Internet Engineering Task Force (IETF)                          R. Asati
Request for Comments: 7552                                  C. Pignataro
Updates: 5036, 6720                                              K. Raza
Category: Standards Track                                          Cisco
ISSN: 2070-1721                                                V. Manral
                                                          Ionos Networks
                                                              R. Papneja
                                                                  Huawei
                                                                May
                                                               June 2015

                        Updates to LDP for IPv6

Abstract

   The Label Distribution Protocol (LDP) specification defines
   procedures to exchange label bindings over either IPv4 or IPv6
   networks, or both.  This document corrects and clarifies the LDP
   behavior when an IPv6 network is used (with or without IPv4).  This
   document updates RFCs 5036 and 6720.

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/rfc7552.

Copyright Notice

   Copyright (c) 2015 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.

   This document may contain material from IETF Documents or IETF
   Contributions published or made publicly available before November
   10, 2008.  The person(s) controlling the copyright in some of this
   material may not have granted the IETF Trust the right to allow
   modifications of such material outside the IETF Standards Process.
   Without obtaining an adequate license from the person(s) controlling
   the copyright in such materials, this document may not be modified
   outside the IETF Standards Process, and derivative works of it may
   not be created outside the IETF Standards Process, except to format
   it for publication as an RFC or to translate it into languages other
   than English.

Table of Contents

   1. Introduction ....................................................4
      1.1. Topology Scenarios for Dual-Stack Environment ..............5
      1.2. Single-Hop vs. Multi-Hop LDP Peering .......................6
   2. Specification Language ..........................................6
   3. LSP Mapping .....................................................7
   4. LDP Identifiers .................................................8
   5. Neighbor Discovery ..............................................8
      5.1. Basic Discovery Mechanism ..................................8
           5.1.1. Maintaining Hello Adjacencies .......................9
      5.2. Extended Discovery Mechanism ..............................10
   6. LDP Session Establishment and Maintenance ......................10
      6.1. Transport Connection Establishment ........................10
           6.1.1. Dual-Stack: Transport Connection Preference
                  and Role of an LSR .................................12
      6.2. LDP Session Maintenance ...................................14
   7. Binding Distribution ...........................................15
      7.1. Address Distribution ......................................15
      7.2. Label Distribution ........................................16
   8. LDP Identifiers and Duplicate Next-Hop Addresses ...............17
   9. LDP TTL Security ...............................................18
   10. IANA Considerations ...........................................18
   11. Security Considerations .......................................19
   12. References ....................................................19
      12.1. Normative References .....................................19
      12.2. Informative References ...................................20
   Appendix A. Additional Considerations .............................21
     A.1. LDPv6 and LDPv4 Interoperability Safety Net ................21
     A.2. Accommodating Implementations Not Compliant with RFC 5036 ..21
     A.3. Why prohibit IPv4-mapped IPv6 addresses in LDP? ............22
     A.4. Why a 32-bit value even for the IPv6 LDP Router Id? ........22
   Acknowledgments ...................................................23
   Contributors ......................................................23
   Authors' Addresses.................................................24

1.  Introduction

   The LDP specification [RFC5036] defines procedures and messages for
   exchanging FEC-label bindings over either IPv4 or IPv6 networks, or
   both (i.e., Dual-stack networks).

   However, RFC 5036 has the following deficiencies (i.e., lacks
   details) in regard to IPv6 usage (with or without IPv4):

   1. Label Switched Path (LSP) Mapping: No rule for mapping a
      particular packet to a particular LSP that has an Address Prefix
      Forwarding Equivalence Class (FEC) element containing the IPv6
      address of the egress router

   2. LDP Identifier: No details specific to IPv6 usage

   3. LDP Discovery: No details for using a particular IPv6 destination
      (multicast) address or the source address

   4. LDP Session Establishment: No rule for handling both IPv4 and IPv6
      Transport Address optional objects in a Hello message, and
      subsequently two IPv4 and IPv6 transport connections

   5. LDP Address Distribution: No rule for advertising IPv4 and/or IPv6
      address bindings over an LDP session

   6. LDP Label Distribution: No rule for advertising IPv4 and/or IPv6
      FEC-label bindings over an LDP session, or for handling the
      coexistence of IPv4 and IPv6 FEC Elements in the same FEC TLV

   7. Next-Hop Address Resolution: No rule for accommodating the usage
      of duplicate link-local IPv6 addresses

   8. LDP Time to Live (TTL) Security: No rule for a built-in
      Generalized TTL Security Mechanism (GTSM) in LDP with IPv6 (this
      is a deficiency in [RFC6720])

   This document addresses the above deficiencies by specifying the
   desired behavior/rules/details for using LDP in IPv6-enabled networks
   (IPv6-only or Dual-stack networks).  This document closes the IPv6
   MPLS gap discussed in Sections 3.2.1, 3.2.2, and 3.3.1.1 of
   [RFC7439].

   Note that this document updates [RFC5036] and [RFC6720].

