Use of Ethernet Control Word RECOMMENDED
draft-ietf-pals-ethernet-cw-07

Abstract

The pseudowire (PW) encapsulation of Ethernet, as defined in RFC 4448, specifies that the use of the control word (CW) is optional. In the absence of the CW an Ethernet pseudowire packet can be misidentified as an IP packet by a label switching router (LSR). This in turn may lead to the selection of the wrong equal-cost-multi-path (ECMP) path for the packet, leading in turn to the misordering of packets. This problem has become more serious due to the deployment of equipment with Ethernet MAC addresses that start with 0x4 or 0x6. The use of the Ethernet PW CW addresses this problem. This document recommends the use of the Ethernet pseudowire control word in all but exceptional circumstances.

This document updates RFC 4448.

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."

This Internet-Draft will expire on January 3, 2019.

Copyright Notice

Copyright (c) 2018 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. Specification of Requirements
• 3. Background
• 4. Recommendation
• 5. Equal Cost Multi-path (ECMP)
• 6. Mitigations
• 7. Operational Considerations
• 8. Security Considerations
• 9. IANA Considerations
• 10. Acknowledgments
• 11. References
• 11.1. Normative References
• 11.2. Informative References
• Authors' Addresses

1. Introduction

The pseudowire(PW) encapsulation of Ethernet, as defined in [RFC4448], specifies that the use of the control word (CW) is optional. It is common for label switching routers (LSRs) to search past the end of the label stack to determine whether the payload is an IP packet, and if the payload is an IP packet, to select the next hop based on the so called “five-tuple” (IP source address, IP destination address, protocol/next-header, transport layer source port and transport layer destination port). In the absence of a PW CW an Ethernet pseudowire packet can be misidentified as an IP packet by a label switching router (LSR) selecting the equal-cost-multi-path (ECMP) path based on the five-tuple. This in turn may lead to the selection of the wrong ECMP path for the packet, leading in turn to the misordering of packets. Further discussion of this topic is published in [RFC4928].

Flow misordering can also happen in a single path scenario when traffic classification and differential forwarding treatment mechanisms are in use. These errors occur when a forwarder incorrectly assumes that the packet is IP and applies forwarding policy based on fields in the PW payload.

IPv4 and IPv6 packets respectively start with the values 0x4 and 0x6. Misidentification can arise if an Ethernet PW packet without a CW is carrying an Ethernet packet with a destination address that starts either of these values.

This problem has recently become more serious for a number of reasons. Firstly, due to the deployment of equipment with Ethernet MAC addresses that start with 0x4 or 0x6 assigned by the IEEE Registration Authority Committee (RAC). Secondly, concerns over privacy have led to the use of MAC address randomization which assigns local MAC addresses randomly for privacy. Random assignment results in addresses starting with one of these two values one time in eight.

The use of the Ethernet PW CW addresses this problem.

This document recommends the use of the Ethernet pseudowire control word in all but exceptional circumstances.

2. Specification of Requirements

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.

3. Background

Ethernet pseudowire encapsulation is specified in [RFC4448]. In particular the reader is drawn to section 4.6, part of which is quoted below for the convenience of the reader:

    "The control word defined in this section is based on the Generic 
    PW MPLS Control Word as defined in [RFC4385].  It provides the 
    ability to sequence individual frames on the PW, avoidance of 
    equal-cost multiple-path load-balancing (ECMP) [RFC2992], and 
    Operations and Management (OAM) mechanisms including VCCV 
    [RFC5085].

    "[RFC4385] states, "If a PW is sensitive to packet misordering 
    and is being carried over an MPLS PSN that uses the contents 
    of the MPLS payload to select the ECMP path, it MUST employ a 
    mechanism which prevents packet misordering." This is necessary
    because ECMP implementations may examine the first nibble after
    the MPLS label stack to determine whether the labeled packet
    is IP or not.  Thus, if the source MAC address of an Ethernet
    frame carried over the PW without a control word present begins
    with 0x4 or 0x6, it could be mistaken for an IPv4 or IPv6
    packet.  This could, depending on the configuration and
    topology of the MPLS network, lead to a situation where all
    packets for a given PW do not follow the same path.  This may
    increase out-of-order frames on a given PW, or cause OAM packets
    to follow a different path than actual traffic (see
    Section 4.4.3, "Frame Ordering").

    "The features that the control word provides may not be needed 
    for a given Ethernet PW.  For example, ECMP may not be present 
    or active on a given MPLS network, strict frame sequencing may 
    not be required, etc.  If this is the case, the control word 
    provides little value and is therefore optional.  Early Ethernet 
    PW implementations have been deployed that do not include a 
    control word or the ability to process one if present.  To 
    aid in backwards compatibility, future implementations MUST 
    be able to send and receive frames without the control word 
    present."

At the time when pseudowires were first deployed, some equipment of commercial significance was unable to process the Ethernet Control Word. In addition, at that time it was considered that no Ethernet MAC address had been issued by the IEEE Registration Authority Committee (RAC) that starts with 0x4 or 0x6, and thus it was thought to be safe to deploy Ethernet PWs without the CW.

