RFC 8614 PROPOSED STANDARD BGP · VPLS · MPLS

VPLS Control Flags: Updated Processing of Control Flags for BGP Virtual Private LAN Service (VPLS)

How BGP VPLS routers should behave when they disagree about control words and sequence numbers.

On rfc-editor.org ↗
PUBLISHED
June 2019
AUTHORS
R. Singh, K. Kompella, S. Palislamovic
STREAM
IETF
BUILDS ON
RFC 4761 — VPLS using BGP
FLAGS CHANGED
C-bit and S-bit
C-BIT MISMATCH
PW comes up without CW
S-BIT MISMATCH
PW SHOULD NOT come up
IANA ACTIONS
None
DOI
10.17487/RFC8614

RFC text

Contents
Original text · RFC Editor View on rfc-editor.org ↗
Internet Engineering Task Force (IETF)                          R. Singh
Request for Comments: 8614                                   K. Kompella
Updates: 4761                                           Juniper Networks
Category: Standards Track                                S. Palislamovic
ISSN: 2070-1721                                                    Nokia
                                                               June 2019


                Updated Processing of Control Flags for
                 BGP Virtual Private LAN Service (VPLS)

Abstract

   This document updates the meaning of the Control Flags field in the
   "Layer2 Info Extended Community" used for BGP Virtual Private LAN
   Service (VPLS) Network Layer Reachability Information (NLRI) as
   defined in RFC 4761.  This document updates RFC 4761.

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 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc8614.

Copyright Notice

   Copyright (c) 2019 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
   (https://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.

Table of Contents

   1. Introduction ....................................................2
      1.1. Terminology ................................................3
   2. Problem Description .............................................3
   3. Updated Meaning of Control Flags in the Layer2 Info Extended
      Community .......................................................3
      3.1. Control Word (C-Bit) .......................................4
      3.2. Sequence Flag (S-Bit) ......................................4
   4. Using Point-to-Multipoint (P2MP) LSPs as Transport for
      BGP VPLS ........................................................5
   5. Illustrative Diagram ............................................6
   6. Treatment of C-Bits and S-Bits in Multihoming Scenarios .........7
      6.1. Control Word (C-Bit) .......................................7
      6.2. Sequence Flag (S-Bit) ......................................7
   7. Security Considerations .........................................8
   8. IANA Considerations .............................................8
   9. References ......................................................8
      9.1. Normative References .......................................8
      9.2. Informative References .....................................9
   Authors' Addresses .................................................9

1.  Introduction

   "Virtual Private LAN Service (VPLS) Using BGP for Auto-Discovery and
   Signaling" [RFC4761] describes the concepts and signaling for using
   the Border Gateway Protocol (BGP) to set up a VPLS.  It specifies the
   BGP VPLS Network Layer Reachability Information (NLRI) by which a
   Provider Edge (PE) router may require other PEs in the same VPLS to
   include (or not) the Control Word (CW) and sequencing information in
   VPLS frames sent to this PE.

   The use of the CW helps prevent the misordering of IPv4 or IPv6
   Pseudowire (PW) traffic over Equal-Cost Multipath (ECMP) paths or
   Link Aggregation Group (LAG) bundles.  [RFC4385] describes the format
   for the CW that may be used over point-to-point PWs and over a VPLS.
   Along with [RFC3985], [RFC4385] also describes sequence number usage
   for VPLS frames.

   However, [RFC4761] does not specify the behavior of PEs in a mixed
   environment where some PEs support CW/sequencing and others do not.

1.1.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
   "OPTIONAL" in this document are to be interpreted as described in
   BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

2.  Problem Description

   [RFC4761] specifies the VPLS BGP NLRI by which a given PE advertises
   the behavior expected by the multiple PEs participating in the same
   VPLS.  The NLRI indicates the VPLS label that the various PE routers,
   which are referred to in the NLRI, should use when forwarding VPLS
   traffic to this PE.  Additionally, by using the Control Flags, this
   PE specifies whether the other PEs (in the same VPLS) should use the
   CW or sequenced delivery for frames forwarded to this PE.  These are
   indicated by the C-bits and the S-bits, respectively, in the Control
   Flags, as specified in Section 3.2.4 in [RFC4761].

   [RFC4761] requires that if the advertising PE sets the C-bits and
   S-bits, the receiving PE MUST, respectively, insert a CW and include
   sequence numbers when forwarding VPLS traffic to the advertising PE.

   However, in a BGP VPLS deployment, there would often be cases where a
   PE receiving the VPLS BGP NLRI may not have the ability to insert a
   CW or include sequencing information inside PW frames.  Thus, the
   behavior of CW processing and sequencing needs to be further
   specified.

