Performing Network Address Translation (NAT)

From nftables wiki
Jump to: navigation, search

The nat chain type allows you to perform NAT. This chain type comes with special semantics:

  • The first packet of a flow is used to look up for a matching rule which sets up the NAT binding for this flow. This also manipulates this first packet accordingly.
  • No rule lookup happens for follow up packets in the flow: the NAT engine uses the NAT binding information already set up by the first packet to perform the packet manipulation.

Adding a NAT rule to a filter type chain will result in an error.

Stateful NAT

The stateful NAT involves the nf_conntrack kernel engine to match/set packet stateful information and will engage according to the state of connections. This is the most common way of performing NAT and the approach we recommend you to follow.

Be aware that with kernel versions before 4.18, you have to register the prerouting/postrouting chains even if you have no rules there since these chain will invoke the NAT engine for the packets coming in the reply direction. The remaining documentation in this article assumes a newer kernel which doesn't require this inconvenience anymore.

Source NAT

If you want to source NAT the traffic that leaves from your local area network to the Internet, you can create a new table nat with the postrouting chain:

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

Then, add the following rule:

% nft add rule nat postrouting ip saddr oif eth0 snat

This matches for all traffic from the network to the interface eth0. The IPv4 address is used as source for the packets that match this rule.

Destination NAT

You need to add the following table and chain configuration:

% nft add table nat
% nft 'add chain nat prerouting { type nat hook prerouting priority -100; }'

Then, you can add the following rule:

% nft 'add rule nat prerouting iif eth0 tcp dport { 80, 443 } dnat'

This redirects the incoming traffic for TCP ports 80 and 443 to


NOTE: masquerade is available starting with Linux Kernel 3.18.

Masquerade is a special case of SNAT, where the source address is automagically set to the address of the output interface. For example:

% nft add rule nat postrouting masquerade

Note that masquerade only makes sense from postrouting chain of NAT type.


NOTE: redirect is available starting with Linux Kernel 3.19.

By using redirect, packets will be forwarded to local machine. Is a special case of DNAT where the destination is the current machine.

% nft add rule nat prerouting redirect

This example redirects 22/tcp traffic to 2222/tcp:

% nft add rule nat prerouting tcp dport 22 redirect to 2222

Note that: redirect only makes sense in a prerouting chain of NAT type.

NAT flags

Since Linux kernel 3.18, you can combine the following flags with your NAT statements:

  • random: randomize source port mapping.
  • fully-random: full port randomization.
  • persistent: gives a client the same source-/destination-address for each connection.

For example:

% nft add rule nat postrouting masquerade random,persistent
% nft add rule nat postrouting ip saddr oif eth0 snat fully-random

Inet family NAT

Since Linux kernel 5.2, there is support for performing stateful NAT in inet family chains. Syntax and semantics are equivalent to ip/ip6 families; the only exception being if IP addresses are specified, a prefix of either ip or ip6 to clarify the address family is required:

% nft add rule inet nat prerouting dnat ip to
% nft add rule inet nat prerouting dnat ip6 to feed::c0fe


You cannot use iptables and nft to perform NAT at the same time before kernel 4.18. So make sure that the iptable_nat module is unloaded:

% rmmod iptable_nat

With later kernels, it is possible to use iptables and nftables nat at the same time. The nat chains are consulted according to their priorities, the first matching rule that adds a nat mapping (dnat, snat, masquerade) is the one that will be used for the connection.

Stateless NAT

This type of NAT just modifies each packet according to your rules without any other state/connection tracking.

This is valid for 1:1 mappings and is faster than stateful NAT. However, it's easy to shoot yourself in the foot. If your environment doesn't require this approach, better stick to stateful NAT.

You have to disable connection tracking for modified packets.

The example below sets IP/port for each packet (also valid in IPv6):

% nft add rule ip raw prerouting ip protocol tcp ip daddr set tcp dport set 10 notrack
% nft add rule ip6 raw prerouting ip6 nexthdr tcp ip6 daddr set fe00::1 tcp dport set 10 notrack

Be sure to check our documentation regarding mangling packets and setting packet connection tracking metainformation.

To use this feature you require nftables >=0.7 and linux kernel >= 4.9.

See also