Bidirectional Forwarding Detection (BFD) for Multihop Paths
How to run BFD liveness checks across several routed hops, where the single-hop tricks no longer work.
Katz, D. and D. Ward, "Bidirectional Forwarding Detection (BFD) for Multihop Paths", RFC 5883, DOI 10.17487/RFC5883, June 2010, <https://www.rfc-editor.org/info/rfc5883>.
@misc{rfc5883,
series = {Request for Comments},
number = 5883,
howpublished = {RFC 5883},
publisher = {RFC Editor},
doi = {10.17487/RFC5883},
url = {https://www.rfc-editor.org/info/rfc5883},
author = {D. Katz and D. Ward},
title = {{Bidirectional Forwarding Detection (BFD) for Multihop Paths}},
pagetotal = 6,
year = 2010,
month = jun,
} RFC text
Contents
Internet Engineering Task Force (IETF) D. Katz
Request for Comments: 5883 D. Ward
Category: Standards Track Juniper Networks
ISSN: 2070-1721 June 2010
Bidirectional Forwarding Detection (BFD) for Multihop Paths
Abstract
This document describes the use of the Bidirectional Forwarding
Detection (BFD) protocol over multihop paths, including
unidirectional links.
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/rfc5883.
Copyright Notice
Copyright (c) 2010 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.
1. Introduction
The Bidirectional Forwarding Detection (BFD) protocol [BFD] defines a
method for liveness detection of arbitrary paths between systems.
The BFD one-hop specification [BFD-1HOP] describes how to use BFD
across single hops of IPv4 and IPv6.
BFD can also be useful on arbitrary paths between systems, which may
span multiple network hops and follow unpredictable paths.
Furthermore, a pair of systems may have multiple paths between them
that may overlap. This document describes methods for using BFD in
such scenarios.
1.1. Conventions Used in This Document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [KEYWORDS].
2. Applicability
Please note that BFD is intended as an Operations, Administration,
and Maintenance (OAM) mechanism for connectivity check and connection
verification. It is applicable for network-based services (e.g.
router-to-router, subscriber-to-gateway, LSP/circuit endpoints, and
service appliance failure detection). In these scenarios it is
required that the operator correctly provision the rates at which BFD
is transmitted to avoid congestion (e.g link, I/O, CPU) and false
failure detection. It is not applicable for application-to-
application failure detection across the Internet because it does not
have sufficient capability to do necessary congestion detection and
avoidance and therefore cannot prevent congestion collapse. Host-to-
host or application-to-application deployment across the Internet
will require the encapsulation of BFD within a transport that
provides "TCP-friendly" [TFRC] behavior.
3. Issues
There are three primary issues in the use of BFD for multihop paths.
The first is security and spoofing; [BFD-1HOP] describes a
lightweight method of avoiding spoofing by requiring a Time to Live
(TTL)/Hop Limit of 255 on both transmit and receive, but this
obviously does not work across multiple hops. The utilization of BFD
authentication addresses this issue.
The second, more subtle, issue is that of demultiplexing multiple BFD
sessions between the same pair of systems to the proper BFD session.
In particular, the first BFD packet received for a session may carry
a Your Discriminator value of zero, resulting in ambiguity as to
which session the packet should be associated. Once the
discriminator values have been exchanged, all further packets are
demultiplexed to the proper BFD session solely by the contents of the
Your Discriminator field.
[BFD-1HOP] addresses this by requiring that multiple sessions
traverse independent physical or logical links -- the first packet is
demultiplexed based on the link over which it was received. In the
more general case, this scheme cannot work, as two paths over which
BFD is running may overlap to an arbitrary degree (including the
first and/or last hop).
Finally, the Echo function MUST NOT be used over multiple hops.
Intermediate hops would route the packets back to the sender, and
connectivity through the entire path would not be possible to verify.
4. Demultiplexing Packets
There are a number of possibilities for addressing the demultiplexing
issue that may be used, depending on the application.
4.1. Totally Arbitrary Paths
It may be desired to use BFD for liveness detection over paths for
which no part of the route is known (or if known, may not be stable).
A straightforward approach to this problem is to limit BFD deployment
to a single session between a source/destination address pair.
Multiple sessions between the same pair of systems must have at least
one endpoint address distinct from one another.
In this scenario, the initial packet is demultiplexed to the
appropriate BFD session based on the source/destination address pair
when Your Discriminator is set to zero.
