ungleich-staticcms/content/u/blog/how-to-route-ipv4-via-ipv6/contents.lr
2020-02-10 19:45:02 +01:00

97 lines
3 KiB
Markdown

title: How to route IPv4 via IPv6
---
pub_date: 2020-02-10
---
author: ungleich network
---
twitter_handle: ungleich
---
_hidden: no
---
_discoverable: yes
---
abstract:
Bringing IPv4 into the IPv6 world
---
body:
Imagine the following: you are running an IPv6 only network. And now
someone asks you to pass IPv4 traffic through it, without tunneling
it. Was sounds crazy at first, is actually quite feasible.
## A short routing recap
Routers have routing tables. The routing tables basically say
"if you receive a packet for this host, send it to that router".
![IP routing](/u/image/ip-routing.png)
The important thing about this process is that the information on
where to send it to, is **not in the packet**.
## How to send IPv4 packets via IPv6
Because the next hop is not written into the IPv4 packet, the router
is free to forward the packet via any method it thinks is the
best. And if that happens to be IPv6 - well, it will forward the IPv4
packet via an IPv6 neighbour.
## A practical example!
In the IPv6 only coworking network in the [Digital
Chalet](/u/projects/digital-chalet/), I can add an IPv4 default route
via the IPv6 router:
```
[root@diamond ~]# ip route add 0/0 nexthop via inet6 fe80::21b:21ff:febb:6934 dev wlp0s20f3
[root@diamond ~]# ip r
default via inet6 fe80::21b:21ff:febb:6934 dev wlp0s20f3
[root@diamond ~]#
```
Now to be able to actually transmit IPv4 packets, I do need a source
IPv4 address. In the current network I can use an address in the
unused 10.0.8.0/22 network, however I'll add it with a /32 mask to
make it clear that there is no interface local route applied:
```
[root@diamond ~]# ip addr add 10.0.8.42/32 dev wlp0s20f3
[root@diamond ~]# ip r
default via inet6 fe80::21b:21ff:febb:6934 dev wlp0s20f3
[root@diamond ~]# ip a sh dev wlp0s20f3
2: wlp0s20f3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 24:ee:9a:54:c3:bf brd ff:ff:ff:ff:ff:ff
inet 10.0.8.42/32 scope global wlp0s20f3
valid_lft forever preferred_lft forever
inet6 2a0a:e5c0:0:4:c6ea:b1a8:ec14:6f35/64 scope global dynamic mngtmpaddr noprefixroute
valid_lft 86400sec preferred_lft 14400sec
inet6 fe80::3b98:cb58:ed02:c25/64 scope link
valid_lft forever preferred_lft forever
[root@diamond ~]#
```
And I can indeed ping another IPv4 address, routed via IPv6!
```
[root@diamond ~]# ping -4 10.0.8.3
PING 10.0.8.3 (10.0.8.3) 56(84) bytes of data.
64 bytes from 10.0.8.3: icmp_seq=1 ttl=64 time=2.37 ms
^C
--- 10.0.8.3 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 2.365/2.365/2.365/0.000 ms
[root@diamond ~]#
```
## Why?
Why would anyone want to do this? It's quite easy: with this you can
route an IPv4 address to an IPv6 only host. This enables IPv6 only
resources to create and send IPv4 packets, even if they don't have
IPv4 routes.
## Do it yourself
If you don't believe us that it is possible, you can test it yourself
on IPv6 only VMs on [IPv6OnlyHosting.com](https://ipv6onlyhosting.com).