idnits 2.17.1 draft-ietf-mptcp-rfc6824bis-13.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- -- The draft header indicates that this document obsoletes RFC6824, but the abstract doesn't seem to directly say this. It does mention RFC6824 though, so this could be OK. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == The document seems to lack the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords -- however, there's a paragraph with a matching beginning. Boilerplate error? (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). -- The document date (February 17, 2019) is 1867 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: 'Data ACK' is mentioned on line 543, but not defined == Missing Reference: 'Checksum' is mentioned on line 544, but not defined == Missing Reference: 'DFIN' is mentioned on line 3567, but not defined ** Obsolete normative reference: RFC 793 (Obsoleted by RFC 9293) ** Downref: Normative reference to an Informational RFC: RFC 6182 -- Possible downref: Non-RFC (?) normative reference: ref. 'SHS' -- Obsolete informational reference (is this intentional?): RFC 6528 (Obsoleted by RFC 9293) -- Obsolete informational reference (is this intentional?): RFC 6824 (Obsoleted by RFC 8684) Summary: 2 errors (**), 0 flaws (~~), 5 warnings (==), 5 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Internet Engineering Task Force A. Ford 3 Internet-Draft Pexip 4 Obsoletes: 6824 (if approved) C. Raiciu 5 Intended status: Standards Track U. Politechnica of Bucharest 6 Expires: August 21, 2019 M. Handley 7 U. College London 8 O. Bonaventure 9 U. catholique de Louvain 10 C. Paasch 11 Apple, Inc. 12 February 17, 2019 14 TCP Extensions for Multipath Operation with Multiple Addresses 15 draft-ietf-mptcp-rfc6824bis-13 17 Abstract 19 TCP/IP communication is currently restricted to a single path per 20 connection, yet multiple paths often exist between peers. The 21 simultaneous use of these multiple paths for a TCP/IP session would 22 improve resource usage within the network and, thus, improve user 23 experience through higher throughput and improved resilience to 24 network failure. 26 Multipath TCP provides the ability to simultaneously use multiple 27 paths between peers. This document presents a set of extensions to 28 traditional TCP to support multipath operation. The protocol offers 29 the same type of service to applications as TCP (i.e., reliable 30 bytestream), and it provides the components necessary to establish 31 and use multiple TCP flows across potentially disjoint paths. 33 This document specifies v1 of Multipath TCP, obsoleting v0 as 34 specified in RFC6824, through clarifications and modifications 35 primarily driven by deployment experience. 37 Status of This Memo 39 This Internet-Draft is submitted in full conformance with the 40 provisions of BCP 78 and BCP 79. 42 Internet-Drafts are working documents of the Internet Engineering 43 Task Force (IETF). Note that other groups may also distribute 44 working documents as Internet-Drafts. The list of current Internet- 45 Drafts is at http://datatracker.ietf.org/drafts/current/. 47 Internet-Drafts are draft documents valid for a maximum of six months 48 and may be updated, replaced, or obsoleted by other documents at any 49 time. It is inappropriate to use Internet-Drafts as reference 50 material or to cite them other than as "work in progress." 52 This Internet-Draft will expire on August 21, 2019. 54 Copyright Notice 56 Copyright (c) 2019 IETF Trust and the persons identified as the 57 document authors. All rights reserved. 59 This document is subject to BCP 78 and the IETF Trust's Legal 60 Provisions Relating to IETF Documents 61 (http://trustee.ietf.org/license-info) in effect on the date of 62 publication of this document. Please review these documents 63 carefully, as they describe your rights and restrictions with respect 64 to this document. Code Components extracted from this document must 65 include Simplified BSD License text as described in Section 4.e of 66 the Trust Legal Provisions and are provided without warranty as 67 described in the Simplified BSD License. 69 Table of Contents 71 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 72 1.1. Design Assumptions . . . . . . . . . . . . . . . . . . . 4 73 1.2. Multipath TCP in the Networking Stack . . . . . . . . . . 5 74 1.3. Terminology . . . . . . . . . . . . . . . . . . . . . . . 6 75 1.4. MPTCP Concept . . . . . . . . . . . . . . . . . . . . . . 7 76 1.5. Requirements Language . . . . . . . . . . . . . . . . . . 8 77 2. Operation Overview . . . . . . . . . . . . . . . . . . . . . 8 78 2.1. Initiating an MPTCP Connection . . . . . . . . . . . . . 9 79 2.2. Associating a New Subflow with an Existing MPTCP 80 Connection . . . . . . . . . . . . . . . . . . . . . . . 10 81 2.3. Informing the Other Host about Another Potential Address 11 82 2.4. Data Transfer Using MPTCP . . . . . . . . . . . . . . . . 12 83 2.5. Requesting a Change in a Path's Priority . . . . . . . . 13 84 2.6. Closing an MPTCP Connection . . . . . . . . . . . . . . . 13 85 2.7. Notable Features . . . . . . . . . . . . . . . . . . . . 14 86 3. MPTCP Protocol . . . . . . . . . . . . . . . . . . . . . . . 15 87 3.1. Connection Initiation . . . . . . . . . . . . . . . . . . 16 88 3.2. Starting a New Subflow . . . . . . . . . . . . . . . . . 23 89 3.3. General MPTCP Operation . . . . . . . . . . . . . . . . . 28 90 3.3.1. Data Sequence Mapping . . . . . . . . . . . . . . . . 30 91 3.3.2. Data Acknowledgments . . . . . . . . . . . . . . . . 33 92 3.3.3. Closing a Connection . . . . . . . . . . . . . . . . 34 93 3.3.4. Receiver Considerations . . . . . . . . . . . . . . . 35 94 3.3.5. Sender Considerations . . . . . . . . . . . . . . . . 37 95 3.3.6. Reliability and Retransmissions . . . . . . . . . . . 37 96 3.3.7. Congestion Control Considerations . . . . . . . . . . 39 97 3.3.8. Subflow Policy . . . . . . . . . . . . . . . . . . . 39 98 3.4. Address Knowledge Exchange (Path Management) . . . . . . 41 99 3.4.1. Address Advertisement . . . . . . . . . . . . . . . . 42 100 3.4.2. Remove Address . . . . . . . . . . . . . . . . . . . 45 101 3.5. Fast Close . . . . . . . . . . . . . . . . . . . . . . . 46 102 3.6. Subflow Reset . . . . . . . . . . . . . . . . . . . . . . 48 103 3.7. Fallback . . . . . . . . . . . . . . . . . . . . . . . . 49 104 3.8. Error Handling . . . . . . . . . . . . . . . . . . . . . 53 105 3.9. Heuristics . . . . . . . . . . . . . . . . . . . . . . . 53 106 3.9.1. Port Usage . . . . . . . . . . . . . . . . . . . . . 54 107 3.9.2. Delayed Subflow Start and Subflow Symmetry . . . . . 54 108 3.9.3. Failure Handling . . . . . . . . . . . . . . . . . . 55 109 4. Semantic Issues . . . . . . . . . . . . . . . . . . . . . . . 56 110 5. Security Considerations . . . . . . . . . . . . . . . . . . . 57 111 6. Interactions with Middleboxes . . . . . . . . . . . . . . . . 60 112 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 63 113 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 64 114 8.1. MPTCP Option Subtypes . . . . . . . . . . . . . . . . . . 64 115 8.2. MPTCP Handshake Algorithms . . . . . . . . . . . . . . . 65 116 8.3. MP_TCPRST Reason Codes . . . . . . . . . . . . . . . . . 66 117 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 67 118 9.1. Normative References . . . . . . . . . . . . . . . . . . 67 119 9.2. Informative References . . . . . . . . . . . . . . . . . 67 120 Appendix A. Notes on Use of TCP Options . . . . . . . . . . . . 71 121 Appendix B. TCP Fast Open and MPTCP . . . . . . . . . . . . . . 72 122 B.1. TFO cookie request with MPTCP . . . . . . . . . . . . . . 72 123 B.2. Data sequence mapping under TFO . . . . . . . . . . . . . 73 124 B.3. Connection establishment examples . . . . . . . . . . . . 74 125 Appendix C. Control Blocks . . . . . . . . . . . . . . . . . . . 76 126 C.1. MPTCP Control Block . . . . . . . . . . . . . . . . . . . 76 127 C.1.1. Authentication and Metadata . . . . . . . . . . . . . 76 128 C.1.2. Sending Side . . . . . . . . . . . . . . . . . . . . 77 129 C.1.3. Receiving Side . . . . . . . . . . . . . . . . . . . 77 130 C.2. TCP Control Blocks . . . . . . . . . . . . . . . . . . . 77 131 C.2.1. Sending Side . . . . . . . . . . . . . . . . . . . . 78 132 C.2.2. Receiving Side . . . . . . . . . . . . . . . . . . . 78 133 Appendix D. Finite State Machine . . . . . . . . . . . . . . . . 78 134 Appendix E. Changes from RFC6184 . . . . . . . . . . . . . . . . 79 135 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 81 137 1. Introduction 139 Multipath TCP (MPTCP) is a set of extensions to regular TCP [RFC0793] 140 to provide a Multipath TCP [RFC6182] service, which enables a 141 transport connection to operate across multiple paths simultaneously. 142 This document presents the protocol changes required to add multipath 143 capability to TCP; specifically, those for signaling and setting up 144 multiple paths ("subflows"), managing these subflows, reassembly of 145 data, and termination of sessions. This is not the only information 146 required to create a Multipath TCP implementation, however. This 147 document is complemented by three others: 149 o Architecture [RFC6182], which explains the motivations behind 150 Multipath TCP, contains a discussion of high-level design 151 decisions on which this design is based, and an explanation of a 152 functional separation through which an extensible MPTCP 153 implementation can be developed. 155 o Congestion control [RFC6356] presents a safe congestion control 156 algorithm for coupling the behavior of the multiple paths in order 157 to "do no harm" to other network users. 159 o Application considerations [RFC6897] discusses what impact MPTCP 160 will have on applications, what applications will want to do with 161 MPTCP, and as a consequence of these factors, what API extensions 162 an MPTCP implementation should present. 164 This document is an update to, and obsoletes, the v0 specification of 165 Multipath TCP [RFC6824]. This document specifies MPTCP v1, which is 166 not backward compatible with MPTCP v0. This document additionally 167 defines version negotiation procedures for implementations that 168 support both versions. 170 1.1. Design Assumptions 172 In order to limit the potentially huge design space, the working 173 group imposed two key constraints on the Multipath TCP design 174 presented in this document: 176 o It must be backwards-compatible with current, regular TCP, to 177 increase its chances of deployment. 179 o It can be assumed that one or both hosts are multihomed and 180 multiaddressed. 182 To simplify the design, we assume that the presence of multiple 183 addresses at a host is sufficient to indicate the existence of 184 multiple paths. These paths need not be entirely disjoint: they may 185 share one or many routers between them. Even in such a situation, 186 making use of multiple paths is beneficial, improving resource 187 utilization and resilience to a subset of node failures. The 188 congestion control algorithms defined in [RFC6356] ensure this does 189 not act detrimentally. Furthermore, there may be some scenarios 190 where different TCP ports on a single host can provide disjoint paths 191 (such as through certain Equal-Cost Multipath (ECMP) implementations 193 [RFC2992]), and so the MPTCP design also supports the use of ports in 194 path identifiers. 196 There are three aspects to the backwards-compatibility listed above 197 (discussed in more detail in [RFC6182]): 199 External Constraints: The protocol must function through the vast 200 majority of existing middleboxes such as NATs, firewalls, and 201 proxies, and as such must resemble existing TCP as far as possible 202 on the wire. Furthermore, the protocol must not assume the 203 segments it sends on the wire arrive unmodified at the 204 destination: they may be split or coalesced; TCP options may be 205 removed or duplicated. 207 Application Constraints: The protocol must be usable with no change 208 to existing applications that use the common TCP API (although it 209 is reasonable that not all features would be available to such 210 legacy applications). Furthermore, the protocol must provide the 211 same service model as regular TCP to the application. 213 Fallback: The protocol should be able to fall back to standard TCP 214 with no interference from the user, to be able to communicate with 215 legacy hosts. 217 The complementary application considerations document [RFC6897] 218 discusses the necessary features of an API to provide backwards- 219 compatibility, as well as API extensions to convey the behavior of 220 MPTCP at a level of control and information equivalent to that 221 available with regular, single-path TCP. 223 Further discussion of the design constraints and associated design 224 decisions are given in the MPTCP Architecture document [RFC6182] and 225 in [howhard]. 227 1.2. Multipath TCP in the Networking Stack 229 MPTCP operates at the transport layer and aims to be transparent to 230 both higher and lower layers. It is a set of additional features on 231 top of standard TCP; Figure 1 illustrates this layering. MPTCP is 232 designed to be usable by legacy applications with no changes; 233 detailed discussion of its interactions with applications is given in 234 [RFC6897]. 236 +-------------------------------+ 237 | Application | 238 +---------------+ +-------------------------------+ 239 | Application | | MPTCP | 240 +---------------+ + - - - - - - - + - - - - - - - + 241 | TCP | | Subflow (TCP) | Subflow (TCP) | 242 +---------------+ +-------------------------------+ 243 | IP | | IP | IP | 244 +---------------+ +-------------------------------+ 246 Figure 1: Comparison of Standard TCP and MPTCP Protocol Stacks 248 1.3. Terminology 250 This document makes use of a number of terms that are either MPTCP- 251 specific or have defined meaning in the context of MPTCP, as follows: 253 Path: A sequence of links between a sender and a receiver, defined 254 in this context by a 4-tuple of source and destination address/ 255 port pairs. 257 Subflow: A flow of TCP segments operating over an individual path, 258 which forms part of a larger MPTCP connection. A subflow is 259 started and terminated similar to a regular TCP connection. 261 (MPTCP) Connection: A set of one or more subflows, over which an 262 application can communicate between two hosts. There is a one-to- 263 one mapping between a connection and an application socket. 265 Data-level: The payload data is nominally transferred over a 266 connection, which in turn is transported over subflows. Thus, the 267 term "data-level" is synonymous with "connection level", in 268 contrast to "subflow-level", which refers to properties of an 269 individual subflow. 271 Token: A locally unique identifier given to a multipath connection 272 by a host. May also be referred to as a "Connection ID". 274 Host: An end host operating an MPTCP implementation, and either 275 initiating or accepting an MPTCP connection. 277 In addition to these terms, note that MPTCP's interpretation of, and 278 effect on, regular single-path TCP semantics are discussed in 279 Section 4. 281 1.4. MPTCP Concept 283 This section provides a high-level summary of normal operation of 284 MPTCP, and is illustrated by the scenario shown in Figure 2. A 285 detailed description of operation is given in Section 3. 287 o To a non-MPTCP-aware application, MPTCP will behave the same as 288 normal TCP. Extended APIs could provide additional control to 289 MPTCP-aware applications [RFC6897]. An application begins by 290 opening a TCP socket in the normal way. MPTCP signaling and 291 operation are handled by the MPTCP implementation. 293 o An MPTCP connection begins similarly to a regular TCP connection. 294 This is illustrated in Figure 2 where an MPTCP connection is 295 established between addresses A1 and B1 on Hosts A and B, 296 respectively. 298 o If extra paths are available, additional TCP sessions (termed 299 MPTCP "subflows") are created on these paths, and are combined 300 with the existing session, which continues to appear as a single 301 connection to the applications at both ends. The creation of the 302 additional TCP session is illustrated between Address A2 on Host A 303 and Address B1 on Host B. 305 o MPTCP identifies multiple paths by the presence of multiple 306 addresses at hosts. Combinations of these multiple addresses 307 equate to the additional paths. In the example, other potential 308 paths that could be set up are A1<->B2 and A2<->B2. Although this 309 additional session is shown as being initiated from A2, it could 310 equally have been initiated from B1. 312 o The discovery and setup of additional subflows will be achieved 313 through a path management method; this document describes a 314 mechanism by which a host can initiate new subflows by using its 315 own additional addresses, or by signaling its available addresses 316 to the other host. 318 o MPTCP adds connection-level sequence numbers to allow the 319 reassembly of segments arriving on multiple subflows with 320 differing network delays. 322 o Subflows are terminated as regular TCP connections, with a four- 323 way FIN handshake. The MPTCP connection is terminated by a 324 connection-level FIN. 326 Host A Host B 327 ------------------------ ------------------------ 328 Address A1 Address A2 Address B1 Address B2 329 ---------- ---------- ---------- ---------- 330 | | | | 331 | (initial connection setup) | | 332 |----------------------------------->| | 333 |<-----------------------------------| | 334 | | | | 335 | (additional subflow setup) | 336 | |--------------------->| | 337 | |<---------------------| | 338 | | | | 339 | | | | 341 Figure 2: Example MPTCP Usage Scenario 343 1.5. Requirements Language 345 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 346 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 347 "OPTIONAL" in this document are to be interpreted as described in 348 BCP 14 RFC 2119 [RFC2119] RFC 8174 [RFC8174] when, and only when, 349 they appear in all capitals, as shown here. 351 2. Operation Overview 353 This section presents a single description of common MPTCP operation, 354 with reference to the protocol operation. This is a high-level 355 overview of the key functions; the full specification follows in 356 Section 3. Extensibility and negotiated features are not discussed 357 here. Considerable reference is made to symbolic names of MPTCP 358 options throughout this section -- these are subtypes of the IANA- 359 assigned MPTCP option (see Section 8), and their formats are defined 360 in the detailed protocol specification that follows in Section 3. 362 A Multipath TCP connection provides a bidirectional bytestream 363 between two hosts communicating like normal TCP and, thus, does not 364 require any change to the applications. However, Multipath TCP 365 enables the hosts to use different paths with different IP addresses 366 to exchange packets belonging to the MPTCP connection. A Multipath 367 TCP connection appears like a normal TCP connection to an 368 application. However, to the network layer, each MPTCP subflow looks 369 like a regular TCP flow whose segments carry a new TCP option type. 370 Multipath TCP manages the creation, removal, and utilization of these 371 subflows to send data. The number of subflows that are managed 372 within a Multipath TCP connection is not fixed and it can fluctuate 373 during the lifetime of the Multipath TCP connection. 