Internet sharing

From ArchWiki
(Redirected from 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 ethtool interface | grep MDI as root helps to figure it.

Configuration

Note: This section assumes that the network device connected to the client computer(s) is named net0 and the network device connected to the internet is internet0.
Tip: You can rename your devices to this scheme using udev#Setting static device names.

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

Warning: Enabling IP forwarding without a properly configured firewall is a security risk.

To check the current packet forwarding settings, run:

# sysctl -a | grep forward

You will note options for controlling forwarding per default, per interface, as well as separate options for IPv4/IPv6 per interface. For detailed description of all available options, see the kernel documentation.

To enable IPv4 and IPv6 packet forwarding, configure sysctl to set these settings:

net.ipv4.ip_forward = 1
net.ipv4.conf.all.forwarding = 1
net.ipv6.conf.all.forwarding = 1
Tip: To enable packet forwarding selectively for a specific interface, use net.ipv[46].conf.interface_name.forwarding=1 instead.
Warning:

This article or section is out of date.

Reason: Most likely outdated. As systemd-networkd 256+ now supports setting IPv4Fowarding on a per link level. Please delete this Note, and below warning on confirmation. (Discuss in Talk:Internet sharing)

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.[1] [2]

To make changes persistent across reboots, see Sysctl#Configuration. You might consider writing settings to a file with a descriptive filename, such as /etc/sysctl.d/30-ipforward.conf.

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

Packet forwarding with systemd-networkd

This article or section is a candidate for merging with systemd-networkd#[Network].

Notes: An arbitrary file naming does not make sense for a generic example. A .network file without a [Match] section does not make sense. Simply say what needs to be set and link to the systemd-networkd#[Network] page. (Discuss in Talk:Internet sharing)

If you are using systemd-networkd to manage your network configuration, you can also persist those settings across reboots:

/etc/systemd/network/20-lan.network
[Network]
IPv4Forwarding=yes
...

This essentially sets the same net.ipv[46].conf.interface_name.forwarding=1 as mentioned in previous section. For IPv6 the configuration is IPv6Forwarding=yes.

Note: As of systemd 256, the IPv6Forwarding option does not behave as documented, see [3].

This sets up packet forwarding for the specific interface only. For internet sharing to properly work, you need to enable packet forwarding on both (all) interfaces where traffic should be routed between. Typically your lan and wan interfaces.

See also systemd-networkd#[Network].

Enable NAT

With iptables

Install the iptables 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
Note: Of course, this also works with a mobile broadband connection (usually called ppp0 on routing PC).

Use -I DOCKER-USER instead of -A FORWARD if you installed docker. [4]

# 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 nftables package. To enable NAT with nftables, you will have to create the postrouting 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 net0 addresses for internet0:

# nft add rule inet nat postrouting oifname internet0 masquerade

Many firewall configurations, like the default /etc/nftables.conf, set the default policy of the 'filter' table's 'forward' chain to 'drop'. In such cases, you will need rules to allow forwarding NAT traffic:

# 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 firewalld 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[5]

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 reload 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. See Router#DNS and DHCP for the available options. Then configure a DHCP client on every client PC, see Network configuration#Network managers.

This article or section needs language, wiki syntax or style improvements. See Help:Style for reference.

Reason: This is not an iptables guide. Expanding the chain with iptables -I might skip other important rules; if you need to script an ON/OFF switch for this, use custom chain with a jump placed carefully in the INPUT chain. (Discuss in Talk:Internet sharing)

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

Alternatively using firewalld

# firewall-cmd --zone=internal --permanent --add-service dns
# firewall-cmd --zone=internal --permanent --add-service dhcp
# firewall-cmd --zone=internal --permanent --add-service dhcpv6

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 192.168.123.100 can also be configured manually. There are many tools available to configure the network accordingly. One prominent example of such a tool is ip(8), see Network configuration#Network management. Alternatively, one can use a .network 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 no route to host 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 docker.service and docker.socket solves this problem.

docker github issue.

See also