RFC 6146






Internet Engineering Task Force (IETF)                        M. Bagnulo
Request for Comments: 6146                                          UC3M
Category: Standards Track                                    P. Matthews
ISSN: 2070-1721                                           Alcatel-Lucent
                                                          I. van Beijnum
                                                          IMDEA Networks
                                                              April 2011


        Stateful NAT64: Network Address and Protocol Translation
                   from IPv6 Clients to IPv4 Servers

Abstract



   This document describes stateful NAT64 translation, which allows
   IPv6-only clients to contact IPv4 servers using unicast UDP, TCP, or
   ICMP.  One or more public IPv4 addresses assigned to a NAT64
   translator are shared among several IPv6-only clients.  When stateful
   NAT64 is used in conjunction with DNS64, no changes are usually
   required in the IPv6 client or the IPv4 server.

Status of This Memo



   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc6146.

















Bagnulo, et al.              Standards Track                    [Page 1]

RFC 6146                     Stateful NAT64                   April 2011


Copyright Notice



   Copyright (c) 2011 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.





































Bagnulo, et al.              Standards Track                    [Page 2]

RFC 6146                     Stateful NAT64                   April 2011


Table of Contents



   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  4
     1.1.  Features of Stateful NAT64 . . . . . . . . . . . . . . . .  5
     1.2.  Overview . . . . . . . . . . . . . . . . . . . . . . . . .  6
       1.2.1.  Stateful NAT64 Solution Elements . . . . . . . . . . .  6
       1.2.2.  Stateful NAT64 Behavior Walk-Through . . . . . . . . .  8
       1.2.3.  Filtering  . . . . . . . . . . . . . . . . . . . . . . 10
   2.  Terminology  . . . . . . . . . . . . . . . . . . . . . . . . . 11
   3.  Stateful NAT64 Normative Specification . . . . . . . . . . . . 14
     3.1.  Binding Information Bases  . . . . . . . . . . . . . . . . 14
     3.2.  Session Tables . . . . . . . . . . . . . . . . . . . . . . 15
     3.3.  Packet Processing Overview . . . . . . . . . . . . . . . . 17
     3.4.  Determining the Incoming Tuple . . . . . . . . . . . . . . 18
     3.5.  Filtering and Updating Binding and Session Information . . 20
       3.5.1.  UDP Session Handling . . . . . . . . . . . . . . . . . 21
         3.5.1.1.  Rules for Allocation of IPv4 Transport
                   Addresses for UDP  . . . . . . . . . . . . . . . . 23
       3.5.2.  TCP Session Handling . . . . . . . . . . . . . . . . . 24
         3.5.2.1.  State Definition . . . . . . . . . . . . . . . . . 24
         3.5.2.2.  State Machine for TCP Processing in the NAT64  . . 25
         3.5.2.3.  Rules for Allocation of IPv4 Transport
                   Addresses for TCP  . . . . . . . . . . . . . . . . 33
       3.5.3.  ICMP Query Session Handling  . . . . . . . . . . . . . 33
       3.5.4.  Generation of the IPv6 Representations of IPv4
               Addresses  . . . . . . . . . . . . . . . . . . . . . . 36
     3.6.  Computing the Outgoing Tuple . . . . . . . . . . . . . . . 36
       3.6.1.  Computing the Outgoing 5-Tuple for TCP, UDP, and
               for ICMP Error Messages Containing a TCP or UDP
               Packets  . . . . . . . . . . . . . . . . . . . . . . . 37
       3.6.2.  Computing the Outgoing 3-Tuple for ICMP Query
               Messages and for ICMP Error Messages Containing an
               ICMP Query . . . . . . . . . . . . . . . . . . . . . . 38
     3.7.  Translating the Packet . . . . . . . . . . . . . . . . . . 38
     3.8.  Handling Hairpinning . . . . . . . . . . . . . . . . . . . 39
   4.  Protocol Constants . . . . . . . . . . . . . . . . . . . . . . 39
   5.  Security Considerations  . . . . . . . . . . . . . . . . . . . 40
     5.1.  Implications on End-to-End Security  . . . . . . . . . . . 40
     5.2.  Filtering  . . . . . . . . . . . . . . . . . . . . . . . . 40
     5.3.  Attacks on NAT64 . . . . . . . . . . . . . . . . . . . . . 41
     5.4.  Avoiding Hairpinning Loops . . . . . . . . . . . . . . . . 42
   6.  Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 43
   7.  Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 43
   8.  References . . . . . . . . . . . . . . . . . . . . . . . . . . 43
     8.1.  Normative References . . . . . . . . . . . . . . . . . . . 43
     8.2.  Informative References . . . . . . . . . . . . . . . . . . 44





Bagnulo, et al.              Standards Track                    [Page 3]

RFC 6146                     Stateful NAT64                   April 2011


1.  Introduction



   This document specifies stateful NAT64, a mechanism for IPv4-IPv6
   transition and IPv4-IPv6 coexistence.  Together with DNS64 [RFC6147],
   these two mechanisms allow an IPv6-only client to initiate
   communications to an IPv4-only server.  They also enable peer-to-peer
   communication between an IPv4 and an IPv6 node, where the
   communication can be initiated when either end uses existing, NAT-
   traversal, peer-to-peer communication techniques, such as Interactive
   Connectivity Establishment (ICE) [RFC5245].  Stateful NAT64 also
   supports IPv4-initiated communications to a subset of the IPv6 hosts
   through statically configured bindings in the stateful NAT64.

   Stateful NAT64 is a mechanism for translating IPv6 packets to IPv4
   packets and vice versa.  The translation is done by translating the
   packet headers according to the IP/ICMP Translation Algorithm defined
   in [RFC6145].  The IPv4 addresses of IPv4 hosts are algorithmically
   translated to and from IPv6 addresses by using the algorithm defined
   in [RFC6052] and an IPv6 prefix assigned to the stateful NAT64 for
   this specific purpose.  The IPv6 addresses of IPv6 hosts are
   translated to and from IPv4 addresses by installing mappings in the
   normal Network Address Port Translation (NAPT) manner [RFC3022].  The
   current specification only defines how stateful NAT64 translates
   unicast packets carrying TCP, UDP, and ICMP traffic.  Multicast
   packets and other protocols, including the Stream Control
   Transmission Protocol (SCTP), the Datagram Congestion Control
   Protocol (DCCP), and IPsec, are out of the scope of this
   specification.

   DNS64 is a mechanism for synthesizing AAAA resource records (RRs)
   from A RRs.  The IPv6 address contained in the synthetic AAAA RR is
   algorithmically generated from the IPv4 address and the IPv6 prefix
   assigned to a NAT64 device by using the same algorithm defined in
   [RFC6052].

   Together, these two mechanisms allow an IPv6-only client (i.e., a
   host with a networking stack that only implements IPv6, a host with a
   networking stack that implements both protocols but with only IPv6
   connectivity, or a host running an IPv6-only application) to initiate
   communications to an IPv4-only server (which is analogous to the
   IPv6-only host above).

   These mechanisms are expected to play a critical role in IPv4-IPv6
   transition and IPv4-IPv6 coexistence.  Due to IPv4 address depletion,
   it is likely that in the future, the new clients will be IPv6-only
   and they will want to connect to the existing IPv4-only servers.  The
   stateful NAT64 and DNS64 mechanisms are easily deployable, since they
   do not require changes to either the IPv6 client or the IPv4 server.



Bagnulo, et al.              Standards Track                    [Page 4]

RFC 6146                     Stateful NAT64                   April 2011


   For basic functionality, the approach only requires the deployment of
   the stateful NAT64 function in the devices connecting an IPv6-only
   network to the IPv4-only network, along with the deployment of a few
   DNS64-enabled name servers accessible to the IPv6-only hosts.  An
   analysis of the application scenarios can be found in [RFC6144].

   For brevity, in the rest of the document, we will refer to the
   stateful NAT64 either as stateful NAT64 or simply as NAT64.

1.1.  Features of Stateful NAT64



   The features of NAT64 are:

   o  NAT64 is compliant with the recommendations for how NATs should
      handle UDP [RFC4787], TCP [RFC5382], and ICMP [RFC5508].  As such,
      NAT64 only supports Endpoint-Independent Mappings and supports
      both Endpoint-Independent and Address-Dependent Filtering.
      Because of the compliance with the aforementioned requirements,
      NAT64 is compatible with current NAT traversal techniques, such as
      ICE [RFC5245], and with other NAT traversal techniques.

   o  In the absence of preexisting state in a NAT64, only IPv6 nodes
      can initiate sessions to IPv4 nodes.  This works for roughly the
      same class of applications that work through IPv4-to-IPv4 NATs.

   o  Depending on the filtering policy used (Endpoint-Independent, or
      Address-Dependent), IPv4-nodes might be able to initiate sessions
      to a given IPv6 node, if the NAT64 somehow has an appropriate
      mapping (i.e., state) for an IPv6 node, via one of the following
      mechanisms:

      *  The IPv6 node has recently initiated a session to the same or
         another IPv4 node.  This is also the case if the IPv6 node has
         used a NAT-traversal technique (such as ICE).

      *  A statically configured mapping exists for the IPv6 node.

   o  IPv4 address sharing: NAT64 allows multiple IPv6-only nodes to
      share an IPv4 address to access the IPv4 Internet.  This helps
      with the forthcoming IPv4 exhaustion.

   o  As currently defined in this NAT64 specification, only TCP, UDP,
      and ICMP are supported.  Support for other protocols (such as
      other transport protocols and IPsec) is to be defined in separate
      documents.






Bagnulo, et al.              Standards Track                    [Page 5]

RFC 6146                     Stateful NAT64                   April 2011


1.2.  Overview



   This section provides a non-normative introduction to NAT64.  This is
   achieved by describing the NAT64 behavior involving a simple setup
   that involves a single NAT64 device, a single DNS64, and a simple
   network topology.  The goal of this description is to provide the
   reader with a general view of NAT64.  It is not the goal of this
   section to describe all possible configurations nor to provide a
   normative specification of the NAT64 behavior.  So, for the sake of
   clarity, only TCP and UDP are described in this overview; the details
   of ICMP, fragmentation, and other aspects of translation are
   purposefully avoided in this overview.  The normative specification
   of NAT64 is provided in Section 3.

   The NAT64 mechanism is implemented in a device that has (at least)
   two interfaces, an IPv4 interface connected to the IPv4 network, and
   an IPv6 interface connected to the IPv6 network.  Packets generated
   in the IPv6 network for a receiver located in the IPv4 network will
   be routed within the IPv6 network towards the NAT64 device.  The
   NAT64 will translate them and forward them as IPv4 packets through
   the IPv4 network to the IPv4 receiver.  The reverse takes place for
   packets generated by hosts connected to the IPv4 network for an IPv6
   receiver.  NAT64, however, is not symmetric.  In order to be able to
   perform IPv6-IPv4 translation, NAT64 requires state.  The state
   contains the binding of an IPv6 address and TCP/UDP port (hereafter
   called an IPv6 transport address) to an IPv4 address and TCP/UDP port
   (hereafter called an IPv4 transport address).

   Such binding state is either statically configured in the NAT64 or it
   is created when the first packet flowing from the IPv6 network to the
   IPv4 network is translated.  After the binding state has been
   created, packets flowing in both directions on that particular flow
   are translated.  The result is that, in the general case, NAT64 only
   supports communications initiated by the IPv6-only node towards an
   IPv4-only node.  Some additional mechanisms (like ICE) or static
   binding configuration can be used to provide support for
   communications initiated by an IPv4-only node to an IPv6-only node.

1.2.1.  Stateful NAT64 Solution Elements



   In this section, we describe the different elements involved in the
   NAT64 approach.

   The main component of the proposed solution is the translator itself.
   The translator has essentially two main parts, the address
   translation mechanism and the protocol translation mechanism.