375 All MPTCP operations are signaled with a TCP option -- a single 376 numerical type for MPTCP, with "sub-types" for each MPTCP message. 377 What follows is a summary of the purpose and rationale of these 378 messages. 380 2.1. Initiating an MPTCP Connection 382 This is the same signaling as for initiating a normal TCP connection, 383 but the SYN, SYN/ACK, and initial ACK (and data) packets also carry 384 the MP_CAPABLE option. This option has a variable length and serves 385 multiple purposes. Firstly, it verifies whether the remote host 386 supports Multipath TCP; secondly, this option allows the hosts to 387 exchange some information to authenticate the establishment of 388 additional subflows. Further details are given in Section 3.1. 390 Host A Host B 391 ------ ------ 392 MP_CAPABLE -> 393 [flags] 394 <- MP_CAPABLE 395 [B's key, flags] 396 ACK + MP_CAPABLE (+ data) -> 397 [A's key, B's key, flags, (data-level details)] 399 Retransmission of the ACK + MP_CAPABLE can occur if it is not known 400 if it has been received. The following diagrams show all possible 401 exchanges for the initial subflow setup to ensure this reliability. 403 Host A (with data to send immediately) Host B 404 ------ ------ 405 MP_CAPABLE -> 406 [flags] 407 <- MP_CAPABLE 408 [B's key, flags] 409 ACK + MP_CAPABLE + data -> 410 [A's key, B's key, flags, data-level details] 412 Host A (with data to send later) Host B 413 ------ ------ 414 MP_CAPABLE -> 415 [flags] 416 <- MP_CAPABLE 417 [B's key, flags] 418 ACK + MP_CAPABLE -> 419 [A's key, B's key, flags] 421 ACK + MP_CAPABLE + data -> 422 [A's key, B's key, flags, data-level details] 424 Host A Host B (sending first) 425 ------ ------ 426 MP_CAPABLE -> 427 [flags] 428 <- MP_CAPABLE 429 [B's key, flags] 430 ACK + MP_CAPABLE -> 431 [A's key, B's key, flags] 433 <- ACK + DSS + data 434 [data-level details] 436 2.2. Associating a New Subflow with an Existing MPTCP Connection 438 The exchange of keys in the MP_CAPABLE handshake provides material 439 that can be used to authenticate the endpoints when new subflows will 440 be set up. Additional subflows begin in the same way as initiating a 441 normal TCP connection, but the SYN, SYN/ACK, and ACK packets also 442 carry the MP_JOIN option. 444 Host A initiates a new subflow between one of its addresses and one 445 of Host B's addresses. The token -- generated from the key -- is 446 used to identify which MPTCP connection it is joining, and the HMAC 447 is used for authentication. The Hash-based Message Authentication 448 Code (HMAC) uses the keys exchanged in the MP_CAPABLE handshake, and 449 the random numbers (nonces) exchanged in these MP_JOIN options. 450 MP_JOIN also contains flags and an Address ID that can be used to 451 refer to the source address without the sender needing to know if it 452 has been changed by a NAT. Further details are in Section 3.2. 454 Host A Host B 455 ------ ------ 456 MP_JOIN -> 457 [B's token, A's nonce, 458 A's Address ID, flags] 459 <- MP_JOIN 460 [B's HMAC, B's nonce, 461 B's Address ID, flags] 462 ACK + MP_JOIN -> 463 [A's HMAC] 465 <- ACK 467 2.3. Informing the Other Host about Another Potential Address 469 The set of IP addresses associated to a multihomed host may change 470 during the lifetime of an MPTCP connection. MPTCP supports the 471 addition and removal of addresses on a host both implicitly and 472 explicitly. If Host A has established a subflow starting at address/ 473 port pair IP#-A1 and wants to open a second subflow starting at 474 address/port pair IP#-A2, it simply initiates the establishment of 475 the subflow as explained above. The remote host will then be 476 implicitly informed about the new address. 478 In some circumstances, a host may want to advertise to the remote 479 host the availability of an address without establishing a new 480 subflow, for example, when a NAT prevents setup in one direction. In 481 the example below, Host A informs Host B about its alternative IP 482 address/port pair (IP#-A2). Host B may later send an MP_JOIN to this 483 new address. The ADD_ADDR option contains a HMAC to authenticate the 484 address as having been sent from the originator of the connection. 485 The receiver of this option echoes it back to the client to indicate 486 successful reception. Further details are in Section 3.4.1. 488 Host A Host B 489 ------ ------ 490 ADD_ADDR -> 491 [Echo-flag=0, 492 IP#-A2, 493 IP#-A2's Address ID, 494 HMAC of IP#-A2] 496 <- ADD_ADDR 497 [Echo-flag=1, 498 IP#-A2, 499 IP#-A2's Address ID, 500 HMAC of IP#-A2] 502 There is a corresponding signal for address removal, making use of 503 the Address ID that is signaled in the add address handshake. 504 Further details in Section 3.4.2. 506 Host A Host B 507 ------ ------ 508 REMOVE_ADDR -> 509 [IP#-A2's Address ID] 511 2.4. Data Transfer Using MPTCP 513 To ensure reliable, in-order delivery of data over subflows that may 514 appear and disappear at any time, MPTCP uses a 64-bit data sequence 515 number (DSN) to number all data sent over the MPTCP connection. Each 516 subflow has its own 32-bit sequence number space, utilising the 517 regular TCP sequence number header, and an MPTCP option maps the 518 subflow sequence space to the data sequence space. In this way, data 519 can be retransmitted on different subflows (mapped to the same DSN) 520 in the event of failure. 522 The Data Sequence Signal (DSS) carries the Data Sequence Mapping. 523 The Data Sequence Mapping consists of the subflow sequence number, 524 data sequence number, and length for which this mapping is valid. 525 This option can also carry a connection-level acknowledgment (the 526 "Data ACK") for the received DSN. 528 With MPTCP, all subflows share the same receive buffer and advertise 529 the same receive window. There are two levels of acknowledgment in 530 MPTCP. Regular TCP acknowledgments are used on each subflow to 531 acknowledge the reception of the segments sent over the subflow 532 independently of their DSN. In addition, there are connection-level 533 acknowledgments for the data sequence space. These acknowledgments 534 track the advancement of the bytestream and slide the receiving 535 window. 537 Further details are in Section 3.3. 539 Host A Host B 540 ------ ------ 541 DSS -> 542 [Data Sequence Mapping] 543 [Data ACK] 544 [Checksum] 546 2.5. Requesting a Change in a Path's Priority 548 Hosts can indicate at initial subflow setup whether they wish the 549 subflow to be used as a regular or backup path -- a backup path only 550 being used if there are no regular paths available. During a 551 connection, Host A can request a change in the priority of a subflow 552 through the MP_PRIO signal to Host B. Further details are in 553 Section 3.3.8. 555 Host A Host B 556 ------ ------ 557 MP_PRIO -> 559 2.6. Closing an MPTCP Connection 561 When a host wants to close an existing subflow, but not the whole 562 connection, it can initiate a regular TCP FIN/ACK exchange. 564 When Host A wants to inform Host B that it has no more data to send, 565 it signals this "DATA_FIN" as part of the Data Sequence Signal (see 566 above). It has the same semantics and behavior as a regular TCP FIN, 567 but at the connection level. Once all the data on the MPTCP 568 connection has been successfully received, then this message is 569 acknowledged at the connection level with a DATA_ACK. Further 570 details are in Section 3.3.3. 572 Host A Host B 573 ------ ------ 574 DATA_SEQUENCE_SIGNAL -> 575 [DATA_FIN] 576 <- (MPTCP DATA_ACK) 578 There is an additional method of connection closure, referred to as 579 "Fast Close", which is analogous to closing a single-path TCP 580 connection with a RST signal. The MP_FASTCLOSE signal is used to 581 indicate to the peer that the connection will be abruptly closed and 582 no data will be accepted anymore. This can be used on an ACK 583 (ensuring reliability of the signal), or a RST (which is not). Both 584 examples are shown in the following diagrams. Further details are in 585 Section 3.5. 587 Host A Host B 588 ------ ------ 589 ACK + MP_FASTCLOSE -> 590 [B's key] 592 [RST on all other subflows] -> 594 <- [RST on all subflows] 596 Host A Host B 597 ------ ------ 598 RST + MP_FASTCLOSE -> 599 [B's key] [on all subflows] 601 <- [RST on all subflows] 603 2.7. Notable Features 605 It is worth highlighting that MPTCP's signaling has been designed 606 with several key requirements in mind: 608 o To cope with NATs on the path, addresses are referred to by 609 Address IDs, in case the IP packet's source address gets changed 610 by a NAT. Setting up a new TCP flow is not possible if the 611 receiver of the SYN is behind a NAT; to allow subflows to be 612 created when either end is behind a NAT, MPTCP uses the ADD_ADDR 613 message. 615 o MPTCP falls back to ordinary TCP if MPTCP operation is not 616 possible, for example, if one host is not MPTCP capable or if a 617 middlebox alters the payload. This is discussed in Section 3.7. 619 o To address the threats identified in [RFC6181], the following 620 steps are taken: keys are sent in the clear in the MP_CAPABLE 621 messages; MP_JOIN messages are secured with HMAC-SHA256 622 ([RFC2104], [SHS]) using those keys; and standard TCP validity 623 checks are made on the other messages (ensuring sequence numbers 624 are in-window [RFC5961]). Residual threats to MPTCP v0 [RFC6824] 625 were identified in [RFC7430], and those affecting the protocol 626 (i.e. modification to ADD_ADDR) have been incorporated in this 627 document. Further discussion of security can be found in 628 Section 5. 630 3. MPTCP Protocol 632 This section describes the operation of the MPTCP protocol, and is 633 subdivided into sections for each key part of the protocol operation. 635 All MPTCP operations are signaled using optional TCP header fields. 636 A single TCP option number ("Kind") has been assigned by IANA for 637 MPTCP (see Section 8), and then individual messages will be 638 determined by a "subtype", the values of which are also stored in an 639 IANA registry (and are also listed in Section 8). As with all TCP 640 options, the Length field is specified in bytes, and includes the 2 641 bytes of Kind and Length. 643 Throughout this document, when reference is made to an MPTCP option 644 by symbolic name, such as "MP_CAPABLE", this refers to a TCP option 645 with the single MPTCP option type, and with the subtype value of the 646 symbolic name as defined in Section 8. This subtype is a 4-bit field 647 -- the first 4 bits of the option payload, as shown in Figure 3. The 648 MPTCP messages are defined in the following sections. 650 1 2 3 651 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 1 652 +---------------+---------------+-------+-----------------------+ 653 | Kind | Length |Subtype| | 654 +---------------+---------------+-------+ | 655 | Subtype-specific data | 656 | (variable length) | 657 +---------------------------------------------------------------+ 659 Figure 3: MPTCP Option Format 661 Those MPTCP options associated with subflow initiation are used on 662 packets with the SYN flag set. Additionally, there is one MPTCP 663 option for signaling metadata to ensure segmented data can be 664 recombined for delivery to the application. 666 The remaining options, however, are signals that do not need to be on 667 a specific packet, such as those for signaling additional addresses. 668 Whilst an implementation may desire to send MPTCP options as soon as 669 possible, it may not be possible to combine all desired options (both 670 those for MPTCP and for regular TCP, such as SACK (selective 671 acknowledgment) [RFC2018]) on a single packet. Therefore, an 672 implementation may choose to send duplicate ACKs containing the 673 additional signaling information. This changes the semantics of a 674 duplicate ACK; these are usually only sent as a signal of a lost 675 segment [RFC5681] in regular TCP. Therefore, an MPTCP implementation 676 receiving a duplicate ACK that contains an MPTCP option MUST NOT 677 treat it as a signal of congestion. Additionally, an MPTCP 678 implementation SHOULD NOT send more than two duplicate ACKs in a row 679 for the purposes of sending MPTCP options alone, in order to ensure 680 no middleboxes misinterpret this as a sign of congestion. 682 Furthermore, standard TCP validity checks (such as ensuring the 683 sequence number and acknowledgment number are within window) MUST be 684 undertaken before processing any MPTCP signals, as described in 685 [RFC5961], and initial subflow sequence numbers SHOULD be generated 686 according to the recommendations in [RFC6528]. 688 3.1. Connection Initiation 690 Connection initiation begins with a SYN, SYN/ACK, ACK exchange on a 691 single path. Each packet contains the Multipath Capable (MP_CAPABLE) 692 MPTCP option (Figure 4). This option declares its sender is capable 693 of performing Multipath TCP and wishes to do so on this particular 694 connection. 696 The MP_CAPABLE exchange in this specification (v1) is different to 697 that specified in v0 [RFC6824]. If a host supports multiple versions 698 of MPTCP, the sender of the MP_CAPABLE option SHOULD signal the 699 highest version number it supports. In return, in its MP_CAPABLE 700 option, the receiver will signal the version number it wishes to use, 701 which MUST be equal to or lower than the version number indicated in 702 the initial MP_CAPABLE. There is a caveat though with respect to 703 this version negotiation with old listeners that only support v0. A 704 listener that supports v0 expects that the MP_CAPABLE option in the 705 SYN-segment includes the initiator's key. If the initiator however 706 already upgraded to v1, it won't include the key in the SYN-segment. 707 Thus, the listener will ignore the MP_CAPABLE of this SYN-segment and 708 reply with a SYN/ACK that does not include an MP_CAPABLE, thus 709 leading to a fallback to regular TCP. An initiator MAY cache this 710 information about a peer and for future connections, MAY choose to 711 attempt using MPTCP v0, if supported, before recording the host as 712 not supporting MPTCP. 714 The MP_CAPABLE option is variable-length, with different fields 715 included depending on which packet the option is used on. The full 716 MP_CAPABLE option is shown in Figure 4. 718 1 2 3 719 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 1 720 +---------------+---------------+-------+-------+---------------+ 721 | Kind | Length |Subtype|Version|A|B|C|D|E|F|G|H| 722 +---------------+---------------+-------+-------+---------------+ 723 | Option Sender's Key (64 bits) | 724 | (if option Length > 4) | 725 | | 726 +---------------------------------------------------------------+ 727 | Option Receiver's Key (64 bits) | 728 | (if option Length > 12) | 729 | | 730 +-------------------------------+-------------------------------+ 731 | Data-Level Length (16 bits) | Checksum (16 bits, optional) | 732 +-------------------------------+-------------------------------+ 734 Figure 4: Multipath Capable (MP_CAPABLE) Option 736 The MP_CAPABLE option is carried on the SYN, SYN/ACK, and ACK packets 737 that start the first subflow of an MPTCP connection, as well as the 738 first packet that carries data, if the initiator wishes to send 739 first. The data carried by each option is as follows, where A = 740 initiator and B = listener. 742 o SYN (A->B): only the first four octets (Length = 4). 744 o SYN/ACK (B->A): B's Key for this connection (Length = 12). 746 o ACK (no data) (A->B): A's Key followed by B's Key (Length = 20). 748 o ACK (with first data) (A->B): A's Key followed by B's Key followed 749 by Data-Level Length, and optional Checksum (Length = 22 or 24). 751 The contents of the option is determined by the SYN and ACK flags of 752 the packet, along with the option's length field. For the diagram 753 shown in Figure 4, "sender" and "receiver" refer to the sender or 754 receiver of the TCP packet (which can be either host). 756 The initial SYN, containing just the MP_CAPABLE header, is used to 757 define the version of MPTCP being requested, as well as exchanging 758 flags to negotiate connection features, described later. 760 This option is used to declare the 64-bit keys that the end hosts 761 have generated for this MPTCP connection. These keys are used to 762 authenticate the addition of future subflows to this connection. 763 This is the only time the key will be sent in clear on the wire 764 (unless "fast close", Section 3.5, is used); all future subflows will 765 identify the connection using a 32-bit "token". This token is a 766 cryptographic hash of this key. The algorithm for this process is 767 dependent on the authentication algorithm selected; the method of 768 selection is defined later in this section. 770 Upon reception of the initial SYN-segment, a stateful server 771 generates a random key and replies with a SYN/ACK. The key's method 772 of generation is implementation specific. The key MUST be hard to 773 guess, and it MUST be unique for the sending host across all its 774 current MPTCP connections. Recommendations for generating random 775 numbers for use in keys are given in [RFC4086]. Connections will be 776 indexed at each host by the token (a one-way hash of the key). 777 Therefore, an implementation will require a mapping from each token 778 to the corresponding connection, and in turn to the keys for the 779 connection. 781 There is a risk that two different keys will hash to the same token. 782 The risk of hash collisions is usually small, unless the host is 783 handling many tens of thousands of connections. Therefore, an 784 implementation SHOULD check its list of connection tokens to ensure 785 there is no collision before sending its key, and if there is, then 786 it should generate a new key. This would, however, be costly for a 787 server with thousands of connections. The subflow handshake 788 mechanism (Section 3.2) will ensure that new subflows only join the 789 correct connection, however, through the cryptographic handshake, as 790 well as checking the connection tokens in both directions, and 791 ensuring sequence numbers are in-window. So in the worst case if 792 there was a token collision, the new subflow would not succeed, but 793 the MPTCP connection would continue to provide a regular TCP service. 795 Since key generation is implementation-specific, there is no 796 requirement that they be simply random numbers. An implementation is 797 free to exchange cryptographic material out-of-band and generate 798 these keys from this, in order to provide additional mechanisms by 799 which to verify the identity of the communicating entities. For 800 example, an implementation could choose to link its MPTCP keys to 801 those used in higher-layer TLS or SSH connections. 