PPPoE & IPv6

hci · Mon Jan 10, 2011 5:45 pm

Is anyone doing this? Is there a HOW-TO or anything on WIKI? Anyone share a sample config?

I do not see where the IP 'pools' supports IPv6? I imagine you assign every user a single IPv6 IP and also route them an entire subnet of IPv6?

Mon Jan 10, 2011 5:53 pm

Address pools does not support IPv6, yet.

hci · Mon Jan 10, 2011 6:00 pm

Can you assign a PPPoE user both an IPv4 and an IPv6 IP at same time?

Mon Jan 10, 2011 6:08 pm

Yes there are two separate parameters for ppp secrets

remote-address - this is for ipv4 address
remote-ipv6-prefix - this one is for ipv6 prefix

prcek · Mon Jan 10, 2011 8:32 pm

Yes there are two separate parameters for ppp secrets

remote-address - this is for ipv4 address
remote-ipv6-prefix - this one is for ipv6 prefix

For radius scenario

remote-address = framed-ip-address
remote-ipv6-prefix = framed-ipv6-prefix

rigt? So in my case, framed-ipv6-prefix reply attribute is ignored by NAS. PPPoE session set fe80::x/64 route entry only. On both sides (NAS and client). I didnt have time to check what is going wrong, but if you have any ideas please let me know, thanks.

prcek

hci · Thu Jan 13, 2011 2:20 am

>>remote-address - this is for ipv4 address

这路线和分配PPPoE客户机的IPv4 IP.

>>remote-ipv6-prefix - this one is for ipv6 prefix

So this routes the IPv6 subnet and assigns the subnet to the PPPoE client? So does it also tell the PPPoE client what its subnet is?

hci · Thu Jan 13, 2011 2:50 am

Interesting Doc I found:

http://tools.ietf.org/html/draft-huang- ... options-00

omidkosari · Mon Jan 17, 2011 6:05 pm

I also interested in RADIUS scenario .

shade · Wed Jan 19, 2011 11:57 am

I made it working with FreeRADIUS.

It is possible to give both IPv4 and IPv6 address in one session.

TABLE radcheck id username attribute op value 1 alex Password == test TABLE radusergroup username groupname priority alex ppp 1 TABLE radgroupreply (common attributes which we give every ppp-user) id groupname attribute op value 1 ppp Framed-Protocol = PPP 2 ppp Service-Type = Framed-User TABLE radreply (user-specific attributes, IPv4 and IPv6 addresses) id username attribute op value 6 alex Framed-IPv6-Prefix = 2001:470:9de7:11::/64 3 alex Framed-IP-Address = 10.9.1.11 4 alex Framed-IP-Netmask = 255.255.255.255

But I cannot find a billing with IPv6 accounting support. Anyone can advice something?

prcek · Sun Jan 23, 2011 9:46 pm

Shade, could you post your /ipv6 address print from pppoe client please? In my case, it shows:

0 DL fe80::20c:42ff:fe0d:e5b/64 bridge1_local no 1 DL fe80::20c:42ff:fe0d:e5f/64 bridge2_wan no 2 DL fe80::a/64 pppoe-out1 no

and there is no sight about framed-ipv6-prefix (xx:xx:1:2710::/64). My route list on pppoe client shows:

0 ADS ::/0 pppoe-out1 1 1 ADS fe80::50/128 pppoe-out1 1

and again, nothing about framed-ipv6-prefix.

The NAS side shows framed-ipv6-prefix after successful pppoe login:

0 A S ::/0 xx:xx:1:3e8::1 1 1 ADS xx:xx:1:2710::/64  1 2 ADS fe80::a/128  1

Is this all correct? I dont think so.

Prcek

shade · Tue Jan 25, 2011 9:11 pm

hm.. client side seems doen't work

I tested it with Windows 7 as client and RB750 (5.0rc5) as server and it is works.