Since that time the RAC has issued Ethernet MAC addresses start with 0x4 or 0x6 and thus the assumption that in practical networks there would be no confusion between an Ethernet PW packet without the CW and an IP packet is no longer correct.

Possibly through the use of unauthorized Ethernet MAC addresses, this assumption has been unsafe for a while, leading some equipment vendors to implement more complex, proprietary, methods to discriminate between Ethernet PW packets and IP packets. Such mechanisms rely on the heuristics of examining the transit packets in trying to find out the exact payload type of the packet and cannot be reliable due to the random nature of the payload carried within such packets.

A posting on the NANOG email list highlighted this problem:

https://mailman.nanog.org/pipermail/nanog/2016-December/089395.html

RFC EDITOR Please delete this paragraph.
Kramdown does not include references when they are only found in literal text so I include them here: [RFC4385] [RFC2992] [RFC5085] as a fixup.

4. Recommendation

The ambiguity between an MPLS payload that is an Ethernet PW and one that is an IP packet is resolved when the Ethernet PW control word is used. This document updates [RFC4448] to state that where both the ingress PE and the egress PE support the Ethernet pseudowire control word, then the CW MUST be used.

Where the application of ECMP to an Ethernet PW traffic is required, and where both the ingress and the egress PEs support [RFC6790] (Entropy Label Indicator/Entropy Label (ELI/EL)) or both the ingress and the egress PEs support [RFC6391] (FAT PW), then either method may be used. The use of both methods on the same PW is not normally necessary and should be avoided unless circumstances require it. In the case of multi-segment PWs, if ELI/EL is used then it SHOULD be used on every segment of the PW. The method by which usage of ELI/EL on every segment is guaranteed is out of scope of this document.

5. Equal Cost Multi-path (ECMP)

Where the volume of traffic on an Ethernet PW is such that ECMP is required then one of two methods may be used:

o Flow-Aware Transport (FAT) of Pseudowires over an MPLS Packet Switched Network specified in [RFC6391], or

o LSP entropy labels specified in [RFC6790]

RFC6391 works by increasing the entropy of the bottom of stack label. It requires that both the ingress and egress provider edge (PE)s support this feature. It also requires that sufficient LSRs on the LSP between the ingress and egress PE be able to select an
ECMP path on an MPLS packet with the resultant stack depth.

RFC6790 works by including an entropy value in the LSP part of the label stack. This requires that the Ingress and Egress PEs support the insertion and removal of the EL and the entropy label indicator, and that sufficient LSRs on the LSP are able to preform ECMP based on the EL.

In both cases there are considerations in getting Operations, Administration, and Maintenance (OAM) packets to follow the same path as a data packet. This is described in detail section 7 of [RFC6391], and section 6 of RFC6790. However in both cases the situation is improved compared to the ECMP behavior in the case where the Ethernet PW CW was not used, since there is currently no known method of getting a PW OAM packet to follow the same path as a PW data packet subjected to ECMP based on the five tuple of the IP payload.

The PW label is pushed before the LSP label. As the EL/ELI labels are part of the LSP layer rather than part of the PW layer, they are pushed after the PW label has been pushed.

6. Mitigations

Where it is not possible to use the Ethernet PW CW, the effects of ECMP can be disabled by carrying the PW over a traffic engineered path that does not subject the payload to load balancing (for example [RFC3209]). However such paths may be subjected to link bundle load balancing and of course the single LSP has to carry the full PW load.

7. Operational Considerations

In some cases, the inclusion of a CW in the PW is determined by equipment configuration. Furthermore, it is possible that the default configuration in such cases is to disable use of the CW. Care needs to be taken to ensure that software that implements this recommendation does not depend on existing configuration settings that prevents the use of control word. It is recommended that platform software emits a rate limited message indicating that CW can be used but is disabled due to existing configuration.

Instead of including a payload type in the packet, MPLS relies on the control plane to signal the payload type that follows the bottom of the label stack. Some LSRs attempt to deduce the packet type by MPLS payload inspection, in some cases looking past the PW CW. If the payload appears to be IP or IP carried in an Ethernet header they perform an ECMP calculation based on what they assume to be the five tuple fields. However deduction of the payload type in this way is not an exact science, and where a packet that is not IP is mistaken for an IP packet the result can be packets delivered out of order. Misordering of this type can be difficult for an operator to diagnose. Operators should be aware when enabling capability that allows information gleaned from packet inspection past the PW CW to be used in any ECMP calculation, that this may cause Ethernet frames to be delivered out of order despite the presence of the CW.

8. Security Considerations

This document expresses a preference for one existing and widely deployed Ethernet PW encapsulation over another. These methods have identical security considerations, which are discussed in [RFC4448]. This document introduces no additional security issues.

9. IANA Considerations

This document makes no IANA requests.

10. Acknowledgments

The authors thank Job Snijders for drawing attention to this problem. The authors also thank Pat Thaler for clarifying the matter of local MAC address assignment. We thank Sasha Vainshtein for his valuable review comments.

11. References

11.1. Normative References

11.2. Informative References

Authors' Addresses