path: root/doc/rfc/rfc4338.txt

Network Working Group                                         C. DeSanti
Request for Comments: 4338                                 Cisco Systems
Obsoletes: 3831, 2625                                         C. Carlson
Category: Standards Track                             QLogic Corporation
                                                                R. Nixon
                                                                  Emulex
                                                            January 2006


                    Transmission of IPv6, IPv4, and
      Address Resolution Protocol (ARP) Packets over Fibre Channel

Status of This Memo

   This document specifies an Internet standards track protocol for the
   Internet community, and requests discussion and suggestions for
   improvements.  Please refer to the current edition of the "Internet
   Official Protocol Standards" (STD 1) for the standardization state
   and status of this protocol.  Distribution of this memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2006).

Abstract

   This document specifies the way of encapsulating IPv6, IPv4, and
   Address Resolution Protocol (ARP) packets over Fibre Channel.  This
   document also specifies the method of forming IPv6 link-local
   addresses and statelessly autoconfigured IPv6 addresses on Fibre
   Channel networks, and a mechanism to perform IPv4 address resolution
   over Fibre Channel networks.

   This document obsoletes RFC 2625 and RFC 3831.

















DeSanti, et al.             Standards Track                     [Page 1]
^L
RFC 4338                 IP over Fibre Channel              January 2006


Table of Contents

   1. Introduction ....................................................3
   2. Summary of Fibre Channel ........................................4
      2.1. Overview ...................................................4
      2.2. Identifiers and Login ......................................5
      2.3. FC Levels and Frame Format .................................5
      2.4. Sequences and Exchanges ....................................6
   3. IP-capable Nx_Ports .............................................7
   4. IPv6, IPv4, and ARP Encapsulation ...............................7
      4.1. FC Sequence Format for IPv6 and IPv4 Packets ...............7
      4.2. FC Sequence Format for ARP Packets .........................9
      4.3. FC Classes of Service .....................................10
      4.4. FC Header Code Points .....................................10
      4.5. FC Network_Header .........................................11
      4.6. LLC/SNAP Header ...........................................12
      4.7. Bit and Byte Ordering .....................................12
      4.8. Maximum Transfer Unit .....................................12
   5. IPv6 Stateless Address Autoconfiguration .......................13
      5.1. IPv6 Interface Identifier and Address Prefix ..............13
      5.2. Generating an Interface ID from a Format 1 N_Port_Name ....14
      5.3. Generating an Interface ID from a Format 2 N_Port_Name ....15
      5.4. Generating an Interface ID from a Format 5 N_Port_Name ....16
      5.5. Generating an Interface ID from an EUI-64 Mapped
           N_Port_Name ...............................................17
   6. Link-local Addresses ...........................................18
   7. ARP Packet Format ..............................................18
   8. Link-layer Address/Hardware Address ............................20
   9. Address Mapping for Unicast ....................................20
      9.1. Overview ..................................................20
      9.2. IPv6 Address Mapping ......................................20
      9.3. IPv4 Address Mapping ......................................21
   10. Address Mapping for Multicast .................................22
   11. Sequence Management ...........................................23
   12. Exchange Management ...........................................23
   13. Interoperability with RFC 2625 ................................24
   14. Security Considerations .......................................25
   15. IANA Considerations ...........................................25
   16. Acknowledgements ..............................................25
   17. Normative References ..........................................26
   18. Informative References ........................................26
   A. Transmission of a Broadcast FC Sequence over FC Topologies
      (Informative) ..................................................28
   B. Validation of the <N_Port_Name, N_Port_ID> Mapping
      (Informative) ..................................................29
   C. Fibre Channel Bit and Byte Numbering Guidance ..................30
   D. Changes from RFC 2625 ..........................................31
   E. Changes from RFC 3831 ..........................................31



DeSanti, et al.             Standards Track                     [Page 2]
^L
RFC 4338                 IP over Fibre Channel              January 2006


1.  Introduction

   Fibre Channel (FC) is a high-speed serial interface technology that
   supports several Upper Layer Protocols including Small Computer
   System Interface (SCSI), IPv6 [IPv6], and IPv4 [IPv4].

   [RFC-2625] defined how to encapsulate IPv4 and Address Resolution
   Protocol (ARP) packets over Fibre Channel for a subset of Fibre
   Channel devices.  This specification enables the support of IPv4 for
   a broader category of Fibre Channel devices.  In addition, this
   specification simplifies [RFC-2625] by removing unused options and
   clarifying current implementations.  This document obsoletes
   [RFC-2625].

   Specific [RFC-2625] limitations that this document aims to resolve
   are the following:

   -  N_Port_Name format restriction.  [RFC-2625] restricts the use of
      IPv4 to Fibre Channel devices having the format 0x1 N_Port_Name,
      but many current implementations use other N_Port_Name formats.

   -  Use of Fibre Channel Address Resolution Protocol (FARP).
      [RFC-2625] requires the support of FARP to map N_Port_Names to
      N_Port_IDs, but many current implementations use other methods,
      such as the Fibre Channel Name Server.

   -  Missing support for IPv4 multicast.  [RFC-2625] does not specify
      how to transmit IPv4 packets with a multicast destination address
      over Fibre Channel.

   [RFC-3831] defines how to encapsulate IPv6 over Fibre Channel and a
   method of forming IPv6 link-local addresses [AARCH] and statelessly
   autoconfigured IPv6 addresses on Fibre Channel networks.  [RFC-3831]
   also describes the content of the Source/Target Link-layer Address
   option used in Neighbor Discovery [DISC] when the messages are
   transmitted on a Fibre Channel network.  This document obsoletes
   [RFC-3831].

   Warning to readers familiar with Fibre Channel: both Fibre Channel
   and IETF standards use the same byte transmission order.  However,
   the bit numbering is different.  See Appendix C for guidance.

   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 [KEYWORDS].






DeSanti, et al.             Standards Track                     [Page 3]
^L
RFC 4338                 IP over Fibre Channel              January 2006


2.  Summary of Fibre Channel

2.1.  Overview

   Fibre Channel (FC) is a gigabit-speed network technology primarily
   used for storage networking.  Fibre Channel is standardized in the
   T11 Technical Committee of the InterNational Committee for
   Information Technology Standards (INCITS), an American National
   Standard Institute (ANSI) accredited standards committee.

   Fibre Channel devices are called Nodes.  Each Node has one or more
   Ports that connect to Ports of other devices.  Fibre Channel may be
   implemented using any combination of the following three topologies:

   -  a point-to-point link between two Ports;
   -  a set of Ports interconnected by a switching network called a
      Fabric, as defined in [FC-FS];
   -  a set of Ports interconnected with a loop topology, as defined in
      [FC-AL-2].

   A Node Port that does not operate in a loop topology is called an
   N_Port.  A Node Port that operates in a loop topology using the
   loop-specific protocols is designated as an NL_Port.  The term
   Nx_Port is used to indicate a Node Port that is capable of operating
   in either mode.

   A Fabric Port that does not operate in a loop topology is called an
   F_Port.  A Fabric Port that operates in a loop topology using the
   loop-specific protocols is designated as an FL_Port.  The term
   Fx_Port is used to indicate a Fabric Port that is capable of
   operating in either mode.

   A Fibre Channel network, built with any combination of the FC
   topologies described above, is a multiaccess network with broadcast
   capabilities.

   From an IPv6 point of view, a Fibre Channel network is an IPv6 Link
   [IPv6].  IP-capable Nx_Ports are what [IPv6] calls Interfaces.

   From an IPv4 point of view, a Fibre Channel network is an IPv4 Local
   Network [IPv4].  IP-capable Nx_Ports are what [IPv4] calls Local
   Network Interfaces.









