OK, I've now done a bit more, and have a working but manually setup config.
First a little more about my setup. I have a home server, hosting as a Xen guest a small Linux machine running as a dedicated router. This is running Debian Squeeze (2.6.32 kernel), PPPoE via a Netgear DM111P modem for internet connection, Shorewall (IPv4 firewall), Shorewall6 (IPv6 firewall), radvd for basic IPv6 client setup, and a tunnel via Hurricane Electric (HE) for IPv6 connectivity.
I have manual IPv6 address assignments for my key boxes (including this router), and in internal DNS server/resolver. So I can connect to "eddie.<mydomain>.co.uk and my internal clients will use IPv4 or IPv6 to access it depending on whether the client has an IPv6 address (not everything supports IPv6), and what it's config says.
Anyway, this is where I've got to :
First step was to add an additional peer to my PPP config.
In [tt]/etc/ppp/peers[/tt] - copy existing peer config (plusnet) to new file (plusnet-ipv6
To this file, add the option +ipv6
Add the login & password to[tt] /etc/ppp/chap-secrets[/tt]
If you are using the default setup done with the PPP config scripts, then you will be using a config called dsl-provider.
Bring down my IPv6 tunnel with Hurricane Electric (HE)
ifdown he-ipv6
Bring down my normal connection with plusnet
ifdown plusnet
Remove HE IPv6 addresses
ip addr del xxxx:xxxx:xxxx::1/64 dev eth0
Stop radvd so clients will stop using their existing IPv6 addresses.
/etc/init.d/radvd stop
And clear my current Shorewall6 config to avoid it getting in the way and confusing things
shorewall6 clear
Add an entry to /etc/network/interfaces
iface plusnet-ipv6 inet ppp
provider plusnet-ipv6
I've deliberately not made this "auto" as I don't want it bringing up except while I'm doing things manually.
We should now be able to bring up the new connection
ifup plusnet-ipv6
and check that it's come up
ifconfig ppp0
ppp0 Link encap:Point-to-Point Protocol
inet addr:81.174.xxx.xxx P-t-P:195.166.128.123 Mask:255.255.255.255
inet6 addr: fe80::xxxx:xxxx:xxxx:xxxx/10 Scope:Link
UP POINTOPOINT RUNNING NOARP MULTICAST MTU:1492 Metric:1
RX packets:770 errors:0 dropped:0 overruns:0 frame:0
TX packets:919 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:3
RX bytes:315291 (307.9 KiB) TX bytes:153018 (149.4 KiB)
The key things at this point are :
1) Our IPv4 has come up correctly and all our IPv4 stuff should be working
2) We have a LInk Local IPv6 address (fe80::…)
We now need a DHCP client for IPv6. I'm used to using the ISC DHCP software, but this does not work on PPP interfaces without patching. So I installed Wide DHCP :
apt-get install wide-dhcpv6-client
Configure the DHCP client to ask for information (see
http://community.plus.net/forum/index.php/topic,96320.msg812216.html#msg812216)interface ppp0 {
# request domain-name-servers;
# request domain-name;
send rapid-commit;
send ia-pd 0;
script "/etc/wide-dhcpv6/dhcp6c-script";
};
id-assoc pd {
prefix-interface lo {
sla-id 3; #interface id to form prefix
sla-len 8; #shift from allocated prefix, ie /56 + 8 = /64, so netmask for lo0 is /64
};
};And test it :
dhcp6c -d -D -f ppp0
You should now get pages and pages of debug information !
Fairly quickly I realise that this DHCP client config does things I don't want. It changes my resolve.conf which is bad since I run my own internal resolver, and it adds an address to lo which I don't need.
So, I changed the config to just :
interface ppp0 {
send rapid-commit;
send ia-pd 0;
};
id-assoc pd {
};All this does is get a PD from Plusnet's server which seems to keep things happy at their end.
Questions :Is DHCP actually required ?
Does it (for example) setup IPv6 routing or whatever via the active link ?
Would things work reliably if I just used DCHP to find out my PD initially, and then hard-coded everything after that ?I then manually configured interface IPs and routes :
ip addr add 2a02:16c8:xxx:xxx0::1/64 dev eth0
ip addr add 2a02:16c8:xxx:xxx7::1/64 dev ppp0
ip -6 route add default via fe80::90:1a00:3a3:8d54 dev ppp0
And tested :
ping6 ipv6.google.com
traceroute6 ipv6.google.com
The address fe80::90:1a00:3a3:8d54 here is the LL address of the remote end as reported by wide-dhcpd.
A quick test reveals that the default route can be simplified :
ip -6 route add default dev ppp0
This means that I don't need to parse the DCHP client output to get the remote endpoint address in order to add a default route.
Starting radvd allows my other machines to self-assign addresses :
/etc/init.d/radvd start
[tt]/etc/radvd.conf[/tt] :
interface eth0 {
AdvSendAdvert on;
MinRtrAdvInterval 3;
MaxRtrAdvInterval 10;
AdvDefaultPreference low;
AdvHomeAgentFlag off;
prefix ::/64
{
AdvOnLink on;
AdvAutonomous on;
AdvRouterAddr off;
};
};By using "prefix ::/64", radvd just gets the public IPs from those already present on the interface. Useful during testing as it is one less config file to alter when switching Pv6 upstream connections.
Now, to sort out a firewall. I'm already running Shorewall6 with my HE connection, so a couple of changes needed :
In /etc/shorewall6/interfaces, remove he-ipv6 and add ppp0 to the net zone
In /etc/shorewall6/rules, add rules to allow in and outbound DHCP6 traffic :
ACCEPT $FW net udp 547
ACCEPT net $FW udp 546
So that's got me a working IPv6 connection using my Linux router.
Now, what to do to automate it all ?As I see it, I need the following :
When ppp0 comes up - start a dhcp6 client on it.
When the dhcp client gets a PD, configure my interfaces, add a default route, restart Shorewall6 and radvd.
If ppp0 goes down, then drop my public addresses and default route, restart Shorewall6 and radvd.
I don't see easy ways to automate much of that at all. Any suggestions welcome !
