rfc1771.txt (131903B)
1 2 3 4 5 6 7 Network Working Group Y. Rekhter 8 Request for Comments: 1771 T.J. Watson Research Center, IBM Corp. 9 Obsoletes: 1654 T. Li 10 Category: Standards Track cisco Systems 11 Editors 12 March 1995 13 14 15 A Border Gateway Protocol 4 (BGP-4) 16 17 Status of this Memo 18 19 This document specifies an Internet standards track protocol for the 20 Internet community, and requests discussion and suggestions for 21 improvements. Please refer to the current edition of the "Internet 22 Official Protocol Standards" (STD 1) for the standardization state 23 and status of this protocol. Distribution of this memo is unlimited. 24 25 Abstract 26 27 This document, together with its companion document, "Application of 28 the Border Gateway Protocol in the Internet", define an inter- 29 autonomous system routing protocol for the Internet. 30 31 1. Acknowledgements 32 33 This document was originally published as RFC 1267 in October 1991, 34 jointly authored by Kirk Lougheed (cisco Systems) and Yakov Rekhter 35 (IBM). 36 37 We would like to express our thanks to Guy Almes (ANS), Len Bosack 38 (cisco Systems), and Jeffrey C. Honig (Cornell University) for their 39 contributions to the earlier version of this document. 40 41 We like to explicitly thank Bob Braden (ISI) for the review of the 42 earlier version of this document as well as his constructive and 43 valuable comments. 44 45 We would also like to thank Bob Hinden, Director for Routing of the 46 Internet Engineering Steering Group, and the team of reviewers he 47 assembled to review the previous version (BGP-2) of this document. 48 This team, consisting of Deborah Estrin, Milo Medin, John Moy, Radia 49 Perlman, Martha Steenstrup, Mike St. Johns, and Paul Tsuchiya, acted 50 with a strong combination of toughness, professionalism, and 51 courtesy. 52 53 54 55 56 57 58 Rekhter & Li [Page 1] 59 60 RFC 1771 BGP-4 March 1995 61 62 63 This updated version of the document is the product of the IETF IDR 64 Working Group with Yakov Rekhter and Tony Li as editors. Certain 65 sections of the document borrowed heavily from IDRP [7], which is the 66 OSI counterpart of BGP. For this credit should be given to the ANSI 67 X3S3.3 group chaired by Lyman Chapin (BBN) and to Charles Kunzinger 68 (IBM Corp.) who was the IDRP editor within that group. We would also 69 like to thank Mike Craren (Proteon, Inc.), Dimitry Haskin (Bay 70 Networks, Inc.), John Krawczyk (Bay Networks, Inc.), and Paul Traina 71 (cisco Systems) for their insightful comments. 72 73 We would like to specially acknowledge numerous contributions by 74 Dennis Ferguson (MCI). 75 76 The work of Yakov Rekhter was supported in part by the National 77 Science Foundation under Grant Number NCR-9219216. 78 79 2. Introduction 80 81 The Border Gateway Protocol (BGP) is an inter-Autonomous System 82 routing protocol. It is built on experience gained with EGP as 83 defined in RFC 904 [1] and EGP usage in the NSFNET Backbone as 84 described in RFC 1092 [2] and RFC 1093 [3]. 85 86 The primary function of a BGP speaking system is to exchange network 87 reachability information with other BGP systems. This network 88 reachability information includes information on the list of 89 Autonomous Systems (ASs) that reachability information traverses. 90 This information is sufficient to construct a graph of AS 91 connectivity from which routing loops may be pruned and some policy 92 decisions at the AS level may be enforced. 93 94 BGP-4 provides a new set of mechanisms for supporting classless 95 interdomain routing. These mechanisms include support for 96 advertising an IP prefix and eliminates the concept of network 97 "class" within BGP. BGP-4 also introduces mechanisms which allow 98 aggregation of routes, including aggregation of AS paths. These 99 changes provide support for the proposed supernetting scheme [8, 9]. 100 101 To characterize the set of policy decisions that can be enforced 102 using BGP, one must focus on the rule that a BGP speaker advertise to 103 its peers (other BGP speakers which it communicates with) in 104 neighboring ASs only those routes that it itself uses. This rule 105 reflects the "hop-by-hop" routing paradigm generally used throughout 106 the current Internet. Note that some policies cannot be supported by 107 the "hop-by-hop" routing paradigm and thus require techniques such as 108 source routing to enforce. For example, BGP does not enable one AS 109 to send traffic to a neighboring AS intending that the traffic take a 110 different route from that taken by traffic originating in the 111 112 113 114 Rekhter & Li [Page 2] 115 116 RFC 1771 BGP-4 March 1995 117 118 119 neighboring AS. On the other hand, BGP can support any policy 120 conforming to the "hop-by-hop" routing paradigm. Since the current 121 Internet uses only the "hop-by-hop" routing paradigm and since BGP 122 can support any policy that conforms to that paradigm, BGP is highly 123 applicable as an inter-AS routing protocol for the current Internet. 124 125 A more complete discussion of what policies can and cannot be 126 enforced with BGP is outside the scope of this document (but refer to 127 the companion document discussing BGP usage [5]). 128 129 BGP runs over a reliable transport protocol. This eliminates the 130 need to implement explicit update fragmentation, retransmission, 131 acknowledgement, and sequencing. Any authentication scheme used by 132 the transport protocol may be used in addition to BGP's own 133 authentication mechanisms. The error notification mechanism used in 134 BGP assumes that the transport protocol supports a "graceful" close, 135 i.e., that all outstanding data will be delivered before the 136 connection is closed. 137 138 BGP uses TCP [4] as its transport protocol. TCP meets BGP's 139 transport requirements and is present in virtually all commercial 140 routers and hosts. In the following descriptions the phrase 141 "transport protocol connection" can be understood to refer to a TCP 142 connection. BGP uses TCP port 179 for establishing its connections. 143 144 This document uses the term `Autonomous System' (AS) throughout. The 145 classic definition of an Autonomous System is a set of routers under 146 a single technical administration, using an interior gateway protocol 147 and common metrics to route packets within the AS, and using an 148 exterior gateway protocol to route packets to other ASs. Since this 149 classic definition was developed, it has become common for a single 150 AS to use several interior gateway protocols and sometimes several 151 sets of metrics within an AS. The use of the term Autonomous System 152 here stresses the fact that, even when multiple IGPs and metrics are 153 used, the administration of an AS appears to other ASs to have a 154 single coherent interior routing plan and presents a consistent 155 picture of what destinations are reachable through it. 156 157 The planned use of BGP in the Internet environment, including such 158 issues as topology, the interaction between BGP and IGPs, and the 159 enforcement of routing policy rules is presented in a companion 160 document [5]. This document is the first of a series of documents 161 planned to explore various aspects of BGP application. Please send 162 comments to the BGP mailing list (bgp@ans.net). 163 164 165 166 167 168 169 170 Rekhter & Li [Page 3] 171 172 RFC 1771 BGP-4 March 1995 173 174 175 3. Summary of Operation 176 177 Two systems form a transport protocol connection between one another. 178 They exchange messages to open and confirm the connection parameters. 179 The initial data flow is the entire BGP routing table. Incremental 180 updates are sent as the routing tables change. BGP does not require 181 periodic refresh of the entire BGP routing table. Therefore, a BGP 182 speaker must retain the current version of the entire BGP routing 183 tables of all of its peers for the duration of the connection. 184 KeepAlive messages are sent periodically to ensure the liveness of 185 the connection. Notification messages are sent in response to errors 186 or special conditions. If a connection encounters an error 187 condition, a notification message is sent and the connection is 188 closed. 189 190 The hosts executing the Border Gateway Protocol need not be routers. 191 A non-routing host could exchange routing information with routers 192 via EGP or even an interior routing protocol. That non-routing host 193 could then use BGP to exchange routing information with a border 194 router in another Autonomous System. The implications and 195 applications of this architecture are for further study. 196 197 If a particular AS has multiple BGP speakers and is providing transit 198 service for other ASs, then care must be taken to ensure a consistent 199 view of routing within the AS. A consistent view of the interior 200 routes of the AS is provided by the interior routing protocol. A 201 consistent view of the routes exterior to the AS can be provided by 202 having all BGP speakers within the AS maintain direct BGP connections 203 with each other. Using a common set of policies, the BGP speakers 204 arrive at an agreement as to which border routers will serve as 205 exit/entry points for particular destinations outside the AS. This 206 information is communicated to the AS's internal routers, possibly 207 via the interior routing protocol. Care must be taken to ensure that 208 the interior routers have all been updated with transit information 209 before the BGP speakers announce to other ASs that transit service is 210 being provided. 211 212 Connections between BGP speakers of different ASs are referred to as 213 "external" links. BGP connections between BGP speakers within the 214 same AS are referred to as "internal" links. Similarly, a peer in a 215 different AS is referred to as an external peer, while a peer in the 216 same AS may be described as an internal peer. 217 218 219 220 221 222 223 224 225 226 Rekhter & Li [Page 4] 227 228 RFC 1771 BGP-4 March 1995 229 230 231 3.1 Routes: Advertisement and Storage 232 233 For purposes of this protocol a route is defined as a unit of 234 information that pairs a destination with the attributes of a path to 235 that destination: 236 237 - Routes are advertised between a pair of BGP speakers in UPDATE 238 messages: the destination is the systems whose IP addresses are 239 reported in the Network Layer Reachability Information (NLRI) 240 field, and the the path is the information reported in the path 241 attributes fields of the same UPDATE message. 242 243 - Routes are stored in the Routing Information Bases (RIBs): 244 namely, the Adj-RIBs-In, the Loc-RIB, and the Adj-RIBs-Out. Routes 245 that will be advertised to other BGP speakers must be present in 246 the Adj-RIB-Out; routes that will be used by the local BGP speaker 247 must be present in the Loc-RIB, and the next hop for each of these 248 routes must be present in the local BGP speaker's forwarding 249 information base; and routes that are received from other BGP 250 speakers are present in the Adj-RIBs-In. 251 252 If a BGP speaker chooses to advertise the route, it may add to or 253 modify the path attributes of the route before advertising it to a 254 peer. 255 256 BGP provides mechanisms by which a BGP speaker can inform its peer 257 that a previously advertised route is no longer available for use. 258 There are three methods by which a given BGP speaker can indicate 259 that a route has been withdrawn from service: 260 261 a) the IP prefix that expresses destinations for a previously 262 advertised route can be advertised in the WITHDRAWN ROUTES field 263 in the UPDATE message, thus marking the associated route as being 264 no longer available for use 265 266 b) a replacement route with the same Network Layer Reachability 267 Information can be advertised, or 268 269 c) the BGP speaker - BGP speaker connection can be closed, which 270 implicitly removes from service all routes which the pair of 271 speakers had advertised to each other. 272 273 274 275 276 277 278 279 280 281 282 Rekhter & Li [Page 5] 283 284 RFC 1771 BGP-4 March 1995 285 286 287 3.2 Routing Information Bases 288 289 The Routing Information Base (RIB) within a BGP speaker consists of 290 three distinct parts: 291 292 a) Adj-RIBs-In: The Adj-RIBs-In store routing information that has 293 been learned from inbound UPDATE messages. Their contents 294 represent routes that are available as an input to the Decision 295 Process. 296 297 b) Loc-RIB: The Loc-RIB contains the local routing information 298 that the BGP speaker has selected by applying its local policies 299 to the routing information contained in its Adj-RIBs-In. 300 301 c) Adj-RIBs-Out: The Adj-RIBs-Out store the information that the 302 local BGP speaker has selected for advertisement to its peers. The 303 routing information stored in the Adj-RIBs-Out will be carried in 304 the local BGP speaker's UPDATE messages and advertised to its 305 peers. 306 307 In summary, the Adj-RIBs-In contain unprocessed routing information 308 that has been advertised to the local BGP speaker by its peers; the 309 Loc-RIB contains the routes that have been selected by the local BGP 310 speaker's Decision Process; and the Adj-RIBs-Out organize the routes 311 for advertisement to specific peers by means of the local speaker's 312 UPDATE messages. 313 314 Although the conceptual model distinguishes between Adj-RIBs-In, 315 Loc-RIB, and Adj-RIBs-Out, this neither implies nor requires that an 316 implementation must maintain three separate copies of the routing 317 information. The choice of implementation (for example, 3 copies of 318 the information vs 1 copy with pointers) is not constrained by the 319 protocol. 320 321 4. Message Formats 322 323 This section describes message formats used by BGP. 324 325 Messages are sent over a reliable transport protocol connection. A 326 message is processed only after it is entirely received. The maximum 327 message size is 4096 octets. All implementations are required to 328 support this maximum message size. The smallest message that may be 329 sent consists of a BGP header without a data portion, or 19 octets. 330 331 332 333 334 335 336 337 338 Rekhter & Li [Page 6] 339 340 RFC 1771 BGP-4 March 1995 341 342 343 4.1 Message Header Format 344 345 Each message has a fixed-size header. There may or may not be a data 346 portion following the header, depending on the message type. The 347 layout of these fields is shown below: 348 349 0 1 2 3 350 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 351 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 352 | | 353 + + 354 | | 355 + + 356 | Marker | 357 + + 358 | | 359 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 360 | Length | Type | 361 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 362 363 Marker: 364 365 This 16-octet field contains a value that the receiver of the 366 message can predict. If the Type of the message is OPEN, or if 367 the OPEN message carries no Authentication Information (as an 368 Optional Parameter), then the Marker must be all ones. 369 Otherwise, the value of the marker can be predicted by some a 370 computation specified as part of the authentication mechanism 371 (which is specified as part of the Authentication Information) 372 used. The Marker can be used to detect loss of synchronization 373 between a pair of BGP peers, and to authenticate incoming BGP 374 messages. 