803 If the server behaves in a stateless manner, it has to generate its 804 own key in a verifiable fashion. This verifiable way of generating 805 the key can be done by using a hash of the 4-tuple, sequence number 806 and a local secret (similar to what is done for the TCP-sequence 807 number [RFC4987]). It will thus be able to verify whether it is 808 indeed the originator of the key echoed back in the later MP_CAPABLE 809 option. As for a stateful server, the tokens SHOULD be checked for 810 uniqueness, however if uniqueness is not met, and there is no way to 811 generate an alternative verifiable key, then the connection MUST fall 812 back to using regular TCP by not sending a MP_CAPABLE in the SYN/ACK. 814 The ACK carries both A's key and B's key. This is the first time 815 that A's key is seen on the wire, although it is expected that A will 816 have generated a key locally before the initial SYN. The echoing of 817 B's key allows B to operate statelessly, as described above. 818 Therefore, A's key must be delivered reliably to B, and in order to 819 do this, the transmission of this packet must be made reliable. 821 If B has data to send first, then the reliable delivery of the 822 ACK+MP_CAPABLE can be inferred by the receipt of this data with a 823 MPTCP Data Sequence Signal (DSS) option (Section 3.3). If, however, 824 A wishes to send data first, it has two options to ensure the 825 reliable delivery of the ACK+MP_CAPABLE. If it immediately has data 826 to send, then the third ACK (with data) would also contain an 827 MP_CAPABLE option with additional data parameters (the Data-Level 828 Length and optional Checksum as shown in Figure 4). If A does not 829 immediately have data to send, it MUST include the MP_CAPABLE on the 830 third ACK, but without the additional data parameters. When A does 831 have data to send, it must repeat the sending of the MP_CAPABLE 832 option from the third ACK, with additional data parameters. This 833 MP_CAPABLE option is in place of the DSS, and simply specifies the 834 data-level length of the payload, and the checksum (if the use of 835 checksums is negotiated). This is the minimal data required to 836 establish a MPTCP connection - it allows validation of the payload, 837 and given it is the first data, the Initial Data Sequence Number 838 (IDSN) is also known (as it is generated from the key, as described 839 below). Conveying the keys on the first data packet allows the TCP 840 reliability mechanisms to ensure the packet is successfully 841 delivered. The receiver will acknowledge this data at the connection 842 level with a Data ACK, as if a DSS option has been received. 844 There could be situations where both A and B attempt to transmit 845 initial data at the same time. For example, if A did not initially 846 have data to send, but then needed to transmit data before it had 847 received anything from B, it would use a MP_CAPABLE option with data 848 parameters (since it would not know if the MP_CAPABLE on the ACK was 849 received). In such a situation, B may also have transmitted data 850 with a DSS option, but it had not yet been received at A. Therefore, 851 B has received data with a MP_CAPABLE mapping after it has sent data 852 with a DSS option. To ensure these situations can be handled, it 853 follows that the data parameters in a MP_CAPABLE are semantically 854 equivalent to those in a DSS option and can be used interchangeably. 855 Similar situations could occur when the MP_CAPABLE with data is lost 856 and retransmitted. Furthermore, in the case of TCP Segmentation 857 Offloading, the MP_CAPABLE with data parameters may be duplicated 858 across multiple packets, and implementations must also be able to 859 cope with duplicate MP_CAPABLE mappings as well as duplicate DSS 860 mappings. 862 Additionally, the MP_CAPABLE exchange allows the safe passage of 863 MPTCP options on SYN packets to be determined. If any of these 864 options are dropped, MPTCP will gracefully fall back to regular 865 single-path TCP, as documented in Section 3.7. If at any point in 866 the handshake either party thinks the MPTCP negotiation is 867 compromised, for example by a middlebox corrupting the TCP options, 868 or unexpected ACK numbers being present, the host MUST stop using 869 MPTCP and no longer include MPTCP options in future TCP packets. The 870 other host will then also fall back to regular TCP using the fall 871 back mechanism. Note that new subflows MUST NOT be established 872 (using the process documented in Section 3.2) until a Data Sequence 873 Signal (DSS) option has been successfully received across the path 874 (as documented in Section 3.3). 876 Like all MPTCP options, the MP_CAPABLE option starts with the Kind 877 and Length to specify the TCP-option kind and its length. Followed 878 by that is the MP_CAPABLE option. The first 4 bits of the first 879 octet in the MP_CAPABLE option (Figure 4) define the MPTCP option 880 subtype (see Section 8; for MP_CAPABLE, this is 0x0), and the 881 remaining 4 bits of this octet specify the MPTCP version in use (for 882 this specification, this is 1). 884 The second octet is reserved for flags, allocated as follows: 886 A: The leftmost bit, labeled "A", SHOULD be set to 1 to indicate 887 "Checksum Required", unless the system administrator has decided 888 that checksums are not required (for example, if the environment 889 is controlled and no middleboxes exist that might adjust the 890 payload). 892 B: The second bit, labeled "B", is an extensibility flag, and MUST be 893 set to 0 for current implementations. This will be used for an 894 extensibility mechanism in a future specification, and the impact 895 of this flag will be defined at a later date. If receiving a 896 message with the 'B' flag set to 1, and this is not understood, 897 then the MP_CAPABLE in this SYN MUST be silently ignored, which 898 triggers a fallback to regular TCP; the sender is expected to 899 retry with a format compatible with this legacy specification. 900 Note that the length of the MP_CAPABLE option, and the meanings of 901 bits "C" through "H", may be altered by setting B=1. 903 C: The third bit, labeled "C", is set to "1" to indicate that the 904 sender of this option will not accept additional MPTCP subflows to 905 the source address and port, and therefore the receiver MUST NOT 906 try to open any additional subflows towards this address and port. 907 This is an efficiency improvement for situations where the sender 908 knows a restriction is in place, for example if the sender is 909 behind a strict NAT, or operating behind a legacy Layer 4 load 910 balancer. 912 D through H: The remaining bits, labeled "D" through "H", are used 913 for crypto algorithm negotiation. In this specification only the 914 rightmost bit, labeled "H", is assigned. Bit "H" indicates the 915 use of HMAC-SHA256 (as defined in Section 3.2). An implementation 916 that only supports this method MUST set bit "H" to 1, and bits "D" 917 through "G" to 0. 919 A crypto algorithm MUST be specified. If flag bits D through H are 920 all 0, the MP_CAPABLE option MUST be treated as invalid and ignored 921 (that is, it must be treated as a regular TCP handshake). 923 The selection of the authentication algorithm also impacts the 924 algorithm used to generate the token and the Initial Data Sequence 925 Number (IDSN). In this specification, with only the SHA-256 926 algorithm (bit "H") specified and selected, the token MUST be a 927 truncated (most significant 32 bits) SHA-256 hash ([SHS], [RFC6234]) 928 of the key. A different, 64-bit truncation (the least significant 64 929 bits) of the SHA-256 hash of the key MUST be used as the IDSN. Note 930 that the key MUST be hashed in network byte order. Also note that 931 the "least significant" bits MUST be the rightmost bits of the 932 SHA-256 digest, as per [SHS]. Future specifications of the use of 933 the crypto bits may choose to specify different algorithms for token 934 and IDSN generation. 936 Both the crypto and checksum bits negotiate capabilities in similar 937 ways. For the Checksum Required bit (labeled "A"), if either host 938 requires the use of checksums, checksums MUST be used. In other 939 words, the only way for checksums not to be used is if both hosts in 940 their SYNs set A=0. This decision is confirmed by the setting of the 941 "A" bit in the third packet (the ACK) of the handshake. For example, 942 if the initiator sets A=0 in the SYN, but the responder sets A=1 in 943 the SYN/ACK, checksums MUST be used in both directions, and the 944 initiator will set A=1 in the ACK. The decision whether to use 945 checksums will be stored by an implementation in a per-connection 946 binary state variable. If A=1 is received by a host that does not 947 want to use checksums, it MUST fall back to regular TCP by ignoring 948 the MP_CAPABLE option as if it was invalid. 950 For crypto negotiation, the responder has the choice. The initiator 951 creates a proposal setting a bit for each algorithm it supports to 1 952 (in this version of the specification, there is only one proposal, so 953 bit "H" will be always set to 1). The responder responds with only 1 954 bit set -- this is the chosen algorithm. The rationale for this 955 behavior is that the responder will typically be a server with 956 potentially many thousands of connections, so it may wish to choose 957 an algorithm with minimal computational complexity, depending on the 958 load. If a responder does not support (or does not want to support) 959 any of the initiator's proposals, it can respond without an 960 MP_CAPABLE option, thus forcing a fallback to regular TCP. 962 The MP_CAPABLE option is only used in the first subflow of a 963 connection, in order to identify the connection; all following 964 subflows will use the "Join" option (see Section 3.2) to join the 965 existing connection. 967 If a SYN contains an MP_CAPABLE option but the SYN/ACK does not, it 968 is assumed that sender of the SYN/ACK is not multipath capable; thus, 969 the MPTCP session MUST operate as a regular, single-path TCP. If a 970 SYN does not contain a MP_CAPABLE option, the SYN/ACK MUST NOT 971 contain one in response. If the third packet (the ACK) does not 972 contain the MP_CAPABLE option, then the session MUST fall back to 973 operating as a regular, single-path TCP. This is to maintain 974 compatibility with middleboxes on the path that drop some or all TCP 975 options. Note that an implementation MAY choose to attempt sending 976 MPTCP options more than one time before making this decision to 977 operate as regular TCP (see Section 3.9). 979 If the SYN packets are unacknowledged, it is up to local policy to 980 decide how to respond. It is expected that a sender will eventually 981 fall back to single-path TCP (i.e., without the MP_CAPABLE option) in 982 order to work around middleboxes that may drop packets with unknown 983 options; however, the number of multipath-capable attempts that are 984 made first will be up to local policy. It is possible that MPTCP and 985 non-MPTCP SYNs could get reordered in the network. Therefore, the 986 final state is inferred from the presence or absence of the 987 MP_CAPABLE option in the third packet of the TCP handshake. If this 988 option is not present, the connection SHOULD fall back to regular 989 TCP, as documented in Section 3.7. 991 The initial data sequence number on an MPTCP connection is generated 992 from the key. The algorithm for IDSN generation is also determined 993 from the negotiated authentication algorithm. In this specification, 994 with only the SHA-256 algorithm specified and selected, the IDSN of a 995 host MUST be the least significant 64 bits of the SHA-256 hash of its 996 key, i.e., IDSN-A = Hash(Key-A) and IDSN-B = Hash(Key-B). This 997 deterministic generation of the IDSN allows a receiver to ensure that 998 there are no gaps in sequence space at the start of the connection. 999 The SYN with MP_CAPABLE occupies the first octet of data sequence 1000 space, although this does not need to be acknowledged at the 1001 connection level until the first data is sent (see Section 3.3). 1003 3.2. Starting a New Subflow 1005 Once an MPTCP connection has begun with the MP_CAPABLE exchange, 1006 further subflows can be added to the connection. Hosts have 1007 knowledge of their own address(es), and can become aware of the other 1008 host's addresses through signaling exchanges as described in 1009 Section 3.4. Using this knowledge, a host can initiate a new subflow 1010 over a currently unused pair of addresses. It is permitted for 1011 either host in a connection to initiate the creation of a new 1012 subflow, but it is expected that this will normally be the original 1013 connection initiator (see Section 3.9 for heuristics). 1015 A new subflow is started as a normal TCP SYN/ACK exchange. The Join 1016 Connection (MP_JOIN) MPTCP option is used to identify the connection 1017 to be joined by the new subflow. It uses keying material that was 1018 exchanged in the initial MP_CAPABLE handshake (Section 3.1), and that 1019 handshake also negotiates the crypto algorithm in use for the MP_JOIN 1020 handshake. 1022 This section specifies the behavior of MP_JOIN using the HMAC-SHA256 1023 algorithm. An MP_JOIN option is present in the SYN, SYN/ACK, and ACK 1024 of the three-way handshake, although in each case with a different 1025 format. 1027 In the first MP_JOIN on the SYN packet, illustrated in Figure 5, the 1028 initiator sends a token, random number, and address ID. 1030 The token is used to identify the MPTCP connection and is a 1031 cryptographic hash of the receiver's key, as exchanged in the initial 1032 MP_CAPABLE handshake (Section 3.1). In this specification, the 1033 tokens presented in this option are generated by the SHA-256 ([SHS], 1034 [RFC6234]) algorithm, truncated to the most significant 32 bits. The 1035 token included in the MP_JOIN option is the token that the receiver 1036 of the packet uses to identify this connection; i.e., Host A will 1037 send Token-B (which is generated from Key-B). Note that the hash 1038 generation algorithm can be overridden by the choice of cryptographic 1039 handshake algorithm, as defined in Section 3.1. 1041 The MP_JOIN SYN sends not only the token (which is static for a 1042 connection) but also random numbers (nonces) that are used to prevent 1043 replay attacks on the authentication method. Recommendations for the 1044 generation of random numbers for this purpose are given in [RFC4086]. 1046 The MP_JOIN option includes an "Address ID". This is an identifier 1047 generated by the sender of the option, used to identify the source 1048 address of this packet, even if the IP header has been changed in 1049 transit by a middlebox. The numeric value of this field is generated 1050 by the sender and must map uniquely to a source IP address for the 1051 sending host. The Address ID allows address removal (Section 3.4.2) 1052 without needing to know what the source address at the receiver is, 1053 thus allowing address removal through NATs. The Address ID also 1054 allows correlation between new subflow setup attempts and address 1055 signaling (Section 3.4.1), to prevent setting up duplicate subflows 1056 on the same path, if an MP_JOIN and ADD_ADDR are sent at the same 1057 time. 1059 The Address IDs of the subflow used in the initial SYN exchange of 1060 the first subflow in the connection are implicit, and have the value 1061 zero. A host MUST store the mappings between Address IDs and 1062 addresses both for itself and the remote host. An implementation 1063 will also need to know which local and remote Address IDs are 1064 associated with which established subflows, for when addresses are 1065 removed from a local or remote host. 1067 The MP_JOIN option on packets with the SYN flag set also includes 4 1068 bits of flags, 3 of which are currently reserved and MUST be set to 1069 zero by the sender. The final bit, labeled "B", indicates whether 1070 the sender of this option wishes this subflow to be used as a backup 1071 path (B=1) in the event of failure of other paths, or whether it 1072 wants it to be used as part of the connection immediately. By 1073 setting B=1, the sender of the option is requesting the other host to 1074 only send data on this subflow if there are no available subflows 1075 where B=0. Subflow policy is discussed in more detail in 1076 Section 3.3.8. 1078 1 2 3 1079 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 1080 +---------------+---------------+-------+-----+-+---------------+ 1081 | Kind | Length = 12 |Subtype| |B| Address ID | 1082 +---------------+---------------+-------+-----+-+---------------+ 1083 | Receiver's Token (32 bits) | 1084 +---------------------------------------------------------------+ 1085 | Sender's Random Number (32 bits) | 1086 +---------------------------------------------------------------+ 1088 Figure 5: Join Connection (MP_JOIN) Option (for Initial SYN) 1090 When receiving a SYN with an MP_JOIN option that contains a valid 1091 token for an existing MPTCP connection, the recipient SHOULD respond 1092 with a SYN/ACK also containing an MP_JOIN option containing a random 1093 number and a truncated (leftmost 64 bits) Hash-based Message 1094 Authentication Code (HMAC). This version of the option is shown in 1095 Figure 6. If the token is unknown, or the host wants to refuse 1096 subflow establishment (for example, due to a limit on the number of 1097 subflows it will permit), the receiver will send back a reset (RST) 1098 signal, analogous to an unknown port in TCP, containing a MP_TCPRST 1099 option (Section 3.6) with a "MPTCP specific error" reason code. 1100 Although calculating an HMAC requires cryptographic operations, it is 1101 believed that the 32-bit token in the MP_JOIN SYN gives sufficient 1102 protection against blind state exhaustion attacks; therefore, there 1103 is no need to provide mechanisms to allow a responder to operate 1104 statelessly at the MP_JOIN stage. 1106 An HMAC is sent by both hosts -- by the initiator (Host A) in the 1107 third packet (the ACK) and by the responder (Host B) in the second 1108 packet (the SYN/ACK). Doing the HMAC exchange at this stage allows 1109 both hosts to have first exchanged random data (in the first two SYN 1110 packets) that is used as the "message". This specification defines 1111 that HMAC as defined in [RFC2104] is used, along with the SHA-256 1112 hash algorithm [SHS] (potentially implemented as in [RFC6234]), thus 1113 generating a 160-bit / 20-octet HMAC. Due to option space 1114 limitations, the HMAC included in the SYN/ACK is truncated to the 1115 leftmost 64 bits, but this is acceptable since random numbers are 1116 used; thus, an attacker only has one chance to guess the HMAC 1117 correctly (if the HMAC is incorrect, the TCP connection is closed, so 1118 a new MP_JOIN negotiation with a new random number is required). 