1.1.  Topology Scenarios for Dual-Stack Environment

   Two Label Switching Routers (LSRs) may involve Basic and/or Extended
   LDP Discovery in IPv6 and/or IPv4 address families in various
   topology scenarios.

   This document addresses the following three topology scenarios in
   which the LSRs may be connected via one or more Dual-stack
   LDP-enabled interfaces (Figure 1), or one or more Single-stack
   LDP-enabled interfaces (Figures 2 and 3):

                          R1------------------R2
                                 IPv4+IPv6

            Figure 1: LSRs Connected via a Dual-Stack Interface

                                   IPv4
                           R1=================R2
                                   IPv6

         Figure 2: LSRs Connected via Two Single-Stack Interfaces

                  R1------------------R2---------------R3
                         IPv4                 IPv6

           Figure 3: LSRs Connected via a Single-Stack Interface

   Note that the topology scenario illustrated in Figure 1 also covers
   the case of a Single-stack LDP-enabled interface (say, IPv4) being
   converted to a Dual-stack LDP-enabled interface (by enabling IPv6
   routing as well as IPv6 LDP), even though the LDP-over-IPv4
   (LDPoIPv4) session may already be established between the LSRs.

   Note that the topology scenario illustrated in Figure 2 also
   covers the case of two routers getting connected via an additional
   Single-stack LDP-enabled interface (IPv6 routing and IPv6 LDP), even
   though the LDPoIPv4 session may already be established between the
   LSRs over the existing interface(s).

   This document also addresses the scenario in which the LSRs do the
   Extended Discovery in IPv6 and/or IPv4 address families:

                                   IPv4
                          R1-------------------R2
                                   IPv6

          Figure 4: LSRs Involving IPv4 and IPv6 Address Families

1.2.  Single-Hop vs. Multi-Hop LDP Peering

   The LDP TTL Security mechanism specified by this document applies
   only to single-hop LDP peering sessions, not to multi-hop LDP peering
   sessions, in line with Section 5.5 of [RFC5082].  [RFC5082] describes
   the Generalized TTL Security Mechanism (GTSM).

   As a consequence, any LDP feature that relies on a multi-hop LDP
   peering session would not work with GTSM and will warrant (statically
   or dynamically) disabling GTSM.  Please see Section 9.

2.  Specification Language

   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 [RFC2119].

   Abbreviations:

      LDP                Label Distribution Protocol

      LDPoIPv4           LDP-over-IPv4 transport connection

      LDPoIPv6           LDP-over-IPv6 transport connection

      FEC                Forwarding Equivalence Class

      TLV                Type Length Value

      LSR                Label Switching Router

      LSP                Label Switched Path

      LSPv4              IPv4-signaled Label Switched Path

      LSPv6              IPv6-signaled Label Switched Path

      AFI                Address Family Identifier
      LDP Id             LDP Identifier

      Single-stack LDP   LDP supporting just one address family
                         (for discovery, session setup, address/label
                         binding exchange, etc.)

      Dual-stack LDP     LDP supporting two address families
                         (for discovery, session setup, address/label
                         binding exchange, etc.)

      Dual-stack LSR     LSR supporting Dual-stack LDP for a peer

      Single-stack LSR   LSR supporting Single-stack LDP for a peer

   Note that an LSR can be a Dual-stack and Single-stack LSR at the same
   time for different peers.  This document loosely uses the term
   "address family" to mean "IP address family".

3.  LSP Mapping

   Section 2.1 of [RFC5036] specifies the procedure for mapping a
   particular packet to a particular LSP using three rules.  Quoting the
   third rule from [RFC5036]:

      If it is known that a packet must traverse a particular egress
      router, and there is an LSP that has an Address Prefix FEC element
      that is a /32 address of that router, then the packet is mapped to
      that LSP.

   This rule is correct for IPv4 (to set up LSPv4), but not for IPv6
   (to set up LSPv6), since an IPv6 router may even have a /64 or /96
   or /128 (or whatever prefix length) address.  Hence, that rule is
   updated here to use IPv4 or IPv6 addresses instead of /32 or /128
   addresses, as shown below:

      If it is known that a packet must traverse a particular egress
      router, and there is an LSP that has an Address Prefix FEC element
      that is an IPv4 or IPv6 address of that router, then the packet is
      mapped to that LSP.

4.  LDP Identifiers

   In line with Section 2.2.2 of [RFC5036], this document specifies the
   usage of a 32-bit (unsigned non-zero integer) LSR Id on an
   IPv6-enabled LSR (with or without Dual-stacking).

   This document also qualifies the first sentence of the last paragraph
   of Section 2.5.2 of [RFC5036] to be per address family.

   From Section 2.5.2 of [RFC5036]:

      An LSR MUST advertise the same transport address in all Hellos
      that advertise the same label space.

   Updated by this document, as follows:

      For a given address family, an LSR MUST advertise the same
      transport address in all Hellos that advertise the same label
      space.

   This rightly enables the per-platform label space to be shared
   between IPv4 and IPv6.

   In summary, this document mandates the usage of a common LDP
   Identifier (the same LSR Id and label space id) for both IPv4 and
   IPv6 address families.

