IPv6 router guide

From Gentoo Wiki
Jump to: navigation, search

Other languages:
This article has been flagged as dirty as of February 3, 2016 for not conforming to the wiki guidelines. It is now grouped in the list of articles that need formatting improvements.


Basic kernel configuration

Any of the 2.6 kernel trees available in Gentoo will easily support IPv6 connections. The new USAGI IPv6 stack is integrated to the kernel since Linux 2.6.0.

root #emerge --ask sys-kernel/gentoo-sources

Now we are ready to enter the kernel source directory and begin our actual kernel configuration.

root #cd /usr/src/linux
root #make menuconfig
This assumes the symlink /usr/src/linux points to the sources that will be used as part of this guide.
KERNEL 'make menuconfig' options
Networking support --->
  Networking options --->
    <*> The IPv6 protocol --->
## (The IPv6 options beneath this one can be useful for many other applications,
## but should not be needed for a basic setup)
## (This option is only required when using ptrtd for 6to4 conversion)
Device Drivers --->
  Network device support --->
    <*> Universal TUN/TAP device driver support

Testing IPv6 support

After enabling the recommended options, recompile the kernel and reboot into the new IPv6-enabled kernel.

If iproute2 is not yet installed, it is seriously recommended to do so. iproute2 is a network configuration suite that contains ip, the famous replacement for ifconfig, route, iptunnel and others...

root #emerge --ask sys-apps/iproute2
Use of ifconfig can cause serious headaches if there are multiple tunnel devices. The tunnels need to be removed in backorder, which means that the latest created must be removed first.

If IPv6 is working, the loopback device should show an IPv6 address:

root #ip -6 addr show lo
1: lo: <LOOPBACK,UP> mtu 16436
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever
    ## (The above lines show things are working)

Before going any further, make sure to add ipv6 to the list of USE variables in make.conf, so that future emerges of packages will include IPv6 support.

Tunnel configuration

Basic configuration

Most ISPs still do not offer any native IPv6 connections. To get around this limitation, there are several "tunnel brokers" around the globe that offer free IPv6 tunnels. This will allow to tunnel all the IPv6 connections through an IPv4 connection.

Broker Location
Hurricane Electric North America, Europe, Asia
Freenet6 US
SixXS Europe (starting from april 2016 SixXS is no more accepting signups, nor tunnel or subnet requests)
Singnet Singapore
Aarnet Australia/South Pacific

Below are two examples for setting up a tunnel with two popular North American tunnels, Hurricane Electric (applies for non-heartbeat tunnels from sixxs.net as well) and Freenet6.

Hurricane Electric

Hurricane Electric (HE for short) offers free IPv6 tunnels and allocates a /64 block of addresses for each customer. It also allows configuration of reverse DNS. Getting a tunnel from HE is as easy as going to https://www.tunnelbroker.net/ and filling out a one page form.

Registration includes listing information like home address and phone number.

After a tunnel is approved and a /64 block is allocated, start to configure the system. HE provides sample configurations based on ifconfig and the iproute utilities. The following two examples assume that the following configuration is used:

Local IPv4 Address (eth0)
HE IPv4 Address
Local IPv6 tunnel Address 2001:470:1F00:FFFF::2
Remote IPv6 tunnel Address 2001:470:1F00:FFFF::1
IPv6 Block 2001:470:1F00:296::/64

Using the sys-apps/iproute2 package and the ip command, do the following.

Create a tunnel between the local (eth0) IPv4 and HE's remote IPv4 address:

root #ip tunnel add he6 mode sit remote local ttl 64 dev eth0

Extract the tunneling overhead from the MTU:

root #ip link set he6 mtu 1280

Bring the tunnel up:

root #ip link set he6 up

Assign the IPv6 address to it:

root #ip addr add 2001:470:1F00:FFFF::2 dev he6

Route all global unicast IPv6 addresses through our 'he6' tunnel device:

root #ip route add 2000::/3 dev he6

The following example shows how to establish this at boot time:

CODE netifrc example
iptunnel_he6="mode sit remote local ttl 64 dev eth0"
routes_he6="default via 2001:470:1F00:FFFF::1 dev he6"

To make this device start on boot:

root #cd /etc/init.d
root #ln -s net.lo net.he6
root #rc-update add net.he6 default
If there is no default policy of ACCEPT for the IPv4 iptables then add:
root #iptables -A INPUT -i eth0 -p ipv6 -j ACCEPT

When tunneling IPv6 over IPv4, the packets will first come through the IPv4 chain before being passed to the IPv6 chain.


Freenet6 is another free tunnel broker. Optional registration only requires a username and a valid email address. They have chosen to turn the tunnel management into a client/server setup and have created the gogoCLIENT client. The client is available in Portage. To install it do:

root #emerge --ask net-misc/gogoc

Now when choosing to connect with authentication, configure gogoCLIENT by editing /etc/gogoc/gogoc.conf. Only the userid and passwd fields need to be changed to match those assigned from Freenet6 and change the gateway server. Below is a sample config file.