1120 The initiator's authentication information is sent in its first ACK 1121 (the third packet of the handshake), as shown in Figure 7. This data 1122 needs to be sent reliably, since it is the only time this HMAC is 1123 sent; therefore, receipt of this packet MUST trigger a regular TCP 1124 ACK in response, and the packet MUST be retransmitted if this ACK is 1125 not received. In other words, sending the ACK/MP_JOIN packet places 1126 the subflow in the PRE_ESTABLISHED state, and it moves to the 1127 ESTABLISHED state only on receipt of an ACK from the receiver. It is 1128 not permitted to send data while in the PRE_ESTABLISHED state. The 1129 reserved bits in this option MUST be set to zero by the sender. 1131 The key for the HMAC algorithm, in the case of the message 1132 transmitted by Host A, will be Key-A followed by Key-B, and in the 1133 case of Host B, Key-B followed by Key-A. These are the keys that 1134 were exchanged in the original MP_CAPABLE handshake. The "message" 1135 for the HMAC algorithm in each case is the concatenations of random 1136 number for each host (denoted by R): for Host A, R-A followed by R-B; 1137 and for Host B, R-B followed by R-A. 1139 1 2 3 1140 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 1141 +---------------+---------------+-------+-----+-+---------------+ 1142 | Kind | Length = 16 |Subtype| |B| Address ID | 1143 +---------------+---------------+-------+-----+-+---------------+ 1144 | | 1145 | Sender's Truncated HMAC (64 bits) | 1146 | | 1147 +---------------------------------------------------------------+ 1148 | Sender's Random Number (32 bits) | 1149 +---------------------------------------------------------------+ 1151 Figure 6: Join Connection (MP_JOIN) Option (for Responding SYN/ACK) 1153 1 2 3 1154 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 1155 +---------------+---------------+-------+-----------------------+ 1156 | Kind | Length = 24 |Subtype| (reserved) | 1157 +---------------+---------------+-------+-----------------------+ 1158 | | 1159 | | 1160 | Sender's HMAC (160 bits) | 1161 | | 1162 | | 1163 +---------------------------------------------------------------+ 1165 Figure 7: Join Connection (MP_JOIN) Option (for Third ACK) 1167 These various MPTCP options fit together to enable authenticated 1168 subflow setup as illustrated in Figure 8. 1170 Host A Host B 1171 ------------------------ ---------- 1172 Address A1 Address A2 Address B1 1173 ---------- ---------- ---------- 1174 | | | 1175 | | SYN + MP_CAPABLE | 1176 |--------------------------------------------->| 1177 |<---------------------------------------------| 1178 | SYN/ACK + MP_CAPABLE(Key-B) | 1179 | | | 1180 | ACK + MP_CAPABLE(Key-A, Key-B) | 1181 |--------------------------------------------->| 1182 | | | 1183 | | SYN + MP_JOIN(Token-B, R-A) | 1184 | |------------------------------->| 1185 | |<-------------------------------| 1186 | | SYN/ACK + MP_JOIN(HMAC-B, R-B) | 1187 | | | 1188 | | ACK + MP_JOIN(HMAC-A) | 1189 | |------------------------------->| 1190 | |<-------------------------------| 1191 | | ACK | 1193 HMAC-A = HMAC(Key=(Key-A+Key-B), Msg=(R-A+R-B)) 1194 HMAC-B = HMAC(Key=(Key-B+Key-A), Msg=(R-B+R-A)) 1196 Figure 8: Example Use of MPTCP Authentication 1198 If the token received at Host B is unknown or local policy prohibits 1199 the acceptance of the new subflow, the recipient MUST respond with a 1200 TCP RST for the subflow. If appropriate, a MP_TCPRST option with a 1201 "Administratively prohibited" reason code (Section 3.6) should be 1202 included. 1204 If the token is accepted at Host B, but the HMAC returned to Host A 1205 does not match the one expected, Host A MUST close the subflow with a 1206 TCP RST. In this, and all following cases of sending a RST in this 1207 section, the sender SHOULD send a MP_TCPRST option (Section 3.6) on 1208 this RST packet with the reason code for a "MPTCP specific error". 1210 If Host B does not receive the expected HMAC, or the MP_JOIN option 1211 is missing from the ACK, it MUST close the subflow with a TCP RST. 1213 If the HMACs are verified as correct, then both hosts have 1214 authenticated each other as being the same peers as existed at the 1215 start of the connection, and they have agreed of which connection 1216 this subflow will become a part. 1218 If the SYN/ACK as received at Host A does not have an MP_JOIN option, 1219 Host A MUST close the subflow with a TCP RST. 1221 This covers all cases of the loss of an MP_JOIN. In more detail, if 1222 MP_JOIN is stripped from the SYN on the path from A to B, and Host B 1223 does not have a listener on the relevant port, it will respond with a 1224 RST in the normal way. If in response to a SYN with an MP_JOIN 1225 option, a SYN/ACK is received without the MP_JOIN option (either 1226 since it was stripped on the return path, or it was stripped on the 1227 outgoing path but Host B responded as if it were a new regular TCP 1228 session), then the subflow is unusable and Host A MUST close it with 1229 a RST. 1231 Note that additional subflows can be created between any pair of 1232 ports (but see Section 3.9 for heuristics); no explicit application- 1233 level accept calls or bind calls are required to open additional 1234 subflows. To associate a new subflow with an existing connection, 1235 the token supplied in the subflow's SYN exchange is used for 1236 demultiplexing. This then binds the 5-tuple of the TCP subflow to 1237 the local token of the connection. A consequence is that it is 1238 possible to allow any port pairs to be used for a connection. 1240 Demultiplexing subflow SYNs MUST be done using the token; this is 1241 unlike traditional TCP, where the destination port is used for 1242 demultiplexing SYN packets. Once a subflow is set up, demultiplexing 1243 packets is done using the 5-tuple, as in traditional TCP. The 1244 5-tuples will be mapped to the local connection identifier (token). 1245 Note that Host A will know its local token for the subflow even 1246 though it is not sent on the wire -- only the responder's token is 1247 sent. 1249 3.3. General MPTCP Operation 1251 This section discusses operation of MPTCP for data transfer. At a 1252 high level, an MPTCP implementation will take one input data stream 1253 from an application, and split it into one or more subflows, with 1254 sufficient control information to allow it to be reassembled and 1255 delivered reliably and in order to the recipient application. The 1256 following subsections define this behavior in detail. 1258 The data sequence mapping and the Data ACK are signaled in the Data 1259 Sequence Signal (DSS) option (Figure 9). Either or both can be 1260 signaled in one DSS, depending on the flags set. The data sequence 1261 mapping defines how the sequence space on the subflow maps to the 1262 connection level, and the Data ACK acknowledges receipt of data at 1263 the connection level. These functions are described in more detail 1264 in the following two subsections. 1266 1 2 3 1267 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 1268 +---------------+---------------+-------+----------------------+ 1269 | Kind | Length |Subtype| (reserved) |F|m|M|a|A| 1270 +---------------+---------------+-------+----------------------+ 1271 | Data ACK (4 or 8 octets, depending on flags) | 1272 +--------------------------------------------------------------+ 1273 | Data sequence number (4 or 8 octets, depending on flags) | 1274 +--------------------------------------------------------------+ 1275 | Subflow Sequence Number (4 octets) | 1276 +-------------------------------+------------------------------+ 1277 | Data-Level Length (2 octets) | Checksum (2 octets) | 1278 +-------------------------------+------------------------------+ 1280 Figure 9: Data Sequence Signal (DSS) Option 1282 The flags, when set, define the contents of this option, as follows: 1284 o A = Data ACK present 1286 o a = Data ACK is 8 octets (if not set, Data ACK is 4 octets) 1288 o M = Data Sequence Number (DSN), Subflow Sequence Number (SSN), 1289 Data-Level Length, and Checksum (if negotiated) present 1291 o m = Data sequence number is 8 octets (if not set, DSN is 4 octets) 1293 The flags 'a' and 'm' only have meaning if the corresponding 'A' or 1294 'M' flags are set; otherwise, they will be ignored. The maximum 1295 length of this option, with all flags set, is 28 octets. 1297 The 'F' flag indicates "DATA_FIN". If present, this means that this 1298 mapping covers the final data from the sender. This is the 1299 connection-level equivalent to the FIN flag in single-path TCP. A 1300 connection is not closed unless there has been a DATA_FIN exchange, 1301 or an implementation-specific, connection-level timeout. The purpose 1302 of the DATA_FIN and the interactions between this flag, the subflow- 1303 level FIN flag, and the data sequence mapping are described in 1304 Section 3.3.3. The remaining reserved bits MUST be set to zero by an 1305 implementation of this specification. 1307 Note that the checksum is only present in this option if the use of 1308 MPTCP checksumming has been negotiated at the MP_CAPABLE handshake 1309 (see Section 3.1). The presence of the checksum can be inferred from 1310 the length of the option. If a checksum is present, but its use had 1311 not been negotiated in the MP_CAPABLE handshake, the checksum field 1312 MUST be ignored. If a checksum is not present when its use has been 1313 negotiated, the receiver MUST close the subflow with a RST as it is 1314 considered broken. This RST SHOULD be accompanied with a MP_TCPRST 1315 option (Section 3.6) with the reason code for a "MPTCP specific 1316 error". 1318 3.3.1. Data Sequence Mapping 1320 The data stream as a whole can be reassembled through the use of the 1321 data sequence mapping components of the DSS option (Figure 9), which 1322 define the mapping from the subflow sequence number to the data 1323 sequence number. This is used by the receiver to ensure in-order 1324 delivery to the application layer. Meanwhile, the subflow-level 1325 sequence numbers (i.e., the regular sequence numbers in the TCP 1326 header) have subflow-only relevance. It is expected (but not 1327 mandated) that SACK [RFC2018] is used at the subflow level to improve 1328 efficiency. 1330 The data sequence mapping specifies a mapping from subflow sequence 1331 space to data sequence space. This is expressed in terms of starting 1332 sequence numbers for the subflow and the data level, and a length of 1333 bytes for which this mapping is valid. This explicit mapping for a 1334 range of data was chosen rather than per-packet signaling to assist 1335 with compatibility with situations where TCP/IP segmentation or 1336 coalescing is undertaken separately from the stack that is generating 1337 the data flow (e.g., through the use of TCP segmentation offloading 1338 on network interface cards, or by middleboxes such as performance 1339 enhancing proxies). It also allows a single mapping to cover many 1340 packets, which may be useful in bulk transfer situations. 1342 A mapping is fixed, in that the subflow sequence number is bound to 1343 the data sequence number after the mapping has been processed. A 1344 sender MUST NOT change this mapping after it has been declared; 1345 however, the same data sequence number can be mapped to by different 1346 subflows for retransmission purposes (see Section 3.3.6). This would 1347 also permit the same data to be sent simultaneously on multiple 1348 subflows for resilience or efficiency purposes, especially in the 1349 case of lossy links. Although the detailed specification of such 1350 operation is outside the scope of this document, an implementation 1351 SHOULD treat the first data that is received at a subflow for the 1352 data sequence space as that which should be delivered to the 1353 application, and any later data for that sequence space should be 1354 ignored. 1356 The data sequence number is specified as an absolute value, whereas 1357 the subflow sequence numbering is relative (the SYN at the start of 1358 the subflow has relative subflow sequence number 0). This is to 1359 allow middleboxes to change the initial sequence number of a subflow, 1360 such as firewalls that undertake Initial Sequence Number (ISN) 1361 randomization. 1363 The data sequence mapping also contains a checksum of the data that 1364 this mapping covers, if use of checksums has been negotiated at the 1365 MP_CAPABLE exchange. Checksums are used to detect if the payload has 1366 been adjusted in any way by a non-MPTCP-aware middlebox. If this 1367 checksum fails, it will trigger a failure of the subflow, or a 1368 fallback to regular TCP, as documented in Section 3.7, since MPTCP 1369 can no longer reliably know the subflow sequence space at the 1370 receiver to build data sequence mappings. 1372 The checksum algorithm used is the standard TCP checksum [RFC0793], 1373 operating over the data covered by this mapping, along with a pseudo- 1374 header as shown in Figure 10. 1376 1 2 3 1377 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 1378 +--------------------------------------------------------------+ 1379 | | 1380 | Data Sequence Number (8 octets) | 1381 | | 1382 +--------------------------------------------------------------+ 1383 | Subflow Sequence Number (4 octets) | 1384 +-------------------------------+------------------------------+ 1385 | Data-Level Length (2 octets) | Zeros (2 octets) | 1386 +-------------------------------+------------------------------+ 1388 Figure 10: Pseudo-Header for DSS Checksum 1390 Note that the data sequence number used in the pseudo-header is 1391 always the 64-bit value, irrespective of what length is used in the 1392 DSS option itself. The standard TCP checksum algorithm has been 1393 chosen since it will be calculated anyway for the TCP subflow, and if 1394 calculated first over the data before adding the pseudo-headers, it 1395 only needs to be calculated once. Furthermore, since the TCP 1396 checksum is additive, the checksum for a DSN_MAP can be constructed 1397 by simply adding together the checksums for the data of each 1398 constituent TCP segment, and adding the checksum for the DSS pseudo- 1399 header. 1401 Note that checksumming relies on the TCP subflow containing 1402 contiguous data; therefore, a TCP subflow MUST NOT use the Urgent 1403 Pointer to interrupt an existing mapping. Further note, however, 1404 that if Urgent data is received on a subflow, it SHOULD be mapped to 1405 the data sequence space and delivered to the application analogous to 1406 Urgent data in regular TCP. 1408 To avoid possible deadlock scenarios, subflow-level processing should 1409 be undertaken separately from that at connection level. Therefore, 1410 even if a mapping does not exist from the subflow space to the data- 1411 level space, the data SHOULD still be ACKed at the subflow (if it is 1412 in-window). This data cannot, however, be acknowledged at the data 1413 level (Section 3.3.2) because its data sequence numbers are unknown. 1414 Implementations MAY hold onto such unmapped data for a short while in 1415 the expectation that a mapping will arrive shortly. Such unmapped 1416 data cannot be counted as being within the connection level receive 1417 window because this is relative to the data sequence numbers, so if 1418 the receiver runs out of memory to hold this data, it will have to be 1419 discarded. If a mapping for that subflow-level sequence space does 1420 not arrive within a receive window of data, that subflow SHOULD be 1421 treated as broken, closed with a RST, and any unmapped data silently 1422 discarded. 1424 Data sequence numbers are always 64-bit quantities, and MUST be 1425 maintained as such in implementations. If a connection is 1426 progressing at a slow rate, so protection against wrapped sequence 1427 numbers is not required, then an implementation MAY include just the 1428 lower 32 bits of the data sequence number in the data sequence 1429 mapping and/or Data ACK as an optimization, and an implementation can 1430 make this choice independently for each packet. An implementation 1431 MUST be able to receive and process both 64-bit or 32-bit sequence 1432 number values, but it is not required that an implementation is able 1433 to send both. 1435 An implementation MUST send the full 64-bit data sequence number if 1436 it is transmitting at a sufficiently high rate that the 32-bit value 1437 could wrap within the Maximum Segment Lifetime (MSL) [RFC7323]. The 1438 lengths of the DSNs used in these values (which may be different) are 1439 declared with flags in the DSS option. Implementations MUST accept a 1440 32-bit DSN and implicitly promote it to a 64-bit quantity by 1441 incrementing the upper 32 bits of sequence number each time the lower 1442 32 bits wrap. A sanity check MUST be implemented to ensure that a 1443 wrap occurs at an expected time (e.g., the sequence number jumps from 1444 a very high number to a very low number) and is not triggered by out- 1445 of-order packets. 1447 As with the standard TCP sequence number, the data sequence number 1448 should not start at zero, but at a random value to make blind session 1449 hijacking harder. This specification requires setting the initial 1450 data sequence number (IDSN) of each host to the least significant 64 1451 bits of the SHA-256 hash of the host's key, as described in 1452 Section 3.1. This is required also in order for the receiver to know 1453 what the expected IDSN is, and thus determine if any initial 1454 connection-level packets are missing; this is particularly relevant 1455 if two subflows start transmitting simultaneously. 1457 A data sequence mapping does not need to be included in every MPTCP 1458 packet, as long as the subflow sequence space in that packet is 1459 covered by a mapping known at the receiver. This can be used to 1460 reduce overhead in cases where the mapping is known in advance; one 1461 such case is when there is a single subflow between the hosts, 1462 another is when segments of data are scheduled in larger than packet- 1463 sized chunks. 1465 An "infinite" mapping can be used to fall back to regular TCP by 1466 mapping the subflow-level data to the connection-level data for the 1467 remainder of the connection (see Section 3.7). This is achieved by 1468 setting the Data-Level Length field of the DSS option to the reserved 1469 value of 0. The checksum, in such a case, will also be set to zero. 1471 3.3.2. Data Acknowledgments 1473 To provide full end-to-end resilience, MPTCP provides a connection- 1474 level acknowledgment, to act as a cumulative ACK for the connection 1475 as a whole. This is the "Data ACK" field of the DSS option 1476 (Figure 9). The Data ACK is analogous to the behavior of the 1477 standard TCP cumulative ACK -- indicating how much data has been 1478 successfully received (with no holes). This is in comparison to the 1479 subflow-level ACK, which acts analogous to TCP SACK, given that there 1480 may still be holes in the data stream at the connection level. The 1481 Data ACK specifies the next data sequence number it expects to 1482 receive. 