IPv6 large packets failing

  • Hello everyone!

    I just registered and got a /64 from Hurricane Electric's Tunnelbroker and I followed these instructions to get it set up on pfSense 2.0 Beta 5: http://iserv.nl/files/pfsense/ipv6/. After getting everything working, the first site I tested was ipv6.google.com; it worked great. Excitedly I went over to test-ipv6.com expecting to see a ton of check marks, at least that was the plan. After letting test-ipv6 do its thing, I got a red X next to IPv6 "large packets." At this point, I noticed that every other ipv6 site I tried (other than google) was not working. I can ping6 sites like he.net, sixxs.net, ect. although as soon as I try to browse to them (in Chrome, Firefox, and/or IE), it times out because these so-called "large packets" are getting dropped somewhere.

    I did some research on large packets and 9 times out of 10 the "easy" fix is to set the MTU on the tunnel interface (gif0) to 1280. I tried this, although it hasn't helped.

    Can anyone shed some light on this problem? I would really like to get IPv6 completely working on my network so I can start messing around with it.

    Thanks in advance,

  • If the mtu of your WAN interface is smaller then 1500 bytes, which is common on pppoe and pppoa then there will be mtu issues.

    What you can do is set the mtu of your client computers to 1472 bytes and then it should just work.

    At my work we have 2 fiber connections that terminate by pppoe, i then need to make a v6 tunnel to he.net which is eating another few more bytes. This means that for everything to work reliably I need to set the mtu on everything below 1472. That's not helping much.

    They've developed path mtu discovery, and that doesn't work for me atleast. Or not on the Amsterdam Tunnel broker.

  • Thank you very much, Databeestje; that did the trick. I have a Verizon FiOS connection that terminates by PPPoE with an MTU less than 1500, so that was definitely my problem. I am now happily exploring and learning about the IPv6 realm.

    Thanks again,