   This document updates the meaning of the Control Flags in the Layer2
   Info Extended Community in the BGP VPLS NLRI.  It also specifies the
   forwarding behavior for a mixed-mode environment where not every PE
   in a VPLS has the ability or the configuration to honor the Control
   Flags received from the PE advertising the BGP NLRI.

3.  Updated Meaning of Control Flags in the Layer2 Info Extended
    Community

   [RFC4761] does not allow for the CW setting to be negotiated.  In a
   typical implementation, if a PE sets the C-bit, it expects to receive
   VPLS frames with a CW and will send frames the same way.  If the PEs
   at the two ends of a PW do not agree on the setting of the C-bit, the
   PW does not come up.  The behavior is similar for the S-bit.

   This memo updates the meaning of the C-bit and the S-bit in the
   Control Flags.

3.1.  Control Word (C-Bit)

   If a PE sets the C-bit in its NLRI, it means that the PE has the
   ability to send and receive frames with a CW.

   -  If the PEs at both ends of a PW set the C-bit, CWs MUST be used in
      both directions of the PW.

   -  If both PEs send a C-bit of 0, CWs MUST NOT be used on the PW.

   These two cases behave as before.

   However, if the PEs at both ends of the PW do not agree on the
   setting of the C-bit, CWs MUST NOT be used in either direction on
   that PW, but the PW MUST NOT be prevented from coming up due to this
   mismatch.  So, the PW will still come up but will not use the CW in
   either direction.  This behavior is changed from the behavior
   described in [RFC4761] where the PW does not come up.

3.2.  Sequence Flag (S-Bit)

   If a PE sets the S-bit in its NLRI, it means that the PE has the
   ability to set sequence numbers as described in Section 4.1 in
   [RFC4385] and process sequence numbers as described in Section 4.2 in
   [RFC4385].

   -  If the PEs at both ends of a PW set the S-bit, non-zero sequence
      numbers MUST be used in both directions of the PW.

   -  If both PEs send an S-bit of 0, sequence numbers MUST NOT be used
      on the PW.

   These two cases behave as before.

   [RFC4761] does not allow for the S-bit setting to be negotiated
   either.  In a typical implementation, if the PE sets the S-bit in the
   advertised NLRI, it expects to receive VPLS frames with non-zero
   sequence numbers and will send outgoing frames over the PW with
   non-zero sequence numbers.

   This memo further specifies the expected behavior when the PEs at the
   ends of the PW advertise differing S-bit values.  If the PEs at both
   ends of the PW do not agree on the setting of the S-bit, then the PW
   SHOULD NOT come up.  This is to avoid running into out-of-sequence
   ordering scenarios when the multiple PEs that are enabling
   multihoming for a site have differing S-bit advertisements as
   described in Section 4.2 in [RFC4385].  However, if a deployment is
   known to not utilize multihoming, a user-configurable way to override
   this recommendation MAY be provided by an implementation whereby the
   PW is allowed to come up.  In that case, the PE advertising the S-bit
   as 0 should set sequence numbers in the frames as 0, and the PW
   receiving the frames should not expect to receive non-zero sequence
   numbers.

4.  Using Point-to-Multipoint (P2MP) LSPs as Transport for BGP VPLS

   BGP VPLS can be used over point-to-point Label Switched Paths (LSPs)
   acting as transport between the VPLS PEs.  Alternately, BGP VPLS may
   also be used over Point-to-Multipoint (P2MP) LSPs with the source of
   the P2MP LSP rooted at the PE advertising the VPLS BGP NLRI.

   In a network that uses P2MP LSPs as transport for a VPLS, there may
   be some PEs that support the CW while others may not.  The behavior
   is similar for the sequencing of VPLS frames.

   In such a setup, a source PE that supports CW should set up two
   different P2MP LSPs such that:

   -  One P2MP LSP will transport CW-marked frames to those PEs that
      advertised the C-bit as 1.

   -  The other P2MP LSP will transport frames without the CW to those
      PEs that advertised the C-bit as 0.

   Using two different P2MP LSPs to deliver frames with and without the
   CW to different PEs ensures that a P2MP root PE honors the C-bit
   advertised by the other P2MP PEs.

   However, the set of leaves on the two P2MP LSPs (rooted at the given
   PE) MUST NOT contain any PEs that advertised a value for the S-bit
   different from what the root PE itself is advertising.  PEs that
   advertised their S-bit values differently (from what the P2MP root PE
   advertised) will not be on either of the P2MP LSPs.  This ensures
   that the P2MP root PE is sending VPLS frames only to those PEs that
   agree on the setting of the S-bit.