DeSanti, et al.             Standards Track                     [Page 4]
^L
RFC 4338                 IP over Fibre Channel              January 2006


2.2.  Identifiers and Login

   Fibre Channel entities are identified by non-volatile 64-bit
   Name_Identifiers.  [FC-FS] defines several formats of
   Name_Identifiers.  The value of the most significant 4 bits defines
   the format of a Name_Identifier.  These Name_Identifiers are referred
   to in a more concise manner as follows:

   -  an Nx_Port's Name_Identifier is called N_Port_Name;
   -  an Fx_Port's Name_Identifier is called F_Port_Name;
   -  a Node's Name_Identifier is called Node_Name;
   -  a Fabric's Name_Identifier is called Fabric_Name.

   An Nx_Port connected to a Fibre Channel network is associated with
   two identifiers, its non-volatile N_Port_Name and a volatile 24-bit
   address called N_Port_ID.  The N_Port_Name is used to identify the
   Nx_Port, and the N_Port_ID is used for communications among Nx_Ports.

   Each Nx_Port acquires an N_Port_ID from the Fabric by performing a
   process called Fabric Login, or FLOGI.  The FLOGI process is used
   also to negotiate several communications parameters between the
   Nx_Port and the Fabric, such as the receive data field size, which
   determines the maximum size of the Fibre Channel frames that may be
   transferred between the Nx_Port and the Fabric.

   Before effective communication may take place between two Nx_Ports,
   they must complete a process called Port Login, or PLOGI.  The PLOGI
   process provides each Nx_Port with the other Nx_Port's N_Port_Name,
   and negotiates several communication parameters, such as the receive
   data field size, which determines the maximum size of the Fibre
   Channel frames that may be transferred between the two Nx_Ports.

   Both Fabric Login and Port Login may be explicit (i.e., performed
   using specific FC control messages called Extended Link Services, or
   ELSes) or implicit (i.e., in which the parameters are specified by
   configuration or other methods).

2.3.  FC Levels and Frame Format

   [FC-FS] describes the Fibre Channel protocol using 5 different
   levels.  The FC-2 and FC-4 levels are relevant for this
   specification.  The FC-2 level defines the FC frame format, the
   transport services, and the control functions necessary for
   information transfer.  The FC-4 level supports Upper Level Protocols,
   such as IPv6, IPv4, and SCSI.  The Fibre Channel frame format is
   shown in figure 1.





DeSanti, et al.             Standards Track                     [Page 5]
^L
RFC 4338                 IP over Fibre Channel              January 2006


      +-----+-----------+-----------+--------//-------+-----+-----+
      |     |           |         Data Field          |     |     |
      | SOF | FC Header |<--------------------------->| CRC | EOF |
      |     |           | Optional  | Frame           |     |     |
      |     |           | Header(s) | Payload         |     |     |
      +-----+-----------+-----------+--------//-------+-----+-----+

                      Figure 1: Fibre Channel Frame Format

   The Start of Frame (SOF) and End of Frame (EOF) are special FC
   transmission words that act as frame delimiters.  The Cyclic
   Redundancy Check (CRC) is 4 octets long and is used to verify the
   integrity of a frame.

   The FC Header is 24 octets long and contains several fields
   associated with the identification and control of the Data Field.

   The Data Field is of variable size, ranging from 0 to 2112 octets,
   and includes the user data in the Frame Payload field and Optional
   Headers.  The currently defined Optional Headers are the following:

   -  ESP_Header;
   -  Network_Header;
   -  Association_Header;
   -  Device_Header.

   The value of the SOF field determines the FC Class of service
   associated with the frame.  Five Classes of service are specified in
   [FC-FS].  They are distinguished primarily by the method of flow
   control between the communicating Nx_Ports and by the level of data
   integrity provided.  A given Fabric or Nx_Port may support one or
   more of the following Classes of service:

   -  Class 1: Dedicated physical connection with delivery confirmation;
   -  Class 2: Frame multiplexed service with delivery confirmation;
   -  Class 3: Datagram service;
   -  Class 4: Fractional bandwidth;
   -  Class 6: Reliable multicast via dedicated connections.

   Classes 3 and 2 are commonly used for storage networking
   applications; Classes 1 and 6 are typically used for specialized
   applications in avionics.  Class 3 is recommended for IPv6, IPv4, and
   ARP (see section 4.3).

2.4.  Sequences and Exchanges

   An application-level payload such as an IPv6 or IPv4 packet is called
   an Information Unit at the FC-4 level of Fibre Channel.  Each FC-4



DeSanti, et al.             Standards Track                     [Page 6]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   Information Unit is mapped to an FC Sequence by the FC-2 level.  An
   FC Sequence consists of one or more FC frames related by the value of
   the Sequence_ID (SEQ_ID) field of the FC Header.

   The architectural maximum data that may be carried by an FC frame is
   2112 octets.  The maximum usable frame size depends on the Fabric and
   Nx_Port implementations and is negotiated during the Login process.
   Whenever an Information Unit to be transmitted exceeds this value,
   the FC-2 level segments it into multiple FC frames, sent as a single
   Sequence.  The receiving Nx_Port reassembles the Sequence of frames
   and delivers a reassembled Information Unit to the FC-4 level.  The
   Sequence Count (SEQ_CNT) field of the FC Header may be used to ensure
   frame ordering.

   Multiple Sequences may be grouped together as belonging to the same
   FC Exchange.  The Exchange is a mechanism used by two Nx_Ports to
   identify and manage an operation between them.  The Exchange is
   opened when the operation is started between the two Nx_Ports, and
   closed when the operation ends.  FC frames belonging to the same
   Exchange are related by the value of the Exchange_ID fields in the FC
   Header.  An Originator Exchange_ID (OX_ID) and a Responder
   Exchange_ID (RX_ID) uniquely identify the Exchange between a pair of
   Nx_Ports.

3.  IP-capable Nx_Ports

   This specification requires an IP-capable Nx_Port to have the
   following properties:

   -  The format of its N_Port_Name MUST be one of 0x1, 0x2, 0x5, 0xC,
      0xD, 0xE, 0xF (see section 5.1);
   -  It MUST support Class 3;
   -  It MUST support continuously increasing SEQ_CNT [FC-FS];
   -  It MUST be able to transmit and receive an FC-4 Information Unit
      at least 1304 octets long (see section 4.1);
   -  It SHOULD support a receive data field size for Device_Data FC
      frames of at least 1024 octets (see section 10).

4.  IPv6, IPv4, and ARP Encapsulation

4.1.  FC Sequence Format for IPv6 and IPv4 Packets

   An IPv6 or IPv4 packet is mapped to an Information Unit at the FC-4
   level of Fibre Channel, which in turn is mapped to an FC Sequence by
   the FC-2 level [FC-FS].  An FC Information Unit containing an IP
   packet MUST carry the FC Network_Header [FC-FS] and the Logical Link
   Control/SubNetwork Access Protocol (LLC/SNAP) header [IEEE-LLC],
   resulting in the FC Information Unit format shown in figure 2.



DeSanti, et al.             Standards Track                     [Page 7]
^L
RFC 4338                 IP over Fibre Channel              January 2006


      +---------------+---------------+---------------+---------------+
      |                                                               |
      +-                                                             -+
      |                        Network_Header                         |
      +-                         (16 octets)                         -+
      |                                                               |
      +-                                                             -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |                        LLC/SNAP header                        |
      +-                          (8 octets)                         -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |                                                               |
      +-                                                             -+
      /                      IPv6 or IPv4 Packet                      /
      /                                                               /
      +-                                                             -+
      |                                                               |
      +---------------+---------------+---------------+---------------+

               Figure 2: FC Information Unit Mapping an IP Packet

   In order to support the minimum IPv6 MTU (i.e., 1280 octets), an
   Nx_Port supporting IP MUST be able to transmit and receive an FC-4
   Information Unit at least 1304 octets long (i.e., 1280 + 8 + 16).

