Setting up NAT64 with tayga on Centos

· by Artem Sidorenko · Read in about 4 min · (773 words)

Maybe you are also playing aroung with IPv6 and want to setup IPv6 only network and asking yourself how to reach the IPv4 Internet? Right, with DNS64 and NAT64. This blog post gives an overview about a such setup on CentOS/RHEL 7 with bind and tayga.

Overview

How does it work? DNS server checks if there is an AAAA entry (IPv6 address). If there is none, but there is an A entry (IPv4 address), DNS server encodes the IPv4 addresses into the configured /96 IPv6 prefix and returns it as AAAA entry. This /96 prefix is routed to the NAT64 gateway. This gateway makes a translation between IPv6 and IPv4.

The DNS64 is supported by the last bind versions and tayga is one of the NAT64 gateway implementations for Linux.

So, you will need a /96 prefix. You can use the reserved prefix 64:ff9b::/96 for this purpose, but keep in mind: you won’t be able to use the translation to the private IPv4 addresses defined by RFC1918. If you want to be able to reach this addresses too, you should use a ULA prefix (and you can also register it in the Sixxs ULA database). In this blog post I assume that you use the 2001:db8:64:ff9b::/96 prefix.

DNS64

Configuration of bind is pretty simple and looks like this:

...

acl "ignore" { // do not apply DNS64 to this addresses
  10/8;
  172.16/12;
};

acl "clients" { // your IPv6 clients
  ::1/128;
  2001:db8:1000:/64; // maybe some network with IPv6 clients
};

options {
...
  listen-on-v6 port 53 { any; };

  dns64 2001:db8:64:ff9b::/96 {
    clients { clients; };
    mapped { !ignore; any; };
    break-dnssec yes; // more on this below
  };
...
};
...

In this configuration we do not use DNS64 for ranges in the ACL ignore (if you are doing dual-stack, this might be the ranges you want to keep IPv4 only). The ACL clients defines the source range of DNS clients, where DNS64 should be used (might be useful for multi-view setups).

Now some words on the break-dnssec option. Jen Linkova did a good explanation of the problem at the RIPE72 (and here a short blog post). If we want to be able to reach the IPv4-only systems with enabled DNSSEC for their DNS entries: we have to break DNSSEC. However, this should be a very small amount of systems on the internet (see the numbers on the slides of Jen Linkova).

NAT64 with tayga

You can install tayga from EPEL repository:

$ yum -y install epel-release
...
$ yum -y install tayga
...

Basic network configuration of CentOS system

Then you should configure the IPv4 and IPv6 transport networks on the CentOS box, we use 192.0.2.0/30 and 2001:db8:20::/64 for this purpose:

  • 192.0.2.1 is the router IP (e.g. your internet router with IPv4 connectivity)
  • 192.0.2.2 is the IP of tayga system
  • 2001:db8:20::1/64 is the router IP (e.g. you router with IPv6 clients)
  • 2001:db8:20::2/64 is the IP of tayga system
# /etc/sysconfig/network-scripts/ifcfg-ens3
DEVICE=ens3
ONBOOT=yes
BOOTPROTO=none
TYPE=Etheret
IPADDR="192.0.2.2"
NETMASK="255.255.255.252"
GATEWAY="192.0.2.1"
IPV6INIT=yes
IPV6FORWARDING=yes
IPV6ADDR=2001:db8:20::2/64
IPV6_DEFAULTGW=2001:db8:20::1/64

If you would like to accept RAs on this interface (e.g. in order to provide IPv6 internet to the CentOS system like described in this blog post), you will have to create a file /sbin/ifup-local and make it executable:

#!/bin/sh
# /sbin/ifup-local
if [[ "$1" == "ens3" ]];
then
  sysctl net.ipv6.conf.ens3.accept_ra=2
fi

The reason for this are the RHEL network scripts (esp. /etc/sysconfig/network-scripts/ifup-ipv6:93-129). You do not have any way to set net.ipv6.conf.ens3.accept_ra=2 in order to accept RAs (IPV6_AUTOCONF=yes sets it to 1 and its not enough in this case).

Routing configuration

Tayga will need some IPv4 network for address translation. We will use 192.0.2.128/25 for this. So you will need following routing configuration on your router(s):

  • IPv4 192.0.2.128/25 via 192.0.2.2
  • IPv6 2001:db8:64:ff9b::/96 via 2001:db8:20::2/64

Enable IPv4 and IPv6 forwarding on the CentOS system:

$ cat > /etc/sysctl.d/forward.conf <<EOF
> net.ipv4.ip_forward=1
> net.ipv6.conf.all.forwarding=1
> EOF
$ sysctl --system
...

Network configuration for Tayga

Tayga needs its own tun interface, which is used as NAT64 gateway. Create following interface and routing configurations:

# /etc/sysconfig/network-scripts/ifcfg-tunnat64
TYPE=Tap  # Do not be confused about tap instead of tun here. Tun interface type gets determined by the interface name
BOOTPROTO=none
DEVICE=tunnat64
ONBOOT=yes
IPV6INIT=yes
IPV6FORWARDING=yes
# /etc/sysconfig/network-scripts/route-tunnat64
192.0.2.128/25 dev tunnat64
# /etc/sysconfig/network-scripts/route6-tunnat64
2001:db8:64:ff9b::/96 dev tunnat64

and now enable the tunnel interface:

$ ifup tunnat64

Tayga configuration

Create the tayga configuration:

# /etc/tayga/default.conf
tun-device tunnat64
ipv4-addr 192.0.2.129
prefix 2001:db8:64:ff9b::/96
dynamic-pool 192.0.2.128/25
data-dir /var/lib/tayga/default

and start tayga:

$ systemctl enable tayga@default
$ systemctl start tayga@default

See too