   The ingress router for the P2MP LSP should send separate NLRIs for
   the cases of using the CW and for not using the CW.

5.  Illustrative Diagram

                                                          -----
                                                         /  A1 \
           ----                                     ____CE1     |
          /    \          --------       --------  /    |       |
         |  A2 CE2-      /        \     /        PE1     \     /
          \    /   \    /          \___/          | \     -----
           ----     ---PE2                        |  \
                       |                          |   \   -----
                       | Service Provider Network |    \ /     \
                       |                          |     CE5  A5
                       |            ___           |   /  \     /
                        \          /   \         PE4_/    -----
                        PE3       /     \       /
                |------/  \-------       -------
         ----  /   |    ----
        /    \/    \   /    \               CE = Customer Edge Device
       |  A3 CE3    --CE4 A4 |              PE = Provider Edge Router
        \    /         \    /
         ----           ----                A<n> = Customer site n

                        Figure 1: Example of a VPLS

   In the above topology, let there be a VPLS configured with the PEs as
   displayed.  Let PE1 be the PE under consideration that is CW enabled
   and sequencing enabled.  Let PE2 and PE3 also be CW enabled and
   sequencing enabled.  Let PE4 not be CW enabled or have the ability to
   include sequence numbers.  PE1 will advertise a VPLS BGP NLRI,
   containing the C/S-bits marked as 1.  PE2 and PE3, on learning of the
   NLRI from PE1, will include the CW and non-zero sequence numbers in
   the VPLS frames being forwarded to PE1 as described in Section 4 in
   [RFC4385].  However, PE4, which does not have the ability to include
   a CW or include non-zero sequence numbers, will not.

   As per [RFC4761], PE1 would expect all other PEs to forward
   CW-containing frames that have non-zero sequence numbers.  That
   expectation cannot be met by PE4 in this example.  Thus, as per
   [RFC4761], the PW between PE1 and PE4 does not come up.

   However, this document addresses how an implementation should support
   BGP VPLS in a network where a subset of the BGP VPLS PEs support the
   CW and/or frame sequencing.  PE1 will not bring up the PW with PE4
   due to the S-bit mismatch, unless overridden by local configuration
   on PE1 and PE4 as specified in Section 3.2.  If PE4 instead was to
   advertise a C-bit of 0 and an S-bit of 1, then the PW between PE1 and
   PE4 would come up despite the CW mismatch.  Additionally, PE1 would
   set up its data plane such that it will strip the CW only for those
   VPLS frames that are received from PEs that have indicated their
   desire to receive CW-marked frames.  So, PE1 will set up its data
   plane to strip the CW only for VPLS frames received from PE2 and PE3,
   and it will expect to process PW frames containing non-zero sequence
   numbers as described in Section 4.2 in [RFC4385].  PE1 will set up
   its data plane to not strip the CW from frames received from PE4, and
   it would expect PE4 to send frames with non-zero sequence numbers.
   All frames sent by PE4 to PE1 over the PW would have a non-zero
   sequence number.

6.  Treatment of C-Bits and S-Bits in Multihoming Scenarios

6.1.  Control Word (C-Bit)

   In a multihomed environment, different PEs may effectively represent
   the same service destination endpoint.  It could be assumed that the
   end-to-end PW establishment process should follow the same rules when
   it comes to CW requirements, meaning that setting the C-bit would be
   enforced equally toward both primary and backup designated
   forwarders.

   However, in the multihoming case, each PW SHOULD be evaluated
   independently.  Assuming the network topology specified in Section 5,
   there could be the case where the PW between PE2 and PE1 could have
   the CW signaled via the extended community and would be used in the
   VPLS frame, while the PE2-to-PE4 PW would not insert the CW in the
   VPLS frame due to a C-bit mismatch.  The multihoming behavior of the
   rest of the PEs should simply follow the rules specified in
   [VPLS-MULTIHOMING].

6.2.  Sequence Flag (S-Bit)

   In a multihomed environment, different PEs may effectively represent
   the same service destination endpoint.  In this case, the rules for
   end-to-end PW establishment SHOULD follow the same behavior as that
   described in Section 3.2 when it comes to S-bit requirements.
   Consider the case described in Section 5 with CE5 having a connection
   to multiple PEs (multihomed) to PE4 and PE1.  The PW's behavior is
   similar to that for the CW scenario such that the S-bit evaluation
   SHOULD be independent per PW.  So, in the case where PE4 does not set
   the S-bit in its advertised NLRI, there is an S-bit mismatch between
   PE1 and PE4.  This mismatch prevents the PW establishment between PE1
   and PE4.  So, only one PW -- between PE1 and PE2 -- would be
   established for the multihomed site shown.  Thus, even though CE5 is
   physically multihomed, due to PE4's lack of support for sending
   frames with non-zero sequence numbers, there would be no PW between
   PE2 and PE4.  CE5 would effectively not be multihomed.

