idnits 2.17.1 draft-ietf-bess-evpn-irb-mcast-05.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 13, 2020) is 1283 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-09) exists of draft-ietf-bess-evpn-df-election-framework-07 -- Possible downref: Non-RFC (?) normative reference: ref. 'EVPN-AR' -- Possible downref: Non-RFC (?) normative reference: ref. 'EVPN-BUM' == Outdated reference: A later version (-15) exists of draft-ietf-bess-evpn-inter-subnet-forwarding-05 == Outdated reference: A later version (-21) exists of draft-ietf-bess-evpn-igmp-mld-proxy-02 Summary: 0 errors (**), 0 flaws (~~), 4 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 BESS W. Lin 3 Internet-Draft Z. Zhang 4 Intended status: Standards Track J. Drake 5 Expires: April 16, 2021 E. Rosen, Ed. 6 Juniper Networks, Inc. 7 J. Rabadan 8 Nokia 9 A. Sajassi 10 Cisco Systems 11 October 13, 2020 13 EVPN Optimized Inter-Subnet Multicast (OISM) Forwarding 14 draft-ietf-bess-evpn-irb-mcast-05 16 Abstract 18 Ethernet VPN (EVPN) provides a service that allows a single Local 19 Area Network (LAN), comprising a single IP subnet, to be divided into 20 multiple "segments". Each segment may be located at a different 21 site, and the segments are interconnected by an IP or MPLS backbone. 22 Intra-subnet traffic (either unicast or multicast) always appears to 23 the endusers to be bridged, even when it is actually carried over the 24 IP or MPLS backbone. When a single "tenant" owns multiple such LANs, 25 EVPN also allows IP unicast traffic to be routed between those LANs. 26 This document specifies new procedures that allow inter-subnet IP 27 multicast traffic to be routed among the LANs of a given tenant, 28 while still making intra-subnet IP multicast traffic appear to be 29 bridged. These procedures can provide optimal routing of the inter- 30 subnet multicast traffic, and do not require any such traffic to 31 leave a given router and then reenter that same router. These 32 procedures also accommodate IP multicast traffic that needs to travel 33 to or from systems that are outside the EVPN domain. 35 Status of This Memo 37 This Internet-Draft is submitted in full conformance with the 38 provisions of BCP 78 and BCP 79. 40 Internet-Drafts are working documents of the Internet Engineering 41 Task Force (IETF). Note that other groups may also distribute 42 working documents as Internet-Drafts. The list of current Internet- 43 Drafts is at https://datatracker.ietf.org/drafts/current/. 45 Internet-Drafts are draft documents valid for a maximum of six months 46 and may be updated, replaced, or obsoleted by other documents at any 47 time. It is inappropriate to use Internet-Drafts as reference 48 material or to cite them other than as "work in progress." 49 This Internet-Draft will expire on April 16, 2021. 51 Copyright Notice 53 Copyright (c) 2020 IETF Trust and the persons identified as the 54 document authors. All rights reserved. 56 This document is subject to BCP 78 and the IETF Trust's Legal 57 Provisions Relating to IETF Documents 58 (https://trustee.ietf.org/license-info) in effect on the date of 59 publication of this document. Please review these documents 60 carefully, as they describe your rights and restrictions with respect 61 to this document. Code Components extracted from this document must 62 include Simplified BSD License text as described in Section 4.e of 63 the Trust Legal Provisions and are provided without warranty as 64 described in the Simplified BSD License. 66 Table of Contents 68 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4 69 1.1. Background . . . . . . . . . . . . . . . . . . . . . . . 4 70 1.1.1. Segments, Broadcast Domains, and Tenants . . . . . . 4 71 1.1.2. Inter-BD (Inter-Subnet) IP Traffic . . . . . . . . . 5 72 1.1.3. EVPN and IP Multicast . . . . . . . . . . . . . . . . 6 73 1.1.4. BDs, MAC-VRFS, and EVPN Service Models . . . . . . . 7 74 1.2. Need for EVPN-aware Multicast Procedures . . . . . . . . 7 75 1.3. Additional Requirements That Must be Met by the Solution 8 76 1.4. Terminology . . . . . . . . . . . . . . . . . . . . . . . 10 77 1.5. Model of Operation: Overview . . . . . . . . . . . . . . 12 78 1.5.1. Control Plane . . . . . . . . . . . . . . . . . . . . 12 79 1.5.2. Data Plane . . . . . . . . . . . . . . . . . . . . . 14 80 2. Detailed Model of Operation . . . . . . . . . . . . . . . . . 17 81 2.1. Supplementary Broadcast Domain . . . . . . . . . . . . . 17 82 2.2. Detecting When a Route is About/For/From a Particular BD 18 83 2.3. Use of IRB Interfaces at Ingress PE . . . . . . . . . . . 21 84 2.4. Use of IRB Interfaces at an Egress PE . . . . . . . . . . 23 85 2.5. Announcing Interest in (S,G) . . . . . . . . . . . . . . 23 86 2.6. Tunneling Frames from Ingress PE to Egress PEs . . . . . 24 87 2.7. Advanced Scenarios . . . . . . . . . . . . . . . . . . . 25 88 3. EVPN-aware Multicast Solution Control Plane . . . . . . . . . 26 89 3.1. Supplementary Broadcast Domain (SBD) and Route Targets . 26 90 3.2. Advertising the Tunnels Used for IP Multicast . . . . . . 27 91 3.2.1. Constructing Routes for the SBD . . . . . . . . . . . 28 92 3.2.2. Ingress Replication . . . . . . . . . . . . . . . . . 28 93 3.2.3. Assisted Replication . . . . . . . . . . . . . . . . 29 94 3.2.3.1. Automatic SBD Matching . . . . . . . . . . . . . 30 95 3.2.4. BIER . . . . . . . . . . . . . . . . . . . . . . . . 30 96 3.2.5. Inclusive P2MP Tunnels . . . . . . . . . . . . . . . 31 97 3.2.5.1. Using the BUM Tunnels as IP Multicast Inclusive 98 Tunnels . . . . . . . . . . . . . . . . . . . . . 31 99 3.2.5.2. Using Wildcard S-PMSI A-D Routes to Advertise 100 Inclusive Tunnels Specific to IP Multicast . . . 33 101 3.2.6. Selective Tunnels . . . . . . . . . . . . . . . . . . 33 102 3.3. Advertising SMET Routes . . . . . . . . . . . . . . . . . 34 103 4. Constructing Multicast Forwarding State . . . . . . . . . . . 37 104 4.1. Layer 2 Multicast State . . . . . . . . . . . . . . . . . 37 105 4.1.1. Constructing the OIF List . . . . . . . . . . . . . . 38 106 4.1.2. Data Plane: Applying the OIF List to an (S,G) Frame . 38 107 4.1.2.1. Eligibility of an AC to Receive a Frame . . . . . 39 108 4.1.2.2. Applying the OIF List . . . . . . . . . . . . . . 39 109 4.2. Layer 3 Forwarding State . . . . . . . . . . . . . . . . 40 110 5. Interworking with non-OISM EVPN-PEs . . . . . . . . . . . . . 41 111 5.1. IPMG Designated Forwarder . . . . . . . . . . . . . . . . 44 112 5.2. Ingress Replication . . . . . . . . . . . . . . . . . . . 44 113 5.2.1. Ingress PE is non-OISM . . . . . . . . . . . . . . . 46 114 5.2.2. Ingress PE is OISM . . . . . . . . . . . . . . . . . 47 115 5.3. P2MP Tunnels . . . . . . . . . . . . . . . . . . . . . . 48 116 6. Traffic to/from Outside the EVPN Tenant Domain . . . . . . . 48 117 6.1. Layer 3 Interworking via EVPN OISM PEs . . . . . . . . . 49 118 6.1.1. General Principles . . . . . . . . . . . . . . . . . 49 119 6.1.2. Interworking with MVPN . . . . . . . . . . . . . . . 52 120 6.1.2.1. MVPN Sources with EVPN Receivers . . . . . . . . 54 121 6.1.2.1.1. Identifying MVPN Sources . . . . . . . . . . 54 122 6.1.2.1.2. Joining a Flow from an MVPN Source . . . . . 54 123 6.1.2.2. EVPN Sources with MVPN Receivers . . . . . . . . 56 124 6.1.2.2.1. General procedures . . . . . . . . . . . . . 56 125 6.1.2.2.2. Any-Source Multicast (ASM) Groups . . . . . . 58 126 6.1.2.2.3. Source on Multihomed Segment . . . . . . . . 58 127 6.1.2.3. Obtaining Optimal Routing of Traffic Between MVPN 128 and EVPN . . . . . . . . . . . . . . . . . . . . 59 129 6.1.2.4. Selecting the MEG SBD-DR . . . . . . . . . . . . 60 130 6.1.3. Interworking with 'Global Table Multicast' . . . . . 60 131 6.1.4. Interworking with PIM . . . . . . . . . . . . . . . . 61 132 6.1.4.1. Source Inside EVPN Domain . . . . . . . . . . . . 61 133 6.1.4.2. Source Outside EVPN Domain . . . . . . . . . . . 63 134 6.2. Interworking with PIM via an External PIM Router . . . . 63 135 7. Using an EVPN Tenant Domain as an Intermediate (Transit) 136 Network for Multicast traffic . . . . . . . . . . . . . . . . 64 137 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 67 138 9. Security Considerations . . . . . . . . . . . . . . . . . . . 67 139 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 67 140 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 67 141 11.1. Normative References . . . . . . . . . . . . . . . . . . 68 142 11.2. Informative References . . . . . . . . . . . . . . . . . 69 143 Appendix A. Integrated Routing and Bridging . . . . . . . . . . 71 144 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 76 146 1. Introduction 148 1.1. Background 150 Ethernet VPN (EVPN) [RFC7432] provides a Layer 2 VPN (L2VPN) 151 solution, which allows IP backbone provider to offer ethernet service 152 to a set of customers, known as "tenants". 154 In this section (as well as in [EVPN-IRB]), we provide some essential 155 background information on EVPN. 157 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 158 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 159 "OPTIONAL" in this document are to be interpreted as described in BCP 160 14 [RFC2119] [RFC8174] when, and only when, they appear in all 161 capitals, as shown here. 163 1.1.1. Segments, Broadcast Domains, and Tenants 165 One of the key concepts of EVPN is the Broadcast Domain (BD). A BD 166 is essentially an emulated ethernet. Each BD belongs to a single 167 tenant. A BD typically consists of multiple ethernet "segments", and 168 each segment may be attached to a different EVPN Provider Edge 169 (EVPN-PE) router. EVPN-PE routers are often referred to as "Network 170 Virtualization Endpoints" or NVEs. However, this document will use 171 the term "EVPN-PE", or, when the context is clear, just "PE". 173 In this document, we use the term "segment" to mean the same as 174 "Ethernet Segment" or "ES" in [RFC7432]. 176 Attached to each segment are "Tenant Systems" (TSes). A TS may be 177 any type of system, physical or virtual, host or router, etc., that 178 can attach to an ethernet. 180 When two TSes are on the same segment, traffic between them does not 181 pass through an EVPN-PE. When two TSes are on different segments of 182 the same BD, traffic between them does pass through an EVPN-PE. 184 When two TSes, say TS1 and TS2 are on the same BD, then: 186 o If TS1 knows the MAC address of TS2, TS1 can send unicast ethernet 187 frames to TS2. TS2 will receive the frames unaltered. 189 o If TS1 broadcasts an ethernet frame, TS2 will receive the 190 unaltered frame. 192 o If TS1 multicasts an ethernet frame, TS2 will receive the 193 unaltered frame, as long as TS2 has been provisioned to receive 194 ethernet multicasts. 196 When we say that TS2 receives an unaltered frame from TS1, we mean 197 that the frame still contains TS1's MAC address, and that no 198 alteration of the frame's payload (and consequently, no alteration of 199 the payload's IP header) has been made. 201 EVPN allows a single segment to be attached to multiple PE routers. 202 This is known as "EVPN multi-homing". Suppose a given segment is 203 attached to both PE1 and PE2, and suppose PE1 receives a frame from 204 that segment. It may be necessary for PE1 to send the frame over the 205 backbone to PE2. EVPN has procedures to ensure that such a frame 206 cannot be sent by PE2 back to its originating segment. This is 207 particularly important for multicast, because a frame arriving at PE1 208 from a given segment will already have been seen by all the systems 209 on that segment that need to see it. If the frame were sent back to 210 the originating segment by PE2, receivers on that segment would 211 receive the packet twice. Even worse, the frame might be sent back 212 to PE1, which could cause an infinite loop. 214 1.1.2. Inter-BD (Inter-Subnet) IP Traffic 216 If a given tenant has multiple BDs, the tenant may wish to allow IP 217 communication among these BDs. Such a set of BDs is known as an 218 "EVPN Tenant Domain" or just a "Tenant Domain". 220 If tenant systems TS1 and TS2 are not in the same BD, then they do 221 not receive unaltered ethernet frames from each other. In order for 222 TS1 to send traffic to TS2, TS1 encapsulates an IP datagram inside an 223 ethernet frame, and uses ethernet to send these frames to an IP 224 router. The router decapsulates the IP datagram, does the IP 225 processing, and re-encapsulates the datagram for ethernet. The MAC 226 source address field now has the MAC address of the router, not of 227 TS1. The TTL field of the IP datagram should be decremented by 228 exactly 1, even if the frame needs to be sent from one PE to another. 229 The structure of the provider's IP backbone is thus hidden from the 230 tenants. 232 EVPN accommodates the need for inter-BD communication within a Tenant 233 Domain by providing an integrated L2/L3 service for unicast IP 234 traffic. EVPN's Integrated Routing and Bridging (IRB) functionality 235 is specified in [EVPN-IRB]. Each BD in a Tenant Domain is assumed to 236 be a single IP subnet, and each IP subnet within a a given Tenant 237 Domain is assumed to be a single BD. EVPN's IRB functionality allows 238 IP traffic to travel from one BD to another, and ensures that proper 239 IP processing (e.g., TTL decrement) is done. 241 A brief overview of IRB, including the notion of an "IRB interface", 242 can be found in Appendix A. As explained there, an IRB interface is 243 a sort of virtual interface connecting an L3 routing instance to a 244 BD. A BD may have multiple attachment circuits (ACs) to a given PE, 245 where each AC connects to a different ethernet segment of the BD. 246 However, these ACs are not visible to the L3 routing function; from 247 the perspective of an L3 routing instance, a PE has just one 248 interface to each BD, viz., the IRB interface for that BD. 250 The "L3 routing instance" depicted in Appendix A is associated with a 251 single Tenant Domain, and may be thought of as an IP-VRF for that 252 Tenant Domain. 254 1.1.3. EVPN and IP Multicast 256 [EVPN-IRB] and [EVPN_IP_Prefix] cover inter-subnet (inter-BD) IP 257 unicast forwarding, but they do not cover inter-subnet IP multicast 258 forwarding. 260 [RFC7432] covers intra-subnet (intra-BD) ethernet multicast. The 261 intra-subnet ethernet multicast procedures of [RFC7432] are used for 262 ethernet Broadcast traffic, for ethernet unicast traffic whose MAC 263 Destination Address field contains an Unknown address, and for 264 ethernet traffic whose MAC Destination Address field contains an 265 ethernet Multicast MAC address. These three classes of traffic are 266 known collectively as "BUM traffic" (Broadcast/Unknown-Unicast/ 267 Multicast), and the procedures for handling BUM traffic are known as 268 "BUM procedures". 270 [IGMP-Proxy] extends the intra-subnet ethernet multicast procedures 271 by adding procedures that are specific to, and optimized for, the use 272 of IP multicast within a subnet. However,that document does not 273 cover inter-subnet IP multicast. 275 The purpose of this document is to specify procedures for EVPN that 276 provide optimized IP multicast functionality within an EVPN tenant 277 domain. This document also specifies procedures that allow IP 278 multicast packets to be sourced from or destined to systems outside 279 the Tenant Domain. We refer to the entire set of these procedures as 280 "OISM" (Optimized Inter-Subnet Multicast) procedures. 282 In order to support the OISM procedures specified in this document, 283 an EVPN-PE MUST also support [EVPN-IRB] and [IGMP-Proxy]. (However, 284 certain of the procedures in [IGMP-Proxy] are modified when OISM is 285 supported.) 287 1.1.4. BDs, MAC-VRFS, and EVPN Service Models 289 [RFC7432] defines the notion of "MAC-VRF". A MAC-VRF contains one or 290 more "Bridge Tables" (see section 3 of [RFC7432] for a discussion of 291 this terminology), each of which represents a single Broadcast 292 Domain. 294 In the IRB model (outlined in Appendix A) a L3 routing instance has 295 one IRB interface per BD, NOT one per MAC-VRF. This document does 296 not distinguish between a "Broadcast Domain" and a "Bridge Table", 297 and will use the terms interchangeably (or will use the acronym "BD" 298 to refer to either). The way the BDs are grouped into MAC-VRFs is 299 not relevant to the procedures specified in this document. 301 Section 6 of [RFC7432] also defines several different EVPN service 302 models: 304 o In the "vlan-based service", each MAC-VRF contains one "bridge 305 table", where the bridge table corresponds to a particular Virtual 306 LAN (VLAN). (See section 3 of [RFC7432] for a discussion of this 307 terminology.) Thus each VLAN is treated as a BD. 309 o In the "vlan bundle service", each MAC-VRF contains one bridge 310 table, where the bridge table corresponds to a set of VLANs. Thus 311 a set of VLANs are treated as constituting a single BD. 313 o In the "vlan-aware bundle service", each MAC-VRF may contain 314 multiple bridge tables, where each bridge table corresponds to one 315 BD. If a MAC-VRF contains several bridge tables, then it 316 corresponds to several BDs. 318 The procedures of this document are intended to work for all these 319 service models. 321 1.2. Need for EVPN-aware Multicast Procedures 323 Inter-subnet IP multicast among a set of BDs can be achieved, in a 324 non-optimal manner, without any specific EVPN procedures. For 325 instance, if a particular tenant has n BDs among which he wants to 326 send IP multicast traffic, he can simply attach a conventional 327 multicast router to all n BDs. Or more generally, as long as each BD 328 has at least one IP multicast router, and the IP multicast routers 329 communicate multicast control information with each other, 330 conventional IP multicast procedures will work normally, and no 331 special EVPN functionality is needed. 333 However, that technique does not provide optimal routing for 334 multicast. In conventional multicast routing, for a given multicast 335 flow, there is only one multicast router on each BD that is permitted 336 to send traffic of that flow to the BD. If that BD has receivers for 337 a given flow, but the source of the flow is not on that BD, then the 338 flow must pass through that multicast router. This leads to the 339 "hair-pinning" problem described (for unicast) in Appendix A. 341 For example, consider an (S,G) flow that is sourced by a TS S and 342 needs to be received by TSes R1 and R2. Suppose S is on a segment of 343 BD1, R1 is on a segment of BD2, but both are attached to PE1. 344 Suppose also that the tenant has a multicast router, attached to a 345 segment of BD1 and to a segment of BD2. However, the segments to 346 which that router is attached are both attached to PE2. Then the 347 flow from S to R would have to follow the path: 348 S-->PE1-->PE2-->Tenant Multicast Router-->PE2-->PE1-->R1. Obviously, 349 the path S-->PE1-->R would be preferred. 351 Now suppose that there is a second receiver, R2. R2 is attached to a 352 third BD, BD3. However, it is attached to a segment of BD3 that is 353 attached to PE1. And suppose also that the Tenant Multicast Router 354 is attached to a segment of BD3 that attaches to PE2. In this case, 355 the Tenant Multicast Router will make two copies of the packet, one 356 for BD2 and one for BD3. PE2 will send both copies back to PE1. Not 357 only is the routing sub-optimal, but PE2 sends multiple copies of the 358 same packet to PE1. This is a further sub-optimality. 360 This is only an example; many more examples of sub-optimal multicast 361 routing can easily be given. To eliminate sub-optimal routing and 362 extra copies, it is necessary to have a multicast solution that is 363 EVPN-aware, and that can use its knowledge of the internal structure 364 of a Tenant Domain to ensure that multicast traffic gets routed 365 optimally. The procedures of this document allow us to avoid all 366 such sub-optimalities when routing inter-subnet multicasts within a 367 Tenant Domain. 369 1.3. Additional Requirements That Must be Met by the Solution 371 In addition to providing optimal routing of multicast flows within a 372 Tenant Domain, the EVPN-aware multicast solution is intended to 373 satisfy the following requirements: 375 o The solution must integrate well with the procedures specified in 376 [IGMP-Proxy]. That is, an integrated set of procedures must 377 handle both intra-subnet multicast and inter-subnet multicast. 379 o With regard to intra-subnet multicast, the solution MUST maintain 380 the integrity of multicast ethernet service. This means: 382 * If a source and a receiver are on the same subnet, the MAC 383 source address (SA) of the multicast frame sent by the source 384 will not get rewritten. 386 * If a source and a receiver are on the same subnet, no IP 387 processing of the ethernet payload is done. The IP TTL is not 388 decremented, the header checksum is not changed, no 389 fragmentation is done, etc. 391 o On the other hand, if a source and a receiver are on different 392 subnets, the frame received by the receiver will not have the MAC 393 Source address of the source, as the frame will appear to have 394 come from a multicast router. Also, proper processing of the IP 395 header is done, e.g., TTL decrement by 1, header checksum 396 modification, possibly fragmentation, etc. 398 o If a Tenant Domain contains several BDs, it MUST be possible for a 399 multicast flow (even when the multicast group address is an "any 400 source multicast" (ASM) address), to have sources in one of those 401 BDs and receivers in one or more of the other BDs, without 402 requiring the presence of any system performing PIM Rendezvous 403 Point (RP) functions ([RFC7761]). Multicast throughout a Tenant 404 Domain must not require the tenant systems to be aware of any 405 underlying multicast infrastructure. 407 o Sometimes a MAC address used by one TS on a particular BD is also 408 used by another TS on a different BD. Inter-subnet routing of 409 multicast traffic MUST NOT make any assumptions about the 410 uniqueness of a MAC address across several BDs. 412 o If two EVPN-PEs attached to the same Tenant Domain both support 413 the OISM procedures, each may receive inter-subnet multicasts from 414 the other, even if the egress PE is not attached to any segment of 415 the BD from which the multicast packets are being sourced. It 416 MUST NOT be necessary to provision the egress PE with knowledge of 417 the ingress BD. 419 o There must be a procedure that that allows EVPN-PE routers 420 supporting OISM procedures to send/receive multicast traffic to/ 421 from EVPN-PE routers that support only [RFC7432], but that do not 422 support the OISM procedures or even the procedures of [EVPN-IRB]. 