Bagnulo, et al.              Standards Track                    [Page 6]

RFC 6146                     Stateful NAT64                   April 2011


   Protocol translation from an IPv4 packet header to an IPv6 packet
   header and vice versa is performed according to the IP/ICMP
   Translation Algorithm [RFC6145].

   Address translation maps IPv6 transport addresses to IPv4 transport
   addresses and vice versa.  In order to create these mappings, the
   NAT64 has two pools of addresses: an IPv6 address pool (to represent
   IPv4 addresses in the IPv6 network) and an IPv4 address pool (to
   represent IPv6 addresses in the IPv4 network).

   The IPv6 address pool is one or more IPv6 prefixes assigned to the
   translator itself.  Hereafter, we will call the IPv6 address pool
   Pref64::/n; in the case there is more than one prefix assigned to the
   NAT64, the comments made about Pref64::/n apply to each of them.
   Pref64::/n will be used by the NAT64 to construct IPv4-Converted IPv6
   addresses as defined in [RFC6052].  Due to the abundance of IPv6
   address space, it is possible to assign one or more Pref64::/n, each
   of them being equal to or even bigger than the size of the whole IPv4
   address space.  This allows each IPv4 address to be mapped into a
   different IPv6 address by simply concatenating a Pref64::/n with the
   IPv4 address being mapped and a suffix.  The provisioning of the
   Pref64::/n as well as the address format are defined in [RFC6052].

   The IPv4 address pool is a set of IPv4 addresses, normally a prefix
   assigned by the local administrator.  Since IPv4 address space is a
   scarce resource, the IPv4 address pool is small and typically not
   sufficient to establish permanent one-to-one mappings with IPv6
   addresses.  So, except for the static/manually created ones, mappings
   using the IPv4 address pool will be created and released dynamically.
   Moreover, because of the IPv4 address scarcity, the usual practice
   for NAT64 is likely to be the binding of IPv6 transport addresses
   into IPv4 transport addresses, instead of IPv6 addresses into IPv4
   addresses directly, enabling a higher utilization of the limited IPv4
   address pool.  This implies that NAT64 performs both address and port
   translation.

   Because of the dynamic nature of the IPv6-to-IPv4 address mapping and
   the static nature of the IPv4-to-IPv6 address mapping, it is far
   simpler to allow communications initiated from the IPv6 side toward
   an IPv4 node, whose address is algorithmically mapped into an IPv6
   address, than communications initiated from IPv4-only nodes to an
   IPv6 node.  In that case, an IPv4 address needs to be associated with
   the IPv6 node's address dynamically.

   Using a mechanism such as DNS64, an IPv6 client obtains an IPv6
   address that embeds the IPv4 address of the IPv4 server and sends a
   packet to that IPv6 address.  The packets are intercepted by the
   NAT64 device, which associates an IPv4 transport address out of its



Bagnulo, et al.              Standards Track                    [Page 7]

RFC 6146                     Stateful NAT64                   April 2011


   IPv4 pool to the IPv6 transport address of the initiator, creating
   binding state, so that reply packets can be translated and forwarded
   back to the initiator.  The binding state is kept while packets are
   flowing.  Once the flow stops, and based on a timer, the IPv4
   transport address is returned to the IPv4 address pool so that it can
   be reused for other communications.

   To allow an IPv6 initiator to do a DNS lookup to learn the address of
   the responder, DNS64 [RFC6147] is used to synthesize AAAA RRs from
   the A RRs.  The IPv6 addresses contained in the synthetic AAAA RRs
   contain a Pref64::/n assigned to the NAT64 and the IPv4 address of
   the responder.  The synthetic AAAA RRs are passed back to the IPv6
   initiator, which will initiate an IPv6 communication with an IPv6
   address associated to the IPv4 receiver.  The packet will be routed
   to the NAT64 device, which will create the IPv6-to-IPv4 address
   mapping as described before.

1.2.2.  Stateful NAT64 Behavior Walk-Through



   In this section, we provide a simple example of the NAT64 behavior.
   We consider an IPv6 node that is located in an IPv6-only site and
   that initiates a TCP connection to an IPv4-only node located in the
   IPv4 network.

   The scenario for this case is depicted in the following figure:

             +---------------------+         +---------------+
             |IPv6 network         |         |    IPv4       |
             |           |  +-------------+  |  network      |
             |           |--| Name server |--|               |
             |           |  | with DNS64  |  |  +----+       |
             |  +----+   |  +-------------+  |  | H2 |       |
             |  | H1 |---|         |         |  +----+       |
             |  +----+   |      +-------+    |  192.0.2.1    |
             |2001:db8::1|------| NAT64 |----|               |
             |           |      +-------+    |               |
             |           |         |         |               |
             +---------------------+         +---------------+

   The figure above shows an IPv6 node H1 with an IPv6 address
   2001:db8::1 and an IPv4 node H2 with IPv4 address 192.0.2.1.  H2 has
   h2.example.com as its Fully Qualified Domain Name (FQDN).

   A NAT64 connects the IPv6 network to the IPv4 network.  This NAT64
   uses the Well-Known Prefix 64:ff9b::/96 defined in [RFC6052] to
   represent IPv4 addresses in the IPv6 address space and a single IPv4
   address 203.0.113.1 assigned to its IPv4 interface.  The routing is




Bagnulo, et al.              Standards Track                    [Page 8]

RFC 6146                     Stateful NAT64                   April 2011


   configured in such a way that the IPv6 packets addressed to a
   destination address in 64:ff9b::/96 are routed to the IPv6 interface
   of the NAT64 device.

   Also shown is a local name server with DNS64 functionality.  The
   local name server uses the Well-Known Prefix 64:ff9b::/96 to create
   the IPv6 addresses in the synthetic RRs.

   For this example, assume the typical DNS situation where IPv6 hosts
   have only stub resolvers, and the local name server does the
   recursive lookups.

   The steps by which H1 establishes communication with H2 are:

   1.  H1 performs a DNS query for h2.example.com and receives the
       synthetic AAAA RR from the local name server that implements the
       DNS64 functionality.  The AAAA record contains an IPv6 address
       formed by the Well-Known Prefix and the IPv4 address of H2 (i.e.,
       64:ff9b::192.0.2.1).

   2.  H1 sends a TCP SYN packet to H2.  The packet is sent from a
       source transport address of (2001:db8::1,1500) to a destination
       transport address of (64:ff9b::192.0.2.1,80), where the ports are
       set by H1.

   3.  The packet is routed to the IPv6 interface of the NAT64 (since
       IPv6 routing is configured that way).



   4.  The NAT64 receives the packet and performs the following actions:



       *  The NAT64 selects an unused port (e.g., 2000) on its IPv4
          address 203.0.113.1 and creates the mapping entry
          (2001:db8::1,1500) <--> (203.0.113.1,2000)

       *  The NAT64 translates the IPv6 header into an IPv4 header using
          the IP/ICMP Translation Algorithm [RFC6145].

       *  The NAT64 includes (203.0.113.1,2000) as the source transport
          address in the packet and (192.0.2.1,80) as the destination
          transport address in the packet.  Note that 192.0.2.1 is
          extracted directly from the destination IPv6 address of the
          received IPv6 packet that is being translated.  The
          destination port 80 of the translated packet is the same as
          the destination port of the received IPv6 packet.

   5.  The NAT64 sends the translated packet out of its IPv4 interface
       and the packet arrives at H2.






Bagnulo, et al.              Standards Track                    [Page 9]

RFC 6146                     Stateful NAT64                   April 2011


   6.  H2 node responds by sending a TCP SYN+ACK packet with the
       destination transport address (203.0.113.1,2000) and source
       transport address (192.0.2.1,80).

   7.  Since the IPv4 address 203.0.113.1 is assigned to the IPv4
       interface of the NAT64 device, the packet is routed to the NAT64
       device, which will look for an existing mapping containing
       (203.0.113.1,2000).  Since the mapping (2001:db8::1,1500) <-->
       (203.0.113.1,2000) exists, the NAT64 performs the following
       operations:

       *  The NAT64 translates the IPv4 header into an IPv6 header using
          the IP/ICMP Translation Algorithm [RFC6145].

       *  The NAT64 includes (2001:db8::1,1500) as the destination
          transport address in the packet and (64:ff9b::192.0.2.1,80) as
          the source transport address in the packet.  Note that
          192.0.2.1 is extracted directly from the source IPv4 address
          of the received IPv4 packet that is being translated.  The
          source port 80 of the translated packet is the same as the
          source port of the received IPv4 packet.

   8.  The translated packet is sent out of the IPv6 interface to H1.



   The packet exchange between H1 and H2 continues, and packets are
   translated in the different directions as previously described.

   It is important to note that the translation still works if the IPv6
   initiator H1 learns the IPv6 representation of H2's IPv4 address
   (i.e., 64:ff9b::192.0.2.1) through some scheme other than a DNS
   lookup.  This is because the DNS64 processing does NOT result in any
   state being installed in the NAT64 and because the mapping of the
   IPv4 address into an IPv6 address is the result of concatenating the
   Well-Known Prefix to the original IPv4 address.

1.2.3.  Filtering



   NAT64 may do filtering, which means that it only allows a packet in
   through an interface under certain circumstances.  The NAT64 can
   filter IPv6 packets based on the administrative rules to create
   entries in the binding and session tables.  The filtering can be
   flexible and general, but the idea of the filtering is to provide the
   administrators necessary control to avoid denial-of-service (DoS)
   attacks that would result in exhaustion of the NAT64's IPv4 address,
   port, memory, and CPU resources.  Filtering techniques of incoming
   IPv6 packets are not specific to the NAT64 and therefore are not
   described in this specification.




Bagnulo, et al.              Standards Track                   [Page 10]

RFC 6146                     Stateful NAT64                   April 2011


   Filtering of IPv4 packets, on the other hand, is tightly coupled to
   the NAT64 state and therefore is described in this specification.  In
   this document, we consider that the NAT64 may do no filtering, or it
   may filter incoming IPv4 packets.

   NAT64 filtering of incoming IPv4 packets is consistent with the
   recommendations of [RFC4787] and [RFC5382].  Because of that, the
   NAT64 as specified in this document supports both Endpoint-
   Independent Filtering and Address-Dependent Filtering, both for TCP
   and UDP as well as filtering of ICMP packets.

   If a NAT64 performs Endpoint-Independent Filtering of incoming IPv4
   packets, then an incoming IPv4 packet is dropped unless the NAT64 has
   state for the destination transport address of the incoming IPv4
   packet.

   If a NAT64 performs Address-Dependent Filtering of incoming IPv4
   packets, then an incoming IPv4 packet is dropped unless the NAT64 has
   state involving the destination transport address of the IPv4
   incoming packet and the particular source IP address of the incoming
   IPv4 packet.

2.  Terminology



   This section provides a definitive reference for all the terms used
   in this document.

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in RFC 2119 [RFC2119].

   The following additional terms are used in this document:

   3-Tuple:  The tuple (source IP address, destination IP address, ICMP
      Identifier).  A 3-tuple uniquely identifies an ICMP Query session.
      When an ICMP Query session flows through a NAT64, each session has
      two different 3-tuples: one with IPv4 addresses and one with IPv6
      addresses.

   5-Tuple:  The tuple (source IP address, source port, destination IP
      address, destination port, transport protocol).  A 5-tuple
      uniquely identifies a UDP/TCP session.  When a UDP/TCP session
      flows through a NAT64, each session has two different 5-tuples:
      one with IPv4 addresses and one with IPv6 addresses.







Bagnulo, et al.              Standards Track                   [Page 11]

RFC 6146                     Stateful NAT64                   April 2011


   BIB:  Binding Information Base.  A table of bindings kept by a NAT64.
      Each NAT64 has a BIB for each translated protocol.  An
      implementation compliant to this document would have a BIB for
      TCP, one for UDP, and one for ICMP Queries.  Additional BIBs would
      be added to support other protocols, such as SCTP.