CODE gogoc.conf example

Testing the connection

Now that the tunnel is configured, test the connection. The easiest way to do this is to use the ping6 utility and try to ping an IPv6 host.

root #emerge --ask iputils
user $ping6 www.kame.net
PING www.kame.net(orange.kame.net) 56 data bytes
64 bytes from orange.kame.net: icmp_seq=1 ttl=52 time=290 ms
64 bytes from orange.kame.net: icmp_seq=2 ttl=52 time=277 ms
64 bytes from orange.kame.net: icmp_seq=3 ttl=52 time=280 ms
64 bytes from orange.kame.net: icmp_seq=4 ttl=52 time=279 ms
64 bytes from orange.kame.net: icmp_seq=5 ttl=52 time=277 ms
--- www.kame.net ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4038ms
rtt min/avg/max/mdev = 277.040/281.041/290.046/4.699 ms

Further work is currently in progress to add better IPv6 support to the network init scripts. To know the status of this and/or help out, please email latexer@gentoo.org.

IPv6 support in applications

Re-emerging packages

Unless USE="ipv6" was already set in /etc/portage/make.conf previously, it is probably necessary to re-emerge a bunch of packages to compile in IPv6 support for them. To get a list of all the installed packages which are affected by USE flag changes, use Portage's --newuse (-N) option:

root #emerge -uDNav @world

When many USE flags have been changed, the list could be quite long. It's suggested to keep the system up-to-date, so it won't hurt if all affected packages are rebuilt.

Some packages (erroneously) detect IPv6 support automatically and hence have no ipv6 USE flag. Thus not all packages, which should support IPv6, will support it if they have not been compiled with an IPv6 enabled kernel.

IPv6 specific packages

There are a few packages which specifically deal with IPv6 items. Most of these are located in /usr/portage/net-misc.

Package Description
net-misc/ipv6calc Converts an IPv6 address to a compressed format
net-misc/netcat6 netcat version that supports IPv6 and IPv4
dev-perl/Socket6 IPv6 related part of the C socket.h defines and structure manipulators

DNS setup

IPv6 and DNS

Just as DNS for IPv4 uses A records, DNS for IPv6 uses AAAA records. (This is because IPv4 is an address space of 2^32 while IPv6 is an address space of 2^128). For reverse DNS, the INT standard is deprecated but still widely supported. ARPA is the latest standard. Support for the ARPA format will be described here.

BIND configuration

Recent versions of BIND include excellent IPv6 support. This section will assume at least minimal knowledge about the configuration and use of BIND. We will assume that bind is not running in a chroot. If this assumption is wrong, simply append the chroot prefix to most of the paths in the following section.

First add entries for both forward and reverse DNS zone files in /etc/bind/named.conf.

FILE /etc/bind/named.confnamed.conf entries
## (We allow bind to listen to IPv6 addresses.
## Using 'any' is the only way to do it prior to bind-9.3)
options {
    listen-on-v6 { any; }
## (This will provide the forward DNS for the domain 'ipv6-rules.com':)
zone "ipv6-rules.com" IN {
    type master;
    file "pri/ipv6-rules.com";
## (This format for reverse DNS is "bitwise." It's done by taking the IPv6 prefix,
## reversing the order of the numbers and putting a period between each number)
zone "" {
        type master;
        file "pri/rev-ipv6-rules.com.arpa";

Now we must create those zone files and add entries for all of our hosts:

FILE /etc/bind/pri/ipv6-rules.com
$TTL    2h
@       IN      SOA     ipv6-rules.com. webmaster.ipv6-rules.com.  (
                                2003052501 ; Serial
                                28800      ; Refresh
                                14400      ; Retry
                                3600000    ; Expire
                                86400 )    ; Minimum
                NS      ns1.ipv6-rules.com
IN      AAAA    2001:470:1f00:296::1 ; address for ipv6-rules.com
host1   IN      AAAA    2001:470:1f00:296::2 ; address for host1.ipv6-rules.com
host2   IN      AAAA    2001:470:1f00:296::3:3 ; address for host2.ipv6-rules.com
FILE /etc/bind/pri/ipv6-rules.com.arpa
$TTL 3d ; Default TTL (bind 8 needs this, bind 9 ignores it)
@       IN SOA ipv6-rules.com. webmaster.ipv6-rules.com. (
                        2003052501      ; Serial number (YYYYMMdd)
                        24h             ; Refresh time
                        30m             ; Retry time
                        2d              ; Expire time
                        3d )            ; Default TTL
        IN      NS     ns1.ipv6-rules.com.
; IPv6 PTR entries
$ORIGIN IN      PTR     ipv6-rules.com. IN      PTR     host1.ipv6-rules.com. IN      PTR     host2.ipv6-rules.com.

DJBDNS configuration

There are currently some third-party patches to DJBDNS available at http://www.fefe.de/dns/ that allow it to do IPv6 name serving. DJBDNS can be installed with these patches by emerging it with ipv6 in the USE variable.

