Recursive PCP
Apple Inc.
1 Infinite Loop
Cupertino
California
95014
USA
+1 408 974 3207
cheshire@apple.com
PCP working group
The Port Control Protocol (PCP) allows clients to request
explicit dynamic inbound and outbound port mappings in their their
closest on-path NAT, Firewall, or other middlebox. However, in
today's world, there may be more than one NAT on the path between
a client and the public Internet.
This document describes how the closest on-path middlebox
generates a corresponding upstream PCP request to the next closest
on-path middlebox, to request an appropriate explicit dynamic port
mapping in that middlebox too. Applied recursively, this generates
the necessary chain of port mappings in any number of middleboxes
on the path between the client and the public Internet.
When NAT Port Mapping Protcol was first
created in 2004, a common network configuration was that a residential
customer received a single public routable IPv4 address from their ISP,
and had a single NAT gateway serving multiple computers in their home.
Consequently, creating appropriate mappings in that single NAT gateway
was sufficient to provide full Internet connectivity.
In today's world, with public routable IPv4 addresses becoming
less readily available, it is increasingly common for customers to
receive a private address from their ISP, and the ISP uses a NAT
gateway of its own to translate those packets before sending them
out onto the public Internet. This means that there is likely
to be more than on NAT on the path between client machines and
the public Internet:
If a residential customer receives a translated address
from their ISP, and then installs their own residential NAT
gateway to share that address between multiple client devices
in their home, then there are at least two NAT gateways on the
path between client devices and the public Internet.
If a mobile phone customer receives a translated address
from their mobile phone carrier, and uses "Personal Hotspot"
or "Internet Sharing" software on their mobile phone to make
Wi-Fi Internet access available to other client devices, then
there are at least two NAT gateways on the path between those
client devices and the public Internet.
If a hotel guest connects a portable Wi-Fi gateway, such as
an Apple AirPort Express, to their hotel room Ethernet port to
share their room's Internet connection between their phone, their
iPad, and their laptop computer, then packets from the client devices
may traverse the hotel guest's portable NAT, the hotel network's
NAT, and the ISP's NAT before reaching the public Internet.
While it is possible, in theory, that client devices could somehow
discover all the NATs on the path, and communicate with each one
separately using Port Control Protocol
(NAT-PMP's IETF Standards Track successor), in practice it's not
clear how client devices would reliably learn this information.
Since the NAT gateways are installed and operated by different
individuals and organizations, no single entity has knowledge of
all the NATs on the path. Also, even if a client device could
somehow know all the NATs on the path, requiring a client device
to communicate separately with all of them imposes unreasonable
complexity on PCP clients, many of which are expected to be simple
low-cost devices.
In addition, this goes against the spirit of NAT gateways.
The main purpose of a NAT gateway is to make multiple local
client devices making outgoing TCP connections to appear,
from the point of view of everything upstream of the NAT gateway,
to be a single client device making outgoing TCP connections.
In the same spirit, it makes sense for a PCP-capable NAT gateway
to make multiple local client devices requesting port mappings to
appear, from the point of view of everything upstream of the NAT
gateway, to be a single client device requesting port mappings.
This document specifies how a PCP-capable NAT gateway
uses Recursive PCP to create the appearance of being a single
device, from the point of view of the upstream network.
Upon receipt of a PCP request from a local PCP client, a
Recursive PCP server first examines its local mapping table to see
if it already has a valid active mapping matching the Internal
Address and Internal Port (and in the case of PEER requests,
remote peer) given in the request. If so, it sends a reply giving
the outermost External Address and Port (previously learned using
Recursive PCP, as described below).
If the Recursive PCP server does not already have a valid
active mapping for this request, then it allocates an available
port on its external interface. We assume for the sake of this
description that the address of its external interface is itself
a private address, subject to translation by an upstream NAT.
The Recursive PCP server then constructs an appropriate
corresponding PCP request of its own, and sends it to its upstream
NAT. In the upstream PCP request, the Internal Address and
Internal Port are the Recursive PCP server's own external address
and port just allocated for this mapping. The Suggested External
Address and Port in the upstream PCP request may be copied from
the original PCP request. How the Recursive PCP server knows the
destination IP address for its upstream PCP request is outside the
scope of this document, but this may be achieved in a zero-configuration
manner using PCP Anycast.