   Endpoint-Independent Mapping:  In NAT64, using the same mapping for
      all the sessions involving a given IPv6 transport address of an
      IPv6 host (irrespectively of the transport address of the IPv4
      host involved in the communication).  Endpoint-Independent Mapping
      is important for peer-to-peer communication.  See [RFC4787] for
      the definition of the different types of mappings in IPv4-to-IPv4
      NATs.

   Filtering, Endpoint-Independent:  The NAT64 only filters incoming
      IPv4 packets destined to a transport address for which there is no
      state in the NAT64, regardless of the source IPv4 transport
      address.  The NAT forwards any packets destined to any transport
      address for which it has state.  In other words, having state for
      a given transport address is sufficient to allow any packets back
      to the internal endpoint.  See [RFC4787] for the definition of the
      different types of filtering in IPv4-to-IPv4 NATs.

   Filtering, Address-Dependent:  The NAT64 filters incoming IPv4
      packets destined to a transport address for which there is no
      state (similar to the Endpoint-Independent Filtering).
      Additionally, the NAT64 will filter out incoming IPv4 packets
      coming from a given IPv4 address X and destined for a transport
      address for which it has state if the NAT64 has not sent packets
      to X previously (independently of the port used by X).  In other
      words, for receiving packets from a specific IPv4 endpoint, it is
      necessary for the IPv6 endpoint to send packets first to that
      specific IPv4 endpoint's IP address.

   Hairpinning:  Having a packet do a "U-turn" inside a NAT and come
      back out the same side as it arrived on.  If the destination IPv6
      address and its embedded IPv4 address are both assigned to the
      NAT64 itself, then the packet is being sent to another IPv6 host
      connected to the same NAT64.  Such a packet is called a 'hairpin
      packet'.  A NAT64 that forwards hairpin packets back to the IPv6
      host is defined as supporting "hairpinning".  Hairpinning support
      is important for peer-to-peer applications, as there are cases
      when two different hosts on the same side of a NAT can only
      communicate using sessions that hairpin through the NAT.  Hairpin
      packets can be either TCP or UDP.  More detailed explanation of
      hairpinning and examples for the UDP case can be found in
      [RFC4787].




Bagnulo, et al.              Standards Track                   [Page 12]

RFC 6146                     Stateful NAT64                   April 2011


   ICMP Query packet:  ICMP packets that are not ICMP error messages.
      For ICMPv6, ICMPv6 Query Messages are the ICMPv6 Informational
      messages as defined in [RFC4443].  For ICMPv4, ICMPv4 Query
      messages are all ICMPv4 messages that are not ICMPv4 error
      messages.

   Mapping or Binding:  A mapping between an IPv6 transport address and
      a IPv4 transport address or a mapping between an (IPv6 address,
      ICMPv6 Identifier) pair and an (IPv4 address, ICMPv4 Identifier)
      pair.  Used to translate the addresses and ports / ICMP
      Identifiers of packets flowing between the IPv6 host and the IPv4
      host.  In NAT64, the IPv4 address and port / ICMPv4 Identifier is
      always one assigned to the NAT64 itself, while the IPv6 address
      and port / ICMPv6 Identifier belongs to some IPv6 host.

   Session:  The flow of packets between two different hosts.  This may
      be TCP, UDP, or ICMP Queries.  In NAT64, typically one host is an
      IPv4 host, and the other one is an IPv6 host.  However, due to
      hairpinning, both hosts might be IPv6 hosts.

   Session table:  A table of sessions kept by a NAT64.  Each NAT64 has
      three session tables, one for TCP, one for UDP, and one for ICMP
      Queries.

   Stateful NAT64:  A function that has per-flow state that translates
      IPv6 packets to IPv4 packets and vice versa, for TCP, UDP, and
      ICMP.  The NAT64 uses binding state to perform the translation
      between IPv6 and IPv4 addresses.  In this document, we also refer
      to stateful NAT64 simply as NAT64.

   Stateful NAT64 device:  The device where the NAT64 function is
      executed.  In this document, we also refer to stateful NAT64
      device simply as NAT64 device.

   Transport Address:  The combination of an IPv6 or IPv4 address and a
      port.  Typically written as (IP address,port), e.g.,
      (192.0.2.15,8001).

   Tuple:  Refers to either a 3-tuple or a 5-tuple as defined above.

   For a detailed understanding of this document, the reader should also
   be familiar with NAT terminology [RFC4787].









Bagnulo, et al.              Standards Track                   [Page 13]

RFC 6146                     Stateful NAT64                   April 2011


3.  Stateful NAT64 Normative Specification

   A NAT64 is a device with at least one IPv6 interface and at least one
   IPv4 interface.  Each NAT64 device MUST have at least one unicast /n
   IPv6 prefix assigned to it, denoted Pref64::/n.  Additional
   considerations about the Pref64::/n are presented in Section 3.5.4.
   A NAT64 MUST have one or more unicast IPv4 addresses assigned to it.

   A NAT64 uses the following conceptual dynamic data structures:

   o  UDP Binding Information Base

   o  UDP Session Table

   o  TCP Binding Information Base

   o  TCP Session Table

   o  ICMP Query Binding Information Base

   o  ICMP Query Session Table

   These tables contain information needed for the NAT64 processing.
   The actual division of the information into six tables is done in
   order to ease the description of the NAT64 behavior.  NAT64
   implementations are free to use different data structures but they
   MUST store all the required information, and the externally visible
   outcome MUST be the same as the one described in this document.

   The notation used is the following: uppercase letters are IPv4
   addresses; uppercase letters with a prime(') are IPv6 addresses;
   lowercase letters are ports; IPv6 prefixes of length n are indicated
   by "P::/n"; mappings are indicated as "(X,x) <--> (Y',y)".

3.1.  Binding Information Bases



   A NAT64 has three Binding Information Bases (BIBs): one for TCP, one
   for UDP, and one for ICMP Queries.  In the case of UDP and TCP BIBs,
   each BIB entry specifies a mapping between an IPv6 transport address
   and an IPv4 transport address:

      (X',x) <--> (T,t)

   where X' is some IPv6 address, T is an IPv4 address, and x and t are
   ports.  T will always be one of the IPv4 addresses assigned to the
   NAT64.  The BIB has then two columns: the BIB IPv6 transport address
   and the BIB IPv4 transport address.  A given IPv6 or IPv4 transport
   address can appear in at most one entry in a BIB: for example,



Bagnulo, et al.              Standards Track                   [Page 14]

RFC 6146                     Stateful NAT64                   April 2011


   (2001:db8::17, 49832) can appear in at most one TCP and at most one
   UDP BIB entry.  TCP and UDP have separate BIBs because the port
   number space for TCP and UDP are distinct.  If the BIBs are
   implemented as specified in this document, it results in
   Endpoint-Independent Mappings in the NAT64.  The information in the
   BIBs is also used to implement Endpoint-Independent Filtering.
   (Address-Dependent Filtering is implemented using the session tables
   described below.)

   In the case of the ICMP Query BIB, each ICMP Query BIB entry
   specifies a mapping between an (IPv6 address, ICMPv6 Identifier) pair
   and an (IPv4 address, ICMPv4 Identifier) pair.

      (X',i1) <--> (T,i2)

   where X' is some IPv6 address, T is an IPv4 address, i1 is an ICMPv6
   Identifier, and i2 is an ICMPv4 Identifier.  T will always be one of
   the IPv4 addresses assigned to the NAT64.  A given (IPv6 or IPv4
   address, ICMPv6 or ICMPv4 Identifier) pair can appear in at most one
   entry in the ICMP Query BIB.

   Entries in any of the three BIBs can be created dynamically as the
   result of the flow of packets as described in Section 3.5, but they
   can also be created manually by an administrator.  NAT64
   implementations SHOULD support manually configured BIB entries for
   any of the three BIBs.  Dynamically created entries are deleted from
   the corresponding BIB when the last session associated with the BIB
   entry is removed from the session table.  Manually configured BIB
   entries are not deleted when there is no corresponding Session Table
   Entry and can only be deleted by the administrator.

3.2.  Session Tables



   A NAT64 also has three session tables: one for TCP sessions, one for
   UDP sessions, and one for ICMP Query sessions.  Each entry keeps
   information on the state of the corresponding session.  In the TCP
   and UDP session tables, each entry specifies a mapping between a pair
   of IPv6 transport addresses and a pair of IPv4 transport addresses:

      (X',x),(Y',y) <--> (T,t),(Z,z)

   where X' and Y' are IPv6 addresses, T and Z are IPv4 addresses, and
   x, y, z, and t are ports.  T will always be one of the IPv4 addresses
   assigned to the NAT64.  Y' is always the IPv6 representation of the
   IPv4 address Z, so Y' is obtained from Z using the algorithm applied
   by the NAT64 to create IPv6 representations of IPv4 addresses. y will
   always be equal to z.




Bagnulo, et al.              Standards Track                   [Page 15]

RFC 6146                     Stateful NAT64                   April 2011


   For each TCP or UDP Session Table Entry (STE), there are then five
   columns.  The terminology used for the STE columns is from the
   perspective of an incoming IPv6 packet being translated into an
   outgoing IPv4 packet.  The columns are:

      The STE source IPv6 transport address; (X',x) in the example
      above.

      The STE destination IPv6 transport address; (Y',y) in the example
      above.

      The STE source IPv4 transport address; (T,t) in the example above.

      The STE destination IPv4 transport address; (Z,z) in the example
      above.

      The STE lifetime.

   In the ICMP Query session table, each entry specifies a mapping
   between a 3-tuple of IPv6 source address, IPv6 destination address,
   and ICMPv6 Identifier and a 3-tuple of IPv4 source address, IPv4
   destination address, and ICMPv4 Identifier:

      (X',Y',i1) <--> (T,Z,i2)

   where X' and Y' are IPv6 addresses, T and Z are IPv4 addresses, i1 is
   an ICMPv6 Identifier, and i2 is an ICMPv4 Identifier.  T will always
   be one of the IPv4 addresses assigned to the NAT64.  Y' is always the
   IPv6 representation of the IPv4 address Z, so Y' is obtained from Z
   using the algorithm applied by the NAT64 to create IPv6
   representations of IPv4 addresses.

   For each ICMP Query Session Table Entry (STE), there are then seven
   columns:

      The STE source IPv6 address; X' in the example above.

      The STE destination IPv6 address; Y' in the example above.

      The STE ICMPv6 Identifier; i1 in the example above.

      The STE source IPv4 address; T in the example above.

      The STE destination IPv4 address; Z in the example above.

      The STE ICMPv4 Identifier; i2 in the example above.

      The STE lifetime.



Bagnulo, et al.              Standards Track                   [Page 16]

RFC 6146                     Stateful NAT64                   April 2011


3.3.  Packet Processing Overview



   The NAT64 uses the session state information to determine when the
   session is completed, and also uses session information for Address-
   Dependent Filtering.  A session can be uniquely identified by either
   an incoming tuple or an outgoing tuple.

   For each TCP or UDP session, there is a corresponding BIB entry,
   uniquely specified by either the source IPv6 transport address (in
   the IPv6 --> IPv4 direction) or the destination IPv4 transport
   address (in the IPv4 --> IPv6 direction).  For each ICMP Query
   session, there is a corresponding BIB entry, uniquely specified by
   either the source IPv6 address and ICMPv6 Identifier (in the IPv6 -->
   IPv4 direction) or the destination IPv4 address and the ICMPv4
   Identifier (in the IPv4 --> IPv6 direction).  However, for all the
   BIBs, a single BIB entry can have multiple corresponding sessions.
   When the last corresponding session is deleted, if the BIB entry was
   dynamically created, the BIB entry is deleted.