7.  Security Considerations

   This document updates the behavior specified in [RFC4761].  The
   security considerations discussed in [RFC4761] apply.  This document
   essentially addresses BGP VPLS behavior for PEs when the C-bit value,
   the S-bit value, or both values advertised by a given PE are
   different from what another PE in the VPLS is advertising.  Any
   bit-flipping media errors leading to causing this mismatch of
   C/S-bits between PEs do not adversely affect the availability of the
   PWs.  Rather, they cause CWs to not be used or cause the
   NLRI-advertising PE to not expect non-zero sequenced frames, for the
   C-bit and the S-bit, respectively, being mismatched across PEs.  This
   is no worse than the previous behavior where any bit-flipping media
   errors leading to a mismatch of the C/S-bits between PEs would cause
   the PW to not come up.

8.  IANA Considerations

   This document has no IANA actions.

9.  References

9.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <https://www.rfc-editor.org/info/rfc2119>.

   [RFC4761]  Kompella, K., Ed. and Y. Rekhter, Ed., "Virtual Private
              LAN Service (VPLS) Using BGP for Auto-Discovery and
              Signaling", RFC 4761, DOI 10.17487/RFC4761, January 2007,
              <https://www.rfc-editor.org/info/rfc4761>.

   [RFC4385]  Bryant, S., Swallow, G., Martini, L., and D. McPherson,
              "Pseudowire Emulation Edge-to-Edge (PWE3) Control Word for
              Use over an MPLS PSN", RFC 4385, DOI 10.17487/RFC4385,
              February 2006, <https://www.rfc-editor.org/info/rfc4385>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in
              RFC 2119 Key Words", BCP 14, RFC 8174,
              DOI 10.17487/RFC8174, May 2017,
              <https://www.rfc-editor.org/info/rfc8174>.

9.2.  Informative References

   [RFC3985]  Bryant, S., Ed. and P. Pate, Ed., "Pseudo Wire Emulation
              Edge-to-Edge (PWE3) Architecture", RFC 3985,
              DOI 10.17487/RFC3985, March 2005,
              <https://www.rfc-editor.org/info/rfc3985>.

   [VPLS-MULTIHOMING]
              Kothari, B., Kompella, K., Henderickx, W., Balus, F.,
              and J. Uttaro, "BGP based Multi-homing in Virtual
              Private LAN Service", Work in Progress,
              draft-ietf-bess-vpls-multihoming-03, March 2019.

Authors' Addresses

   Ravi Singh
   Juniper Networks
   1133 Innovation Way
   Sunnyvale, CA  94089
   United States of America

   Email: ravis@juniper.net


   Kireeti Kompella
   Juniper Networks
   1133 Innovation Way
   Sunnyvale, CA  94089
   United States of America

   Email: kireeti@juniper.net


   Senad Palislamovic
   Nokia
   600 Mountain Avenue
   Murray Hill, NJ  07974-0636
   United States of America

   Email: Senad.palislamovic@nokia.com

AI ELI5

On this page
PROBLEMRFC 4761 lets a router demand that peers use a control word (CW) and sequence numbers, but does not say what to do when a peer cannot, so a mismatched pseudowire simply does not come up.
MECHANISMA C-bit mismatch no longer blocks the pseudowire: it comes up without the CW in either direction. An S-bit mismatch SHOULD still keep it down, though an override MAY be configured where there is no multihoming.
WHY IT MATTERSNetworks that mix routers with and without CW support can still build VPLS pseudowires. The RFC updates RFC 4761 and is a Proposed Standard.

§1Introduction and Problem

RFC 4761 describes how BGP sets up a VPLS (Virtual Private LAN Service). In the BGP VPLS NLRI, a Provider Edge (PE) router uses Control Flags to say whether other PEs in the same VPLS should include a Control Word (CW) and sequence numbers in frames sent to it. These are the C-bits and S-bits.

The CW helps prevent misordering of IPv4 or IPv6 pseudowire traffic over Equal-Cost Multipath (ECMP) paths or Link Aggregation Group (LAG) bundles. RFC 4761 requires that when the advertising PE sets the C-bit and S-bit, the receiving PE MUST insert a CW and include sequence numbers. It says nothing about a mixed environment where some PEs cannot do this, and that is the gap RFC 8614 fills.