375 376 Length: 377 378 This 2-octet unsigned integer indicates the total length of the 379 message, including the header, in octets. Thus, e.g., it 380 allows one to locate in the transport-level stream the (Marker 381 field of the) next message. The value of the Length field must 382 always be at least 19 and no greater than 4096, and may be 383 further constrained, depending on the message type. No 384 "padding" of extra data after the message is allowed, so the 385 Length field must have the smallest value required given the 386 rest of the message. 387 388 389 390 391 392 393 394 Rekhter & Li [Page 7] 395 396 RFC 1771 BGP-4 March 1995 397 398 399 Type: 400 401 This 1-octet unsigned integer indicates the type code of the 402 message. The following type codes are defined: 403 404 1 - OPEN 405 2 - UPDATE 406 3 - NOTIFICATION 407 4 - KEEPALIVE 408 409 4.2 OPEN Message Format 410 411 After a transport protocol connection is established, the first 412 message sent by each side is an OPEN message. If the OPEN message is 413 acceptable, a KEEPALIVE message confirming the OPEN is sent back. 414 Once the OPEN is confirmed, UPDATE, KEEPALIVE, and NOTIFICATION 415 messages may be exchanged. 416 417 In addition to the fixed-size BGP header, the OPEN message contains 418 the following fields: 419 420 0 1 2 3 421 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 422 +-+-+-+-+-+-+-+-+ 423 | Version | 424 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 425 | My Autonomous System | 426 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 427 | Hold Time | 428 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 429 | BGP Identifier | 430 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 431 | Opt Parm Len | 432 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 433 | | 434 | Optional Parameters | 435 | | 436 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 437 438 Version: 439 440 This 1-octet unsigned integer indicates the protocol version 441 number of the message. The current BGP version number is 4. 442 443 My Autonomous System: 444 445 This 2-octet unsigned integer indicates the Autonomous System 446 number of the sender. 447 448 449 450 Rekhter & Li [Page 8] 451 452 RFC 1771 BGP-4 March 1995 453 454 455 Hold Time: 456 457 This 2-octet unsigned integer indicates the number of seconds 458 that the sender proposes for the value of the Hold Timer. Upon 459 receipt of an OPEN message, a BGP speaker MUST calculate the 460 value of the Hold Timer by using the smaller of its configured 461 Hold Time and the Hold Time received in the OPEN message. The 462 Hold Time MUST be either zero or at least three seconds. An 463 implementation may reject connections on the basis of the Hold 464 Time. The calculated value indicates the maximum number of 465 seconds that may elapse between the receipt of successive 466 KEEPALIVE, and/or UPDATE messages by the sender. 467 468 BGP Identifier: 469 470 This 4-octet unsigned integer indicates the BGP Identifier of 471 the sender. A given BGP speaker sets the value of its BGP 472 Identifier to an IP address assigned to that BGP speaker. The 473 value of the BGP Identifier is determined on startup and is the 474 same for every local interface and every BGP peer. 475 476 Optional Parameters Length: 477 478 This 1-octet unsigned integer indicates the total length of the 479 Optional Parameters field in octets. If the value of this field 480 is zero, no Optional Parameters are present. 481 482 Optional Parameters: 483 484 This field may contain a list of optional parameters, where 485 each parameter is encoded as a <Parameter Type, Parameter 486 Length, Parameter Value> triplet. 487 488 0 1 489 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 490 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-... 491 | Parm. Type | Parm. Length | Parameter Value (variable) 492 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-... 493 494 Parameter Type is a one octet field that unambiguously 495 identifies individual parameters. Parameter Length is a one 496 octet field that contains the length of the Parameter Value 497 field in octets. Parameter Value is a variable length field 498 that is interpreted according to the value of the Parameter 499 Type field. 500 501 502 503 504 505 506 Rekhter & Li [Page 9] 507 508 RFC 1771 BGP-4 March 1995 509 510 511 This document defines the following Optional Parameters: 512 513 a) Authentication Information (Parameter Type 1): 514 515 This optional parameter may be used to authenticate a BGP 516 peer. The Parameter Value field contains a 1-octet 517 Authentication Code followed by a variable length 518 Authentication Data. 519 520 0 1 2 3 4 5 6 7 8 521 +-+-+-+-+-+-+-+-+ 522 | Auth. Code | 523 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 524 | | 525 | Authentication Data | 526 | | 527 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 528 529 Authentication Code: 530 531 This 1-octet unsigned integer indicates the 532 authentication mechanism being used. Whenever an 533 authentication mechanism is specified for use within 534 BGP, three things must be included in the 535 specification: 536 537 - the value of the Authentication Code which indicates 538 use of the mechanism, 539 - the form and meaning of the Authentication Data, and 540 - the algorithm for computing values of Marker fields. 541 542 Note that a separate authentication mechanism may be 543 used in establishing the transport level connection. 544 545 Authentication Data: 546 547 The form and meaning of this field is a variable- 548 length field depend on the Authentication Code. 549 550 The minimum length of the OPEN message is 29 octets (including 551 message header). 552 553 554 555 556 557 558 559 560 561 562 Rekhter & Li [Page 10] 563 564 RFC 1771 BGP-4 March 1995 565 566 567 4.3 UPDATE Message Format 568 569 UPDATE messages are used to transfer routing information between BGP 570 peers. The information in the UPDATE packet can be used to construct 571 a graph describing the relationships of the various Autonomous 572 Systems. By applying rules to be discussed, routing information 573 loops and some other anomalies may be detected and removed from 574 inter-AS routing. 575 576 An UPDATE message is used to advertise a single feasible route to a 577 peer, or to withdraw multiple unfeasible routes from service (see 578 3.1). An UPDATE message may simultaneously advertise a feasible route 579 and withdraw multiple unfeasible routes from service. The UPDATE 580 message always includes the fixed-size BGP header, and can optionally 581 include the other fields as shown below: 582 583 +-----------------------------------------------------+ 584 | Unfeasible Routes Length (2 octets) | 585 +-----------------------------------------------------+ 586 | Withdrawn Routes (variable) | 587 +-----------------------------------------------------+ 588 | Total Path Attribute Length (2 octets) | 589 +-----------------------------------------------------+ 590 | Path Attributes (variable) | 591 +-----------------------------------------------------+ 592 | Network Layer Reachability Information (variable) | 593 +-----------------------------------------------------+ 594 595 Unfeasible Routes Length: 596 597 This 2-octets unsigned integer indicates the total length of 598 the Withdrawn Routes field in octets. Its value must allow the 599 length of the Network Layer Reachability Information field to 600 be determined as specified below. 601 602 A value of 0 indicates that no routes are being withdrawn from 603 service, and that the WITHDRAWN ROUTES field is not present in 604 this UPDATE message. 605 606 Withdrawn Routes: 607 608 This is a variable length field that contains a list of IP 609 address prefixes for the routes that are being withdrawn from 610 service. Each IP address prefix is encoded as a 2-tuple of the 611 form <length, prefix>, whose fields are described below: 612 613 614 615 616 617 618 Rekhter & Li [Page 11] 619 620 RFC 1771 BGP-4 March 1995 621 622 623 +---------------------------+ 624 | Length (1 octet) | 625 +---------------------------+ 626 | Prefix (variable) | 627 +---------------------------+ 628 629 The use and the meaning of these fields are as follows: 630 631 a) Length: 632 633 The Length field indicates the length in bits of the IP 634 address prefix. A length of zero indicates a prefix that 635 matches all IP addresses (with prefix, itself, of zero 636 octets). 637 638 b) Prefix: 639 640 The Prefix field contains IP address prefixes followed by 641 enough trailing bits to make the end of the field fall on an 642 octet boundary. Note that the value of trailing bits is 643 irrelevant. 644 645 Total Path Attribute Length: 646 647 This 2-octet unsigned integer indicates the total length of the 648 Path Attributes field in octets. Its value must allow the 649 length of the Network Layer Reachability field to be determined 650 as specified below. 651 652 A value of 0 indicates that no Network Layer Reachability 653 Information field is present in this UPDATE message. 654 655 Path Attributes: 656 657 A variable length sequence of path attributes is present in 658 every UPDATE. Each path attribute is a triple <attribute type, 659 attribute length, attribute value> of variable length. 660 661 Attribute Type is a two-octet field that consists of the 662 Attribute Flags octet followed by the Attribute Type Code 663 octet. 664 665 0 1 666 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 667 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 668 | Attr. Flags |Attr. Type Code| 669 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 670 671 672 673 674 Rekhter & Li [Page 12] 675 676 RFC 1771 BGP-4 March 1995 677 678 679 The high-order bit (bit 0) of the Attribute Flags octet is the 680 Optional bit. It defines whether the attribute is optional (if 681 set to 1) or well-known (if set to 0). 682 683 The second high-order bit (bit 1) of the Attribute Flags octet 684 is the Transitive bit. It defines whether an optional 685 attribute is transitive (if set to 1) or non-transitive (if set 686 to 0). For well-known attributes, the Transitive bit must be 687 set to 1. (See Section 5 for a discussion of transitive 688 attributes.) 689 690 The third high-order bit (bit 2) of the Attribute Flags octet 691 is the Partial bit. It defines whether the information 692 contained in the optional transitive attribute is partial (if 693 set to 1) or complete (if set to 0). For well-known attributes 694 and for optional non-transitive attributes the Partial bit must 695 be set to 0. 696 697 The fourth high-order bit (bit 3) of the Attribute Flags octet 698 is the Extended Length bit. It defines whether the Attribute 699 Length is one octet (if set to 0) or two octets (if set to 1). 700 Extended Length may be used only if the length of the attribute 701 value is greater than 255 octets. 702 703 The lower-order four bits of the Attribute Flags octet are . 704 unused. They must be zero (and must be ignored when received). 705 706 The Attribute Type Code octet contains the Attribute Type Code. 707 Currently defined Attribute Type Codes are discussed in Section 708 5. 709 710 If the Extended Length bit of the Attribute Flags octet is set 711 to 0, the third octet of the Path Attribute contains the length 712 of the attribute data in octets. 713 714 If the Extended Length bit of the Attribute Flags octet is set 715 to 1, then the third and the fourth octets of the path 716 attribute contain the length of the attribute data in octets. 717 718 The remaining octets of the Path Attribute represent the 719 attribute value and are interpreted according to the Attribute 720 Flags and the Attribute Type Code. The supported Attribute Type 721 Codes, their attribute values and uses are the following: 722 723 724 725 726 727 728 729 730 Rekhter & Li [Page 13] 731 732 RFC 1771 BGP-4 March 1995 733 734 735 a) ORIGIN (Type Code 1): 736 737 ORIGIN is a well-known mandatory attribute that defines the 738 origin of the path information. The data octet can assume 739 the following values: 740 741 Value Meaning 742 743 0 IGP - Network Layer Reachability Information 744 is interior to the originating AS 745 746 1 EGP - Network Layer Reachability Information 747 learned via EGP 748 749 2 INCOMPLETE - Network Layer Reachability 750 Information learned by some other means 751 752 Its usage is defined in 5.1.1 753 754 b) AS_PATH (Type Code 2): 755 756 AS_PATH is a well-known mandatory attribute that is composed 757 of a sequence of AS path segments. Each AS path segment is 758 represented by a triple <path segment type, path segment 759 length, path segment value>. 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 Rekhter & Li [Page 14] 787 788 RFC 1771 BGP-4 March 1995 789 790 791 The path segment type is a 1-octet long field with the 792 following values defined: 793 794 Value Segment Type 795 796 1 AS_SET: unordered set of ASs a route in the 797 UPDATE message has traversed 798 799 2 AS_SEQUENCE: ordered set of ASs a route in 800 the UPDATE message has traversed 801 802 The path segment length is a 1-octet long field containing 803 the number of ASs in the path segment value field. 804 805 The path segment value field contains one or more AS 806 numbers, each encoded as a 2-octets long field. 807 808 Usage of this attribute is defined in 5.1.2. 809 810 c) NEXT_HOP (Type Code 3): 811 812 This is a well-known mandatory attribute that defines the IP 813 address of the border router that should be used as the next 814 hop to the destinations listed in the Network Layer 815 Reachability field of the UPDATE message. 816 817 Usage of this attribute is defined in 5.1.3. 818 819 d) MULTI_EXIT_DISC (Type Code 4): 820 821 This is an optional non-transitive attribute that is a four 822 octet non-negative integer. The value of this attribute may 823 be used by a BGP speaker's decision process to discriminate 824 among multiple exit points to a neighboring autonomous 825 system. 826 827 Its usage is defined in 5.1.4. 828 829 e) LOCAL_PREF (Type Code 5): 830 831 LOCAL_PREF is a well-known discretionary attribute that is a 832 four octet non-negative integer. It is used by a BGP speaker 833 to inform other BGP speakers in its own autonomous system of 834 the originating speaker's degree of preference for an 835 advertised route. Usage of this attribute is described in 836 5.1.5. 837 838 839 840 841 842 Rekhter & Li [Page 15] 843 844 RFC 1771 BGP-4 March 1995 845 846 847 f) ATOMIC_AGGREGATE (Type Code 6) 848 849 ATOMIC_AGGREGATE is a well-known discretionary attribute of 850 length 0. It is used by a BGP speaker to inform other BGP 851 speakers that the local system selected a less specific 852 route without selecting a more specific route which is 853 included in it. Usage of this attribute is described in 854 5.1.6. 855 856 g) AGGREGATOR (Type Code 7) 857 858 AGGREGATOR is an optional transitive attribute of length 6. 