5.  Neighbor Discovery

   If Dual-stack LDP is enabled (i.e., LDP enabled in both IPv6 and IPv4
   address families) on an interface or for a targeted neighbor, then
   the LSR MUST transmit both IPv6 and IPv4 LDP (Link or targeted)
   Hellos and include the same LDP Identifier (assuming per-platform
   label space usage) in them.

   If Single-stack LDP is enabled (i.e., LDP enabled in either an IPv6
   or IPv4 address family), then the LSR MUST transmit either IPv6 or
   IPv4 LDP (Link or targeted) Hellos, respectively.

5.1.  Basic Discovery Mechanism

   Section 2.4.1 of [RFC5036] defines the Basic Discovery mechanism for
   directly connected LSRs.  Following this mechanism, LSRs periodically
   send LDP Link Hellos destined to the "all routers on this subnet"
   group multicast IP address.

   Interestingly enough, per the IPv6 addressing architecture [RFC4291],
   IPv6 has three "all routers on this subnet" multicast addresses:

      ff01:0:0:0:0:0:0:2   = Interface-local scope

      ff02:0:0:0:0:0:0:2   = Link-local scope

      ff05:0:0:0:0:0:0:2   = Site-local scope

   [RFC5036] does not specify which particular IPv6 "all routers on this
   subnet" group multicast IP address should be used by LDP Link Hellos.

   This document specifies the usage of link-local scope (i.e.,
   ff02:0:0:0:0:0:0:2) as the destination multicast IP address in IPv6
   LDP Link Hellos.  An LDP Link Hello packet received on any of the
   other destination addresses MUST be dropped.  Additionally, the
   link-local IPv6 address MUST be used as the source IP address in IPv6
   LDP Link Hellos.

   Also, the LDP Link Hello packets MUST have their IPv6 Hop Limit set
   to 255, be checked for the same upon receipt (before any LDP-specific
   processing), and be handled as specified in Section 3 of [RFC5082].
   The built-in inclusion of GTSM automatically protects IPv6 LDP from
   off-link attacks.

   More importantly, if an interface is a Dual-stack LDP interface
   (i.e., LDP enabled in both IPv6 and IPv4 address families), then the
   LSR MUST periodically transmit both IPv6 and IPv4 LDP Link Hellos
   (using the same LDP Identifier per Section 4) on that interface and
   be able to receive them.  This facilitates discovery of IPv6-only,
   IPv4-only, and Dual-stack peers on the interface's subnet and ensures
   successful subsequent peering using the appropriate (address family)
   transport on a multi-access or broadcast interface.

5.1.1.  Maintaining Hello Adjacencies

   In the case of a Dual-stack LDP-enabled interface, the LSR SHOULD
   maintain Link Hello adjacencies for both IPv4 and IPv6 address
   families.  This document, however, allows an LSR to maintain
   Receive-side Link Hello adjacencies only for the address family that
   has been used for the establishment of the LDP session (whether an
   LDPoIPv4 or LDPoIPv6 session).

5.2.  Extended Discovery Mechanism

   The Extended Discovery mechanism (defined in Section 2.4.2 of
   [RFC5036]), in which the targeted LDP Hellos are sent to a unicast
   IPv6 address destination, requires only one IPv6-specific
   consideration: the link-local IPv6 addresses MUST NOT be used as the
   targeted LDP Hello packet's source or destination addresses.

6.  LDP Session Establishment and Maintenance

   Section 2.5.1 of [RFC5036] defines a two-step process for LDP session
   establishment, once the neighbor discovery has completed (i.e., LDP
   Hellos have been exchanged):

   1. Transport connection establishment

   2. Session initialization

   Section 6.1 discusses the LDP considerations for IPv6 and/or
   Dual-stacking in the context of session establishment, whereas
   Section 6.2 discusses the LDP considerations for IPv6 and/or
   Dual-stacking in the context of session maintenance.

6.1.  Transport Connection Establishment

   Section 2.5.2 of [RFC5036] specifies the use of a Transport Address
   optional object (TLV) in LDP Hello messages to convey the transport
   (IP) address; however, it does not specify the behavior of LDP if
   both IPv4 and IPv6 Transport Address objects (TLVs) are sent in a
   Hello message or separate Hello messages.  More importantly, it does
   not specify whether both IPv4 and IPv6 transport connections should
   be allowed if both IPv4 and IPv6 Hello adjacencies were present prior
   to session establishment.

   This document specifies the following:

   1. An LSR MUST NOT send a Hello message containing both IPv4 and IPv6
      Transport Address optional objects.  In other words, there MUST be
      at most one Transport Address optional object in a Hello message.
      An LSR MUST include only the transport address whose address
      family is the same as that of the IP packet carrying the Hello
      message.

   2. An LSR SHOULD accept the Hello message that contains both IPv4 and
      IPv6 Transport Address optional objects but MUST use only the
      transport address whose address family is the same as that of the
      IP packet carrying the Hello message.  An LSR SHOULD accept only
      the first Transport Address optional object for a given address
      family in the received Hello message and ignore the rest if the
      LSR receives more than one Transport Address optional object for a
      given address family.