423 However, when interworking with such routers (which we call 424 "non-OISM PE routers"), optimal routing may not be achievable. 426 o It MUST be possible to support scenarios in which multicast flows 427 with sources inside a Tenant Domain have "external" receivers, 428 i.e., receivers that are outside the domain. It must also be 429 possible to support scenarios where multicast flows with external 430 sources (sources outside the Tenant Domain) have receivers inside 431 the domain. 433 This presupposes that unicast routes to multicast sources outside 434 the domain can be distributed to EVPN-PEs attached to the domain, 435 and that unicast routes to multicast sources within the domain can 436 be distributed outside the domain. 438 Of particular importance are the scenario in which the external 439 sources and/or receivers are reachable via L3VPN/MVPN, and the 440 scenario in which external sources and/or receivers are reachable 441 via IP/PIM. 443 The solution for external interworking MUST allow for deployment 444 scenarios in which EVPN does not need to export a host route for 445 every multicast source. 447 o The solution for external interworking must not presuppose that 448 the same tunneling technology is used within both the EVPN domain 449 and the external domain. For example, MVPN interworking must be 450 possible when MVPN is using MPLS P2MP tunneling, and EVPN is using 451 Ingress Replication or VXLAN tunneling. 453 o The solution must not be overly dependent on the details of a 454 small set of use cases, but must be adaptable to new use cases as 455 they arise. (That is, the solution must be robust.) 457 1.4. Terminology 459 In this document we make frequent use of the following terminology: 461 o OISM: Optimized Inter-Subnet Multicast. EVPN-PEs that follow the 462 procedures of this document will be known as "OISM" PEs. EVPN-PEs 463 that do not follow the procedures of this document will be known 464 as "non-OISM" PEs. 466 o IP Multicast Packet: An IP packet whose IP Destination Address 467 field is a multicast address that is not a link-local address. 468 (Link-local addresses are IPv4 addresses in the 224/8 range and 469 IPv6 address in the FF02/16 range.) 471 o IP Multicast Frame: An ethernet frame whose payload is an IP 472 multicast packet (as defined above). 474 o (S,G) Multicast Packet: An IP multicast packet whose IP Source 475 Address field contains S and whose IP Destination Address field 476 contains G. 478 o (S,G) Multicast Frame: An IP multicast frame whose payload 479 contains S in its IP Source Address field and G in its IP 480 Destination Address field. 482 o Broadcast Domain (BD): an emulated ethernet, such that two systems 483 on the same BD will receive each other's link-local broadcasts. 485 Note that EVPN supports service models in which a single EVPN 486 Instance (EVI) contains only one BD, and service models in which a 487 single EVI contains multiple BDs. Both types of service model are 488 supported by this draft. In all models, a given BD belongs to 489 only one EVI. 491 o Designated Forwarder (DF). As defined in [RFC7432], an ethernet 492 segment may be multi-homed (attached to more than one PE). An 493 ethernet segment may also contain multiple BDs, of one or more 494 EVIs. For each such EVI, one of the PEs attached to the segment 495 becomes that EVI's DF for that segment. Since a BD may belong to 496 only one EVI, we can speak unambiguously of the BD's DF for a 497 given segment. 499 When the text makes it clear that we are speaking in the context 500 of a given BD, we will frequently use the term "a segment's DF" to 501 mean the given BD's DF for that segment. 503 o AC: Attachment Circuit. An AC connects the bridging function of 504 an EVPN-PE to an ethernet segment of a particular BD. ACs are not 505 visible at the router (L3) layer. 507 If a given ethernet segment, attached to a given PE, contains n 508 BDs, we will say that the PE has n ACs to that segment. 510 o L3 Gateway: An L3 Gateway is a PE that connects an EVPN tenant 511 domain to an external multicast domain by performing both the OISM 512 procedures and the Layer 3 multicast procedures of the external 513 domain. 515 o PEG (PIM/EVPN Gateway): A L3 Gateway that connects an EVPN Tenant 516 Domain to an external multicast domain whose Layer 3 multicast 517 procedures are those of PIM ([RFC7761]). 519 o MEG (MVPN/EVPN Gateway): A L3 Gateway that connects an EVPN Tenant 520 Domain to an external multicast domain whose Layer 3 multicast 521 procedures are those of MVPN ([RFC6513], [RFC6514]). 523 o IPMG (IP Multicast Gateway): A PE that is used for interworking 524 OISM EVPN-PEs with non-OISM EVPN-PEs. 526 o DR (Designated Router): A PE that has special responsibilities for 527 handling multicast on a given BD. 529 o FHR (First Hop Router): The FHR is a PIM router ([RFC7761]) with 530 special responsibilities. It is the first multicast router to see 531 (S,G) packets from source S, and if G is an "Any Source Multicast 532 (ASM)" group, the FHR is responsible for sending PIM Register 533 messages to the PIM Rendezvous Point for group G. 535 o LHR (Last Hop Router): The LHR is a PIM router ([RFC7761]) with 536 special responsibilities. Generally it is attached to a LAN, and 537 it determines whether there are any hosts on the LAN that need to 538 receive a given multicast flow. If so, it creates and sends the 539 PIM Join messages that are necessary to draw the flow. 541 o EC (Extended Community). A BGP Extended Communities attribute 542 ([RFC4360], [RFC7153]) is a BGP path attribute that consists of 543 one or more extended communities. 545 o RT (Route Target): A Route Target is a particular kind of BGP 546 Extended Community. A BGP Extended Community consists of a type 547 field, a sub-type field, and a value field. Certain type/sub-type 548 combinations indicate that a particular Extended Community is an 549 RT. RT1 and RT2 are considered to be the same RT if and only if 550 they have the same type, same sub-type, and same value fields. 552 o Use of the "C-" prefix. In many documents on VPN multicast, the 553 prefix "C-" appears before any address or wildcard that refers to 554 an address or addresses in a tenant's address space, rather than 555 to an address of addresses in the address space of the backbone 556 network. This document omits the "C-" prefix in many cases where 557 it is clear from the context that the reference is to the tenant's 558 address space. 560 This document also assumes familiarity with the terminology of 561 [RFC4364], [RFC6514], [RFC7432], [RFC7761], [IGMP-Proxy], 562 [EVPN_IP_Prefix] and [EVPN-BUM]. 564 1.5. Model of Operation: Overview 566 1.5.1. Control Plane 568 In this section, and in the remainder of this document, we assume the 569 reader is familiar with the procedures of IGMP/MLD (see [RFC2236] and 570 [RFC2710]), by which hosts announce their interest in receiving 571 particular multicast flows. 573 Consider a Tenant Domain consisting of a set of k BDs: BD1, ..., BDk. 574 To support the OISM procedures, each Tenant Domain must also be 575 associated with a "Supplementary Broadcast Domain" (SBD). An SBD is 576 treated in the control plane as a real BD, but it does not have any 577 ACs. The SBD has several uses; these will be described later in this 578 document (see Section 2.1 and Section 3). 580 Each PE that attaches to one or more of the BDs in a given tenant 581 domain will be provisioned to recognize that those BDs are part of 582 the same Tenant Domain. Note that a given PE does not need to be 583 configured with all the BDs of a given Tenant Domain. In general, a 584 PE will only be attached to a subset of the BDs in a given Tenant 585 Domain, and will be configured only with that subset of BDs. 586 However, each PE attached to a given Tenant Domain must be configured 587 with the SBD for that Tenant Domain. 589 Suppose a particular segment of a particular BD is attached to PE1. 590 [RFC7432] specifies that PE1 must originate an Inclusive Multicast 591 Ethernet Tag (IMET) route for that BD, and that the IMET route must 592 be propagated to all other PEs attached to the same BD. If the given 593 segment contains a host that has interest in receiving a particular 594 multicast flow, either an (S,G) flow or a (*,G) flow, PE1 will learn 595 of that interest by participating in the IGMP/MLD procedures, as 596 specified in [IGMP-Proxy]. In this case, we will say that: 598 o PE1 is interested in receiving the flow; 600 o The AC attaching the interested host to PE1 is also said to be 601 interested in the flow; 603 o The BD containing an AC that is interested in a particular flow is 604 also said to be interested in that flow. 606 Once PE1 determines that it has an AC that is interested in receiving 607 a particular flow or set of flows, it originates one or more 608 Selective Multicast Ethernet Tag (SMET) route to advertise that 609 interest. 611 Note that each IMET or SMET route is "for" a particular BD. The 612 notion of a route being "for" a particular BD is explained in 613 Section 2.2. 615 When OISM is being supported, the procedures of [IGMP-Proxy], are 616 modified as follows: 618 o The IMET route originated by a particular PE for a particular BD 619 is distributed to all other PEs attached to the Tenant Domain 620 containing that BD, even to those PEs that are not attached to 621 that particular BD. 623 o The SMET routes originated by a particular PE are originated on a 624 per-Tenant-Domain basis, rather than on a per-BD basis. That is, 625 the SMET routes are considered to be for the Tenant Domain's SBD, 626 rather than for any of its ordinary BDs. These SMET routes are 627 distributed to all the PEs attached to the Tenant Domain. 629 In this way, each PE attached to a given Tenant Domain learns, 630 from each other PE attached to the same Tenant Domain, the set of 631 flows that are of interest to each of those other PEs. 633 An OISM PE that is provisioned with several BDs in the same Tenant 634 Domain MUST originate an IMET route for each such BD. To indicate 635 its support of [IGMP-Proxy], it SHOULD attach the EVPN Multicast 636 Flags Extended Community to each such IMET route, but it MUST attach 637 the EC to at least one such IMET route. 639 Suppose PE1 is provisioned with both BD1 and BD2, and is provisioned 640 to consider them to be part of the same Tenant Domain. It is 641 possible that PE1 will receive from PE2 both an IMET route for BD1 642 and an IMET route for BD2. If either of these IMET routes has the 643 EVPN Multicast Flags Extended Community, PE1 MUST assume that PE2 is 644 supporting the procedures of [IGMP-Proxy] for ALL BDs in the Tenant 645 Domain. 647 If a PE supports OISM functionality, it indicates that by setting the 648 "OISM-supported" flag in the Multicast Flags Extended Community that 649 it attaches to some or all of its IMET routes. An OISM PE SHOULD 650 attach this EC with the OISM-supported flag set to all the IMET 651 routes it originates. However, if PE1 imports IMET routes from PE2, 652 and at least one of PE2's IMET routes indicates that PE2 is an OISM 653 PE, PE1 MUST assume that PE2 is following OISM procedures. 655 1.5.2. Data Plane 657 Suppose PE1 has an AC to a segment in BD1, and PE1 receives from that 658 AC an (S,G) multicast frame (as defined in Section 1.4). 660 There may be other ACs of PE1 on which TSes have indicated an 661 interest (via IGMP/MLD) in receiving (S,G) multicast packets. PE1 is 662 responsible for sending the received multicast packet out those ACs. 663 There are two cases to consider: 665 o Intra-Subnet Forwarding: In this case, an attachment AC with 666 interest in (S,G) is connected to a segment that is part of the 667 source BD, BD1. If the segment is not multi-homed, or if PE1 is 668 the Designated Forwarder (DF) (see [RFC7432]) for that segment, 669 PE1 sends the multicast frame on that AC without changing the MAC 670 SA. The IP header is not modified at all; in particular, the TTL 671 is not decremented. 673 o Inter-Subnet Forwarding: An AC with interest in (S,G) is connected 674 to a segment of BD2, where BD2 is different than BD1. If PE1 is 675 the DF for that segment (or if the segment is not multi-homed), 676 PE1 decapsulates the IP multicast packet, performs any necessary 677 IP processing (including TTL decrement), then re-encapsulates the 678 packet appropriately for BD2. PE1 then sends the packet on the 679 AC. Note that after re-encapsulation, the MAC SA will be PE1's 680 MAC address on BD2. The IP TTL will have been decremented by 1. 682 In addition, there may be other PEs that are interested in (S,G) 683 traffic. Suppose PE2 is such a PE. Then PE1 tunnels a copy of the 684 IP multicast frame (with its original MAC SA, and with no alteration 685 of the payload's IP header) to PE2. The tunnel encapsulation 686 contains information that PE2 can use to associate the frame with an 687 "apparent source BD". If the actual source BD of the frame is BD1, 688 then: 690 o If PE2 is attached to BD1, the tunnel encapsulation used to send 691 the frame to PE2 will cause PE2 to identify BD1 as the apparent 692 source BD. 694 o If PE2 is not attached to BD1, the tunnel encapsulation used to 695 send the frame to PE2 will cause PE2 to identify the SBD as the 696 apparent source BD. 698 Note that the tunnel encapsulation used for a particular BD will have 699 been advertised in an IMET route or S-PMSI route ([EVPN-BUM]) for 700 that BD. That route carries a PMSI Tunnel attribute, which specifies 701 how packets originating from that BD are encapsulated. This 702 information enables the PE receiving a tunneled packet to identify 703 the apparent source BD as stated above. See Section 3.2 for more 704 details. 706 When PE2 receives the tunneled frame, it will forward it on any of 707 its ACs that have interest in (S,G). 709 If PE2 determines from the tunnel encapsulation that the apparent 710 source BD is BD1, then 712 o For those ACs that connect PE2 to BD1, the intra-subnet forwarding 713 procedure described above is used, except that it is now PE2, not 714 PE1, carrying out that procedure. Unmodified EVPN procedures from 716 [RFC7432] are used to ensure that a packet originating from a 717 multi-homed segment is never sent back to that segment. 719 o For those ACs that do not connect to BD1, the inter-subnet 720 forwarding procedure described above is used, except that it is 721 now PE2, not PE1, carrying out that procedure. 723 If the tunnel encapsulation identifies the apparent source BD as the 724 SBD, PE2 applies the inter-subnet forwarding procedures described 725 above to all of its ACs that have interest in the flow. 727 These procedures ensure that an IP multicast frame travels from its 728 ingress PE to all egress PEs that are interested in receiving it. 729 While in transit, the frame retains its original MAC SA, and the 730 payload of the frame retains its original IP header. Note that in 731 all cases, when an IP multicast packet is sent from one BD to 732 another, these procedures cause its TTL to be decremented by 1. 734 So far we have assumed that an IP multicast packet arrives at its 735 ingress PE over an AC that belongs to one of the BDs in a given 736 Tenant Domain. However, it is possible for a packet to arrive at its 737 ingress PE in other ways. Since an EVPN-PE supporting IRB has an 738 IP-VRF, it is possible that the IP-VRF will have a "VRF interface" 739 that is not an IRB interface. For example, there might be a VRF 740 interface that is actually a physical link to an external ethernet 741 switch, or to a directly attached host, or to a router. When an 742 EVPN-PE, say PE1, receives a packet through such means, we will say 743 that the packet has an "external" source (i.e., a source "outside the 744 Tenant Domain"). There are also other scenarios in which a multicast 745 packet might have an external source, e.g., it might arrive over an 746 MVPN tunnel from an L3VPN PE. In such cases, we will still refer to 747 PE1 as the "ingress EVPN-PE". 749 When an EVPN-PE, say PE1, receives an externally sourced multicast 750 packet, and there are receivers for that packet inside the Tenant 751 Domain, it does the following: 753 o Suppose PE1 has an AC in BD1 that has interest in (S,G). Then PE1 754 encapsulates the packet for BD1, filling in the MAC SA field with 755 PE1's own MAC address on BD1. It sends the resulting frame on the 756 AC. 758 o Suppose some other EVPN-PE, say PE2, has interest in (S,G). PE1 759 encapsulates the packet for ethernet, filling in the MAC SA field 760 with PE1's own MAC address on the SBD. PE1 then tunnels the 761 packet to PE2. The tunnel encapsulation will identify the 762 apparent source BD as the SBD. Since the apparent source BD is 763 the SBD, PE2 will know to treat the frame as an inter-subnet 764 multicast. 766 When ingress replication is used to transmit IP multicast frames from 767 an ingress EVPN-PE to a set of egress PEs, then of course the ingress 768 PE has to send multiple copies of the frame. Each copy is the 769 original ethernet frame; decapsulation and IP processing take place 770 only at the egress PE. 772 If a Point-to-Multipoint (P2MP) tree or BIER ([EVPN-BIER]) is used to 773 transmit an IP multicast frame from an ingress PE to a set of egress 774 PEs, then the ingress PE only has to send one copy of the frame to 775 each of its next hops. Again, each egress PE receives the original 776 frame and does any necessary IP processing. 778 2. Detailed Model of Operation 780 The model described in Section 1.5.2 can be expressed more precisely 781 using the notion of "IRB interface" (see Appendix A). For a given 782 Tenant Domain: 784 o A given PE has one IRB for each BD to which it is attached. This 785 IRB interface connects L3 routing to that BD. When IP multicast 786 packets are sent or received on the IRB interfaces, the semantics 787 of the interface is modified from the semantics described in 788 Appendix A. See Section 2.3 for the details of the modification. 790 o Each PE also has an IRB interface that connects L3 routing to the 791 SBD. The semantics of this interface is different than the 792 semantics of the IRB interface to the real BDs. See Section 2.3. 794 In this section we assume that PIM is not enabled on the IRB 795 interfaces. In general, it is not necessary to enable PIM on the IRB 796 interfaces unless there are PIM routers on one of the Tenant Domain's 797 BDs, or unless there is some other scenario requiring a Tenant 798 Domain's L3 routing instance to become a PIM adjacency of some other 799 system. These cases will be discussed in Section 7. 801 2.1. Supplementary Broadcast Domain 803 Suppose a given Tenant Domain contains three BDs (BD1, BD2, BD3) and 804 two PEs (PE1, PE2). PE1 attaches to BD1 and BD2, while PE2 attaches 805 to BD2 and BD3. 807 To carry out the procedures described above, all the PEs attached to 808 the Tenant Domain must be provisioned with the SBD for that tenant 809 domain. A Route Target (RT) must be associated with the SBD, and 810 provisioned on each of those PEs. We will refer to that RT as the 811 "SBD-RT". 813 A Tenant Domain is also configured with an IP-VRF ([EVPN-IRB]), and 814 the IP-VRF is associated with an RT. This RT MAY be the same as the 815 SBD-RT. 817 Suppose an (S,G) multicast frame originating on BD1 has a receiver on 818 BD3. PE1 will transmit the packet to PE2 as a frame, and the 819 encapsulation will identify the frame's source BD as BD1. Since PE2 820 is not provisioned with BD1, it will treat the packet as if its 821 source BD were the SBD. That is, a packet can be transmitted from 822 BD1 to BD3 even though its ingress PE is not configured for BD3, and/ 823 or its egress PE is not configured for BD1. 825 EVPN supports service models in which a given EVPN Instance (EVI) can 826 contain only one BD. It also supports service models in which a 827 given EVI can contain multiple BDs. No matter which service model is 828 being used for a particular tenant, it is highly RECOMMENDED that an 829 EVI containing only the SBD be provisioned for that tenant. 831 If, for some reason, it is not feasible to provision an EVI that 832 contains only the SBD, it is possible to put the SBD in an EVI that 833 contains other BDs. However, in that case, the SBD-RT MUST be 834 different than the RT associated with any other BD. Otherwise the 835 procedures of this document (as detailed in Sections 2.2 and 3.1) 836 will not produce correct results. 838 2.2. Detecting When a Route is About/For/From a Particular BD 840 In this document, we frequently say that a particular multicast route 841 is "about" a particular BD, or is "from" a particular BD, or is "for" 842 a particular BD or is "related to" a particular BD or "is associated 843 with" a particular BD. These terms are used interchangeably. 844 Subsequent sections of this document explain when various routes must 845 be originated for particular BDs. In this section, we explain how 846 the PE originating a route marks the route to indicate which BD it is 847 about. We also explain how a PE receiving the route determines which 848 BD the route is about. 850 In EVPN, each BD is assigned a Route Target (RT). An RT is a BGP 851 extended community that can be attached to the BGP routes used by the 852 EVPN control plane. In some EVPN service models, each BD is assigned 853 a unique RT. In other service models, a set of BDs (all in the same 854 EVI) may be assigned the same RT. The RT that is assigned to the SBD 855 is called the "SBD-RT". 857 In those service models that allow a set of BDs to share a single RT, 858 each BD is assigned a non-zero Tag ID. The Tag ID appears in the 859 Network Layer Reachability Information (NLRI) of many of the BGP 860 routes that are used by the EVPN control plane. 862 A given route may be about the SBD, or about an "ordinary BD" (a BD 863 that is not the SBD). An RT that has been assigned to an ordinary BD 864 will be known as an "ordinary BD-RT". 866 When constructing an IMET, SMET, S-PMSI ([EVPN-BUM]), or Leaf 867 ([EVPN-BUM]) route that is about a given BD, the following rules 868 apply: 870 o If the route is about an ordinary BD, say BD1, then 872 * the route MUST carry the ordinary BD-RT associated with BD1, 873 and 875 * the route MUST NOT carry any RT that is associated with an 876 ordinary BD other than BD1. 