I've found that the default route might dissapear when the parent interface goes down. I have not debugged that.

Yeah I think that is a 2.0 thing. I'm not exactly sure what the criteria are for showing up there, but it should be gateways listed there, I thought. Looking a bit deeper we already have a ticket open for that: http://redmine.pfsense.org/issues/785 But it's set to 'future' since it's more of a convenience and could be done via static routes as well.

The remco bressers post is indeed hackery. And as you also mention this is very different from the current IPv6 code in my branch which is far more turn key. I just setup another v6 box at work, and it's becoming easier, just plug in some addresses and it works. That was the whole intention of pfSense. Your advice on configuring proper IPv6 over fiddling around is a good suggestion. It's far more valuable and less likely to break your internet in interesting ways. :-)

This sometimes happens, it's reproduceable, I have not investigated.

Have not done this since moved tunnel endpoint to my pfsense box.. Much better way to do it!!  Now you have firewall control at pfsense, etc. But to allow the tunnel endpoint to be box on lan side of your pfsense box, allow ipv6 traffic and put in the lan side IP address you want to be the endpoint, example 192.168.1.100 or something, whatever the IP of your box is that is going to be the endpoint of your ipv6 tunnel. [image: ipv6behindpfsense.jpg] [image: ipv6behindpfsense.jpg_thumb]

I have not attempted a gateway pool yet. And that doesn't work with a gateway pool anyhow, any attempt to send out traffic from a HE address through the Sixxs router will fail. The only way to make that work is by performing NAT and there is no provision for that. You can not add multiple v6 subnets on the same interface either. The only trick you could theoretically use here is performing network prefix translation. So if you have 2 /64 subnets, one from each ISP you can then translate this network when traffic leaves the network. I do this by assigning a ULA range on the LAN, then create 2 mapping, 1 for each WAN with each network prefix. This so that traffic leaving either interface is using the netblock from the correct ISP. Not much different from ipv4 nat, but subtly different.

Oh yes - I forgot that part :-) One of the reasons why some of the big ones are dragging their feet is also that they earn A LOT of money from charging extra for official IP addresses. How many addresses should a non commercial have? Give each one a /64 subnet - don't try to create artificial limits. It is a new world - deal with it. If someone want to set up a server at home - let them do it. I think ISP's should realise that what they provied is a connection. What this connection is used for - reading the newspaper, watching TV, listening to radio, online purchases, research, running your own server for your blog - that should all be up to the person paying for the line. You can differenciate on the guaranteed quality you are offering - commercial customers could get two lines in through different connection points so it is less likely they will loose connection - home users will obviously only have one line and multiple single points of failure. Also, a home user line is always oversold - I don't know any ISP that actually has enough bandwith to give full speed to all customers at the same time. A commercial customer could receive minimum guarantees of bandwith/speed. There are ways to differentiate commercial and home users. But it should never be on what they can do with their line. That would be like Ford denying you to carry tools that you use at work in your car because it was not bought as a commercial vehicle.

I've decided on the options disabled, managed, unmanaged, combined. Fix forthcoming later.

@iFloris: There is a script on the he.net website to change your ip. By installing cron and calling the script every 24 (or so) hours, you can sort of automatically update your endpoint. That sounds like something that would make a good candidate for a dyndns update type.

@databeestje: Outside of the snapshots I don't have these binaries Download a snapshot, extract the racoon, setkey and racoonctl binaries from /usr/local/sbin and upload those to your install thanks for the quick reply… As I was driving into work after posting this, I thought of trying that in-case there is not binaries to download.

2.0-RC1-IPv6 (i386) built on Tue Mar 15 13:14:13 EDT 2011 Not entirely sure if this is an IPv6 thing or just an RC2 thing… searched a bit but didnt find anything on it over in that forum. Since last Thursday, i'm not sowing any wan-out data (IPv6 or 4) in my rrd traffic or packets graphs. Nothing sticking out at me in the logs either. Anyone else seeing this?

Create a 1.2.3 vm, create 2 routes to a lan IP. Upgrade that and it should trigger it

ofcourse linksys and dd-wrt are not really comparable. I've just setup a other customer from the bounty forum with a working IPv6 setup. With a gif tunnel or static addressing on the wan it does work with v6. With recent fixes I've made it should be safe to run either v4 or v6 on the same interface. I fixed the issue that made it drop it's v4 default route. If you do not mind giving me remote access to that install to develop working ipv6 support then I'd love to get that working. It is probably the most significant building block for IPv6 deployment on cable internet networks. And there are a lot of those. If you have questions, feel free to climb into the email to seth.mos@dds.nl and we'll see what we can arrange. I've just made some fixes to filter.inc for dhcpd, the webui listen port on the v6 address and a DNS rebinding attack fix. I've also added v6 network support for the ipsec phase2 page. Maybe I can fix the phase1 too.

Well, it's not uncommon for rules to reload if it has something like hostnames and other factors. I think what's biting here is that firewall rules for proper multi-wan support are missing. When that falls through the reply-to is skipped and the default route is relied on. I'll have a quick look if I can make the filter process intelligent enough to find out the proper v6 wan gateway.

It appears there is no support in DHCP v6 for static IP address assignments based on "raw" MAC address but assignments can be based on various forms of UIDs, some of which can be derived from a MAC address.

I installed net/dhcp6 now and playing with it. IPv6 is new for me.

Same as before, don't toggle port forward. Add a firewall rule on the IPv6 WAN interface that allows traffic from Any to IPv6 address of the client behind it with the port you want to give them access to.

No interest at this point, directing resources elsewhere.

You can, I just use the last octet, which is already good for 65k hosts or so. I like the addresses simplere over autoconfig