   3. An LSR MUST send separate Hello messages (each containing either
      an IPv4 or IPv6 Transport Address optional object) for each IP
      address family if Dual-stack LDP is enabled (for an interface or
      neighbor).

   4. An LSR MUST use a global unicast IPv6 address in an IPv6 Transport
      Address optional object of outgoing targeted Hellos and check for
      the same in incoming targeted Hellos (i.e., MUST discard the
      targeted Hello if it failed the check).

   5. An LSR MUST prefer using a global unicast IPv6 address in an
      IPv6 Transport Address optional object of outgoing Link Hellos if
      it had to choose between a global unicast IPv6 address and a
      unique-local or link-local IPv6 address.

   6. A Single-stack LSR MUST establish either an LDPoIPv4 or LDPoIPv6
      session with a remote LSR as per the enabled address family.

   7. A Dual-stack LSR MUST NOT initiate or accept the request for a TCP
      connection for a new LDP session with a remote LSR if it already
      has an LDPoIPv4 or LDPoIPv6 session for the same LDP Identifier
      established with that remote LSR.

      This means that only one transport connection is established,
      regardless of IPv6 and/or IPv4 Hello adjacencies present between
      two LSRs.

   8. A Dual-stack LSR SHOULD prefer establishing an LDPoIPv6 session
      (instead of an LDPoIPv4 session) with a remote Dual-stack LSR by
      following the 'transport connection role' determination logic in
      Section 6.1.1.

      Additionally, to ensure the above preference in the case where
      Dual-stack LDP is enabled on an interface, it would be desirable
      that IPv6 LDP Link Hellos are transmitted before IPv4 LDP Link
      Hellos, particularly when an interface is coming into service or
      being reconfigured.

6.1.1.  Dual-Stack: Transport Connection Preference and Role of an LSR

   Section 2.5.2 of [RFC5036] specifies the rules for determining
   active/passive roles in setting up a TCP connection.  These rules are
   clear for Single-stack LDP but not for Dual-stack LDP, in which an
   LSR may assume different roles for different address families,
   causing the LDP session to not get established.

   To ensure a deterministic transport connection (active/passive) role
   in the case of Dual-stack LDP, this document specifies that the
   Dual-stack LSR conveys its transport connection preference in every
   LDP Hello message.  This preference is encoded in a new TLV, named
   the "Dual-Stack capability" TLV, as defined below:

     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|0|  Dual-Stack capability    |        Length                 |
    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
    |TR     |        Reserved       |     MBZ                       |
    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

                    Figure 5: Dual-Stack Capability TLV

   Where:

      U and F bits: 1 and 0 (as specified by [RFC5036])

      Dual-Stack capability: TLV code point (Ox0701)

      TR:   Transport Connection Preference

            This document defines the following two values:

               0100: LDPoIPv4 connection

               0110: LDPoIPv6 connection (default)

      Reserved

            This field is reserved.  It MUST be set to zero on
            transmission and ignored on receipt.

   A Dual-stack LSR (i.e., an LSR supporting Dual-stack LDP for a peer)
   MUST include the Dual-Stack capability TLV in all of its LDP Hellos
   and MUST set the "TR" field to announce its preference for either an
   LDPoIPv4 or LDPoIPv6 transport connection for that peer.  The default
   preference is LDPoIPv6.

   A Dual-stack LSR MUST always check for the presence of the Dual-Stack
   capability TLV in the received Hello messages and take appropriate
   action, as follows:

   1. If the Dual-Stack capability TLV is present and the remote
      preference does not match the local preference (or does not get
      recognized), then the LSR MUST discard the Hello message and log
      an error.

      If an LDP session was already in place, then the LSR MUST send a
      fatal Notification message with status code of 'Transport
      Connection Mismatch' (0x00000032) and reset the session.

   2. If the Dual-Stack capability TLV is present and the remote
      preference matches the local preference, then:

      a) If TR=0100 (LDPoIPv4), then determine the active/passive roles
         for the TCP connection using an IPv4 transport address as
         defined in Section 2.5.2 of RFC 5036.

      b) If TR=0110 (LDPoIPv6), then determine the active/passive roles
         for the TCP connection by using an IPv6 transport address as
         defined in Section 2.5.2 of RFC 5036.

   3. If the Dual-Stack capability TLV is NOT present and

      a) only IPv4 Hellos are received, then the neighbor is deemed as a
         legacy IPv4-only LSR (supporting Single-stack LDP); hence, an
         LDPoIPv4 session SHOULD be established (similar to that of 2a
         above).

         However, if IPv6 Hellos are also received at any time during
         the life of the session from that neighbor, then the neighbor
         is deemed as a noncompliant Dual-stack LSR (similar to that of
         3c below), resulting in any established LDPoIPv4 session being
         reset and a fatal Notification message being sent (with status
         code of 'Dual-Stack Noncompliance', 0x00000033).

      b) only IPv6 Hellos are received, then the neighbor is deemed as
         an IPv6-only LSR (supporting Single-stack LDP) and an LDPoIPv6
         session SHOULD be established (similar to that of 2b above).