   The FC ESP_Header [FC-FS] MAY be used to secure the FC frames
   composing an IP FC Sequence.  Other FC Optional Headers MUST NOT be
   used in an IP FC Sequence.

   An IP FC Sequence often consists of more than one frame, all frames
   having the same TYPE (see section 4.4).  The first frame of the
   Sequence MUST include the FC Network_Header and the LLC/SNAP header.
   The other frames MUST NOT include them, as shown in figure 3.

                       First Frame of an IP FC Sequence
   +-----------+-------------------+-----------------+-------//--------+
   | FC Header | FC Network_Header | LLC/SNAP header | First chunk of  |
   |           |                   |                 | the IP Packet   |
   +-----------+-------------------+-----------------+-------//--------+

         Subsequent Frames of an IP FC Sequence
   +-----------+-----------------//--------------------+
   | FC Header |   Additional chunk of the IP Packet   |
   +-----------+----------------//---------------------+

               Figure 3: Optional Headers in an IP FC Sequence



DeSanti, et al.             Standards Track                     [Page 8]
^L
RFC 4338                 IP over Fibre Channel              January 2006


4.2.  FC Sequence Format for ARP Packets

   An ARP packet is mapped to an Information Unit at the FC-4 level of
   Fibre Channel, which in turn is mapped to an FC Sequence by the FC-2
   level.  An FC Information Unit containing an ARP packet MUST carry
   the FC Network_Header [FC-FS] and the LLC/SNAP header [IEEE-LLC],
   resulting in the FC Information Unit format shown in figure 4.

      +---------------+---------------+---------------+---------------+
      |                                                               |
      +-                                                             -+
      |                        Network_Header                         |
      +-                         (16 octets)                         -+
      |                                                               |
      +-                                                             -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |                        LLC/SNAP header                        |
      +-                          (8 octets)                         -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |                                                               |
      +-                                                             -+
      /                           ARP Packet                          /
      /                                                               /
      +-                                                             -+
      |                                                               |
      +---------------+---------------+---------------+---------------+

               Figure 4: FC Information Unit Mapping an ARP Packet

   Given the limited size of an ARP packet (see section 7), an FC
   Sequence carrying an ARP packet MUST be mapped to a single FC frame
   that MUST include the FC Network_Header and the LLC/SNAP header.

   The FC ESP_Header [FC-FS] MAY be used to secure an FC frame carrying
   an ARP packet.  Other FC Optional Headers MUST NOT be used in an FC
   frame carrying an ARP packet.













DeSanti, et al.             Standards Track                     [Page 9]
^L
RFC 4338                 IP over Fibre Channel              January 2006


4.3.  FC Classes of Service

   This specification uses FC Class 3.  The following types of packets
   MUST be mapped in Class 3 FC frames:

   -  multicast IPv6 packets;
   -  multicast/broadcast IPv4 packets;
   -  Control Protocol packets (e.g., ARP packets; IPv6 packets carrying
      ICMPv6 [ICMPv6], Neighbor Discovery [DISC], or Multicast Listener
      Discovery [MLDv2] messages; IPv4 packets carrying ICMP [ICMPv4] or
      IGMP [IGMPv3] messages; IPv6 and IPv4 Routing Protocols packets).

   Other IPv6 and IPv4 packets (i.e., unicast IP packets carrying data
   traffic) SHOULD be mapped in Class 3 FC frames as well.  Support for
   reception of IPv4 or IPv6 packets mapped in FC frames of any Class
   other than Class 3 is OPTIONAL; receivers MAY ignore them.

4.4.  FC Header Code Points

   The fields of the Fibre Channel Header are shown in figure 5.  The
   D_ID and S_ID fields contain, respectively, the destination N_Port_ID
   and the source N_Port_ID.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |     R_CTL     |                      D_ID                     |
      +---------------+---------------+---------------+---------------+
      |  CS_CTL/Prio  |                      S_ID                     |
      +---------------+---------------+---------------+---------------+
      |     TYPE      |                     F_CTL                     |
      +---------------+---------------+---------------+---------------+
      |    SEQ_ID     |    DF_CTL     |            SEQ_CNT            |
      +---------------+---------------+---------------+---------------+
      |             OX_ID             |             RX_ID             |
      +---------------+---------------+---------------+---------------+
      |                           Parameter                           |
      +---------------+---------------+---------------+---------------+

                           Figure 5: FC Header Format

   To encapsulate IPv6 and IPv4 over Fibre Channel, the following code
   points apply.  When a single value is listed without further
   qualification, that value MUST be used:

   -  R_CTL: 0x04 (Device_Data frame with Unsolicited Data Information
      Category [FC-FS]);
   -  TYPE: 0x05 (IP over Fibre Channel);



DeSanti, et al.             Standards Track                    [Page 10]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   -  CS_CTL/Prio: 0x00 is the default, see [FC-FS] for other values;
   -  DF_CTL: 0x20 (Network_Header) for the first FC frame of an IPv6 or
      IPv4 Sequence, 0x00 for the following FC frames.  If the FC
      ESP_Header is used, then 0x60 for the first FC frame of an IPv6 or
      IPv4 Sequence, 0x40 for the following FC frames;
   -  F_CTL, SEQ_ID, SEQ_CNT, OX_ID, RX_ID: see section 11, section 12,
      and [FC-FS] for additional requirements;
   -  Parameter: if Relative Offset [FC-FS] is not used, the content of
      this field MUST be ignored by the receiver, and SHOULD be set to
      zero by the sender.  If Relative Offset is used, see [FC-FS].

   To encapsulate ARP over Fibre Channel, the following code points
   apply.  When a single value is listed without further qualification,
   that value MUST be used:

   -  R_CTL: 0x04 (Device_Data frame with Unsolicited Data Information
      Category [FC-FS]);
   -  TYPE: 0x05 (IP over Fibre Channel);
   -  CS_CTL/Prio: 0x00 is the default, see [FC-FS] for other values;
   -  DF_CTL: 0x20 (Network_Header).  If the FC ESP_Header is used, then
      0x60;
   -  F_CTL, SEQ_ID, SEQ_CNT, OX_ID, RX_ID: see section 11, section 12,
      and [FC-FS] for additional requirements;
   -  Parameter: SHOULD be set to zero.

4.5.  FC Network_Header

   The fields of the FC Network_Header are shown in figure 6.  For use
   with IPv6, IPv4, and ARP, the N_Port_Names formats MUST be one of
   0x1, 0x2, 0x5, 0xC, 0xD, 0xE, 0xF [FC-FS].

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      +-                   Destination N_Port_Name                   -+
      |                                                               |
      +---------------------------------------------------------------+
      |                                                               |
      +-                     Source N_Port_Name                      -+
      |                                                               |
      +---------------------------------------------------------------+

                       Figure 6: FC Network_Header Format







DeSanti, et al.             Standards Track                    [Page 11]
^L
RFC 4338                 IP over Fibre Channel              January 2006


4.6.  LLC/SNAP Header

   The fields of the LLC/SNAP header [IEEE-LLC] are shown in figure 7.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |     DSAP      |     SSAP      |     CTRL      |      OUI      |
      +---------------+---------------+---------------+---------------+
      |              OUI              |              PID              |
      +---------------+---------------+---------------+---------------+

                        Figure 7: LLC/SNAP Header Format

   To encapsulate IPv6, IPv4, and ARP over Fibre Channel, the following
   code points MUST be used:

   -  DSAP: 0xAA;
   -  SSAP: 0xAA;
   -  CTRL: 0x03;
   -  OUI:  0x000000;
   -  PID:  0x86DD for IPv6, 0x0800 for IPv4, 0x0806 for ARP.

4.7.  Bit and Byte Ordering

   IPv6, IPv4, and ARP packets are mapped to the FC-4 level using the
   big-endian byte ordering that corresponds to the standard network
   byte order or canonical form.

