in the man page for more information on the fields.
DNS and DHCP
Dnsmasq can to be configured to be a DHCP server with a configuration similar to the following:
interface=intern0 # make dnsmasq listen for requests only on intern0 (our LAN) #no-dhcp-interface=intern0 # optionally disable the DHCP functionality of dnsmasq and use systemd-networkd instead expand-hosts # add a domain to simple hostnames in /etc/hosts domain=foo.bar # allow fully qualified domain names for DHCP hosts (needed when # "expand-hosts" is used) dhcp-range=10.0.0.2,10.0.0.255,255.255.255.0,1h # defines a DHCP-range for the LAN: # from 10.0.0.2 to .255 with a subnet mask of 255.255.255.0 and a # DHCP lease of 1 hour (change to your own preferences)
Somewhere below, you will notice you can also add "static" DHCP leases, i.e. assign an IP-address to the MAC-address of a computer on the LAN. This way, whenever the computer requests a new lease, it will get the same IP. That is very useful for network servers with a DNS record. You can also deny certain MACs from obtaining an IP.
DHCP via systemd-networkd
Time to tie the two network interfaces together.
First of all, we need to allow packets to hop from one network interface to the other. For this one needs to have packet forwarding enabled in kernel via Internet sharing#Enable packet forwarding for details.. See
net.**forwarding is set correctly (i.e. is
1), packets still need to be properly sent and received. Hence, it is necessary to translate the IP addresses between the outward facing network and the subnet used locally. The technique is called masquerading . We also need two forwarding rules to keep connections going and enable LAN to WAN forwarding. For this task, we are going to use iptables:
*nat :PREROUTING ACCEPT [0:0] :INPUT ACCEPT [0:0] :OUTPUT ACCEPT [0:0] :POSTROUTING ACCEPT [0:0] -A POSTROUTING -o extern0 -j MASQUERADE COMMIT *filter :INPUT ACCEPT [0:0] :FORWARD ACCEPT [0:0] :OUTPUT ACCEPT [0:0] -A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT -A FORWARD -i intern0 -o extern0 -j ACCEPT COMMIT
If you're connecting via PPPoE, you'll also need to clamp mss to pmtu in order the prevent fragmentation from happening:
*mangle :PREROUTING ACCEPT [0:0] :INPUT ACCEPT [0:0] :FORWARD ACCEPT [0:0] :OUTPUT ACCEPT [0:0] :POSTROUTING ACCEPT [0:0] -A FORWARD -o ppp0 -p tcp -m tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu COMMIT
Start and enable
iptables.service. The router should now be fully functional and route your traffic. However, since it is facing the public Internet, it makes sense to additionally secure it using a Simple stateful firewall.
Amend or create the previously discussed network configuration for
intern0 to include the
IPMasquerade=true option in the
[Network] section. This configuration will implicitly enable packet forwarding on all interfaces, see . See systemd-networkd#[DHCPServer] for an example configuration.
Connection sharing with shorewall
See Shorewall for a detailed configuration guide.
Unique Local Addresses
You can use your router in IPv6 mode even if you do not have an IPv6 address from your ISP. Unless you disable IPv6, all interfaces should have been assigned a unique
For internal networking the block
fc00::/7 has been reserved. These addresses are guaranteed to be unique and non-routable from the open Internet. Addresses that belong to the
fc00::/7 block are called Unique Local Addresses. To get started generate a ULA /64 block to use in your network. For this example we will use
fd00:aaaa:bbbb:cccc::/64. Firstly, we must assign a static IPv6 on the internal interface. Modify the
intern0-profile we created above to include the following line:
This will add the ULA to the internal interface. As far as the router goes, this is all you need to configure.
Global Unicast Addresses
If your ISP or WAN network can access the IPv6 Internet, you can additionally assign global link addresses to your router and propagate them through SLAAC to your internal network. The global unicast prefix is usually either static or provided through prefix delegation.
Static IPv6 prefix
If your ISP has provided you with a static prefix, then edit
/etc/netctl/extern0-profile and simply add the IPv6 and the IPv6 prefix (usually /64) you have been provided
You can use this in addition to the ULA address described above.
Acquiring IPv6 prefix via DHCPv6-PD
If your ISP handles IPv6 via prefix delegation, then you can follow the instructions in the IPv6#Prefix delegation (DHCPv6-PD) on how to properly configure your router. Following the conventions of this article, the WAN interface is
ppp0 if you are connecting through PPPoE) and the LAN interface is
Router Advertisement and Stateless Autoconfiguration (SLAAC)
To properly hand out IPv6s to the network clients, we will need to use an advertising daemon. Follow the details of the main IPv6 article on how to set up
radvd. According to this guide's convention, the LAN-facing interface is
intern0. You can either advertise all prefixes or choose which prefixes will be assigned to the local network.
The above configuration of shorewall does not include UPnP support. Use of UPnP is discouraged as it may make the gateway vulnerable to attacks from within the LAN. However, some applications require this to function correctly.
Read the Shorewall guide on UPnP for more information.
OpenSSH can be used to administer your router remotely. This is useful for running it in headless mode (no monitor or input devices).
Caching web proxy
To use the router as a time server, see System time#Time synchronization for available Network Time Protocol (NTP) server implementations.
Then, configure shorewall or iptables to allow NTP traffic in and out.
Traffic shaping is very useful, especially when you are not the only one on the LAN. The idea is to assign a priority to different types of traffic. Interactive traffic (ssh, online gaming) probably needs the highest priority, while P2P traffic can do with the lowest. Then there is everything in between.