Matt-J.co.uk : Ramblings

Hello again my minions!

Randomly started an interesting project last night, while thinking about wanting access to my NAS and network via wifi (but secure) and also better, more reliable remote access to my pc’s even when they are behind a nat or two.

My Current Layout Involves a Custom WRT54G Router performing NAT onto the 200.x subnet, and then connected to my multi-homed linux server/router that routes and firewall’s between the 200.x subnet for internet, my personal 1.x subnet, and the rest of my houses 2.x subnet (martin’s NAS, Sams server etc) (My server also serves apache, samba etc etc but that dosnt matter here)

I decided the way to go was OpenVPN, as I am very comfortable with its configuration, and prefer a fully routed vpn over port by port client less ssl (web based) access solutions.

I Set up a PKI based openvpn system. allowing a client to logon with the correctly signed certificate and private key pair, instead of a username/password combination.

The wifi point is built into my nat router, and so by default, wireless clients get assigned an IP in the ‘internet only’ 200.x range with access to the web via NAT but on the wrong side of the linux firewall to gain access to the network. (It’s still WPA2 -PSK Protected, but I guess thats just my paranoia :P)

After setting up my laptop with a client configuration which tries to establish a tunnel to 192.168.200.2 (linux server from 200.x subnet.. only openvpn allowed through firewall) first, and an internet route able IP if that should be unreachable (I’ll explain that in a mo)

I also configured openvpn to push route information for my subnets to clients at logon, minimizing configuration.

I started the openvpn server and connected my client. Everything worked first time. I am still using 192.168.200.1 as my default gateway, (as i dont want internet traffic having to route in and out of my net again) however trying to connect to my webserver or file shares on any of my pc’s works perfectly, and speedy enough over 54Mbps WIFI.

With all this working, My plan was then to purchase another fixed IP address on my dedicated server (this webserver) and use IPTables to route all traffic to that destination IP to my Linux Firewall/Router down a Static p2pVPN tunnel between the webserver and my linux server.

Then I had a better idea, as forwarding everything (especially UDP) in IPtables was a horrible way to do things as i’d most probably have to use NAT, which seemed to lack elegance somewhat.

I created the p2p Openvpn tunnel as planned, but instead of adding the new internet ip address to my webservers network card (ip addr add blah.blah.blah.blah dev eth0) and then IPtables-ing the traffic down the tunnel. I instead added the ip address to the p2p VPN tunnels tun interface on my (local.. in my room) side, then added a static route to my webserver informing it that the new internet IP address i had purhased was reachable at 10.0.0.2 (the linux server side of the p2p vpn tunnel) and finally enabled IP forwarding on my webserver.

So Far so good! I had an actual internet IP available to me in my room now, behind university’s NAT, properly routed back to the web! I opened a UMTS connection to the internet through my phone + bluetooth + laptop, and tried pinging my new IP, I did not get replies, But loading wireshark on my linux server did show the ping requests were getting to me down the tunnel!

However (And I should have seen this coming.. but I was tired as it was now about 5am) my server replied.. to an internet  IP… and so of course… it routed via the default route.. which was my local NATed internet connection, so no wonder the replies did not get through. I still had a bit to go before getting a full TCP session to occur!

So now I needed a way to change the default route depending on where the traffic was going to / coming from.

A quick refresher on ‘ip rule’ toolset within iproute2 software and I was able to create a separate routing table called table_tunnel. In it I placed the single route ‘ip route add default via 10.0.0.1 tun1′. I then Created a rule that told the system that any IP Data with a source address of  my new internet IP. Should traverse the ‘table_tunnel’ routing table to find out where to go, and so consequently, all data coming in to my new internet IP, also leaves via the tunnel, allowing proper TCP communication, while allowing my normal internet browsing to continue working normally via the standard ‘Main’ and ‘Local’ Routing tables.

Success! I finally got ping replies from my dial-up (UMTS) laptop!

I then just had to script all the changes to make them survive reboots. Daemonise the openvpn processes, and tighten up the firewall on the incoming p2p vpn tunnel to only allow the openvpn UDP port (as thats what i will be doing all my remote access over, so why leave anything else open, when openvpn will give me access into my subnet anyway.)

And so now, When I start openvpn on my laptop. If im not connected to my WIFI (Ie, somewhere around the world on the internet), it cannot connect to 192.168.200.2, and so tries the internet IP.. which ends up at the same openvpn server on my local network, but via a real internet IP and a bit of clever routing!

Downside: Ive been up all night
Upside: I can access all my data wherever I am… and even if i move house in the future, that internet IP will still allow me to access my systems, as theres nothing dependant on the local home internet IP/details ;)

Also… should I want to host other things from home to the public (quickly set up apache/ftp/voice server.. Its as easy as changing one IPTables rule!)

I Have intertubes in my room!!!!

//Matt