4.8.  Maximum Transfer Unit

   The default MTU size for IPv6 packets over Fibre Channel is 65280
   octets.  Large IPv6 packets are mapped to a Sequence of FC frames
   (see section 2.4).  This size may be reduced by a Router
   Advertisement [DISC] containing an MTU option that specifies a
   smaller MTU, or by manual configuration of each Nx_Port.  However, as
   required by [IPv6], the MTU MUST NOT be lower than 1280 octets.  If a
   Router Advertisement received on an Nx_Port has an MTU option
   specifying an MTU larger than 65280, or larger than a manually
   configured value, that MTU option MAY be logged to system management
   but MUST be otherwise ignored.

   As the default MTU size far exceeds the message sizes typically used
   in the Internet, an IPv6 over FC implementation SHOULD implement Path
   MTU Discovery [PMTUD6], or at least maintain different MTU values for
   on-link and off-link destinations.





DeSanti, et al.             Standards Track                    [Page 12]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   For correct operation of IPv6 in a routed environment, it is
   critically important to configure an appropriate MTU option in Router
   Advertisements.

   For correct operation of IPv6 when mixed media (e.g., Ethernet and
   Fibre Channel) are bridged together, the smallest MTU of all the
   media must be advertised by routers in an MTU option.  If there are
   no routers present, this MTU must be manually configured in each node
   that is connected to a medium with a default MTU larger than the
   smallest MTU.

   The default MTU size for IPv4 packets over Fibre Channel is 65280
   octets.  Large IPv4 packets are mapped to a Sequence of FC frames
   (see section 2.4).  This size may be reduced by manual configuration
   of each Nx_Port or by the Path MTU Discovery technique [PMTUD4].

5.  IPv6 Stateless Address Autoconfiguration

5.1.  IPv6 Interface Identifier and Address Prefix

   The IPv6 Interface ID [AARCH] for an Nx_Port is based on the EUI-64
   address [EUI64] derived from the Nx_Port's N_Port_Name.  The IPv6
   Interface Identifier is obtained by complementing the Universal/Local
   (U/L) bit of the OUI field of the derived EUI-64 address.  The U/L
   bit has no function in Fibre Channel; however, it has to be properly
   handled when a Name_Identifier is converted to an EUI-64 address.

   [FC-FS] specifies a method to map format 0x1 (IEEE 48-bit address),
   0x2 (IEEE Extended), or 0x5 (IEEE Registered) FC Name_Identifiers in
   EUI-64 addresses.  This allows the usage of these Name_Identifiers to
   support IPv6.  [FC-FS] also defines EUI-64 mapped FC Name_Identifiers
   (formats 0xC, 0xD, 0xE, and 0xF) that are derived from an EUI-64
   address.  It is possible to reverse this address mapping to obtain
   the original EUI-64 address in order to support IPv6.

   IPv6 stateless address autoconfiguration MUST be performed as
   specified in [ACONF].  An IPv6 Address Prefix used for stateless
   address autoconfiguration of an Nx_Port MUST have a length of 64
   bits.












DeSanti, et al.             Standards Track                    [Page 13]
^L
RFC 4338                 IP over Fibre Channel              January 2006


5.2.  Generating an Interface ID from a Format 1 N_Port_Name

   The Name_Identifier format 0x1 is shown in figure 8.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |0 0 0 1|         0x000         |              OUI              |
      +-------+-------+---------------+---------------+---------------+
      |      OUI      |                      VSID                     |
      +---------------+---------------+---------------+---------------+

                      Figure 8: Format 0x1 Name_Identifier

   The EUI-64 address derived from this Name_Identifier has the format
   shown in figure 9 [FC-FS].

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |         OUI with complemented U/L bit         |0 0 0 1|  VSID |
      +---------------+---------------+-------+-------+-------+-------+
      |                   VSID                |         0x000         |
      +---------------+---------------+-------+-------+---------------+

           Figure 9: EUI-64 Address from a Format 0x1 Name_Identifier

   The IPv6 Interface Identifier is obtained from this EUI-64 address by
   complementing the U/L bit in the OUI field.  Therefore, the OUI in
   the IPv6 Interface ID is exactly as in the FC Name_Identifier.  The
   resulting IPv6 Interface Identifier has local scope [AARCH] and the
   format shown in figure 10.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                      OUI                      |0 0 0 1|  VSID |
      +---------------+---------------+-------+-------+-------+-------+
      |                   VSID                |         0x000         |
      +---------------+---------------+-------+-------+---------------+

         Figure 10: IPv6 Interface ID from a Format 0x1 Name_Identifier

   As an example, the FC Name_Identifier 0x10-00-34-63-46-AB-CD-EF
   generates the IPv6 Interface Identifier 3463:461A:BCDE:F000.






DeSanti, et al.             Standards Track                    [Page 14]
^L
RFC 4338                 IP over Fibre Channel              January 2006


5.3.  Generating an Interface ID from a Format 2 N_Port_Name

   The Name_Identifier format 0x2 is shown in figure 11.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |0 0 1 0|    Vendor Specific    |              OUI              |
      +-------+-------+---------------+---------------+---------------+
      |      OUI      |                      VSID                     |
      +---------------+---------------+---------------+---------------+

                      Figure 11: Format 0x2 Name_Identifier

   The EUI-64 address derived from this Name_Identifier has the format
   shown in figure 12 [FC-FS].

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |         OUI with complemented U/L bit         |0 0 1 0|  VSID |
      +---------------+-----------------------+-------+-------+-------+
      |                   VSID                |    Vendor Specific    |
      +---------------+-----------------------+-------+---------------+

           Figure 12: EUI-64 Address from a Format 0x2 Name_Identifier

   The IPv6 Interface Identifier is obtained from this EUI-64 address by
   complementing the U/L bit in the OUI field.  Therefore, the OUI in
   the IPv6 Interface ID is exactly as in the FC Name_Identifier.  The
   resulting IPv6 Interface Identifier has local scope [AARCH] and the
   format shown in figure 13.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                      OUI                      |0 0 1 0|  VSID |
      +---------------+-----------------------+-------+-------+-------+
      |                   VSID                |    Vendor Specific    |
      +---------------+-----------------------+-------+---------------+

         Figure 13: IPv6 Interface ID from a Format 0x2 Name_Identifier

   As an example, the FC Name_Identifier 0x27-89-34-63-46-AB-CD-EF
   generates the IPv6 Interface Identifier 3463:462A:BCDE:F789.






DeSanti, et al.             Standards Track                    [Page 15]
^L
RFC 4338                 IP over Fibre Channel              January 2006


5.4.  Generating an Interface ID from a Format 5 N_Port_Name

   The Name_Identifier format 0x5 is shown in figure 14.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |0 1 0 1|                      OUI                      |  VSID |
      +-------+-------+---------------+---------------+-------+-------+
      |                             VSID                              |
      +---------------+---------------+---------------+---------------+

                      Figure 14: Format 0x5 Name_Identifier

   The EUI-64 address derived from this Name_Identifier has the format
   shown in figure 15 [FC-FS].

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |         OUI with complemented U/L bit         |0 1 0 1|  VSID |
      +---------------+---------------+---------------+-------+-------+
      |                             VSID                              |
      +---------------+---------------+---------------+---------------+

           Figure 15: EUI-64 Address from a Format 0x5 Name_Identifier

   The IPv6 Interface Identifier is obtained from this EUI-64 address
   complementing the U/L bit in the OUI field.  Therefore, the OUI in
   the IPv6 Interface ID is exactly as in the FC Name_Identifier.  The
   resulting IPv6 Interface Identifier has local scope [AARCH] and the
   format shown in figure 16.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                      OUI                      |0 1 0 1|  VSID |
      +---------------+---------------+---------------+-------+-------+
      |                             VSID                              |
      +---------------+---------------+---------------+---------------+

         Figure 16: IPv6 Interface ID from a Format 0x5 Name_Identifier

   As an example, the FC Name_Identifier 0x53-46-34-6A-BC-DE-F7-89
   generates the IPv6 Interface Identifier 3463:465A:BCDE:F789.






