Using several IPs (IP Unnumbered) over one PPPoE session

[meteor713]

I am using RP-PPPoE Client & Server.
I have question about "using several IPs over one PPPoE" in the linux
box.

For some Asian FTTH (Fiber To The Home) service supports several
public IPs via one PPPoE Tunnel.
(They say, it's IP unnumbered service)

Once, the PPPoE session is established, the router will be a DHCP
server in order to distribute IP addresses to client PCs.

For example...
The one PPPoE session's is as below.

Peer IP: 10.1.1.1
Local IP: 210.10.1.0 / 255.255.255.252 (PPPoE Server give IP subnet
!)

So, the router is runnning on Bridge mode, and the bridge interface's
IP address is 210.10.1.1
Then the router will give rest of IPs (210.10.1.2~6) to
PC's inside of the network.

The PC can have real public IP and can do more things than Private IP
with NAT.


So~~~~ Can I do this with Linux Box ?

[Clifford Kite]

Post by meteor713
I am using RP-PPPoE Client & Server.
I have question about "using several IPs over one PPPoE" in the linux
box.
For some Asian FTTH (Fiber To The Home) service supports several
public IPs via one PPPoE Tunnel.
(They say, it's IP unnumbered service)
Once, the PPPoE session is established, the router will be a DHCP
server in order to distribute IP addresses to client PCs.
For example...
The one PPPoE session's is as below.
Peer IP: 10.1.1.1
Local IP: 210.10.1.0 / 255.255.255.252 (PPPoE Server give IP subnet
!)
So, the router is runnning on Bridge mode, and the bridge interface's
IP address is 210.10.1.1

I'm not sure why the router would be in Bridge mode using an IP address
considering that you are using PPPoE. But then I'm not familiar with
exactly how a bridge between networks functions.

Post by meteor713
Then the router will give rest of IPs (210.10.1.2~6) to
PC's inside of the network.

How sure are you about "210.10.1.2~6" IP addresses? That would seem to
imply a 210.10.1.0/28 subnet (255.255.255.248).

Post by meteor713
The PC can have real public IP and can do more things than Private IP
with NAT.
So~~~~ Can I do this with Linux Box ?

I *think* so. But the "Local IP" does _appear_ to be a subnet
specification, 210.10.1.0/30. In that case you should have exactly
2 IP addresses you can use, 210.10.1.1 and 210.10.1.2. The network
address would be 210.10.1.0 and the broadcast address 210.10.1.3.
And if the bridge does indeed use 210.10.1.1 then that leaves you
with just one IP address.

Be aware (you probably already are) that I have no hands-on experience
that could help to make this reply authoritative.

[meteor713]

Post by Clifford Kite

Post by meteor713
I am using RP-PPPoE Client & Server.
I have question about "using several IPs over one PPPoE" in the linux
box.
For some Asian FTTH (Fiber To The Home) service supports several
public IPs via one PPPoE Tunnel.
(They say, it's IP unnumbered service)
Once, the PPPoE session is established, the router will be a DHCP
server in order to distribute IP addresses to client PCs.
For example...
The one PPPoE session's is as below.
Peer IP: 10.1.1.1
Local IP: 210.10.1.0 / 255.255.255.252 (PPPoE Server give IP subnet
!)
So, the router is runnning on Bridge mode, and the bridge interface's
IP address is 210.10.1.1

I'm not sure why the router would be in Bridge mode using an IP address
considering that you are using PPPoE. But then I'm not familiar with
exactly how a bridge between networks functions.

The router must be assigned a IP from the given 6 IPs. An External ethernet
(eth1)is connected to ADSL Modem, and An Internal Ethernet(eth0) is connected
Internal HUB which connects several PCs. Those PCs must have public IP address
for special pplication.
in the Bridge Mode , eth0 and eth1 is working same as Switch HUB's port.
PPPoE Session is P-to-P tunnel, so Router must have one IP.
The router's IP is a IP address of both (?) eth0 and eth1.