878 o If the route is about the SBD, the route MUST carry the SBD-RT, 879 and MUST NOT carry any RT that is associated with any other BD. 881 o As detailed in subsequent sections, under certain circumstances a 882 route that is about BD1 may carry both the RT of BD1 and also the 883 SBD-RT. 885 The IMET route for the SBD MUST carry an Multicast Flags Extended 886 Community, in which an "OISM SBD" flag is set. 888 The IMET route for a BD other than the SBD SHOULD carry an EVI-RT EC 889 as defined in [IGMP-Proxy]. The EC is constructed from the SBD-RT, 890 to indicate the BD's corresponding SBD. This allows all PEs to check 891 that they have consistent SBD provisioning and allow an AR-replicator 892 to automatically determine a BD's corresponding SBD w/o any 893 provisioning, as explained in Section 3.2.3.1. 895 When receiving an IMET, SMET, S-PMSI or Leaf route, it is necessary 896 for the receiving PE to determine the BD to which the route belongs. 897 This is done by examining the RTs carried by the route, as well as 898 the Tag ID field of the route's NLRI. There are several cases to 899 consider. Some of these cases are error cases that arise when the 900 route has not been properly constructed. 902 When one of the error cases is detected, the route MUST be regarded 903 as a malformed route, and the "treat-as-withdraw" procedure of 904 [RFC7606] MUST be applied. Note though that these error cases are 905 only detectable by EVPN procedures at the receiving PE; BGP 906 procedures at intermediate nodes will generally not detect the 907 existence of such error cases, and in general SHOULD NOT attempt to 908 do so. 910 Case 1: The receiving PE recognizes more than one of the route's RTs 911 as being an SBD-RT (i.e., the route carries SBD-RTs of more 912 than one Tenant Domain). 914 This is an error case; the route has not been properly 915 constructed. 917 Case 2: The receiving PE recognizes one of the route's RTs as being 918 associated with an ordinary BD, and recognizes one of the 919 route's other RTs as being associated with a different 920 ordinary BD. 922 This is an error case; the route has not been properly 923 constructed. 925 Case 3: The receiving PE recognizes one of the route's RTs as being 926 associated with an ordinary BD in a particular Tenant 927 Domain, and recognizes another of the route's RTs as being 928 associated with the SBD of a different Tenant Domain. 930 This is an error case; the route has not been properly 931 constructed. 933 Case 4: The receiving PE does not recognize any of the route's RTs 934 as being associated with an ordinary BD in any of its tenant 935 domains, but does recognize one of the RTs as the SBD-RT of 936 one of its Tenant Domains. 938 In this case, receiving PE associates the route with the SBD 939 of that Tenant Domain. This association is made even if the 940 Tag ID field of the route's NLRI is not the Tag ID of the 941 SBD. 943 This is a normal use case where either (a) the route is for 944 a BD to which the receiving PE is not attached, or (b) the 945 route is for the SBD. In either case, the receiving PE 946 associates the route with the SBD. 948 Case 5: The receiving PE recognizes exactly one of the RTs as an 949 ordinary BD-RT that is associated with one of the PE's EVIs, 950 say EVI-1. The receiving PE also recognizes one of the RTs 951 as being the SBD-RT of the Tenant Domain containing EVI-1. 953 In this case, the route is associated with the BD in EVI-1 954 that is identified (in the context of EVI-1) by the Tag ID 955 field of the route's NLRI. (If EVI-1 contains only a single 956 BD, the Tag ID is likely to be zero.) 958 This is the case where the route is for a BD to which the 959 receiving PE is attached, but the route also carries the 960 SBD-RT. In this case, the receiving PE associates the route 961 with the ordinary BD, not with the SBD. 963 N.B.: According to the above rules, the mapping from BD to RT is a 964 many-to-one or one-to-one mapping. A route that an EVPN-PE 965 originates for a particular BD carries that BD's RT, and an EVPN-PE 966 that receives the route associates it with a BD as described above. 967 However, RTs are not used only to help identify the BD to which a 968 route belongs; they may also used by BGP to determine the path along 969 which the route is distributed, and to determine which PEs receive 970 the route. There may be cases where it is desirable to originate a 971 route about a particular BD, but have that route distributed to only 972 some of the EVPN-PEs attached to that BD. Or one might want the 973 route distributed to some intermediate set of systems, where it might 974 be modified or replaced before being propagated further. Such 975 situations are outside the scope of this document. 977 Additionally, there may be situations where it is desirable to 978 exchange routes among two or more different Tenant Domains ("EVPN 979 Extranet"). Such situations are outside the scope of this document. 981 2.3. Use of IRB Interfaces at Ingress PE 983 When an (S,G) multicast frame is received from an AC belonging to a 984 particular BD, say BD1: 986 1. The frame is sent unchanged to other EVPN-PEs that are interested 987 in (S,G) traffic. The encapsulation used to send the frame to 988 the other EVPN-PEs depends on the tunnel type being used for 989 multicast transmission. (For our purposes, we consider Ingress 990 Replication (IR), Assisted Replication (AR) and BIER to be 991 "tunnel types", even though IR, AR and BIER do not actually use 992 P2MP tunnels.) At the egress PE, the apparent source BD of the 993 frame can be inferred from the tunnel encapsulation. If the 994 egress PE is not attached to the actual source BD, it will infer 995 that the apparent source BD is the SBD. 997 Note that the the inter-PE transmission of a multicast frame 998 among EVPN-PEs of the same Tenant Domain does NOT involve the IRB 999 interfaces, as long as the multicast frame was received over an 1000 AC attached to one of the Tenant Domain's BDs. 1002 2. The frame is also sent up the IRB interface that attaches BD1 to 1003 the Tenant Domain's L3 routing instance in this PE. That is, the 1004 L3 routing instance, behaving as if it were a multicast router, 1005 receives the IP multicast frames that arrive at the PE from its 1006 local ACs. The L3 routing instance decapsulates the frame's 1007 payload to extract the IP multicast packet, decrements the IP 1008 TTL, adjusts the header checksum, and does any other necessary IP 1009 processing (e.g., fragmentation). 1011 3. The L3 routing instance keeps track of which BDs have local 1012 receivers for (S,G) traffic. (A "local receiver" is a TS, 1013 reachable via a local AC, that has expressed interest in (S,G) 1014 traffic.) If the L3 routing instance has an IRB interface to 1015 BD2, and it knows that BD2 has a LOCAL receiver interested in 1016 (S,G) traffic, it encapsulates the packet in an ethernet header 1017 for BD2, putting its own MAC address in the MAC SA field. Then 1018 it sends the packet down the IRB interface to BD2. 1020 If a packet is sent from the L3 routing instance to a particular BD 1021 via the IRB interface (step 3 in the above list), and if the BD in 1022 question is NOT the SBD, the packet is sent ONLY to LOCAL ACs of that 1023 BD. If the packet needs to go to other PEs, it has already been sent 1024 to them in step 1. Note that this is a change in the IRB interface 1025 semantics from what is described in [EVPN-IRB] and Figure 2. 1027 If a given locally attached segment is multi-homed, existing EVPN 1028 procedures ensure that a packet is not sent by a given PE to that 1029 segment unless the PE is the DF for that segment. Those procedures 1030 also ensure that a packet is never sent by a PE to its segment of 1031 origin. Thus EVPN segment multi-homing is fully supported; duplicate 1032 delivery to a segment or looping on a segment are thereby prevented, 1033 without the need for any new procedures to be defined in this 1034 document. 1036 What if an IP multicast packet is received from outside the tenant 1037 domain? For instance, perhaps PE1's IP-VRF for a particular tenant 1038 domain also has a physical interface leading to an external switch, 1039 host, or router, and PE1 receives an IP multicast packet or frame on 1040 that interface. Or perhaps the packet is from an L3VPN, or a 1041 different EVPN Tenant Domain. 1043 Such a packet is first processed by the L3 routing instance, which 1044 decrements TTL and does any other necessary IP processing. Then the 1045 packet is sent into the Tenant Domain by sending it down the IRB 1046 interface to the SBD of that Tenant Domain. This requires 1047 encapsulating the packet in an ethernet header. The MAC SA field 1048 will contain the PE's own MAC on the SBD. 1050 An IP multicast packet sent by the L3 routing instance down the IRB 1051 interface to the SBD is treated as if it had arrived from a local AC, 1052 and steps 1-3 are applied. Note that the semantics of sending a 1053 packet down the IRB interface to the SBD are thus slightly different 1054 than the semantics of sending a packet down other IRB interfaces. IP 1055 multicast packets sent down the SBD's IRB interface may be 1056 distributed to other PEs, but IP multicast packets sent down other 1057 IRB interfaces are distributed only to local ACs. 1059 If a PE sends a link-local multicast packet down the SBD IRB 1060 interface, that packet will be distributed (as an ethernet frame) to 1061 other PEs of the Tenant Domain, but will not appear on any of the 1062 actual BDs. 1064 2.4. Use of IRB Interfaces at an Egress PE 1066 Suppose an egress EVPN-PE receives an (S,G) multicast frame from the 1067 frame's ingress EVPN-PE. As described above, the packet will arrive 1068 as an ethernet frame over a tunnel from the ingress PE, and the 1069 tunnel encapsulation will identify the source BD of the ethernet 1070 frame. 1072 We define the notion of the frame's "apparent source BD" as follows. 1073 If the egress PE is attached to the actual source BD, the actual 1074 source BD is the apparent source BD. If the egress PE is not 1075 attached to the actual source BD, the SBD is the apparent source BD. 1077 The egress PE now takes the following steps: 1079 1. If the egress PE has ACs belonging to the apparent source BD of 1080 the frame, it sends the frame unchanged to any ACs of that BD 1081 that have interest in (S,G) packets. The MAC SA of the frame is 1082 not modified, and the IP header of the frame's payload is not 1083 modified in any way. 1085 2. The frame is also sent to the L3 routing instance by being sent 1086 up the IRB interface that attaches the L3 routing instance to the 1087 apparent source BD. Steps 2 and 3 of Section 2.3 are then 1088 applied. 1090 2.5. Announcing Interest in (S,G) 1092 [IGMP-Proxy] defines procedures used by an egress PE to announce its 1093 interest in a multicast flow or set of flows. If an egress PE 1094 determines it has LOCAL receivers in a particular BD, say BD1, that 1095 are interested in a particular set of flows, it originates one or 1096 more SMET routes for BD1. Each SMET route specifies a particular 1097 (S,G) or (*,G) flow. By originating an SMET route for BD1, a PE is 1098 announcing "I have receivers for (S,G) or (*,G) in BD1". Such an 1099 SMET route carries the Route Target (RT) for BD1, ensuring that it 1100 will be distributed to all PEs that are attached to BD1. 1102 The OISM procedures for originating SMET routes differ slightly from 1103 those in [IGMP-Proxy]. In most cases, the SMET routes are considered 1104 to be for the SBD, rather than for the BD containing local receivers. 1105 These SMET routes carry the SBD-RT, and do not carry any ordinary BD- 1106 RT. Details on the processing of SMET routes can be found in 1107 Section 3.3. 1109 Since the SMET routes carry the SBD-RT, every ingress PE attached to 1110 a particular Tenant Domain will learn of all other PEs (attached to 1111 the same Tenant Domain) that have interest in a particular set of 1112 flows. Note that a PE that receives a given SMET route does not 1113 necessarily have any BDs (other than the SBD) in common with the PE 1114 that originates that SMET route. 1116 If all the sources and receivers for a given (*,G) are in the Tenant 1117 Domain, inter-subnet "Any Source Multicast" traffic will be properly 1118 routed without requiring any Rendezvous Points, shared trees, or 1119 other complex aspects of multicast routing infrastructure. Suppose, 1120 for example, that: 1122 o PE1 has a local receiver, on BD1, for (*,G) 1124 o PE2 has a local source, on BD2, for (*,G). 1126 PE1 will originate an SMET(*,G) route for the SBD, and PE2 will 1127 receive that route, even if PE2 is not attached to BD1. PE2 will 1128 thus know to forward (S,G) traffic to PE1. PE1 does not need to do 1129 any "source discovery". (This does assume that source S does not 1130 send the same (S,G) datagram on two different BDs, and that the 1131 Tenant Domain does not contain two or more sources with the same IP 1132 address S. The use of multicast sources that have IP "anycast" 1133 addresses is outside the scope of this document.) 1135 If some PE attached to the Tenant Domain does not support [IGMP- 1136 Proxy], it will be assumed to be interested in all flows. Whether a 1137 particular remote PE supports [IGMP-Proxy] is determined by the 1138 presence of the Multicast Flags Extended Community in its IMET route; 1139 this is specified in [IGMP-Proxy]. 1141 2.6. Tunneling Frames from Ingress PE to Egress PEs 1143 [RFC7432] specifies the procedures for setting up and using "BUM 1144 tunnels". A BUM tunnel is a tunnel used to carry traffic on a 1145 particular BD if that traffic is (a) broadcast traffic, or (b) 1146 unicast traffic with an unknown MAC DA, or (c) ethernet multicast 1147 traffic. 1149 This document allows the BUM tunnels to be used as the default 1150 tunnels for transmitting IP multicast frames. It also allows a 1151 separate set of tunnels to be used, instead of the BUM tunnels, as 1152 the default tunnels for carrying IP multicast frames. Let's call 1153 these "IP Multicast Tunnels". 1155 When the tunneling is done via Ingress Replication or via BIER, this 1156 difference is of no significance. However, when P2MP tunnels are 1157 used, there is a significant advantage to having separate IP 1158 multicast tunnels. 1160 Other things being equal, it is desirable for an ingress PE to 1161 transmit a copy of a given (S,G) multicast frame on only one P2MP 1162 tunnel. All egress PEs interested in (S,G) packets then have to join 1163 that tunnel. If the source BD and PE for an (S,G) frame are BD1 an 1164 PE1 respectively, and if PE2 has receivers on BD2 for (S,G), then PE2 1165 must join the P2MP LSP on which PE1 transmits the (S,G) frame. PE2 1166 must join this P2MP LSP even if PE2 is not attached to the source BD 1167 (BD1). If PE1 were transmitting the multicast frame on its BD1 BUM 1168 tunnel, then PE2 would have to join the BD1 BUM tunnel, even though 1169 PE2 has no BD1 attachment circuits. This would cause PE2 to pull all 1170 the BUM traffic from BD1, most of which it would just have to 1171 discard. Thus we RECOMMEND that the default IP multicast tunnels be 1172 distinct from the BUM tunnels. 1174 Notwithstanding the above, link local IP multicast traffic MUST 1175 always be carried on the BUM tunnels, and ONLY on the BUM tunnels. 1176 Link local IP multicast traffic consists of IPv4 traffic with a 1177 destination address prefix of 224/8 and IPv6 traffic with a 1178 destination address prefix of FF02/16. In this document, the terms 1179 "IP multicast packet" and "IP multicast frame" are defined in 1180 Section 1.4 so as to exclude the link-local traffic. 1182 Note that it is also possible to use "selective tunnels" to carry 1183 particular multicast flows (see Section 3.2). When an (S,G) frame is 1184 transmitted on a selective tunnel, it is not transmitted on the BUM 1185 tunnel or on the default IP Multicast tunnel. 1187 2.7. Advanced Scenarios 1189 There are some deployment scenarios that require special procedures: 1191 1. Some multicast sources or receivers are attached to PEs that 1192 support [RFC7432], but do not support this document or 1193 [EVPN-IRB]. To interoperate with these "non-OISM PEs", it is 1194 necessary to have one or more gateway PEs that interface the 1195 tunnels discussed in this document with the BUM tunnels of the 1196 legacy PEs. This is discussed in Section 5. 1198 2. Sometimes multicast traffic originates from outside the EVPN 1199 domain, or needs to be sent outside the EVPN domain. This is 1200 discussed in Section 6. An important special case of this, 1201 integration with MVPN, is discussed in Section 6.1.2. 1203 3. In some scenarios, one or more of the tenant systems is a PIM 1204 router, and the Tenant Domain is used for as a transit network 1205 that is part of a larger multicast domain. This is discussed in 1206 Section 7. 1208 3. EVPN-aware Multicast Solution Control Plane 1210 3.1. Supplementary Broadcast Domain (SBD) and Route Targets 1212 As discussed in Section 2.1, every Tenant Domain is associated with a 1213 single Supplementary Broadcast Domain (SBD). Recall that a Tenant 1214 Domain is defined to be a set of BDs that can freely send and receive 1215 IP multicast traffic to/from each other. If an EVPN-PE has one or 1216 more ACs in a BD of a particular Tenant Domain, and if the EVPN-PE 1217 supports the procedures of this document, that EVPN-PE MUST be 1218 provisioned with the SBD of that Tenant Domain. 1220 At each EVPN-PE attached to a given Tenant Domain, there is an IRB 1221 interface leading from the L3 routing instance of that Tenant Domain 1222 to the SBD. However, the SBD has no ACs. 1224 Each SBD is provisioned with a Route Target (RT). All the EVPN-PEs 1225 supporting a given SBD are provisioned with that RT as an import RT. 1226 That RT MUST NOT be the same as the RT associated with any other BD. 1228 We will use the term "SBD-RT" to denote the RT has has been assigned 1229 to the SBD. Routes carrying this RT will be propagated to all 1230 EVPN-PEs in the same Tenant Domain as the originator. 1232 Section 2.2 specifies the rules by which an EVPN-PE that receives a 1233 route determines whether a received route "belongs to" a particular 1234 ordinary BD or SBD. 1236 Section 2.2 also specifies additional rules that must be following 1237 when constructing routes that belong to a particular BD, including 1238 the SBD. 1240 The SBD SHOULD be in an EVPN Instance (EVI) of its own. Even if the 1241 SBD is not in an EVI of its own, the SBD-RT MUST be different than 1242 the RT associated with any other BD. This restriction is necessary 1243 in order for the rules of Sections 2.2 and 3.1 to work correctly. 1245 Note that an SBD, just like any other BD, is associated on each 1246 EVPN-PE with a MAC-VRF. Per [RFC7432], each MAC-VRF is associated 1247 with a Route Distinguisher (RD). When constructing a route that is 1248 "about" an SBD, an EVPN-PE will place the RD of the associated 1249 MAC-VRF in the "Route Distinguisher" field of the NLRI. (If the 1250 Tenant Domain has several MAC-VRFs on a given PE, the EVPN-PE has a 1251 choice of which RD to use.) 1253 If Assisted Replication (AR, see [EVPN-AR]) is used, each 1254 AR-REPLICATOR for a given Tenant Domain must be provisioned with the 1255 SBD of that Tenant Domain, even if the AR-REPLICATOR does not have 1256 any L3 routing instance. 1258 3.2. Advertising the Tunnels Used for IP Multicast 1260 The procedures used for advertising the tunnels that carry IP 1261 multicast traffic depend upon the type of tunnel being used. If the 1262 tunnel type is neither Ingress Replication, Assisted Replication, nor 1263 BIER, there are procedures for advertising both "inclusive tunnels" 1264 and "selective tunnels". 1266 When IR, AR or BIER are used to transmit IP multicast packets across 1267 the core, there are no P2MP tunnels. Once an ingress EVPN-PE 1268 determines the set of egress EVPN-PEs for a given flow, the IMET 1269 routes contain all the information needed to transport packets of 1270 that flow to the egress PEs. 1272 If AR is used, the ingress EVPN-PE is also an AR-LEAF and the IMET 1273 route coming from the selected AR-REPLICATOR contains the information 1274 needed. The AR-REPLICATOR will behave as an ingress EVPN-PE when 1275 sending a flow to the egress EVPN-PEs. 1277 If the tunneling technique requires P2MP tunnels to be set up (e.g., 1278 RSVP-TE P2MP, mLDP, PIM), some of the tunnels may be selective 1279 tunnels and some may be inclusive tunnels. 1281 Selective P2MP tunnels are always advertised by the ingress PE using 1282 S-PMSI A-D routes ([EVPN-BUM]). 1284 For inclusive tunnels, there is a choice between using a BD's 1285 ordinary "BUM tunnel" [RFC7432] as the default inclusive tunnel for 1286 carrying IP multicast traffic, or using a separate IP multicast 1287 tunnel as the default inclusive tunnel for carrying IP multicast. In 1288 the former case, the inclusive tunnel is advertised in an IMET route. 1289 In the latter case, the inclusive tunnel is advertised in a (C-*,C-*) 1290 S-PMSI A-D route ([EVPN-BUM]). Details may be found in subsequent 1291 sections. 1293 3.2.1. Constructing Routes for the SBD 1295 There are situations in which an EVPN-PE needs to originate IMET, 1296 SMET, and/or SPMSI routes for the SBD. Throughout this document, we 1297 will refer to such routes respectively as "SBD-IMET routes", 1298 "SBD-SMET routes", and "SBD-SPMSI routes". Subsequent sections 1299 detail the conditions under which these routes need to be originated. 1301 When an EVPN-PE needs to originate an SBD-IMET, SBD-SMET, or 1302 SBD-SPMSI route, it constructs the route as follows: 1304 o the RD field of the route's NLRI is set to the RD of the MAC-VRF 1305 that is associated with the SBD; 1307 o the SBD-RT is attached to the route; 1309 o the "Tag ID" field of the route's NLRI is set to the Tag ID that 1310 has been assigned to the SBD. This is most likely 0 if a 1311 VLAN-based or VLAN-bundle service is being used, but non-zero if a 1312 VLAN-aware bundle service is being used. 1314 3.2.2. Ingress Replication 1316 When Ingress Replication (IR) is used to transport IP multicast 1317 frames of a given Tenant Domain, each EVPN-PE attached to that Tenant 1318 Domain MUST originate an SBD-IMET route (see Section 3.2.1). 