The BGP specification (RFC 4271) mandates a maximum BGP message size of 4,096 octets. As BGP is extended to support new Address Family Identifiers (AFIs), Subsequent AFIs (SAFIs), and other features, there is a need to extend the maximum message size beyond 4,096 octets. This document updates the BGP specification by extending the maximum message size from 4,096 octets to 65,535 octets for all messages except for OPEN and KEEPALIVE messages.

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/rfc8654.

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.

The BGP specification [RFC 4271] mandates a maximum BGP message size of 4,096 octets. As BGP is extended to support new AFIs, SAFIs, and other capabilities (e.g., BGPsec [RFC 8205] and BGP - Link State (BGP-LS) [RFC 7752]), there is a need to extend the maximum message size beyond 4,096 octets. This document provides an extension to BGP to extend the message size limit from 4,096 octets to 65,535 octets for all messages except for OPEN and KEEPALIVE messages.

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 [RFC 2119] [RFC 8174] when, and only when, they appear in all capitals, as shown here.

A BGP message over 4,096 octets in length is a BGP Extended Message. BGP Extended Messages have a maximum message size of 65,535 octets. The smallest message that may be sent is a BGP KEEPALIVE, which consists of 19 octets.

The BGP Extended Message Capability is a new BGP capability [RFC 5492] defined with Capability Code 6 and Capability Length 0. To advertise the BGP Extended Message Capability to a peer, a BGP speaker uses BGP Capabilities Advertisement [RFC 5492]. By advertising the BGP Extended Message Capability to a peer, a BGP speaker conveys that it is able to receive and properly handle BGP Extended Messages (see Section 4). Peers that wish to use the BGP Extended Message Capability MUST support error handling for BGP UPDATE messages per [RFC 7606].

The BGP Extended Message Capability applies to all messages except for OPEN and KEEPALIVE messages. These exceptions reduce the complexity of providing backward compatibility. A BGP speaker that is capable of receiving BGP Extended Messages SHOULD advertise the BGP Extended Message Capability to its peers using BGP Capabilities Advertisement [RFC 5492]. A BGP speaker MAY send BGP Extended Messages to a peer only if the BGP Extended Message Capability was received from that peer. An implementation that advertises the BGP Extended Message Capability MUST be capable of receiving a message with a length up to and including 65,535 octets. Applications generating information that might be encapsulated within BGP messages MUST limit the size of their payload to take the maximum message size into account. If a BGP message with a length greater than 4,096 octets is received by a BGP listener who has not advertised the BGP Extended Message Capability, the listener will generate a NOTIFICATION with the Error Subcode set to Bad Message Length (RFC 4271, Section 6.1). A BGP UPDATE will (if allowed by policy, best path, etc.) typically propagate throughout the BGP-speaking Internet and hence to BGP speakers that may not support BGP Extended Messages. Therefore, an announcement in a BGP Extended Message where the size of the attribute set plus the NLRI is larger than 4,096 octets may cause lack of reachability. A BGP speaker that has advertised the BGP Extended Message Capability to its peers may receive an UPDATE from one of its peers that produces an ongoing announcement that is larger than 4,096 octets. When propagating that UPDATE onward to a neighbor that has not advertised the BGP Extended Message Capability, the speaker SHOULD try to reduce the outgoing message size by removing attributes eligible under the "attribute discard" approach of [RFC 7606]. If the message is still too big, then it must not be sent to the neighbor (RFC 4271, Section 9.2). Additionally, if the NLRI was previously advertised to that peer, it must be withdrawn from service (RFC 4271, Section 9.1.3). If an Autonomous System (AS) has multiple internal BGP speakers and also has multiple external BGP neighbors, care must be taken to ensure a consistent view within the AS in order to present a consistent external view. In the context of BGP Extended Messages, a consistent view can only be guaranteed if all the Internal BGP (iBGP) speakers advertise the BGP Extended Message Capability. If that is not the case, then the operator should consider whether or not the BGP Extended Message Capability should be advertised to external peers. During the incremental deployment of BGP Extended Messages and use of the "attribute discard" approach of [RFC 7606] in an iBGP mesh or with External BGP (eBGP) peers, the operator should monitor any routes dropped and any discarded attributes.