859 The attribute contains the last AS number that formed the 860 aggregate route (encoded as 2 octets), followed by the IP 861 address of the BGP speaker that formed the aggregate route 862 (encoded as 4 octets). Usage of this attribute is described 863 in 5.1.7 864 865 Network Layer Reachability Information: 866 867 This variable length field contains a list of IP address 868 prefixes. The length in octets of the Network Layer 869 Reachability Information is not encoded explicitly, but can be 870 calculated as: 871 872 UPDATE message Length - 23 - Total Path Attributes Length - 873 Unfeasible Routes Length 874 875 where UPDATE message Length is the value encoded in the fixed- 876 size BGP header, Total Path Attribute Length and Unfeasible 877 Routes Length are the values encoded in the variable part of 878 the UPDATE message, and 23 is a combined length of the fixed- 879 size BGP header, the Total Path Attribute Length field and the 880 Unfeasible Routes Length field. 881 882 Reachability information is encoded as one or more 2-tuples of 883 the form <length, prefix>, whose fields are described below: 884 885 +---------------------------+ 886 | Length (1 octet) | 887 +---------------------------+ 888 | Prefix (variable) | 889 +---------------------------+ 890 891 892 893 894 895 896 897 898 Rekhter & Li [Page 16] 899 900 RFC 1771 BGP-4 March 1995 901 902 903 The use and the meaning of these fields are as follows: 904 905 a) Length: 906 907 The Length field indicates the length in bits of the IP 908 address prefix. A length of zero indicates a prefix that 909 matches all IP addresses (with prefix, itself, of zero 910 octets). 911 912 b) Prefix: 913 914 The Prefix field contains IP address prefixes followed by 915 enough trailing bits to make the end of the field fall on an 916 octet boundary. Note that the value of the trailing bits is 917 irrelevant. 918 919 The minimum length of the UPDATE message is 23 octets -- 19 octets 920 for the fixed header + 2 octets for the Unfeasible Routes Length + 2 921 octets for the Total Path Attribute Length (the value of Unfeasible 922 Routes Length is 0 and the value of Total Path Attribute Length is 923 0). 924 925 An UPDATE message can advertise at most one route, which may be 926 described by several path attributes. All path attributes contained 927 in a given UPDATE messages apply to the destinations carried in the 928 Network Layer Reachability Information field of the UPDATE message. 929 930 An UPDATE message can list multiple routes to be withdrawn from 931 service. Each such route is identified by its destination (expressed 932 as an IP prefix), which unambiguously identifies the route in the 933 context of the BGP speaker - BGP speaker connection to which it has 934 been previously been advertised. 935 936 An UPDATE message may advertise only routes to be withdrawn from 937 service, in which case it will not include path attributes or Network 938 Layer Reachability Information. Conversely, it may advertise only a 939 feasible route, in which case the WITHDRAWN ROUTES field need not be 940 present. 941 942 4.4 KEEPALIVE Message Format 943 944 BGP does not use any transport protocol-based keep-alive mechanism to 945 determine if peers are reachable. Instead, KEEPALIVE messages are 946 exchanged between peers often enough as not to cause the Hold Timer 947 to expire. A reasonable maximum time between KEEPALIVE messages 948 would be one third of the Hold Time interval. KEEPALIVE messages 949 MUST NOT be sent more frequently than one per second. An 950 implementation MAY adjust the rate at which it sends KEEPALIVE 951 952 953 954 Rekhter & Li [Page 17] 955 956 RFC 1771 BGP-4 March 1995 957 958 959 messages as a function of the Hold Time interval. 960 961 If the negotiated Hold Time interval is zero, then periodic KEEPALIVE 962 messages MUST NOT be sent. 963 964 KEEPALIVE message consists of only message header and has a length of 965 19 octets. 966 967 4.5 NOTIFICATION Message Format 968 969 A NOTIFICATION message is sent when an error condition is detected. 970 The BGP connection is closed immediately after sending it. 971 972 In addition to the fixed-size BGP header, the NOTIFICATION message 973 contains the following fields: 974 975 0 1 2 3 976 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 977 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 978 | Error code | Error subcode | Data | 979 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + 980 | | 981 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 982 983 Error Code: 984 985 This 1-octet unsigned integer indicates the type of 986 NOTIFICATION. The following Error Codes have been defined: 987 988 Error Code Symbolic Name Reference 989 990 1 Message Header Error Section 6.1 991 992 2 OPEN Message Error Section 6.2 993 994 3 UPDATE Message Error Section 6.3 995 996 4 Hold Timer Expired Section 6.5 997 998 5 Finite State Machine Error Section 6.6 999 1000 6 Cease Section 6.7 1001 1002 Error subcode: 1003 1004 This 1-octet unsigned integer provides more specific 1005 information about the nature of the reported error. Each Error 1006 Code may have one or more Error Subcodes associated with it. 1007 1008 1009 1010 Rekhter & Li [Page 18] 1011 1012 RFC 1771 BGP-4 March 1995 1013 1014 1015 If no appropriate Error Subcode is defined, then a zero 1016 (Unspecific) value is used for the Error Subcode field. 1017 1018 Message Header Error subcodes: 1019 1020 1 - Connection Not Synchronized. 1021 2 - Bad Message Length. 1022 3 - Bad Message Type. 1023 1024 OPEN Message Error subcodes: 1025 1026 1 - Unsupported Version Number. 1027 2 - Bad Peer AS. 1028 3 - Bad BGP Identifier. ' 1029 4 - Unsupported Optional Parameter. 1030 5 - Authentication Failure. 1031 6 - Unacceptable Hold Time. 1032 1033 UPDATE Message Error subcodes: 1034 1035 1 - Malformed Attribute List. 1036 2 - Unrecognized Well-known Attribute. 1037 3 - Missing Well-known Attribute. 1038 4 - Attribute Flags Error. 1039 5 - Attribute Length Error. 1040 6 - Invalid ORIGIN Attribute 1041 7 - AS Routing Loop. 1042 8 - Invalid NEXT_HOP Attribute. 1043 9 - Optional Attribute Error. 1044 10 - Invalid Network Field. 1045 11 - Malformed AS_PATH. 1046 1047 Data: 1048 1049 This variable-length field is used to diagnose the reason for 1050 the NOTIFICATION. The contents of the Data field depend upon 1051 the Error Code and Error Subcode. See Section 6 below for more 1052 details. 1053 1054 Note that the length of the Data field can be determined from 1055 the message Length field by the formula: 1056 1057 Message Length = 21 + Data Length 1058 1059 The minimum length of the NOTIFICATION message is 21 octets 1060 (including message header). 1061 1062 1063 1064 1065 1066 Rekhter & Li [Page 19] 1067 1068 RFC 1771 BGP-4 March 1995 1069 1070 1071 5. Path Attributes 1072 1073 This section discusses the path attributes of the UPDATE message. 1074 1075 Path attributes fall into four separate categories: 1076 1077 1. Well-known mandatory. 1078 2. Well-known discretionary. 1079 3. Optional transitive. 1080 4. Optional non-transitive. 1081 1082 Well-known attributes must be recognized by all BGP implementations. 1083 Some of these attributes are mandatory and must be included in every 1084 UPDATE message. Others are discretionary and may or may not be sent 1085 in a particular UPDATE message. 1086 1087 All well-known attributes must be passed along (after proper 1088 updating, if necessary) to other BGP peers. 1089 1090 In addition to well-known attributes, each path may contain one or 1091 more optional attributes. It is not required or expected that all 1092 BGP implementations support all optional attributes. The handling of 1093 an unrecognized optional attribute is determined by the setting of 1094 the Transitive bit in the attribute flags octet. Paths with 1095 unrecognized transitive optional attributes should be accepted. If a 1096 path with unrecognized transitive optional attribute is accepted and 1097 passed along to other BGP peers, then the unrecognized transitive 1098 optional attribute of that path must be passed along with the path to 1099 other BGP peers with the Partial bit in the Attribute Flags octet set 1100 to 1. If a path with recognized transitive optional attribute is 1101 accepted and passed along to other BGP peers and the Partial bit in 1102 the Attribute Flags octet is set to 1 by some previous AS, it is not 1103 set back to 0 by the current AS. Unrecognized non-transitive optional 1104 attributes must be quietly ignored and not passed along to other BGP 1105 peers. 1106 1107 New transitive optional attributes may be attached to the path by the 1108 originator or by any other AS in the path. If they are not attached 1109 by the originator, the Partial bit in the Attribute Flags octet is 1110 set to 1. The rules for attaching new non-transitive optional 1111 attributes will depend on the nature of the specific attribute. The 1112 documentation of each new non-transitive optional attribute will be 1113 expected to include such rules. (The description of the 1114 MULTI_EXIT_DISC attribute gives an example.) All optional attributes 1115 (both transitive and non-transitive) may be updated (if appropriate) 1116 by ASs in the path. 1117 1118 1119 1120 1121 1122 Rekhter & Li [Page 20] 1123 1124 RFC 1771 BGP-4 March 1995 1125 1126 1127 The sender of an UPDATE message should order path attributes within 1128 the UPDATE message in ascending order of attribute type. The 1129 receiver of an UPDATE message must be prepared to handle path 1130 attributes within the UPDATE message that are out of order. 1131 1132 The same attribute cannot appear more than once within the Path 1133 Attributes field of a particular UPDATE message. 1134 1135 5.1 Path Attribute Usage 1136 1137 The usage of each BGP path attributes is described in the following 1138 clauses. 1139 1140 5.1.1 ORIGIN 1141 1142 ORIGIN is a well-known mandatory attribute. The ORIGIN attribute 1143 shall be generated by the autonomous system that originates the 1144 associated routing information. It shall be included in the UPDATE 1145 messages of all BGP speakers that choose to propagate this 1146 information to other BGP speakers. 1147 1148 5.1.2 AS_PATH 1149 1150 AS_PATH is a well-known mandatory attribute. This attribute 1151 identifies the autonomous systems through which routing information 1152 carried in this UPDATE message has passed. The components of this 1153 list can be AS_SETs or AS_SEQUENCEs. 1154 1155 When a BGP speaker propagates a route which it has learned from 1156 another BGP speaker's UPDATE message, it shall modify the route's 1157 AS_PATH attribute based on the location of the BGP speaker to which 1158 the route will be sent: 1159 1160 a) When a given BGP speaker advertises the route to another BGP 1161 speaker located in its own autonomous system, the advertising 1162 speaker shall not modify the AS_PATH attribute associated with the 1163 route. 1164 1165 b) When a given BGP speaker advertises the route to a BGP speaker 1166 located in a neighboring autonomous system, then the advertising 1167 speaker shall update the AS_PATH attribute as follows: 1168 1169 1) if the first path segment of the AS_PATH is of type 1170 AS_SEQUENCE, the local system shall prepend its own AS number 1171 as the last element of the sequence (put it in the leftmost 1172 position). 1173 1174 1175 1176 1177 1178 Rekhter & Li [Page 21] 1179 1180 RFC 1771 BGP-4 March 1995 1181 1182 1183 2) if the first path segment of the AS_PATH is of type AS_SET, 1184 the local system shall prepend a new path segment of type 1185 AS_SEQUENCE to the AS_PATH, including its own AS number in that 1186 segment. 1187 1188 When a BGP speaker originates a route then: 1189 1190 a) the originating speaker shall include its own AS number in 1191 the AS_PATH attribute of all UPDATE messages sent to BGP 1192 speakers located in neighboring autonomous systems. (In this 1193 case, the AS number of the originating speaker's autonomous 1194 system will be the only entry in the AS_PATH attribute). 1195 1196 b) the originating speaker shall include an empty AS_PATH 1197 attribute in all UPDATE messages sent to BGP speakers located 1198 in its own autonomous system. (An empty AS_PATH attribute is 1199 one whose length field contains the value zero). 1200 1201 5.1.3 NEXT_HOP 1202 1203 The NEXT_HOP path attribute defines the IP address of the border 1204 router that should be used as the next hop to the destinations listed 1205 in the UPDATE message. If a border router belongs to the same AS as 1206 its peer, then the peer is an internal border router. Otherwise, it 1207 is an external border router. A BGP speaker can advertise any 1208 internal border router as the next hop provided that the interface 1209 associated with the IP address of this border router (as specified in 1210 the NEXT_HOP path attribute) shares a common subnet with both the 1211 local and remote BGP speakers. A BGP speaker can advertise any 1212 external border router as the next hop, provided that the IP address 1213 of this border router was learned from one of the BGP speaker's 1214 peers, and the interface associated with the IP address of this 1215 border router (as specified in the NEXT_HOP path attribute) shares a 1216 common subnet with the local and remote BGP speakers. A BGP speaker 1217 needs to be able to support disabling advertisement of external 1218 border routers. 1219 1220 A BGP speaker must never advertise an address of a peer to that peer 1221 as a NEXT_HOP, for a route that the speaker is originating. A BGP 1222 speaker must never install a route with itself as the next hop. 1223 1224 When a BGP speaker advertises the route to a BGP speaker located in 1225 its own autonomous system, the advertising speaker shall not modify 1226 the NEXT_HOP attribute associated with the route. When a BGP speaker 1227 receives the route via an internal link, it may forward packets to 1228 the NEXT_HOP address if the address contained in the attribute is on 1229 a common subnet with the local and remote BGP speakers. 1230 1231 1232 1233 1234 Rekhter & Li [Page 22] 1235 1236 RFC 1771 BGP-4 March 1995 1237 1238 1239 5.1.4 MULTI_EXIT_DISC 1240 1241 The MULTI_EXIT_DISC attribute may be used on external (inter-AS) 1242 links to discriminate among multiple exit or entry points to the same 1243 neighboring AS. The value of the MULTI_EXIT_DISC attribute is a four 1244 octet unsigned number which is called a metric. All other factors 1245 being equal, the exit or entry point with lower metric should be 1246 preferred. If received over external links, the MULTI_EXIT_DISC 1247 attribute may be propagated over internal links to other BGP speakers 1248 within the same AS. The MULTI_EXIT_DISC attribute is never 1249 propagated to other BGP speakers in neighboring AS's. 1250 1251 5.1.5 LOCAL_PREF 1252 1253 LOCAL_PREF is a well-known discretionary attribute that shall be 1254 included in all UPDATE messages that a given BGP speaker sends to the 1255 other BGP speakers located in its own autonomous system. A BGP 1256 speaker shall calculate the degree of preference for each external 1257 route and include the degree of preference when advertising a route 1258 to its internal peers. The higher degree of preference should be 1259 preferred. A BGP speaker shall use the degree of preference learned 1260 via LOCAL_PREF in its decision process (see section 9.1.1). 1261 1262 A BGP speaker shall not include this attribute in UPDATE messages 1263 that it sends to BGP speakers located in a neighboring autonomous 1264 system. If it is contained in an UPDATE message that is received from 1265 a BGP speaker which is not located in the same autonomous system as 1266 the receiving speaker, then this attribute shall be ignored by the 1267 receiving speaker. 