DeSanti, et al.             Standards Track                    [Page 16]
^L
RFC 4338                 IP over Fibre Channel              January 2006


5.5.  Generating an Interface ID from an EUI-64 Mapped N_Port_Name

   The EUI-64 mapped Name_Identifiers formats (formats 0xC through 0xF)
   are derived from an EUI-64 address by compressing the OUI field of
   such addresses.  The compression is performed by removing the
   Universal/Local and Individual/Group bits from the OUI, and by
   putting bits 0 to 5 of the OUI in the first octet of the
   Name_Identifier, and bits 8 to 23 of the OUI in the second and third
   octet of the Name_Identifier, as shown in figure 17.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |1 1| OUI[0..5] |           OUI[8..23]          |      VSID     |
      +---+-----------+---------------+---------------+---------------+
      |                             VSID                              |
      +---------------+---------------+---------------+---------------+

                Figure 17: EUI-64 Mapped Name_Identifiers Format

   The EUI-64 address used to generate the Name_Identifier shown in
   figure 17 has the format shown in figure 18.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | OUI[0..5] |0 0|           OUI[8..23]          |      VSID     |
      +-----------+---+---------------+---------------+---------------+
      |                             VSID                              |
      +---------------+---------------+---------------+---------------+

       Figure 18: EUI-64 Address from an EUI-64 Mapped Name_Identifier

   The IPv6 Interface Identifier is obtained from this EUI-64 address by
   complementing the U/L bit in the OUI field.  The resulting IPv6
   Interface Identifier has global scope [AARCH] and the format shown in
   figure 19.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      | OUI[0..5] |1 0|           OUI[8..23]          |      VSID     |
      +-----------+---+---------------+---------------+---------------+
      |                             VSID                              |
      +---------------+---------------+---------------+---------------+

      Figure 19: IPv6 Interface ID from an EUI-64 Mapped Name_Identifier




DeSanti, et al.             Standards Track                    [Page 17]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   As an example, the FC Name_Identifier 0xCD-63-46-AB-01-25-78-9A
   generates the IPv6 Interface Identifier 3663:46AB:0125:789A.

6.  Link-local Addresses

   The IPv6 link-local address [AARCH] for an Nx_Port is formed by
   appending the Interface Identifier (as defined in section 5) to the
   prefix FE80::/64.  The resulting address is shown in figure 20.

        10 bits            54 bits                  64 bits
      +----------+-----------------------+----------------------------+
      |1111111010|         (zeros)       |    Interface Identifier    |
      +----------+-----------------------+----------------------------+

                    Figure 20: IPv6 Link-local Address Format

7.  ARP Packet Format

   The Address Resolution Protocol defined in [ARP] is designed to be a
   general purpose protocol, to accommodate many network technologies
   and many Upper Layer Protocols.

   [RFC-2625] chose to use for Fibre Channel the same ARP packet format
   used for Ethernet networks.  In order to do that, [RFC-2625]
   restricted the use of IPv4 to Nx_Ports having N_Port_Name format 0x1.
   Although this may have been a reasonable choice at that time, today
   there are Nx_Ports with an N_Port_Name format other than 0x1 in
   widespread use.

   This specification accommodates Nx_Ports with N_Port_Names of a
   format different from 0x1 by defining a Fibre Channel specific
   version of the ARP protocol (FC ARP), carrying both N_Port_Name and
   N_Port_ID as Hardware (HW) Address.

   IANA has registered the number 18 (decimal) to identify Fibre Channel
   as ARP HW type.  The FC ARP packet format is shown in figure 21.  The
   length of the FC ARP packet is 40 octets.














DeSanti, et al.             Standards Track                    [Page 18]
^L
RFC 4338                 IP over Fibre Channel              January 2006


       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |        HW Type = 0x0012       |       Protocol = 0x0800       |
      +---------------+---------------+---------------+---------------+
      |  HW Len = 12  | Proto Len = 4 |            Opcode             |
      +---------------+---------------+---------------+---------------+
      |                                                               |
      +-                                                             -+
      |                      HW Address of Sender                     |
      +-                                                             -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |                   Protocol Address of Sender                  |
      +---------------+---------------+---------------+---------------+
      |                                                               |
      +-                                                             -+
      |                      HW Address of Target                     |
      +-                                                             -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |                   Protocol Address of Target                  |
      +---------------+---------------+---------------+---------------+

                         Figure 21: FC ARP Packet Format

   The following code points MUST be used with FC ARP:

   -  HW Type:   0x0012 (Fibre Channel);
   -  Protocol:  0x0800 (IPv4);
   -  HW Len:    12 (Length in octets of the HW Address);
   -  Proto Len: 4  (Length in octets of the Protocol Address);
   -  Opcode:    0x0001 for ARP Request, 0x0002 for ARP Reply [ARP];
   -  HW Address of Sender: the HW Address (see section 8) of the
      Requester in an ARP Request, or the HW Address of the Responder in
      an ARP Reply;
   -  Protocol Address of Sender: the IPv4 address of the Requester in
      an ARP Request, or that of the Responder in an ARP Reply;
   -  HW Address of Target: set to zero in an ARP Request, and to the HW
      Address (see section 8) of the Requester in an ARP Reply;
   -  Protocol Address of Target: the IPv4 address of the Responder in
      an ARP Request, or that of the Requester in an ARP Reply.









DeSanti, et al.             Standards Track                    [Page 19]
^L
RFC 4338                 IP over Fibre Channel              January 2006


8.  Link-layer Address/Hardware Address

   The Link-layer Address used in the Source/Target Link-layer Address
   option (see section 9.2) and the Hardware Address used in FC ARP (see
   section 7) have the same format, shown in figure 22.

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                                                               |
      +-                         N_Port_Name                         -+
      |                                                               |
      +---------------+---------------+---------------+---------------+
      |   Reserved    |                   N_Port_ID                   |
      +---------------+---------------+---------------+---------------+

                 Figure 22: Link-layer Address/HW Address Format

   Reserved fields MUST be set to zero when transmitting, and MUST be
   ignored when receiving.

9.  Address Mapping for Unicast

9.1.  Overview

   An Nx_Port has two kinds of Fibre Channel addresses:

   -  a non-volatile 64-bit address, called N_Port_Name;
   -  a volatile 24-bit address, called N_Port_ID.

   The N_Port_Name is used to uniquely identify the Nx_Port, and the
   N_Port_ID is used to route frames to the Nx_Port.  Both FC addresses
   are required to resolve an IPv6 or IPv4 unicast address.  The fact
   that the N_Port_ID is volatile implies that an Nx_Port MUST validate
   the mapping between its N_Port_Name and N_Port_ID when certain Fibre
   Channel events occur (see Appendix B).

9.2.  IPv6 Address Mapping

   The procedure for mapping IPv6 unicast addresses into Fibre Channel
   link-layer addresses uses the Neighbor Discovery Protocol [DISC].
   The Source/Target Link-layer Address option has the format shown in
   figure 23 when the link layer is Fibre Channel.








DeSanti, et al.             Standards Track                    [Page 20]
^L
RFC 4338                 IP over Fibre Channel              January 2006


       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |     Type      |  Length = 2   |                               |
      +---------------+---------------+                              -+
      |                                                               |
      +-                     Link-layer Address                      -+
      |                                                               |
      +-                              +---------------+---------------+
      |                               |            Padding            |
      +---------------+---------------+---------------+---------------+

    Figure 23: Source/Target Link-layer Address Option for Fibre Channel

      Type:               1 for Source Link-layer address.
                          2 for Target Link-layer address.

