Original article is from fedoramagazine.org
Our example router in this example has two network interfaces:
eth0: public internet connectivity eth1: private LAN (192.168.3.1/24)
We want machines on the private LAN to route their traffic through the router to the public internet via NAT. Also, we want clients on the LAN to get their IP addresses assigned automatically. Network configuration
All of the systemd-networkd configuration files live within /etc/systemd/network and we need to create that directory:
# mkdir /etc/systemd/network
We need to write a network configuration file for our public interface that systemd-networkd can read. Open up /etc/systemd/network/eth0.network and write these lines:
[Match] Name=eth0 [Network] Address=PUBLIC_IP_ADDRESS/CIDR Gateway=GATEWAY DNS=37.235.1.174 DNS=64.6.64.6 IPForward=yes
If we break this configuration file down, we’re telling systemd-networkd to apply this configuration to any devices that are called eth0. Also, we’re specifying a public IP address and CIDR mask (like /24 or /22) so that the interface can be configured. The gateway address will be added to the routing table. We’ve also provided DNS servers to use with systemd-resolved (more on that later).
I added
IPForward=yes
so that systemd-networkd will automatically enable forwarding for the interface via sysctl. (That always seems to be the step I forget when I build a Linux router.)
Let’s do the same for our LAN interface. Create this configuration file and store it as /etc/systemd/network/eth1.network:
[Match] Name=eth1 [Network] Address=192.168.3.1/24 IPForward=yes
We don’t need to specify a gateway address here because this interface will be the gateway for the LAN.
If we’re planning to use systemd-networkd, we need to ensure that it runs instead of traditional network scripts or NetworkManager:
# systemctl disable network # systemctl disable NetworkManager # systemctl enable systemd-networkd
Also, let’s be sure to use systemd-resolved to handle our /etc/resolv.conf:
# systemctl enable systemd-resolved # systemctl start systemd-resolved # rm -f /etc/resolv.conf # ln -s /run/systemd/resolve/resolv.conf /etc/resolv.conf
We’re now set to reboot! It’s possible to bring up systemd-networkd without rebooting but I’d rather verify with a reboot now than get goosed with a broken network after a reboot later.
Once your router is back up, run
# networkctl
and verify that you have routable in the output for both interfaces:
[root@router ~]# networkctl IDX LINK TYPE OPERATIONAL SETUP 1 lo loopback carrier unmanaged 2 eth0 ether routable configured 3 eth1 ether routable configured
Now that both network interfaces are online, we need something to tell our clients about the IP configuration they should be using. There are plenty of good options here, but I prefer dnsmasq. It has served me well over the years and it provides some handy features along with DHCP, such as DNS caching, TFTP and IPv6 router announcements.
Let’s install dnsmasq and enable it at boot:
# dnf -y install dnsmasq # systemctl enable dnsmasq
Open /etc/dnsmasq.conf in your favorite text editor and edit a few lines:
dhcp-authoritative
This tells dnsmasq that it’s the exclusive DHCP server on the network and that it should answer all requests
interface=and add eth1 on the end (should look like when you’re done) interface=eth1
dhcp-range line and change it to: dhcp-range=192.168.3.50,192.168.3.150,12h
We’re giving clients 12 hour leases on 192.168.3.0/24
systemctl start dnsmasq
We’re almost done! Now it’s time to tell iptables to masquerade any packets from our LAN to the internet. But wait, it’s 2015 (or later) and we have tools like firewall-cmd to do that for us in Fedora.
Let’s enable masquerading, allow DNS, and allow DHCP traffic. We can then make the state permanent:
#firewall-cmd --add-masquerade # firewall-cmd --add-service=dns --add-service=dhcp # firewall-cmd --runtime-to-permanent
Put a client machine on your LAN network and you should be able to ping some public sites from the client:
[root@client ~]# ping -c 4 icanhazip.com PING icanhazip.com (104.238.141.75) 56(84) bytes of data. 64 bytes from lax.icanhazip.com (104.238.141.75): icmp_seq=1 ttl=52 time=69.8 ms 64 bytes from lax.icanhazip.com (104.238.141.75): icmp_seq=2 ttl=52 time=69.7 ms 64 bytes from lax.icanhazip.com (104.238.141.75): icmp_seq=3 ttl=52 time=69.6 ms 64 bytes from lax.icanhazip.com (104.238.141.75): icmp_seq=4 ttl=52 time=69.7 ms --- icanhazip.com ping statistics --- 4 packets transmitted, 4 received, 0% packet loss, time 3005ms rtt min/avg/max/mdev = 69.659/69.758/69.874/0.203 ms
If you need to adjust your network configuration, just run </code># systemctl restart systemd-networkd</code> afterwards. I’ve found that it’s quite intelligent about the network devices and it won’t reconfigure anything that hasn’t changed.
The networkctl command is very powerful. Check out the status and lldp functions to get more information about your network devices and the networks they’re connected to.
When something goes wrong, look in your systemd journal:
[root@router ~]# journalctl -u systemd-networkd -- Logs begin at Fri 2015-07-31 01:22:38 UTC, end at Fri 2015-07-31 02:11:24 UTC. -- Jul 31 01:46:14 router systemd[1]: Starting Network Service... Jul 31 01:46:14 router systemd-networkd[286]: Enumeration completed Jul 31 01:46:14 router systemd[1]: Started Network Service. Jul 31 01:46:15 router systemd-networkd[286]: eth1 : link configured Jul 31 01:46:15 router systemd-networkd[286]: eth0 : gained carrier Jul 31 01:46:15 router systemd-networkd[286]: eth0 : link configured Jul 31 01:46:16 router systemd-networkd[286]: eth1 : gained carrier