Post by Clifford Kite

Post by meteor713
Then the router will give rest of IPs (210.10.1.2~6) to
PC's inside of the network.

How sure are you about "210.10.1.2~6" IP addresses? That would seem to
imply a 210.10.1.0/28 subnet (255.255.255.248).

You right! In Japan, They say PPPoE IP-8 Service.

Post by Clifford Kite

Post by meteor713
The PC can have real public IP and can do more things than Private IP
with NAT.
So~~~~ Can I do this with Linux Box ?

I *think* so. But the "Local IP" does _appear_ to be a subnet
specification, 210.10.1.0/30. In that case you should have exactly
2 IP addresses you can use, 210.10.1.1 and 210.10.1.2. The network
address would be 210.10.1.0 and the broadcast address 210.10.1.3.
And if the bridge does indeed use 210.10.1.1 then that leaves you
with just one IP address.
Be aware (you probably already are) that I have no hands-on experience
that could help to make this reply authoritative.

My question is about how to receive a subnet address from PPPoE session.
In the RFC of IPCP, There's no Subnet related portion.

Anyone has experience with PPPoE session with subnet local IP assignments?

Thanks Clifford.

[Clifford Kite]

You don't need to reply to this if you feel it wouldn't be productive,
I won't be offended.

Post by meteor713

Post by Clifford Kite
How sure are you about "210.10.1.2~6" IP addresses? That would seem to
imply a 210.10.1.0/28 subnet (255.255.255.248).

^^^^^^^^^^^^^
This should be 210.10.1.0/29.

...

Post by meteor713
My question is about how to receive a subnet address from PPPoE session.
In the RFC of IPCP, There's no Subnet related portion.

That's correct, IPCP doesn't negotiate subnets, and, AFAIK, neither
does PPPoE.

Why would you need to get a subnet address from the ISP? It seems to
me that the subnet should be static when you are allowed to have more
than one IP address. But maybe not.

Does the router IP address assigned by the ISP sometimes change between
connections?

If it does change then the IP address assigned to the router PPP
interface for each connection must be in whatever subnet is assigned,
and might always be the network address plus 1, as it was for the
"Local IP: 210.10.1.0 / 255.255.255.252" subnet.

Now let's try to clarify some things for me by backing up to

Post by meteor713

Post by Clifford Kite

Post by meteor713
The one PPPoE session's is as below.
Peer IP: 10.1.1.1
Local IP: 210.10.1.0 / 255.255.255.252 (PPPoE Server give IP subnet
!)
So, the router is runnning on Bridge mode, and the bridge interface's
IP address is 210.10.1.1

Can you identify which process generates the "Peer IP" and the "Local
IP" messages?

Where do these messages appear, in a terminal window or in a log file?

The "Local IP: 210.10.1.0 / 255.255.255.252" does appear to be a subnet
specification. If you are allowed 6 host IP addresses then why is this
subnet showing up when it allows only 2 host IP addresses?

Isn't 210.10.1.1 the router's PPP interface local IP address rather than
the (Ethernet) bridge's IP address?

It could be useful if we could see the outputs of ifconfig and "route -n"
during a PPPoE connection.

[meteor713]

Post by Clifford Kite

Post by meteor713

Post by Clifford Kite
How sure are you about "210.10.1.2~6" IP addresses? That would seem to
imply a 210.10.1.0/28 subnet (255.255.255.248).

^^^^^^^^^^^^^
This should be 210.10.1.0/29.

Yep~ You absolutely right!

Post by Clifford Kite

Post by meteor713
My question is about how to receive a subnet address from PPPoE session.
In the RFC of IPCP, There's no Subnet related portion.

That's correct, IPCP doesn't negotiate subnets, and, AFAIK, neither
does PPPoE.
Why would you need to get a subnet address from the ISP? It seems to
me that the subnet should be static when you are allowed to have more
than one IP address. But maybe not.
Does the router IP address assigned by the ISP sometimes change between
connections?