1320 The SBD-IMET route MUST carry a PMSI Tunnel attribute (PTA), and the 1321 MPLS label field of the PTA MUST specify a downstream-assigned MPLS 1322 label that maps uniquely (in the context of the originating EVPN-PE) 1323 to the SBD. 1325 Following the procedures of [RFC7432], an EVPN-PE MUST also originate 1326 an IMET route for each BD to which it is attached. Each of these 1327 IMET routes carries a PTA specifying a downstream-assigned label that 1328 maps uniquely, in the context of the originating EVPN-PE, to the BD 1329 in question. These IMET routes need not carry the SBD-RT. 1331 When an ingress EVPN-PE needs to use IR to send an IP multicast frame 1332 from a particular source BD to an egress EVPN-PE, the ingress PE 1333 determines whether the egress PE has originated an IMET route for 1334 that BD. If so, that IMET route contains the MPLS label that the 1335 egress PE has assigned to the source BD. The ingress PE uses that 1336 label when transmitting the packet to the egress PE. Otherwise, the 1337 ingress PE uses the label that the egress PE has assigned to the SBD 1338 (in the SBD-IMET route originated by the egress). 1340 Note that the set of IMET routes originated by a given egress PE, and 1341 installed by a given ingress PE, may change over time. If the egress 1342 PE withdraws its IMET route for the source BD, the ingress PE MUST 1343 stop using the label carried in that IMET route, and instead MUST use 1344 the label carried in the SBD-IMET route from that egress PE. 1345 Implementors must also take into account that an IMET route from a 1346 particular PE for a particular BD may arrive after that PE's SBD-IMET 1347 route. 1349 3.2.3. Assisted Replication 1351 When Assisted Replication is used to transport IP multicast frames of 1352 a given Tenant Domain, each EVPN-PE (including the AR-REPLICATOR) 1353 attached to the Tenant Domain MUST originate an SBD-IMET route (see 1354 Section 3.2.1). 1356 An AR-REPLICATOR attached to a given Tenant Domain is considered to 1357 be an EVPN-PE of that Tenant Domain. It is attached to all the BDs 1358 in the Tenant Domain, but it does not necessarily have L3 routing 1359 instances. 1361 As with Ingress Replication, the SBD-IMET route carries a PTA where 1362 the MPLS label field specifies the downstream-assigned MPLS label 1363 that identifies the SBD. However, the AR-REPLICATOR and AR-LEAF 1364 EVPN-PEs will set the PTA's flags differently, as per [EVPN-AR]. 1366 In addition, each EVPN-PE originates an IMET route for each BD to 1367 which it is attached. As in the case of Ingress Replication, these 1368 routes carry the downstream-assigned MPLS labels that identify the 1369 BDs and do not carry the SBD-RT. 1371 When an ingress EVPN-PE, acting as AR-LEAF, needs to send an IP 1372 multicast frame from a particular source BD to an egress EVPN-PE, the 1373 ingress PE determines whether there is any AR-REPLICATOR that 1374 originated an IMET route for that BD. After the AR-REPLICATOR 1375 selection (if there are more than one), the AR-LEAF uses the label 1376 contained in the IMET route of the AR-REPLICATOR when transmitting 1377 packets to it. The AR-REPLICATOR receives the packet and, based on 1378 the procedures specified in [EVPN-AR] and in Section 3.2.2 of this 1379 document, transmits the packets to the egress EVPN-PEs using the 1380 labels contained in the received IMET routes for either the source BD 1381 or the SBD. 1383 If an ingress AR-LEAF for a given BD has not received any IMET route 1384 for that BD from an AR-REPLICATOR, the ingress AR-LEAF follows the 1385 procedures in Section 3.2.2. 1387 3.2.3.1. Automatic SBD Matching 1389 Each PE needs to know a BD's corresponding SBD. Configuring that 1390 information in each BD is one way but it requires repetitive 1391 configuration and consistency check (to make sure that all the BDs of 1392 the same tenant are configured with the same SBD). A better way is 1393 to configure the SBD info in the L3 routing instance so that all 1394 related BDs will derive the SBD information. 1396 An AR-replicator also needs to know same information, though it does 1397 not necessarily have an L3 routing instance. However from the EVI-RT 1398 EC in a BD's IMET route, an AR-replicator can derive the 1399 corresponding SBD of that BD w/o any configuration. 1401 3.2.4. BIER 1403 When BIER is used to transport multicast packets of a given Tenant 1404 Domain, and a given EVPN-PE attached to that Tenant Domain is a 1405 possible ingress EVPN-PE for traffic originating outside that Tenant 1406 Domain, the given EVPN-PE MUST originate an SBD-IMET route, (see 1407 Section 3.2.1). 1409 In addition, IMET routes that are originated for other BDs in the 1410 Tenant Domain MUST carry the SBD-RT. 1412 Each IMET route (including but not limited to the SBD-IMET route) 1413 MUST carry a PMSI Tunnel attribute (PTA). The MPLS label field of 1414 the PTA MUST specify an upstream-assigned MPLS label that maps 1415 uniquely (in the context of the originating EVPN-PE) to the BD for 1416 which the route is originated. 1418 Suppose an ingress EVPN-PE, PE1, needs to use BIER to tunnel an IP 1419 multicast frame to a set of egress EVPN-PEs. And suppose the frame's 1420 source BD is BD1. The frame is encapsulated as follows: 1422 o A four-octet MPLS label stack entry ([RFC3032]) is prepended to 1423 the frame. The Label field is set to the upstream-assigned label 1424 that PE1 has assigned to BD1. 1426 o The resulting MPLS packet is then encapsulated in a BIER 1427 encapsulation ([RFC8296], [EVPN-BIER]). The BIER BitString is set 1428 to identify the egress EVPN-PEs. The BIER "proto" field is set to 1429 the value for "MPLS packet with upstream-assigned label at top of 1430 stack". 1432 Note: It is possible that the packet being tunneled from PE1 1433 originated outside the Tenant Domain. In this case, the actual 1434 source BD (BD1) is considered to be the SBD, and the 1435 upstream-assigned label it carries will be the label that PE1 1436 assigned to the SBD, and advertised in its SBD-IMET route. 1438 Suppose an egress PE, PE2, receives such a BIER packet. The BFIR-id 1439 field of the BIER header allows PE2 to determine that the ingress PE 1440 is PE1. There are then two cases to consider: 1442 1. PE2 has received and installed an IMET route for BD1 from PE1. 1444 In this case, the BIER packet will be carrying the 1445 upstream-assigned label that is specified in the PTA of that IMET 1446 route. This enables PE2 to determine the "apparent source BD" 1447 (as defined in Section 2.4). 1449 2. PE2 has not received and installed an IMET route for BD1 from 1450 PE1. 1452 In this case, PE2 will not recognize the upstream-assigned label 1453 carried in the BIER packet. PE2 MUST discard the packet. 1455 Further details on the use of BIER to support EVPN can be found in 1456 [EVPN-BIER]. 1458 3.2.5. Inclusive P2MP Tunnels 1460 3.2.5.1. Using the BUM Tunnels as IP Multicast Inclusive Tunnels 1462 The procedures in this section apply only when 1464 (a) it is desired to use the BUM tunnels to carry IP multicast 1465 traffic across the backbone, and 1467 (b) the BUM tunnels are P2MP tunnels (i.e., neither IR, AR, nor BIER 1468 are being used to transport the BUM traffic). 1470 In this case, an IP multicast frame (whether inter-subnet or 1471 intra-subnet) will be carried across the backbone in the BUM tunnel 1472 belonging to its source BD. Each EVPN-PE attached to a given Tenant 1473 Domain needs to join the BUM tunnels for every BD in the Tenant 1474 Domain, even those BDs to which the EVPN-PE is not locally attached. 1475 This ensures that an IP multicast packet from any source BD can reach 1476 all PEs attached to the Tenant Domain. 1478 Note that this will cause all the BUM traffic from a given BD in a 1479 Tenant Domain to be sent to all PEs that attach to that Tenant 1480 Domain, even the PEs that don't attach to the given BD. To avoid 1481 this, it is RECOMMENDED that the BUM tunnels not be used as IP 1482 Multicast inclusive tunnels, and that the procedures of 1483 Section 3.2.5.2 be used instead. 1485 If a PE is a possible ingress EVPN-PE for traffic originating outside 1486 the Tenant Domain, the PE MUST originate an SBD-IMET route (see 1487 Section 3.2.1). This route MUST carry a PTA specifying the P2MP 1488 tunnel used for transmitting IP multicast packets that originate 1489 outside the tenant domain. All EVPN-PEs of the Tenant Domain MUST 1490 join the tunnel specified in the PTA of an SBD-IMET route: 1492 o If the tunnel is an RSVP-TE P2MP tunnel, the originator of the 1493 route MUST use RSVP-TE P2MP procedures to add each PE of the 1494 Tenant Domain to the tunnel, even PEs that have not originated an 1495 SBD-IMET route. 1497 o If the tunnel is an mLDP or PIM tunnel, each PE importing the 1498 SBD-IMET route MUST add itself to the tunnel, using mLDP or PIM 1499 procedures, respectively. 1501 Whether or not a PE originates an SBD-IMET route, it will of course 1502 originate an IMET route for each BD to which it is attached. Each of 1503 these IMET routes MUST carry the SBD-RT, as well as the RT for the BD 1504 to which it belongs. 1506 If a received IMET route is not the SBD-IMET route, it will also be 1507 carrying the RT for its source BD. The route's NLRI will carry the 1508 Tag ID for the source BD. From the RT and the Tag ID, any PE 1509 receiving the route can determine the route's source BD. 1511 If the MPLS label field of the PTA contains zero, the specified P2MP 1512 tunnel is used only to carry frames of a single source BD. 1514 If the MPLS label field of the PTA does not contain zero, it MUST 1515 contain an upstream-assigned MPLS label that maps uniquely (in the 1516 context of the originating EVPN-PE) to the source BD (or, in the case 1517 of an SBD-IMET route, to the SBD). The tunnel may then be used to 1518 carry frames of multiple source BDs. The apparent source BD of a 1519 particular packet is inferred from the label carried by the packet. 1521 IP multicast traffic originating outside the Tenant Domain is 1522 transmitted with the label corresponding to the SBD, as specified in 1523 the ingress EVPN-PE's SBD-IMET route. 1525 3.2.5.2. Using Wildcard S-PMSI A-D Routes to Advertise Inclusive 1526 Tunnels Specific to IP Multicast 1528 The procedures of this section apply when (and only when) it is 1529 desired to transmit IP multicast traffic on an inclusive tunnel, but 1530 not on the same tunnel used to transmit BUM traffic. 1532 However, these procedures do NOT apply when the tunnel type is 1533 Ingress Replication or BIER, EXCEPT in the case where it is necessary 1534 to interwork between non-OISM PEs and OISM PEs, as specified in 1535 Section 5. 1537 Each EVPN-PE attached to the given Tenant Domain MUST originate an 1538 SBD-SPMSI A-D route. The NLRI of that route MUST contain (C-*,C-*) 1539 (see [RFC6625]). Additional rules for constructing that route are 1540 given in Section 3.2.1. 1542 In addition, an EVPN-PE MUST originate an S-PMSI A-D route containing 1543 (C-*,C-*) in its NLRI for each of the other BDs, in the given Tenant 1544 Domain, to which it is attached. All such routes MUST carry the 1545 SBD-RT. This ensures that those routes are imported by all EVPN-PEs 1546 attached to the Tenant Domain. 1548 A PE receiving these routes follows the procedures of Section 2.2 to 1549 determine which BD the route is for. 1551 If the MPLS label field of the PTA contains zero, the specified 1552 tunnel is used only to carry frames of a single source BD. 1554 If the MPLS label field of the PTA does not contain zero, it MUST 1555 specify an upstream-assigned MPLS label that maps uniquely (in the 1556 context of the originating EVPN-PE) to the source BD. The tunnel may 1557 be used to carry frames of multiple source BDs, and the apparent 1558 source BD for a particular packet is inferred from the label carried 1559 by the packet. 1561 The EVPN-PE advertising these S-PMSI A-D route routes is specifying 1562 the default tunnel that it will use (as ingress PE) for transmitting 1563 IP multicast packets. The upstream-assigned label allows an egress 1564 PE to determine the apparent source BD of a given packet. 1566 3.2.6. Selective Tunnels 1568 An ingress EVPN-PE for a given multicast flow or set of flows can 1569 always assign the flow to a particular P2MP tunnel by originating an 1570 S-PMSI A-D route whose NLRI identifies the flow or set of flows. The 1571 NLRI of the route could be (C-*,C-G), or (C-S,C-G). The S-PMSI A-D 1572 route MUST carry the SBD-RT, so that it is imported by all EVPN-PEs 1573 attached to the Tenant Domain. 1575 An S-PMSI A-D route is "for" a particular source BD. It MUST carry 1576 the RT associated with that BD, and it MUST have the Tag ID for that 1577 BD in its NLRI. 1579 When an EVPN-PE imports an S-PMSI A-D route, it applies the rules of 1580 Section 2.2 to associate the route with a particular BD. 1582 Each such route MUST contain a PTA, as specified in Section 3.2.5.2. 1584 An egress EVPN-PE interested in the specified flow or flows MUST join 1585 the specified tunnel. Procedures for joining the specified tunnel 1586 are specific to the tunnel type. (Note that if the tunnel type is 1587 RSVP-TE P2MP LSP, the Leaf Information Required (LIR) flag of the PTA 1588 SHOULD NOT be set. An ingress OISM PE knows which OISM EVPN PEs are 1589 interested in any given flow, and hence can add them to the RSVP-TE 1590 P2MP tunnel that carries such flows.) 1592 If the PTA does not specify a non-zero MPLS label, the apparent 1593 source BD of any packets that arrive on that tunnel is considered to 1594 be the BD associated with the route that carries the PTA. If the PTA 1595 does specify a non-zero MPLS label, the apparent source BD of any 1596 packets that arrive on that tunnel carrying the specified label is 1597 considered to be the BD associated with the route that carries the 1598 PTA. 1600 It should be noted that when either IR or BIER is used, there is no 1601 need for an ingress PE to use S-PMSI A-D routes to assign specific 1602 flows to selective tunnels. The procedures of Section 3.3, along 1603 with the procedures of Section 3.2.2, Section 3.2.3, or 1604 Section 3.2.4, provide the functionality of selective tunnels without 1605 the need to use S-PMSI A-D routes. 1607 3.3. Advertising SMET Routes 1609 [IGMP-Proxy] allows an egress EVPN-PE to express its interest in a 1610 particular multicast flow or set of flows by originating an SMET 1611 route. The NLRI of the SMET route identifies the flow or set of 1612 flows as (C-*,C-*) or (C-*,C-G) or (C-S,C-G). 1614 Each SMET route belongs to a particular BD. The Tag ID for the BD 1615 appears in the NLRI of the route, and the route carries the RT 1616 associated that that BD. From this pair, other EVPN-PEs 1617 can identify the BD to which a received SMET route belongs. 1618 (Remember though that the route may be carrying multiple RTs.) 1619 There are three cases to consider: 1621 o Case 1: It is known that no BD of a Tenant Domain contains a 1622 multicast router. 1624 In this case, an egress PE advertises its interest in a flow or 1625 set of flows by originating an SMET route that belongs to the SBD. 1626 We refer to this as an SBD-SMET route. The SBD-SMET route carries 1627 the SBD-RT, and has the Tag ID for the SBD in its NLRI. SMET 1628 routes for the individual BDs are not needed, because there is no 1629 need for a PE that receives an SMET route to send a corresponding 1630 IGMP Join message out any of its ACs. 1632 o Case 2: It is known that more than one BD of a Tenant Domain may 1633 contain a multicast router. 1635 This is very like Case 1. An egress PE advertises its interest in 1636 a flow or set of flows by originating an SBD-SMET route. The 1637 SBD-SMET route carries the SBD-RT, and has the Tag ID for the SBD 1638 in its NLRI. 1640 In this case, it is important to be sure that SMET routes for the 1641 individual BDs are not originated. Suppose, for example, that PE1 1642 had local receivers for a given flow on both BD1 and BD2, and that 1643 it originated SMET routes for both those BDs. Then PEs receiving 1644 those SMET routes might send IGMP Joins on both those BDs. This 1645 could cause externally sourced multicast traffic to enter the 1646 Tenant Domain at both BDs, which could result in duplication of 1647 data. 1649 N.B.: If it is possible that more than one BD contains a tenant 1650 multicast router, then in order to receive multicast data 1651 originating from outside EVPN, the PEs MUST follow the procedures 1652 of Section 6. 1654 o Case 3: It is known that only a single BD of a Tenant Domain 1655 contains a multicast router. 1657 Suppose that an egress PE is attached to a BD on which there might 1658 be a tenant multicast router. (The tenant router is not 1659 necessarily on a segment that is attached to that PE.) And 1660 suppose that the PE has one or more ACs attached to that BD which 1661 are interested in a given multicast flow. In this case, IN 1662 ADDITION to the SMET route for the SBD, the egress PE MAY 1663 originate an SMET route for that BD. This will enable the ingress 1664 PE(s) to send IGMP/MLD messages on ACs for the BD, as specified in 1665 [IGMP-Proxy]. As long as that is the only BD on which there is a 1666 tenant multicast router, there is no possibility of duplication of 1667 data. 1669 This document does not specify procedures for dynamically determining 1670 which of the three cases applies to a given deployment; the PEs of a 1671 given Tenant Domain MUST be provisioned to know which case applies. 1673 As detailed in [IGMP-Proxy], an SMET route carries a Multicast Flags 1674 EC containing flags indicating whether it is to result in the 1675 propagation of IGMP v1, v2, or v3 messages on the ACs of the BD to 1676 which the SMET route belongs. These flags SHOULD be set to zero in 1677 an SBD-SMET route. 1679 Note that a PE only needs to originate the set of SBD-SMET routes 1680 that are needed to pull in all the traffic in which it is interested. 1681 Suppose PE1 has ACs attached to BD1 that are interested in (C-*,C-G) 1682 traffic, and ACs attached to BD2 that are interested in (C-S,C-G) 1683 traffic. A single SBD-SMET route specifying (C-*,C-G) will pull in 1684 all the necessary flows. 1686 As another example, suppose the ACs attached to BD1 are interested in 1687 (C-*,C-G) but not in (C-S,C-G), while the ACs attached to BD2 are 1688 interested in (C-S,C-G). A single SBD-SMET route specifying 1689 (C-*,C-G) will pull in all the necessary flows. 1691 In other words, to determine the set of SBD-SMET routes that have to 1692 be sent for a given C-G, the PE has to merge the IGMP/MLD state for 1693 all the BDs (of the given Tenant Domain) to which it is attached. 1695 Per [IGMP-Proxy], importing an SMET route for a particular BD will 1696 cause IGMP/MLD state to be instantiated for the IRB interface to that 1697 BD. This applies as well when the BD is the SBD. 1699 However, traffic that originates in one of the actual BDs of a 1700 particular Tenant Domain MUST NOT be sent down the IRB interface that 1701 connects the L3 routing instance of that Tenant Domain to the SBD. 1702 That would cause duplicate delivery of traffic, since such traffic 1703 will have already been distributed throughout the Tenant Domain. 1704 Therefore, when setting up the IGMP/MLD state based on SBD-SMET 1705 routes, care must be taken to ensure that the IRB interface to the 1706 SBD is not added to the Outgoing Interface (OIF) list if the traffic 1707 originates within the Tenant Domain. 1709 There are some multicast scenarios that make use of "anycast 1710 sources". For example, two different sources may share the same 1711 anycast IP address, say S1, and each may transmit an (S1,G) multicast 1712 flow. In such a scenario, the two (S1,G) flows are typically 1713 identical. Ordinary PIM procedures will cause only one the flows to 1714 be delivered to each receiver that has expressed interest in either 1715 (*,G) or (S1,G). However, the OISM procedures described in this 1716 document will result in both of the (S1,G) flows being distributed in 1717 the Tenant Domain, and duplicate delivery will result. Therefore, if 1718 there are receivers for (*,G) in a given Tenant Domain, there MUST 1719 NOT be anycast sources for G within that Tenant Domain. (This 1720 restriction can be lifted by defining additional procedures; however 1721 that is outside the scope of this document.) 1723 4. Constructing Multicast Forwarding State 1725 4.1. Layer 2 Multicast State 1727 An EVPN-PE maintains "layer 2 multicast state" for each BD to which 1728 it is attached. 1730 Let PE1 be an EVPN-PE, and BD1 be a BD to which it is attached. At 1731 PE1, BD1's layer 2 multicast state for a given (C-S,C-G) or (C-*,C-G) 1732 governs the disposition of an IP multicast packet that is received by 1733 BD1's layer 2 multicast function on an EVPN-PE. 1735 An IP multicast (S,G) packet is considered to have been received by 1736 BD1's layer 2 multicast function in PE1 in the following cases: 1738 o The packet is the payload of an ethernet frame received by PE1 1739 from an AC that attaches to BD1. 1741 o The packet is the payload of an ethernet frame whose apparent 1742 source BD is BD1, and which is received by the PE1 over a tunnel 1743 from another EVPN-PE. 1745 o The packet is received from BD1's IRB interface (i.e., has been 1746 transmitted by PE1's L3 routing instance down BD1's IRB 1747 interface). 1749 According to the procedures of this document, all transmission of IP 1750 multicast packets from one EVPN-PE to another is done at layer 2. 1751 That is, the packets are transmitted as ethernet frames, according to 1752 the layer 2 multicast state. 1754 Each layer 2 multicast state (S,G) or (*,G) contains a set "output 1755 interfaces" (OIF list). The disposition of an (S,G) multicast frame 1756 received by BD1's layer 2 multicast function is determined as 1757 follows: 1759 o The OIF list is taken from BD1's layer 2 (S,G) state, or if there 1760 is no such (S,G) state, then from BD1's (*,G) state. (If neither 1761 state exists, the OIF list is considered to be null.) 1763 o The rules of Section 4.1.2 are applied to the OIF list. This will 1764 generally result in the frame being transmitted to some, but not 1765 all, elements of the OIF list. 1767 Note that there is no RPF check at layer 2. 