   The NAT64 will receive packets through its interfaces.  These packets
   can be either IPv6 packets or IPv4 packets, and they may carry TCP
   traffic, UDP traffic, or ICMP traffic.  The processing of the packets
   will be described next.  In the case that the processing is common to
   all the aforementioned types of packets, we will refer to the packet
   as the incoming IP packet in general.  In the case that the
   processing is specific to IPv6 packets, we will explicitly refer to
   the incoming packet as an incoming IPv6 packet; analogous terminology
   will apply in the case of processing that is specific to IPv4
   packets.

   The processing of an incoming IP packet takes the following steps:

   1.  Determining the incoming tuple

   2.  Filtering and updating binding and session information

   3.  Computing the outgoing tuple

   4.  Translating the packet

   5.  Handling hairpinning

   The details of these steps are specified in the following
   subsections.







Bagnulo, et al.              Standards Track                   [Page 17]

RFC 6146                     Stateful NAT64                   April 2011


   This breakdown of the NAT64 behavior into processing steps is done
   for ease of presentation.  A NAT64 MAY perform the steps in a
   different order or MAY perform different steps, but the externally
   visible outcome MUST be the same as the one described in this
   document.

3.4.  Determining the Incoming Tuple



   This step associates an incoming tuple with every incoming IP packet
   for use in subsequent steps.  In the case of TCP, UDP, and ICMP error
   packets, the tuple is a 5-tuple consisting of the source IP address,
   source port, destination IP address, destination port, and transport
   protocol.  In case of ICMP Queries, the tuple is a 3-tuple consisting
   of the source IP address, destination IP address, and ICMP
   Identifier.

   If the incoming IP packet contains a complete (un-fragmented) UDP or
   TCP protocol packet, then the 5-tuple is computed by extracting the
   appropriate fields from the received packet.

   If the incoming packet is a complete (un-fragmented) ICMP Query
   message (i.e., an ICMPv4 Query message or an ICMPv6 Informational
   message), the 3-tuple is the source IP address, the destination IP
   address, and the ICMP Identifier.

   If the incoming IP packet contains a complete (un-fragmented) ICMP
   error message containing a UDP or a TCP packet, then the incoming
   5-tuple is computed by extracting the appropriate fields from the IP
   packet embedded inside the ICMP error message.  However, the role of
   source and destination is swapped when doing this: the embedded
   source IP address becomes the destination IP address in the incoming
   5-tuple, the embedded source port becomes the destination port in the
   incoming 5-tuple, etc.  If it is not possible to determine the
   incoming 5-tuple (perhaps because not enough of the embedded packet
   is reproduced inside the ICMP message), then the incoming IP packet
   MUST be silently discarded.

   If the incoming IP packet contains a complete (un-fragmented) ICMP
   error message containing an ICMP error message, then the packet is
   silently discarded.

   If the incoming IP packet contains a complete (un-fragmented) ICMP
   error message containing an ICMP Query message, then the incoming
   3-tuple is computed by extracting the appropriate fields from the IP
   packet embedded inside the ICMP error message.  However, the role of
   source and destination is swapped when doing this: the embedded
   source IP address becomes the destination IP address in the incoming
   3-tuple, the embedded destination IP address becomes the source



Bagnulo, et al.              Standards Track                   [Page 18]

RFC 6146                     Stateful NAT64                   April 2011


   address in the incoming 3-tuple, and the embedded ICMP Identifier is
   used as the ICMP Identifier of the incoming 3-tuple.  If it is not
   possible to determine the incoming 3-tuple (perhaps because not
   enough of the embedded packet is reproduced inside the ICMP message),
   then the incoming IP packet MUST be silently discarded.

   If the incoming IP packet contains a fragment, then more processing
   may be needed.  This specification leaves open the exact details of
   how a NAT64 handles incoming IP packets containing fragments, and
   simply requires that the external behavior of the NAT64 be compliant
   with the following conditions:

      The NAT64 MUST handle fragments.  In particular, NAT64 MUST handle
      fragments arriving out of order, conditional on the following:

      *  The NAT64 MUST limit the amount of resources devoted to the
         storage of fragmented packets in order to protect from DoS
         attacks.

      *  As long as the NAT64 has available resources, the NAT64 MUST
         allow the fragments to arrive over a time interval.  The time
         interval SHOULD be configurable and the default value MUST be
         of at least FRAGMENT_MIN.

      *  The NAT64 MAY require that the UDP, TCP, or ICMP header be
         completely contained within the fragment that contains fragment
         offset equal to zero.

      For incoming packets carrying TCP or UDP fragments with a non-zero
      checksum, NAT64 MAY elect to queue the fragments as they arrive
      and translate all fragments at the same time.  In this case, the
      incoming tuple is determined as documented above to the un-
      fragmented packets.  Alternatively, a NAT64 MAY translate the
      fragments as they arrive, by storing information that allows it to
      compute the 5-tuple for fragments other than the first.  In the
      latter case, subsequent fragments may arrive before the first, and
      the rules (in the bulleted list above) about how the NAT64 handles
      (out-of-order) fragments apply.

      For incoming IPv4 packets carrying UDP packets with a zero
      checksum, if the NAT64 has enough resources, the NAT64 MUST
      reassemble the packets and MUST calculate the checksum.  If the
      NAT64 does not have enough resources, then it MUST silently
      discard the packets.  The handling of fragmented and un-fragmented
      UDP packets with a zero checksum as specified above deviates from
      that specified in [RFC6145].





Bagnulo, et al.              Standards Track                   [Page 19]

RFC 6146                     Stateful NAT64                   April 2011


      Implementers of NAT64 should be aware that there are a number of
      well-known attacks against IP fragmentation; see [RFC1858] and
      [RFC3128].  Implementers should also be aware of additional issues
      with reassembling packets at high rates, described in [RFC4963].

   If the incoming packet is an IPv6 packet that contains a protocol
   other than TCP, UDP, or ICMPv6 in the last Next Header, then the
   packet SHOULD be discarded and, if the security policy permits, the
   NAT64 SHOULD send an ICMPv6 Destination Unreachable error message
   with Code 4 (Port Unreachable) to the source address of the received
   packet.  NOTE: This behavior may be updated by future documents that
   define how other protocols such as SCTP or DCCP are processed by
   NAT64.

   If the incoming packet is an IPv4 packet that contains a protocol
   other than TCP, UDP, or ICMPv4, then the packet SHOULD be discarded
   and, if the security policy permits, the NAT64 SHOULD send an ICMPv4
   Destination Unreachable error message with Code 2 (Protocol
   Unreachable) to the source address of the received packet.  NOTE:
   This behavior may be updated by future documents that define how
   other protocols such as SCTP or DCCP are processed by NAT64.

3.5.  Filtering and Updating Binding and Session Information



   This step updates binding and session information stored in the
   appropriate tables.  This step may also filter incoming packets, if
   desired.

   The details of this step depend on the protocol, i.e., UDP, TCP, or
   ICMP.  The behaviors for UDP, TCP, and ICMP Queries are described in
   Section 3.5.1, Section 3.5.2, and Section 3.5.3, respectively.  For
   the case of ICMP error messages, they do not affect in any way either
   the BIBs or the session tables, so there is no processing resulting
   from these messages in this section.  ICMP error message processing
   continues in Section 3.6.

   Irrespective of the transport protocol used, the NAT64 MUST silently
   discard all incoming IPv6 packets containing a source address that
   contains the Pref64::/n.  This is required in order to prevent
   hairpinning loops as described in Section 5.  In addition, the NAT64
   MUST only process incoming IPv6 packets that contain a destination
   address that contains Pref64::/n.  Likewise, the NAT64 MUST only
   process incoming IPv4 packets that contain a destination address that
   belongs to the IPv4 pool assigned to the NAT64.







Bagnulo, et al.              Standards Track                   [Page 20]

RFC 6146                     Stateful NAT64                   April 2011


3.5.1.  UDP Session Handling



   The following state information is stored for a UDP session:

      Binding:(X',x),(Y',y) <--> (T,t),(Z,z)

      Lifetime: a timer that tracks the remaining lifetime of the UDP
      session.  When the timer expires, the UDP session is deleted.  If
      all the UDP sessions corresponding to a dynamically created UDP
      BIB entry are deleted, then the UDP BIB entry is also deleted.

   An IPv6 incoming packet with an incoming tuple with source transport
   address (X',x) and destination transport address (Y',y) is processed
   as follows:

      The NAT64 searches for a UDP BIB entry that contains the BIB IPv6
      transport address that matches the IPv6 source transport address
      (X',x).  If such an entry does not exist, the NAT64 tries to
      create a new entry (if resources and policy permit).  The source
      IPv6 transport address of the packet (X',x) is used as the BIB
      IPv6 transport address, and the BIB IPv4 transport address is set
      to (T,t), which is allocated using the rules defined in
      Section 3.5.1.1.  The result is a BIB entry as follows: (X',x)
      <--> (T,t).

      The NAT64 searches for the Session Table Entry corresponding to
      the incoming 5-tuple.  If no such entry is found, the NAT64 tries
      to create a new entry (if resources and policy permit).  The
      information included in the session table is as follows:

      *  The STE source IPv6 transport address is set to (X',x), i.e.,
         the source IPv6 transport address contained in the received
         IPv6 packet.

      *  The STE destination IPv6 transport address is set to (Y',y),
         i.e., the destination IPv6 transport address contained in the
         received IPv6 packet.

      *  The STE source IPv4 transport address is extracted from the
         corresponding UDP BIB entry, i.e., it is set to (T,t).

      *  The STE destination IPv4 transport is set to (Z(Y'),y), y being
         the same port as the STE destination IPv6 transport address and
         Z(Y') being algorithmically generated from the IPv6 destination
         address (i.e., Y') using the reverse algorithm (see
         Section 3.5.4).





Bagnulo, et al.              Standards Track                   [Page 21]

RFC 6146                     Stateful NAT64                   April 2011


      The result is a Session Table Entry as follows:
      (X',x),(Y',y) <--> (T,t),(Z(Y'),y)

      The NAT64 sets (or resets) the timer in the Session Table Entry to
      the maximum session lifetime.  The maximum session lifetime MAY be
      configurable, and the default SHOULD be at least UDP_DEFAULT.  The
      maximum session lifetime MUST NOT be less than UDP_MIN.  The
      packet is translated and forwarded as described in the following
      sections.

   An IPv4 incoming packet, with an incoming tuple with source IPv4
   transport address (W,w) and destination IPv4 transport address (T,t)
   is processed as follows:

      The NAT64 searches for a UDP BIB entry that contains the BIB IPv4
      transport address matching (T,t), i.e., the IPv4 destination
      transport address in the incoming IPv4 packet.  If such an entry
      does not exist, the packet MUST be dropped.  An ICMP error message
      with Type 3 (Destination Unreachable) MAY be sent to the original
      sender of the packet.

      If the NAT64 applies Address-Dependent Filters on its IPv4
      interface, then the NAT64 checks to see if the incoming packet is
      allowed according to the Address-Dependent Filtering rule.  To do
      this, it searches for a Session Table Entry with an STE source
      IPv4 transport address equal to (T,t), i.e., the destination IPv4
      transport address in the incoming packet, and STE destination IPv4
      address equal to W, i.e., the source IPv4 address in the incoming
      packet.  If such an entry is found (there may be more than one),
      packet processing continues.  Otherwise, the packet is discarded.
      If the packet is discarded, then an ICMP error message MAY be sent
      to the original sender of the packet.  The ICMP error message, if
      sent, has Type 3 (Destination Unreachable) and Code 13
      (Communication Administratively Prohibited).

      In case the packet is not discarded in the previous processing
      (either because the NAT64 is not filtering or because the packet
      is compliant with the Address-Dependent Filtering rule), then the
      NAT64 searches for the Session Table Entry containing the STE
      source IPv4 transport address equal to (T,t) and the STE
      destination IPv4 transport address equal to (W,w).  If no such
      entry is found, the NAT64 tries to create a new entry (if
      resources and policy permit).  In case a new UDP Session Table
      Entry is created, it contains the following information:

      *  The STE source IPv6 transport address is extracted from the
         corresponding UDP BIB entry.