I tried RB750(5.0rc8) as client and RB750 (5.0rc5) as server and it seems doesn't work. I don't see neither IPv6 address no IPv6 route, neither on client's side no on server's side.

shade · Tue Jan 25, 2011 9:30 pm

sorry, on server side I see route

[admin@MikroTik] > /ipv6 route print Flags: X - disabled, A - active, D - dynamic, C - connect, S - static, r - rip, o - ospf, b - bgp, U - unreachable # DST-ADDRESS GATEWAY DISTANCE 0 A S ::/0 2001:470:1f0a:1b4e::1 1 1 ADC 2001:470:1f0a:1b4e::/64 he-ipv6 0 2 ADC 2001:470:1f0b:1b4e::/64 ether1-gateway 0 3 A S 2001:470:9de7:1::/64 2001:470:1f0b:1b4e::10 1 4 ADS 2001:470:9de7:11::/64  1 5 ADS fe80::b/128  1 [admin@MikroTik] > ping 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 HOST SIZE TTL TIME STATUS 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 64 0ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 64 0ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 64 0ms echo reply sent=3 received=3 packet-loss=0% min-rtt=0ms avg-rtt=0ms max-rtt=0ms [admin@MikroTik] >

but on client side only fe80:* and it is

[admin@MikroTik] > /ipv6 route print Flags: X - disabled, A - active, D - dynamic, C - connect, S - static, r - rip, o - ospf, b - bgp, U - unreachable # DST-ADDRESS GATEWAY DISTANCE 0 ADS ::/0 pptp-out1 1 1 X S ::/0 2001:470:1f0b:1b4e::1 1 2 ADS fe80::4e/128 pptp-out1 1 [admin@MikroTik] > ping 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 HOST SIZE TTL TIME STATUS fe80::4e 104 64 1ms destination unreachable fe80::4e 104 64 0ms destination unreachable fe80::4e 104 64 1ms destination unreachable sent=3 received=0 packet-loss=100% [admin@MikroTik] >

shade · Tue Jan 25, 2011 9:40 pm

But I think that's is normal, because routers do not have clientside stateless configuration (and it is documented somewhere)... and you must add address manually, this way works!

[admin@MikroTik] > /ipv6 address add address=2001:470:9de7:11::10/64 interface=pptp-out1 [admin@MikroTik] > ping 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 HOST SIZE TTL TIME STATUS 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 1ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 1ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 1ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 1ms echo reply sent=4 received=4 packet-loss=0% min-rtt=1ms avg-rtt=1ms max-rtt=1ms [admin@MikroTik] > /ipv6 address print Flags: X - disabled, I - invalid, D - dynamic, G - global, L - link-local # ADDRESS INTERFACE ADVERTISE 0 XG 2001:470:1f0b:1b4e::10/64 ether1-gateway no 1 XG 2001:470:9de7:1::1/64 ether2-local-master yes 2 DL fe80::20c:42ff:fe80:b2c6/64 ether2-local-master no 3 DL fe80::20c:42ff:fe80:b2c8/64 ether4-local-slave no 4 DL fe80::20c:42ff:fe80:b2c9/64 ether5-local-slave no 5 DL fe80::20c:42ff:fe80:b2c5/64 vlan3 no 6 DL fe80::20c:42ff:fe80:b2c5/64 vlan10 no 7 DL fe80::20c:42ff:fe80:b2c5/64 vlan1 no 8 DL fe80::20c:42ff:fe80:b2c5/64 ether1-gateway no 9 DL fe80::20c:42ff:fe80:b2c5/64 br0 no 10 DL fe80::b/64 pptp-out1 no 11 G 2001:470:9de7:11::10/64 pptp-out1 yes [admin@MikroTik] > /ipv6 address disable numbers=11 [admin@MikroTik] > /ipv6 address enable numbers=11 [admin@MikroTik] > ping 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 HOST SIZE TTL TIME STATUS 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 1ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 1ms echo reply 2001:470:1f0b:1b4e:4a5b:39ff:fe06:72a1 56 63 0ms echo reply sent=3 received=3 packet-loss=0% min-rtt=0ms avg-rtt=0ms max-rtt=1ms [admin@MikroTik] >

janisk · Wed Jan 26, 2011 9:19 am

RouterOS PPPoE client is not yet capable to get IPv6 prefix from PPPoE-server.