1769 4.1.1. Constructing the OIF List 1771 In this document, we have extended the procedures of [IGMP-Proxy] so 1772 that IMET and SMET routes for a particular BD are distributed not 1773 just to PEs that attach to that BD, but to PEs that attach to any BD 1774 in the Tenant Domain. In this way, each PE attached to a given 1775 Tenant Domain learns, from each other PE attached to the same Tenant 1776 Domain, the set of flows that are of interest to each of those other 1777 PEs. (If some PE attached to the Tenant Domain does not support 1778 [IGMP-Proxy], it will be assumed to be interested in all flows. 1779 Whether a particular remote PE supports [IGMP-Proxy] is determined by 1780 the presence of an Extended Community in its IMET route; this is 1781 specified in [IGMP-Proxy].) If a set of remote PEs are interested in 1782 a particular flow, the tunnels used to reach those PEs are added to 1783 the OIF list of the multicast states corresponding to that flow. 1785 An EVPN-PE may run IGMP/MLD procedures on each of its ACs, in order 1786 to determine the set of flows of interest to each AC. (An AC is said 1787 to be interested in a given flow if it connects to a segment that has 1788 tenant systems interested in that flow.) If IGMP/MLD procedures are 1789 not being run on a given AC, that AC is considered to be interested 1790 in all flows. For each BD, the set of ACs interested in a given flow 1791 is determined, and the ACs of that set are added to the OIF list of 1792 that BD's multicast state for that flow. 1794 The OIF list for each multicast state must also contain the IRB 1795 interface for the BD to which the state belongs. 1797 Implementors should note that the OIF list of a multicast state will 1798 change from time to time as ACs and/or remote PEs either become 1799 interested in, or lose interest in, particular multicast flows. 1801 4.1.2. Data Plane: Applying the OIF List to an (S,G) Frame 1803 When an (S,G) multicast frame is received by the layer 2 multicast 1804 function of a given EVPN-PE, say PE1, its disposition depends (a) the 1805 way it was received, (b) upon the OIF list of the corresponding 1806 multicast state (see Section 4.1.1), (c) upon the "eligibility" of an 1807 AC to receive a given frame (see Section 4.1.2.1 and (d) upon its 1808 apparent source BD (see Section 3.2 for information about determining 1809 the apparent source BD of a frame received over a tunnel from another 1810 PE). 1812 4.1.2.1. Eligibility of an AC to Receive a Frame 1814 A given (S,G) multicast frame is eligible to be transmitted by a 1815 given PE, say PE1, on a given AC, say AC1, only if one of the 1816 following conditions holds: 1818 1. ESI labels are being used, PE1 is the DF for the segment to which 1819 AC1 is connected, and the frame did not originate from that same 1820 segment (as determined by the ESI label), or 1822 2. The ingress PE for the frame is a remote PE, say PE2, local bias 1823 is being used, and PE2 is not connected to the same segment as 1824 AC1. 1826 4.1.2.2. Applying the OIF List 1828 Assume a given (S,G) multicast frame has been received by a given PE, 1829 say PE1. PE1 determines the apparent source BD of the frame, finds 1830 the layer 2 (S,G) state for that BD (or the (*,G) state if there is 1831 no (S,G) state), and takes the OIF list from that state. (Note that 1832 if PE1 is not attached to the actual source BD, the apparent source 1833 BD will be the SBD.) 1835 Suppose PE1 has determined the frame's apparent source BD to be BD1 1836 (which may or may not be the SBD.) There are the following cases to 1837 consider: 1839 1. The frame was received by PE1 from a local AC, say AC1, that 1840 attaches to BD1. 1842 a. The frame MUST be sent out all local ACs of BD1 that appear 1843 in the OIF list, except for AC1 itself. 1845 b. The frame MUST also be delivered to any other EVPN-PEs that 1846 have interest in it. This is achieved as follows: 1848 i. If (a) AR is being used, and (b) PE1 is an AR-LEAF, and 1849 (c) the OIF list is non-null, PE1 MUST send the frame 1850 to the AR-REPLICATOR. 1852 ii. Otherwise the frame MUST be sent on all tunnels in the 1853 OIF list. 1855 c. The frame MUST be sent to the local L3 routing instance by 1856 being sent up the IRB interface of BD1. It MUST NOT be sent 1857 up any other IRB interfaces. 1859 2. The frame was received by PE1 over a tunnel from another PE. 1860 (See Section 3.2 for the rules to determine the apparent source 1861 BD of a packet received from another PE. Note that if PE1 is not 1862 attached to the source BD, it will regard the SBD as the apparent 1863 source BD.) 1865 a. The frame MUST be sent out all local ACs in the OIF list that 1866 connect to BD1 and that are eligible (per Section 4.1.2.1) to 1867 receive the frame. 1869 b. The frame MUST be sent up the IRB interface of the apparent 1870 source BD. (Note that this may be the SBD.) The frame MUST 1871 NOT be sent up any other IRB interfaces. 1873 c. If PE1 is not an AR-REPLICATOR, it MUST NOT send the frame to 1874 any other EVPN-PEs. However, if PE1 is an AR-REPLICATOR, it 1875 MUST send the frame to all tunnels in the OIF list, except 1876 for the tunnel over which the frame was received. 1878 3. The frame was received by PE1 from the BD1 IRB interface (i.e., 1879 the frame has been transmitted by PE1's L3 routing instance down 1880 the BD1 IRB interface), and BD1 is NOT the SBD. 1882 a. The frame MUST be sent out all local ACs in the OIF list that 1883 are eligible (per Section 4.1.2.1 to receive the frame. 1885 b. The frame MUST NOT be sent to any other EVPN-PEs. 1887 c. The frame MUST NOT be sent up any IRB interfaces. 1889 4. The frame was received from the SBD IRB interface (i.e., has been 1890 transmitted by PE1's L3 routing instance down the SBD IRB 1891 interface). 1893 a. The frame MUST be sent on all tunnels in the OIF list. This 1894 causes the frame to be delivered to any other EVPN-PEs that 1895 have interest in it. 1897 b. The frame MUST NOT be sent on any local ACs. 1899 c. The frame MUST NOT be sent up any IRB interfaces. 1901 4.2. Layer 3 Forwarding State 1903 If an EVPN-PE is performing IGMP/MLD procedures on the ACs of a given 1904 BD, it processes those messages at layer 2 to help form the layer 2 1905 multicast state. If also sends those messages up that BD's IRB 1906 interface to the L3 routing instance of a particular tenant domain. 1908 This causes layer 2 (C-S,C-G) or (C-*,C-G) L3 state to be created/ 1909 updated. 1911 A layer 3 multicast state has both an Input Interface (IIF) and an 1912 OIF list. 1914 To set the IIF of an (C-S,C-G) state, the EVPN-PE must determine the 1915 source BD of C-S. This is done by looking up S in the local 1916 MAC-VRF(s) of the given Tenant Domain. 1918 If the source BD is present on the PE, the IIF is set to the IRB 1919 interface that attaches to that BD. Otherwise the IIF is set to the 1920 SBD IRB interface. 1922 For (C-*,C-G) states, traffic can arrive from any BD, so the IIF 1923 needs to be set to a wildcard value meaning "any IRB interface". 1925 The OIF list of these states includes one or more of the IRB 1926 interfaces of the Tenant Domain. In general, maintenance of the OIF 1927 list does not require any EVPN-specific procedures. However, there 1928 is one EVPN-specific rule: 1930 If the IIF is one of the IRB interfaces (or the wild card meaning 1931 "any IRB interface"), then the SBD IRB interface MUST NOT be added 1932 to the OIF list. Traffic originating from within a particular 1933 EVPN Tenant Domain must not be sent down the SBD IRB interface, as 1934 such traffic has already been distributed to all EVPN-PEs attached 1935 to that Tenant Domain. 1937 Please also see Section 6.1.1, which states a modification of this 1938 rule for the case where OISM is interworking with external Layer 3 1939 multicast routing. 1941 5. Interworking with non-OISM EVPN-PEs 1943 It is possible that a given Tenant Domain will be attached to both 1944 OISM PEs and non-OISM PEs. Inter-subnet IP multicast should be 1945 possible and fully functional even if not all PEs attaching to a 1946 Tenant Domain can be upgraded to support OISM functionality. 1948 Note that the non-OISM PEs are not required to have IRB support, or 1949 support for [IGMP-Proxy]. It is however advantageous for the 1950 non-OISM PEs to support [IGMP-Proxy]. 1952 In this section, we will use the following terminology: 1954 o PE-S: the ingress PE for an (S,G) flow. 1956 o PE-R: an egress PE for an (S,G) flow. 1958 o BD-S: the source BD for an (S,G) flow. PE-S must have one or more 1959 ACs attached BD-S, at least one of which attaches to host S. 1961 o BD-R: a BD that contains a host interested in the flow. The host 1962 is attached to PE-R via an AC that belongs to BD-R. 1964 To allow OISM PEs to interwork with non-OISM PEs, a given Tenant 1965 Domain needs to contain one or more "IP Multicast Gateways" (IPMGs). 1966 An IPMG is an OISM PE with special responsibilities regarding the 1967 interworking between OISM and non-OISM PEs. 1969 If a PE is functioning as an IPMG, it MUST signal this fact by 1970 setting the "IPMG" flag in the Multicast Flags EC that it attaches to 1971 its IMET routes. An IPMG SHOULD attach this EC with the IPMG flag 1972 set to all IMET routes it originates. However, if PE1 imports any 1973 IMET route from PE2 that has the EC present with the "IPMG" flag set, 1974 then the PE1 will assume that PE2 is an IPMG. 1976 An IPMG Designated Forwarder (IPMG-DF) selection procedure is used to 1977 ensure that, at any given time, there is exactly one active IPMG-DF 1978 for any given BD. Details of the IPMG-DF selection procedure are in 1979 Section 5.1. The IPMG-DF for a given BD, say BD-S, has special 1980 functions to perform when it receives (S,G) frames on that BD: 1982 o If the frames are from a non-OISM PE-S: 1984 * The IPMG-DF forwards them to OISM PEs that do not attach to 1985 BD-S but have interest in (S,G). 1987 Note that OISM PEs that do attach to BD-S will have received 1988 the frames on the BUM tunnel from the non-OISM PE-S. 1990 * The IPMG-DF forwards them to non-OISM PEs that have interest in 1991 (S,G) on ACs that do not belong to BD-S. 1993 Note that if a non-OISM PE has multiple BDs other than BD-S 1994 with interest in (S,G), it will receive one copy of the frame 1995 for each such BD. This is necessary because the non-OISM PEs 1996 cannot move IP multicast traffic from one BD to another. 1998 o If the frames are from an OISM PE, the IPMG-DF forwards them to 1999 non-OISM PEs that have interest in (S,G) on ACs that do not belong 2000 to BD-S. 2002 If a non-OISM PE has interest in (S,G) on an AC belonging to BD-S, 2003 it will have received a copy of the (S,G) frame, encapsulated for 2004 BD-S, from the OISM PE-S. (See Section 3.2.2.) If the non-OISM 2005 PE has interest in (S,G) on one or more ACs belonging to 2006 BD-R1,...,BD-Rk where the BD-Ri are distinct from BD-S, the 2007 IPMG-DF needs to send it a copy of the frame for BD-Ri. 2009 If an IPMG receives a frame on a BD for which it is not the IPMG-DF, 2010 it just follows normal OISM procedures. 2012 This section specifies several sets of procedures: 2014 o the procedures that the IPMG-DF for a given BD needs to follow 2015 when receiving, on that BD, an IP multicast frame from a non-OISM 2016 PE; 2018 o the procedures that the IPMG-DF for a given BD needs to follow 2019 when receiving, on that BD, an IP multicast frame from an OISM PE; 2021 o the procedures that an OISM PE needs to follow when receiving, on 2022 a given BD, an IP multicast frame from a non-OISM PE, when the 2023 OISM PE is not the IPMG-DF for that BD. 2025 To enable OISM/non-OISM interworking in a given Tenant Domain, the 2026 Tenant Domain MUST have some EVPN-PEs that can function as IPMGs. An 2027 IPMG must be configured with the SBD. It must also be configured 2028 with every BD of the Tenant Domain that exists on any of the non-OISM 2029 PEs of that domain. (Operationally, it may be simpler to configure 2030 the IPMG with all the BDs of the Tenant Domain.) 2032 A non-OISM PE of course only needs to be configured with BDs for 2033 which it has ACs. An OISM PE that is not an IPMG only needs to be 2034 configured with the SBD and with the BDs for which it has ACs. 2036 An IPMG MUST originate a wildcard SMET route (with (C-*,C-*) in the 2037 NLRI) for each BD in the Tenant Domain. This will cause it to 2038 receive all the IP multicast traffic that is sourced in the Tenant 2039 Domain. Note that non-OISM nodes that do not support [IGMP-Proxy] 2040 will send all the multicast traffic from a given BD to all PEs 2041 attached to that BD, even if those PEs do not originate an SMET 2042 route. 2044 The interworking procedures vary somewhat depending upon whether 2045 packets are transmitted from PE to PE via Ingress Replication (IR) or 2046 via Point-to-Multipoint (P2MP) tunnels. We do not consider the use 2047 of BIER in this section, due to the low likelihood of there being a 2048 non-OISM PE that supports BIER. 2050 5.1. IPMG Designated Forwarder 2052 Every PE that is eligible for selection as an IPMG-DF for a 2053 particular BD originates both an IMET route for that BD and an 2054 SBD-IMET route. As stated in Section 5, these SBD-IMET routes carry 2055 a Multicast Flags EC with the IPMG Flag set. 2057 These SBD-IMET routes SHOULD also carry a DF Election EC. The DF 2058 Election EC and its use is specified in ([DF-Election-Framework]). 2059 When the route is originated, the AC-DF bit in the DF Election EC 2060 SHOULD be set to zero. This bit is not used when selecting an 2061 IPMSG-DF, i.e., it MUST be ignored by the receiver of an SBD-IMET 2062 route. 2064 In the context of a given Tenant Domain, to select the IPMG-DF for a 2065 particular BD, say BD1, the IPMGs of the Tenant Domain perform the 2066 following procedure: 2068 o From the set of received SBD-IMET routes for the given tenant 2069 domain, determine the candidate set of PEs that support IPMG 2070 functionality for that domain. 2072 o Eliminate from that candidate set any PEs from which an IMET route 2073 for BD1 has not been received. 2075 o Select a DF Election algorithm as specified in 2076 [DF-Election-Framework]. Some of the possible algorithms can be 2077 found, e.g., in [DF-Election-Framework], [RFC7432], and 2078 [EVPN-DF-WEIGHTED]. 2080 o Apply the DF Election Algorithm (see [DF-Election-Framework]) to 2081 the candidate set of PEs. The "winner' becomes the IPMG-DF for 2082 BD1. 2084 Note that even if a given PE supports MEG (Section 6.1.2) and/or PEG 2085 (Section 6.1.4) functionality, as well as IPMG functionality, its 2086 SBD-IMET routes carry only one DF Election EC. 2088 5.2. Ingress Replication 2090 The procedures of this section are used when Ingress Replication is 2091 used to transmit packets from one PE to another. 2093 When a non-OISM PE-S transmits a multicast frame from BD-S to another 2094 PE, PE-R, PE-S will use the encapsulation specified in the BD-S IMET 2095 route that was originated by PE-R. This encapsulation will include 2096 the label that appears in the "MPLS label" field of the PMSI Tunnel 2097 attribute (PTA) of the IMET route. If the tunnel type is VXLAN, the 2098 "label" is actually a Virtual Network Identifier (VNI); for other 2099 tunnel types, the label is an MPLS label. In either case, we will 2100 speak of the transmitted frames as carrying a label that was assigned 2101 to a particular BD by the PE-R to which the frame is being 2102 transmitted. 2104 To support OISM/non-OISM interworking, an OISM PE-R MUST originate, 2105 for each of its BDs, both an IMET route and an S-PMSI (C-*,C-*) A-D 2106 route. Note that even when IR is being used, interworking between 2107 OISM and non-OISM PEs requires the OISM PEs to follow the rules of 2108 Section 3.2.5.2, as modified below. 2110 Non-OISM PEs will not understand S-PMSI A-D routes. So when a 2111 non-OISM PE-S transmits an IP multicast frame with a particular 2112 source BD to an IPMG, it encapsulates the frame using the label 2113 specified in that IPMG's BD-S IMET route. (This is just the 2114 procedure of [RFC7432].) 2116 The (C-*,C-*) S-PMSI A-D route originated by a given OISM PE will 2117 have a PTA that specifies IR. 2119 o If MPLS tunneling is being used, the MPLS label field SHOULD 2120 contain a non-zero value, and the LIR flag SHOULD be zero. (The 2121 case where the MPLS label field is zero or the LIR flag is set is 2122 outside the scope of this document.) 2124 o If the tunnel encapsulation is VXLAN, the MPLS label field MUST 2125 contain a non-zero value, and the LIR flag MUST be zero. 2127 When an OISM PE-S transmits an IP multicast frame to an IPMG, it will 2128 use the label specified in that IPMG's (C-*,C-*) S-PMSI A-D route. 2130 When a PE originates both an IMET route and a (C-*,C-*) S-PMSI A-D 2131 route, the values of the MPLS label field in the respective PTAs must 2132 be distinct. Further, each MUST map uniquely (in the context of the 2133 originating PE) to the route's BD. 2135 As a result, an IPMG receiving an MPLS-encapsulated IP multicast 2136 frame can always tell by the label whether the frame's ingress PE is 2137 an OISM PE or a non-OISM PE. When an IPMG receives a VXLAN- 2138 encapsulated IP multicast frame it may need to determine the identity 2139 of the ingress PE from the outer IP encapsulation; it can then 2140 determine whether the ingress PE is an OISM PE or a non-OISM PE by 2141 looking the IMET route from that PE. 2143 Suppose an IPMG receives an IP multicast frame from another EVPN-PE 2144 in the Tenant Domain, and the IPMG is not the IPMG-DF for the frame's 2145 source BD. Then the IPMG performs only the ordinary OISM functions; 2146 it does not perform the IPMG-specific functions for that frame. In 2147 the remainder of this section, when we discuss the procedures applied 2148 by an IPMG when it receives an IP multicast frame, we are presuming 2149 that the source BD of the frame is a BD for which the IPMG is the 2150 IPMG-DF. 2152 We have two basic cases to consider: (1) a frame's ingress PE is a 2153 non-OISM node, and (2) a frame's ingress PE is an OISM node. 2155 5.2.1. Ingress PE is non-OISM 2157 In this case, a non-OISM PE, PE-S, has received an (S,G) multicast 2158 frame over an AC that is attached to a particular BD, BD-S. By 2159 virtue of normal EVPN procedures, PE-S has sent a copy of the frame 2160 to every PE-R (both OISM and non-OISM) in the Tenant Domain that is 2161 attached to BD-S. If the non-OISM node supports [IGMP-Proxy], only 2162 PEs that have expressed interest in (S,G) receive the frame. The 2163 IPMG will have expressed interest via a (C-*,C-*) SMET route and thus 2164 receives the frame. 2166 Any OISM PE (including an IPMG) receiving the frame will apply normal 2167 OISM procedures. As a result it will deliver the frame to any of its 2168 local ACs (in BD-S or in any other BD) that have interest in (S,G). 2170 An OISM PE that is also the IPMG-DF for a particular BD, say BD-S, 2171 has additional procedures that it applies to frames received on BD-S 2172 from non-OISM PEs: 2174 1. When the IPMG-DF for BD-S receives an (S,G) frame from a 2175 non-OISM node, it MUST forward a copy of the frame to every OISM 2176 PE that is NOT attached to BD-S but has interest in (S,G). The 2177 copy sent to a given OISM PE-R must carry the label that PE-R 2178 has assigned to the SBD in an S-PMSI A-D route. The IPMG MUST 2179 NOT do any IP processing of the frame's IP payload. TTL 2180 decrement and other IP processing will be done by PE-R, per the 2181 normal OISM procedures. There is no need for the IPMG to 2182 include an ESI label in the frame's tunnel encapsulation, 2183 because it is already known that the frame's source BD has no 2184 presence on PE-R. There is also no need for the IPMG to modify 2185 the frame's MAC SA. 2187 2. In addition, when the IPMG-DF for BD-S receives an (S,G) frame 2188 from a non-OISM node, it may need to forward copies of the frame 2189 to other non-OISM nodes. Before it does so, it MUST decapsulate 2190 the (S,G) packet, and do the IP processing (e.g., TTL 2191 decrement). Suppose PE-R is a non-OISM node that has an AC to 2192 BD-R, where BD-R is not the same as BD-S, and that AC has 2193 interest in (S,G). The IPMG must then encapsulate the (S,G) 2194 packet (after the IP processing has been done) in an ethernet 2195 header. The MAC SA field will have the MAC address of the 2196 IPMG's IRB interface to BD-R. The IPMG then sends the frame to 2197 PE-R. The tunnel encapsulation will carry the label that PE-R 2198 advertised in its IMET route for BD-R. There is no need to 2199 include an ESI label, as the source and destination BDs are 2200 known to be different. 2202 Note that if a non-OISM PE-R has several BDs (other than BD-S) 2203 with local ACs that have interest in (S,G), the IPMG will send 2204 it one copy for each such BD. This is necessary because the 2205 non-OISM PE cannot move packets from one BD to another. 2207 There may be deployment scenarios in which every OISM PE is 2208 configured with every BD that is present on any non-OISM PE. In such 2209 scenarios, the procedures of item 1 above will not actually result in 2210 the transmission of any packets. Hence if it is known a priori that 2211 this deployment scenario exists for a given tenant domain, the 2212 procedures of item 1 above can be disabled. 2214 5.2.2. Ingress PE is OISM 2216 In this case, an OISM PE, PE-S, has received an (S,G) multicast frame 2217 over an AC that attaches to a particular BD, BD-S. 2219 By virtue of receiving all the IMET routes about BD-S, PE-S will know 2220 all the PEs attached to BD-S. By virtue of normal OISM procedures: 2222 o PE-S will send a copy of the frame to every OISM PE-R (including 2223 the IPMG) in the Tenant Domain that is attached to BD-S and has 2224 interest in (S,G). The copy sent to a given PE-R carries the 2225 label that that the PE-R has assigned to BD-S in its (C-*,C-*) 2226 S-PMSI A-D route. 2228 o PE-S will also transmit a copy of the (S,G) frame to every OISM 2229 PE-R that has interest in (S,G) but is not attached to BD-S. The 2230 copy will contain the label that the PE-R has assigned to the SBD. 2231 (As in Section 5.2.1, an IPMG is assumed to have indicated 2232 interest in all multicast flows.) 