         However, if IPv4 Hellos are also received at any time during
         the life of the session from that neighbor, then the neighbor
         is deemed as a noncompliant Dual-stack LSR (similar to that of
         3c below), resulting in any established LDPoIPv6 session being
         reset and a fatal Notification message being sent (with status
         code of 'Dual-Stack Noncompliance', 0x00000033).

      c) both IPv4 and IPv6 Hellos are received, then the neighbor is
         deemed as a noncompliant Dual-stack neighbor and is not allowed
         to have any LDP session.  A Notification message should be sent
         (with status code of 'Dual-Stack Noncompliance', 0x00000033).

   A Dual-stack LSR MUST convey the same transport connection preference
   ("TR" field value) in all (link and targeted) Hellos that advertise
   the same label space to the same peer and/or on the same interface.
   This ensures that two LSRs linked by multiple Hello adjacencies using
   the same label spaces play the same connection establishment role for
   each adjacency.

   A Dual-stack LSR MUST follow Section 2.5.5 of [RFC5036] and check for
   matching Hello messages from the peer (either all Hellos also include
   the Dual-Stack capability (with the same TR value) or none do).

   A Single-stack LSR does not need to use the Dual-Stack capability in
   Hello messages and SHOULD ignore this capability if received.

   An implementation may provide an option to favor one AFI (say, IPv4)
   over another AFI (say, IPv6) for the TCP transport connection, so as
   to use the favored IP version for the LDP session and force
   deterministic active/passive roles.

   Note: An alternative to this new capability TLV could be a new Flag
   value in an LDP Hello message; however, it would be used even in
   Single-stack IPv6 LDP networks and linger on forever, even though
   Dual-stack will not.  Hence, the idea of this alternative has been
   discarded.

6.2.  LDP Session Maintenance

   This document specifies that two LSRs maintain a single LDP session,
   regardless of the number of Link or targeted Hello adjacencies
   between them, as described in Section 6.1.  This is independent of
   whether:

   - they are connected via a Dual-stack LDP-enabled interface(s) or via
     two (or more) Single-stack LDP-enabled interfaces;

   - a Single-stack LDP-enabled interface is converted to a Dual-stack
     LDP-enabled interface (see Figure 1) on either LSR;

   - an additional Single-stack or Dual-stack LDP-enabled interface is
     added or removed between two LSRs (see Figure 2).

   If the last Hello adjacency for a given address family goes down
   (e.g., due to Dual-stack LDP-enabled interfaces being converted into
   Single-stack LDP-enabled interfaces on one LSR) and that address
   family is the same as the one used in the transport connection, then
   the transport connection (LDP session) MUST be reset.  Otherwise, the
   LDP session MUST stay intact.

   If the LDP session is torn down for whatever reason (LDP disabled for
   the corresponding transport, Hello adjacency expiry, preference
   mismatch, etc.), then the LSRs SHOULD initiate the establishment of a
   new LDP session as per the procedures described in Section 6.1 of
   this document.

7.  Binding Distribution

   LSRs by definition can be enabled for Dual-stack LDP globally and/or
   per peer so as to exchange the address and label bindings for both
   IPv4 and IPv6 address families, independent of any LDPoIPv4 or
   LDPoIPv6 session between them.

   However, there might be some legacy LSRs that are fully compliant
   with RFC 5036 for IPv4 but are noncompliant for IPv6 (for example,
   see Section 3.5.5.1 of RFC 5036), causing them to reset the session
   upon receiving IPv6 address bindings or IPv6 FEC (Prefix) label
   bindings from a peer compliant with this document.  This is somewhat
   undesirable, as clarified further in Appendices A.1 and A.2.

   To help maintain backward compatibility (i.e., accommodate IPv4-only
   LDP implementations that may not be compliant with RFC 5036,
   Section 3.5.5.1), this specification requires that an LSR MUST NOT
   send any IPv6 bindings to a peer if the peer has been determined to
   be a legacy LSR.

   The Dual-Stack capability TLV, which is defined in Section 6.1.1, is
   also used to determine whether or not a peer is a legacy (IPv4-only
   Single-stack) LSR.

7.1.  Address Distribution

   An LSR MUST NOT advertise (via an Address message) any IPv4-mapped
   IPv6 addresses (as defined in Section 2.5.5.2 of [RFC4291]) and MUST
   ignore such addresses if ever received.  Please see Appendix A.3.

   If an LSR is enabled with Single-stack LDP for any peer, then it MUST
   advertise (via an Address message) its local IP addresses as per the
   enabled address family to that peer and process received Address
   messages containing IP addresses as per the enabled address family
   from that peer.

   If an LSR is enabled with Dual-stack LDP for a peer and

   1. does not find the Dual-Stack capability TLV in the incoming IPv4
      LDP Hello messages from that peer, then the LSR MUST NOT advertise
      its local IPv6 addresses to the peer.