A BGP speaker that has the ability to use BGP Extended Messages but has not advertised the BGP Extended Message Capability, presumably due to configuration, MUST NOT accept a BGP Extended Message. A speaker MUST NOT implement a more liberal policy accepting BGP Extended Messages. A BGP speaker that does not advertise the BGP Extended Message Capability might also genuinely not support BGP Extended Messages. Such a speaker will follow the error-handling procedures of [RFC 4271] if it receives a BGP Extended Message. Similarly, any speaker that treats an improper BGP Extended Message as a fatal error MUST follow the error-handling procedures of [RFC 4271]. Error handling for UPDATE messages, as specified in Section 6.3 of RFC 4271, is unchanged. However, if a NOTIFICATION is to be sent to a BGP speaker that has not advertised the BGP Extended Message Capability, the size of the message MUST NOT exceed 4,096 octets. It is RECOMMENDED that BGP protocol developers and implementers are conservative in their application and use of BGP Extended Messages. Future protocol specifications MUST describe how to handle peers that can only accommodate 4,096 octet messages.

[RFC 4271] states "The value of the Length field MUST always be at least 19 and no greater than 4096." This document changes the latter number to 65,535 for all messages except for OPEN and KEEPALIVE messages. Section 6.1 of RFC 4271 specifies raising an error if the length of a message is over 4,096 octets. For all messages except for OPEN and KEEPALIVE messages, if the receiver has advertised the BGP Extended Message Capability, this document raises that limit to 65,535.

IANA has made the following allocation in the "Capability Codes" registry: Table 1: Addition to "Capability Codes" Registry

This extension to BGP does not change BGP's underlying security issues [RFC 4272]. Due to increased memory requirements for buffering, there may be increased exposure to resource exhaustion, intentional or unintentional. If a remote speaker is able to craft a large BGP Extended Message to send on a path where one or more peers do not support BGP Extended Messages, peers that support BGP Extended Messages may:

• act to reduce the outgoing message (see Section 4) and, in doing so, cause an attack by discarding attributes one or more of its peers may be expecting. The attributes eligible under the "attribute discard" approach must have no effect on route selection or installation [RFC 7606].
• act to reduce the outgoing message (see Section 4) and, in doing so, allow a downgrade attack. This would only affect the attacker's message, where 'downgrade' has questionable meaning.
• incur resource load (processing, message resizing, etc.) when reformatting the large messages.

[RFC2119]

S. Bradner, "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>.

[RFC4271]

Y. Rekhter, T. Li, and S. Hares, "A Border Gateway Protocol 4 (BGP-4)", RFC 4271, DOI 10.17487/RFC4271, January 2006,
<https://www.rfc-editor.org/info/rfc4271>.

[RFC5492]

J. Scudder, and R. Chandra, "Capabilities Advertisement with BGP-4", RFC 5492, DOI 10.17487/RFC5492, February 2009,
<https://www.rfc-editor.org/info/rfc5492>.

[RFC7606]

E. Chen, J. Scudder, P. Mohapatra, and K. Patel, "Revised Error Handling for BGP UPDATE Messages", RFC 7606, DOI 10.17487/RFC7606, August 2015,
<https://www.rfc-editor.org/info/rfc7606>.

[RFC8174]

B. Leiba, "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>.

[RFC4272]

S. Murphy, "BGP Security Vulnerabilities Analysis", RFC 4272, DOI 10.17487/RFC4272, January 2006,
<https://www.rfc-editor.org/info/rfc4272>.

[RFC7752]

H. Gredler, J. Medved, S. Previdi, A. Farrel, and S. Ray, "North-Bound Distribution of Link-State and Traffic Engineering (TE) Information Using BGP", RFC 7752, DOI 10.17487/RFC7752, March 2016,
<https://www.rfc-editor.org/info/rfc7752>.

[RFC8205]

M. Lepinski, and K. Sriram, "BGPsec Protocol Specification", RFC 8205, DOI 10.17487/RFC8205, September 2017,
<https://www.rfc-editor.org/info/rfc8205>.

The authors thank Alvaro Retana for an amazing review; Enke Chen, Susan Hares, John Scudder, John Levine, and Job Snijders for their input; and Oliver Borchert and Kyehwan Lee for their implementations and testing.

Randy Bush

Keyur Patel

Dave Ward