2234 o PE-S will also transmit a copy of the (S,G) frame to every 2235 non-OISM PE-R that is attached to BD-S. It does this using the 2236 label advertised by that PE-R in its IMET route for BD-S. 2238 The PE-Rs follow their normal procedures. An OISM PE that receives 2239 the (S,G) frame on BD-S applies the OISM procedures to deliver the 2240 frame to its local ACs, as necessary. A non-OISM PE that receives 2241 the (S,G) frame on BD-S delivers the frame only to its local BD-S 2242 ACs, as necessary. 2244 Suppose that a non-OISM PE-R has interest in (S,G) on a BD, BD-R, 2245 that is different than BD-S. If the non-OISM PE-R is attached to 2246 BD-S, the OISM PE-S will send forward it the original (S,G) multicast 2247 frame, but the non-OISM PE-R will not be able to send the frame to 2248 ACs that are not in BD-S. If PE-R is not even attached to BD-S, the 2249 OISM PE-S will not send it a copy of the frame at all, because PE-R 2250 is not attached to the SBD. In these cases, the IPMG needs to relay 2251 the (S,G) multicast traffic from OISM PE-S to non-OISM PE-R. 2253 When the IPMG-DF for BD-S receives an (S,G) frame from an OISM PE-S, 2254 it has to forward it to every non-OISM PE-R that that has interest in 2255 (S,G) on a BD-R that is different than BD-S. The IPMG MUST 2256 decapsulate the IP multicast packet, do the IP processing, re- 2257 encapsulate it for BD-R (changing the MAC SA to the IPMG's own MAC 2258 address on BD-R), and send a copy of the frame to PE-R. Note that a 2259 given non-OISM PE-R will receive multiple copies of the frame, if it 2260 has multiple BDs on which there is interest in the frame. 2262 5.3. P2MP Tunnels 2264 When IR is used to distribute the multicast traffic among the 2265 EVPN-PEs, the procedures of Section 5.2 ensure that there will be no 2266 duplicate delivery of multicast traffic. That is, no egress PE will 2267 ever send a frame twice on any given AC. If P2MP tunnels are being 2268 used to distribute the multicast traffic, it is necessary have 2269 additional procedures to prevent duplicate delivery. 2271 At the present time, it is not clear that there will be a use case in 2272 which OISM nodes need to interwork with non-OISM nodes that use P2MP 2273 tunnels. If it is determined that there is such a use case, 2274 procedures for it will be included in a future revision of this 2275 document. 2277 6. Traffic to/from Outside the EVPN Tenant Domain 2279 In this section, we discuss scenarios where a multicast source 2280 outside a given EVPN Tenant Domain sends traffic to receivers inside 2281 the domain (as well as, possibly, to receivers outside the domain). 2282 This requires the OISM procedures to interwork with various layer 3 2283 multicast routing procedures. 2285 We assume in this section that the Tenant Domain is not being used as 2286 an intermediate transit network for multicast traffic; that is, we do 2287 not consider the case where the Tenant Domain contains multicast 2288 routers that will receive traffic from sources outside the domain and 2289 forward the traffic to receivers outside the domain. The transit 2290 scenario is considered in Section 7. 2292 We can divide the non-transit scenarios into two classes: 2294 1. One or more of the EVPN PE routers provide the functionality 2295 needed to interwork with layer 3 multicast routing procedures. 2297 2. A single BD in the Tenant Domain contains external multicast 2298 routers ("tenant multicast routers"), and those tenant multicast 2299 routers are used to interwork, on behalf of the entire Tenant 2300 Domain, with layer 3 multicast routing procedures. 2302 6.1. Layer 3 Interworking via EVPN OISM PEs 2304 6.1.1. General Principles 2306 Sometimes it is necessary to interwork an EVPN Tenant Domain with an 2307 external layer 3 multicast domain (the "external domain"). This is 2308 needed to allow EVPN tenant systems to receive multicast traffic from 2309 sources ("external sources") outside the EVPN Tenant Domain. It is 2310 also needed to allow receivers ("external receivers") outside the 2311 EVPN Tenant Domain to receive traffic from sources inside the Tenant 2312 Domain. 2314 In order to allow interworking between an EVPN Tenant Domain and an 2315 external domain, one or more OISM PEs must be "L3 Gateways". An L3 2316 Gateway participates both in the OISM procedures and in the L3 2317 multicast routing procedures of the external domain. 2319 An L3 Gateway that has interest in receiving (S,G) traffic must be 2320 able to determine the best route to S. If an L3 Gateway has interest 2321 in (*,G), it must be able to determine the best route to G's RP. In 2322 these interworking scenarios, the L3 Gateway must be running a layer 2323 3 unicast routing protocol. Via this protocol, it imports unicast 2324 routes (either IP routes or VPN-IP routes) from routers other than 2325 EVPN PEs. And since there may be multicast sources inside the EVPN 2326 Tenant Domain, the EVPN PEs also need to export, either as IP routes 2327 or as VPN-IP routes (depending upon the external domain), unicast 2328 routes to those sources. 2330 When selecting the best route to a multicast source or RP, an L3 2331 Gateway might have a choice between an EVPN route and an IP/VPN-IP 2332 route. When such a choice exists, the L3 Gateway SHOULD always 2333 prefer the EVPN route. This will ensure that when traffic originates 2334 in the Tenant Domain and has a receiver in the Tenant Domain, the 2335 path to that receiver will remain within the EVPN Tenant Domain, even 2336 if the source is also reachable via a routed path. This also 2337 provides protection against sub-optimal routing that might occur if 2338 two EVPN PEs export IP/VPN-IP routes and each imports the other's IP/ 2339 VPN-IP routes. 2341 Section 4.2 discusses the way layer 3 multicast states are 2342 constructed by OISM PEs. These layer 3 multicast states have IRB 2343 interfaces as their IIF and OIF list entries, and are the basis for 2344 interworking OISM with other layer 3 multicast procedures such as 2345 MVPN or PIM. From the perspective of the layer 3 multicast 2346 procedures running in a given L3 Gateway, an EVPN Tenant Domain is a 2347 set of IRB interfaces. 2349 When interworking an EVPN Tenant Domain with an external domain, the 2350 L3 Gateway's layer 3 multicast states will not only have IRB 2351 interfaces as IIF and OIF list entries, but also other "interfaces" 2352 that lead outside the Tenant Domain. For example, when interworking 2353 with MVPN, the multicast states may have MVPN tunnels as well as IRB 2354 interfaces as IIF or OIF list members. When interworking with PIM, 2355 the multicast states may have PIM-enabled non-IRB interfaces as IIF 2356 or OIF list members. 2358 As long as a Tenant Domain is not being used as an intermediate 2359 transit network for IP multicast traffic, it is not necessary to 2360 enable PIM on its IRB interfaces. 2362 In general, an L3 Gateway has the following responsibilities: 2364 o It exports, to the external domain, unicast routes to those 2365 multicast sources in the EVPN Tenant Domain that are locally 2366 attached to the L3 Gateway. 2368 o It imports, from the external domain, unicast routes to multicast 2369 sources that are in the external domain. 2371 o It executes the procedures necessary to draw externally sourced 2372 multicast traffic that is of interest to locally attached 2373 receivers in the EVPN Tenant Domain. When such traffic is 2374 received, the traffic is sent down the IRB interfaces of the BDs 2375 on which the locally attached receivers reside. 2377 One of the L3 Gateways in a given Tenant Domain becomes the "DR" for 2378 the SBD. (See Section 6.1.2.4.) This L3 gateway has the following 2379 additional responsibilities: 2381 o It exports, to the external domain, unicast routes to multicast 2382 sources that in the EVPN Tenant Domain that are not locally 2383 attached to any L3 gateway. 2385 o It imports, from the external domain, unicast routes to multicast 2386 sources that are in the external domain. 2388 o It executes the procedures necessary to draw externally sourced 2389 multicast traffic that is of interest to receivers in the EVPN 2390 Tenant Domain that are not locally attached to an L3 gateway. 2391 When such traffic is received, the traffic is sent down the SBD 2392 IRB interface. OISM procedures already described in this document 2393 will then ensure that the IP multicast traffic gets distributed 2394 throughout the Tenant Domain to any EVPN PEs that have interest in 2395 it. Thus to an OISM PE that is not an L3 gateway the externally 2396 sourced traffic will appear to have been sourced on the SBD. 2398 In order for this to work, some special care is needed when an L3 2399 gateway creates or modifies a layer 3 (*,G) multicast state. Suppose 2400 group G has both external sources (sources outside the EVPN Tenant 2401 Domain) and internal sources (sources inside the EVPN tenant domain). 2402 Section 4.2 states that when there are internal sources, the SBD IRB 2403 interface must not be added to the OIF list of the (*,G) state. 2404 Traffic from internal sources will already have been delivered to all 2405 the EVPN PEs that have interest in it. However, if the OIF list of 2406 the (*,G) state does not contain its SBD IRB interface, then traffic 2407 from external sources will not get delivered to other EVPN PEs. 2409 One way of handling this is the following. When a L3 gateway 2410 receives (S,G) traffic from other than an IRB interface, and the 2411 traffic corresponds to a layer 3 (*,G) state, the L3 gateway can 2412 create (S,G) state. The IIF will be set to the external interface 2413 over which the traffic is expected. The OIF list will contain the 2414 SBD IRB interface, as well as the IRB interfaces of any other BDs 2415 attached to the PEG DR that have locally attached receivers with 2416 interest in the (S,G) traffic. The (S,G) state will ensure that the 2417 external traffic is sent down the SBD IRB interface. The following 2418 text will assume this procedure; however other implementation 2419 techniques may also be possible. 2421 If a particular BD is attached to several L3 Gateways, one of the L3 2422 Gateways becomes the DR for that BD. (See Section 6.1.2.4.) If the 2423 interworking scenario requires FHR functionality, it is generally the 2424 DR for a particular BD that is responsible for performing that 2425 functionality on behalf of the source hosts on that BD. (E.g., if 2426 the interworking scenario requires that PIM Register messages be sent 2427 by a FHR, the DR for a given BD would send the PIM Register messages 2428 for sources on that BD.) Note though that the DR for the SBD does 2429 not perform FHR functionality on behalf of external sources. 2431 An optional alternative is to have each L3 gateway perform FHR 2432 functionality for locally attached sources. Then the DR would only 2433 have to perform FHR functionality on behalf of sources that are 2434 locally attached to itself AND sources that are not attached to any 2435 L3 gateway. 2437 N.B.: If it is possible that more than one BD contains a tenant 2438 multicast router, then a PE receiving an SMET route for that BD MUST 2439 NOT reconstruct IGMP Join Reports from the SMET route, and MUST NOT 2440 transmit any such IGMP Join Reports on its local ACs attaching to 2441 that BD. Otherwise, multicast traffic may be duplicated. 2443 6.1.2. Interworking with MVPN 2445 In this section, we specify the procedures necessary to allow EVPN 2446 PEs running OISM procedures to interwork with L3VPN PEs that run BGP- 2447 based MVPN ([RFC6514]) procedures. More specifically, the procedures 2448 herein allow a given EVPN Tenant Domain to become part of an L3VPN/ 2449 MVPN, and support multicast flows where either: 2451 o The source of a given multicast flow is attached to an ethernet 2452 segment whose BD is part of an EVPN Tenant Domain, and one or more 2453 receivers of the flow are attached to the network via L3VPN/MVPN. 2454 (Other receivers may be attached to the network via EVPN.) 2456 o The source of a given multicast flow is attached to the network 2457 via L3VPN/MVPN, and one or more receivers of the flow are attached 2458 to an ethernet segment that is part of an EVPN tenant domain. 2459 (Other receivers may be attached via L3VPN/MVPN.) 2461 In this interworking model, existing L3VPN/MVPN PEs are unaware that 2462 certain sources or receivers are part of an EVPN Tenant Domain. The 2463 existing L3VPN/MVPN nodes run only their standard procedures and are 2464 entirely unaware of EVPN. Interworking is achieved by having some or 2465 all of the EVPN PEs function as L3 Gateways running L3VPN/MVPN 2466 procedures, as detailed in the following sub-sections. 2468 In this section, we assume that there are no tenant multicast routers 2469 on any of the EVPN-attached ethernet segments. (There may of course 2470 be multicast routers in the L3VPN.) Consideration of the case where 2471 there are tenant multicast routers is deferred till Section 7.) 2473 To support MVPN/EVPN interworking, we introduce the notion of an 2474 MVPN/EVPN Gateway, or MEG. 2476 A MEG is an L3 Gateway (see Section 6.1.1), hence is both an OISM PE 2477 and an L3VPN/MVPN PE. For a given EVPN Tenant Domain it will have an 2478 IP-VRF. If the Tenant Domain is part of an L3VPN/MVPN, the IP-VRF 2479 also serves as an L3VPN VRF ([RFC4364]). The IRB interfaces of the 2480 IP-VRF are considered to be "VRF interfaces" of the L3VPN VRF. The 2481 L3VPN VRF may also have other local VRF interfaces that are not EVPN 2482 IRB interfaces. 2484 The VRF on the MEG will import VPN-IP routes ([RFC4364]) from other 2485 L3VPN Provider Edge (PE) routers. It will also export VPN-IP routes 2486 to other L3VPN PE routers. In order to do so, it must be 2487 appropriately configured with the Route Targets used in the L3VPN to 2488 control the distribution of the VPN-IP routes. These Route Targets 2489 will in general be different than the Route Targets used for 2490 controlling the distribution of EVPN routes, as there is no need to 2491 distribute EVPN routes to L3VPN-only PEs and no reason to distribute 2492 L3VPN/MVPN routes to EVPN-only PEs. 2494 Note that the RDs in the imported VPN-IP routes will not necessarily 2495 conform to the EVPN rules (as specified in [RFC7432]) for creating 2496 RDs. Therefore a MEG MUST NOT expect the RDs of the VPN-IP routes to 2497 be of any particular format other than what is required by the L3VPN/ 2498 MVPN specifications. 2500 The VPN-IP routes that a MEG exports to L3VPN are subnet routes and/ 2501 or host routes for the multicast sources that are part of the EVPN 2502 tenant domain. The exact set of routes that need to be exported is 2503 discussed in Section 6.1.2.2. 2505 Each IMET route originated by a MEG SHOULD carry a Multicast Flags 2506 Extended Community with the "MEG" flag set, indicating that the 2507 originator of the IMET route is a MEG. However, PE1 will consider 2508 PE2 to be a MEG if PE1 imports at least one IMET route from PE2 that 2509 carries the Multicast Flags EC with the MEG flag set. 2511 All the MEGs of a given Tenant Domain attach to the SBD of that 2512 domain, and one of them is selected to be the SBD's Designated Router 2513 (the "MEG SBD-DR") for the domain. The selection procedure is 2514 discussed in Section 6.1.2.4. 2516 In this model of operation, MVPN procedures and EVPN procedures are 2517 largely independent. In particular, there is no assumption that MVPN 2518 and EVPN use the same kind of tunnels. Thus no special procedures 2519 are needed to handle the common scenarios where, e.g., EVPN uses 2520 VXLAN tunnels but MVPN uses MPLS P2MP tunnels, or where EVPN uses 2521 Ingress Replication but MVPN uses MPLS P2MP tunnels. 2523 Similarly, no special procedures are needed to prevent duplicate data 2524 delivery on ethernet segments that are multi-homed. 2526 The MEG does have some special procedures (described below) for 2527 interworking between EVPN and MVPN; these have to do with selection 2528 of the Upstream PE for a given multicast source, with the exporting 2529 of VPN-IP routes, and with the generation of MVPN C-multicast routes 2530 triggered by the installation of SMET routes. 2532 6.1.2.1. MVPN Sources with EVPN Receivers 2534 6.1.2.1.1. Identifying MVPN Sources 2536 Consider a multicast source S. It is possible that a MEG will import 2537 both an EVPN unicast route to S and a VPN-IP route (or an ordinary IP 2538 route), where the prefix length of each route is the same. In order 2539 to draw (S,G) multicast traffic for any group G, the MEG SHOULD use 2540 the EVPN route rather than the VPN-IP or IP route to determine the 2541 "Upstream PE" (see section 5 of [RFC6513]). 2543 Doing so ensures that when an EVPN tenant system desires to receive a 2544 multicast flow from another EVPN tenant system, the traffic from the 2545 source to that receiver stays within the EVPN domain. This prevents 2546 problems that might arise if there is a unicast route via L3VPN to S, 2547 but no multicast routers along the routed path. This also prevents 2548 problem that might arise as a result of the fact that the MEGs will 2549 import each others' VPN-IP routes. 2551 In the Section 6.1.2.1.2, we describe the procedures to be used when 2552 the selected route to S is a VPN-IP route. 2554 6.1.2.1.2. Joining a Flow from an MVPN Source 2556 Consider a tenant system, R, on a particular BD, BD-R. Suppose R 2557 wants to receive (S,G) multicast traffic, where source S is not 2558 attached to any PE in the EVPN Tenant Domain, but is attached to an 2559 MVPN PE. 2561 o Suppose R is on a singly homed ethernet segment of BD-R, and that 2562 segment is attached to PE1, where PE1 is a MEG. PE1 learns via 2563 IGMP/MLD listening that R is interested in (S,G). PE1 determines 2564 from its VRF that there is no route to S within the Tenant Domain 2565 (i.e., no EVPN RT-2 route with S's IP address), but that there is 2566 a route to S via L3VPN (i.e., the VRF contains a subnet or host 2567 route to S that was received as a VPN-IP route). PE1 thus 2568 originates (if it hasn't already) an MVPN C-multicast Source Tree 2569 Join(S,G) route. The route is constructed according to normal 2570 MVPN procedures. 2572 The layer 2 multicast state is constructed as specified in 2573 Section 4.1. 2575 In the layer 3 multicast state, the IIF is the appropriate MVPN 2576 tunnel, and the IRB interface to BD-R is added to the OIF list. 2578 When PE1 receives (S,G) traffic from the appropriate MVPN tunnel, 2579 it performs IP processing of the traffic, and then sends the 2580 traffic down its IRB interface to BD-R. Following normal OISM 2581 procedures, the (S,G) traffic will be encapsulated for ethernet 2582 and sent out the AC to which R is attached. 2584 o Suppose R is on a singly homed ethernet segment of BD-R, and that 2585 segment is attached to PE1, where PE1 is an OISM PE but is NOT a 2586 MEG. PE1 learns via IGMP/MLD listening that R is interested in 2587 (S,G). PE1 follows normal OISM procedures, originating an SBD- 2588 SMET route for (S,G); this route will be received by all the MEGs 2589 of the Tenant Domain, including the MEG SBD-DR. The MEG SBD-DR 2590 can determine from PE1's IMET routes whether PE1 is itself a MEG. 2591 If PE1 is not a MEG, the MEG SBD-DR will originate (if it hasn't 2592 already) an MVPN C-multicast Source Tree Join(S,G) route. This 2593 will cause the MEG SBD-DR to receive (S,G) traffic on an MVPN 2594 tunnel. 2596 The layer 2 multicast state is constructed as specified in 2597 Section 4.1. 2599 In the layer 3 multicast state, the IIF is the appropriate MVPN 2600 tunnel, and the IRB interface to the SBD is added to the OIF list. 2602 When the MEG SBD-DR receives (S,G) traffic on an MVPN tunnel, it 2603 performs IP processing of the traffic, and the sends the traffic 2604 down its IRB interface to the SBD. Following normal OISM 2605 procedures, the traffic will be encapsulated for ethernet and 2606 delivered to all PEs in the Tenant Domain that have interest in 2607 (S,G), including PE1. 2609 o If R is on a multi-homed ethernet segment of BD-R, one of the PEs 2610 attached to the segment will be its DF (following normal EVPN 2611 procedures), and the DF will know (via IGMP/MLD listening or the 2612 procedures of [IGMP-Proxy]) that a tenant system reachable via one 2613 of its local ACs to BD-R is interested in (S,G) traffic. The DF 2614 is responsible for originating an SBD-SMET route for (S,G), 2615 following normal OISM procedures. If the DF is a MEG, it MUST 2616 originate the corresponding MVPN C-multicast Source Tree Join(S,G) 2617 route; if the DF is not a MEG, the MEG SBD-DR SBD MUST originate 2618 the C-multicast route when it receives the SMET route. 2620 Optionally, if the non-DF is a MEG, it MAY originate the 2621 corresponding MVPN C-multicast Source Tree Join(S,G) route. This 2622 will cause the traffic to flow to both the DF and the non-DF, but 2623 only the DF will forward the traffic out an AC. This allows for 2624 quicker recovery if the DF's local AC to R fails. 2626 o If R is attached to a non-OISM PE, it will receive the traffic via 2627 an IPMG, as specified in Section 5. 2629 If an EVPN-attached receiver is interested in (*,G) traffic, and if 2630 it is possible for there to be sources of (*,G) traffic that are 2631 attached only to L3VPN nodes, the MEGs will have to know the group- 2632 to-RP mappings. That will enable them to originate MVPN C-multicast 2633 Shared Tree Join(*,G) routes and to send them towards the RP. (Since 2634 we are assuming in this section that there are no tenant multicast 2635 routers attached to the EVPN Tenant Domain, the RP must be attached 2636 via L3VPN. Alternatively, the MEG itself could be configured to 2637 function as an RP for group G.) 2639 The layer 2 multicast states are constructed as specified in 2640 Section 4.1. 2642 In the layer 3 (*,G) multicast state, the IIF is the appropriate MVPN 2643 tunnel. A MEG will add to the (*,G) OIF list its IRB interfaces for 2644 any BDs containing locally attached receivers. If there are 2645 receivers attached to other EVPN PEs, then whenever (S,G) traffic 2646 from an external source matches a (*,G) state, the MEG will create 2647 (S,G) state, with the MVPN tunnel as the IIF, the OIF list copied 2648 from the (*,G) state, and the SBD IRB interface added to the OIF 2649 list. (Please see the discussion in Section 6.1.1 regarding the 2650 inclusion of the SBD IRB interface in a (*,G) state; the SBD IRB 2651 interface is used in the OIF list only for traffic from external 2652 sources.) 