Bagnulo, et al.              Standards Track                   [Page 22]

RFC 6146                     Stateful NAT64                   April 2011


      *  The STE destination IPv6 transport address is set to (Y'(W),w),
         w being the same port w as the source IPv4 transport address
         and Y'(W) being the IPv6 representation of W, generated using
         the algorithm described in Section 3.5.4.

      *  The STE source IPv4 transport address is set to (T,t), i.e.,
         the destination IPv4 transport addresses contained in the
         received IPv4 packet.

      *  The STE destination IPv4 transport is set to (W,w), i.e., the
         source IPv4 transport addresses contained in the received IPv4
         packet.

      The NAT64 sets (or resets) the timer in the Session Table Entry to
      the maximum session lifetime.  The maximum session lifetime MAY be
      configurable, and the default SHOULD be at least UDP_DEFAULT.  The
      maximum session lifetime MUST NOT be less than UDP_MIN.  The
      packet is translated and forwarded as described in the following
      sections.

3.5.1.1.  Rules for Allocation of IPv4 Transport Addresses for UDP



   When a new UDP BIB entry is created for a source transport address of
   (S',s), the NAT64 allocates an IPv4 transport address for this BIB
   entry as follows:

      If there exists some other BIB entry containing S' as the IPv6
      address and mapping it to some IPv4 address T, then the NAT64
      SHOULD use T as the IPv4 address.  Otherwise, use any IPv4 address
      of the IPv4 pool assigned to the NAT64 to be used for translation.

      If the port s is in the Well-Known port range 0-1023, and the
      NAT64 has an available port t in the same port range, then the
      NAT64 SHOULD allocate the port t.  If the NAT64 does not have a
      port available in the same range, the NAT64 MAY assign a port t
      from another range where it has an available port.  (This behavior
      is recommended in REQ 3-a of [RFC4787].)

      If the port s is in the range 1024-65535, and the NAT64 has an
      available port t in the same port range, then the NAT64 SHOULD
      allocate the port t.  If the NAT64 does not have a port available
      in the same range, the NAT64 MAY assign a port t from another
      range where it has an available port.  (This behavior is
      recommended in REQ 3-a of [RFC4787].)

      The NAT64 SHOULD preserve the port parity (odd/even), as per
      Section 4.2.2 of [RFC4787]).




Bagnulo, et al.              Standards Track                   [Page 23]

RFC 6146                     Stateful NAT64                   April 2011


      In all cases, the allocated IPv4 transport address (T,t) MUST NOT
      be in use in another entry in the same BIB, but can be in use in
      other BIBs (e.g., the UDP and TCP BIBs).

   If it is not possible to allocate an appropriate IPv4 transport
   address or create a BIB entry, then the packet is discarded.  The
   NAT64 SHOULD send an ICMPv6 Destination Unreachable error message
   with Code 3 (Address Unreachable).

3.5.2.  TCP Session Handling



   In this section, we describe how the TCP BIB and session table are
   populated.  We do so by defining the state machine that the NAT64
   uses for TCP.  We first describe the states and the information
   contained in them, and then we describe the actual state machine and
   state transitions.

3.5.2.1.  State Definition



   The following state information is stored for a TCP session:

      Binding:(X',x),(Y',y) <--> (T,t),(Z,z)

      Lifetime: a timer that tracks the remaining lifetime of the TCP
      session.  When the timer expires, the TCP session is deleted.  If
      all the TCP sessions corresponding to a TCP BIB entry are deleted,
      then the dynamically created TCP BIB entry is also deleted.

   Because the TCP session inactivity lifetime is set to at least 2
   hours and 4 minutes (as per [RFC5382]), it is important that each TCP
   Session Table Entry corresponds to an existing TCP session.  In order
   to do that, for each TCP session established, the TCP connection
   state is tracked using the following state machine.

   The states are as follows:

      CLOSED: Analogous to [RFC0793], CLOSED is a fictional state
      because it represents the state when there is no state for this
      particular 5-tuple, and therefore no connection.

      V4 INIT: An IPv4 packet containing a TCP SYN was received by the
      NAT64, implying that a TCP connection is being initiated from the
      IPv4 side.  The NAT64 is now waiting for a matching IPv6 packet
      containing the TCP SYN in the opposite direction.







Bagnulo, et al.              Standards Track                   [Page 24]

RFC 6146                     Stateful NAT64                   April 2011


      V6 INIT: An IPv6 packet containing a TCP SYN was received,
      translated, and forwarded by the NAT64, implying that a TCP
      connection is being initiated from the IPv6 side.  The NAT64 is
      now waiting for a matching IPv4 packet containing the TCP SYN in
      the opposite direction.

      ESTABLISHED: Represents an open connection, with data able to flow
      in both directions.

      V4 FIN RCV: An IPv4 packet containing a TCP FIN was received by
      the NAT64, data can still flow in the connection, and the NAT64 is
      waiting for a matching TCP FIN in the opposite direction.

      V6 FIN RCV: An IPv6 packet containing a TCP FIN was received by
      the NAT64, data can still flow in the connection, and the NAT64 is
      waiting for a matching TCP FIN in the opposite direction.

      V6 FIN + V4 FIN RCV: Both an IPv4 packet containing a TCP FIN and
      an IPv6 packet containing an TCP FIN for this connection were
      received by the NAT64.  The NAT64 keeps the connection state alive
      and forwards packets in both directions for a short period of time
      to allow remaining packets (in particular, the ACKs) to be
      delivered.

      TRANS: The lifetime of the state for the connection is set to
      TCP_TRANS minutes either because a packet containing a TCP RST was
      received by the NAT64 for this connection or simply because the
      lifetime of the connection has decreased and there are only
      TCP_TRANS minutes left.  The NAT64 will keep the state for the
      connection for TCP_TRANS minutes, and if no other data packets for
      that connection are received, the state for this connection is
      then terminated.

3.5.2.2.  State Machine for TCP Processing in the NAT64



   The state machine used by the NAT64 for the TCP session processing is
   depicted next.  The described state machine handles all TCP segments
   received through the IPv6 and IPv4 interface.  There is one state
   machine per TCP connection that is potentially established through
   the NAT64.  After bootstrapping of the NAT64 device, all TCP sessions
   are in CLOSED state.  As we mention above, the CLOSED state is a
   fictional state when there is no state for that particular connection
   in the NAT64.  It should be noted that there is one state machine per
   connection, so only packets belonging to a given connection are
   inputs to the state machine associated to that connection.  In other
   words, when in the state machine below we state that a packet is
   received, it is implicit that the incoming 5-tuple of the data packet
   matches to the one of the state machine.



Bagnulo, et al.              Standards Track                   [Page 25]

RFC 6146                     Stateful NAT64                   April 2011


   A TCP segment with the SYN flag set that is received through the IPv6
   interface is called a V6 SYN, similarly, V4 SYN, V4 FIN, V6 FIN, V6
   FIN + V4 FIN, V6 RST, and V4 RST.

   The figure presents a simplified version of the state machine; refer
   to the text for the full specification of the state machine.

                                      +-----------------------------+
                                      |                             |
                                      V                             |
                       V6       +------+      V4                    |
                  +----SYN------|CLOSED|-----SYN------+             |
                  |             +------+              |             |
                  |                ^                  |             |
                  |                |TCP_TRANS T.O.    |             |
                  V                |                  V             |
              +-------+         +-------+          +-------+        |
              |V6 INIT|         | TRANS |          |V4 INIT|        |
              +-------+         +-------+          +-------+        |
                 |               |    ^               |             |
                 |         data pkt   |               |             |
                 |               |  V4 or V6 RST      |             |
                 |               |  TCP_EST T.O.      |             |
              V4 SYN             V    |              V6 SYN         |
                 |          +--------------+          |             |
                 +--------->| ESTABLISHED  |<---------+             |
                            +--------------+                        |
                              |           |                         |
                          V4 FIN       V6 FIN                       |
                              |           |                         |
                              V           V                         |
                      +---------+       +----------+                |
                      | V4 FIN  |       |  V6 FIN  |                |
                      |   RCV   |       |    RCV   |                |
                      +---------+       +----------+                |
                              |           |                         |
                          V6 FIN       V4 FIN                 TCP_TRANS
                              |           |                        T.O.
                              V           V                         |
                         +---------------------+                    |
                         | V4 FIN + V6 FIN RCV |--------------------+
                         +---------------------+

   We next describe the state information and the transitions.







Bagnulo, et al.              Standards Track                   [Page 26]

RFC 6146                     Stateful NAT64                   April 2011


   *** CLOSED ***

   If a V6 SYN is received with an incoming tuple with source transport
   address (X',x) and destination transport address (Y',y) (this is the
   case of a TCP connection initiated from the IPv6 side), the
   processing is as follows:

   1.  The NAT64 searches for a TCP BIB entry that matches the IPv6
       source transport address (X',x).

          If such an entry does not exist, the NAT64 tries to create a
          new BIB entry (if resources and policy permit).  The BIB IPv6
          transport address is set to (X',x), i.e., the source IPv6
          transport address of the packet.  The BIB IPv4 transport
          address is set to an IPv4 transport address allocated using
          the rules defined in Section 3.5.2.3.  The processing of the
          packet continues as described in bullet 2.

          If the entry already exists, then the processing continues as
          described in bullet 2.

   2.  Then the NAT64 tries to create a new TCP session entry in the TCP
       session table (if resources and policy permit).  The information
       included in the session table is as follows:

          The STE source IPv6 transport address is set to (X',x), i.e.,
          the source transport address contained in the received V6 SYN
          packet.

          The STE destination IPv6 transport address is set to (Y',y),
          i.e., the destination transport address contained in the
          received V6 SYN packet.

          The STE source IPv4 transport address is set to the BIB IPv4
          transport address of the corresponding TCP BIB entry.

          The STE destination IPv4 transport address contains the port y
          (i.e., the same port as the IPv6 destination transport
          address) and the IPv4 address that is algorithmically
          generated from the IPv6 destination address (i.e., Y') using
          the reverse algorithm as specified in Section 3.5.4.

          The lifetime of the TCP Session Table Entry is set to at least
          TCP_TRANS (the transitory connection idle timeout as defined
          in [RFC5382]).

   3.  The state of the session is moved to V6 INIT.




Bagnulo, et al.              Standards Track                   [Page 27]

RFC 6146                     Stateful NAT64                   April 2011


   4.  The NAT64 translates and forwards the packet as described in the
       following sections.

   If a V4 SYN packet is received with an incoming tuple with source
   IPv4 transport address (Y,y) and destination IPv4 transport address
   (X,x) (this is the case of a TCP connection initiated from the IPv4
   side), the processing is as follows:

      If the security policy requires silently dropping externally
      initiated TCP connections, then the packet is silently discarded.

      Else, if the destination transport address contained in the
      incoming V4 SYN (i.e., X,x) is not in use in the TCP BIB, then:

         The NAT64 tries to create a new Session Table Entry in the TCP
         session table (if resources and policy permit), containing the
         following information:

         +  The STE source IPv4 transport address is set to (X,x), i.e.,
            the destination transport address contained in the V4 SYN.

         +  The STE destination IPv4 transport address is set to (Y,y),
            i.e., the source transport address contained in the V4 SYN.

         +  The STE transport IPv6 source address is left unspecified
            and may be populated by other protocols that are out of the
            scope of this specification.

         +  The STE destination IPv6 transport address contains the port
            y (i.e., the same port as the STE destination IPv4 transport
            address) and the IPv6 representation of Y (i.e., the IPv4
            address of the STE destination IPv4 transport address),
            generated using the algorithm described in Section 3.5.4.

         The state is moved to V4 INIT.