1484 The Data ACK, as for the DSN, can be sent as the full 64-bit value, 1485 or as the lower 32 bits. If data is received with a 64-bit DSN, it 1486 MUST be acknowledged with a 64-bit Data ACK. If the DSN received is 1487 32 bits, it is valid for the implementation to choose whether to send 1488 a 32-bit or 64-bit Data ACK. 1490 The Data ACK proves that the data, and all required MPTCP signaling, 1491 has been received and accepted by the remote end. One key use of the 1492 Data ACK signal is that it is used to indicate the left edge of the 1493 advertised receive window. As explained in Section 3.3.4, the 1494 receive window is shared by all subflows and is relative to the Data 1495 ACK. Because of this, an implementation MUST NOT use the RCV.WND 1496 field of a TCP segment at the connection level if it does not also 1497 carry a DSS option with a Data ACK field. Furthermore, separating 1498 the connection-level acknowledgments from the subflow level allows 1499 processing to be done separately, and a receiver has the freedom to 1500 drop segments after acknowledgment at the subflow level, for example, 1501 due to memory constraints when many segments arrive out of order. 1503 An MPTCP sender MUST NOT free data from the send buffer until it has 1504 been acknowledged by both a Data ACK received on any subflow and at 1505 the subflow level by all subflows on which the data was sent. The 1506 former condition ensures liveness of the connection and the latter 1507 condition ensures liveness and self-consistence of a subflow when 1508 data needs to be retransmitted. Note, however, that if some data 1509 needs to be retransmitted multiple times over a subflow, there is a 1510 risk of blocking the sending window. In this case, the MPTCP sender 1511 can decide to terminate the subflow that is behaving badly by sending 1512 a RST, using an appropriate MP_TCPRST (Section 3.6) error code. 1514 The Data ACK MAY be included in all segments; however, optimizations 1515 SHOULD be considered in more advanced implementations, where the Data 1516 ACK is present in segments only when the Data ACK value advances, and 1517 this behavior MUST be treated as valid. This behavior ensures the 1518 sender buffer is freed, while reducing overhead when the data 1519 transfer is unidirectional. 1521 3.3.3. Closing a Connection 1523 In regular TCP, a FIN announces the receiver that the sender has no 1524 more data to send. In order to allow subflows to operate 1525 independently and to keep the appearance of TCP over the wire, a FIN 1526 in MPTCP only affects the subflow on which it is sent. This allows 1527 nodes to exercise considerable freedom over which paths are in use at 1528 any one time. The semantics of a FIN remain as for regular TCP; 1529 i.e., it is not until both sides have ACKed each other's FINs that 1530 the subflow is fully closed. 1532 When an application calls close() on a socket, this indicates that it 1533 has no more data to send; for regular TCP, this would result in a FIN 1534 on the connection. For MPTCP, an equivalent mechanism is needed, and 1535 this is referred to as the DATA_FIN. 1537 A DATA_FIN is an indication that the sender has no more data to send, 1538 and as such can be used to verify that all data has been successfully 1539 received. A DATA_FIN, as with the FIN on a regular TCP connection, 1540 is a unidirectional signal. 1542 The DATA_FIN is signaled by setting the 'F' flag in the Data Sequence 1543 Signal option (Figure 9) to 1. A DATA_FIN occupies 1 octet (the 1544 final octet) of the connection-level sequence space. Note that the 1545 DATA_FIN is included in the Data-Level Length, but not at the subflow 1546 level: for example, a segment with DSN 80, and Data-Level Length 11, 1547 with DATA_FIN set, would map 10 octets from the subflow into data 1548 sequence space 80-89, the DATA_FIN is DSN 90; therefore, this segment 1549 including DATA_FIN would be acknowledged with a DATA_ACK of 91. 1551 Note that when the DATA_FIN is not attached to a TCP segment 1552 containing data, the Data Sequence Signal MUST have a subflow 1553 sequence number of 0, a Data-Level Length of 1, and the data sequence 1554 number that corresponds with the DATA_FIN itself. The checksum in 1555 this case will only cover the pseudo-header. 1557 A DATA_FIN has the semantics and behavior as a regular TCP FIN, but 1558 at the connection level. Notably, it is only DATA_ACKed once all 1559 data has been successfully received at the connection level. Note, 1560 therefore, that a DATA_FIN is decoupled from a subflow FIN. It is 1561 only permissible to combine these signals on one subflow if there is 1562 no data outstanding on other subflows. Otherwise, it may be 1563 necessary to retransmit data on different subflows. Essentially, a 1564 host MUST NOT close all functioning subflows unless it is safe to do 1565 so, i.e., until all outstanding data has been DATA_ACKed, or until 1566 the segment with the DATA_FIN flag set is the only outstanding 1567 segment. 1569 Once a DATA_FIN has been acknowledged, all remaining subflows MUST be 1570 closed with standard FIN exchanges. Both hosts SHOULD send FINs on 1571 all subflows, as a courtesy to allow middleboxes to clean up state 1572 even if an individual subflow has failed. It is also encouraged to 1573 reduce the timeouts (Maximum Segment Lifetime) on subflows at end 1574 hosts after receiving a DATA_FIN. In particular, any subflows where 1575 there is still outstanding data queued (which has been retransmitted 1576 on other subflows in order to get the DATA_FIN acknowledged) MAY be 1577 closed with a RST with MP_TCPRST (Section 3.6) error code for "too 1578 much outstanding data". 1580 A connection is considered closed once both hosts' DATA_FINs have 1581 been acknowledged by DATA_ACKs. 1583 As specified above, a standard TCP FIN on an individual subflow only 1584 shuts down the subflow on which it was sent. If all subflows have 1585 been closed with a FIN exchange, but no DATA_FIN has been received 1586 and acknowledged, the MPTCP connection is treated as closed only 1587 after a timeout. This implies that an implementation will have 1588 TIME_WAIT states at both the subflow and connection levels (see 1589 Appendix D). This permits "break-before-make" scenarios where 1590 connectivity is lost on all subflows before a new one can be re- 1591 established. 1593 3.3.4. Receiver Considerations 1595 Regular TCP advertises a receive window in each packet, telling the 1596 sender how much data the receiver is willing to accept past the 1597 cumulative ack. The receive window is used to implement flow 1598 control, throttling down fast senders when receivers cannot keep up. 1600 MPTCP also uses a unique receive window, shared between the subflows. 1601 The idea is to allow any subflow to send data as long as the receiver 1602 is willing to accept it. The alternative, maintaining per subflow 1603 receive windows, could end up stalling some subflows while others 1604 would not use up their window. 1606 The receive window is relative to the DATA_ACK. As in TCP, a 1607 receiver MUST NOT shrink the right edge of the receive window (i.e., 1608 DATA_ACK + receive window). The receiver will use the data sequence 1609 number to tell if a packet should be accepted at the connection 1610 level. 1612 When deciding to accept packets at subflow level, regular TCP checks 1613 the sequence number in the packet against the allowed receive window. 1614 With multipath, such a check is done using only the connection-level 1615 window. A sanity check SHOULD be performed at subflow level to 1616 ensure that the subflow and mapped sequence numbers meet the 1617 following test: SSN - SUBFLOW_ACK <= DSN - DATA_ACK, where SSN is the 1618 subflow sequence number of the received packet and SUBFLOW_ACK is the 1619 RCV.NXT (next expected sequence number) of the subflow (with the 1620 equivalent connection-level definitions for DSN and DATA_ACK). 1622 In regular TCP, once a segment is deemed in-window, it is put either 1623 in the in-order receive queue or in the out-of-order queue. In 1624 Multipath TCP, the same happens but at the connection level: a 1625 segment is placed in the connection level in-order or out-of-order 1626 queue if it is in-window at both connection and subflow levels. The 1627 stack still has to remember, for each subflow, which segments were 1628 received successfully so that it can ACK them at subflow level 1629 appropriately. Typically, this will be implemented by keeping per 1630 subflow out-of-order queues (containing only message headers, not the 1631 payloads) and remembering the value of the cumulative ACK. 1633 It is important for implementers to understand how large a receiver 1634 buffer is appropriate. The lower bound for full network utilization 1635 is the maximum bandwidth-delay product of any one of the paths. 1636 However, this might be insufficient when a packet is lost on a slower 1637 subflow and needs to be retransmitted (see Section 3.3.6). A tight 1638 upper bound would be the maximum round-trip time (RTT) of any path 1639 multiplied by the total bandwidth available across all paths. This 1640 permits all subflows to continue at full speed while a packet is 1641 fast-retransmitted on the maximum RTT path. Even this might be 1642 insufficient to maintain full performance in the event of a 1643 retransmit timeout on the maximum RTT path. It is for future study 1644 to determine the relationship between retransmission strategies and 1645 receive buffer sizing. 1647 3.3.5. Sender Considerations 1649 The sender remembers receiver window advertisements from the 1650 receiver. It should only update its local receive window values when 1651 the largest sequence number allowed (i.e., DATA_ACK + receive window) 1652 increases, on the receipt of a DATA_ACK. This is important to allow 1653 using paths with different RTTs, and thus different feedback loops. 1655 MPTCP uses a single receive window across all subflows, and if the 1656 receive window was guaranteed to be unchanged end-to-end, a host 1657 could always read the most recent receive window value. However, 1658 some classes of middleboxes may alter the TCP-level receive window. 1659 Typically, these will shrink the offered window, although for short 1660 periods of time it may be possible for the window to be larger 1661 (however, note that this would not continue for long periods since 1662 ultimately the middlebox must keep up with delivering data to the 1663 receiver). Therefore, if receive window sizes differ on multiple 1664 subflows, when sending data MPTCP SHOULD take the largest of the most 1665 recent window sizes as the one to use in calculations. This rule is 1666 implicit in the requirement not to reduce the right edge of the 1667 window. 1669 The sender MUST also remember the receive windows advertised by each 1670 subflow. The allowed window for subflow i is (ack_i, ack_i + 1671 rcv_wnd_i), where ack_i is the subflow-level cumulative ACK of 1672 subflow i. This ensures data will not be sent to a middlebox unless 1673 there is enough buffering for the data. 1675 Putting the two rules together, we get the following: a sender is 1676 allowed to send data segments with data-level sequence numbers 1677 between (DATA_ACK, DATA_ACK + receive_window). Each of these 1678 segments will be mapped onto subflows, as long as subflow sequence 1679 numbers are in the allowed windows for those subflows. Note that 1680 subflow sequence numbers do not generally affect flow control if the 1681 same receive window is advertised across all subflows. They will 1682 perform flow control for those subflows with a smaller advertised 1683 receive window. 1685 The send buffer MUST, at a minimum, be as big as the receive buffer, 1686 to enable the sender to reach maximum throughput. 1688 3.3.6. Reliability and Retransmissions 1690 The data sequence mapping allows senders to resend data with the same 1691 data sequence number on a different subflow. When doing this, a host 1692 MUST still retransmit the original data on the original subflow, in 1693 order to preserve the subflow integrity (middleboxes could replay old 1694 data, and/or could reject holes in subflows), and a receiver will 1695 ignore these retransmissions. While this is clearly suboptimal, for 1696 compatibility reasons this is sensible behavior. Optimizations could 1697 be negotiated in future versions of this protocol. Note also that 1698 this property would also permit a sender to always send the same 1699 data, with the same data sequence number, on multiple subflows, if it 1700 so desired for reliability reasons. 1702 This protocol specification does not mandate any mechanisms for 1703 handling retransmissions, and much will be dependent upon local 1704 policy (as discussed in Section 3.3.8). One can imagine aggressive 1705 connection-level retransmissions policies where every packet lost at 1706 subflow level is retransmitted on a different subflow (hence, wasting 1707 bandwidth but possibly reducing application-to-application delays), 1708 or conservative retransmission policies where connection-level 1709 retransmits are only used after a few subflow-level retransmission 1710 timeouts occur. 1712 It is envisaged that a standard connection-level retransmission 1713 mechanism would be implemented around a connection-level data queue: 1714 all segments that haven't been DATA_ACKed are stored. A timer is set 1715 when the head of the connection-level is ACKed at subflow level but 1716 its corresponding data is not ACKed at data level. This timer will 1717 guard against failures in retransmission by middleboxes that 1718 proactively ACK data. 1720 The sender MUST keep data in its send buffer as long as the data has 1721 not been acknowledged at both connection level and on all subflows on 1722 which it has been sent. In this way, the sender can always 1723 retransmit the data if needed, on the same subflow or on a different 1724 one. A special case is when a subflow fails: the sender will 1725 typically resend the data on other working subflows after a timeout, 1726 and will keep trying to retransmit the data on the failed subflow 1727 too. The sender will declare the subflow failed after a predefined 1728 upper bound on retransmissions is reached (which MAY be lower than 1729 the usual TCP limits of the Maximum Segment Life), or on the receipt 1730 of an ICMP error, and only then delete the outstanding data segments. 1732 Multiple retransmissions are triggers that will indicate that a 1733 subflow performs badly and could lead to a host resetting the subflow 1734 with a RST. However, additional research is required to understand 1735 the heuristics of how and when to reset underperforming subflows. 1736 For example, a highly asymmetric path may be misdiagnosed as 1737 underperforming. A RST for this purpose SHOULD be accompanied with 1738 an "Unacceptable performance" MP_TCPRST option (Section 3.6). 1740 3.3.7. Congestion Control Considerations 1742 Different subflows in an MPTCP connection have different congestion 1743 windows. To achieve fairness at bottlenecks and resource pooling, it 1744 is necessary to couple the congestion windows in use on each subflow, 1745 in order to push most traffic to uncongested links. One algorithm 1746 for achieving this is presented in [RFC6356]; the algorithm does not 1747 achieve perfect resource pooling but is "safe" in that it is readily 1748 deployable in the current Internet. By this, we mean that it does 1749 not take up more capacity on any one path than if it was a single 1750 path flow using only that route, so this ensures fair coexistence 1751 with single-path TCP at shared bottlenecks. 1753 It is foreseeable that different congestion controllers will be 1754 implemented for MPTCP, each aiming to achieve different properties in 1755 the resource pooling/fairness/stability design space, as well as 1756 those for achieving different properties in quality of service, 1757 reliability, and resilience. 1759 Regardless of the algorithm used, the design of the MPTCP protocol 1760 aims to provide the congestion control implementations sufficient 1761 information to take the right decisions; this information includes, 1762 for each subflow, which packets were lost and when. 1764 3.3.8. Subflow Policy 1766 Within a local MPTCP implementation, a host may use any local policy 1767 it wishes to decide how to share the traffic to be sent over the 1768 available paths. 1770 In the typical use case, where the goal is to maximize throughput, 1771 all available paths will be used simultaneously for data transfer, 1772 using coupled congestion control as described in [RFC6356]. It is 1773 expected, however, that other use cases will appear. 1775 For instance, a possibility is an 'all-or-nothing' approach, i.e., 1776 have a second path ready for use in the event of failure of the first 1777 path, but alternatives could include entirely saturating one path 1778 before using an additional path (the 'overflow' case). Such choices 1779 would be most likely based on the monetary cost of links, but may 1780 also be based on properties such as the delay or jitter of links, 1781 where stability (of delay or bandwidth) is more important than 1782 throughput. Application requirements such as these are discussed in 1783 detail in [RFC6897]. 1785 The ability to make effective choices at the sender requires full 1786 knowledge of the path "cost", which is unlikely to be the case. It 1787 would be desirable for a receiver to be able to signal their own 1788 preferences for paths, since they will often be the multihomed party, 1789 and may have to pay for metered incoming bandwidth. 1791 Whilst fine-grained control may be the most powerful solution, that 1792 would require some mechanism such as overloading the Explicit 1793 Congestion Notification (ECN) signal [RFC3168], which is undesirable, 1794 and it is felt that there would not be sufficient benefit to justify 1795 an entirely new signal. Therefore, the MP_JOIN option (see 1796 Section 3.2) contains the 'B' bit, which allows a host to indicate to 1797 its peer that this path should be treated as a backup path to use 1798 only in the event of failure of other working subflows (i.e., a 1799 subflow where the receiver has indicated B=1 SHOULD NOT be used to 1800 send data unless there are no usable subflows where B=0). 1802 In the event that the available set of paths changes, a host may wish 1803 to signal a change in priority of subflows to the peer (e.g., a 1804 subflow that was previously set as backup should now take priority 1805 over all remaining subflows). Therefore, the MP_PRIO option, shown 1806 in Figure 11, can be used to change the 'B' flag of the subflow on 1807 which it is sent. 1809 Another use of the MP_PRIO option is to set the 'B' flag on a subflow 1810 to cleanly retire its use before closing it and removing it with 1811 REMOVE_ADDR Section 3.4.2, for example to support make-before-break 1812 session continuity, where new subflows are added before the 1813 previously used ones are closed. 