1268 1269 5.1.6 ATOMIC_AGGREGATE 1270 1271 ATOMIC_AGGREGATE is a well-known discretionary attribute. If a BGP 1272 speaker, when presented with a set of overlapping routes from one of 1273 its peers (see 9.1.4), selects the less specific route without 1274 selecting the more specific one, then the local system shall attach 1275 the ATOMIC_AGGREGATE attribute to the route when propagating it to 1276 other BGP speakers (if that attribute is not already present in the 1277 received less specific route). A BGP speaker that receives a route 1278 with the ATOMIC_AGGREGATE attribute shall not remove the attribute 1279 from the route when propagating it to other speakers. A BGP speaker 1280 that receives a route with the ATOMIC_AGGREGATE attribute shall not 1281 make any NLRI of that route more specific (as defined in 9.1.4) when 1282 advertising this route to other BGP speakers. A BGP speaker that 1283 receives a route with the ATOMIC_AGGREGATE attribute needs to be 1284 cognizant of the fact that the actual path to destinations, as 1285 specified in the NLRI of the route, while having the loop-free 1286 property, may traverse ASs that are not listed in the AS_PATH 1287 1288 1289 1290 Rekhter & Li [Page 23] 1291 1292 RFC 1771 BGP-4 March 1995 1293 1294 1295 attribute. 1296 1297 5.1.7 AGGREGATOR 1298 1299 AGGREGATOR is an optional transitive attribute which may be included 1300 in updates which are formed by aggregation (see Section 9.2.4.2). A 1301 BGP speaker which performs route aggregation may add the AGGREGATOR 1302 attribute which shall contain its own AS number and IP address. 1303 1304 6. BGP Error Handling. 1305 1306 This section describes actions to be taken when errors are detected 1307 while processing BGP messages. 1308 1309 When any of the conditions described here are detected, a 1310 NOTIFICATION message with the indicated Error Code, Error Subcode, 1311 and Data fields is sent, and the BGP connection is closed. If no 1312 Error Subcode is specified, then a zero must be used. 1313 1314 The phrase "the BGP connection is closed" means that the transport 1315 protocol connection has been closed and that all resources for that 1316 BGP connection have been deallocated. Routing table entries 1317 associated with the remote peer are marked as invalid. The fact that 1318 the routes have become invalid is passed to other BGP peers before 1319 the routes are deleted from the system. 1320 1321 Unless specified explicitly, the Data field of the NOTIFICATION 1322 message that is sent to indicate an error is empty. 1323 1324 6.1 Message Header error handling. 1325 1326 All errors detected while processing the Message Header are indicated 1327 by sending the NOTIFICATION message with Error Code Message Header 1328 Error. The Error Subcode elaborates on the specific nature of the 1329 error. 1330 1331 The expected value of the Marker field of the message header is all 1332 ones if the message type is OPEN. The expected value of the Marker 1333 field for all other types of BGP messages determined based on the 1334 presence of the Authentication Information Optional Parameter in the 1335 BGP OPEN message and the actual authentication mechanism (if the 1336 Authentication Information in the BGP OPEN message is present). If 1337 the Marker field of the message header is not the expected one, then 1338 a synchronization error has occurred and the Error Subcode is set to 1339 Connection Not Synchronized. 1340 1341 1342 1343 1344 1345 1346 Rekhter & Li [Page 24] 1347 1348 RFC 1771 BGP-4 March 1995 1349 1350 1351 If the Length field of the message header is less than 19 or greater 1352 than 4096, or if the Length field of an OPEN message is less than 1353 the minimum length of the OPEN message, or if the Length field of an 1354 UPDATE message is less than the minimum length of the UPDATE message, 1355 or if the Length field of a KEEPALIVE message is not equal to 19, or 1356 if the Length field of a NOTIFICATION message is less than the 1357 minimum length of the NOTIFICATION message, then the Error Subcode is 1358 set to Bad Message Length. The Data field contains the erroneous 1359 Length field. 1360 1361 If the Type field of the message header is not recognized, then the 1362 Error Subcode is set to Bad Message Type. The Data field contains 1363 the erroneous Type field. 1364 1365 6.2 OPEN message error handling. 1366 1367 All errors detected while processing the OPEN message are indicated 1368 by sending the NOTIFICATION message with Error Code OPEN Message 1369 Error. The Error Subcode elaborates on the specific nature of the 1370 error. 1371 1372 If the version number contained in the Version field of the received 1373 OPEN message is not supported, then the Error Subcode is set to 1374 Unsupported Version Number. The Data field is a 2-octet unsigned 1375 integer, which indicates the largest locally supported version number 1376 less than the version the remote BGP peer bid (as indicated in the 1377 received OPEN message). 1378 1379 If the Autonomous System field of the OPEN message is unacceptable, 1380 then the Error Subcode is set to Bad Peer AS. The determination of 1381 acceptable Autonomous System numbers is outside the scope of this 1382 protocol. 1383 1384 If the Hold Time field of the OPEN message is unacceptable, then the 1385 Error Subcode MUST be set to Unacceptable Hold Time. An 1386 implementation MUST reject Hold Time values of one or two seconds. 1387 An implementation MAY reject any proposed Hold Time. An 1388 implementation which accepts a Hold Time MUST use the negotiated 1389 value for the Hold Time. 1390 1391 If the BGP Identifier field of the OPEN message is syntactically 1392 incorrect, then the Error Subcode is set to Bad BGP Identifier. 1393 Syntactic correctness means that the BGP Identifier field represents 1394 a valid IP host address. 1395 1396 If one of the Optional Parameters in the OPEN message is not 1397 recognized, then the Error Subcode is set to Unsupported Optional 1398 Parameters. 1399 1400 1401 1402 Rekhter & Li [Page 25] 1403 1404 RFC 1771 BGP-4 March 1995 1405 1406 1407 If the OPEN message carries Authentication Information (as an 1408 Optional Parameter), then the corresponding authentication procedure 1409 is invoked. If the authentication procedure (based on Authentication 1410 Code and Authentication Data) fails, then the Error Subcode is set to 1411 Authentication Failure. 1412 1413 6.3 UPDATE message error handling. 1414 1415 All errors detected while processing the UPDATE message are indicated 1416 by sending the NOTIFICATION message with Error Code UPDATE Message 1417 Error. The error subcode elaborates on the specific nature of the 1418 error. 1419 1420 Error checking of an UPDATE message begins by examining the path 1421 attributes. If the Unfeasible Routes Length or Total Attribute 1422 Length is too large (i.e., if Unfeasible Routes Length + Total 1423 Attribute Length + 23 exceeds the message Length), then the Error 1424 Subcode is set to Malformed Attribute List. 1425 1426 If any recognized attribute has Attribute Flags that conflict with 1427 the Attribute Type Code, then the Error Subcode is set to Attribute 1428 Flags Error. The Data field contains the erroneous attribute (type, 1429 length and value). 1430 1431 If any recognized attribute has Attribute Length that conflicts with 1432 the expected length (based on the attribute type code), then the 1433 Error Subcode is set to Attribute Length Error. The Data field 1434 contains the erroneous attribute (type, length and value). 1435 1436 If any of the mandatory well-known attributes are not present, then 1437 the Error Subcode is set to Missing Well-known Attribute. The Data 1438 field contains the Attribute Type Code of the missing well-known 1439 attribute. 1440 1441 If any of the mandatory well-known attributes are not recognized, 1442 then the Error Subcode is set to Unrecognized Well-known Attribute. 1443 The Data field contains the unrecognized attribute (type, length and 1444 value). 1445 1446 If the ORIGIN attribute has an undefined value, then the Error 1447 Subcode is set to Invalid Origin Attribute. The Data field contains 1448 the unrecognized attribute (type, length and value). 1449 1450 If the NEXT_HOP attribute field is syntactically incorrect, then the 1451 Error Subcode is set to Invalid NEXT_HOP Attribute. The Data field 1452 contains the incorrect attribute (type, length and value). Syntactic 1453 correctness means that the NEXT_HOP attribute represents a valid IP 1454 host address. Semantic correctness applies only to the external BGP 1455 1456 1457 1458 Rekhter & Li [Page 26] 1459 1460 RFC 1771 BGP-4 March 1995 1461 1462 1463 links. It means that the interface associated with the IP address, as 1464 specified in the NEXT_HOP attribute, shares a common subnet with the 1465 receiving BGP speaker and is not the IP address of the receiving BGP 1466 speaker. If the NEXT_HOP attribute is semantically incorrect, the 1467 error should be logged, and the the route should be ignored. In this 1468 case, no NOTIFICATION message should be sent. 1469 1470 The AS_PATH attribute is checked for syntactic correctness. If the 1471 path is syntactically incorrect, then the Error Subcode is set to 1472 Malformed AS_PATH. 1473 1474 If an optional attribute is recognized, then the value of this 1475 attribute is checked. If an error is detected, the attribute is 1476 discarded, and the Error Subcode is set to Optional Attribute Error. 1477 The Data field contains the attribute (type, length and value). 1478 1479 If any attribute appears more than once in the UPDATE message, then 1480 the Error Subcode is set to Malformed Attribute List. 1481 1482 The NLRI field in the UPDATE message is checked for syntactic 1483 validity. If the field is syntactically incorrect, then the Error 1484 Subcode is set to Invalid Network Field. 1485 1486 6.4 NOTIFICATION message error handling. 1487 1488 If a peer sends a NOTIFICATION message, and there is an error in that 1489 message, there is unfortunately no means of reporting this error via 1490 a subsequent NOTIFICATION message. Any such error, such as an 1491 unrecognized Error Code or Error Subcode, should be noticed, logged 1492 locally, and brought to the attention of the administration of the 1493 peer. The means to do this, however, lies outside the scope of this 1494 document. 1495 1496 6.5 Hold Timer Expired error handling. 1497 1498 If a system does not receive successive KEEPALIVE and/or UPDATE 1499 and/or NOTIFICATION messages within the period specified in the Hold 1500 Time field of the OPEN message, then the NOTIFICATION message with 1501 Hold Timer Expired Error Code must be sent and the BGP connection 1502 closed. 1503 1504 6.6 Finite State Machine error handling. 1505 1506 Any error detected by the BGP Finite State Machine (e.g., receipt of 1507 an unexpected event) is indicated by sending the NOTIFICATION message 1508 with Error Code Finite State Machine Error. 1509 1510 1511 1512 1513 1514 Rekhter & Li [Page 27] 1515 1516 RFC 1771 BGP-4 March 1995 1517 1518 1519 6.7 Cease. 1520 1521 In absence of any fatal errors (that are indicated in this section), 1522 a BGP peer may choose at any given time to close its BGP connection 1523 by sending the NOTIFICATION message with Error Code Cease. However, 1524 the Cease NOTIFICATION message must not be used when a fatal error 1525 indicated by this section does exist. 1526 1527 6.8 Connection collision detection. 1528 1529 If a pair of BGP speakers try simultaneously to establish a TCP 1530 connection to each other, then two parallel connections between this 1531 pair of speakers might well be formed. We refer to this situation as 1532 connection collision. Clearly, one of these connections must be 1533 closed. 1534 1535 Based on the value of the BGP Identifier a convention is established 1536 for detecting which BGP connection is to be preserved when a 1537 collision does occur. The convention is to compare the BGP 1538 Identifiers of the peers involved in the collision and to retain only 1539 the connection initiated by the BGP speaker with the higher-valued 1540 BGP Identifier. 1541 1542 Upon receipt of an OPEN message, the local system must examine all of 1543 its connections that are in the OpenConfirm state. A BGP speaker may 1544 also examine connections in an OpenSent state if it knows the BGP 1545 Identifier of the peer by means outside of the protocol. If among 1546 these connections there is a connection to a remote BGP speaker whose 1547 BGP Identifier equals the one in the OPEN message, then the local 1548 system performs the following collision resolution procedure: 1549 1550 1. The BGP Identifier of the local system is compared to the BGP 1551 Identifier of the remote system (as specified in the OPEN 1552 message). 1553 1554 2. If the value of the local BGP Identifier is less than the 1555 remote one, the local system closes BGP connection that already 1556 exists (the one that is already in the OpenConfirm state), and 1557 accepts BGP connection initiated by the remote system. 1558 1559 3. Otherwise, the local system closes newly created BGP connection 1560 (the one associated with the newly received OPEN message), and 1561 continues to use the existing one (the one that is already in the 1562 OpenConfirm state). 1563 1564 Comparing BGP Identifiers is done by treating them as (4-octet 1565 long) unsigned integers. 1566 1567 1568 1569 1570 Rekhter & Li [Page 28] 1571 1572 RFC 1771 BGP-4 March 1995 1573 1574 1575 A connection collision with an existing BGP connection that is in 1576 Established states causes unconditional closing of the newly 1577 created connection. Note that a connection collision cannot be 1578 detected with connections that are in Idle, or Connect, or Active 1579 states. 1580 1581 Closing the BGP connection (that results from the collision 1582 resolution procedure) is accomplished by sending the NOTIFICATION 1583 message with the Error Code Cease. 1584 1585 7. BGP Version Negotiation. 1586 1587 BGP speakers may negotiate the version of the protocol by making 1588 multiple attempts to open a BGP connection, starting with the highest 1589 version number each supports. If an open attempt fails with an Error 1590 Code OPEN Message Error, and an Error Subcode Unsupported Version 1591 Number, then the BGP speaker has available the version number it 1592 tried, the version number its peer tried, the version number passed 1593 by its peer in the NOTIFICATION message, and the version numbers that 1594 it supports. If the two peers do support one or more common 1595 versions, then this will allow them to rapidly determine the highest 1596 common version. In order to support BGP version negotiation, future 1597 versions of BGP must retain the format of the OPEN and NOTIFICATION 1598 messages. 1599 1600 8. BGP Finite State machine. 1601 1602 This section specifies BGP operation in terms of a Finite State 1603 Machine (FSM). Following is a brief summary and overview of BGP 1604 operations by state as determined by this FSM. A condensed version 1605 of the BGP FSM is found in Appendix 1. 1606 1607 Initially BGP is in the Idle state. 