Some routers in a network can use a special numbered envelope and some cannot. The old rules did not say what to do when they met.

§3.1Control Word (C-Bit) Behavior

A PE that sets the C-bit in its NLRI says it can send and receive frames with a CW. Under RFC 4761 the setting cannot be negotiated, and if the two ends of a pseudowire disagreed, the pseudowire did not come up.

  • If both ends set the C-bit, CWs MUST be used in both directions.
  • If both ends send a C-bit of 0, CWs MUST NOT be used.
  • If the ends disagree, CWs MUST NOT be used in either direction, but the pseudowire MUST NOT be prevented from coming up.

The last case is the change: the pseudowire now comes up without the CW.

If both want the special envelope, they use it. If only one wants it, they skip it and still talk.

§3.2Sequence Flag (S-Bit) Behavior

A PE that sets the S-bit says it can set and process sequence numbers as described in RFC 4385. If both ends set the S-bit, non-zero sequence numbers MUST be used in both directions. If both send 0, sequence numbers MUST NOT be used.

If the ends disagree, the pseudowire SHOULD NOT come up. This avoids out-of-sequence ordering when several PEs multihome a site and advertise different S-bit values. Where a deployment is known not to use multihoming, an implementation MAY offer a user-configurable override that lets the pseudowire come up. In that case the PE advertising the S-bit as 0 should set sequence numbers in its frames to 0, and the receiver should not expect non-zero sequence numbers.

Numbering the pages matters more than the envelope, because mixed-up pages can confuse things. So by default, if they disagree, they do not connect.

WHAT HAPPENS WHEN THE TWO ENDS DISAGREE
FlagRFC 4761RFC 8614
C-bit (control word) THIS RFCPseudowire does not come upComes up, CW not used in either direction
S-bit (sequencing) THIS RFCPseudowire does not come upSHOULD NOT come up; override MAY be offered

§4P2MP LSPs as Transport

BGP VPLS can run over point-to-point LSPs or over Point-to-Multipoint (P2MP) LSPs rooted at the PE that advertises the NLRI. When some PEs support the CW and others do not, a source PE that supports it should set up two P2MP LSPs: one carrying CW-marked frames to PEs that advertised the C-bit as 1, and one carrying frames without the CW to PEs that advertised 0.

The leaves of the two LSPs MUST NOT include any PE whose S-bit differs from the root PE's. Such PEs are on neither LSP. The ingress router should send separate NLRIs for the with-CW and without-CW cases.

The sender runs two delivery routes, one for receivers who want the envelope and one for those who do not. Receivers who disagree about page numbers get neither.

§5Worked Example with Four PEs

The RFC's example has PE1, PE2 and PE3 enabled for CW and sequencing, and PE4 with neither. PE1 advertises C and S bits of 1. PE2 and PE3 send the CW and non-zero sequence numbers to PE1, while PE4 cannot. Under RFC 4761 the PW between PE1 and PE4 would not come up.

Under RFC 8614, PE1 still will not bring up the PW with PE4 because of the S-bit mismatch, unless local configuration on both overrides that. If PE4 instead advertised a C-bit of 0 and an S-bit of 1, the PW would come up despite the CW mismatch. PE1 would then strip the CW only from frames received from PE2 and PE3, and expect non-zero sequence numbers from all of them, including PE4.

One router cannot use envelopes or page numbers, so it is left out. If it could at least number pages, it could join without the envelope.

§6Multihoming

In a multihomed setup, different PEs may represent the same service endpoint. For the C-bit, each PW SHOULD be evaluated independently: the PW between PE2 and PE1 could use the CW while the PE2-to-PE4 PW does not, because of a C-bit mismatch. Other multihoming behavior follows the VPLS multihoming work in progress.

For the S-bit, the S-bit rules of section 3.2 apply and evaluation SHOULD also be per PW. In the example, if PE4 does not set the S-bit, no PW forms between PE1 and PE4, so CE5 would effectively not be multihomed.

Each connection is judged on its own. One bad match can leave a customer with only one working connection instead of two.

§7Security Considerations

The security considerations of RFC 4761 apply. The RFC notes that bit-flipping media errors that cause a C-bit or S-bit mismatch do not adversely affect the availability of pseudowires under the new rules: they cause CWs not to be used, or the advertising PE not to expect non-zero sequenced frames. It says this is no worse than before, when such a mismatch kept the PW down. There are no IANA actions.

If a flag gets garbled by accident, the worst case is that the special envelope is skipped.

§8References