         The lifetime of the STE entry is set to TCP_INCOMING_SYN as per
         [RFC5382], and the packet is stored.  The result is that the
         NAT64 will not drop the packet based on the filtering, nor
         create a BIB entry.  Instead, the NAT64 will only create the
         Session Table Entry and store the packet.  The motivation for
         this is to support simultaneous open of TCP connections.

      If the destination transport address contained in the incoming V4
      SYN (i.e., X,x) is in use in the TCP BIB, then:






Bagnulo, et al.              Standards Track                   [Page 28]

RFC 6146                     Stateful NAT64                   April 2011


         The NAT64 tries to create a new Session Table Entry in the TCP
         session table (if resources and policy permit), containing the
         following information:

         +  The STE source IPv4 transport address is set to (X,x), i.e.,
            the destination transport address contained in the V4 SYN.

         +  The STE destination IPv4 transport address is set to (Y,y),
            i.e., the source transport address contained in the V4 SYN.

         +  The STE transport IPv6 source address is set to the IPv6
            transport address contained in the corresponding TCP BIB
            entry.

         +  The STE destination IPv6 transport address contains the port
            y (i.e., the same port as the STE destination IPv4 transport
            address) and the IPv6 representation of Y (i.e., the IPv4
            address of the STE destination IPv4 transport address),
            generated using the algorithm described in Section 3.5.4.

         The state is moved to V4 INIT.

         If the NAT64 is performing Address-Dependent Filtering, the
         lifetime of the STE entry is set to TCP_INCOMING_SYN as per
         [RFC5382], and the packet is stored.  The motivation for
         creating the Session Table Entry and storing the packet
         (instead of simply dropping the packet based on the filtering)
         is to support simultaneous open of TCP connections.

         If the NAT64 is not performing Address-Dependent Filtering, the
         lifetime of the STE is set to at least TCP_TRANS (the
         transitory connection idle timeout as defined in [RFC5382]),
         and it translates and forwards the packet as described in the
         following sections.

   For any other packet belonging to this connection:

      If there is a corresponding entry in the TCP BIB, the packet
      SHOULD be translated and forwarded if the security policy allows
      doing so.  The state remains unchanged.

      If there is no corresponding entry in the TCP BIB, the packet is
      silently discarded.








Bagnulo, et al.              Standards Track                   [Page 29]

RFC 6146                     Stateful NAT64                   April 2011


   *** V4 INIT ***

   If a V6 SYN is received with incoming tuple with source transport
   address (X',x) and destination transport address (Y',y), then the
   lifetime of the TCP Session Table Entry is set to at least the
   maximum session lifetime.  The value for the maximum session lifetime
   MAY be configurable, but it MUST NOT be less than TCP_EST (the
   established connection idle timeout as defined in [RFC5382]).  The
   default value for the maximum session lifetime SHOULD be set to
   TCP_EST.  The packet is translated and forwarded.  The state is moved
   to ESTABLISHED.

   If the lifetime expires, an ICMP Port Unreachable error (Type 3, Code
   3) containing the IPv4 SYN packet stored is sent back to the source
   of the v4 SYN, the Session Table Entry is deleted, and the state is
   moved to CLOSED.

   For any other packet, the packet SHOULD be translated and forwarded
   if the security policy allows doing so.  The state remains unchanged.

   *** V6 INIT ***

   If a V4 SYN is received (with or without the ACK flag set), with an
   incoming tuple with source IPv4 transport address (Y,y) and
   destination IPv4 transport address (X,x), then the state is moved to
   ESTABLISHED.  The lifetime of the TCP Session Table Entry is set to
   at least the maximum session lifetime.  The value for the maximum
   session lifetime MAY be configurable, but it MUST NOT be less than
   TCP_EST (the established connection idle timeout as defined in
   [RFC5382]).  The default value for the maximum session lifetime
   SHOULD be set to TCP_EST.  The packet is translated and forwarded.

   If the lifetime expires, the Session Table Entry is deleted, and the
   state is moved to CLOSED.

   If a V6 SYN packet is received, the packet is translated and
   forwarded.  The lifetime of the TCP Session Table Entry is set to at
   least TCP_TRANS.  The state remains unchanged.

   For any other packet, the packet SHOULD be translated and forwarded
   if the security policy allows doing so.  The state remains unchanged.

   *** ESTABLISHED ***

   If a V4 FIN packet is received, the packet is translated and
   forwarded.  The state is moved to V4 FIN RCV.





Bagnulo, et al.              Standards Track                   [Page 30]

RFC 6146                     Stateful NAT64                   April 2011


   If a V6 FIN packet is received, the packet is translated and
   forwarded.  The state is moved to V6 FIN RCV.

   If a V4 RST or a V6 RST packet is received, the packet is translated
   and forwarded.  The lifetime is set to TCP_TRANS and the state is
   moved to TRANS.  (Since the NAT64 is uncertain whether the peer will
   accept the RST packet, instead of moving the state to CLOSED, it
   moves to TRANS, which has a shorter lifetime.  If no other packets
   are received for this connection during the short timer, the NAT64
   assumes that the peer has accepted the RST packet and moves to
   CLOSED.  If packets keep flowing, the NAT64 assumes that the peer has
   not accepted the RST packet and moves back to the ESTABLISHED state.
   This is described below in the TRANS state processing description.)

   If any other packet is received, the packet is translated and
   forwarded.  The lifetime of the TCP Session Table Entry is set to at
   least the maximum session lifetime.  The value for the maximum
   session lifetime MAY be configurable, but it MUST NOT be less than
   TCP_EST (the established connection idle timeout as defined in
   [RFC5382]).  The default value for the maximum session lifetime
   SHOULD be set to TCP_EST.  The state remains unchanged as
   ESTABLISHED.

   If the lifetime expires, then the NAT64 SHOULD send a probe packet
   (as defined next) to at least one of the endpoints of the TCP
   connection.  The probe packet is a TCP segment for the connection
   with no data.  The sequence number and the acknowledgment number are
   set to zero.  All flags but the ACK flag are set to zero.  The state
   is moved to TRANS.

      Upon the reception of this probe packet, the endpoint will reply
      with an ACK containing the expected sequence number for that
      connection.  It should be noted that, for an active connection,
      each of these probe packets will generate one packet from each end
      involved in the connection, since the reply of the first point to
      the probe packet will generate a reply from the other endpoint.

   *** V4 FIN RCV ***

   If a V6 FIN packet is received, the packet is translated and
   forwarded.  The lifetime is set to TCP_TRANS.  The state is moved to
   V6 FIN + V4 FIN RCV.

   If any packet other than the V6 FIN is received, the packet is
   translated and forwarded.  The lifetime of the TCP Session Table
   Entry is set to at least the maximum session lifetime.  The value for
   the maximum session lifetime MAY be configurable, but it MUST NOT be




Bagnulo, et al.              Standards Track                   [Page 31]

RFC 6146                     Stateful NAT64                   April 2011


   less than TCP_EST (the established connection idle timeout as defined
   in [RFC5382]).  The default value for the maximum session lifetime
   SHOULD be set to TCP_EST.  The state remains unchanged as V4 FIN RCV.

   If the lifetime expires, the Session Table Entry is deleted, and the
   state is moved to CLOSED.

   *** V6 FIN RCV ***

   If a V4 FIN packet is received, the packet is translated and
   forwarded.  The lifetime is set to TCP_TRANS.  The state is moved to
   V6 FIN + V4 FIN RCV.

   If any packet other than the V4 FIN is received, the packet is
   translated and forwarded.  The lifetime of the TCP Session Table
   Entry is set to at least the maximum session lifetime.  The value for
   the maximum session lifetime MAY be configurable, but it MUST NOT be
   less than TCP_EST (the established connection idle timeout as defined
   in [RFC5382]).  The default value for the maximum session lifetime
   SHOULD be set to TCP_EST.  The state remains unchanged as V6 FIN RCV.

   If the lifetime expires, the Session Table Entry is deleted and the
   state is moved to CLOSED.

   *** V6 FIN + V4 FIN RCV ***

   All packets are translated and forwarded.

   If the lifetime expires, the Session Table Entry is deleted and the
   state is moved to CLOSED.

   *** TRANS ***

   If a packet other than a RST packet is received, the lifetime of the
   TCP Session Table Entry is set to at least the maximum session
   lifetime.  The value for the maximum session lifetime MAY be
   configurable, but it MUST NOT be less than TCP_EST (the established
   connection idle timeout as defined in [RFC5382]).  The default value
   for the maximum session lifetime SHOULD be set to TCP_EST.  The state
   is moved to ESTABLISHED.

   If the lifetime expires, the Session Table Entry is deleted and the
   state is moved to CLOSED.








Bagnulo, et al.              Standards Track                   [Page 32]

RFC 6146                     Stateful NAT64                   April 2011


3.5.2.3.  Rules for Allocation of IPv4 Transport Addresses for TCP



   When a new TCP BIB entry is created for a source transport address of
   (S',s), the NAT64 allocates an IPv4 transport address for this BIB
   entry as follows:

      If there exists some other BIB entry in any of the BIBs that
      contains S' as the IPv6 address and maps it to some IPv4 address
      T, then T SHOULD be used as the IPv4 address.  Otherwise, use any
      IPv4 address of the IPv4 pool assigned to the NAT64 to be used for
      translation.

      If the port s is in the Well-Known port range 0-1023, and the
      NAT64 has an available port t in the same port range, then the
      NAT64 SHOULD allocate the port t.  If the NAT64 does not have a
      port available in the same range, the NAT64 MAY assign a port t
      from another range where it has an available port.

      If the port s is in the range 1024-65535, and the NAT64 has an
      available port t in the same port range, then the NAT64 SHOULD
      allocate the port t.  If the NAT64 does not have a port available
      in the same range, the NAT64 MAY assign a port t from another
      range where it has an available port.

      In all cases, the allocated IPv4 transport address (T,t) MUST NOT
      be in use in another entry in the same BIB, but can be in use in
      other BIBs (e.g., the UDP and TCP BIBs).

   If it is not possible to allocate an appropriate IPv4 transport
   address or create a BIB entry, then the packet is discarded.  The
   NAT64 SHOULD send an ICMPv6 Destination Unreachable error message
   with Code 3 (Address Unreachable).

3.5.3.  ICMP Query Session Handling



   The following state information is stored for an ICMP Query session
   in the ICMP Query session table:

      Binding:(X',Y',i1) <--> (T,Z,i2)

      Lifetime: a timer that tracks the remaining lifetime of the ICMP
      Query session.  When the timer expires, the session is deleted.
      If all the ICMP Query sessions corresponding to a dynamically
      created ICMP Query BIB entry are deleted, then the ICMP Query BIB
      entry is also deleted.






Bagnulo, et al.              Standards Track                   [Page 33]

RFC 6146                     Stateful NAT64                   April 2011


   An incoming ICMPv6 Informational packet with IPv6 source address X',
   IPv6 destination address Y', and ICMPv6 Identifier i1 is processed as
   follows:

      If the local security policy determines that ICMPv6 Informational
      packets are to be filtered, the packet is silently discarded.
      Else, the NAT64 searches for an ICMP Query BIB entry that matches
      the (X',i1) pair.  If such an entry does not exist, the NAT64
      tries to create a new entry (if resources and policy permit) with
      the following data:

      *  The BIB IPv6 address is set to X' (i.e., the source IPv6
         address of the IPv6 packet).

      *  The BIB ICMPv6 Identifier is set to i1 (i.e., the ICMPv6
         Identifier).

      *  If there exists another BIB entry in any of the BIBs that
         contains the same IPv6 address X' and maps it to an IPv4
         address T, then use T as the BIB IPv4 address for this new
         entry.  Otherwise, use any IPv4 address assigned to the IPv4
         interface.

      *  Any available value is used as the BIB ICMPv4 Identifier, i.e.,
         any identifier value for which no other entry exists with the
         same (IPv4 address, ICMPv4 Identifier) pair.