1815 1 2 3 1816 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 1817 +---------------+---------------+-------+-----+-+ 1818 | Kind | Length |Subtype| |B| 1819 +---------------+---------------+-------+-----+-+ 1821 Figure 11: Change Subflow Priority (MP_PRIO) Option 1823 It should be noted that the backup flag is a request from a data 1824 receiver to a data sender only, and the data sender SHOULD adhere to 1825 these requests. A host cannot assume that the data sender will do 1826 so, however, since local policies -- or technical difficulties -- may 1827 override MP_PRIO requests. Note also that this signal applies to a 1828 single direction, and so the sender of this option could choose to 1829 continue using the subflow to send data even if it has signaled B=1 1830 to the other host. 1832 3.4. Address Knowledge Exchange (Path Management) 1834 We use the term "path management" to refer to the exchange of 1835 information about additional paths between hosts, which in this 1836 design is managed by multiple addresses at hosts. For more detail of 1837 the architectural thinking behind this design, see the MPTCP 1838 Architecture document [RFC6182]. 1840 This design makes use of two methods of sharing such information, and 1841 both can be used on a connection. The first is the direct setup of 1842 new subflows, already described in Section 3.2, where the initiator 1843 has an additional address. The second method, described in the 1844 following subsections, signals addresses explicitly to the other host 1845 to allow it to initiate new subflows. The two mechanisms are 1846 complementary: the first is implicit and simple, while the explicit 1847 is more complex but is more robust. Together, the mechanisms allow 1848 addresses to change in flight (and thus support operation through 1849 NATs, since the source address need not be known), and also allow the 1850 signaling of previously unknown addresses, and of addresses belonging 1851 to other address families (e.g., both IPv4 and IPv6). 1853 Here is an example of typical operation of the protocol: 1855 o An MPTCP connection is initially set up between address/port A1 of 1856 Host A and address/port B1 of Host B. If Host A is multihomed and 1857 multiaddressed, it can start an additional subflow from its 1858 address A2 to B1, by sending a SYN with a Join option from A2 to 1859 B1, using B's previously declared token for this connection. 1860 Alternatively, if B is multihomed, it can try to set up a new 1861 subflow from B2 to A1, using A's previously declared token. In 1862 either case, the SYN will be sent to the port already in use for 1863 the original subflow on the receiving host. 1865 o Simultaneously (or after a timeout), an ADD_ADDR option 1866 (Section 3.4.1) is sent on an existing subflow, informing the 1867 receiver of the sender's alternative address(es). The recipient 1868 can use this information to open a new subflow to the sender's 1869 additional address. In our example, A will send ADD_ADDR option 1870 informing B of address/port A2. The mix of using the SYN-based 1871 option and the ADD_ADDR option, including timeouts, is 1872 implementation specific and can be tailored to agree with local 1873 policy. 1875 o If subflow A2-B1 is successfully set up, Host B can use the 1876 Address ID in the Join option to correlate this with the ADD_ADDR 1877 option that will also arrive on an existing subflow; now B knows 1878 not to open A2-B1, ignoring the ADD_ADDR. Otherwise, if B has not 1879 received the A2-B1 MP_JOIN SYN but received the ADD_ADDR, it can 1880 try to initiate a new subflow from one or more of its addresses to 1881 address A2. This permits new sessions to be opened if one host is 1882 behind a NAT. 1884 Other ways of using the two signaling mechanisms are possible; for 1885 instance, signaling addresses in other address families can only be 1886 done explicitly using the Add Address option. 1888 3.4.1. Address Advertisement 1890 The Add Address (ADD_ADDR) MPTCP option announces additional 1891 addresses (and optionally, ports) on which a host can be reached 1892 (Figure 12). This option can be used at any time during a 1893 connection, depending on when the sender wishes to enable multiple 1894 paths and/or when paths become available. As with all MPTCP signals, 1895 the receiver MUST undertake standard TCP validity checks, e.g. 1896 [RFC5961], before acting upon it. 1898 Every address has an Address ID that can be used for uniquely 1899 identifying the address within a connection for address removal. The 1900 Address ID is also used to identify MP_JOIN options (see Section 3.2) 1901 relating to the same address, even when address translators are in 1902 use. The Address ID MUST uniquely identify the address for the 1903 sender of the option (within the scope of the connection), but the 1904 mechanism for allocating such IDs is implementation specific. 1906 All address IDs learned via either MP_JOIN or ADD_ADDR SHOULD be 1907 stored by the receiver in a data structure that gathers all the 1908 Address ID to address mappings for a connection (identified by a 1909 token pair). In this way, there is a stored mapping between Address 1910 ID, observed source address, and token pair for future processing of 1911 control information for a connection. Note that an implementation 1912 MAY discard incoming address advertisements at will, for example, for 1913 avoiding updating mapping state, or because advertised addresses are 1914 of no use to it (for example, IPv6 addresses when it has IPv4 only). 1915 Therefore, a host MUST treat address advertisements as soft state, 1916 and it MAY choose to refresh advertisements periodically. 1918 This option is shown in Figure 12. The illustration is sized for 1919 IPv4 addresses. For IPv6, the length of the address will be 16 1920 octets (instead of 4). 1922 The 2 octets that specify the TCP port number to use are optional and 1923 their presence can be inferred from the length of the option. 1924 Although it is expected that the majority of use cases will use the 1925 same port pairs as used for the initial subflow (e.g., port 80 1926 remains port 80 on all subflows, as does the ephemeral port at the 1927 client), there may be cases (such as port-based load balancing) where 1928 the explicit specification of a different port is required. If no 1929 port is specified, MPTCP SHOULD attempt to connect to the specified 1930 address on the same port as is already in use by the subflow on which 1931 the ADD_ADDR signal was sent; this is discussed in more detail in 1932 Section 3.9. 1934 The Truncated HMAC present in this Option is the rightmost 64 bits of 1935 an HMAC, negotiated and calculated in the same way as for MP_JOIN as 1936 described in Section 3.2. For this specification of MPTCP, as there 1937 is only one hash algorithm option specified, this will be HMAC as 1938 defined in [RFC2104], using the SHA-256 hash algorithm [SHS], 1939 implemented as in [RFC6234]. In the same way as for MP_JOIN, the key 1940 for the HMAC algorithm, in the case of the message transmitted by 1941 Host A, will be Key-A followed by Key-B, and in the case of Host B, 1942 Key-B followed by Key-A. These are the keys that were exchanged in 1943 the original MP_CAPABLE handshake. The message for the HMAC is the 1944 Address ID, IP Address, and Port which precede the HMAC in the 1945 ADD_ADDR option. If the port is not present in the ADD_ADDR option, 1946 the HMAC message will nevertheless include two octets of value zero. 1947 The rationale for the HMAC is to prevent unauthorized entities from 1948 injecting ADD_ADDR signals in an attempt to hijack a connection. 1949 Note that additionally the presence of this HMAC prevents the address 1950 being changed in flight unless the key is known by an intermediary. 1951 If a host receives an ADD_ADDR option for which it cannot validate 1952 the HMAC, it SHOULD silently ignore the option. 1954 A set of four flags are present after the subtype and before the 1955 Address ID. Only the rightmost bit - labelled 'E' - is assigned in 1956 this specification. The other bits are currently unassigned and MUST 1957 be set to zero by a sender and MUST be ignored by the receiver. 1959 The 'E' flag exists to provide reliability for this option. Because 1960 this option will often be sent on pure ACKs, there is no guarantee of 1961 reliability. Therefore, a receiver receiving a fresh ADD_ADDR option 1962 (where E=0), will send the same option back to the sender, but not 1963 including the HMAC, and with E=1, to indicate receipt. The lack of 1964 this echo can be used by the initial ADD_ADDR sender to retransmit 1965 the ADD_ADDR according to local policy. 1967 1 2 3 1968 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 1969 +---------------+---------------+-------+-------+---------------+ 1970 | Kind | Length |Subtype|(rsv)|E| Address ID | 1971 +---------------+---------------+-------+-------+---------------+ 1972 | Address (IPv4 - 4 octets / IPv6 - 16 octets) | 1973 +-------------------------------+-------------------------------+ 1974 | Port (2 octets, optional) | | 1975 +-------------------------------+ | 1976 | Truncated HMAC (8 octets, if E=0) | 1977 | +-------------------------------+ 1978 | | 1979 +-------------------------------+ 1981 Figure 12: Add Address (ADD_ADDR) Option 1983 Due to the proliferation of NATs, it is reasonably likely that one 1984 host may attempt to advertise private addresses [RFC1918]. It is not 1985 desirable to prohibit this, since there may be cases where both hosts 1986 have additional interfaces on the same private network, and a host 1987 MAY want to advertise such addresses. The MP_JOIN handshake to 1988 create a new subflow (Section 3.2) provides mechanisms to minimize 1989 security risks. The MP_JOIN message contains a 32-bit token that 1990 uniquely identifies the connection to the receiving host. If the 1991 token is unknown, the host will return with a RST. In the unlikely 1992 event that the token is valid at the receiving host, subflow setup 1993 will continue, but the HMAC exchange must occur for authentication. 1994 This will fail, and will provide sufficient protection against two 1995 unconnected hosts accidentally setting up a new subflow upon the 1996 signal of a private address. Further security considerations around 1997 the issue of ADD_ADDR messages that accidentally misdirect, or 1998 maliciously direct, new MP_JOIN attempts are discussed in Section 5. 2000 A host that receives an ADD_ADDR but finds a connection set up to 2001 that IP address and port number is unsuccessful SHOULD NOT perform 2002 further connection attempts to this address/port combination for this 2003 connection. A sender that wants to trigger a new incoming connection 2004 attempt on a previously advertised address/port combination can 2005 therefore refresh ADD_ADDR information by sending the option again. 2007 A host can therefore send an ADD_ADDR message with an already 2008 assigned Address ID, but the Address MUST be the same as previously 2009 assigned to this Address ID. A new ADD_ADDR may have the same, or 2010 different, port number. If the port number is different, the 2011 receiving host SHOULD try to set up a new subflow to this new 2012 address/port combination. 2014 A host wishing to replace an existing Address ID MUST first remove 2015 the existing one (Section 3.4.2). 2017 During normal MPTCP operation, it is unlikely that there will be 2018 sufficient TCP option space for ADD_ADDR to be included along with 2019 those for data sequence numbering (Section 3.3.1). Therefore, it is 2020 expected that an MPTCP implementation will send the ADD_ADDR option 2021 on separate ACKs. As discussed earlier, however, an MPTCP 2022 implementation MUST NOT treat duplicate ACKs with any MPTCP option, 2023 with the exception of the DSS option, as indications of congestion 2024 [RFC5681], and an MPTCP implementation SHOULD NOT send more than two 2025 duplicate ACKs in a row for signaling purposes. 2027 3.4.2. Remove Address 2029 If, during the lifetime of an MPTCP connection, a previously 2030 announced address becomes invalid (e.g., if the interface 2031 disappears), the affected host SHOULD announce this so that the peer 2032 can remove subflows related to this address. A host MAY also choose 2033 to announce that a valid IP address should not be used any longer, 2034 for example for make-before-break session continuity. 2036 This is achieved through the Remove Address (REMOVE_ADDR) option 2037 (Figure 13), which will remove a previously added address (or list of 2038 addresses) from a connection and terminate any subflows currently 2039 using that address. 2041 For security purposes, if a host receives a REMOVE_ADDR option, it 2042 must ensure the affected path(s) are no longer in use before it 2043 instigates closure. The receipt of REMOVE_ADDR SHOULD first trigger 2044 the sending of a TCP keepalive [RFC1122] on the path, and if a 2045 response is received the path SHOULD NOT be removed. If the path is 2046 found to still be alive, the receiving host SHOULD no longer use the 2047 specified address for future connections, but it is the 2048 responsibility of the host which sent the REMOVE_ADDR to shut down 2049 the subflow. The requesting host MAY also use MP_PRIO 2050 (Section 3.3.8) to request a path is no longer used, before removal. 2051 Typical TCP validity tests on the subflow (e.g., ensuring sequence 2052 and ACK numbers are correct) MUST also be undertaken. An 2053 implementation can use indications of these test failures as part of 2054 intrusion detection or error logging. 2056 The sending and receipt (if no keepalive response was received) of 2057 this message SHOULD trigger the sending of RSTs by both hosts on the 2058 affected subflow(s) (if possible), as a courtesy to cleaning up 2059 middlebox state, before cleaning up any local state. 2061 Address removal is undertaken by ID, so as to permit the use of NATs 2062 and other middleboxes that rewrite source addresses. If there is no 2063 address at the requested ID, the receiver will silently ignore the 2064 request. 2066 A subflow that is still functioning MUST be closed with a FIN 2067 exchange as in regular TCP, rather than using this option. For more 2068 information, see Section 3.3.3. 2070 1 2 3 2071 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 2072 +---------------+---------------+-------+-------+---------------+ 2073 | Kind | Length = 3+n |Subtype|(resvd)| Address ID | ... 2074 +---------------+---------------+-------+-------+---------------+ 2075 (followed by n-1 Address IDs, if required) 2077 Figure 13: Remove Address (REMOVE_ADDR) Option 2079 3.5. Fast Close 2081 Regular TCP has the means of sending a reset (RST) signal to abruptly 2082 close a connection. With MPTCP, a regular RST only has the scope of 2083 the subflow and will only close the concerned subflow but not affect 2084 the remaining subflows. MPTCP's connection will stay alive at the 2085 data level, in order to permit break-before-make handover between 2086 subflows. It is therefore necessary to provide an MPTCP-level 2087 "reset" to allow the abrupt closure of the whole MPTCP connection, 2088 and this is the MP_FASTCLOSE option. 2090 MP_FASTCLOSE is used to indicate to the peer that the connection will 2091 be abruptly closed and no data will be accepted anymore. The reasons 2092 for triggering an MP_FASTCLOSE are implementation specific. Regular 2093 TCP does not allow sending a RST while the connection is in a 2094 synchronized state [RFC0793]. Nevertheless, implementations allow 2095 the sending of a RST in this state, if, for example, the operating 2096 system is running out of resources. In these cases, MPTCP should 2097 send the MP_FASTCLOSE. This option is illustrated in Figure 14. 2099 1 2 3 2100 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 2101 +---------------+---------------+-------+-----------------------+ 2102 | Kind | Length |Subtype| (reserved) | 2103 +---------------+---------------+-------+-----------------------+ 2104 | Option Receiver's Key | 2105 | (64 bits) | 2106 | | 2107 +---------------------------------------------------------------+ 2109 Figure 14: Fast Close (MP_FASTCLOSE) Option 2111 If Host A wants to force the closure of an MPTCP connection, it has 2112 two different options: 2114 o Option A (ACK) : Host A sends an ACK containing the MP_FASTCLOSE 2115 option on one subflow, containing the key of Host B as declared in 2116 the initial connection handshake. On all the other subflows, Host 2117 A sends a regular TCP RST to close these subflows, and tears them 2118 down. Host A now enters FASTCLOSE_WAIT state. 2120 o Option R (RST) : Host A sends a RST containing the MP_FASTCLOSE 2121 option on all subflows, containing the key of Host B as declared 2122 in the initial connection handshake. Host A can tear the subflows 2123 and the connection down immediately. 2125 If host A decides to force the closure by using Option A and sending 2126 an ACK with the MP_FASTCLOSE option, the connection shall proceed as 2127 follows: 2129 o Upon receipt of an ACK with MP_FASTCLOSE by Host B, containing the 2130 valid key, Host B answers on the same subflow with a TCP RST and 2131 tears down all subflows also through sending TCP RST signals. 2132 Host B can now close the whole MPTCP connection (it transitions 2133 directly to CLOSED state). 2135 o As soon as Host A has received the TCP RST on the remaining 2136 subflow, it can close this subflow and tear down the whole 2137 connection (transition from FASTCLOSE_WAIT to CLOSED states). If 2138 Host A receives an MP_FASTCLOSE instead of a TCP RST, both hosts 2139 attempted fast closure simultaneously. Host A should reply with a 2140 TCP RST and tear down the connection. 2142 o If Host A does not receive a TCP RST in reply to its MP_FASTCLOSE 2143 after one retransmission timeout (RTO) (the RTO of the subflow 2144 where the MP_FASTCLOSE has been sent), it SHOULD retransmit the 2145 MP_FASTCLOSE. The number of retransmissions SHOULD be limited to 2146 avoid this connection from being retained for a long time, but 2147 this limit is implementation specific. A RECOMMENDED number is 3. 2148 If no TCP RST is received in response, Host A SHOULD send a TCP 2149 RST with the MP_FASTCLOSE option itself when it releases state in 2150 order to clear any remaining state at middleboxes. 2152 If however host A decides to force the closure by using Option R and 2153 sending a RST with the MP_FASTCLOSE option, Host B will act as 2154 follows: Upon receipt of a RST with MP_FASTCLOSE, containing the 2155 valid key, Host B tears down all subflows by sending a TCP RST. Host 2156 B can now close the whole MPTCP connection (it transitions directly 2157 to CLOSED state). 2159 3.6. Subflow Reset 2161 An implementation of MPTCP may also need to send a regular TCP RST to 2162 force the closure of a subflow. A host sends a TCP RST in order to 2163 close a subflow or reject an attempt to open a subflow (MP_JOIN). In 2164 order to inform the receiving host why a subflow is being closed or 2165 rejected, the TCP RST packet MAY include the MP_TCPRST Option. The 2166 host MAY use this information to decide, for example, whether it 2167 tries to re-establish the subflow immediately, later, or never. 