   2. finds the Dual-Stack capability TLV in the incoming IPv4 (or IPv6)
      LDP Hello messages from that peer, then it MUST advertise (via an
      Address message) its local IPv4 and IPv6 addresses to that peer.

   3. does not find the Dual-Stack capability TLV in the incoming IPv6
      LDP Hello messages, then it MUST advertise (via an Address
      message) only its local IPv6 addresses to that peer.

      This last point helps to maintain forward compatibility (no need
      to require this TLV in the case of IPv6 Single-stack LDP).

7.2.  Label Distribution

   An LSR MUST NOT allocate and MUST NOT advertise FEC-label bindings
   for link-local or IPv4-mapped IPv6 addresses (defined in
   Section 2.5.5.2 of [RFC4291]), and it MUST ignore such bindings if
   ever received.  Please see Appendix A.3.

   If an LSR is enabled with Single-stack LDP for any peer, then it MUST
   advertise (via a Label Mapping message) FEC-label bindings for the
   enabled address family to that peer and process received FEC-label
   bindings for the enabled address family from that peer.

   If an LSR is enabled with Dual-stack LDP for a peer and

   1. does not find the Dual-Stack capability TLV in the incoming IPv4
      LDP Hello messages from that peer, then the LSR MUST NOT advertise
      IPv6 FEC-label bindings to the peer (even if IP capability
      negotiation for the IPv6 address family was done).

   2. finds the Dual-Stack capability TLV in the incoming IPv4 (or IPv6)
      LDP Hello messages from that peer, then it MUST advertise
      FEC-label bindings for both IPv4 and IPv6 address families to that
      peer.

   3. does not find the Dual-Stack capability TLV in the incoming IPv6
      LDP Hello messages, then it MUST advertise FEC-label bindings for
      IPv6 address families to that peer.

      This last point helps to maintain forward compatibility (no need
      to require this TLV for IPv6 Single-stack LDP).

   An LSR MAY further constrain the advertisement of FEC-label bindings
   for a particular address family by negotiating the IP capability for
   a given address family, as specified in [RFC7473].  This allows an
   LSR pair to neither advertise nor receive the undesired FEC-label
   bindings on a per-address-family basis to a peer.

   If an LSR is configured to change an interface or peer from
   Single-stack LDP to Dual-stack LDP, then an LSR SHOULD use Typed
   Wildcard FEC procedures [RFC5918] to request the label bindings for
   the enabled address family.  This helps to relearn the label bindings
   that may have been discarded before, without resetting the session.

8.  LDP Identifiers and Duplicate Next-Hop Addresses

   RFC 5036, Section 2.7 specifies the logic for mapping the IP routing
   next hop (of a given FEC) to an LDP peer so as to find the correct
   label entry for that FEC.  The logic involves using the IP routing
   next-hop address as an index into the (peer address) database (which
   is populated by the Address message containing a mapping between each
   peer's local addresses and its LDP Identifier) to determine the LDP
   peer.

   However, this logic is insufficient to deal with duplicate IPv6
   (link-local) next-hop addresses used by two or more peers.  The
   reason is that all interior IPv6 routing protocols (can) use
   link-local IPv6 addresses as the IP routing next hops, and
   "IP Version 6 Addressing Architecture" [RFC4291] allows a link-local
   IPv6 address to be used on more than one link.

   Hence, this logic is extended by this specification to use not only
   the IP routing next-hop address but also the IP routing next-hop
   interface to uniquely determine the LDP peer(s).  The next-hop
   address-based LDP peer mapping is to be done through the LDP peer
   address database (populated by Address messages received from the LDP
   peers), whereas next-hop interface-based LDP peer mapping is to be
   done through the LDP Hello adjacency/interface database (populated by
   Hello messages received from the LDP peers).

   This extension solves the problem of two or more peers using the same
   link-local IPv6 address (in other words, duplicate peer addresses) as
   the IP routing next hops.

   Lastly, for better scale and optimization, an LSR may advertise only
   the link-local IPv6 addresses in the Address message, assuming that
   the peer uses only the link-local IPv6 addresses as static and/or
   dynamic IP routing next hops.

9.  LDP TTL Security

   This document recommends enabling mandates the use of the Generalized TTL Security
   Mechanism (GTSM) [RFC6720] for LDP, as specified in [RFC6720], LDP Link Hello packets over IPv6 (see
   Section 5.1).

   This document further recommends enabling GTSM for the LDP/TCP
   transport connection over IPv6 (i.e., LDPoIPv6).  The inclusion of  This GTSM inclusion
   is intended to automatically protect IPv6 LDP peering sessions from
   off-link attacks.

   [RFC6720] allows for the implementation to statically (via
   configuration) and/or dynamically override the default behavior
   (enable/disable GTSM) on a per-peer basis.  Such an option could be
   set on either LSR in a peering session (since GTSM negotiation would
   ultimately disable GTSM between the LSR and its peer(s)).