This approach is appropriate for general connectivity detection
between systems over routed paths and is also useful for OSPF Virtual
Links [OSPFv2] [OSPFv3].
4.2. Out-of-Band Discriminator Signaling
Another approach to the demultiplexing problem is to signal the
discriminator values in each direction through an out-of-band
mechanism prior to establishing the BFD session. Once learned, the
discriminators are sent as usual in the BFD Control packets; no
packets with Your Discriminator set to zero are ever sent. This
method is used by the BFD MPLS specification [BFD-MPLS].
This approach is advantageous because it allows BFD to be directed by
other system components that have knowledge of the paths in use, and
from the perspective of BFD implementation it is very simple.
The disadvantage is that it requires at least some level of BFD-
specific knowledge in parts of the system outside of BFD.
4.3. Unidirectional Links
Unidirectional links are classified as multihop paths because the
return path (which should exist at some level in order to make the
link useful) may be arbitrary, and the return paths for BFD sessions
protecting parallel unidirectional links may overlap or even be
identical. (If two unidirectional links, one in each direction, are
to carry a single BFD session, this can be done using the single-hop
approach.)
Either of the two methods outlined earlier may be used in the
unidirectional link case, but a more general solution can be found
strictly within BFD and without addressing limitations.
The approach is similar to the one-hop specification, since the
unidirectional link is a single hop. Let's define the two systems as
the Unidirectional Sender and the Unidirectional Receiver. In this
approach, the Unidirectional Sender MUST operate in the Active role
(as defined in the base BFD specification), and the Unidirectional
Receiver MUST operate in the Passive role.
In the Passive role, by definition, the Unidirectional Receiver does
not transmit any BFD Control packets until it learns the
discriminator value in use by the other system (upon receipt of the
first BFD Control packet). The Unidirectional Receiver demultiplexes
the first packet to the proper BFD session based on the physical or
logical link over which it was received. This allows the receiver to
learn the remote discriminator value, which it then echoes back to
the sender in its own (arbitrarily routed) BFD Control packet, after
which time all packets are demultiplexed solely by discriminator.
5. Encapsulation
The encapsulation of BFD Control packets for multihop application in
IPv4 and IPv6 is identical to that defined in [BFD-1HOP], except that
the UDP destination port MUST have a value of 4784. This can aid in
the demultiplexing and internal routing of incoming BFD packets.
6. Authentication
By their nature, multihop paths expose BFD to spoofing. As the
number of hops increases, the exposure to attack grows. As such,
implementations of BFD SHOULD utilize cryptographic authentication
over multihop paths to help mitigate denial-of-service attacks.
7. IANA Considerations
Port 4784 has been assigned by IANA for use with BFD Multihop
Control.
8. Security Considerations
As the number of hops increases, BFD becomes further exposed to
attack. The use of strong forms of authentication is strongly
encouraged.
No additional security issues are raised in this document beyond
those that exist in the referenced BFD documents.
9. References
9.1. Normative References
[BFD] Katz, D. and D. Ward, "Bidirectional Forwarding
Detection", RFC 5880, June 2010.
[BFD-1HOP] Katz, D. and D. Ward, "Bidirectional Forwarding Detection
(BFD) for IPv4 and IPv6 (Single Hop)", RFC 5881, June
2010.
[KEYWORDS] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
9.2. Informative References
[BFD-MPLS] Aggarwal, R., Kompella, K., Nadeau, T., and G. Swallow,
"Bidirectional Forwarding Detection (BFD) for MPLS Label
Switched Paths (LSPs)", RFC 5884, June 2010.
[OSPFv2] Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998.
[OSPFv3] Coltun, R., Ferguson, D., Moy, J., and A. Lindem, "OSPF
for IPv6", RFC 5340, July 2008.
[TFRC] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP
Friendly Rate Control (TFRC): Protocol Specification", RFC
5348, September 2008.
Authors' Addresses
Dave Katz
Juniper Networks
1194 N. Mathilda Ave.
Sunnyvale, CA 94089-1206
USA
Phone: +1-408-745-2000
EMail: dkatz@juniper.net
Dave Ward
Juniper Networks
1194 N. Mathilda Ave.
Sunnyvale, CA 94089-1206
USA
Phone: +1-408-745-2000
EMail: dward@juniper.net
AI ELI5
On this page
§1BFD Beyond a Single Hop
The Bidirectional Forwarding Detection (BFD) protocol, defined in RFC 5880, detects liveness on arbitrary paths between systems. The one-hop specification, RFC 5881, covers single hops of IPv4 and IPv6. RFC 5883 covers paths that span multiple network hops and follow unpredictable routes. A pair of systems may also have several paths that overlap.