In Japan, FTTH service (NTT etc.) gives the dynamic subnet. The given
subnet lasts quite long time, but if I reset the router, I get
different subnet.
If you search Google with "PPPoE IP-8", you will see the Japanese
service about "OCN IP-8" service.

Post by Clifford Kite

Post by meteor713

Post by Clifford Kite

Post by meteor713
Peer IP: 10.1.1.1
Local IP: 210.10.1.0 / 255.255.255.252 (PPPoE Server give IP subnet
!)
So, the router is runnning on Bridge mode, and the bridge interface's
IP address is 210.10.1.1

Can you identify which process generates the "Peer IP" and the "Local
IP" messages?

I meant PPPoE connection.
ifconfig ppp1, it will... actually I want to see those, but the
result is

ppp1 Link encap:Point-to-Point Protocol
inet addr:210.10.1.0 P-t-P:10.1.1.1 Mask:255.255.255.255
UP POINTOPOINT RUNNING NOARP MULTICAST MTU:1492 Metric:1
RX packets:129 errors:0 dropped:0 overruns:0 frame:0
TX packets:316 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:3
RX bytes:8042 (7.8 Kb) TX bytes:18176 (17.7 Kb)

Post by Clifford Kite
Isn't 210.10.1.1 the router's PPP interface local IP address rather than
the (Ethernet) bridge's IP address?

That is another problem.
I setup eth0 and eth1 as bridge br0.
After negotiate PPPoE, 210.10.1.1 ( In real, 210.10.1.0 assigned...
anyway) IP is assigned to br0, not eth1. But RP-PPPoE just does that.

Anyway, here is a full story why I tried to do this.
NTT, or OCN service in Japan, They offer FTTH service, it's over
50Mbps in quite a low price. Sorrowfully, we must buy the router in
quite an expensive price. the router's activity is just connect PPPoE
on one ethernet (WAN) and responses dhcp client on the other ethernet
(LAN). That's all. So I want to replace it with Linux box

In compact, the problem counts two.
1. How can RP-PPPoE get subnet local ip.
2. How can PPPoE assign his IP to br0.

I think I can get etherreal capture file of THE ROUTER's PPPoE nego
packets in this week.

Thank Clifford for your follow-ups.

[Clifford Kite]

Post by meteor713
In Japan, FTTH service (NTT etc.) gives the dynamic subnet. The given
subnet lasts quite long time, but if I reset the router, I get
different subnet.
If you search Google with "PPPoE IP-8", you will see the Japanese
service about "OCN IP-8" service.

Okay, I did a search on google, and then did another search limiting
the language to English. :-/

There was a page at one site which described, in moderate detail,
an ISP PPPoE implementation that was very similar to your situation,
http://www.dslreports.com/faq/10245 for anyone interested. The
information on that page, together with your comments and an faq page
from http://www.tldp.org/HOWTO/DSL-HOWTO/appendix.html, section

8.3.2. Fiber in the Loop (IFITL or FTTC, and FTTH)

are the basis for comments that follow. Now I understand much more than
when we started - but not really enough. This problem has turned out to
much more complex than I thought it was in the beginning.

...

Post by meteor713
ppp1 Link encap:Point-to-Point Protocol
inet addr:210.10.1.0 P-t-P:10.1.1.1 Mask:255.255.255.255
UP POINTOPOINT RUNNING NOARP MULTICAST MTU:1492 Metric:1
RX packets:129 errors:0 dropped:0 overruns:0 frame:0
TX packets:316 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:3
RX bytes:8042 (7.8 Kb) TX bytes:18176 (17.7 Kb)

Post by Clifford Kite
Isn't 210.10.1.1 the router's PPP interface local IP address rather than
the (Ethernet) bridge's IP address?