Not all record types are support yet with these patches. In particular, NS and MX records are not supported.
root #emerge --ask djbdns

After djbdns is installed, it can be setup by running tinydns-setup and answering a few questions about which IP addresses to bind to, where to install tinydns, etc.

root #tinydns-setup

Assuming we've installed tinydns into /var/tinydns, we can now edit /var/tinydns/root/data. This file will contain all the data needed to get tinydns handling DNS for the IPv6 delegation.

CODE sample data file
## (*.ipv6-rules.com is authoritatively handled by
## (Authoritative reverse DNS for 2001:470:1f00:296::/64)
## (Specify the IPs for host1 and host2)
## (Point www to host1)

Lines prefixed with a 6 will have both an AAAA and a PTR record created. Those prefixed with a 3 will only have an AAAA record created. Besides manually editing the data file, it is possible to use the scripts add-host6 and add-alias6 to add new entries. After changes are made to the data file, simply run make from /var/tinydns/root. This will create /var/tinydns/root/data.cfb, which tinydns will use as its source of information for DNS requests.

IPv6 router

Configure routing

Further configuration is required if we want to use our system as a router for other clients wishing to connect to the outside world with IPv6. We need to enable forwarding of IPv6 packets. We can do this in one of two ways.

  • Either we set the value 1 in the forwarding pseudo-file:
root #echo 1 > /proc/sys/net/ipv6/conf/all/forwarding
  • Or we use the sysctl command:
root #sysctl -w net.ipv6.conf.all.forwarding=1
The radvd init script explained in the next chapter enables (and disables) forwarding, making the next step unnecessary.

To enable forwarding at boot, edit /etc/sysctl.conf and add the following line.

CODE sysctl.conf addition
## (When using radvd, this setting is not needed)

Traffic should now be forwarded from this box through the tunnel we've established with our broker.

To assign IPv6 addresses to clients, the IPv6 specification allows for both stateless and stateful IP assignment. Stateless assignment uses a process called Router Advertisement and allows clients to obtain an IP and a default route by simply bringing an interface up. It is called "stateless" because there is no record of IPs assigned and the host they are assigned to. Stateful assignment is handled by DHCPv6. It is "stateful" because the server keeps a state of the clients who've requested IPs and received them.

Stateless configuration

Stateless configuration is easily accomplished using the Router Advertisement Daemon, or radvd:

root #emerge --ask radvd

After having emerged radvd, we need to create /etc/radvd/radvd.conf that contains information about what IP block to assign IPs from. Here is a sample radvd.conf file using the prefix we've been assigned from our tunnel broker.

CODE Sample radvd.conf
interface eth1
        ## (Send advertisement messages to other hosts)
        AdvSendAdvert on;
        ## (Fragmentation is bad(tm))
        AdvLinkMTU 1280;
        MaxRtrAdvInterval 300;
        ## (IPv6 subnet prefix we've been assigned by our PoP)
        prefix 2001:470:1F00:296::/64
                AdvOnLink on;
                AdvAutonomous on;
Make sure the interface on the first line is correct so that router advertisements are broadcasted to the intranet and not to the ISP!

Further information is available in man radvd.conf. We can now start radvd and set it to start at boot.

root #/etc/init.d/radvd start
root #rc-update add radvd default

Stateful configuration

To have a stateful configuration, install and configure net-misc/dibbler.

root #emerge --ask dibbler

Now we must configure the dibbler client by editing /etc/dibbler/client.conf.

CODE Sample dibbler client configuration
iface ppp0 {
	rapid-commit yes
	option dns-server

We can now start the dibbler client, and configure it to start at boot.

root #/etc/init.d/dibbler-client start
root #rc-update add dibbler-client default

IPv6 clients

Using radvd

Clients behind this router should now be able to connect to the rest of the net via IPv6. If using radvd, configuring hosts should be as easy as bringing the interface up. (This is probably already done by the net.ethX init scripts).

root #ip link set eth0 up
root #ip addr show eth0
1: eth0: <BROADCAST,MULTICAST,UP> mtu 1400 qdisc pfifo_fast qlen 1000
    link/ether 00:01:03:2f:27:89 brd ff:ff:ff:ff:ff:ff
    inet6 2001:470:1f00:296:209:6bff:fe06:b7b4/128 scope global
       valid_lft forever preferred_lft forever
    inet6 fe80::209:6bff:fe06:b7b4/64 scope link
       valid_lft forever preferred_lft forever
    inet6 ff02::1/128 scope global
       valid_lft forever preferred_lft forever

Should this not work ensure that the IPv6 firewall is allowing ICMPv6 packets through:

root #ip6tables -A INPUT -p icmpv6 -j ACCEPT

External resources

There are many excellent resources online pertaining to IPv6.

On IRC, try the #ipv6 channel on Freenode. Connect to the Freenode servers using an IPv6 enabled client by connecting to irc.ipv6.freenode.net.

This page is based on a document formerly found on our main website gentoo.org.
The following people contributed to the original document: Peter Johanson, Jorge Paulo, Camille Huot, Pasi Valminen, , Markos Chandras (Hwoarang)
They are listed here because wiki history does not allow for any external attribution. If you edit the wiki article, please do not add yourself here; your contributions are recorded on each article's associated history page.