It gets link local address and that is it.

prcek · Wed Jan 26, 2011 10:27 pm

RouterOS PPPoE client is not yet capable to get IPv6 prefix from PPPoE-server.

It gets link local address and that is it.

"not yet" means "still in progress" or could we take it as a final feature? Thank you shade and janisk.

prcek

m4rk0 · 星期四,2011年1月27日12:18

Is anyone doing this? Is there a HOW-TO or anything on WIKI? Anyone share a sample config?

I do not see where the IP 'pools' supports IPv6? I imagine you assign every user a single IPv6 IP and also route them an entire subnet of IPv6?

What about NAT 1:1?
PPPoE (local IPv4) / GATEWAY bind local IPv4 to public IPv6?

prcek · Thu Jan 27, 2011 10:18 am

RouterOS PPPoE client is not yet capable to get IPv6 prefix from PPPoE-server.

It gets link local address and that is it.

I have spent some time thinking about it. Client RouterOS can´t know which of interfaces on the router is local and that´s why RouterOS can´t add the IPv6 address to right (local) interface or bridge. We can do it by hand, that´s no problem here of course. But it could be handy to know, what IPv6 prefix is waiting for client, so we have not to ask the ISP. Client RouterOS PPPoE should add (disabled) IPv6 address to (for example pppoe) interface and next step admin of the client router will decide which interface of the router is local and change (and enable) IPv6 adress to this interface.

prcek

janisk · Thu Jan 27, 2011 10:23 am

as i wrote before - router is not yet capable to receive any IPv6 prefix of any kind in any way. Feature of router receiving prefix is in our TO-DO list.

Thu Jan 27, 2011 10:33 am

Thanks Janis,

We all look forward to this feature being completed.

hci · Fri Feb 04, 2011 1:31 am

IANA ran out of IPv4

http://teamarin.net/2011/02/03/iana-ipv ... depletion/

hci · Fri Feb 04, 2011 2:04 am

With IPv4 we assign PPPoE users a single IPv4. We also have proxy ARP enable on the PPPoE servers Internet gateway facing interface. When the gateway does an ARP request the PPPoE server answers on behalf of the PPPoE user with that IP.

How does this work on PPPoE user with an IPv6 prefix?

fato · Fri Feb 18, 2011 1:30 am

Pote tha to ftiaksete gamo tin panagia sas?

m4rk0 · Wed Feb 23, 2011 11:20 am

IANA ran out of IPv4

http://teamarin.net/2011/02/03/iana-ipv ... depletion/

And we just need to wait

When bigger countries transfer to IPv6 they will free IPv4 space

Beccara · Fri Feb 25, 2011 11:11 am

Say what? Your going to be waiting a looooong time for v4 space to release and it wont be much use as the tipping point to release it will be when most content is on v6

omidkosari · Wed Jul 13, 2011 11:17 am

rigt? So in my case, framed-ipv6-prefix reply attribute is ignored by NAS. PPPoE session set fe80::x/64 route entry only. On both sides (NAS and client). I didnt have time to check what is going wrong, but if you have any ideas please let me know, thanks.

My client is windows 7and i have same problem . with mikrotiks own remote-ipv6-prefix everything works fine but with radius framed-ipv6-prefix the router ignores it .

radius server log shows

Framed-IPv6-Prefix: ['2001:470:9de7:11::/64']

mikrotik log shows

Jun/26/2011 16:18:31 radius,debug,packet Framed-IPv6-Prefix = 0x326130303a316365303a303a31303a3a Jun/26/2011 16:18:31 radius,debug,packet 2f3634

routeros version 5.4 . any trick ?!

janisk · Wed Jul 13, 2011 1:08 pm

RouterOS expects string that contains ipv6 prefix in form /

mpoisel · Fri Sep 09, 2011 1:36 pm

RouterOS expects string that contains ipv6 prefix in form /

could you provide some example please?

docmarius · Wed Nov 02, 2011 9:43 pm

Now RouterOS 5.8 supports IPv6 pools.
Any chances for PPPv6 with PD on next release?

hci · Fri Nov 04, 2011 1:10 am

Now RouterOS 5.8 supports IPv6 pools.
Any chances for PPPv6 with PD on next release?