   LDP Link Hello packets MUST have their IPv6 Hop Limit set to 255 and
   be checked for the same upon receipt before any further processing,
   as per Section 3 of [RFC5082].

10.  IANA Considerations

   This document defines a new optional parameter for the LDP Hello
   message and two new status codes for the LDP Notification message.

   The "Dual-Stack capability" parameter has been assigned a code point
   (0x0701) from the "TLV Type Name Space" registry.  IANA has allocated
   this code point from the IETF Consensus range 0x0700-0x07ff for the
   Dual-Stack capability TLV.

   The 'Transport Connection Mismatch' status code has been assigned a
   code point (0x00000032) from the "Status Code Name Space" registry.
   IANA has allocated this code point from the IETF Consensus range and
   marked the E bit column with a '1'.

   The 'Dual-Stack Noncompliance' status code has been assigned a code
   point (0x00000033) from the "Status Code Name Space" registry.  IANA
   has allocated this code point from the IETF Consensus range and
   marked the E bit column with a '1'.

11.  Security Considerations

   The extensions defined in this document only clarify the behavior of
   LDP; they do not define any new protocol procedures.  Hence, this
   document does not add any new security issues to LDP.

   While the security issues relevant for [RFC5036] are relevant for
   this document as well, this document reduces the chances of off-link
   attacks when using an IPv6 transport connection by including the use
   of GTSM procedures [RFC5082].  Please see Section 9 for LDP TTL
   Security details.

   Moreover, this document allows the use of IPsec [RFC4301] for IPv6
   protection; hence, LDP can benefit from the additional security as
   specified in [RFC7321] as well as [RFC5920].

12.  References

12.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, March 1997,
              <http://www.rfc-editor.org/info/rfc2119>.

   [RFC4291]  Hinden, R. and S. Deering, "IP Version 6 Addressing
              Architecture", RFC 4291, February 2006,
              <http://www.rfc-editor.org/info/rfc4291>.

   [RFC5036]  Andersson, L., Ed., Minei, I., Ed., and B. Thomas, Ed.,
              "LDP Specification", RFC 5036, October 2007,
              <http://www.rfc-editor.org/info/rfc5036>.

   [RFC5082]  Gill, V., Heasley, J., Meyer, D., Savola, P., Ed., and C.
              Pignataro, "The Generalized TTL Security Mechanism
              (GTSM)", RFC 5082, October 2007,
              <http://www.rfc-editor.org/info/rfc5082>.

   [RFC5918]  Asati, R., Minei, I., and B. Thomas, "Label Distribution
              Protocol (LDP) 'Typed Wildcard' Forward Equivalence Class
              (FEC)", RFC 5918, August 2010,
              <http://www.rfc-editor.org/info/rfc5918>.

12.2.  Informative References

   [RFC4038]  Shin, M-K., Ed., Hong, Y-G., Hagino, J., Savola, P., and
              E. Castro, "Application Aspects of IPv6 Transition",
              RFC 4038, March 2005,
              <http://www.rfc-editor.org/info/rfc4038>.

   [RFC4301]  Kent, S. and K. Seo, "Security Architecture for the
              Internet Protocol", RFC 4301, December 2005,
              <http://www.rfc-editor.org/info/rfc4301>.

   [RFC5340]  Coltun, R., Ferguson, D., Moy, J., and A. Lindem, "OSPF
              for IPv6", RFC 5340, July 2008,
              <http://www.rfc-editor.org/info/rfc5340>.

   [RFC5920]  Fang, L., Ed., "Security Framework for MPLS and GMPLS
              Networks", RFC 5920, July 2010,
              <http://www.rfc-editor.org/info/rfc5920>.

   [RFC6286]  Chen, E. and J. Yuan, "Autonomous-System-Wide Unique BGP
              Identifier for BGP-4", RFC 6286, June 2011,
              <http://www.rfc-editor.org/info/rfc6286>.

   [RFC6720]  Pignataro, C. and R. Asati, "The Generalized TTL Security
              Mechanism (GTSM) for the Label Distribution Protocol
              (LDP)", RFC 6720, August 2012,
              <http://www.rfc-editor.org/info/rfc6720>.

   [RFC7321]  McGrew, D. and P. Hoffman, "Cryptographic Algorithm
              Implementation Requirements and Usage Guidance for
              Encapsulating Security Payload (ESP) and Authentication
              Header (AH)", RFC 7321, August 2014,
              <http://www.rfc-editor.org/info/rfc7321>.

   [RFC7439]  George, W., Ed., and C. Pignataro, Ed., "Gap Analysis for
              Operating IPv6-Only MPLS Networks", RFC 7439,
              January 2015, <http://www.rfc-editor.org/info/rfc7439>.

   [RFC7473]  Raza, K. and S. Boutros, "Controlling State Advertisements
              of Non-negotiated LDP Applications", RFC 7473, March 2015,
              <http://www.rfc-editor.org/info/rfc7473>.

Appendix A.  Additional Considerations

A.1.  LDPv6 and LDPv4 Interoperability Safety Net

   It is not safe to assume that implementations compliant with RFC 5036
   have supported the handling of an IPv6 address family (IPv6
   FEC-label) in a Label Mapping message all along.

