]> Juniper Networks

jgs@juniper.net

Juniper Networks

kireeti@juniper.net

Cisco Systems

satyamoh@cisco.com

AT&T

ju1738@att.com

Comcast

Bin_Wen@comcast.com

General Internet Engineering Task Force BGP MPLS This specification defines the Entropy Label Capability Attribute version 3 (ELCv3), a BGP attribute that can be used to inform an LSP ingress router about an LSP egress router's ability to process entropy labels. This version of the attribute corrects a specification error in the first version, and an improper code point reuse in the second.

defines the Entropy Label Capability attribute (ELC), an optional, transitive BGP path attribute. For correct operation, it is necessary that any intermediate node modifying the next hop of a route must remove the ELC unless the node so doing is able to process entropy labels. Sadly, these requirements cannot be fulfilled with the ELC as specified, because it is an optional, transitive attribute: by definition, a node that does not support the ELC will propagate the attribute. But such a node might be exactly the one that we desire to remove it. For this reason, deprecated the attribute. Roughly concurrently with the development and advancement of RFC 7447, Juniper Networks began shipping routing code that implements what is documented in and dubbed Entropy Label Capability version 2 (ELCv2). That implementation uses the code point that was assigned by RFC 6790 and deprecated by RFC 7447. At time of writing, the functionality is in use in operational networks. The present specification is based on ELCv2 but moves to a new, previously unallocated, code point. A related solution to the problem of signaling entropy label capability is . That specification is based on an optional, non-transitive path attribute. In contrast, ELCv3 (and ELCv2) is based on an optional, transitive path attribute. This expands the deployment options available -- in many cases (for example, route reflectors) it's fine that an intermediate node does propagate an ELCv3 even if it doesn't itself have the ability to process entropy labels. In order to prevent use of the signaled information beyond the intended perimeter (the problem that led to the deprecation of ELC, and which is inherently solved by 's use of a non-transitive attribute), in this specification we take the approach of carrying a copy of the next hop information in the ELCv3. This allows the node processing it to know if it can rely on the information carried therein, while still allowing it to be propagated by all intermediate nodes.

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 when, and only when, they appear in all capitals, as shown here.

The Entropy Label Capability Path Attribute, Version 3 (ELCv3) is an optional, transitive BGP attribute with type code TBD1. The ELCv3 has as its data a network layer address, representing the next hop of the route the ELCv3 accompanies. The ELCv3 signals a useful optimization, so it is desirable to make it transitive; the next hop data is to ensure correctness if it traverses BGP speakers that do not understand the ELCv3. The Attribute Data field of the ELCv3 path attribute is encoded as shown below:

The meanings of the fields are as given in Section 3 of . When BGP is used for distributing labeled Network Layer Reachability Information (NLRI) as described in, for example, , the route may include the ELCv3 as part of the Path Attributes. The inclusion of this attribute with a route indicates that the egress of the associated Label Switched Path (LSP) can process entropy labels as an egress Label Switched Router (LSR) for that route -- see . Below, we refer to this for brevity as being "EL-capable."

When a BGP speaker S has a route R it wishes to advertise with next hop N to its peer, it MUST NOT include the ELCv3 attribute except if it knows that the egress of the associated LSP L is EL-capable. Specifically, this will be true if S: Is itself the egress, and knows itself to be EL-capable, or Is re-advertising a BGP route it received with a valid ELCv3 attribute, and is not changing the value of N, or Is re-advertising a BGP route it received with a valid ELCv3 attribute, and is changing the value of N, and knows (for example, through configuration) that the router represented by N is either the LSP egress and is EL-capable, or that it will process the outer label(s) without processing the entropy label below, as with a transit LSR, or Is redistributing a route learned from another protocol, and that other protocol conveyed the knowledge that the egress of L was EL-capable (for example, this might be known through the LDP ELC TLV, ). In any event, when sending an ELCv3, S MUST set the data portion of the ELCv3 to be equal to N, using the encoding given in . The ELCv3 MAY be advertised with routes that are labeled, such as those using SAFI 4 . It MUST NOT be advertised with unlabeled routes. We note that due to the nature of BGP optional transitive path attributes, any BGP speaker that does not implement this specification will propagate the ELCv3, the requirements of this section notwithstanding. However, such a speaker will not update the data part of the ELCv3.

When a BGP speaker receives an unlabeled route that includes the ELCv3, it MUST discard the ELCv3. When a BGP speaker receives a labeled route that includes the ELCv3, it MUST compare the address given in the ELCv3's data portion to the next hop of the route. If the two are equal, the egress of the LSP supports entropy labels, which implies that the receiving BGP speaker, if acting as ingress, MAY insert an entropy label below the advertised label, as per . If the two are not equal, either some intermediate router that does not implement this specification modified the next hop, or some router on the path had an incorrect implementation. In either case, the action taken is the same: the ELCv3 MUST be discarded. The Partial bit MAY be inspected -- if it is equal to zero, then the mismatch must have been caused by an incorrect implementation, and the error MAY be logged. When a BGP speaker receives a route that includes an ELCv3 whose Attribute Length is less than 4, whose Attribute Length is not equal to 4 plus the value encoded in the Length of Next Hop Network Address carried in the Attribute Data, or whose Attribute Data is otherwise inconsistent with the encoding specified in , it MUST discard the ELCv3.

IANA is requested to make a new allocation in the BGP Path Attributes registry: Value = TBD1 Code = ELCv3 Reference = (this document)

Insertion of an ELCv3 by an attacker could cause forwarding to fail. Deletion of an ELCv3 by an attacker could cause one path in the network to be overutilized and another to be underutilized. However, we note that an attacker able to accomplish either of these (below, an "on-path attacker") could equally insert or remove any other BGP path attribute or message. The former attack described above denies service for a given route, which can be accomplished by an on-path attacker in any number of ways even absent ELCv3. The latter attack defeats an optimization but nothing more; it seems dubious that an attacker would go to the trouble of doing so rather than launching some more damaging attack. In sum, the ELCv3 attribute creates no significant issues beyond those analyzed in .

&RFC2119; &RFC4271; &RFC4760; &RFC6790; &RFC8174; &RFC4272; &RFC7447; &RFC8277; &I-D.ietf-idr-next-hop-capability; &I-D.scudder-bgp-entropy-label;

Backward Compatibility As was noted in , there are networks in which ELCv2 (documented in ) is already in use. Any node that sends the ELCv2 format may also include an ELCv3 per , so that both formats are sent. When a BGP speaker that supports the ELCv2 format receives a labeled route that includes both ELCv2 and ELCv3 it should evaluate each attribute as per their documented use ( in the case of this document). If, after evaluation, either ELCv2 or ELCv3 is determined to be valid, the route may be considered EL-capable.

-->

Contributors Cisco Systems

sekrier@cisco.com

Thanks to Alia Atlas, Bruno Decraene, Martin Djernaes, John Drake, Adrian Farrell, Keyur Patel, Toby Rees, and Ravi Singh, for their discussion of this issue. Particular thanks to Kevin Wang for his many valuable contributions.