Internet sharing

This article explains how to share the internet connection from one machine to other(s).

Requirements

The machine acting as server should have an additional network device. That network device requires a functional data link layer to the machine(s) that are going to receive internet access:

To be able to share internet to several machines a switch can provide the data link layer connection.
A wireless device can share access to several machines as well, see Software access point first for this case.
If you are sharing to only one machine, a crossover cable is sufficient. In case one of the two computers' ethernet cards has MDI-X capability, a crossover cable is not necessary and a regular ethernet cable can be used. Executing as root helps to figure it.

Configuration

This section assumes that the network device connected to the client computer(s) is named and the network device connected to the internet as .

All configuration is done on the server computer, except for the final step of #Assigning IP addresses to the client PC(s).

Static IP address

On the server computer, assign a static IPv4 address to the interface connected to the other machines. The first 3 bytes of this address cannot be exactly the same as those of another interface, unless both interfaces have netmasks strictly greater than /24.

# ip link set up dev net0
# ip addr add 192.168.123.100/24 dev net0 # arbitrary address

To have your static IP assigned at boot, you can use a network manager.

Enable packet forwarding

Check the current packet forwarding settings:

# sysctl -a | grep forward

You will note that options exist for controlling forwarding per default, per interface, as well as separate options for IPv4/IPv6 per interface.

Enter this command to temporarily enable packet forwarding at runtime:

# sysctl net.ipv4.ip_forward=1

Warning: If the system uses systemd-networkd to control the network interfaces, a per-interface setting for IPv4 is not possible, i.e. systemd logic propagates any configured forwarding into a global (for all interfaces) setting for IPv4. The advised work-around is to use a firewall to forbid forwarding again on selective interfaces. See the systemd.network(5) manual page for more information. The IPForward=kernel semantics introduced in a previous systemd release 220/221 to honor kernel settings does not apply anymore.

Edit to make the previous change persistent after a reboot for all interfaces:

/etc/sysctl.d/30-ipforward.conf

net.ipv4.ip_forward=1
net.ipv6.conf.default.forwarding=1
net.ipv6.conf.all.forwarding=1

Afterwards it is advisable to double-check forwarding is enabled as required after a reboot.

With iptables

Install the package. Use iptables to enable NAT:

# iptables -t nat -A POSTROUTING -o internet0 -j MASQUERADE
# iptables -A FORWARD -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
# iptables -A FORWARD -i net0 -o internet0 -j ACCEPT

Use instead of if you installed docker.

# iptables -t nat -A POSTROUTING -o internet0 -j MASQUERADE
# iptables -I DOCKER-USER 1 -m conntrack --ctstate RELATED,ESTABLISHED -j ACCEPT
# iptables -I DOCKER-USER 2 -i net0 -o internet0 -j ACCEPT

If connected via PPPoE, clamp mss to pmtu in order to prevent fragmentation:

# iptables -t mangle -A FORWARD -o ppp0 -p tcp -m tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu

Read the iptables article for more information (especially saving the rule and applying it automatically on boot). There is also an excellent guide on iptables Simple stateful firewall.

With nftables

Install the package. To enable NAT with nftables, you will have to create the chain in a new/existing table:

# nft add table inet nat
# nft add chain inet nat postrouting '{ type nat hook postrouting priority 100 ; }'

After that, you have to masquerade the addresses for :

# nft add rule inet nat postrouting oifname internet0 masquerade

You may want to add some more firewall restrictions on the forwarding (assuming the filter table already exists, like configured in nftables#Server):

# nft add chain inet filter forward '{ type filter hook forward priority 0; policy drop; }'
# nft add rule inet filter forward ct state related,established accept
# nft add rule inet filter forward iifname net0 oifname internet0 accept

You can find more information on NAT in nftables in the nftables Wiki. If you want to make these changes permanent, follow the instructions on nftables

With firewalld

Install the package. firewalld is a firewall daemon which relies on nftables or iptables. First change the firewalld zones of network interfaces:

# firewall-cmd --zone=external --change-interface=internet0 --permanent
# firewall-cmd --zone=internal --change-interface=net0 --permanent

Then add a new policy to let traffic flow between the internal and external zone:

# firewall-cmd --permanent --new-policy int2ext
# firewall-cmd --permanent --policy int2ext --add-ingress-zone internal
# firewall-cmd --permanent --policy int2ext --add-egress-zone external
# firewall-cmd --permanent --policy int2ext --set-target ACCEPT
# firewall-cmd --reload

Tip: You can use stricter policy rules than bare ACCEPT as illustrated in the Firewall Rules section of the firewalld concept page

For example, to allow only nodes in 192.168.2.0/24 to access the internet, do:

firewall-cmd --permanent --policy int2ext --add-rich-rule='rule family=ipv4 source address=192.168.2.0/24 accept' Do not forget to relaod rules afterwards:

firewall-cmd --reload

Assigning IP addresses to the client PC(s)

If you are planning to regularly have several machines using the internet shared by this machine, then is a good idea to install a DHCP server, such as dhcpd or dnsmasq. Then configure a DHCP client (e.g. dhcpcd) on every client PC.

Incoming connections to UDP port 67 has to be allowed for DHCP server. It also necessary to allow incoming connections to UDP/TCP port 53 for DNS requests.

# iptables -I INPUT -p udp --dport 67 -i net0 -j ACCEPT
# iptables -I INPUT -p udp --dport 53 -s 192.168.123.0/24 -j ACCEPT
# iptables -I INPUT -p tcp --dport 53 -s 192.168.123.0/24 -j ACCEPT

If you are not planning to use this setup regularly, you can manually add an IP to each client instead.

Manually adding an IP

Instead of using DHCP, a static IP address and a default route via can also be configured manually. There are many tools available to configure the network accordingly. One prominent example of such a tool is , see Network configuration#Network management. Alternatively, one can use a file, see Systemd-networkd#Wired adapter using a static IP to setup a static IP.

Configure a DNS server for each client, see Domain name resolution for details.

That is it. The client PC should now have Internet.

Troubleshooting

If you are able to connect the two PCs but cannot send data (for example, if the client PC makes a DHCP request to the server PC, the server PC receives the request and offers an IP to the client, but the client does not accept it, timing out instead), check that you do not have other iptables rules interfering.

Clients cannot access the internet or cannot connect

Symptoms might also include: Clients get host is down when pinging host, gets when pinging devices outside the LAN (that should be forwarded by NAT), DHCP offers not crossing a bridge, ...

It is known that docker may cause these problems. Simply disabling and solves this problem.

docker github issue.

gollark: Decaying vegetable communications are generally subsumed by their matter reassemblers, and they can physically transfer objects via miniaturized avian/apian carriers or railguns for high-bandwidth communications.

gollark: They have onboard passive-aggression neural networks massively surpassing human performance.

gollark: Oh, absolutely.

gollark: They don't actually have foreheads.

gollark: That would be inefficient, although they can encode some data as very small jitters in their position.