   If a router upgraded per this specification advertised both IPv4 and
   IPv6 FECs in the same Label Mapping message, then an IPv4-only peer
   (not knowing how to process such a message) may abort processing the
   entire Label Mapping message (thereby discarding even the IPv4
   FEC-labels), as per Section 3.4.1.1 of [RFC5036].

   This would result in LDPv6 being somewhat undeployable in existing
   production networks.

   Section 7 of this document provides a good safety net and makes LDPv6
   incrementally deployable without making any such assumption on the
   routers' support for IPv6 FEC processing in current production
   networks.

A.2.  Accommodating Implementations Not Compliant with RFC 5036

   It is not safe to assume that implementations have been [RFC5036]
   compliant in gracefully handling an IPv6 address family (IPv6 Address
   List TLV) in an Address message all along.

   If a router upgraded per this specification advertised IPv6 addresses
   (with or without IPv4 addresses) in an Address message, then an
   IPv4-only peer (not knowing how to process such a message) may not
   follow Section 3.5.5.1 of [RFC5036] and may tear down the LDP
   session.

   This would result in LDPv6 being somewhat undeployable in existing
   production networks.

   Sections 6 and 7 of this document provide a good safety net and make
   LDPv6 incrementally deployable without making any such assumption on
   the routers' support for IPv6 FEC processing in current production
   networks.

A.3.  Why prohibit IPv4-mapped IPv6 addresses in LDP?

   Per discussion with the 6MAN and V6OPS working groups, the
   overwhelming consensus was to not promote IPv4-mapped IPv6 addresses
   appearing in the routing table, as well as in LDP (address and label)
   databases.

   Also, [RFC4038], Section 4.2 suggests that IPv4-mapped IPv6-addressed
   packets should never appear on the wire.

A.4.  Why a 32-bit value even for the IPv6 LDP Router Id?

   The first four octets of the LDP Identifier, the 32-bit LSR Id (i.e.,
   LDP router Id), identify the LSR and provide a globally unique value
   within the MPLS network, regardless of the address family used for
   the LDP session.

   Please note that the 32-bit LSR Id value would not map to any IPv4
   address in an IPv6-only LSR (i.e., Single-stack), nor would there be
   an expectation of it being IP routable or DNS resolvable.  In IPv4
   deployments, the LSR Id is typically derived from an IPv4 address,
   generally assigned to a loopback interface.  In IPv6-only
   deployments, this 32-bit LSR Id must be derived by some other means
   that guarantees global uniqueness within the MPLS network, similar to
   that of the BGP Identifier [RFC6286] and the OSPF router Id
   [RFC5340].

   This document reserves 0.0.0.0 as the LSR Id and prohibits its usage
   with IPv6, in line with the OSPF router Id in OSPF version 3
   [RFC5340].

Acknowledgments

   We acknowledge the authors of [RFC5036], since some text in this
   document is borrowed from [RFC5036].

   Thanks to Bob Thomas for providing critical feedback to improve this
   document early on.

   Many thanks to Eric Rosen, Lizhong Jin, Bin Mo, Mach Chen, Shane
   Amante, Pranjal Dutta, Mustapha Aissaoui, Matthew Bocci, Mark Tinka,
   Tom Petch, Kishore Tiruveedhula, Manoj Dutta, Vividh Siddha, Qin Wu,
   Simon Perreault, Brian E. Carpenter, Santosh Esale, Danial Johari,
   and Loa Andersson for thoroughly reviewing this document and for
   providing insightful comments and multiple improvements.

Contributors

   The following individuals contributed to this document:

   Nagendra Kumar
   Cisco Systems, Inc.
   7200 Kit Creek Road
   Research Triangle Park, NC  27709, United States
   EMail: naikumar@cisco.com

   Andre Pelletier
   Cisco Systems, Inc.
   2000 Innovation Drive
   Kanata, ON K2K-3E8, Canada
   EMail: apelleti@cisco.com

Authors' Addresses

   Rajiv Asati
   Cisco Systems, Inc.
   7025 Kit Creek Road
   Research Triangle Park, NC  27709-4987
   United States

   EMail: rajiva@cisco.com

   Carlos Pignataro
   Cisco Systems, Inc.
   7200 Kit Creek Road
   Research Triangle Park, NC  27709-4987
   United States

   EMail: cpignata@cisco.com

   Kamran Raza
   Cisco Systems, Inc.
   2000 Innovation Drive
   Ottawa, ON  K2K-3E8
   Canada

   EMail: skraza@cisco.com

   Vishwas Manral
   Ionos Networks
   4100 Moorpark Ave., Ste. #122
   San Jose, CA  95117
   United States
   Phone: +1 408 447 1497

   EMail: vishwas@ionosnetworks.com

   Rajiv Papneja
   Huawei Technologies
   2330 Central Expressway
   Santa Clara, CA  95050
   United States
   Phone: +1 571 926 8593

   EMail: rajiv.papneja@huawei.com