2169 1 2 3 2170 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 2171 +---------------+---------------+-------+-----------------------+ 2172 | Kind | Length |Subtype|U|V|W|T| Reason | 2173 +---------------+---------------+-------+-----------------------+ 2175 Figure 15: TCP RST Reason (MP_TCPRST) Option 2177 The MP_TCPRST option contains a reason code that allows the sender of 2178 the option to provide more information about the reason for the 2179 termination of the subflow. Using 12 bits of option space, the first 2180 four bits are reserved for flags (only one of which is currently 2181 defined), and the remaining octet is used to express a reason code 2182 for this subflow termination, from which a receiver MAY infer 2183 information about the usability of this path. 2185 The "T" flag is used by the sender to indicate whether the error 2186 condition that is reported is Transient (T bit set to 1) or Permanent 2187 (T bit set to 0). If the error condition is considered to be 2188 Transient by the sender of the RST segment, the recipient of this 2189 segment MAY try to reestablish a subflow for this connection over the 2190 failed path. The time at which a receiver may try to re-establish 2191 this is implementation-specific, but SHOULD take into account the 2192 properties of the failure defined by the following reason code. If 2193 the error condition is considered to be permanent, the receiver of 2194 the RST segment SHOULD NOT try to reestablish a subflow for this 2195 connection over this path. The "U", "V" and "W" flags are not 2196 defined by this specification and are reserved for future use. An 2197 implementation of this specification MUST set these flags to 0, and a 2198 receiver MUST ignore them. 2200 The "Reason" code is an 8-bit field that indicates the reason for the 2201 termination of the subflow. The following codes are defined in this 2202 document: 2204 o Unspecified error (code 0x0). This is the default error implying 2205 the subflow is no longer available. The presence of this option 2206 shows that the RST was generated by a MPTCP-aware device. 2208 o MPTCP specific error (code 0x01). An error has been detected in 2209 the processing of MPTCP options. This is the usual reason code to 2210 return in the cases where a RST is being sent to close a subflow 2211 for reasons of an invalid response. 2213 o Lack of resources (code 0x02). This code indicates that the 2214 sending host does not have enough resources to support the 2215 terminated subflow. 2217 o Administratively prohibited (code 0x03). This code indicates that 2218 the requested subflow is prohibited by the policies of the sending 2219 host. 2221 o Too much outstanding data (code 0x04). This code indicates that 2222 there is an excessive amount of data that need to be transmitted 2223 over the terminated subflow while having already been acknowledged 2224 over one or more other subflows. This may occur if a path has 2225 been unavailable for a short period and it is more efficient to 2226 reset and start again than it is to retransmit the queued data. 2228 o Unacceptable performance (code 0x05). This code indicates that 2229 the performance of this subflow was too low compared to the other 2230 subflows of this Multipath TCP connection. 2232 o Middlebox interference (code 0x06). Middlebox interference has 2233 been detected over this subflow making MPTCP signaling invalid. 2234 For example, this may be sent if the checksum does not validate. 2236 3.7. Fallback 2238 Sometimes, middleboxes will exist on a path that could prevent the 2239 operation of MPTCP. MPTCP has been designed in order to cope with 2240 many middlebox modifications (see Section 6), but there are still 2241 some cases where a subflow could fail to operate within the MPTCP 2242 requirements. These cases are notably the following: the loss of 2243 MPTCP options on a path, and the modification of payload data. If 2244 such an event occurs, it is necessary to "fall back" to the previous, 2245 safe operation. This may be either falling back to regular TCP or 2246 removing a problematic subflow. 2248 At the start of an MPTCP connection (i.e., the first subflow), it is 2249 important to ensure that the path is fully MPTCP capable and the 2250 necessary MPTCP options can reach each host. The handshake as 2251 described in Section 3.1 SHOULD fall back to regular TCP if either of 2252 the SYN messages do not have the MPTCP options: this is the same, and 2253 desired, behavior in the case where a host is not MPTCP capable, or 2254 the path does not support the MPTCP options. When attempting to join 2255 an existing MPTCP connection (Section 3.2), if a path is not MPTCP 2256 capable and the MPTCP options do not get through on the SYNs, the 2257 subflow will be closed according to the MP_JOIN logic. 2259 There is, however, another corner case that should be addressed. 2260 That is one of MPTCP options getting through on the SYN, but not on 2261 regular packets. This can be resolved if the subflow is the first 2262 subflow, and thus all data in flight is contiguous, using the 2263 following rules. 2265 A sender MUST include a DSS option with data sequence mapping in 2266 every segment until one of the sent segments has been acknowledged 2267 with a DSS option containing a Data ACK. Upon reception of the 2268 acknowledgment, the sender has the confirmation that the DSS option 2269 passes in both directions and may choose to send fewer DSS options 2270 than once per segment. 2272 If, however, an ACK is received for data (not just for the SYN) 2273 without a DSS option containing a Data ACK, the sender determines the 2274 path is not MPTCP capable. In the case of this occurring on an 2275 additional subflow (i.e., one started with MP_JOIN), the host MUST 2276 close the subflow with a RST, which SHOULD contain a MP_TCPRST option 2277 (Section 3.6) with a "Middlebox interferance" reason code. 2279 In the case of such an ACK being received on the first subflow (i.e., 2280 that started with MP_CAPABLE), before any additional subflows are 2281 added, the implementation MUST drop out of an MPTCP mode, back to 2282 regular TCP. The sender will send one final data sequence mapping, 2283 with the Data-Level Length value of 0 indicating an infinite mapping 2284 (to inform the other end in case the path drops options in one 2285 direction only), and then revert to sending data on the single 2286 subflow without any MPTCP options. 2288 If a subflow breaks during operation, e.g. if it is re-routed and 2289 MPTCP options are no longer permitted, then once this is detected (by 2290 the subflow-level receive buffer filling up), the subflow SHOULD be 2291 treated as broken and closed with a RST, since no data can be 2292 delivered to the application layer, and no fallback signal can be 2293 reliably sent. This RST SHOULD include the MP_TCPRST option 2294 (Section 3.6) with a "Middlebox interferance" reason code. 2296 These rules should cover all cases where such a failure could happen: 2297 whether it's on the forward or reverse path and whether the server or 2298 the client first sends data. If lost options on data packets occur 2299 on any other subflow apart from the initial subflow, it should be 2300 treated as a standard path failure. The data would not be DATA_ACKed 2301 (since there is no mapping for the data), and the subflow can be 2302 closed with a RST, containing a MP_TCPRST option (Section 3.6) with a 2303 "Middlebox interferance" reason code. 2305 So far this section has discussed the lost of MPTCP options, either 2306 initially, or during the course of the connection. As described in 2307 Section 3.3, each portion of data for which there is a mapping is 2308 protected by a checksum, if checksums have been negotiated. This 2309 mechanism is used to detect if middleboxes have made any adjustments 2310 to the payload (added, removed, or changed data). A checksum will 2311 fail if the data has been changed in any way. This will also detect 2312 if the length of data on the subflow is increased or decreased, and 2313 this means the data sequence mapping is no longer valid. The sender 2314 no longer knows what subflow-level sequence number the receiver is 2315 genuinely operating at (the middlebox will be faking ACKs in return), 2316 and it cannot signal any further mappings. Furthermore, in addition 2317 to the possibility of payload modifications that are valid at the 2318 application layer, there is the possibility that such modifications 2319 could be triggered across MPTCP segment boundaries, corrupting the 2320 data. Therefore, all data from the start of the segment that failed 2321 the checksum onwards is not trustworthy. 2323 Note that if checksum usage has not been negotiated, this fallback 2324 mechanism cannot be used unless there is some higher or lower layer 2325 signal to inform the MPTCP implementation that the payload has been 2326 tampered with. 2328 When multiple subflows are in use, the data in flight on a subflow 2329 will likely involve data that is not contiguously part of the 2330 connection-level stream, since segments will be spread across the 2331 multiple subflows. Due to the problems identified above, it is not 2332 possible to determine what adjustment has done to the data (notably, 2333 any changes to the subflow sequence numbering). Therefore, it is not 2334 possible to recover the subflow, and the affected subflow must be 2335 immediately closed with a RST, featuring an MP_FAIL option 2336 (Figure 16), which defines the data sequence number at the start of 2337 the segment (defined by the data sequence mapping) that had the 2338 checksum failure. Note that the MP_FAIL option requires the use of 2339 the full 64-bit sequence number, even if 32-bit sequence numbers are 2340 normally in use in the DSS signals on the path. 2342 1 2 3 2343 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 2344 +---------------+---------------+-------+----------------------+ 2345 | Kind | Length=12 |Subtype| (reserved) | 2346 +---------------+---------------+-------+----------------------+ 2347 | | 2348 | Data Sequence Number (8 octets) | 2349 | | 2350 +--------------------------------------------------------------+ 2352 Figure 16: Fallback (MP_FAIL) Option 2354 The receiver of this option MUST discard all data following the data 2355 sequence number specified. Failed data MUST NOT be DATA_ACKed and so 2356 will be retransmitted on other subflows (Section 3.3.6). 2358 A special case is when there is a single subflow and it fails with a 2359 checksum error. If it is known that all unacknowledged data in 2360 flight is contiguous (which will usually be the case with a single 2361 subflow), an infinite mapping can be applied to the subflow without 2362 the need to close it first, and essentially turn off all further 2363 MPTCP signaling. In this case, if a receiver identifies a checksum 2364 failure when there is only one path, it will send back an MP_FAIL 2365 option on the subflow-level ACK, referring to the data-level sequence 2366 number of the start of the segment on which the checksum error was 2367 detected. The sender will receive this, and if all unacknowledged 2368 data in flight is contiguous, will signal an infinite mapping. This 2369 infinite mapping will be a DSS option (Section 3.3) on the first new 2370 packet, containing a data sequence mapping that acts retroactively, 2371 referring to the start of the subflow sequence number of the most 2372 recent segment that was known to be delivered intact (i.e. was 2373 successfully DATA_ACKed). From that point onwards, data can be 2374 altered by a middlebox without affecting MPTCP, as the data stream is 2375 equivalent to a regular, legacy TCP session. Whilst in theory paths 2376 may only be damaged in one direction, and the MP_FAIL signal affects 2377 only one direction of traffic, for implementation simplicity, the 2378 receiver of an MP_FAIL MUST also respond with an MP_FAIL in the 2379 reverse direction and entirely revert to a regular TCP session. 2381 In the rare case that the data is not contiguous (which could happen 2382 when there is only one subflow but it is retransmitting data from a 2383 subflow that has recently been uncleanly closed), the receiver MUST 2384 close the subflow with a RST with MP_FAIL. The receiver MUST discard 2385 all data that follows the data sequence number specified. The sender 2386 MAY attempt to create a new subflow belonging to the same connection, 2387 and, if it chooses to do so, SHOULD place the single subflow 2388 immediately in single-path mode by setting an infinite data sequence 2389 mapping. This mapping will begin from the data-level sequence number 2390 that was declared in the MP_FAIL. 2392 After a sender signals an infinite mapping, it MUST only use subflow 2393 ACKs to clear its send buffer. This is because Data ACKs may become 2394 misaligned with the subflow ACKs when middleboxes insert or delete 2395 data. The receive SHOULD stop generating Data ACKs after it receives 2396 an infinite mapping. 2398 When a connection has fallen back with an infinite mapping, only one 2399 subflow can send data; otherwise, the receiver would not know how to 2400 reorder the data. In practice, this means that all MPTCP subflows 2401 will have to be terminated except one. Once MPTCP falls back to 2402 regular TCP, it MUST NOT revert to MPTCP later in the connection. 2404 It should be emphasized that MPTCP is not attempting to prevent the 2405 use of middleboxes that want to adjust the payload. An MPTCP-aware 2406 middlebox could provide such functionality by also rewriting 2407 checksums. 2409 3.8. Error Handling 2411 In addition to the fallback mechanism as described above, the 2412 standard classes of TCP errors may need to be handled in an MPTCP- 2413 specific way. Note that changing semantics -- such as the relevance 2414 of a RST -- are covered in Section 4. Where possible, we do not want 2415 to deviate from regular TCP behavior. 2417 The following list covers possible errors and the appropriate MPTCP 2418 behavior: 2420 o Unknown token in MP_JOIN (or HMAC failure in MP_JOIN ACK, or 2421 missing MP_JOIN in SYN/ACK response): send RST (analogous to TCP's 2422 behavior on an unknown port) 2424 o DSN out of window (during normal operation): drop the data, do not 2425 send Data ACKs 2427 o Remove request for unknown address ID: silently ignore 2429 3.9. Heuristics 2431 There are a number of heuristics that are needed for performance or 2432 deployment but that are not required for protocol correctness. In 2433 this section, we detail such heuristics. Note that discussion of 2434 buffering and certain sender and receiver window behaviors are 2435 presented in Sections 3.3.4 and 3.3.5, as well as retransmission in 2436 Section 3.3.6. 2438 3.9.1. Port Usage 2440 Under typical operation, an MPTCP implementation SHOULD use the same 2441 ports as already in use. In other words, the destination port of a 2442 SYN containing an MP_JOIN option SHOULD be the same as the remote 2443 port of the first subflow in the connection. The local port for such 2444 SYNs SHOULD also be the same as for the first subflow (and as such, 2445 an implementation SHOULD reserve ephemeral ports across all local IP 2446 addresses), although there may be cases where this is infeasible. 2447 This strategy is intended to maximize the probability of the SYN 2448 being permitted by a firewall or NAT at the recipient and to avoid 2449 confusing any network monitoring software. 2451 There may also be cases, however, where a host wishes to signal that 2452 a specific port should be used, and this facility is provided in the 2453 ADD_ADDR option as documented in Section 3.4.1. It is therefore 2454 feasible to allow multiple subflows between the same two addresses 2455 but using different port pairs, and such a facility could be used to 2456 allow load balancing within the network based on 5-tuples (e.g., some 2457 ECMP implementations [RFC2992]). 2459 3.9.2. Delayed Subflow Start and Subflow Symmetry 2461 Many TCP connections are short-lived and consist only of a few 2462 segments, and so the overheads of using MPTCP outweigh any benefits. 2463 A heuristic is required, therefore, to decide when to start using 2464 additional subflows in an MPTCP connection. We expect that 2465 experience gathered from deployments will provide further guidance on 2466 this, and will be affected by particular application characteristics 2467 (which are likely to change over time). However, a suggested 2468 general-purpose heuristic that an implementation MAY choose to employ 2469 is as follows. Results from experimental deployments are needed in 2470 order to verify the correctness of this proposal. 2472 If a host has data buffered for its peer (which implies that the 2473 application has received a request for data), the host opens one 2474 subflow for each initial window's worth of data that is buffered. 2476 Consideration should also be given to limiting the rate of adding new 2477 subflows, as well as limiting the total number of subflows open for a 2478 particular connection. A host may choose to vary these values based 2479 on its load or knowledge of traffic and path characteristics. 2481 Note that this heuristic alone is probably insufficient. Traffic for 2482 many common applications, such as downloads, is highly asymmetric and 2483 the host that is multihomed may well be the client that will never 2484 fill its buffers, and thus never use MPTCP according to this 2485 heuristic. Advanced APIs that allow an application to signal its 2486 traffic requirements would aid in these decisions. 2488 An additional time-based heuristic could be applied, opening 2489 additional subflows after a given period of time has passed. This 2490 would alleviate the above issue, and also provide resilience for low- 2491 bandwidth but long-lived applications. 2493 Another issue is that both communicating hosts may simultaneously try 2494 to set up a subflow between the same pair of addresses. This leads 2495 to an inefficient use of resources. 2497 If the the same ports are used on all subflows, as recommended above, 2498 then standard TCP simultaneous open logic should take care of this 2499 situation and only one subflow will be established between the 2500 address pairs. However, this relies on the same ports being used at 2501 both end hosts. If a host does not support TCP simultaneous open, it 2502 is RECOMMENDED that some element of randomization is applied to the 2503 time to wait before opening new subflows, so that only one subflow is 2504 created between a given address pair. If, however, hosts signal 2505 additional ports to use (for example, for leveraging ECMP on-path), 2506 this heuristic is not appropriate. 2508 This section has shown some of the considerations that an implementer 2509 should give when developing MPTCP heuristics, but is not intended to 2510 be prescriptive. 2512 3.9.3. Failure Handling 2514 Requirements for MPTCP's handling of unexpected signals have been 2515 given in Section 3.8. There are other failure cases, however, where 2516 a hosts can choose appropriate behavior. 