1608 1609 Idle state: 1610 1611 In this state BGP refuses all incoming BGP connections. No 1612 resources are allocated to the peer. In response to the Start 1613 event (initiated by either system or operator) the local system 1614 initializes all BGP resources, starts the ConnectRetry timer, 1615 initiates a transport connection to other BGP peer, while 1616 listening for connection that may be initiated by the remote 1617 BGP peer, and changes its state to Connect. The exact value of 1618 the ConnectRetry timer is a local matter, but should be 1619 sufficiently large to allow TCP initialization. 1620 1621 If a BGP speaker detects an error, it shuts down the connection 1622 and changes its state to Idle. Getting out of the Idle state 1623 1624 1625 1626 Rekhter & Li [Page 29] 1627 1628 RFC 1771 BGP-4 March 1995 1629 1630 1631 requires generation of the Start event. If such an event is 1632 generated automatically, then persistent BGP errors may result 1633 in persistent flapping of the speaker. To avoid such a 1634 condition it is recommended that Start events should not be 1635 generated immediately for a peer that was previously 1636 transitioned to Idle due to an error. For a peer that was 1637 previously transitioned to Idle due to an error, the time 1638 between consecutive generation of Start events, if such events 1639 are generated automatically, shall exponentially increase. The 1640 value of the initial timer shall be 60 seconds. The time shall 1641 be doubled for each consecutive retry. 1642 1643 Any other event received in the Idle state is ignored. 1644 1645 Connect state: 1646 1647 In this state BGP is waiting for the transport protocol 1648 connection to be completed. 1649 1650 If the transport protocol connection succeeds, the local system 1651 clears the ConnectRetry timer, completes initialization, sends 1652 an OPEN message to its peer, and changes its state to OpenSent. 1653 1654 If the transport protocol connect fails (e.g., retransmission 1655 timeout), the local system restarts the ConnectRetry timer, 1656 continues to listen for a connection that may be initiated by 1657 the remote BGP peer, and changes its state to Active state. 1658 1659 In response to the ConnectRetry timer expired event, the local 1660 system restarts the ConnectRetry timer, initiates a transport 1661 connection to other BGP peer, continues to listen for a 1662 connection that may be initiated by the remote BGP peer, and 1663 stays in the Connect state. 1664 1665 Start event is ignored in the Active state. 1666 1667 In response to any other event (initiated by either system or 1668 operator), the local system releases all BGP resources 1669 associated with this connection and changes its state to Idle. 1670 1671 Active state: 1672 1673 In this state BGP is trying to acquire a peer by initiating a 1674 transport protocol connection. 1675 1676 If the transport protocol connection succeeds, the local system 1677 clears the ConnectRetry timer, completes initialization, sends 1678 an OPEN message to its peer, sets its Hold Timer to a large 1679 1680 1681 1682 Rekhter & Li [Page 30] 1683 1684 RFC 1771 BGP-4 March 1995 1685 1686 1687 value, and changes its state to OpenSent. A Hold Timer value 1688 of 4 minutes is suggested. 1689 1690 In response to the ConnectRetry timer expired event, the local 1691 system restarts the ConnectRetry timer, initiates a transport 1692 connection to other BGP peer, continues to listen for a 1693 connection that may be initiated by the remote BGP peer, and 1694 changes its state to Connect. 1695 1696 If the local system detects that a remote peer is trying to 1697 establish BGP connection to it, and the IP address of the 1698 remote peer is not an expected one, the local system restarts 1699 the ConnectRetry timer, rejects the attempted connection, 1700 continues to listen for a connection that may be initiated by 1701 the remote BGP peer, and stays in the Active state. 1702 1703 Start event is ignored in the Active state. 1704 1705 In response to any other event (initiated by either system or 1706 operator), the local system releases all BGP resources 1707 associated with this connection and changes its state to Idle. 1708 1709 OpenSent state: 1710 1711 In this state BGP waits for an OPEN message from its peer. 1712 When an OPEN message is received, all fields are checked for 1713 correctness. If the BGP message header checking or OPEN 1714 message checking detects an error (see Section 6.2), or a 1715 connection collision (see Section 6.8) the local system sends a 1716 NOTIFICATION message and changes its state to Idle. 1717 1718 If there are no errors in the OPEN message, BGP sends a 1719 KEEPALIVE message and sets a KeepAlive timer. The Hold Timer, 1720 which was originally set to a large value (see above), is 1721 replaced with the negotiated Hold Time value (see section 4.2). 1722 If the negotiated Hold Time value is zero, then the Hold Time 1723 timer and KeepAlive timers are not started. If the value of 1724 the Autonomous System field is the same as the local Autonomous 1725 System number, then the connection is an "internal" connection; 1726 otherwise, it is "external". (This will effect UPDATE 1727 processing as described below.) Finally, the state is changed 1728 to OpenConfirm. 1729 1730 If a disconnect notification is received from the underlying 1731 transport protocol, the local system closes the BGP connection, 1732 restarts the ConnectRetry timer, while continue listening for 1733 connection that may be initiated by the remote BGP peer, and 1734 goes into the Active state. 1735 1736 1737 1738 Rekhter & Li [Page 31] 1739 1740 RFC 1771 BGP-4 March 1995 1741 1742 1743 If the Hold Timer expires, the local system sends NOTIFICATION 1744 message with error code Hold Timer Expired and changes its 1745 state to Idle. 1746 1747 In response to the Stop event (initiated by either system or 1748 operator) the local system sends NOTIFICATION message with 1749 Error Code Cease and changes its state to Idle. 1750 1751 Start event is ignored in the OpenSent state. 1752 1753 In response to any other event the local system sends 1754 NOTIFICATION message with Error Code Finite State Machine Error 1755 and changes its state to Idle. 1756 1757 Whenever BGP changes its state from OpenSent to Idle, it closes 1758 the BGP (and transport-level) connection and releases all 1759 resources associated with that connection. 1760 1761 OpenConfirm state: 1762 1763 In this state BGP waits for a KEEPALIVE or NOTIFICATION 1764 message. 1765 1766 If the local system receives a KEEPALIVE message, it changes 1767 its state to Established. 1768 1769 If the Hold Timer expires before a KEEPALIVE message is 1770 received, the local system sends NOTIFICATION message with 1771 error code Hold Timer Expired and changes its state to Idle. 1772 1773 If the local system receives a NOTIFICATION message, it changes 1774 its state to Idle. 1775 1776 If the KeepAlive timer expires, the local system sends a 1777 KEEPALIVE message and restarts its KeepAlive timer. 1778 1779 If a disconnect notification is received from the underlying 1780 transport protocol, the local system changes its state to Idle. 1781 1782 In response to the Stop event (initiated by either system or 1783 operator) the local system sends NOTIFICATION message with 1784 Error Code Cease and changes its state to Idle. 1785 1786 Start event is ignored in the OpenConfirm state. 1787 1788 In response to any other event the local system sends 1789 NOTIFICATION message with Error Code Finite State Machine Error 1790 and changes its state to Idle. 1791 1792 1793 1794 Rekhter & Li [Page 32] 1795 1796 RFC 1771 BGP-4 March 1995 1797 1798 1799 Whenever BGP changes its state from OpenConfirm to Idle, it 1800 closes the BGP (and transport-level) connection and releases 1801 all resources associated with that connection. 1802 1803 Established state: 1804 1805 In the Established state BGP can exchange UPDATE, NOTIFICATION, 1806 and KEEPALIVE messages with its peer. 1807 1808 If the local system receives an UPDATE or KEEPALIVE message, it 1809 restarts its Hold Timer, if the negotiated Hold Time value is 1810 non-zero. 1811 1812 If the local system receives a NOTIFICATION message, it changes 1813 its state to Idle. 1814 1815 If the local system receives an UPDATE message and the UPDATE 1816 message error handling procedure (see Section 6.3) detects an 1817 error, the local system sends a NOTIFICATION message and 1818 changes its state to Idle. 1819 1820 If a disconnect notification is received from the underlying 1821 transport protocol, the local system changes its state to Idle. 1822 1823 If the Hold Timer expires, the local system sends a 1824 NOTIFICATION message with Error Code Hold Timer Expired and 1825 changes its state to Idle. 1826 1827 If the KeepAlive timer expires, the local system sends a 1828 KEEPALIVE message and restarts its KeepAlive timer. 1829 1830 Each time the local system sends a KEEPALIVE or UPDATE message, 1831 it restarts its KeepAlive timer, unless the negotiated Hold 1832 Time value is zero. 1833 1834 In response to the Stop event (initiated by either system or 1835 operator), the local system sends a NOTIFICATION message with 1836 Error Code Cease and changes its state to Idle. 1837 1838 Start event is ignored in the Established state. 1839 1840 In response to any other event, the local system sends 1841 NOTIFICATION message with Error Code Finite State Machine Error 1842 and changes its state to Idle. 1843 1844 Whenever BGP changes its state from Established to Idle, it 1845 closes the BGP (and transport-level) connection, releases all 1846 resources associated with that connection, and deletes all 1847 1848 1849 1850 Rekhter & Li [Page 33] 1851 1852 RFC 1771 BGP-4 March 1995 1853 1854 1855 routes derived from that connection. 1856 1857 9. UPDATE Message Handling 1858 1859 An UPDATE message may be received only in the Established state. 1860 When an UPDATE message is received, each field is checked for 1861 validity as specified in Section 6.3. 1862 1863 If an optional non-transitive attribute is unrecognized, it is 1864 quietly ignored. If an optional transitive attribute is 1865 unrecognized, the Partial bit (the third high-order bit) in the 1866 attribute flags octet is set to 1, and the attribute is retained for 1867 propagation to other BGP speakers. 1868 1869 If an optional attribute is recognized, and has a valid value, then, 1870 depending on the type of the optional attribute, it is processed 1871 locally, retained, and updated, if necessary, for possible 1872 propagation to other BGP speakers. 1873 1874 If the UPDATE message contains a non-empty WITHDRAWN ROUTES field, 1875 the previously advertised routes whose destinations (expressed as IP 1876 prefixes) contained in this field shall be removed from the Adj-RIB- 1877 In. This BGP speaker shall run its Decision Process since the 1878 previously advertised route is not longer available for use. 1879 1880 If the UPDATE message contains a feasible route, it shall be placed 1881 in the appropriate Adj-RIB-In, and the following additional actions 1882 shall be taken: 1883 1884 i) If its Network Layer Reachability Information (NLRI) is identical 1885 to the one of a route currently stored in the Adj-RIB-In, then the 1886 new route shall replace the older route in the Adj-RIB-In, thus 1887 implicitly withdrawing the older route from service. The BGP speaker 1888 shall run its Decision Process since the older route is no longer 1889 available for use. 1890 1891 ii) If the new route is an overlapping route that is included (see 1892 9.1.4) in an earlier route contained in the Adj-RIB-In, the BGP 1893 speaker shall run its Decision Process since the more specific route 1894 has implicitly made a portion of the less specific route unavailable 1895 for use. 1896 1897 iii) If the new route has identical path attributes to an earlier 1898 route contained in the Adj-RIB-In, and is more specific (see 9.1.4) 1899 than the earlier route, no further actions are necessary. 1900 1901 iv) If the new route has NLRI that is not present in any of the 1902 routes currently stored in the Adj-RIB-In, then the new route shall 1903 1904 1905 1906 Rekhter & Li [Page 34] 1907 1908 RFC 1771 BGP-4 March 1995 1909 1910 1911 be placed in the Adj-RIB-In. The BGP speaker shall run its Decision 1912 Process. 1913 1914 v) If the new route is an overlapping route that is less specific 1915 (see 9.1.4) than an earlier route contained in the Adj-RIB-In, the 1916 BGP speaker shall run its Decision Process on the set of destinations 1917 described only by the less specific route. 1918 1919 9.1 Decision Process 1920 1921 The Decision Process selects routes for subsequent advertisement by 1922 applying the policies in the local Policy Information Base (PIB) to 1923 the routes stored in its Adj-RIB-In. The output of the Decision 1924 Process is the set of routes that will be advertised to all peers; 1925 the selected routes will be stored in the local speaker's Adj-RIB- 1926 Out. 1927 1928 The selection process is formalized by defining a function that takes 1929 the attribute of a given route as an argument and returns a non- 1930 negative integer denoting the degree of preference for the route. 1931 The function that calculates the degree of preference for a given 1932 route shall not use as its inputs any of the following: the 1933 existence of other routes, the non-existence of other routes, or the 1934 path attributes of other routes. Route selection then consists of 1935 individual application of the degree of preference function to each 1936 feasible route, followed by the choice of the one with the highest 1937 degree of preference. 1938 1939 The Decision Process operates on routes contained in each Adj-RIB-In, 1940 and is responsible for: 1941 1942 - selection of routes to be advertised to BGP speakers located in 1943 the local speaker's autonomous system 1944 1945 - selection of routes to be advertised to BGP speakers located in 1946 neighboring autonomous systems 1947 1948 - route aggregation and route information reduction 1949 1950 The Decision Process takes place in three distinct phases, each 1951 triggered by a different event: 1952 1953 a) Phase 1 is responsible for calculating the degree of preference 1954 for each route received from a BGP speaker located in a 1955 neighboring autonomous system, and for advertising to the other 1956 BGP speakers in the local autonomous system the routes that have 1957 the highest degree of preference for each distinct destination. 1958 1959 1960 1961 1962 Rekhter & Li [Page 35] 1963 1964 RFC 1771 BGP-4 March 1995 1965 1966 1967 b) Phase 2 is invoked on completion of phase 1. It is responsible 1968 for choosing the best route out of all those available for each 1969 distinct destination, and for installing each chosen route into 1970 the appropriate Loc-RIB. 1971 1972 c) Phase 3 is invoked after the Loc-RIB has been modified. It is 1973 responsible for disseminating routes in the Loc-RIB to each peer 1974 located in a neighboring autonomous system, according to the 1975 policies contained in the PIB. Route aggregation and information 1976 reduction can optionally be performed within this phase. 1977 1978 9.1.1 Phase 1: Calculation of Degree of Preference 1979 1980 The Phase 1 decision function shall be invoked whenever the local BGP 1981 speaker receives an UPDATE message from a peer located in a 1982 neighboring autonomous system that advertises a new route, a 1983 replacement route, or a withdrawn route. 