2654 Normal MVPN procedures will then result in the MEG getting the (*,G) 2655 traffic from all the multicast sources for G that are attached via 2656 L3VPN. This traffic arrives on MVPN tunnels. When the MEG removes 2657 the traffic from these tunnels, it does the IP processing. If there 2658 are any receivers on a given BD, BD-R, that are attached via local 2659 EVPN ACs, the MEG sends the traffic down its BD-R IRB interface. If 2660 there are any other EVPN PEs that are interested in the (*,G) 2661 traffic, the MEG sends the traffic down the SBD IRB interface. 2662 Normal OISM procedures then distribute the traffic as needed to other 2663 EVPN-PEs. 2665 6.1.2.2. EVPN Sources with MVPN Receivers 2667 6.1.2.2.1. General procedures 2669 Consider the case where an EVPN tenant system S is sending IP 2670 multicast traffic to group G, and there is a receiver R for the (S,G) 2671 traffic that is attached to the L3VPN, but not attached to the EVPN 2672 Tenant Domain. (We assume in this document that the L3VPN/MVPN-only 2673 nodes will not have any special procedures to deal with the case 2674 where a source is inside an EVPN domain.) 2676 In this case, an L3VPN PE through which R can be reached has to send 2677 an MVPN C-multicast Join(S,G) route to one of the MEGs that is 2678 attached to the EVPN Tenant Domain. For this to happen, the L3VPN PE 2679 must have imported a VPN-IP route for S (either a host route or a 2680 subnet route) from a MEG. 2682 If a MEG determines that there is multicast source transmitting on 2683 one of its ACs, the MEG SHOULD originate a VPN-IP host route for that 2684 source. This determination SHOULD be made by examining the IP 2685 multicast traffic that arrives on the ACs. (It MAY be made by 2686 provisioning.) A MEG SHOULD NOT export a VPN-IP host route for any 2687 IP address that is not known to be a multicast source (unless it has 2688 some other reason for exporting such a route). The VPN-IP host route 2689 for a given multicast source MUST be withdrawn if the source goes 2690 silent for a configurable period of time, or if it can be determined 2691 that the source is no longer reachable via a local AC. 2693 A MEG SHOULD also originate a VPN-IP subnet route for each of the BDs 2694 in the Tenant Domain. 2696 VPN-IP routes exported by a MEG must carry any attributes or extended 2697 communities that are required by L3VPN and MVPN. In particular, a 2698 VPN-IP route exported by a MEG must carry a VRF Route Import Extended 2699 Community corresponding to the IP-VRF from which it is imported, and 2700 a Source AS Extended Community. 2702 As a result, if S is attached to a MEG, the L3VPN nodes will direct 2703 their MVPN C-multicast Join routes to that MEG. Normal MVPN 2704 procedures will cause the traffic to be delivered to the L3VPN nodes. 2705 The layer 3 multicast state for (S,G) will have the MVPN tunnel on 2706 its OIF list. The IIF will be the IRB interface leading to the BD 2707 containing S. 2709 If S is not attached to a MEG, the L3VPN nodes will direct their 2710 C-multicast Join routes to whichever MEG appears to be on the best 2711 route to S's subnet. Upon receiving the C-multicast Join, that MEG 2712 will originate an EVPN SMET route for (S,G). As a result, the MEG 2713 will receive the (S,G) traffic at layer 2 via the OISM procedures. 2714 The (S,G) traffic will be sent up the appropriate IRB interface, and 2715 the layer 3 MVPN procedures will ensure that the traffic is delivered 2716 to the L3VPN nodes that have requested it. The layer 3 multicast 2717 state for (S,G) will have the MVPN tunnel in the OIF list, and the 2718 IIF will be one of the following: 2720 o If S belongs to a BD that is attached to the MEG, the IIF will be 2721 the IRB interface to that BD; 2723 o Otherwise the IIF will be the SBD IRB interface. 2725 Note that this works even if S is attached to a non-OISM PE, per the 2726 procedures of Section 5. 2728 6.1.2.2.2. Any-Source Multicast (ASM) Groups 2730 Suppose the MEG SBD-DR learns that one of the PEs in its Tenant 2731 Domain is interested in (*,G), traffic, where G is an Any-Source 2732 Multicast (ASM) group. If there are no tenant multicast routers, the 2733 MEG SBD-DR SHOULD perform the "First Hop Router" (FHR) functionality 2734 for group G on behalf of the Tenant Domain, as described in 2735 [RFC7761]. This means that the MEG SBD-DR must know the identity of 2736 the Rendezvous Point (RP) for each group, must send Register messages 2737 to the Rendezvous Point, etc. 2739 If the MEG SBD-DR is to be the FHR for the Tenant Domain, it must see 2740 all the multicast traffic that is sourced from within the domain and 2741 destined to an ASM group address. The MEG can ensure this by 2742 originating an SBD-SMET route for (*,*). 2744 (As a possible optimization, an SBD-SMET route for (*, "any ASM 2745 group") may be defined in a future revision of this draft.) 2747 In some deployment scenarios, it may be preferred that the MEG that 2748 receives the (S,G) traffic over an AC be the one provides the FHR 2749 functionality. This behavior is OPTIONAL. If this option is used, 2750 it MUST be ensured that the MEG DR does not provide the FHR 2751 functionality for (S,G) traffic that is attached to another MEG; FHR 2752 functionality for (S,G) traffic from a particular source S MUST be 2753 provided by only a single router. 2755 Other deployment scenarios are also possible. For example, one might 2756 want to configure the MEGs to themselves be RPs. In this case, the 2757 RPs would have to exchange with each other information about which 2758 sources are active. The method exchanging such information is 2759 outside the scope of this document. 2761 6.1.2.2.3. Source on Multihomed Segment 2763 Suppose S is attached to a segment that is all-active multi-homed to 2764 PEl and PE2. If S is transmitting to two groups, say G1 and G2, it 2765 is possible that PE1 will receive the (S,G1) traffic from S while PE2 2766 receives the (S,G2) traffic from S. 2768 This creates an issue for MVPN/EVPN interworking, because there is no 2769 way to cause L3VPN/MVPN nodes to select PE1 as the ingress PE for 2770 (S,G1) traffic while selecting PE2 as the ingress PE for (S,G2) 2771 traffic. 2773 However, the following procedure ensures that the IP multicast 2774 traffic will still flow, even if the L3VPN/MVPN nodes picks the 2775 "wrong" EVPN-PE as the Upstream PE for (say) the (S,G1) traffic. 2777 Suppose S is on an ethernet segment, belonging to BD1, that is 2778 multi-homed to both PE1 and PE2, where PE1 is a MEG. And suppose 2779 that IP multicast traffic from S to G travels over the AC that 2780 attaches the segment to PE2 . If PE1 receives a C-multicast Source 2781 Tree Join (S,G) route, it MUST originate an SMET route for (S,G). 2782 Normal OISM procedures will then cause PE2 to send the (S,G) traffic 2783 to PE1 on an EVPN IP multicast tunnel. Normal OISM procedures will 2784 also cause PE1 to send the (S,G) traffic up its BD1 IRB interface. 2785 Normal MVPN procedures will then cause PE1 to forward the traffic on 2786 an MVPN tunnel. In this case, the routing is not optimal, but the 2787 traffic does flow correctly. 2789 6.1.2.3. Obtaining Optimal Routing of Traffic Between MVPN and EVPN 2791 The routing of IP multicast traffic between MVPN nodes and EVPN nodes 2792 will be optimal as long as there is a MEG along the optimal route. 2793 There are various deployment strategies that can be used to obtain 2794 optimal routing between MVPN and EVPN. 2796 In one such scenario, a Tenant Domain will have a small number of 2797 strategically placed MEGs. For example, a Data Center may have a 2798 small number of MEGs that connect it to a wide-area network. Then 2799 the optimal route into or out of the Data Center would be through the 2800 MEGs. 2802 In this scenario, the MEGs do not need to originate VPN-IP host 2803 routes for the multicast sources, they only need to originate VPN-IP 2804 subnet routes. The internal structure of the EVPN is completely 2805 hidden from the MVPN node. EVPN actions such as MAC Mobility and 2806 Mass Withdrawal ([RFC7432]) have zero impact on the MVPN control 2807 plane. 2809 While this deployment scenario provides the most optimal routing and 2810 has the least impact on the installed based of MVPN nodes, it does 2811 complicate network planning considerations. 2813 Another way of providing routing that is close to optimal is to turn 2814 each EVPN PE into a MEG. Then routing of MVPN-to-EVPN traffic is 2815 optimal. However, routing of EVPN-to-MVPN traffic is not guaranteed 2816 to be optimal when a source host is on a multi-homed ethernet segment 2817 (as discussed in Section 6.1.2.2.) 2819 The obvious disadvantage of this method is that it requires every 2820 EVPN PE to be a MEG. 2822 The procedures specified in this document allow an operator to add 2823 MEG functionality to any subset of his EVPN OISM PEs. This allows an 2824 operator to make whatever trade-offs he deems appropriate between 2825 optimal routing and MEG deployment. 2827 6.1.2.4. Selecting the MEG SBD-DR 2829 Every PE that is eligible for selection as the MEG SBD-DR originates 2830 an SBD-IMET route. As stated in Section 5, these SBD-IMET routes 2831 carry a Multicast Flags EC with the MEG Flag set. 2833 These SBD-IMET routes SHOULD also carry a DF Election EC. The DF 2834 Election EC and its use is specified in ([DF-Election-Framework]). 2835 When the route is originated, the AC-DF bit in the DF Election EC 2836 SHOULD be set to zero. This bit is not used when selecting a MEG 2837 SBD-DR, i.e., it MUST be ignored by the receiver of an SBD-IMET 2838 route. 2840 In the context of a given Tenant Domain, to select the MEG SBD-DR, 2841 the MEGs of the Tenant Domain perform the following procedure: 2843 o From the set of received SBD-IMET routes for the given tenant 2844 domain, determine he candidate set of PEs that support MEG 2845 functionality for that domain. 2847 o Select a DF Election algorithm as specified in 2848 [DF-Election-Framework]. Some of the possible algorithms can be 2849 found, e.g., in [RFC7432], [DF-Election-Framework], and 2850 [EVPN-DF-WEIGHTED]. 2852 o Apply the DF Election Algorithm (see [DF-Election-Framework]) to 2853 the candidate set of PEs. The "winner" becomes the MEG SBD-DR. 2855 Note that if a given PE supports IPMG (Section 6.1.2) or PEG 2856 (Section 6.1.4) functionality as well as MEG functionality, its 2857 SBD-IMET routes carry only one DF Election EC. 2859 6.1.3. Interworking with 'Global Table Multicast' 2861 If multicast service to the outside sources and/or receivers is 2862 provided via the BGP-based "Global Table Multicast" (GTM) procedures 2863 of [RFC7716], the procedures of Section 6.1.2 can easily be adapted 2864 for EVPN/GTM interworking. The way to adapt the MVPN procedures to 2865 GTM is explained in [RFC7716]. 2867 6.1.4. Interworking with PIM 2869 As we have been discussing, there may be receivers in an EVPN tenant 2870 domain that are interested in multicast flows whose sources are 2871 outside the EVPN Tenant Domain. Or there may be receivers outside an 2872 EVPN Tenant Domain that are interested in multicast flows whose 2873 sources are inside the Tenant Domain. 2875 If the outside sources and/or receivers are part of an MVPN, 2876 interworking procedures are covered in Section 6.1.2. 2878 There are also cases where an external source or receiver are 2879 attached via IP, and the layer 3 multicast routing is done via PIM. 2880 In this case, the interworking between the "PIM domain" and the EVPN 2881 tenant domain is done at L3 Gateways that perform "PIM/EVPN Gateway" 2882 (PEG) functionality. A PEG is very similar to a MEG, except that its 2883 layer 3 multicast routing is done via PIM rather than via BGP. 2885 If external sources or receivers for a given group are attached to a 2886 PEG via a layer 3 interface, that interface should be treated as a 2887 VRF interface attached to the Tenant Domain's L3VPN VRF. The layer 3 2888 multicast routing instance for that Tenant Domain will either run PIM 2889 on the VRF interface or will listen for IGMP/MLD messages on that 2890 interface. If the external receiver is attached elsewhere on an IP 2891 network, the PE has to enable PIM on its interfaces to the backbone 2892 network. In both cases, the PE needs to perform PEG functionality, 2893 and its IMET routes must carry the Multicast Flags EC with the PEG 2894 flag set. 2896 For each BD on which there is a multicast source or receiver, one of 2897 the PEGs will becomes the PEG DR. DR selection can be done using the 2898 same procedures specified in Section 6.1.2.4, except with "PEG" 2899 substituted for "MEG". 2901 As long as there are no tenant multicast routers within the EVPN 2902 Tenant Domain, the PEGs do not need to run PIM on their IRB 2903 interfaces. 2905 6.1.4.1. Source Inside EVPN Domain 2907 If a PEG receives a PIM Join(S,G) from outside the EVPN tenant 2908 domain, it may find it necessary to create (S,G) state. The PE needs 2909 to determine whether S is within the Tenant Domain. If S is not 2910 within the EVPN Tenant Domain, the PE carries out normal layer 3 2911 multicast routing procedures. If S is within the EVPN tenant domain, 2912 the IIF of the (S,G) state is set as follows: 2914 o if S is on a BD that is attached to the PE, the IIF is the PE's 2915 IRB interface to that BD; 2917 o if S is not on a BD that is attached to the PE, the IIF is the 2918 PE's IRB interface to the SBD. 2920 When the PE creates such an (S,G) state, it MUST originate (if it 2921 hasn't already) an SBD-SMET route for (S,G). This will cause it to 2922 pull the (S,G) traffic via layer 2. When the traffic arrives over an 2923 EVPN tunnel, it gets sent up an IRB interface where the layer 3 2924 multicast routing determines the packet's disposition. The SBD-SMET 2925 route is withdrawn when the (S,G) state no longer exists (unless 2926 there is some other reason for not withdrawing it). 2928 If there are no tenant multicast routers with the EVPN tenant domain, 2929 there cannot be an RP in the Tenant Domain, so a PEG does not have to 2930 handle externally arriving PIM Join(*,G) messages. 2932 The PEG DR for a particular BD MUST act as the a First Hop Router for 2933 that BD. It will examine all (S,G) traffic on the BD, and whenever G 2934 is an ASM group, the PEG DR will send Register messages to the RP for 2935 G. This means that the PEG DR will need to pull all the (S,G) 2936 traffic originating on a given BD, by originating an SMET (*,*) route 2937 for that BD. If a PEG DR is the DR for all the BDS, in SHOULD 2938 originate just an SBD-SMET (*,*) route rather than an SMET (*,*) 2939 route for each BD. 2941 The rules for exporting IP routes to multicast sources are the same 2942 as those specified for MEGs in Section 6.1.2.2, except that the 2943 exported routes will be IP routes rather than VPN-IP routes, and it 2944 is not necessary to attach the VRF Route Import EC or the Source AS 2945 EC. 2947 When a source is on a multi-homed segment, the same issue discussed 2948 in Section 6.1.2.2.3 exists. Suppose S is on an ethernet segment, 2949 belonging to BD1, that is multi-homed to both PE1 and PE2, where PE1 2950 is a PEG. And suppose that IP multicast traffic from S to G travels 2951 over the AC that attaches the segment to PE2. If PE1 receives an 2952 external PIM Join (S,G) route, it MUST originate an SMET route for 2953 (S,G). Normal OISM procedures will cause PE2 to send the (S,G) 2954 traffic to PE1 on an EVPN IP multicast tunnel. Normal OISM 2955 procedures will also cause PE1 to send the (S,G) traffic up its BD1 2956 IRB interface. Normal PIM procedures will then cause PE1 to forward 2957 the traffic along a PIM tree. In this case, the routing is not 2958 optimal, but the traffic does flow correctly. 2960 6.1.4.2. Source Outside EVPN Domain 2962 By means of normal OISM procedures, a PEG learns whether there are 2963 receivers in the Tenant Domain that are interested in receiving (*,G) 2964 or (S,G) traffic. The PEG must determine whether S (or the RP for G) 2965 is outside the EVPN Tenant Domain. If so, and if there is a receiver 2966 on BD1 interested in receiving such traffic, the PEG DR for BD1 is 2967 responsible for originating a PIM Join(S,G) or Join(*,G) control 2968 message. 2970 An alternative would be to allow any PEG that is directly attached to 2971 a receiver to originate the PIM Joins. Then the PEG DR would only 2972 have to originate PIM Joins on behalf of receivers that are not 2973 attached to a PEG. However, if this is done, it is necessary for the 2974 PEGs to run PIM on all their IRB interfaces, so that the PIM Assert 2975 procedures can be used to prevent duplicate delivery to a given BD. 2977 The IIF for the layer 3 (S,G) or (*,G) state is determined by normal 2978 PIM procedures. If a receiver is on BD1, and the PEG DR is attached 2979 to BD1, its IRB interface to BD1 is added to the OIF list. This 2980 ensures that any receivers locally attached to the PEG DR will 2981 receive the traffic. If there are receivers attached to other EVPN 2982 PEs, then whenever (S,G) traffic from an external source matches a 2983 (*,G) state, the PEG will create (S,G) state. The IIF will be set to 2984 whatever external interface the traffic is expected to arrive on 2985 (copied from the (*,G) state), the OIF list is copied from the (*,G) 2986 state, and the SBD IRB interface added to the OIF list. 2988 6.2. Interworking with PIM via an External PIM Router 2990 Section 6.1 describes how to use an OISM PE router as the gateway to 2991 a non-EVPN multicast domain, when the EVPN tenant domain is not being 2992 used as an intermediate transit network for multicast. An 2993 alternative approach is to have one or more external PIM routers 2994 (perhaps operated by a tenant) on one of the BDs of the tenant 2995 domain. We will refer to this BD as the "gateway BD". 2997 In this model: 2999 o The EVPN Tenant Domain is treated as a stub network attached to 3000 the external PIM routers. 3002 o The external PIM routers follow normal PIM procedures, and provide 3003 the FHR and LHR functionality for the entire Tenant Domain. 3005 o The OISM PEs do not run PIM. 3007 o There MUST NOT be more than one gateway BD. 3009 o If an OISM PE not attached to the gateway BD has interest in a 3010 given multicast flow, it conveys that interest, following normal 3011 OISM procedures, by originating an SBD-SMET route for that flow. 3013 o If a PE attached to the gateway BD receives an SBD-SMET, it may 3014 need to generate and transmit a corresponding IGMP/MLD Join out 3015 one or more of its ACs. (Procedures for generating an IGMP/MLD 3016 Join as a result of receiving an SMET route are given in 3017 [IGMP-Proxy].) The PE MUST know which BD is the Gateway BD and 3018 MUST NOT transmit an IGMP/MLD Join to any other BDs. Furthermore, 3019 even if a particular AC is part of that BD, the PE SHOULD NOT 3020 transmit an IGMP/MLD Join on that AC unless that an external PIM 3021 route is attached via that AC. 3023 As a result, IGMP/MLD messages will seen by the external PIM 3024 routers on the gateway BD, and those external PIM routers will 3025 send PIM Join messages externally as required. Traffic of the 3026 given multicast flow will then be received by one of the external 3027 PIM routers, and that traffic will be forwarded by that router to 3028 the gateway BD. 3030 The normal OISM procedures will then cause the given multicast 3031 flow to be tunneled to any PEs of the EVPN Tenant Domain that have 3032 interest in the flow. PEs attached to the gateway BD will see the 3033 flow as originating from the gateway BD, other PEs will see the 3034 flow as originating from the SBD. 3036 o An OISM PE attached to a gateway BD MUST set its layer 2 multicast 3037 state to indicate that each AC to the gateway BD has interest in 3038 all multicast flows. It MUST also originate an SMET route for 3039 (*,*). The procedures for originating SMET routes are discussed 3040 in Section 2.5. 3042 This will cause the OISM PEs attached to the gateway BD to receive 3043 all the IP multicast traffic that is sourced within the EVPN 3044 tenant domain, and to transmit that traffic to the gateway BD, 3045 where the external PIM routers will see it. This enables the 3046 external PIM routers to perform FHR functions on behalf of the 3047 entire Tenant Domain. (Of course, if the gateway BD has a 3048 multi-homed segment, only the PE that is the DF for that segment 3049 will transmit the multicast traffic to the segment.) 3051 7. Using an EVPN Tenant Domain as an Intermediate (Transit) Network for 3052 Multicast traffic 3054 In this section, we consider the scenario where one or more BDs of an 3055 EVPN Tenant Domain are being used to carry IP multicast traffic for 3056 which the source and at least one receiver are not part the tenant 3057 domain. That is, one or more BDs of the Tenant Domain are 3058 intermediate "links" of a larger multicast tree created by PIM. 3060 We define a "tenant multicast router" as a multicast router, running 3061 PIM, that is: 3063 1. attached to one or more BDs of the Tenant Domain, but 3065 2. is not an EVPN PE router. 3067 In order an EVPN Tenant Domain to be used as a transit network for IP 3068 multicast, one or more of its BDs must have tenant multicast routers, 3069 and an OISM PE that attaching to such a BD MUST be provisioned to 3070 enable PIM on its IRB interface to that BD. (This is true even if 3071 none of the tenant routers is on a segment attached to the PE.) 3072 Further, all the OISM PEs (even ones not attached to a BD with tenant 3073 multicast routers) MUST be provisioned to enable PIM on their SBD IRB 3074 interfaces. 3076 If PIM is enabled on a particular BD, the DR Selection procedure of 3077 Section 6.1.2.4 MUST be replaced by the normal PIM DR Election 3078 procedure of [RFC7761]. Note that this may result in one of the 3079 tenant routers being selected as the DR, rather than one of the OISM 3080 PE routers. In this case, First Hop Router and Last Hop Router 3081 functionality will not be performed by any of the EVPN PEs. 3083 A PIM control message on a particular BD is considered to be a 3084 link-local multicast message, and as such is sent transparently from 3085 PE to PE via the BUM tunnel for that BD. This is true whether the 3086 control message was received from an AC, or whether it was received 3087 from the local layer 3 routing instance via an IRB interface. 