BFD is an Operations, Administration, and Maintenance (OAM) mechanism for connectivity checks and connection verification. It suits network-based services such as router-to-router, subscriber-to-gateway, LSP or circuit endpoints, and service appliance failure detection. Operators must provision the BFD transmit rates correctly to avoid congestion and false failure detection.
It is not applicable to application-to-application failure detection across the Internet, because it cannot detect and avoid congestion and so cannot prevent congestion collapse. Such use would need BFD encapsulated in a transport with TCP-friendly behavior.
BFD is like two friends tapping a signal back and forth to show they are still there. It is meant for neighbors in the same network, not for strangers across the whole Internet.
§3Three Problems with Multihop Paths
First, spoofing. The one-hop specification avoids spoofed packets by requiring a Time to Live (TTL) or Hop Limit of 255 on send and receive. That cannot work across several hops, so BFD authentication is used instead.
Second, demultiplexing. The first packet of a session may carry a Your Discriminator of zero, so the receiver cannot tell which session it belongs to. Once discriminators are exchanged, packets are matched solely by that field. The one-hop spec solves this by requiring sessions to use independent links, so the first packet is matched by its incoming link. With multihop paths, two paths can overlap to any degree, including the first or last hop, so that scheme fails.
Third, the Echo function MUST NOT be used over multiple hops. Intermediate hops would route the packets back to the sender, so connectivity through the whole path could not be verified.
There are three snags. Anyone could fake a message, it is hard to tell which conversation a first message belongs to, and a message that bounces back early proves nothing about the far end.
§4Demultiplexing Packets
The RFC gives three approaches, to be chosen according to the application.
Totally arbitrary paths. Limit deployment to a single session per source and destination address pair. Several sessions between the same two systems must differ in at least one endpoint address. The first packet, with Your Discriminator zero, is matched by that address pair. This suits general connectivity detection over routed paths and OSPF Virtual Links.
Out-of-band discriminator signaling. The discriminators are exchanged by another mechanism before the session starts, so no packet with Your Discriminator zero is ever sent. This is used by the BFD MPLS specification. It is simple for BFD and lets other components that know the paths direct it, but it needs some BFD-specific knowledge outside BFD.
Unidirectional links. These count as multihop paths because the return path may be arbitrary and may overlap with others. The Unidirectional Sender MUST operate in the Active role and the Unidirectional Receiver MUST operate in the Passive role. The Passive receiver stays silent until the first Control packet arrives, matches it using the physical or logical link it came in on, and learns the sender's discriminator. It then echoes that value back in its own Control packet over any route, and from then on matching uses discriminators alone.
If two friends want to talk but cannot tell their chats apart, they can use different phone numbers, agree on secret codes ahead of time, or let one speak first so the other learns the code.
| Approach | How the first packet is matched | Suited to |
|---|---|---|
| Arbitrary paths | Source and destination address pair | General connectivity, OSPF Virtual Links |
| Out-of-band signaling | Discriminators learned beforehand | Components that know the paths, such as BFD for MPLS |
| Unidirectional links | Incoming link, with sender Active and receiver Passive | Unidirectional links |
- First Control packetMatched to the session by the incoming link.
- Control packet echoing the discriminatorSent over any route. After this, matching uses discriminators only.
§5Encapsulation on Port 4784
Multihop BFD Control packets are encapsulated in IPv4 and IPv6 exactly as in the one-hop specification, with one difference: the UDP destination port MUST be 4784. A separate port helps with demultiplexing and internal routing of incoming BFD packets. IANA has assigned port 4784 for BFD Multihop Control.
Multihop BFD gets its own mailbox number, 4784, so these messages do not get mixed up with the single-hop kind.
§6Authentication and Security
Multihop paths expose BFD to spoofing, and the exposure grows with the number of hops. Implementations SHOULD therefore use cryptographic authentication over multihop paths to help mitigate denial-of-service attacks. The RFC strongly encourages strong forms of authentication.
It raises no security issues beyond those in the referenced BFD documents.
The farther a message travels, the easier it is for a stranger to fake it, so each message should carry a secret signature.