1984 1985 The Phase 1 decision function is a separate process which completes 1986 when it has no further work to do. 1987 1988 The Phase 1 decision function shall lock an Adj-RIB-In prior to 1989 operating on any route contained within it, and shall unlock it after 1990 operating on all new or unfeasible routes contained within it. 1991 1992 For each newly received or replacement feasible route, the local BGP 1993 speaker shall determine a degree of preference. If the route is 1994 learned from a BGP speaker in the local autonomous system, either the 1995 value of the LOCAL_PREF attribute shall be taken as the degree of 1996 preference, or the local system shall compute the degree of 1997 preference of the route based on preconfigured policy information. If 1998 the route is learned from a BGP speaker in a neighboring autonomous 1999 system, then the degree of preference shall be computed based on 2000 preconfigured policy information. The exact nature of this policy 2001 information and the computation involved is a local matter. The 2002 local speaker shall then run the internal update process of 9.2.1 to 2003 select and advertise the most preferable route. 2004 2005 9.1.2 Phase 2: Route Selection 2006 2007 The Phase 2 decision function shall be invoked on completion of Phase 2008 1. The Phase 2 function is a separate process which completes when 2009 it has no further work to do. The Phase 2 process shall consider all 2010 routes that are present in the Adj-RIBs-In, including those received 2011 from BGP speakers located in its own autonomous system and those 2012 received from BGP speakers located in neighboring autonomous systems. 2013 2014 2015 2016 2017 2018 Rekhter & Li [Page 36] 2019 2020 RFC 1771 BGP-4 March 1995 2021 2022 2023 The Phase 2 decision function shall be blocked from running while the 2024 Phase 3 decision function is in process. The Phase 2 function shall 2025 lock all Adj-RIBs-In prior to commencing its function, and shall 2026 unlock them on completion. 2027 2028 If the NEXT_HOP attribute of a BGP route depicts an address to which 2029 the local BGP speaker doesn't have a route in its Loc-RIB, the BGP 2030 route SHOULD be excluded from the Phase 2 decision function. 2031 2032 For each set of destinations for which a feasible route exists in the 2033 Adj-RIBs-In, the local BGP speaker shall identify the route that has: 2034 2035 a) the highest degree of preference of any route to the same set 2036 of destinations, or 2037 2038 b) is the only route to that destination, or 2039 2040 c) is selected as a result of the Phase 2 tie breaking rules 2041 specified in 9.1.2.1. 2042 2043 The local speaker SHALL then install that route in the Loc-RIB, 2044 replacing any route to the same destination that is currently being 2045 held in the Loc-RIB. The local speaker MUST determine the immediate 2046 next hop to the address depicted by the NEXT_HOP attribute of the 2047 selected route by performing a lookup in the IGP and selecting one of 2048 the possible paths in the IGP. This immediate next hop MUST be used 2049 when installing the selected route in the Loc-RIB. If the route to 2050 the address depicted by the NEXT_HOP attribute changes such that the 2051 immediate next hop changes, route selection should be recalculated as 2052 specified above. 2053 2054 Unfeasible routes shall be removed from the Loc-RIB, and 2055 corresponding unfeasible routes shall then be removed from the Adj- 2056 RIBs-In. 2057 2058 9.1.2.1 Breaking Ties (Phase 2) 2059 2060 In its Adj-RIBs-In a BGP speaker may have several routes to the same 2061 destination that have the same degree of preference. The local 2062 speaker can select only one of these routes for inclusion in the 2063 associated Loc-RIB. The local speaker considers all equally 2064 preferable routes, both those received from BGP speakers located in 2065 neighboring autonomous systems, and those received from other BGP 2066 speakers located in the local speaker's autonomous system. 2067 2068 The following tie-breaking procedure assumes that for each candidate 2069 route all the BGP speakers within an autonomous system can ascertain 2070 the cost of a path (interior distance) to the address depicted by the 2071 2072 2073 2074 Rekhter & Li [Page 37] 2075 2076 RFC 1771 BGP-4 March 1995 2077 2078 2079 NEXT_HOP attribute of the route. Ties shall be broken according to 2080 the following algorithm: 2081 2082 a) If the local system is configured to take into account 2083 MULTI_EXIT_DISC, and the candidate routes differ in their 2084 MULTI_EXIT_DISC attribute, select the route that has the lowest 2085 value of the MULTI_EXIT_DISC attribute. 2086 2087 b) Otherwise, select the route that has the lowest cost (interior 2088 distance) to the entity depicted by the NEXT_HOP attribute of the 2089 route. If there are several routes with the same cost, then the 2090 tie-breaking shall be broken as follows: 2091 2092 - if at least one of the candidate routes was advertised by the 2093 BGP speaker in a neighboring autonomous system, select the 2094 route that was advertised by the BGP speaker in a neighboring 2095 autonomous system whose BGP Identifier has the lowest value 2096 among all other BGP speakers in neighboring autonomous systems; 2097 2098 - otherwise, select the route that was advertised by the BGP 2099 speaker whose BGP Identifier has the lowest value. 2100 2101 9.1.3 Phase 3: Route Dissemination 2102 2103 The Phase 3 decision function shall be invoked on completion of Phase 2104 2, or when any of the following events occur: 2105 2106 a) when routes in a Loc-RIB to local destinations have changed 2107 2108 b) when locally generated routes learned by means outside of BGP 2109 have changed 2110 2111 c) when a new BGP speaker - BGP speaker connection has been 2112 established 2113 2114 The Phase 3 function is a separate process which completes when it 2115 has no further work to do. The Phase 3 Routing Decision function 2116 shall be blocked from running while the Phase 2 decision function is 2117 in process. 2118 2119 All routes in the Loc-RIB shall be processed into a corresponding 2120 entry in the associated Adj-RIBs-Out. Route aggregation and 2121 information reduction techniques (see 9.2.4.1) may optionally be 2122 applied. 2123 2124 For the benefit of future support of inter-AS multicast capabilities, 2125 a BGP speaker that participates in inter-AS multicast routing shall 2126 advertise a route it receives from one of its external peers and if 2127 2128 2129 2130 Rekhter & Li [Page 38] 2131 2132 RFC 1771 BGP-4 March 1995 2133 2134 2135 it installs it in its Loc-RIB, it shall advertise it back to the peer 2136 from which the route was received. For a BGP speaker that does not 2137 participate in inter-AS multicast routing such an advertisement is 2138 optional. When doing such an advertisement, the NEXT_HOP attribute 2139 should be set to the address of the peer. An implementation may also 2140 optimize such an advertisement by truncating information in the 2141 AS_PATH attribute to include only its own AS number and that of the 2142 peer that advertised the route (such truncation requires the ORIGIN 2143 attribute to be set to INCOMPLETE). In addition an implementation is 2144 not required to pass optional or discretionary path attributes with 2145 such an advertisement. 2146 2147 When the updating of the Adj-RIBs-Out and the Forwarding Information 2148 Base (FIB) is complete, the local BGP speaker shall run the external 2149 update process of 9.2.2. 2150 2151 9.1.4 Overlapping Routes 2152 2153 A BGP speaker may transmit routes with overlapping Network Layer 2154 Reachability Information (NLRI) to another BGP speaker. NLRI overlap 2155 occurs when a set of destinations are identified in non-matching 2156 multiple routes. Since BGP encodes NLRI using IP prefixes, overlap 2157 will always exhibit subset relationships. A route describing a 2158 smaller set of destinations (a longer prefix) is said to be more 2159 specific than a route describing a larger set of destinations (a 2160 shorted prefix); similarly, a route describing a larger set of 2161 destinations (a shorter prefix) is said to be less specific than a 2162 route describing a smaller set of destinations (a longer prefix). 2163 2164 The precedence relationship effectively decomposes less specific 2165 routes into two parts: 2166 2167 - a set of destinations described only by the less specific 2168 route, and 2169 2170 - a set of destinations described by the overlap of the less 2171 specific and the more specific routes 2172 2173 When overlapping routes are present in the same Adj-RIB-In, the more 2174 specific route shall take precedence, in order from more specific to 2175 least specific. 2176 2177 The set of destinations described by the overlap represents a portion 2178 of the less specific route that is feasible, but is not currently in 2179 use. If a more specific route is later withdrawn, the set of 2180 destinations described by the overlap will still be reachable using 2181 the less specific route. 2182 2183 2184 2185 2186 Rekhter & Li [Page 39] 2187 2188 RFC 1771 BGP-4 March 1995 2189 2190 2191 If a BGP speaker receives overlapping routes, the Decision Process 2192 shall take into account the semantics of the overlapping routes. In 2193 particular, if a BGP speaker accepts the less specific route while 2194 rejecting the more specific route from the same peer, then the 2195 destinations represented by the overlap may not forward along the ASs 2196 listed in the AS_PATH attribute of that route. Therefore, a BGP 2197 speaker has the following choices: 2198 2199 a) Install both the less and the more specific routes 2200 2201 b) Install the more specific route only 2202 2203 c) Install the non-overlapping part of the less specific 2204 route only (that implies de-aggregation) 2205 2206 d) Aggregate the two routes and install the aggregated route 2207 2208 e) Install the less specific route only 2209 2210 f) Install neither route 2211 2212 If a BGP speaker chooses e), then it should add ATOMIC_AGGREGATE 2213 attribute to the route. A route that carries ATOMIC_AGGREGATE 2214 attribute can not be de-aggregated. That is, the NLRI of this route 2215 can not be made more specific. Forwarding along such a route does 2216 not guarantee that IP packets will actually traverse only ASs listed 2217 in the AS_PATH attribute of the route. If a BGP speaker chooses a), 2218 it must not advertise the more general route without the more 2219 specific route. 2220 2221 9.2 Update-Send Process 2222 2223 The Update-Send process is responsible for advertising UPDATE 2224 messages to all peers. For example, it distributes the routes chosen 2225 by the Decision Process to other BGP speakers which may be located in 2226 either the same autonomous system or a neighboring autonomous system. 2227 rules for information exchange between BGP speakers located in 2228 different autonomous systems are given in 9.2.2; rules for 2229 information exchange between BGP speakers located in the same 2230 autonomous system are given in 9.2.1. 2231 2232 Distribution of routing information between a set of BGP speakers, 2233 all of which are located in the same autonomous system, is referred 2234 to as internal distribution. 2235 2236 2237 2238 2239 2240 2241 2242 Rekhter & Li [Page 40] 2243 2244 RFC 1771 BGP-4 March 1995 2245 2246 2247 9.2.1 Internal Updates 2248 2249 The Internal update process is concerned with the distribution of 2250 routing information to BGP speakers located in the local speaker's 2251 autonomous system. 2252 2253 When a BGP speaker receives an UPDATE message from another BGP 2254 speaker located in its own autonomous system, the receiving BGP 2255 speaker shall not re-distribute the routing information contained in 2256 that UPDATE message to other BGP speakers located in its own 2257 autonomous system. 2258 2259 When a BGP speaker receives a new route from a BGP speaker in a 2260 neighboring autonomous system, it shall advertise that route to all 2261 other BGP speakers in its autonomous system by means of an UPDATE 2262 message if any of the following conditions occur: 2263 2264 1) the degree of preference assigned to the newly received route 2265 by the local BGP speaker is higher than the degree of preference 2266 that the local speaker has assigned to other routes that have been 2267 received from BGP speakers in neighboring autonomous systems, or 2268 2269 2) there are no other routes that have been received from BGP 2270 speakers in neighboring autonomous systems, or 2271 2272 3) the newly received route is selected as a result of breaking a 2273 tie between several routes which have the highest degree of 2274 preference, and the same destination (the tie-breaking procedure 2275 is specified in 9.2.1.1). 2276 2277 When a BGP speaker receives an UPDATE message with a non-empty 2278 WITHDRAWN ROUTES field, it shall remove from its Adj-RIB-In all 2279 routes whose destinations was carried in this field (as IP prefixes). 2280 The speaker shall take the following additional steps: 2281 2282 1) if the corresponding feasible route had not been previously 2283 advertised, then no further action is necessary 2284 2285 2) if the corresponding feasible route had been previously 2286 advertised, then: 2287 2288 i) if a new route is selected for advertisement that has the 2289 same Network Layer Reachability Information as the unfeasible 2290 routes, then the local BGP speaker shall advertise the 2291 replacement route 2292 2293 ii) if a replacement route is not available for advertisement, 2294 then the BGP speaker shall include the destinations of the 2295 2296 2297 2298 Rekhter & Li [Page 41] 2299 2300 RFC 1771 BGP-4 March 1995 2301 2302 2303 unfeasible route (in form of IP prefixes) in the WITHDRAWN 2304 ROUTES field of an UPDATE message, and shall send this message 2305 to each peer to whom it had previously advertised the 2306 corresponding feasible route. 2307 2308 All feasible routes which are advertised shall be placed in the 2309 appropriate Adj-RIBs-Out, and all unfeasible routes which are 2310 advertised shall be removed from the Adj-RIBs-Out. 2311 2312 9.2.1.1 Breaking Ties (Internal Updates) 2313 2314 If a local BGP speaker has connections to several BGP speakers in 2315 neighboring autonomous systems, there will be multiple Adj-RIBs-In 2316 associated with these peers. These Adj-RIBs-In might contain several 2317 equally preferable routes to the same destination, all of which were 2318 advertised by BGP speakers located in neighboring autonomous systems. 2319 The local BGP speaker shall select one of these routes according to 2320 the following rules: 2321 2322 a) If the candidate route differ only in their NEXT_HOP and 2323 MULTI_EXIT_DISC attributes, and the local system is configured to 2324 take into account MULTI_EXIT_DISC attribute, select the routes 2325 that has the lowest value of the MULTI_EXIT_DISC attribute. 2326 2327 b) If the local system can ascertain the cost of a path to the 2328 entity depicted by the NEXT_HOP attribute of the candidate route, 2329 select the route with the lowest cost. 2330 2331 c) In all other cases, select the route that was advertised by the 2332 BGP speaker whose BGP Identifier has the lowest value. 