      Length:             2 (in units of 8 octets).

      Padding:            MUST be set to zero when transmitting,
                          MUST be ignored when receiving.

      Link-layer Address: the Nx_Port's Link-layer Address (see section
      8).

9.3.  IPv4 Address Mapping

   The procedure for mapping IPv4 unicast addresses into Fibre Channel
   link-layer addresses uses the FC ARP protocol, as specified in
   section 7 and [ARP].  A source Nx_Port that has to send IPv4 packets
   to a destination Nx_Port, known by its IPv4 address, MUST perform the
   following steps:

   1) The source Nx_Port first consults its local mapping tables for a
      mapping <destination IPv4 address, N_Port_Name, N_Port_ID>.

   2) If such a mapping is found, and a valid Port Login is in place
      with the destination Nx_Port, then the source Nx_Port sends the
      IPv4 packets to the destination Nx_Port using the retrieved
      N_Port_ID as D_ID.

   3) If such a mapping is not found, or a valid Port Login is not in
      place with the destination Nx_Port, then the source Nx_Port sends
      a broadcast FC ARP Request (see section 10) to its connected FC
      network.






DeSanti, et al.             Standards Track                    [Page 21]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   4) When a broadcast FC ARP Request is received by the Nx_Port with
      the matching IPv4 address, that Nx_Port caches the information
      carried in the FC ARP Request in its local mapping tables and
      generates a unicast FC ARP Reply.  If a valid Port Login to the
      Nx_Port that sent the broadcast FC ARP Request does not exist, the
      Nx_Port MUST perform such a Port Login, and then use it for the
      unicast reply.  The N_Port_ID to which the Port Login is directed
      is taken from the N_Port_ID field of the Sender HW Address field
      in the received FC ARP packet.

   5) If no Nx_Port has the matching IPv4 address, no unicast FC ARP
      Reply is returned.

10.  Address Mapping for Multicast

   IPv6 multicast packets, IPv4 multicast/broadcast packets, and ARP
   broadcast packets MUST be mapped to FC Sequences addressed to the
   broadcast N_Port_ID 0xFFFFFF, sent in FC Class 3 in a unidirectional
   Exchange (see section 12).  Appendix A specifies how to transmit a
   Class 3 broadcast FC Sequence over various Fibre Channel topologies.
   The Destination N_Port_Name field of the FC Network_Header MUST be
   set to the value:

   -  for broadcast ARP and IPv4 packets: 0x10-00-FF-FF-FF-FF-FF-FF;
   -  for multicast IPv6 packets: 0x10-00-33-33-XX-YY-ZZ-QQ, where
      XX-YY-ZZ-QQ are the 4 least significant octets of the multicast
      destination IPv6 address;
   -  for multicast IPv4 packets: 0x10-00-01-00-5E-XX-YY-ZZ, where the
      23 least significant bits of XX-YY-ZZ are the 23 least significant
      bits of the multicast destination IPv4 address and the most
      significant bit of XX-YY-ZZ is set to zero.

   An Nx_Port supporting IPv6 or IPv4 MUST be able to map a received
   broadcast Class 3 Device_Data FC frame to an implicit Port Login
   context in order to handle IPv6 multicast packets, IPv4 multicast or
   broadcast packets, and ARP broadcast packets.  The receive data field
   size of this implicit Port Login MUST be the same across all the
   Nx_Ports connected to the same Fabric, otherwise FC broadcast
   transmission does not work.  In order to reduce the need for FC
   Sequence segmentation, the receive data field size of this implicit
   Port Login SHOULD be 1024 octets.  This receive data field size
   requirement applies to broadcast Device_Data FC frames, not to ELSes.

   Receiving an FC Sequence carrying an IPv6 multicast packet, an IPv4
   multicast/broadcast packet, or an FC ARP broadcast packet triggers
   some additional processing by the Nx_Port when that IPv6, IPv4, or
   FC ARP packet requires a unicast reply.  In this case, if a valid
   Port Login to the Nx_Port that sent the multicast or broadcast packet



DeSanti, et al.             Standards Track                    [Page 22]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   does not exist, the Nx_Port MUST perform such a Port Login, and then
   use it for the unicast reply.  In the case of Neighbor Discovery
   messages [DISC], the N_Port_ID to which the Port Login is directed is
   taken from the N_Port_ID field of the Source Link-layer Address in
   the received Neighbor Discovery message.  In the case of FC ARP
   messages, the N_Port_ID to which the Port Login is directed is taken
   from the N_Port_ID field of the Sender HW Address field in the
   received FC ARP packet.

   As an example, if a received broadcast FC Sequence carries an IPv6
   multicast unsolicited Router Advertisement [DISC], the receiving
   Nx_Port processes it simply by passing the carried IPv6 packet to the
   IPv6 layer.  Instead, if a received broadcast FC Sequence carries an
   IPv6 multicast solicitation message [DISC] requiring a unicast reply,
   and no valid Port Login exists with the Nx_Port sender of the
   multicast packet, then a Port Login MUST be performed in order to
   send the unicast reply message.  If a received broadcast FC Sequence
   carries an IPv6 multicast solicitation message [DISC] requiring a
   multicast reply, the reply is sent to the broadcast N_Port_ID
   0xFFFFFF.

11.  Sequence Management

   FC Sequences carrying IPv6, IPv4, or ARP packets are REQUIRED to be
   non-streamed [FC-FS].  In order to avoid missing FC frame aliasing by
   Sequence_ID reuse, an Nx_Port supporting IPv6 or IPv4 is REQUIRED to
   use continuously increasing SEQ_CNT [FC-FS].  Each Exchange MUST
   start by setting SEQ_CNT to zero in the first frame; every frame
   transmitted after that MUST increment the previous SEQ_CNT by one.
   The Continue Sequence Condition field in the F_CTL field of the FC
   Header MUST be set to zero [FC-FS].

12.  Exchange Management

   To transmit IPv6, IPv4, or ARP packets to another Nx_Port or to a
   multicast/broadcast address, an Nx_Port MUST use dedicated
   unidirectional Exchanges (i.e., Exchanges dedicated to IPv6, IPv4, or
   ARP packet transmission and that do not transfer Sequence
   Initiative).  As such, the Sequence Initiative bit in the F_CTL field
   of the FC Header MUST be set to zero [FC-FS].  The RX_ID field of the
   FC Header MUST be set to 0xFFFF.

   Unicast FC Sequences carrying unicast Control Protocol packets (e.g.,
   ARP packets; IPv6 packets carrying ICMPv6 [ICMPv6], Neighbor
   Discovery [DISC], or Multicast Listener Discovery [MLDv2] messages;
   IPv4 packets carrying ICMP [ICMPv4] or IGMP [IGMPv3] messages) SHOULD
   be sent in short-lived unidirectional Exchanges (i.e., Exchanges
   containing only one Sequence, in which both the First_Sequence and



DeSanti, et al.             Standards Track                    [Page 23]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   Last_Sequence bits in the F_CTL field of the FC Header are set to one
   [FC-FS]).  Unicast FC Sequences carrying other IPv6 and IPv4 packets
   (i.e., unicast IP packets carrying data traffic) MUST be sent in a
   long-lived unidirectional Exchange (i.e., an Exchange containing one
   or more Sequences).  IP multicast packets MUST NOT be carried in
   unicast FC Sequences (see section 10).

   Broadcast FC Sequences carrying multicast or broadcast Control
   Protocol packets (e.g., ARP packets; IPv6 packets carrying ICMPv6
   [ICMPv6], Neighbor Discovery [DISC], or Multicast Listener Discovery
   [MLDv2] messages; IPv4 packets carrying ICMP [ICMPv4] or IGMP
   [IGMPv3] messages) MUST be sent in short-lived unidirectional
   Exchanges.  Broadcast FC Sequences carrying other IPv6 or IPv4
   multicast traffic (i.e., multicast IP packets carrying data traffic)
   MAY be sent in long-lived unidirectional Exchanges to enable a more
   efficient multicast distribution.