      The NAT64 searches for an ICMP Query Session Table Entry
      corresponding to the incoming 3-tuple (X',Y',i1).  If no such
      entry is found, the NAT64 tries to create a new entry (if
      resources and policy permit).  The information included in the new
      Session Table Entry is as follows:

      *  The STE IPv6 source address is set to X' (i.e., the address
         contained in the received IPv6 packet).

      *  The STE IPv6 destination address is set to Y' (i.e., the
         address contained in the received IPv6 packet).

      *  The STE ICMPv6 Identifier is set to i1 (i.e., the identifier
         contained in the received IPv6 packet).

      *  The STE IPv4 source address is set to the IPv4 address
         contained in the corresponding BIB entry.

      *  The STE ICMPv4 Identifier is set to the IPv4 identifier
         contained in the corresponding BIB entry.




Bagnulo, et al.              Standards Track                   [Page 34]

RFC 6146                     Stateful NAT64                   April 2011


      *  The STE IPv4 destination address is algorithmically generated
         from Y' using the reverse algorithm as specified in
         Section 3.5.4.

      The NAT64 sets (or resets) the timer in the session table entry to
      the maximum session lifetime.  By default, the maximum session
      lifetime is ICMP_DEFAULT.  The maximum lifetime value SHOULD be
      configurable.  The packet is translated and forwarded as described
      in the following sections.

   An incoming ICMPv4 Query packet with source IPv4 address Y,
   destination IPv4 address X, and ICMPv4 Identifier i2 is processed as
   follows:

      The NAT64 searches for an ICMP Query BIB entry that contains X as
      the IPv4 address and i2 as the ICMPv4 Identifier.  If such an
      entry does not exist, the packet is dropped.  An ICMP error
      message MAY be sent to the original sender of the packet.  The
      ICMP error message, if sent, has Type 3, Code 1 (Host
      Unreachable).

      If the NAT64 filters on its IPv4 interface, then the NAT64 checks
      to see if the incoming packet is allowed according to the Address-
      Dependent Filtering rule.  To do this, it searches for a Session
      Table Entry with an STE source IPv4 address equal to X, an STE
      ICMPv4 Identifier equal to i2, and a STE destination IPv4 address
      equal to Y.  If such an entry is found (there may be more than
      one), packet processing continues.  Otherwise, the packet is
      discarded.  If the packet is discarded, then an ICMP error message
      MAY be sent to the original sender of the packet.  The ICMP error
      message, if sent, has Type 3 (Destination Unreachable) and Code 13
      (Communication Administratively Prohibited).

      In case the packet is not discarded in the previous processing
      steps (either because the NAT64 is not filtering or because the
      packet is compliant with the Address-Dependent Filtering rule),
      then the NAT64 searches for a Session Table Entry with an STE
      source IPv4 address equal to X, an STE ICMPv4 Identifier equal to
      i2, and a STE destination IPv4 address equal to Y.  If no such
      entry is found, the NAT64 tries to create a new entry (if
      resources and policy permit) with the following information:

      *  The STE source IPv4 address is set to X.

      *  The STE ICMPv4 Identifier is set to i2.

      *  The STE destination IPv4 address is set to Y.




Bagnulo, et al.              Standards Track                   [Page 35]

RFC 6146                     Stateful NAT64                   April 2011


      *  The STE source IPv6 address is set to the IPv6 address of the
         corresponding BIB entry.

      *  The STE ICMPv6 Identifier is set to the ICMPv6 Identifier of
         the corresponding BIB entry.

      *  The STE destination IPv6 address is set to the IPv6
         representation of the IPv4 address of Y, generated using the
         algorithm described in Section 3.5.4.

      *  The NAT64 sets (or resets) the timer in the session table entry
         to the maximum session lifetime.  By default, the maximum
         session lifetime is ICMP_DEFAULT.  The maximum lifetime value
         SHOULD be configurable.  The packet is translated and forwarded
         as described in the following sections.

3.5.4.  Generation of the IPv6 Representations of IPv4 Addresses



   NAT64 supports multiple algorithms for the generation of the IPv6
   representation of an IPv4 address and vice versa.  The constraints
   imposed on the generation algorithms are the following:

      The algorithm MUST be reversible, i.e., it MUST be possible to
      derive the original IPv4 address from the IPv6 representation.

      The input for the algorithm MUST be limited to the IPv4 address,
      the IPv6 prefix (denoted Pref64::/n) used in the IPv6
      representations, and optionally a set of stable parameters that
      are configured in the NAT64 (such as a fixed string to be used as
      a suffix).

         If we note n the length of the prefix Pref64::/n, then n MUST
         be less than or equal to 96.  If a Pref64::/n is configured
         through any means in the NAT64 (such as manually configured, or
         other automatic means not specified in this document), the
         default algorithm MUST use this prefix.  If no prefix is
         available, the algorithm SHOULD use the Well-Known Prefix
         (64:ff9b::/96) defined in [RFC6052].

   NAT64 MUST support the algorithm for generating IPv6 representations
   of IPv4 addresses defined in Section 2.3 of [RFC6052].  The
   aforementioned algorithm SHOULD be used as default algorithm.

3.6.  Computing the Outgoing Tuple



   This step computes the outgoing tuple by translating the IP addresses
   and port numbers or ICMP Identifier in the incoming tuple.




Bagnulo, et al.              Standards Track                   [Page 36]

RFC 6146                     Stateful NAT64                   April 2011


   In the text below, a reference to a BIB means the TCP BIB, the UDP
   BIB, or the ICMP Query BIB, as appropriate.

      NOTE: Not all addresses are translated using the BIB.  BIB entries
      are used to translate IPv6 source transport addresses to IPv4
      source transport addresses, and IPv4 destination transport
      addresses to IPv6 destination transport addresses.  They are NOT
      used to translate IPv6 destination transport addresses to IPv4
      destination transport addresses, nor to translate IPv4 source
      transport addresses to IPv6 source transport addresses.  The
      latter cases are handled by applying the algorithmic
      transformation described in Section 3.5.4.  This distinction is
      important; without it, hairpinning doesn't work correctly.

3.6.1.  Computing the Outgoing 5-Tuple for TCP, UDP, and for ICMP Error
        Messages Containing a TCP or UDP Packets



   The transport protocol in the outgoing 5-tuple is always the same as
   that in the incoming 5-tuple.  When translating from IPv4 ICMP to
   IPv6 ICMP, the protocol number in the last next header field in the
   protocol chain is set to 58 (IPv6-ICMP).  When translating from IPv6
   ICMP to IPv4 ICMP, the protocol number in the protocol field of the
   IP header is set to 1 (ICMP).

   When translating in the IPv6 --> IPv4 direction, let the source and
   destination transport addresses in the incoming 5-tuple be (S',s) and
   (D',d), respectively.  The outgoing source transport address is
   computed as follows: if the BIB contains an entry (S',s) <--> (T,t),
   then the outgoing source transport address is (T,t).

   The outgoing destination address is computed algorithmically from D'
   using the address transformation described in Section 3.5.4.

   When translating in the IPv4 --> IPv6 direction, let the source and
   destination transport addresses in the incoming 5-tuple be (S,s) and
   (D,d), respectively.  The outgoing source transport address is
   computed as follows:

      The outgoing source transport address is generated from S using
      the address transformation algorithm described in Section 3.5.4.

      The BIB table is searched for an entry (X',x) <--> (D,d), and if
      one is found, the outgoing destination transport address is set to
      (X',x).







Bagnulo, et al.              Standards Track                   [Page 37]

RFC 6146                     Stateful NAT64                   April 2011


3.6.2.  Computing the Outgoing 3-Tuple for ICMP Query Messages and for
        ICMP Error Messages Containing an ICMP Query



   When translating in the IPv6 --> IPv4 direction, let the source and
   destination addresses in the incoming 3-tuple be S' and D',
   respectively, and the ICMPv6 Identifier be i1.  The outgoing source
   address is computed as follows: the BIB contains an entry (S',i1)
   <--> (T,i2), then the outgoing source address is T and the ICMPv4
   Identifier is i2.

   The outgoing IPv4 destination address is computed algorithmically
   from D' using the address transformation described in Section 3.5.4.

   When translating in the IPv4 --> IPv6 direction, let the source and
   destination addresses in the incoming 3-tuple be S and D,
   respectively, and the ICMPv4 Identifier is i2.  The outgoing source
   address is generated from S using the address transformation
   algorithm described in Section 3.5.4.  The BIB is searched for an
   entry containing (X',i1) <--> (D,i2), and, if found, the outgoing
   destination address is X' and the outgoing ICMPv6 Identifier is i1.

3.7.  Translating the Packet



   This step translates the packet from IPv6 to IPv4 or vice versa.

   The translation of the packet is as specified in Sections 4 and 5 of
   the IP/ICMP Translation Algorithm [RFC6145], with the following
   modifications:

   o  When translating an IP header (Sections 4.1 and 5.1 of [RFC6145]),
      the source and destination IP address fields are set to the source
      and destination IP addresses from the outgoing tuple as determined
      in Section 3.6.

   o  When the protocol following the IP header is TCP or UDP, then the
      source and destination ports are modified to the source and
      destination ports from the outgoing 5-tuple.  In addition, the TCP
      or UDP checksum must also be updated to reflect the translated
      addresses and ports; note that the TCP and UDP checksum covers the
      pseudo-header that contains the source and destination IP
      addresses.  An algorithm for efficiently updating these checksums
      is described in [RFC3022].

   o  When the protocol following the IP header is ICMP and it is an
      ICMP Query message, the ICMP Identifier is set to the one from the
      outgoing 3-tuple as determined in Section 3.6.2.





Bagnulo, et al.              Standards Track                   [Page 38]

RFC 6146                     Stateful NAT64                   April 2011


   o  When the protocol following the IP header is ICMP and it is an
      ICMP error message, the source and destination transport addresses
      in the embedded packet are set to the destination and source
      transport addresses from the outgoing 5-tuple (note the swap of
      source and destination).

   The size of outgoing packets as well and the potential need for
   fragmentation is done according to the behavior defined in the IP/
   ICMP Translation Algorithm [RFC6145].

3.8.  Handling Hairpinning



   If the destination IP address of the translated packet is an IPv4
   address assigned to the NAT64 itself, then the packet is a hairpin
   packet.  Hairpin packets are processed as follows:

   o  The outgoing 5-tuple becomes the incoming 5-tuple.

   o  The packet is treated as if it was received on the outgoing
      interface.

   o  Processing of the packet continues at step 2 -- "Filtering and
      Updating Binding and Session Information" (Section 3.5).

4.  Protocol Constants

   UDP_MIN: 2 minutes (as defined in [RFC4787])

   UDP_DEFAULT: 5 minutes (as defined in [RFC4787])

   TCP_TRANS: 4 minutes (as defined in [RFC5382])

   TCP_EST: 2 hours (The minimum lifetime for an established TCP session
   defined in [RFC5382] is 2 hours and 4 minutes, which is achieved by
   adding the 2 hours with this timer and the 4 minutes with the
   TCP_TRANS timer.)

   TCP_INCOMING_SYN: 6 seconds (as defined in [RFC5382])

   FRAGMENT_MIN: 2 seconds

   ICMP_DEFAULT: 60 seconds (as defined in [RFC5508])









Bagnulo, et al.              Standards Track                   [Page 39]

RFC 6146                     Stateful NAT64                   April 2011


5.  Security Considerations

5.1.  Implications on End-to-End Security



   Any protocols that protect IP header information are essentially
   incompatible with NAT64.  This implies that end-to-end IPsec
   verification will fail when the Authentication Header (AH) is used
   (both transport and tunnel mode) and when ESP is used in transport
   mode.  This is inherent in any network-layer translation mechanism.
   End-to-end IPsec protection can be restored, using UDP encapsulation
   as described in [RFC3948].  The actual extensions to support IPsec
   are out of the scope of this document.