2333 2334 9.2.2 External Updates 2335 2336 The external update process is concerned with the distribution of 2337 routing information to BGP speakers located in neighboring autonomous 2338 systems. As part of Phase 3 route selection process, the BGP speaker 2339 has updated its Adj-RIBs-Out and its Forwarding Table. All newly 2340 installed routes and all newly unfeasible routes for which there is 2341 no replacement route shall be advertised to BGP speakers located in 2342 neighboring autonomous systems by means of UPDATE message. 2343 2344 Any routes in the Loc-RIB marked as unfeasible shall be removed. 2345 Changes to the reachable destinations within its own autonomous 2346 system shall also be advertised in an UPDATE message. 2347 2348 2349 2350 2351 2352 2353 2354 Rekhter & Li [Page 42] 2355 2356 RFC 1771 BGP-4 March 1995 2357 2358 2359 9.2.3 Controlling Routing Traffic Overhead 2360 2361 The BGP protocol constrains the amount of routing traffic (that is, 2362 UPDATE messages) in order to limit both the link bandwidth needed to 2363 advertise UPDATE messages and the processing power needed by the 2364 Decision Process to digest the information contained in the UPDATE 2365 messages. 2366 2367 9.2.3.1 Frequency of Route Advertisement 2368 2369 The parameter MinRouteAdvertisementInterval determines the minimum 2370 amount of time that must elapse between advertisement of routes to a 2371 particular destination from a single BGP speaker. This rate limiting 2372 procedure applies on a per-destination basis, although the value of 2373 MinRouteAdvertisementInterval is set on a per BGP peer basis. 2374 2375 Two UPDATE messages sent from a single BGP speaker that advertise 2376 feasible routes to some common set of destinations received from BGP 2377 speakers in neighboring autonomous systems must be separated by at 2378 least MinRouteAdvertisementInterval. Clearly, this can only be 2379 achieved precisely by keeping a separate timer for each common set of 2380 destinations. This would be unwarranted overhead. Any technique which 2381 ensures that the interval between two UPDATE messages sent from a 2382 single BGP speaker that advertise feasible routes to some common set 2383 of destinations received from BGP speakers in neighboring autonomous 2384 systems will be at least MinRouteAdvertisementInterval, and will also 2385 ensure a constant upper bound on the interval is acceptable. 2386 2387 Since fast convergence is needed within an autonomous system, this 2388 procedure does not apply for routes receives from other BGP speakers 2389 in the same autonomous system. To avoid long-lived black holes, the 2390 procedure does not apply to the explicit withdrawal of unfeasible 2391 routes (that is, routes whose destinations (expressed as IP prefixes) 2392 are listed in the WITHDRAWN ROUTES field of an UPDATE message). 2393 2394 This procedure does not limit the rate of route selection, but only 2395 the rate of route advertisement. If new routes are selected multiple 2396 times while awaiting the expiration of MinRouteAdvertisementInterval, 2397 the last route selected shall be advertised at the end of 2398 MinRouteAdvertisementInterval. 2399 2400 9.2.3.2 Frequency of Route Origination 2401 2402 The parameter MinASOriginationInterval determines the minimum amount 2403 of time that must elapse between successive advertisements of UPDATE 2404 messages that report changes within the advertising BGP speaker's own 2405 autonomous systems. 2406 2407 2408 2409 2410 Rekhter & Li [Page 43] 2411 2412 RFC 1771 BGP-4 March 1995 2413 2414 2415 9.2.3.3 Jitter 2416 2417 To minimize the likelihood that the distribution of BGP messages by a 2418 given BGP speaker will contain peaks, jitter should be applied to the 2419 timers associated with MinASOriginationInterval, Keepalive, and 2420 MinRouteAdvertisementInterval. A given BGP speaker shall apply the 2421 same jitter to each of these quantities regardless of the 2422 destinations to which the updates are being sent; that is, jitter 2423 will not be applied on a "per peer" basis. 2424 2425 The amount of jitter to be introduced shall be determined by 2426 multiplying the base value of the appropriate timer by a random 2427 factor which is uniformly distributed in the range from 0.75 to 1.0. 2428 2429 9.2.4 Efficient Organization of Routing Information 2430 2431 Having selected the routing information which it will advertise, a 2432 BGP speaker may avail itself of several methods to organize this 2433 information in an efficient manner. 2434 2435 9.2.4.1 Information Reduction 2436 2437 Information reduction may imply a reduction in granularity of policy 2438 control - after information is collapsed, the same policies will 2439 apply to all destinations and paths in the equivalence class. 2440 2441 The Decision Process may optionally reduce the amount of information 2442 that it will place in the Adj-RIBs-Out by any of the following 2443 methods: 2444 2445 a) Network Layer Reachability Information (NLRI): 2446 2447 Destination IP addresses can be represented as IP address 2448 prefixes. In cases where there is a correspondence between the 2449 address structure and the systems under control of an autonomous 2450 system administrator, it will be possible to reduce the size of 2451 the NLRI carried in the UPDATE messages. 2452 2453 b) AS_PATHs: 2454 2455 AS path information can be represented as ordered AS_SEQUENCEs or 2456 unordered AS_SETs. AS_SETs are used in the route aggregation 2457 algorithm described in 9.2.4.2. They reduce the size of the 2458 AS_PATH information by listing each AS number only once, 2459 regardless of how many times it may have appeared in multiple 2460 AS_PATHs that were aggregated. 2461 2462 2463 2464 2465 2466 Rekhter & Li [Page 44] 2467 2468 RFC 1771 BGP-4 March 1995 2469 2470 2471 An AS_SET implies that the destinations listed in the NLRI can be 2472 reached through paths that traverse at least some of the 2473 constituent autonomous systems. AS_SETs provide sufficient 2474 information to avoid routing information looping; however their 2475 use may prune potentially feasible paths, since such paths are no 2476 longer listed individually as in the form of AS_SEQUENCEs. In 2477 practice this is not likely to be a problem, since once an IP 2478 packet arrives at the edge of a group of autonomous systems, the 2479 BGP speaker at that point is likely to have more detailed path 2480 information and can distinguish individual paths to destinations. 2481 2482 9.2.4.2 Aggregating Routing Information 2483 2484 Aggregation is the process of combining the characteristics of 2485 several different routes in such a way that a single route can be 2486 advertised. Aggregation can occur as part of the decision process 2487 to reduce the amount of routing information that will be placed in 2488 the Adj-RIBs-Out. 2489 2490 Aggregation reduces the amount of information that a BGP speaker must 2491 store and exchange with other BGP speakers. Routes can be aggregated 2492 by applying the following procedure separately to path attributes of 2493 like type and to the Network Layer Reachability Information. 2494 2495 Routes that have the following attributes shall not be aggregated 2496 unless the corresponding attributes of each route are identical: 2497 MULTI_EXIT_DISC, NEXT_HOP. 2498 2499 Path attributes that have different type codes can not be aggregated 2500 together. Path of the same type code may be aggregated, according to 2501 the following rules: 2502 2503 ORIGIN attribute: If at least one route among routes that are 2504 aggregated has ORIGIN with the value INCOMPLETE, then the 2505 aggregated route must have the ORIGIN attribute with the value 2506 INCOMPLETE. Otherwise, if at least one route among routes that are 2507 aggregated has ORIGIN with the value EGP, then the aggregated 2508 route must have the origin attribute with the value EGP. In all 2509 other case the value of the ORIGIN attribute of the aggregated 2510 route is INTERNAL. 2511 2512 AS_PATH attribute: If routes to be aggregated have identical 2513 AS_PATH attributes, then the aggregated route has the same AS_PATH 2514 attribute as each individual route. 2515 2516 For the purpose of aggregating AS_PATH attributes we model each AS 2517 within the AS_PATH attribute as a tuple <type, value>, where 2518 "type" identifies a type of the path segment the AS belongs to 2519 2520 2521 2522 Rekhter & Li [Page 45] 2523 2524 RFC 1771 BGP-4 March 1995 2525 2526 2527 (e.g. AS_SEQUENCE, AS_SET), and "value" is the AS number. If the 2528 routes to be aggregated have different AS_PATH attributes, then 2529 the aggregated AS_PATH attribute shall satisfy all of the 2530 following conditions: 2531 2532 - all tuples of the type AS_SEQUENCE in the aggregated AS_PATH 2533 shall appear in all of the AS_PATH in the initial set of routes 2534 to be aggregated. 2535 2536 - all tuples of the type AS_SET in the aggregated AS_PATH shall 2537 appear in at least one of the AS_PATH in the initial set (they 2538 may appear as either AS_SET or AS_SEQUENCE types). 2539 2540 - for any tuple X of the type AS_SEQUENCE in the aggregated 2541 AS_PATH which precedes tuple Y in the aggregated AS_PATH, X 2542 precedes Y in each AS_PATH in the initial set which contains Y, 2543 regardless of the type of Y. 2544 2545 - No tuple with the same value shall appear more than once in 2546 the aggregated AS_PATH, regardless of the tuple's type. 2547 2548 An implementation may choose any algorithm which conforms to these 2549 rules. At a minimum a conformant implementation shall be able to 2550 perform the following algorithm that meets all of the above 2551 conditions: 2552 2553 - determine the longest leading sequence of tuples (as defined 2554 above) common to all the AS_PATH attributes of the routes to be 2555 aggregated. Make this sequence the leading sequence of the 2556 aggregated AS_PATH attribute. 2557 2558 - set the type of the rest of the tuples from the AS_PATH 2559 attributes of the routes to be aggregated to AS_SET, and append 2560 them to the aggregated AS_PATH attribute. 2561 2562 - if the aggregated AS_PATH has more than one tuple with the 2563 same value (regardless of tuple's type), eliminate all, but one 2564 such tuple by deleting tuples of the type AS_SET from the 2565 aggregated AS_PATH attribute. 2566 2567 Appendix 6, section 6.8 presents another algorithm that satisfies 2568 the conditions and allows for more complex policy configurations. 2569 2570 ATOMIC_AGGREGATE: If at least one of the routes to be aggregated 2571 has ATOMIC_AGGREGATE path attribute, then the aggregated route 2572 shall have this attribute as well. 2573 2574 2575 2576 2577 2578 Rekhter & Li [Page 46] 2579 2580 RFC 1771 BGP-4 March 1995 2581 2582 2583 AGGREGATOR: All AGGREGATOR attributes of all routes to be 2584 aggregated should be ignored. 2585 2586 9.3 Route Selection Criteria 2587 2588 Generally speaking, additional rules for comparing routes among 2589 several alternatives are outside the scope of this document. There 2590 are two exceptions: 2591 2592 - If the local AS appears in the AS path of the new route being 2593 considered, then that new route cannot be viewed as better than 2594 any other route. If such a route were ever used, a routing loop 2595 would result. 2596 2597 - In order to achieve successful distributed operation, only 2598 routes with a likelihood of stability can be chosen. Thus, an AS 2599 must avoid using unstable routes, and it must not make rapid 2600 spontaneous changes to its choice of route. Quantifying the terms 2601 "unstable" and "rapid" in the previous sentence will require 2602 experience, but the principle is clear. 2603 2604 9.4 Originating BGP routes 2605 2606 A BGP speaker may originate BGP routes by injecting routing 2607 information acquired by some other means (e.g. via an IGP) into BGP. 2608 A BGP speaker that originates BGP routes shall assign the degree of 2609 preference to these routes by passing them through the Decision 2610 Process (see Section 9.1). These routes may also be distributed to 2611 other BGP speakers within the local AS as part of the Internal update 2612 process (see Section 9.2.1). The decision whether to distribute non- 2613 BGP acquired routes within an AS via BGP or not depends on the 2614 environment within the AS (e.g. type of IGP) and should be controlled 2615 via configuration. 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 Rekhter & Li [Page 47] 2635 2636 RFC 1771 BGP-4 March 1995 2637 2638 2639 Appendix 1. BGP FSM State Transitions and Actions. 2640 2641 This Appendix discusses the transitions between states in the BGP FSM 2642 in response to BGP events. The following is the list of these states 2643 and events when the negotiated Hold Time value is non-zero. 2644 2645 BGP States: 2646 2647 1 - Idle 2648 2 - Connect 2649 3 - Active 2650 4 - OpenSent 2651 5 - OpenConfirm 2652 6 - Established 2653 2654 BGP Events: 2655 2656 1 - BGP Start 2657 2 - BGP Stop 2658 3 - BGP Transport connection open 2659 4 - BGP Transport connection closed 2660 5 - BGP Transport connection open failed 2661 6 - BGP Transport fatal error 2662 7 - ConnectRetry timer expired 2663 8 - Hold Timer expired 2664 9 - KeepAlive timer expired 2665 10 - Receive OPEN message 2666 11 - Receive KEEPALIVE message 2667 12 - Receive UPDATE messages 2668 13 - Receive NOTIFICATION message 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 Rekhter & Li [Page 48] 2691 2692 RFC 1771 BGP-4 March 1995 2693 2694 2695 The following table describes the state transitions of the BGP FSM 2696 and the actions triggered by these transitions. 2697 2698 2699 Event Actions Message Sent Next State 2700 -------------------------------------------------------------------- 2701 Idle (1) 2702 1 Initialize resources none 2 2703 Start ConnectRetry timer 2704 Initiate a transport connection 2705 others none none 1 2706 2707 Connect(2) 2708 1 none none 2 2709 3 Complete initialization OPEN 4 2710 Clear ConnectRetry timer 2711 5 Restart ConnectRetry timer none 3 2712 7 Restart ConnectRetry timer none 2 2713 Initiate a transport connection 2714 others Release resources none 1 2715 2716 Active (3) 2717 1 none none 3 2718 3 Complete initialization OPEN 4 2719 Clear ConnectRetry timer 2720 5 Close connection 3 2721 Restart ConnectRetry timer 2722 7 Restart ConnectRetry timer none 2 2723 Initiate a transport connection 2724 others Release resources none 1 2725 2726 OpenSent(4) 2727 1 none none 4 2728 4 Close transport connection none 3 2729 Restart ConnectRetry timer 2730 6 Release resources none 1 2731 10 Process OPEN is OK KEEPALIVE 5 2732 Process OPEN failed NOTIFICATION 1 2733 others Close transport connection NOTIFICATION 1 2734 Release resources 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 Rekhter & Li [Page 49] 2747 2748 RFC 1771 BGP-4 March 1995 2749 2750 2751 OpenConfirm (5) 2752 1 none none 5 2753 4 Release resources none 1 2754 6 Release resources none 1 2755 9 Restart KeepAlive timer KEEPALIVE 5 2756 11 Complete initialization none 6 2757 Restart Hold Timer 2758 13 Close transport connection 1 2759 Release resources 2760 others Close transport connection NOTIFICATION 1 2761 Release resources 2762 2763 Established (6) 2764 1 none none 6 2765 4 Release resources none 1 2766 6 Release resources none 1 2767 9 Restart KeepAlive timer KEEPALIVE 6 2768 11 Restart Hold Timer KEEPALIVE 6 2769 12 Process UPDATE is OK UPDATE 6 2770 Process UPDATE failed NOTIFICATION 1 2771 13 Close transport connection 1 2772 Release resources 2773 others Close transport connection NOTIFICATION 1 2774 Release resources 2775 --------------------------------------------------------------------- 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 Rekhter & Li [Page 50] 2803 2804 RFC 1771 BGP-4 March 1995 2805 2806 2807 The following is a condensed version of the above state transition 2808 table. 