That is another problem.
I setup eth0 and eth1 as bridge br0.
After negotiate PPPoE, 210.10.1.1 ( In real, 210.10.1.0 assigned...
anyway) IP is assigned to br0, not eth1. But RP-PPPoE just does that.

So it really is the bridge's (br0's) IP address. And apparently the
210.10.1.0 works as the IP address of the PPP interface. This is very
different from a PPPoE connection with a single public IP address, and
not yet common in the United States.

Post by meteor713
Anyway, here is a full story why I tried to do this.
NTT, or OCN service in Japan, They offer FTTH service, it's over
50Mbps in quite a low price. Sorrowfully, we must buy the router in
quite an expensive price. the router's activity is just connect PPPoE
on one ethernet (WAN) and responses dhcp client on the other ethernet
(LAN). That's all. So I want to replace it with Linux box

I'm fairly sure that you need *some* device in addition to the Linux
box to transform the protocol used on the outside lines to the Ethernet
protocol the Linux box uses. The device is commonly called a modem.
Sometimes the modem also functions as a router and/or a bridge.

Post by meteor713
In compact, the problem counts two.
1. How can RP-PPPoE get subnet local ip.
2. How can PPPoE assign his IP to br0.

Regrettably, it's become evident that I'm far below an experience level
required to provide any decent answers, if any exist. Unless there is a
reply from someone that has actually devised and implemented a workaround
that eliminates the router (by using a less expensive modem) then the
router appears to be necessary.

Personally I don't see why SNAT implemented with iptables shouldn't
work just fine for the LAN hosts, but you seem to think otherwise.
Well - perhaps the usable bandwidth allocated to a single public IP
address would be less than that for a subnet of public IP addresses.
And I'm sure you must have a good reason to want public addresses.

[David Efflandt]

Post by meteor713
I am using RP-PPPoE Client & Server.
I have question about "using several IPs over one PPPoE" in the linux
box.
For some Asian FTTH (Fiber To The Home) service supports several
public IPs via one PPPoE Tunnel.
(They say, it's IP unnumbered service)
Once, the PPPoE session is established, the router will be a DHCP
server in order to distribute IP addresses to client PCs.
For example...
The one PPPoE session's is as below.
Peer IP: 10.1.1.1
Local IP: 210.10.1.0 / 255.255.255.252 (PPPoE Server give IP subnet
!)
So, the router is runnning on Bridge mode, and the bridge interface's
IP address is 210.10.1.1
Then the router will give rest of IPs (210.10.1.2~6) to
PC's inside of the network.
The PC can have real public IP and can do more things than Private IP
with NAT.
So~~~~ Can I do this with Linux Box ?

My ISP (SBC ameritech.net) routes all IPs for a static block to the
connecting PPPoE IP (which has netmask 255.255.255.255 and just host and
default routes to gateway) and then the router (or Linux) routes them in
from there. Internal boxes on that public subnet have IP of router as
their default gateway. Not sure whether our sdsl at work uses PPPoE, but
the ISP similarly routes any of our public IPs to WAN IP of router and
internal boxes would use IP of router as default gateway. In either case,
I believe that is just normal routing.

Linux can act as dhcp server (dhcpd as opposed to dhcpcd, which is a
client).

Whether you need to do bridging may depend whether it just means setting a
modem/router to bridge mode so it transparently passes pppoe from Linux,
or whether you need to configure Linux as a bridge for tcp/ip traffic.
Linux can do briding, but proxy arp may be simpler if you do not need to
pass broadcast traffic. For example the following would allow Linux to
answer arp requests on ppp0 for any other routes it has and fairly
transparently pass the traffic through:

echo 1 > /proc/sys/net/ipv4/conf/ppp0/proxy_arp

But you might also need to do that for the LAN interface if DHCP on public
LAN boxes currently shows the remote ppp0 IP as gateway, instead of router
IP as gateway.