   Reasons to terminate a long-lived Exchange include the termination of
   Port Login and the completion of the IP communication.  A long-lived
   Exchange MAY be terminated by setting the Last_Sequence bit in the
   F_CTL field of the FC Header to one, or via the ABTS (Abort Sequence)
   protocol [FC-FS].  A long-lived Exchange SHOULD NOT be terminated by
   transmitting the LOGO ELS, since this may terminate active Exchanges
   on other FC-4s [FC-FS].

13.  Interoperability with RFC 2625

   The IPv4 encapsulation defined in this document, along with Exchange
   and Sequence management, are as defined in [RFC-2625].
   Implementations following this specification are expected to
   interoperate with implementations compliant to [RFC-2625] for IPv4
   packet transmission and reception.

   The main difference between this document and [RFC-2625] is in the
   address resolution procedure.  [RFC-2625] uses the Ethernet format of
   the ARP protocol and requires all Nx_Ports to have a format 0x1
   N_Port_Name.  This specification defines a Fibre Channel format for
   the ARP protocol that supports all commonly used N_Port_Names.  In
   addition, this specification does not use FARP [RFC-2625].

   An Nx_Port following this specification, and not having a format 0x1
   N_Port_Name, is able to interoperate with an [RFC-2625]
   implementation by manually configuring the mapping <destination IPv4
   address, N_Port_Name, N_Port_ID> on the involved Nx_Ports.  Through
   this manual configuration, the ARP protocol does not need to be
   performed.  However, IPv4 communication is not possible if the
   [RFC-2625] implementation strictly enforces the requirement for
   Nx_Ports to use N_Port_Names of format 0x1.



DeSanti, et al.             Standards Track                    [Page 24]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   An Nx_Port following this specification, and having a format 0x1
   N_Port_Name, is able to interoperate with an [RFC-2625]
   implementation by manually configuring the mapping <destination IPv4
   address, N_Port_Name, N_Port_ID> on the involved Nx_Ports, or by
   performing the IPv4 address resolution in compatibility mode, as
   described below:

   -  When IPv4 address resolution is attempted, the Nx_Port MUST send
      two ARP Requests, the first one according to the FC ARP format and
      the second one according to the Ethernet ARP format.  If only an
      Ethernet ARP Reply is received, it provides the N_Port_Name of the
      Nx_Port having the destination IPv4 address.  The N_Port_ID
      associated with the N_Port_Name received in an Ethernet ARP Reply
      may be retrieved from the S_ID field of the received ARP Reply, or
      by querying the Fibre Channel Name Server;
   -  The Nx_Port MUST respond to a received Ethernet ARP Request with
      an Ethernet ARP Reply;
   -  The Nx_Port MAY respond to FARP Requests [RFC-2625].

   The reception of a particular format of ARP message does not imply
   that the sending Nx_Port will continue to use the same format later.

   Support of compatibility mode is REQUIRED by each implementation.
   The use of compatibility mode MUST be administratively configurable.

14.  Security Considerations

   IPv6, IPv4, and ARP do not introduce any additional security concerns
   beyond those that already exist within the Fibre Channel protocols.
   Zoning techniques based on FC Name Server masking (soft zoning) do
   not work with IPv6 and IPv4, because IPv6 and IPv4 over Fibre Channel
   do not use the FC Name Server.  The FC ESP_Header [FC-FS] may be used
   to secure the FC frames composing FC Sequences carrying IPv6, IPv4,
   and ARP packets.  All the techniques defined to secure IP traffic at
   the IP layer may be used in a Fibre Channel environment.

15.  IANA Considerations

   The directory of ARP parameters has been updated to reference this
   document for hardware type 18.

16.  Acknowledgements

   The authors would like to acknowledge the ANSI INCITS T11.3 Task
   Group members who reviewed this document as well as the authors of
   [RFC-2625] and [RFC-3831].  The authors also thank the IMSS WG and
   Brian Haberman for their review and comments.




DeSanti, et al.             Standards Track                    [Page 25]
^L
RFC 4338                 IP over Fibre Channel              January 2006


17.  Normative References

   [FC-FS]     ANSI INCITS 373-2003, "Fibre Channel - Framing and
               Signaling (FC-FS)".

   [FC-AL-2]   ANSI INCITS 332-1999, "Fibre Channel - Arbitrated Loop-2
               (FC-AL-2)".

   [IPv6]      Deering, S. and R. Hinden, "Internet Protocol, Version 6
               (IPv6) Specification", RFC 2460, December 1998.

   [AARCH]     Hinden, R. and S. Deering, "Internet Protocol Version 6
               (IPv6) Addressing Architecture", RFC 3513, April 2003.

   [ACONF]     Thomson, S. and T. Narten, "IPv6 Stateless Address
               Autoconfiguration", RFC 2462, December 1998.

   [DISC]      Narten, T., Nordmark, E., and W. Simpson, "Neighbor
               Discovery for IP Version 6 (IPv6)", RFC 2461, December
               1998.

   [PMTUD6]    McCann, J., Deering, S., and J. Mogul, "Path MTU
               Discovery for IP version 6", RFC 1981, August 1996.

   [IPv4]      Postel, J., "Internet Protocol", STD 5, RFC 791,
               September 1981.

   [ARP]       Plummer, D., "Ethernet Address Resolution Protocol: Or
               converting network protocol addresses to 48.bit Ethernet
               address for transmission on Ethernet hardware", STD 37,
               RFC 826, November 1982.

   [IEEE-LLC]  IEEE Std 802-2001, "IEEE Standard for Local and
               Metropolitan Area Networks: Overview and Architecture".

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

18.  Informative References

   [RFC-3831]  DeSanti, C., "Transmission of IPv6 Packets over Fibre
               Channel", RFC 3831, July 2004.

   [RFC-2625]  Rajagopal, M., Bhagwat, R., and W. Rickard, "IP and ARP
               over Fibre Channel", RFC 2625, June 1999.

   [MLDv2]     Vida, R. and L. Costa, "Multicast Listener Discovery
               Version 2 (MLDv2) for IPv6", RFC 3810, June 2004.



DeSanti, et al.             Standards Track                    [Page 26]
^L
RFC 4338                 IP over Fibre Channel              January 2006


   [IGMPv3]    Cain, B., Deering, S., Kouvelas, I., Fenner, B., and A.
               Thyagarajan, "Internet Group Management Protocol, Version
               3", RFC 3376, October 2002.

   [PMTUD4]    Mogul, J. and S. Deering, "Path MTU discovery", RFC 1191,
               November 1990.

   [ICMPv6]    Conta, A. and S. Deering, "Internet Control Message
               Protocol (ICMPv6) for the Internet Protocol Version 6
               (IPv6) Specification", RFC 2463, December 1998.

   [ICMPv4]    Postel, J., "Internet Control Message Protocol", STD 5,
               RFC 792, September 1981.

   [EUI64]     "Guidelines For 64-bit Global Identifier (EUI-64)
               Registration Authority",
               http://standards.ieee.org/regauth/oui/tutorials/
               EUI64.html

































DeSanti, et al.             Standards Track                    [Page 27]
^L
RFC 4338                 IP over Fibre Channel              January 2006


A.  Transmission of a Broadcast FC Sequence over FC Topologies
    (Informative)

A.1.  Point-to-Point Topology

   No particular mechanisms are required for this case.  The Nx_Port
   connected at the other side of the cable receives the broadcast FC
   Sequence having D_ID 0xFFFFFF.

A.2.  Private Loop Topology

   An NL_Port attached to a private loop must transmit a Class 3
   broadcast FC Sequence by using the OPN(fr) primitive signal
   [FC-AL-2].

