I’m trying to set up a new router, and abandon all of the outdated practices I’ve used before. Part of that (in fact, the main part) is the idea of using a dual stack network.
From my ISP, I have one IPv4 address (78.130.165.174, received via ethernet cable and DHCPv4) and just one router, and per the instructions in the wiki, I have created a tunnel at HE. The IPv4 part works fine, but the IPv6 one doesn’t. Possibly because on 5.19, the instructions don’t follow as nicely as described there. Or maybe there’s something else I don’t understand yet.
I’m trying to do everything with DHCP servers and/or IPv6 stateless configuration (anything, as long as end user devices can be configured automatically) via a wireless interface. I have a Windows 7 machine for testing client connectivity. The way I have it now, I’ve been able to give it an IPv6 with stateless configuration, and I can ping the HE gateway, but I can’t ping my own router, nor do I ever see any gateway and DNS assigned. Needless to say I can’t even ping any site, as DNS resolution doesn’t even occur.
Here’s an export on my current configuration (actual IPs and everything):
If you look at the tunnel broker config page you will see your routed prefix which is different than your client and server prefix (off by one). You need to add two IP addresses, one externally that is your client ipv6 address, and an address in your routed prefix to an internal interface. Your default route(s) then need to point to your server address as part of the tunnel.
I hadn’t noticed the difference in the routed addresses. Thanks.
But despite the adjustments, there’s still no IPv6 internet, and no gateway and DNS on the Win7 machine. The only noticeable difference is that the Win7 machine now receives an address in the routed subnet, as it should.
I should note that the key difference from the wiki tutorial is that the gateway in “/ipv6 route”, can’t be specified as an IP. It can only be specified as an interface, or a link-local address, with interface attached to it (which, let’s face it, isn’t any more useful than just the interface name).
You are getting the address via RA and I don’t think that the windows ipv6 stack supports dns configuration except via static or DHCP, which is not available in the ROS implementation. (I use bind). You don’t have to have ipv6 addressable dns servers, but it does help if you want full ipv6 support. Just add them statically if you can’t run a copy of bind locally to hand out ipv6 dns server options.
You will need to set those two routes for the tunnel to work (one of the two might work, but I use both).
They are both basically default gateway routes with the 2000::/3 is the old legacy reference and ::/0 being the current reference. You just need to point them to the tunnel server address.
Setting the DNS server statically doesn’t seem to work. In fact, setting the IP and gateway statically sort of broke the whole thing (no pings to anything).
I’ll retry later with a different, Windows 8 machine. I remember there were some changes in it with regards to IPv6. Hopefully, it includes a proper DHCP client and/or stateless DNS/gateway configuration. If even that doesn’t work, I guess I’ll hold out IPv6, since MikroTik woudn’t be ready yet. Damn…