What is "PPPv6 with PD"?

jadu · Fri Nov 04, 2011 11:12 am

PPPoE IPv6 PD=Prefix Delegation

brainy · Wed Nov 30, 2011 4:25 pm

rigt? So in my case, framed-ipv6-prefix reply attribute is ignored by NAS. PPPoE session set fe80::x/64 route entry only. On both sides (NAS and client). I didnt have time to check what is going wrong, but if you have any ideas please let me know, thanks.

My client is windows 7and i have same problem . with mikrotiks own remote-ipv6-prefix everything works fine but with radius framed-ipv6-prefix the router ignores it .

radius server log shows
Code:Select all
Framed-IPv6-Prefix: ['2001:470:9de7:11::/64']
mikrotik log shows
Code:Select all
Jun/26/2011 16:18:31 radius,debug,packet Framed-IPv6-Prefix = 0x326130303a316365303a303a31303a3a Jun/26/2011 16:18:31 radius,debug,packet 2f3634
routeros version 5.4 . any trick ?!

Hi, i have the same problem currently. Did you manage to get that fixed?

JanZorz · Thu Dec 01, 2011 3:41 pm

Hi,

in ROS 5.9 we have DHCPv6 client, that is able to receive PD. Fine, great job, we were all waiting for that.

What I'm missing is the configuration option to which ethernet port(s) we automatically provision first (and consequent) /64 subnets.

On Linux you have this option in dhcp6c config:

config 'interface' 'lan'
option 'enabled' '1'
option 'sla_id' '1'
option 'sla_len' '8'

Anyone any idea?

thnx, Jan Zorz

janisk · Fri Dec 02, 2011 11:45 am

assignment of acquired address on one of the interfaces with /64 and advertise is coming.

JanZorz · Fri Dec 02, 2011 1:52 pm

assignment of acquired address on one of the interfaces with /64 and advertise is coming.

Can you predict when? We are testing IPv6 over pppoe for our ISP and having DHCPv6 client with PD is great, but we need PD assignment to interfaces in order to render your device useful

If there is any ROS beta image with this functionality I would be happy to test.

Many thnx, Jan Zorz

janisk · Fri Dec 02, 2011 3:06 pm

if client is receiving /64 prefix, you could get the value and assign this address using script

因为没有testing-build可用this feature automated in dhcp-client.

JanZorz · Fri Dec 02, 2011 3:21 pm

if client is receiving /64 prefix, you could get the value and assign this address using script

因为没有testing-build可用this feature automated in dhcp-client.

No way, /64 PD is a very wrong thing to do for ISP.

They do /56 for residential user and /48 for business - as it is supposed to be.

We'll wait 'till you get this feature in ROS and use OpenWRT meantime for testing...

Thnx very much, Jan Zorz

P.S: I would be very happy if you can notify me (or send ROS image) when this comes to beta. Thnx.

janisk · Fri Dec 02, 2011 3:30 pm

you do not need addresses for intermediate network stuff, just the end equipment of users should get the address. There you give out addresses with /64

Anyway, nothing that script cannot do.

JanZorz · Fri Dec 02, 2011 3:39 pm

you do not need addresses for intermediate network stuff, just the end equipment of users should get the address. There you give out addresses with /64

Anyway, nothing that script cannot do.

We actually put global IPv6 addresses on eth interfaces, just for later troubleshooting purposes. Problem is, that PD currently changes everytime you connect and our ISP is fixing that (as this is not a good practice). Would be too much of a hassle to script all that probably (or not?)

We'll wait for non-scripting solution

Thnx, Jan

docmarius · Sat Dec 03, 2011 11:45 pm

Wow, we are a step closer...

There seems to be one single issue open... ROS should support SLAAC (RFC 4862) for ipv6 address assignement on the PPP connection. I think this is the only missing link.

hci · Sun Dec 04, 2011 7:14 pm

if client is receiving /64 prefix, you could get the value and assign this address using script

因为没有testing-build可用this feature automated in dhcp-client.

No way, /64 PD is a very wrong thing to do for ISP.

They do /56 for residential user and /48 for business - as it is supposed to be.