3089 A PIM Join/Prune message contains three fields that are relevant to 3090 the present discussion: 3092 o Upstream Neighbor 3094 o Group Address (G) 3096 o Source Address (S), omitted in the case of (*,G) Join/Prune 3097 messages. 3099 We will generally speak of a PIM Join as a "Join(S,G)" or a 3100 "Join(*,G)" message, and will use the term "Join(X,G)" to mean 3101 "either Join(S,G) or Join(*,G)". In the context of a Join(X,G), we 3102 will use the term "X" to mean "S in the case of (S,G), or G's RP in 3103 the case of (*,G)". 3105 Suppose BD1 contains two tenant multicast routers, C1 and C2. 3106 Suppose C1 is on a segment attached to PE1, and C2 is on a segment 3107 attached to PE2. When C1 sends a PIM Join(X,G) to BD1, the Upstream 3108 Neighbor field might be set to either PE1, PE2, or C2. C1 chooses 3109 the Upstream Neighbor based on its unicast routing. Typically, it 3110 will choose as the Upstream Neighbor the PIM router on BD1 that is 3111 "closest" (according to the unicast routing) to X. Note that this 3112 will not necessarily be PE1. PE1 may not even be visible to the 3113 unicast routing algorithm used by the tenant routers. Even if it is, 3114 it is unlikely to be the PIM router that is closest to X. So we need 3115 to consider the following two cases: 3117 1. C1 sends a PIM Join(X,G) to BD1, with PE1 as the Upstream 3118 Neighbor. 3120 PE1's PIM routing instance will see the Join arrive on the BD1 3121 IRB interface. If X is not within the Tenant Domain, PE1 3122 handles the Join according to normal PIM procedures. This will 3123 generally result in PE1 selecting an Upstream Neighbor and 3124 sending it a Join(X,G). 3126 If X is within the Tenant Domain, but is attached to some other 3127 PE, PE1 sends (if it hasn't already) an SBD-SMET route for 3128 (X,G). The IIF of the layer 3 (X,G) state will be the SBD IRB 3129 interface, and the OIF list will include the IRB interface to 3130 BD1. 3132 The SBD-SMET route will pull the (X,G) traffic to PE1, and the 3133 (X,G) state will result in the (X,G) traffic being forwarded to 3134 C1. 3136 If X is within the Tenant Domain, but is attached to PE1 itself, 3137 no SBD-SMET route is sent. The IIF of the layer 3 (X,G) state 3138 will be the IRB interface to X's BD, and the OIF list will 3139 include the IRB interface to BD1. 3141 2. C1 sends a PIM Join(X,G) to BD1, with either PE2 or C2 as the 3142 Upstream Neighbor. 3144 PE1's PIM routing instance will see the Join arrive on the BD1 3145 IRB interface. If neither X nor Upstream Neighbor is within the 3146 tenant domain, PE1 handles the Join according to normal PIM 3147 procedures. This will NOT result in PE1 sending a Join(X,G). 3149 If either X or Upstream Neighbor is within the Tenant Domain, 3150 PE1 sends (if it hasn't already) an SBD-SMET route for (X,G). 3151 The IIF of the layer 3 (X,G) state will be the SBD IRB 3152 interface, and the OIF list will include the IRB interface to 3153 BD1. 3155 The SBD-SMET route will pull the (X,G) traffic to PE1, and the 3156 (X,G) state will result in the (X,G) traffic being forwarded to 3157 C1. 3159 8. IANA Considerations 3161 IANA is requested to assign new flags in the "Multicast Flags 3162 Extended Community Flags" registry. These flags are: 3164 o IPMG 3166 o MEG 3168 o PEG 3170 o OISM SBD 3172 o OISM-supported 3174 9. Security Considerations 3176 This document uses protocols and procedures defined in the normative 3177 references, and inherits the security considerations of those 3178 references. 3180 This document adds flags or Extended Communities (ECs) to a number of 3181 BGP routes, in order to signal that particular nodes support the 3182 OISM, IPMG, MEG, and/or PEG functionalities that are defined in this 3183 document. Incorrect addition, removal, or modification of those 3184 flags and/or ECs will cause the procedures defined herein to 3185 malfunction, in which case loss or diversion of data traffic is 3186 possible. 3188 10. Acknowledgements 3190 The authors thank Vikram Nagarajan and Princy Elizabeth for their 3191 work on Section 6.2 and Section 3.2.3.1. The authors also benefited 3192 tremendously from discussions with Aldrin Isaac on EVPN multicast 3193 optimizations. 3195 11. References 3196 11.1. Normative References 3198 [DF-Election-Framework] 3199 Rabadan, J., Mohanty, S., Sajassi, A., Drake, J., Nagaraj, 3200 K., and S. Sathappan, "Framework for EVPN Designated 3201 Forwarder Election Extensibility", internet-draft draft- 3202 ietf-bess-evpn-df-election-framework-07.txt, December 3203 2018. 3205 [EVPN-AR] Rabadan, J., Ed., "Optimized Ingress Replication solution 3206 for EVPN", internet-draft ietf-bess-evpn-optimized-ir- 3207 06.txt, October 2018. 3209 [EVPN-BUM] 3210 Zhang, Z., Lin, W., Rabadan, J., and K. Patel, "Updates on 3211 EVPN BUM Procedures", internet-draft ietf-bess-evpn-bum- 3212 procedure-updates-05.txt, December 2018. 3214 [EVPN-IRB] 3215 Sajassi, A., Salam, S., Thoria, S., Drake, J., and J. 3216 Rabadan, "Integrated Routing and Bridging in EVPN", 3217 internet-draft draft-ietf-bess-evpn-inter-subnet- 3218 forwarding-05.txt, July 2018. 3220 [EVPN_IP_Prefix] 3221 Rabadan, J., Henderickx, W., Drake, J., Lin, W., and A. 3222 Sajassi, "IP Prefix Advertisement in EVPN", internet- 3223 draft ietf-bess-evpn-prefix-advertisement-11.txt, May 3224 2018. 3226 [IGMP-Proxy] 3227 Sajassi, A., Thoria, S., Patel, K., Yeung, D., Drake, J., 3228 and W. Lin, "IGMP and MLD Proxy for EVPN", internet-draft 3229 draft-ietf-bess-evpn-igmp-mld-proxy-02.txt, June 2018. 3231 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 3232 Requirement Levels", BCP 14, RFC 2119, 3233 DOI 10.17487/RFC2119, March 1997, 3234 . 3236 [RFC2236] Fenner, W., "Internet Group Management Protocol, Version 3237 2", RFC 2236, DOI 10.17487/RFC2236, November 1997, 3238 . 3240 [RFC2710] Deering, S., Fenner, W., and B. Haberman, "Multicast 3241 Listener Discovery (MLD) for IPv6", RFC 2710, 3242 DOI 10.17487/RFC2710, October 1999, 3243 . 3245 [RFC3032] Rosen, E., Tappan, D., Fedorkow, G., Rekhter, Y., 3246 Farinacci, D., Li, T., and A. Conta, "MPLS Label Stack 3247 Encoding", RFC 3032, DOI 10.17487/RFC3032, January 2001, 3248 . 3250 [RFC4360] Sangli, S., Tappan, D., and Y. Rekhter, "BGP Extended 3251 Communities Attribute", RFC 4360, DOI 10.17487/RFC4360, 3252 February 2006, . 3254 [RFC6625] Rosen, E., Ed., Rekhter, Y., Ed., Hendrickx, W., and R. 3255 Qiu, "Wildcards in Multicast VPN Auto-Discovery Routes", 3256 RFC 6625, DOI 10.17487/RFC6625, May 2012, 3257 . 3259 [RFC7153] Rosen, E. and Y. Rekhter, "IANA Registries for BGP 3260 Extended Communities", RFC 7153, DOI 10.17487/RFC7153, 3261 March 2014, . 3263 [RFC7432] Sajassi, A., Ed., Aggarwal, R., Bitar, N., Isaac, A., 3264 Uttaro, J., Drake, J., and W. Henderickx, "BGP MPLS-Based 3265 Ethernet VPN", RFC 7432, DOI 10.17487/RFC7432, February 3266 2015, . 3268 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 3269 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 3270 May 2017, . 3272 11.2. Informative References 3274 [EVPN-BIER] 3275 Zhang, Z., Przygienda, T., Sajassi, A., and J. Rabadan, 3276 "EVPN BUM Using BIER", internet-draft ietf-bier-evpn- 3277 01.txt, April 2018. 3279 [EVPN-DF-WEIGHTED] 3280 Rabadan, J., Sathappan, S., Przygienda, T., Lin, W., 3281 Drake, J., Sajassi, A., and S. Mohanty, "Preference-based 3282 EVPN DF Election", internet-draft ietf-bess-evpn-pref-df- 3283 03.txt, December 2018. 3285 [RFC4364] Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private 3286 Networks (VPNs)", RFC 4364, DOI 10.17487/RFC4364, February 3287 2006, . 3289 [RFC6513] Rosen, E., Ed. and R. Aggarwal, Ed., "Multicast in MPLS/ 3290 BGP IP VPNs", RFC 6513, DOI 10.17487/RFC6513, February 3291 2012, . 3293 [RFC6514] Aggarwal, R., Rosen, E., Morin, T., and Y. Rekhter, "BGP 3294 Encodings and Procedures for Multicast in MPLS/BGP IP 3295 VPNs", RFC 6514, DOI 10.17487/RFC6514, February 2012, 3296 . 3298 [RFC7606] Chen, E., Ed., Scudder, J., Ed., Mohapatra, P., and K. 3299 Patel, "Revised Error Handling for BGP UPDATE Messages", 3300 RFC 7606, DOI 10.17487/RFC7606, August 2015, 3301 . 3303 [RFC7716] Zhang, J., Giuliano, L., Rosen, E., Ed., Subramanian, K., 3304 and D. Pacella, "Global Table Multicast with BGP Multicast 3305 VPN (BGP-MVPN) Procedures", RFC 7716, 3306 DOI 10.17487/RFC7716, December 2015, 3307 . 3309 [RFC7761] Fenner, B., Handley, M., Holbrook, H., Kouvelas, I., 3310 Parekh, R., Zhang, Z., and L. Zheng, "Protocol Independent 3311 Multicast - Sparse Mode (PIM-SM): Protocol Specification 3312 (Revised)", STD 83, RFC 7761, DOI 10.17487/RFC7761, March 3313 2016, . 3315 [RFC8296] Wijnands, IJ., Ed., Rosen, E., Ed., Dolganow, A., 3316 Tantsura, J., Aldrin, S., and I. Meilik, "Encapsulation 3317 for Bit Index Explicit Replication (BIER) in MPLS and Non- 3318 MPLS Networks", RFC 8296, DOI 10.17487/RFC8296, January 3319 2018, . 3321 Appendix A. Integrated Routing and Bridging 3323 This Appendix provides a short tutorial on the interaction of routing 3324 and bridging. First it shows the traditional model, where bridging 3325 and routing are performed in separate boxes. Then it shows the model 3326 specified in [EVPN-IRB], where a single box contains both routing and 3327 bridging functions. The latter model is presupposed in the body of 3328 this document. 3330 Figure 1 shows a "traditional" router that only does routing and has 3331 no L2 bridging capabilities. There are two LANs, LAN1 and LAN2. 3332 LAN1 is realized by switch1, LAN2 by switch2. The router has an 3333 interface, "lan1" that attaches to LAN1 (via switch1) and an 3334 interface "lan2" that attachs to LAN2 (via switch2). Each intreface 3335 is configured, as an IP interface, with an IP address and a subnet 3336 mask. 3338 +-------+ +--------+ +-------+ 3339 | | lan1| |lan2 | | 3340 H1 -----+Switch1+--------+ Router1+--------+Switch2+------H3 3341 | | | | | | 3342 H2 -----| | | | | | 3343 +-------+ +--------+ +-------+ 3344 |_________________| |__________________| 3345 LAN1 LAN2 3347 Figure 1: Conventional Router with LAN Interfaces 3349 IP traffic (unicast or multicast) that remains within a single subnet 3350 never reaches the router. For instance, if H1 emits an ethernet 3351 frame with H2's MAC address in the ethernet destination address 3352 field, the frame will go from H1 to Switch1 to H2, without ever 3353 reaching the router. Since the frame is never seen by a router, the 3354 IP datagram within the frame remains entirely unchanged; e.g., its 3355 TTL is not decremented. The ethernet Source and Destination MAC 3356 addresses are not changed either. 3358 If H1 wants to send a unicast IP datagram to H3, which is on a 3359 different subnet, H1 has to be configured with the IP address of a 3360 "default router". Let's assume that H1 is configured with an IP 3361 address of Router1 as its default router address. H1 compares H3's 3362 IP address with its own IP address and IP subnet mask, and determines 3363 that H3 is on a different subnet. So the packet has to be routed. 3364 H1 uses ARP to map Router1's IP address to a MAC address on LAN1. H1 3365 then encapsulates the datagram in an ethernet frame, using router1's 3366 MAC address as the destination MAC address, and sends the frame to 3367 Router1. 3369 Router1 then receives the frame over its lan1 interface. Router1 3370 sees that the frame is addressed to it, so it removes the ethernet 3371 encapsulation and processes the IP datagram. The datagram is not 3372 addressed to Router1, so it must be forwarded further. Router1 does 3373 a lookup of the datagram's IP destination field, and determines that 3374 the destination (H3) can be reached via Router1's lan2 interface. 3375 Router1 now performs the IP processing of the datagram: it decrements 3376 the IP TTL, adjusts the IP header checksum (if present), may fragment 3377 the packet is necessary, etc. Then the datagram (or its fragments) 3378 are encapsulated in an ethernet header, with Router1's MAC address on 3379 LAN2 as the MAC Source Address, and H3's MAC address on LAN2 (which 3380 Router1 determines via ARP) as the MAC Destination Address. Finally 3381 the packet is sent out the lan2 interface. 3383 If H1 has an IP multicast datagram to send (i.e., an IP datagram 3384 whose Destination Address field is an IP Multicast Address), it 3385 encapsulates it in an ethernet frame whose MAC Destination Address is 3386 computed from the IP Destination Address. 3388 If H2 is a receiver for that multicast address, H2 will receive a 3389 copy of the frame, unchanged, from H1. The MAC Source Address in the 3390 ethernet encapsulation does not change, the IP TTL field does not get 3391 decremented, etc. 3393 If H3 is a receiver for that multicast address, the datagram must be 3394 routed to H3. In order for this to happen, Router1 must be 3395 configured as a multicast router, and it must accept traffic sent to 3396 ethernet multicast addresses. Router1 will receive H1's multicast 3397 frame on its lan1 interface, will remove the ethernet encapsulation, 3398 and will determine how to dispatch the IP datagram based on Router1's 3399 multicast forwarding states. If Router1 knows that there is a 3400 receiver for the multicast datagram on LAN2, makes a copy of the 3401 datagram, decrements the TTL (and performs any other necessary IP 3402 processing), then encapsulates the datagram in ethernet frame for 3403 LAN2. The MAC Source Address for this frame will be Router1's MAC 3404 Source Address on LAN2. The MAC Destination Address is computed from 3405 the IP Destination Address. Finally, the frame is sent out Router1's 3406 LAN2 interface. 3408 Figure 2 shows an Integrated Router/Bridge that supports the routing/ 3409 bridging integration model of [EVPN-IRB]. 3411 +------------------------------------------+ 3412 | Integrated Router/Bridge | 3414 +-------+ +--------+ +-------+ 3415 | | IRB1| L3 |IRB2 | | 3416 H1 -----+ BD1 +--------+Routing +--------+ BD2 +------H3 3417 | | |Instance| | | 3418 H2 -----| | | | | | 3419 +-------+ +--------+ +-------+ 3420 |___________________| |____________________| 3421 LAN1 LAN2 3423 Figure 2: Integrated Router/Bridge 3425 In Figure 2, a single box consists of one or more "L3 Routing 3426 Instances". The routing/forwarding tables of a given routing 3427 instance is known as an IP-VRF ([EVPN-IRB]). In the context of EVPN, 3428 it is convenient to think of each routing instance as representing 3429 the routing of a particular tenant. Each IP-VRF is attached to one 3430 or more interfaces. 3432 When several EVPN PEs have a routing instance of the same tenant 3433 domain, those PEs advertise IP routes to the attached hosts. This is 3434 done as specified in [EVPN-IRB]. 3436 The integrated router/bridge shown in Figure 2 also attaches to a 3437 number of "Broadcast Domains" (BDs). Each BD performs the functions 3438 that are performed by the bridges in Figure 1. To the L3 routing 3439 instance, each BD appears to be a LAN. The interface attaching a 3440 particular BD to a particular IP-VRF is known as an "IRB Interface". 3441 From the perspective of L3 routing, each BD is a subnet. Thus each 3442 IRB interface is configured with a MAC address (which is the router's 3443 MAC address on the corresponding LAN), as well as an IP address and 3444 subnet mask. 3446 The integrated router/bridge shown in Figure 2 may have multiple ACs 3447 to each BD. These ACs are visible only to the bridging function, not 3448 to the routing instance. To the L3 routing instance, there is just 3449 one "interface" to each BD. 3451 If the L3 routing instance represents the IP routing of a particular 3452 tenant, the BDs attached to that routing instance are BDs belonging 3453 to that same tenant. 3455 Bridging and routing now proceed exactly as in the case of Figure 1, 3456 except that BD1 replaces Switch1, BD2 replaces Switch2, interface 3457 IRB1 replaces interface lan1, and interface IRB2 replaces interface 3458 lan2. 3460 It is important to understand that an IRB interface connects an L3 3461 routing instance to a BD, NOT to a "MAC-VRF". (See [RFC7432] for the 3462 definition of "MAC-VRF".) A MAC-VRF may contain several BDs, as long 3463 as no MAC address appears in more than one BD. From the perspective 3464 of the L3 routing instance, each individual BD is an individual IP 3465 subnet; whether each BD has its own MAC-VRF or not is irrelevant to 3466 the L3 routing instance. 3468 Figure 3 illustrates IRB when a pair of BDs (subnets) are attached to 3469 two different PE routers. In this example, each BD has two segments, 3470 and one segment of each BD is attached to one PE router. 3472 +------------------------------------------+ 3473 | Integrated Router/Bridges | 3475 +-------+ +--------+ +-------+ 3476 | | IRB1| |IRB2 | | 3477 H1 -----+ BD1 +--------+ PE1 +--------+ BD2 +------H3 3478 |(Seg-1)| |(L3 Rtg)| |(Seg-1)| 3479 H2 -----| | | | | | 3480 +-------+ +--------+ +-------+ 3481 |___________________| | |____________________| 3482 LAN1 | LAN2 3483 | 3484 | 3485 +-------+ +--------+ +-------+ 3486 | | IRB1| |IRB2 | | 3487 H4 -----+ BD1 +--------+ PE2 +--------+ BD2 +------H5 3488 |(Seg-2)| |(L3 Rtg)| |(Seg-2)| 3489 | | | | | | 3490 +-------+ +--------+ +-------+ 3492 Figure 3: Integrated Router/Bridges with Distributed Subnet 3494 If H1 needs to send an IP packet to H4, it determines from its IP 3495 address and subnet mask that H4 is on the same subnet as H1. 3496 Although H1 and H4 are not attached to the same PE router, EVPN 3497 provides ethernet communication among all hosts that are on the same 3498 BD. H1 thus uses ARP to find H4's MAC address, and sends an ethernet 3499 frame with H4's MAC address in the Destination MAC address field. 3500 The frame is received at PE1, but since the Destination MAC address 3501 is not PE1's MAC address, PE1 assumes that the frame is to remain on 3502 BD1. Therefore the packet inside the frame is NOT decapsulated, and 3503 is NOT send up the IRB interface to PE1's routing instance. Rather, 3504 standard EVPN intra-subnet procedures (as detailed in [RFC7432] are 3505 used to deliver the frame to PE2, which then sends it to H4. 3507 If H1 needs to send an IP packet to H5, it determines from its IP 3508 address and subnet mask that H5 is NOT on the same subnet as H1. 3509 Assuming that H1 has been configured with the IP address of PE1 as 3510 its default router, H1 sends the packet in an ethernet frame with 3511 PE1's MAC address in its Destination MAC Address field. PE1 receives 3512 the frame, and sees that the frame is addressed to it. PE1 thus 3513 sends the frame up its IRB1 interface to the L3 routing instance. 3514 Appropriate IP processing is done (e.g., TTL decrement). The L3 3515 routing instance determines that the "next hop" for H5 is PE2, so the 3516 packet is encapsulated (e.g., in MPLS) and sent across the backbone 3517 to PE2's routing instance. PE2 will see that the packet's 3518 destination, H5, is on BD2 segment-2, and will send the packet down 3519 its IRB2 interface. This causes the IP packet to be encapsulated in 3520 an ethernet frame with PE2's MAC address (on BD2) in the Source 3521 Address field and H5's MAC address in the Destination Address field. 3523 Note that if H1 has an IP packet to send to H3, the forwarding of the 3524 packet is handled entirely within PE1. PE1's routing instance sees 3525 the packet arrive on its IRB1 interface, and then transmits the 3526 packet by sending it down its IRB2 interface. 3528 Often, all the hosts in a particular Tenant Domain will be 3529 provisioned with the same value of the default router IP address. 3530 This IP address can be assigned, as an "anycast address", to all the 3531 EVPN PEs attached to that Tenant Domain. Thus although all hosts are 3532 provisioned with the same "default router address", the actual 3533 default router for a given host will be one of the PEs that is 3534 attached to the same ethernet segment as the host. This provisioning 3535 method ensures that IP packets from a given host are handled by the 3536 closest EVPN PE that supports IRB. 3538 In the topology of Figure 3, one could imagine that H1 is configured 3539 with a default router address that belongs to PE2 but not to PE1. 3540 Inter-subnet routing would still work, but IP packets from H1 to H3 3541 would then follow the non-optimal path H1-->PE1-->PE2-->PE1-->H3. 3542 Sending traffic on this sort of path, where it leaves a router and 3543 then comes back to the same router, is sometimes known as 3544 "hairpinning". Similarly, if PE2 supports IRB but PE1 dos not, the 3545 same non-optimal path from H1 to H3 would have to be followed. To 3546 avoid hairpinning, each EVPN PE needs to support IRB. 3548 It is worth pointing out the way IRB interfaces interact with 3549 multicast traffic. Referring again to Figure 3, suppose PE1 and PE2 3550 are functioning as IP multicast routers. Suppose also that H3 3551 transmits a multicast packet, and both H1 and H4 are interested in 3552 receiving that packet. PE1 will receive the packet from H3 via its 3553 IRB2 interface. The ethernet encapsulation from BD2 is removed, the 3554 IP header processing is done, and the packet is then reencapsulated 3555 for BD1, with PE1's MAC address in the MAC Source Address field. 3556 Then the packet is sent down the IRB1 interface. Layer 2 procedures 3557 (as defined in [RFC7432] would then be used to deliver a copy of the 3558 packet locally to H1, and remotely to H4. 3560 Please be aware that his document modifies the semantics, described 3561 in the previous paragraph, of sending/receiving multicast traffic on 3562 an IRB interface. This is explained in Section 1.5.1 and subsequent 3563 sections. 3565 Authors' Addresses 3567 Wen Lin 3568 Juniper Networks, Inc. 3569 10 Technology Park Drive 3570 Westford, Massachusetts 01886 3571 United States 3573 EMail: wlin@juniper.net 3575 Zhaohui Zhang 3576 Juniper Networks, Inc. 3577 10 Technology Park Drive 3578 Westford, Massachusetts 01886 3579 United States 3581 EMail: zzhang@juniper.net 3583 John Drake 3584 Juniper Networks, Inc. 3585 1194 N. Mathilda Ave 3586 Sunnyvale, CA 94089 3587 United States 3589 EMail: jdrake@juniper.net 3591 Eric C. Rosen (editor) 3592 Juniper Networks, Inc. 3593 10 Technology Park Drive 3594 Westford, Massachusetts 01886 3595 United States 3597 EMail: erosen52@gmail.com 3598 Jorge Rabadan 3599 Nokia 3600 777 E. Middlefield Road 3601 Mountain View, CA 94043 3602 United States 3604 EMail: jorge.rabadan@nokia.com 3606 Ali Sajassi 3607 Cisco Systems 3608 170 West Tasman Drive 3609 San Jose, CA 95134 3610 United States 3612 EMail: sajassi@cisco.com