5.2.  Filtering



   NAT64 creates binding state using packets flowing from the IPv6 side
   to the IPv4 side.  In accordance with the procedures defined in this
   document following the guidelines defined in [RFC4787], a NAT64 MUST
   offer "Endpoint-Independent Mapping".  This means:

      For any IPv6 packet with source (S'1,s1) and destination
      (Pref64::D1,d1) that creates an external mapping to (S1,s1v4),
      (D1,d1), for any subsequent packet from (S'1,s1) to
      (Pref64::D2,d2) that creates an external mapping to (S2,s2v4),
      (D2,d2), within a given binding timer window,

      (S1,s1v4) = (S2,s2v4) for all values of D2,d2

   Implementations MAY also provide support for "Address-Dependent
   Mapping" as also defined in this document and following the
   guidelines defined in [RFC4787].

   The security properties, however, are determined by which packets the
   NAT64 filter allows in and which it does not.  The security
   properties are determined by the filtering behavior and filtering
   configuration in the filtering portions of the NAT64, not by the
   address mapping behavior.  For example:

      Without filtering - When "Endpoint-Independent Mapping" is used in
      NAT64, once a binding is created in the IPv6 ---> IPv4 direction,
      packets from any node on the IPv4 side destined to the IPv6
      transport address will traverse the NAT64 gateway and be forwarded
      to the IPv6 transport address that created the binding.  However,

      With filtering - When "Endpoint-Independent Mapping" is used in
      NAT64, once a binding is created in the IPv6 ---> IPv4 direction,
      packets from any node on the IPv4 side destined to the IPv6
      transport address will first be processed against the filtering



Bagnulo, et al.              Standards Track                   [Page 40]

RFC 6146                     Stateful NAT64                   April 2011


      rules.  If the source IPv4 address is permitted, the packets will
      be forwarded to the IPv6 transport address.  If the source IPv4
      address is explicitly denied -- or the default policy is to deny
      all addresses not explicitly permitted -- then the packet will be
      discarded.  A dynamic filter may be employed whereby the filter
      will only allow packets from the IPv4 address to which the
      original packet that created the binding was sent.  This means
      that only the IPv4 addresses to which the IPv6 host has initiated
      connections will be able to reach the IPv6 transport address, and
      no others.  This essentially narrows the effective operation of
      the NAT64 device to an "Address-Dependent Mapping" behavior,
      though not by its mapping behavior, but instead by its filtering
      behavior.

   As currently specified, the NAT64 only requires filtering traffic
   based on the 5-tuple.  In some cases (e.g., statically configured
   mappings), this may make it easy for an attacker to guess.  An
   attacker need not be able to guess other fields, e.g., the TCP
   sequence number, to get a packet through the NAT64.  While such
   traffic might be dropped by the final destination, it does not
   provide additional mitigations against bandwidth/CPU attacks
   targeting the internal network.  To avoid this type of abuse, a NAT64
   MAY keep track of the sequence number of TCP packets in order to
   verify the proper sequencing of exchanged segments, in particular,
   those of the SYNs and the FINs.

5.3.  Attacks on NAT64



   The NAT64 device itself is a potential victim of different types of
   attacks.  In particular, the NAT64 can be a victim of DoS attacks.
   The NAT64 device has a limited number of resources that can be
   consumed by attackers creating a DoS attack.  The NAT64 has a limited
   number of IPv4 addresses that it uses to create the bindings.  Even
   though the NAT64 performs address and port translation, it is
   possible for an attacker to consume all the IPv4 transport addresses
   by sending IPv6 packets with different source IPv6 transport
   addresses.  This attack can only be launched from the IPv6 side,
   since IPv4 packets are not used to create binding state.  DoS attacks
   can also affect other limited resources available in the NAT64 such
   as memory or link capacity.  For instance, it is possible for an
   attacker to launch a DoS attack on the memory of the NAT64 device by
   sending fragments that the NAT64 will store for a given period.  If
   the number of fragments is high enough, the memory of the NAT64 could
   be exhausted.  Similarly, a DoS attack against the NAT64 can be
   crafted by sending either V4 or V6 SYN packets that consume memory in
   the form of session and/or binding table entries.  In the case of
   IPv4 SYNs the situation is aggravated by the requirement to also
   store the data packets for a given amount of time, requiring more



Bagnulo, et al.              Standards Track                   [Page 41]

RFC 6146                     Stateful NAT64                   April 2011


   memory from the NAT64 device.  NAT64 devices MUST implement proper
   protection against such attacks, for instance, allocating a limited
   amount of memory for fragmented packet storage as specified in
   Section 3.4.

   Another consideration related to NAT64 resource depletion refers to
   the preservation of binding state.  Attackers may try to keep a
   binding state alive forever by sending periodic packets that refresh
   the state.  In order to allow the NAT64 to defend against such
   attacks, the NAT64 MAY choose not to extend the session entry
   lifetime for a specific entry upon the reception of packets for that
   entry through the external interface.  As described in the framework
   document [RFC6144], the NAT64 can be deployed in multiple scenarios,
   in some of which the Internet side is the IPv6 one, and in others of
   which the Internet side is the IPv4 one.  It is then important to
   properly set which is the Internet side of the NAT64 in each specific
   configuration.

5.4.  Avoiding Hairpinning Loops



   If an IPv6-only client can guess the IPv4 binding address that will
   be created, it can use the IPv6 representation of that address as the
   source address for creating this binding.  Then, any packet sent to
   the binding's IPv4 address could loop in the NAT64.  This is
   prevented in the current specification by filtering incoming packets
   containing Pref64::/n in the source address, as described below.

   Consider the following example:

   Suppose that the IPv4 pool is 192.0.2.0/24

   Then, the IPv6-only client sends this to NAT64:

      Source: [Pref64::192.0.2.1]:500

      Destination: any

   The NAT64 allocates 192.0.2.1:500 as the IPv4 binding address.  Now
   anything sent to 192.0.2.1:500, be it a hairpinned IPv6 packet or an
   IPv4 packet, could loop.

   It is not hard to guess the IPv4 address that will be allocated.
   First, the attacker creates a binding and uses (for example) Simple
   Traversal of the UDP Protocol through NAT (STUN) [RFC5389] to learn
   its external IPv4 address.  New bindings will always have this
   address.  Then, it uses a source port in the range 1-1023.  This will
   increase the chances to 1/512 (since range and parity are preserved
   by NAT64 in UDP).



Bagnulo, et al.              Standards Track                   [Page 42]

RFC 6146                     Stateful NAT64                   April 2011


   In order to address this vulnerability, the NAT64 MUST drop IPv6
   packets whose source address is in Pref64::/n, as defined in
   Section 3.5.

6.  Contributors



   George Tsirtsis
      Qualcomm
      tsirtsis@googlemail.com

   Greg Lebovitz
      Juniper
      gregory.ietf@gmail.com

   Simon Perreault
      Viagenie
      simon.perreault@viagenie.ca

7.  Acknowledgements



   Dave Thaler, Dan Wing, Alberto Garcia-Martinez, Reinaldo Penno,
   Ranjana Rao, Lars Eggert, Senthil Sivakumar, Zhen Cao, Xiangsong Cui,
   Mohamed Boucadair, Dong Zhang, Bryan Ford, Kentaro Ebisawa, Charles
   Perkins, Magnus Westerlund, Ed Jankiewicz, David Harrington, Peter
   McCann, Julien Laganier, Pekka Savola, and Joao Damas reviewed the
   document and provided useful comments to improve it.

   The content of the document was improved thanks to discussions with
   Christian Huitema, Fred Baker, and Jari Arkko.

   Marcelo Bagnulo and Iljitsch van Beijnum are partly funded by
   Trilogy, a research project supported by the European Commission
   under its Seventh Framework Program.

8.  References

8.1.  Normative References



   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC4443]  Conta, A., Deering, S., and M. Gupta, "Internet Control
              Message Protocol (ICMPv6) for the Internet Protocol
              Version 6 (IPv6) Specification", RFC 4443, March 2006.

   [RFC4787]  Audet, F. and C. Jennings, "Network Address Translation
              (NAT) Behavioral Requirements for Unicast UDP", BCP 127,
              RFC 4787, January 2007.



Bagnulo, et al.              Standards Track                   [Page 43]

RFC 6146                     Stateful NAT64                   April 2011


   [RFC5382]  Guha, S., Biswas, K., Ford, B., Sivakumar, S., and P.
              Srisuresh, "NAT Behavioral Requirements for TCP", BCP 142,
              RFC 5382, October 2008.

   [RFC5508]  Srisuresh, P., Ford, B., Sivakumar, S., and S. Guha, "NAT
              Behavioral Requirements for ICMP", BCP 148, RFC 5508,
              April 2009.

   [RFC6052]  Bao, C., Huitema, C., Bagnulo, M., Boucadair, M., and X.
              Li, "IPv6 Addressing of IPv4/IPv6 Translators", RFC 6052,
              October 2010.

   [RFC6145]  Li, X., Bao, C., and F. Baker, "IP/ICMP Translation
              Algorithm", RFC 6145, April 2011.

8.2.  Informative References



   [RFC0793]  Postel, J., "Transmission Control Protocol", STD 7,
              RFC 793, September 1981.

   [RFC1858]  Ziemba, G., Reed, D., and P. Traina, "Security
              Considerations for IP Fragment Filtering", RFC 1858,
              October 1995.

   [RFC3022]  Srisuresh, P. and K. Egevang, "Traditional IP Network
              Address Translator (Traditional NAT)", RFC 3022,
              January 2001.

   [RFC3128]  Miller, I., "Protection Against a Variant of the Tiny
              Fragment Attack (RFC 1858)", RFC 3128, June 2001.

   [RFC3948]  Huttunen, A., Swander, B., Volpe, V., DiBurro, L., and M.
              Stenberg, "UDP Encapsulation of IPsec ESP Packets",
              RFC 3948, January 2005.

   [RFC4963]  Heffner, J., Mathis, M., and B. Chandler, "IPv4 Reassembly
              Errors at High Data Rates", RFC 4963, July 2007.

   [RFC5245]  Rosenberg, J., "Interactive Connectivity Establishment
              (ICE): A Protocol for Network Address Translator (NAT)
              Traversal for Offer/Answer Protocols", RFC 5245,
              April 2010.

   [RFC5389]  Rosenberg, J., Mahy, R., Matthews, P., and D. Wing,
              "Session Traversal Utilities for NAT (STUN)", RFC 5389,
              October 2008.





Bagnulo, et al.              Standards Track                   [Page 44]

RFC 6146                     Stateful NAT64                   April 2011


   [RFC6144]  Baker, F., Li, X., Bao, C., and K. Yin, "Framework for
              IPv4/IPv6 Translation", RFC 6144, April 2011.

   [RFC6147]  Bagnulo, M., Sullivan, A., Matthews, P., and I. van
              Beijnum, "DNS64: DNS extensions for Network Address
              Translation from IPv6 Clients to IPv4 Servers", RFC 6147,
              April 2011.

Authors' Addresses



   Marcelo Bagnulo
   UC3M
   Av. Universidad 30
   Leganes, Madrid  28911
   Spain

   Phone: +34-91-6249500
   EMail: marcelo@it.uc3m.es
   URI:   http://www.it.uc3m.es/marcelo


   Philip Matthews
   Alcatel-Lucent
   600 March Road
   Ottawa, Ontario
   Canada

   Phone: +1 613-592-4343 x224
   EMail: philip_matthews@magma.ca


   Iljitsch van Beijnum
   IMDEA Networks
   Avda. del Mar Mediterraneo, 22
   Leganes, Madrid  28918
   Spain

   EMail: iljitsch@muada.com













Bagnulo, et al.              Standards Track                   [Page 45]