I would think that /64 for residential and /48 for business would be fine. Not sure why home users would need to subnet?

vegard · Sun Dec 04, 2011 7:46 pm

I would think that /64 for residential and /48 for business would be fine. Not sure why home users would need to subnet?

Dude. V6 is introduced to give everyone enough addresses. There are enough addresses available and no reason to conserve.

It's not difficult to imagine that users would like one subnet for private stuff, guest hotspot, gaming devices etc.

JanZorz · Mon Dec 05, 2011 11:17 am

I would think that /64 for residential and /48 for business would be fine. Not sure why home users would need to subnet?

Dude. V6 is introduced to give everyone enough addresses. There are enough addresses available and no reason to conserve.

It's not difficult to imagine that users would like one subnet for private stuff, guest hotspot, gaming devices etc.

完全正确。请避免/ 64 PD给最终用户would like to avoid users inventing disgusting mechanisms used for splitting /64 in more subnets. Users will use more subnets, as already mentioned, for guest wifi, private wifi, sensors, local net, media devices etc etc. It's up to vendors, what they'll recommend, but currently IPv6 subnetting looks good. At least /56 for residential users is advisable, if not /48.

For this reason Mark Townsley and I also proposed a RIPE IPv6 policy change to RIPE community - extending initial allocation size (/32) to /29 without additional documentation. Why? Because small ISPs that would like to implement 6RD with one 6RD domain will be able to give out just /64 with /32 initial allocation (32bits+32bits). So, this is so wrong that we proposed a policy change.

http://www.ripe.net/ripe/policies/proposals/2011-04

Please don't use /64 as PD for users. Fullstop

Cheers, Jan Zorz

marekm · Wed Dec 14, 2011 11:16 pm

For this reason Mark Townsley and I also proposed a RIPE IPv6 policy change to RIPE community - extending initial allocation size (/32) to /29 without additional documentation. Why? Because small ISPs that would like to implement 6RD with one 6RD domain will be able to give out just /64 with /32 initial allocation (32bits+32bits). So, this is so wrong that we proposed a policy change.

http://www.ripe.net/ripe/policies/proposals/2011-04

Please don't use /64 as PD for users. Fullstop

Cheers, Jan Zorz

How about the /48 IPv6 PI assignments - with ~200 customers in a rural area I'm still a bit too small to become a LIR (for cost reasons), so can't get a /32 but only a /48 IPv6 PI (currently have /23 IPv4 PI for BGP with two uplinks, and just starting to look into IPv6) - is there a chance to propose something to allow making them a little larger, like /44 or /40? As it is now, I can get a /48 and at best give customers something like /60 (for up to sixteen /64 subnets per customer, looks like a reasonable compromise) to have some room for growth (with /56, there are only 256 of them in a /48). The RIPE IPv6 policy is probably still in flux, at least that's the impression I got when asked my sponsoring LIR for IPv6 PI space.

JanZorz · Mon Feb 27, 2012 9:52 am

How about the /48 IPv6 PI assignments - with ~200 customers in a rural area I'm still a bit too small to become a LIR (for cost reasons), so can't get a /32 but only a /48 IPv6 PI (currently have /23 IPv4 PI for BGP with two uplinks, and just starting to look into IPv6) - is there a chance to propose something to allow making them a little larger, like /44 or /40? As it is now, I can get a /48 and at best give customers something like /60 (for up to sixteen /64 subnets per customer, looks like a reasonable compromise) to have some room for growth (with /56, there are only 256 of them in a /48). The RIPE IPv6 policy is probably still in flux, at least that's the impression I got when asked my sponsoring LIR for IPv6 PI space.

According to RIPE policy, you *MUST NOT* use IPv6 PI space for assignments to *any* customers. IPv6 PI space is only for your infrastructure, meant for enterprise companies that are multihomed.

You don't get IPv6 PI space if you even mention subsequent customers of yours. Once one of my friends wanted IPv6 PI for small datacenter and told to RIPE IPRA that he'll host servers from other people in his datacenter and give one IPv6 address to one server (all servers in one /64 flat subnet) - and he was refused with explanation, that PI space must not be used for assignments to customers. For that you must be LIR.

Jan

Who is online