   1) The source NL_Port first sends an Open Broadcast Replicate
      (OPN(fr)) primitive signal, forcing all the NL_Ports in the loop
      (except itself) to replicate the frames that they receive while
      examining the FC Header's D_ID field.

   2) The source NL_Port then removes the OPN(fr) signal when it returns
      to it.

   3) The source NL_Port then sends the Class 3 broadcast FC Sequence
      having D_ID 0xFFFFFF.

A.3.  Public Loop Topology

   An NL_Port attached to a public loop must not use the OPN(fr)
   primitive signal.  Rather, it must send the Class 3 broadcast FC
   Sequence having D_ID 0xFFFFFF to the FL_Port at AL_PA = 0x00
   [FC-AL-2].

   The Fabric propagates the broadcast to all other FC_Ports [FC-FS],
   including the FL_Port that the broadcast arrives on.  This includes
   all F_Ports, and other FL_Ports.

   Each FL_Port propagates the broadcast by using the primitive signal
   OPN(fr), in order to prepare the loop to receive the broadcast
   sequence.

A.4.  Fabric Topology

   An N_Port connected to an F_Port must transmit the Class 3 broadcast
   FC Sequence having D_ID 0xFFFFFF to the F_Port.  The Fabric
   propagates the broadcast to all other FC_Ports [FC-FS].





DeSanti, et al.             Standards Track                    [Page 28]
^L
RFC 4338                 IP over Fibre Channel              January 2006


B.  Validation of the <N_Port_Name, N_Port_ID> Mapping
    (Informative)

B.1.  Overview

   At all times, the <N_Port_Name, N_Port_ID> mapping must be valid
   before use.

   After an FC link interruption occurs, the N_Port_ID of an Nx_Port may
   change, as well as the N_Port_IDs of all other Nx_Ports that have
   previously performed Port Login with this Nx_Port.  Because of this,
   address validation is required after a Loop Initialization Primitive
   Sequence (LIP) in a loop topology [FC-AL-2] or after Not_Operational
   Primitive Sequence / Offline Primitive Sequence (NOS/OLS) in a
   point-to-point topology [FC-FS].

   N_Port_IDs do not change as a result of Link Reset (LR) [FC-FS];
   thus, address validation is not required in this case.

B.2.  FC Layer Address Validation in a Point-to-Point Topology

   No validation is required after Link Reset (LR).  In a point-to-point
   topology, NOS/OLS causes implicit Logout of each N_Port and after an
   NOS/OLS each N_Port must again perform a Port Login [FC-FS].

B.3.  FC Layer Address Validation in a Private Loop Topology

   After a LIP [FC-AL-2], an NL_Port must not transmit any data to
   another NL_Port until the address of the other port has been
   validated.  The validation consists of completing the Address
   Discovery procedure with the ADISC ELS [FC-FS].

   If the three FC addresses (N_Port_ID, N_Port_Name, Node_Name) of a
   logged remote NL_Port exactly match the values prior to the LIP, then
   any active Exchange with that NL_Port may continue.

   If any of the three FC addresses has changed, then the remote NL_Port
   must be logged out.

   If an NL_Port's N_Port_ID changes after a LIP, then all active
   logged-in NL_Ports must be logged out.










DeSanti, et al.             Standards Track                    [Page 29]
^L
RFC 4338                 IP over Fibre Channel              January 2006


B.4.  FC Layer Address Validation in a Public Loop Topology

   A Fabric Address Notification (FAN) ELS may be sent by the Fabric to
   all known previously logged-in NL_Ports following an initialization
   event.  Therefore, after a LIP [FC-AL-2], NL_Ports may wait for this
   notification to arrive, or they may perform an FLOGI.

   If the F_Port_Name and Fabric_Name contained in the FAN ELS or FLOGI
   response exactly match the values before the LIP and if the AL_PA
   [FC-AL-2] obtained by the NL_Port is the same as the one before the
   LIP, then the port may resume all Exchanges.  If not, then FLOGI must
   be performed with the Fabric and all logged-in Nx_Ports must be
   logged out.

   A public loop NL_Port must perform the private loop validation as
   specified in section B.3 to any NL_Port on the local loop that has an
   N_Port_ID of the form 0x00-00-XX (i.e., to any private loop NL_Port).

B.5.  FC Layer Address Validation in a Fabric Topology

   No validation is required after Link Reset (LR).

   After NOS/OLS, an N_Port must perform FLOGI.  If, after FLOGI, the
   N_Port's N_Port_ID, the F_Port_Name, and the Fabric_Name are the same
   as before the NOS/OLS, then the N_Port may resume all Exchanges.  If
   not, all logged-in Nx_Ports must be logged out [FC-FS].

C.  Fibre Channel Bit and Byte Numbering Guidance

   Both Fibre Channel and IETF standards use the same byte transmission
   order.  However, the bit numbering is different.

   Fibre Channel bit numbering can be observed if the data structure
   heading shown in figure 24 is cut and pasted at the top of the
   figures present in this document.

         3                   2                   1                   0
       1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                     Figure 24: Fibre Channel Bit Numbering










DeSanti, et al.             Standards Track                    [Page 30]
^L
RFC 4338                 IP over Fibre Channel              January 2006


D.  Changes from RFC 2625

   -  Nx_Ports with N_Port_Name format 0x2, 0x5, 0xC, 0xD, 0xE, and 0xF
      are supported, in addition to format 0x1;
   -  An IP-capable Nx_Port MUST support Class 3;
   -  An IP-capable Nx_Port MUST support continuously increasing
      SEQ_CNT;
   -  An IP-capable Nx_Port SHOULD support a receive data field size for
      Device_Data FC frames of at least 1024 octets;
   -  The FC ESP_Header MAY be used;
   -  FC Classes of services other than 3 are not recommended;
   -  Defined a new FC ARP format;
   -  Removed support for FARP because some FC implementations do not
      tolerate receiving broadcast ELSes;
   -  Added support for IPv4 multicast;
   -  Clarified the usage of the CS_CTL and Parameter fields of the FC
      Header;
   -  Clarified the usage of FC Classes of service;
   -  Clarified the usage of FC Sequences and Exchanges.

E.  Changes from RFC 3831

   -  Clarified the usage of the CS_CTL and Parameter fields of the FC
      Header;
   -  Clarified the usage of FC Classes of service;
   -  Clarified and updated the mapping of IPv6 multicast on Fibre
      Channel;
   -  Clarified the usage of FC Sequences and Exchanges;
   -  Clarified and updated the format of the Neighbor Discovery
      Link-layer option for Fibre Channel.





















DeSanti, et al.             Standards Track                    [Page 31]
^L
RFC 4338                 IP over Fibre Channel              January 2006


Authors' Addresses

   Claudio DeSanti
   Cisco Systems, Inc.
   170 W. Tasman Dr.
   San Jose, CA 95134
   USA

   Phone:  +1 408 853-9172
   EMail:  cds@cisco.com


   Craig W. Carlson
   QLogic Corporation
   6321 Bury Drive
   Eden Prairie, MN 55346
   USA

   Phone:  +1 952 932-4064
   EMail:  craig.carlson@qlogic.com


   Robert Nixon
   Emulex
   3333 Susan Street
   Costa Mesa, CA 92626
   USA

   Phone:  +1 714 885-3525
   EMail:  bob.nixon@emulex.com





















DeSanti, et al.             Standards Track                    [Page 32]
^L
RFC 4338                 IP over Fibre Channel              January 2006


Full Copyright Statement

   Copyright (C) The Internet Society (2006).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
   ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
   INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
   INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
   WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Intellectual Property

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at
   ietf-ipr@ietf.org.

Acknowledgement

   Funding for the RFC Editor function is provided by the IETF
   Administrative Support Activity (IASA).







DeSanti, et al.             Standards Track                    [Page 33]
^L