Upon receipt of a PCP reply giving the outermost (i.e. publicly
routable) External Address, Port and Lifetime, the Recursive PCP
server records this information in its own mapping table and
relays the information to the requesting downstream PCP client in a PCP reply.
The Recursive PCP server therefore records, among other things,
the following information in its mapping table:
Client's Internal Address and Port.
External Address and Port allocated by this Recursive PCP server.
Outermost External Address and Port allocated by the upstream PCP server.
Mapping lifetime (also dictated by the upstream PCP server).
A Recursive PCP server SHOULD implement Optimized Hairpin
Routing. What this means is the following:
If a Recursive PCP server observes an outgoing packet
arriving on its internal interface that is addressed to
an External Address and Port appearing in the NAT gateway's
own mapping table, then the NAT gateway SHOULD (after creating a
new outbound mapping if one does not already exist) rewrite the
packet appropriately and deliver it to the internal client
currently allocated that External Address and Port.
If a Recursive PCP server observes an outgoing packet
arriving on its internal interface which is addressed to an
Outermost External Address and Port appearing in the NAT
gateway's own mapping table, then the NAT gateway SHOULD do
likewise: create a new outbound mapping if one does not
already exist, and then rewrite the packet appropriately and
deliver it to the internal client currently allocated that
Outermost External Address and Port. This is not necessary
for successful communication, but for efficiency. Without
this Optimized Hairpin Routing, the packet will be delivered
all the way to the outermost NAT gateway, which will then
perform standard hairpin translation and send it back.
Using knowledge of the Outermost External Address and Port,
this rewriting can be anticipated and performed locally, which
will typically offer higher throughput and lower latency than
sending it all the way to the outermost NAT gateway and back.
The protocol specified is described as "recursive" because
of the following properties:
Although the description above refers to an incoming
PCP request being received from a local PCP client, that
local PCP client could itself be a Recursive PCP server
relaying a request on behalf of one of its own local downstream PCP clients
(which could itself be another Recursive PCP server, and so on).
The fact that the Recursive PCP server receiving the request
does not need to be aware of this or take any special action,
is an important simplifying property of the protocol.
The purpose of a NAT gateway is to make many local client
devices appear to be a single client device, and the purpose
of a Recursive PCP server is to make many local client
devices making PCP requests appear to be a single client
device making PCP requests.
Although the description above suggests that the upstream
PCP server may be the final outermost NAT gateway, in fact that
upstream PCP server could itself be another Recursive PCP
server making requests to its own upstream PCP server,
and relaying back the corresponding replies.
Any recursive algorithm needs a mechanism to terminate the
recursion at the appropriate point. This termination of recursion
can be achieved in a variety of ways:
An ISP's NAT gateway could be configured to know
that it is the outermost NAT gateway, and consequently
does not need to relay PCP requests upstream. In fact,
it may be the case that many large-scale NATs of the kind
used by ISPs may simply not implement Recursive PCP,
thereby naturally terminating the recursion at that point.
A NAT gateway could determine automatically that if its
external address is not one of the known private addresses
then
its external address is a public routable IP address, and
consequently it does not need to relay PCP requests upstream.
A NAT gateway could attempt sending PCP requests upstream,
and upon failing to receive any positive reply (e.g. receiving
ICMP host unreachable, ICMP port unreachable, or a timeout)
conclude that it does not need to relay PCP requests upstream.
No IANA actions are required by this document.
No new security concerns are raised by use of Recursive PCP.
Since the purpose of a NAT gateway is to enable multiple client devices
to appear as a single client device to the upstream network,
a NAT gateway implementing Recursive PCP maintains this property,
appearing to the upstream network to be a single client device
using PCP to request port mappings for itself.
Whether those port mappings are for multiple processes running on
multiple CPUs connected via an internal bus in a single computer,
or multiple processes running on multiple CPUs connected via an IP
network, is transparent to the external network.
Port Control Protocol (PCP)
The Port Control Protocol allows an IPv6 or IPv4
host to control how incoming IPv6 or IPv4 packets are
translated and forwarded by a network address translator (NAT)
or simple firewall, and also allows a host to optimize its
outgoing NAT keepalive messages.
PCP Anycast Address
NAT Port Mapping Protocol (NAT-PMP)