2809 2810 2811 Events| Idle | Connect | Active | OpenSent | OpenConfirm | Estab 2812 | (1) | (2) | (3) | (4) | (5) | (6) 2813 |-------------------------------------------------------------- 2814 1 | 2 | 2 | 3 | 4 | 5 | 6 2815 | | | | | | 2816 2 | 1 | 1 | 1 | 1 | 1 | 1 2817 | | | | | | 2818 3 | 1 | 4 | 4 | 1 | 1 | 1 2819 | | | | | | 2820 4 | 1 | 1 | 1 | 3 | 1 | 1 2821 | | | | | | 2822 5 | 1 | 3 | 3 | 1 | 1 | 1 2823 | | | | | | 2824 6 | 1 | 1 | 1 | 1 | 1 | 1 2825 | | | | | | 2826 7 | 1 | 2 | 2 | 1 | 1 | 1 2827 | | | | | | 2828 8 | 1 | 1 | 1 | 1 | 1 | 1 2829 | | | | | | 2830 9 | 1 | 1 | 1 | 1 | 5 | 6 2831 | | | | | | 2832 10 | 1 | 1 | 1 | 1 or 5 | 1 | 1 2833 | | | | | | 2834 11 | 1 | 1 | 1 | 1 | 6 | 6 2835 | | | | | | 2836 12 | 1 | 1 | 1 | 1 | 1 | 1 or 6 2837 | | | | | | 2838 13 | 1 | 1 | 1 | 1 | 1 | 1 2839 | | | | | | 2840 --------------------------------------------------------------- 2841 2842 2843 Appendix 2. Comparison with RFC1267 2844 2845 BGP-4 is capable of operating in an environment where a set of 2846 reachable destinations may be expressed via a single IP prefix. The 2847 concept of network classes, or subnetting is foreign to BGP-4. To 2848 accommodate these capabilities BGP-4 changes semantics and encoding 2849 associated with the AS_PATH attribute. New text has been added to 2850 define semantics associated with IP prefixes. These abilities allow 2851 BGP-4 to support the proposed supernetting scheme [9]. 2852 2853 To simplify configuration this version introduces a new attribute, 2854 LOCAL_PREF, that facilitates route selection procedures. 2855 2856 2857 2858 Rekhter & Li [Page 51] 2859 2860 RFC 1771 BGP-4 March 1995 2861 2862 2863 The INTER_AS_METRIC attribute has been renamed to be MULTI_EXIT_DISC. 2864 A new attribute, ATOMIC_AGGREGATE, has been introduced to insure that 2865 certain aggregates are not de-aggregated. Another new attribute, 2866 AGGREGATOR, can be added to aggregate routes in order to advertise 2867 which AS and which BGP speaker within that AS caused the aggregation. 2868 2869 To insure that Hold Timers are symmetric, the Hold Time is now 2870 negotiated on a per-connection basis. Hold Times of zero are now 2871 supported. 2872 2873 Appendix 3. Comparison with RFC 1163 2874 2875 All of the changes listed in Appendix 2, plus the following. 2876 2877 To detect and recover from BGP connection collision, a new field (BGP 2878 Identifier) has been added to the OPEN message. New text (Section 2879 6.8) has been added to specify the procedure for detecting and 2880 recovering from collision. 2881 2882 The new document no longer restricts the border router that is passed 2883 in the NEXT_HOP path attribute to be part of the same Autonomous 2884 System as the BGP Speaker. 2885 2886 New document optimizes and simplifies the exchange of the information 2887 about previously reachable routes. 2888 2889 Appendix 4. Comparison with RFC 1105 2890 2891 All of the changes listed in Appendices 2 and 3, plus the following. 2892 2893 Minor changes to the RFC1105 Finite State Machine were necessary to 2894 accommodate the TCP user interface provided by 4.3 BSD. 2895 2896 The notion of Up/Down/Horizontal relations present in RFC1105 has 2897 been removed from the protocol. 2898 2899 The changes in the message format from RFC1105 are as follows: 2900 2901 1. The Hold Time field has been removed from the BGP header and 2902 added to the OPEN message. 2903 2904 2. The version field has been removed from the BGP header and 2905 added to the OPEN message. 2906 2907 3. The Link Type field has been removed from the OPEN message. 2908 2909 4. The OPEN CONFIRM message has been eliminated and replaced with 2910 implicit confirmation provided by the KEEPALIVE message. 2911 2912 2913 2914 Rekhter & Li [Page 52] 2915 2916 RFC 1771 BGP-4 March 1995 2917 2918 2919 5. The format of the UPDATE message has been changed 2920 significantly. New fields were added to the UPDATE message to 2921 support multiple path attributes. 2922 2923 6. The Marker field has been expanded and its role broadened to 2924 support authentication. 2925 2926 Note that quite often BGP, as specified in RFC 1105, is referred 2927 to as BGP-1, BGP, as specified in RFC 1163, is referred to as 2928 BGP-2, BGP, as specified in RFC1267 is referred to as BGP-3, and 2929 BGP, as specified in this document is referred to as BGP-4. 2930 2931 Appendix 5. TCP options that may be used with BGP 2932 2933 If a local system TCP user interface supports TCP PUSH function, then 2934 each BGP message should be transmitted with PUSH flag set. Setting 2935 PUSH flag forces BGP messages to be transmitted promptly to the 2936 receiver. 2937 2938 If a local system TCP user interface supports setting precedence for 2939 TCP connection, then the BGP transport connection should be opened 2940 with precedence set to Internetwork Control (110) value (see also 2941 [6]). 2942 2943 Appendix 6. Implementation Recommendations 2944 2945 This section presents some implementation recommendations. 2946 2947 6.1 Multiple Networks Per Message 2948 2949 The BGP protocol allows for multiple address prefixes with the same 2950 AS path and next-hop gateway to be specified in one message. Making 2951 use of this capability is highly recommended. With one address prefix 2952 per message there is a substantial increase in overhead in the 2953 receiver. Not only does the system overhead increase due to the 2954 reception of multiple messages, but the overhead of scanning the 2955 routing table for updates to BGP peers and other routing protocols 2956 (and sending the associated messages) is incurred multiple times as 2957 well. One method of building messages containing many address 2958 prefixes per AS path and gateway from a routing table that is not 2959 organized per AS path is to build many messages as the routing table 2960 is scanned. As each address prefix is processed, a message for the 2961 associated AS path and gateway is allocated, if it does not exist, 2962 and the new address prefix is added to it. If such a message exists, 2963 the new address prefix is just appended to it. If the message lacks 2964 the space to hold the new address prefix, it is transmitted, a new 2965 message is allocated, and the new address prefix is inserted into the 2966 new message. When the entire routing table has been scanned, all 2967 2968 2969 2970 Rekhter & Li [Page 53] 2971 2972 RFC 1771 BGP-4 March 1995 2973 2974 2975 allocated messages are sent and their resources released. Maximum 2976 compression is achieved when all the destinations covered by the 2977 address prefixes share a gateway and common path attributes, making 2978 it possible to send many address prefixes in one 4096-byte message. 2979 2980 When peering with a BGP implementation that does not compress 2981 multiple address prefixes into one message, it may be necessary to 2982 take steps to reduce the overhead from the flood of data received 2983 when a peer is acquired or a significant network topology change 2984 occurs. One method of doing this is to limit the rate of updates. 2985 This will eliminate the redundant scanning of the routing table to 2986 provide flash updates for BGP peers and other routing protocols. A 2987 disadvantage of this approach is that it increases the propagation 2988 latency of routing information. By choosing a minimum flash update 2989 interval that is not much greater than the time it takes to process 2990 the multiple messages this latency should be minimized. A better 2991 method would be to read all received messages before sending updates. 2992 2993 6.2 Processing Messages on a Stream Protocol 2994 2995 BGP uses TCP as a transport mechanism. Due to the stream nature of 2996 TCP, all the data for received messages does not necessarily arrive 2997 at the same time. This can make it difficult to process the data as 2998 messages, especially on systems such as BSD Unix where it is not 2999 possible to determine how much data has been received but not yet 3000 processed. 3001 3002 One method that can be used in this situation is to first try to read 3003 just the message header. For the KEEPALIVE message type, this is a 3004 complete message; for other message types, the header should first be 3005 verified, in particular the total length. If all checks are 3006 successful, the specified length, minus the size of the message 3007 header is the amount of data left to read. An implementation that 3008 would "hang" the routing information process while trying to read 3009 from a peer could set up a message buffer (4096 bytes) per peer and 3010 fill it with data as available until a complete message has been 3011 received. 3012 3013 6.3 Reducing route flapping 3014 3015 To avoid excessive route flapping a BGP speaker which needs to 3016 withdraw a destination and send an update about a more specific or 3017 less specific route shall combine them into the same UPDATE message. 3018 3019 3020 3021 3022 3023 3024 3025 3026 Rekhter & Li [Page 54] 3027 3028 RFC 1771 BGP-4 March 1995 3029 3030 3031 6.4 BGP Timers 3032 3033 BGP employs five timers: ConnectRetry, Hold Time, KeepAlive, 3034 MinASOriginationInterval, and MinRouteAdvertisementInterval The 3035 suggested value for the ConnectRetry timer is 120 seconds. The 3036 suggested value for the Hold Time is 90 seconds. The suggested value 3037 for the KeepAlive timer is 30 seconds. The suggested value for the 3038 MinASOriginationInterval is 15 seconds. The suggested value for the 3039 MinRouteAdvertisementInterval is 30 seconds. 3040 3041 An implementation of BGP MUST allow these timers to be configurable. 3042 3043 6.5 Path attribute ordering 3044 3045 Implementations which combine update messages as described above in 3046 6.1 may prefer to see all path attributes presented in a known order. 3047 This permits them to quickly identify sets of attributes from 3048 different update messages which are semantically identical. To 3049 facilitate this, it is a useful optimization to order the path 3050 attributes according to type code. This optimization is entirely 3051 optional. 3052 3053 6.6 AS_SET sorting 3054 3055 Another useful optimization that can be done to simplify this 3056 situation is to sort the AS numbers found in an AS_SET. This 3057 optimization is entirely optional. 3058 3059 6.7 Control over version negotiation 3060 3061 Since BGP-4 is capable of carrying aggregated routes which cannot be 3062 properly represented in BGP-3, an implementation which supports BGP-4 3063 and another BGP version should provide the capability to only speak 3064 BGP-4 on a per-peer basis. 3065 3066 6.8 Complex AS_PATH aggregation 3067 3068 An implementation which chooses to provide a path aggregation 3069 algorithm which retains significant amounts of path information may 3070 wish to use the following procedure: 3071 3072 For the purpose of aggregating AS_PATH attributes of two routes, 3073 we model each AS as a tuple <type, value>, where "type" identifies 3074 a type of the path segment the AS belongs to (e.g. AS_SEQUENCE, 3075 AS_SET), and "value" is the AS number. Two ASs are said to be the 3076 same if their corresponding <type, value> tuples are the same. 3077 3078 3079 3080 3081 3082 Rekhter & Li [Page 55] 3083 3084 RFC 1771 BGP-4 March 1995 3085 3086 3087 The algorithm to aggregate two AS_PATH attributes works as 3088 follows: 3089 3090 a) Identify the same ASs (as defined above) within each AS_PATH 3091 attribute that are in the same relative order within both 3092 AS_PATH attributes. Two ASs, X and Y, are said to be in the 3093 same order if either: 3094 3095 - X precedes Y in both AS_PATH attributes, or - Y precedes X 3096 in both AS_PATH attributes. 3097 3098 b) The aggregated AS_PATH attribute consists of ASs identified 3099 in (a) in exactly the same order as they appear in the AS_PATH 3100 attributes to be aggregated. If two consecutive ASs identified 3101 in (a) do not immediately follow each other in both of the 3102 AS_PATH attributes to be aggregated, then the intervening ASs 3103 (ASs that are between the two consecutive ASs that are the 3104 same) in both attributes are combined into an AS_SET path 3105 segment that consists of the intervening ASs from both AS_PATH 3106 attributes; this segment is then placed in between the two 3107 consecutive ASs identified in (a) of the aggregated attribute. 3108 If two consecutive ASs identified in (a) immediately follow 3109 each other in one attribute, but do not follow in another, then 3110 the intervening ASs of the latter are combined into an AS_SET 3111 path segment; this segment is then placed in between the two 3112 consecutive ASs identified in (a) of the aggregated attribute. 3113 3114 If as a result of the above procedure a given AS number appears 3115 more than once within the aggregated AS_PATH attribute, all, but 3116 the last instance (rightmost occurrence) of that AS number should 3117 be removed from the aggregated AS_PATH attribute. 3118 3119 References 3120 3121 [1] Mills, D., "Exterior Gateway Protocol Formal Specification", RFC 3122 904, BBN, April 1984. 3123 3124 [2] Rekhter, Y., "EGP and Policy Based Routing in the New NSFNET 3125 Backbone", RFC 1092, T.J. Watson Research Center, February 1989. 3126 3127 [3] Braun, H-W., "The NSFNET Routing Architecture", RFC 1093, 3128 MERIT/NSFNET Project, February 1989. 3129 3130 [4] Postel, J., "Transmission Control Protocol - DARPA Internet 3131 Program Protocol Specification", STD 7, RFC 793, DARPA, September 3132 1981. 3133 3134 3135 3136 3137 3138 Rekhter & Li [Page 56] 3139 3140 RFC 1771 BGP-4 March 1995 3141 3142 3143 [5] Rekhter, Y., and P. Gross, "Application of the Border Gateway 3144 Protocol in the Internet", RFC 1772, T.J. Watson Research Center, 3145 IBM Corp., MCI, March 1995. 3146 3147 [6] Postel, J., "Internet Protocol - DARPA Internet Program Protocol 3148 Specification", STD 5, RFC 791, DARPA, September 1981. 3149 3150 [7] "Information Processing Systems - Telecommunications and 3151 Information Exchange between Systems - Protocol for Exchange of 3152 Inter-domain Routeing Information among Intermediate Systems to 3153 Support Forwarding of ISO 8473 PDUs", ISO/IEC IS10747, 1993 3154 3155 [8] Fuller, V., Li, T., Yu, J., and K. Varadhan, "Classless Inter- 3156 Domain Routing (CIDR): an Address Assignment and Aggregation 3157 Strategy", RFC 1519, BARRNet, cisco, MERIT, OARnet, September 3158 1993 3159 3160 [9] Rekhter, Y., Li, T., "An Architecture for IP Address Allocation 3161 with CIDR", RFC 1518, T.J. Watson Research Center, cisco, 3162 September 1993 3163 3164 Security Considerations 3165 3166 Security issues are not discussed in this document. 3167 3168 Editors' Addresses 3169 3170 Yakov Rekhter 3171 T.J. Watson Research Center IBM Corporation 3172 P.O. Box 704, Office H3-D40 3173 Yorktown Heights, NY 10598 3174 3175 Phone: +1 914 784 7361 3176 EMail: yakov@watson.ibm.com 3177 3178 3179 Tony Li 3180 cisco Systems, Inc. 3181 170 W. Tasman Dr. 3182 San Jose, CA 95134 3183 3184 EMail: tli@cisco.com 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 Rekhter & Li [Page 57] 3195