2518 For example, Section 3.1 suggests that a host SHOULD fall back to 2519 trying regular TCP SYNs after one or more failures of MPTCP SYNs for 2520 a connection. A host may keep a system-wide cache of such 2521 information, so that it can back off from using MPTCP, firstly for 2522 that particular destination host, and eventually on a whole 2523 interface, if MPTCP connections continue failing. 2525 Another failure could occur when the MP_JOIN handshake fails. 2526 Section 3.8 specifies that an incorrect handshake MUST lead to the 2527 subflow being closed with a RST. A host operating an active 2528 intrusion detection system may choose to start blocking MP_JOIN 2529 packets from the source host if multiple failed MP_JOIN attempts are 2530 seen. From the connection initiator's point of view, if an MP_JOIN 2531 fails, it SHOULD NOT attempt to connect to the same IP address and 2532 port during the lifetime of the connection, unless the other host 2533 refreshes the information with another ADD_ADDR option. Note that 2534 the ADD_ADDR option is informational only, and does not guarantee the 2535 other host will attempt a connection. 2537 In addition, an implementation may learn, over a number of 2538 connections, that certain interfaces or destination addresses 2539 consistently fail and may default to not trying to use MPTCP for 2540 these. Behavior could also be learned for particularly badly 2541 performing subflows or subflows that regularly fail during use, in 2542 order to temporarily choose not to use these paths. 2544 4. Semantic Issues 2546 In order to support multipath operation, the semantics of some TCP 2547 components have changed. To aid clarity, this section collects these 2548 semantic changes as a reference. 2550 Sequence number: The (in-header) TCP sequence number is specific to 2551 the subflow. To allow the receiver to reorder application data, 2552 an additional data-level sequence space is used. In this data- 2553 level sequence space, the initial SYN and the final DATA_FIN 2554 occupy 1 octet of sequence space. This is to ensure these signals 2555 are acknowledged at the connection level. There is an explicit 2556 mapping of data sequence space to subflow sequence space, which is 2557 signaled through TCP options in data packets. 2559 ACK: The ACK field in the TCP header acknowledges only the subflow 2560 sequence number, not the data-level sequence space. 2561 Implementations SHOULD NOT attempt to infer a data-level 2562 acknowledgment from the subflow ACKs. This separates subflow- and 2563 connection-level processing at an end host. 2565 Duplicate ACK: A duplicate ACK that includes any MPTCP signaling 2566 (with the exception of the DSS option) MUST NOT be treated as a 2567 signal of congestion. To limit the chances of non-MPTCP-aware 2568 entities mistakenly interpreting duplicate ACKs as a signal of 2569 congestion, MPTCP SHOULD NOT send more than two duplicate ACKs 2570 containing (non-DSS) MPTCP signals in a row. 2572 Receive Window: The receive window in the TCP header indicates the 2573 amount of free buffer space for the whole data-level connection 2574 (as opposed to for this subflow) that is available at the 2575 receiver. This is the same semantics as regular TCP, but to 2576 maintain these semantics the receive window must be interpreted at 2577 the sender as relative to the sequence number given in the 2578 DATA_ACK rather than the subflow ACK in the TCP header. In this 2579 way, the original flow control role is preserved. Note that some 2580 middleboxes may change the receive window, and so a host SHOULD 2581 use the maximum value of those recently seen on the constituent 2582 subflows for the connection-level receive window, and also needs 2583 to maintain a subflow-level window for subflow-level processing. 2585 FIN: The FIN flag in the TCP header applies only to the subflow it 2586 is sent on, not to the whole connection. For connection-level FIN 2587 semantics, the DATA_FIN option is used. 2589 RST: The RST flag in the TCP header applies only to the subflow it 2590 is sent on, not to the whole connection. The MP_FASTCLOSE option 2591 provides the fast close functionality of a RST at the MPTCP 2592 connection level. 2594 Address List: Address list management (i.e., knowledge of the local 2595 and remote hosts' lists of available IP addresses) is handled on a 2596 per-connection basis (as opposed to per subflow, per host, or per 2597 pair of communicating hosts). This permits the application of 2598 per-connection local policy. Adding an address to one connection 2599 (either explicitly through an Add Address message, or implicitly 2600 through a Join) has no implication for other connections between 2601 the same pair of hosts. 2603 5-tuple: The 5-tuple (protocol, local address, local port, remote 2604 address, remote port) presented by kernel APIs to the application 2605 layer in a non-multipath-aware application is that of the first 2606 subflow, even if the subflow has since been closed and removed 2607 from the connection. This decision, and other related API issues, 2608 are discussed in more detail in [RFC6897]. 2610 5. Security Considerations 2612 As identified in [RFC6181], the addition of multipath capability to 2613 TCP will bring with it a number of new classes of threat. In order 2614 to prevent these, [RFC6182] presents a set of requirements for a 2615 security solution for MPTCP. The fundamental goal is for the 2616 security of MPTCP to be "no worse" than regular TCP today, and the 2617 key security requirements are: 2619 o Provide a mechanism to confirm that the parties in a subflow 2620 handshake are the same as in the original connection setup. 2622 o Provide verification that the peer can receive traffic at a new 2623 address before using it as part of a connection. 2625 o Provide replay protection, i.e., ensure that a request to add/ 2626 remove a subflow is 'fresh'. 2628 In order to achieve these goals, MPTCP includes a hash-based 2629 handshake algorithm documented in Sections 3.1 and 3.2. 2631 The security of the MPTCP connection hangs on the use of keys that 2632 are shared once at the start of the first subflow, and are never sent 2633 again over the network (unless used in the fast close mechanism, 2634 Section 3.5). To ease demultiplexing while not giving away any 2635 cryptographic material, future subflows use a truncated cryptographic 2636 hash of this key as the connection identification "token". The keys 2637 are concatenated and used as keys for creating Hash-based Message 2638 Authentication Codes (HMACs) used on subflow setup, in order to 2639 verify that the parties in the handshake are the same as in the 2640 original connection setup. It also provides verification that the 2641 peer can receive traffic at this new address. Replay attacks would 2642 still be possible when only keys are used; therefore, the handshakes 2643 use single-use random numbers (nonces) at both ends -- this ensures 2644 the HMAC will never be the same on two handshakes. Guidance on 2645 generating random numbers suitable for use as keys is given in 2646 [RFC4086] and discussed in Section 3.1. HMAC is also used to secure 2647 the ADD_ADDR option, due to the threats identified in [RFC7430]. 2649 The use of crypto capability bits in the initial connection handshake 2650 to negotiate use of a particular algorithm allows the deployment of 2651 additional crypto mechanisms in the future. Note that this would be 2652 susceptible to bid-down attacks only if the attacker was on-path (and 2653 thus would be able to modify the data anyway). The security 2654 mechanism presented in this document should therefore protect against 2655 all forms of flooding and hijacking attacks discussed in [RFC6181]. 2657 The version negotiation specified in Section 3.1, if differing MPTCP 2658 versions shared a common negotiation format, would allow an on-path 2659 attacker to apply a theoretical bid-down attack. Since the v1 and v0 2660 protocols have a different handshake, such an attack would require 2661 the client to re-establish the connection using v0, and this being 2662 supported by the server. Note that an on-path attacker would have 2663 access to the raw data, negating any other TCP-level security 2664 mechanisms. Also a change from [RFC6824] has removed the subflow 2665 identifier from the MP_PRIO option (Section 3.3.8), to remove the 2666 theoretical attack where a subflow could be placed in "backup" mode 2667 by an attacker. 2669 During normal operation, regular TCP protection mechanisms (such as 2670 ensuring sequence numbers are in-window) will provide the same level 2671 of protection against attacks on individual TCP subflows as exists 2672 for regular TCP today. Implementations will introduce additional 2673 buffers compared to regular TCP, to reassemble data at the connection 2674 level. The application of window sizing will minimize the risk of 2675 denial-of-service attacks consuming resources. 2677 As discussed in Section 3.4.1, a host may advertise its private 2678 addresses, but these might point to different hosts in the receiver's 2679 network. The MP_JOIN handshake (Section 3.2) will ensure that this 2680 does not succeed in setting up a subflow to the incorrect host. 2681 However, it could still create unwanted TCP handshake traffic. This 2682 feature of MPTCP could be a target for denial-of-service exploits, 2683 with malicious participants in MPTCP connections encouraging the 2684 recipient to target other hosts in the network. Therefore, 2685 implementations should consider heuristics (Section 3.9) at both the 2686 sender and receiver to reduce the impact of this. 2688 To further protect against malicious ADD_ADDR messages sent by an 2689 off-path attacker, the ADD_ADDR includes an HMAC using the keys 2690 negotiated during the handshake. This effectively prevents an 2691 attacker from diverting an MPTCP connection through an off-path 2692 ADD_ADDR injection into the stream. 2694 A small security risk could theoretically exist with key reuse, but 2695 in order to accomplish a replay attack, both the sender and receiver 2696 keys, and the sender and receiver random numbers, in the MP_JOIN 2697 handshake (Section 3.2) would have to match. 2699 Whilst this specification defines a "medium" security solution, 2700 meeting the criteria specified at the start of this section and the 2701 threat analysis ([RFC6181]), since attacks only ever get worse, it is 2702 likely that a future Standards Track version of MPTCP would need to 2703 be able to support stronger security. There are several ways the 2704 security of MPTCP could potentially be improved; some of these would 2705 be compatible with MPTCP as defined in this document, whilst others 2706 may not be. For now, the best approach is to get experience with the 2707 current approach, establish what might work, and check that the 2708 threat analysis is still accurate. 2710 Possible ways of improving MPTCP security could include: 2712 o defining a new MPCTP cryptographic algorithm, as negotiated in 2713 MP_CAPABLE. A sub-case could be to include an additional 2714 deployment assumption, such as stateful servers, in order to allow 2715 a more powerful algorithm to be used. 2717 o defining how to secure data transfer with MPTCP, whilst not 2718 changing the signaling part of the protocol. 2720 o defining security that requires more option space, perhaps in 2721 conjunction with a "long options" proposal for extending the TCP 2722 options space (such as those surveyed in [TCPLO]), or perhaps 2723 building on the current approach with a second stage of MPTCP- 2724 option-based security. 2726 o revisiting the working group's decision to exclusively use TCP 2727 options for MPTCP signaling, and instead look at also making use 2728 of the TCP payloads. 2730 MPTCP has been designed with several methods available to indicate a 2731 new security mechanism, including: 2733 o available flags in MP_CAPABLE (Figure 4); 2735 o available subtypes in the MPTCP option (Figure 3); 2737 o the version field in MP_CAPABLE (Figure 4); 2739 6. Interactions with Middleboxes 2741 Multipath TCP was designed to be deployable in the present world. 2742 Its design takes into account "reasonable" existing middlebox 2743 behavior. In this section, we outline a few representative 2744 middlebox-related failure scenarios and show how Multipath TCP 2745 handles them. Next, we list the design decisions multipath has made 2746 to accommodate the different middleboxes. 2748 A primary concern is our use of a new TCP option. Middleboxes should 2749 forward packets with unknown options unchanged, yet there are some 2750 that don't. These we expect will either strip options and pass the 2751 data, drop packets with new options, copy the same option into 2752 multiple segments (e.g., when doing segmentation), or drop options 2753 during segment coalescing. 2755 MPTCP uses a single new TCP option "Kind", and all message types are 2756 defined by "subtype" values (see Section 8). This should reduce the 2757 chances of only some types of MPTCP options being passed, and instead 2758 the key differing characteristics are different paths, and the 2759 presence of the SYN flag. 2761 MPTCP SYN packets on the first subflow of a connection contain the 2762 MP_CAPABLE option (Section 3.1). If this is dropped, MPTCP SHOULD 2763 fall back to regular TCP. If packets with the MP_JOIN option 2764 (Section 3.2) are dropped, the paths will simply not be used. 2766 If a middlebox strips options but otherwise passes the packets 2767 unchanged, MPTCP will behave safely. If an MP_CAPABLE option is 2768 dropped on either the outgoing or the return path, the initiating 2769 host can fall back to regular TCP, as illustrated in Figure 17 and 2770 discussed in Section 3.1. 2772 Subflow SYNs contain the MP_JOIN option. If this option is stripped 2773 on the outgoing path, the SYN will appear to be a regular SYN to Host 2774 B. Depending on whether there is a listening socket on the target 2775 port, Host B will reply either with SYN/ACK or RST (subflow 2776 connection fails). When Host A receives the SYN/ACK it sends a RST 2777 because the SYN/ACK does not contain the MP_JOIN option and its 2778 token. Either way, the subflow setup fails, but otherwise does not 2779 affect the MPTCP connection as a whole. 2781 Host A Host B 2782 | Middlebox M | 2783 | | | 2784 | SYN(MP_CAPABLE) | SYN | 2785 |-------------------|---------------->| 2786 | SYN/ACK | 2787 |<------------------------------------| 2788 a) MP_CAPABLE option stripped on outgoing path 2790 Host A Host B 2791 | SYN(MP_CAPABLE) | 2792 |------------------------------------>| 2793 | Middlebox M | 2794 | | | 2795 | SYN/ACK |SYN/ACK(MP_CAPABLE)| 2796 |<----------------|-------------------| 2797 b) MP_CAPABLE option stripped on return path 2799 Figure 17: Connection Setup with Middleboxes that Strip Options from 2800 Packets 2802 We now examine data flow with MPTCP, assuming the flow is correctly 2803 set up, which implies the options in the SYN packets were allowed 2804 through by the relevant middleboxes. If options are allowed through 2805 and there is no resegmentation or coalescing to TCP segments, 2806 Multipath TCP flows can proceed without problems. 2808 The case when options get stripped on data packets has been discussed 2809 in the Fallback section. If only some MPTCP options are stripped, 2810 behavior is not deterministic. If some data sequence mappings are 2811 lost, the connection can continue so long as mappings exist for the 2812 subflow-level data (e.g., if multiple maps have been sent that 2813 reinforce each other). If some subflow-level space is left unmapped, 2814 however, the subflow is treated as broken and is closed, through the 2815 process described in Section 3.7. MPTCP should survive with a loss 2816 of some Data ACKs, but performance will degrade as the fraction of 2817 stripped options increases. We do not expect such cases to appear in 2818 practice, though: most middleboxes will either strip all options or 2819 let them all through. 2821 We end this section with a list of middlebox classes, their behavior, 2822 and the elements in the MPTCP design that allow operation through 2823 such middleboxes. Issues surrounding dropping packets with options 2824 or stripping options were discussed above, and are not included here: 2826 o NATs [RFC3022] (Network Address (and Port) Translators) change the 2827 source address (and often source port) of packets. This means 2828 that a host will not know its public-facing address for signaling 2829 in MPTCP. Therefore, MPTCP permits implicit address addition via 2830 the MP_JOIN option, and the handshake mechanism ensures that 2831 connection attempts to private addresses [RFC1918], since they are 2832 authenticated, will only set up subflows to the correct hosts. 2833 Explicit address removal is undertaken by an Address ID to allow 2834 no knowledge of the source address. 2836 o Performance Enhancing Proxies (PEPs) [RFC3135] might proactively 2837 ACK data to increase performance. MPTCP, however, relies on 2838 accurate congestion control signals from the end host, and non- 2839 MPTCP-aware PEPs will not be able to provide such signals. MPTCP 2840 will, therefore, fall back to single-path TCP, or close the 2841 problematic subflow (see Section 3.7). 2843 o Traffic Normalizers [norm] may not allow holes in sequence 2844 numbers, and may cache packets and retransmit the same data. 2845 MPTCP looks like standard TCP on the wire, and will not retransmit 2846 different data on the same subflow sequence number. In the event 2847 of a retransmission, the same data will be retransmitted on the 2848 original TCP subflow even if it is additionally retransmitted at 2849 the connection level on a different subflow. 2851 o Firewalls [RFC2979] might perform initial sequence number 2852 randomization on TCP connections. MPTCP uses relative sequence 2853 numbers in data sequence mapping to cope with this. Like NATs, 2854 firewalls will not permit many incoming connections, so MPTCP 2855 supports address signaling (ADD_ADDR) so that a multiaddressed 2856 host can invite its peer behind the firewall/NAT to connect out to 2857 its additional interface. 2859 o Intrusion Detection Systems look out for traffic patterns and 2860 content that could threaten a network. Multipath will mean that 2861 such data is potentially spread, so it is more difficult for an 2862 IDS to analyze the whole traffic, and potentially increases the 2863 risk of false positives. However, a MPTCP-aware IDS can read 2864 tokens to correlate multiple subflows and reassemble them for 2865 analysis. 2867 o Application-level middleboxes such as content-aware firewalls may 2868 alter the payload within a subflow, such as rewriting URIs in HTTP 2869 traffic. MPTCP will detect these using the checksum and close the 2870 affected subflow(s), if there are other subflows that can be used. 2871 If all subflows are affected, multipath will fall back to TCP, 2872 allowing such middleboxes to change the payload. MPTCP-aware 2873 middleboxes should be able to adjust the payload and MPTCP 2874 metadata in order not to break the connection. 2876 In addition, all classes of middleboxes may affect TCP traffic in the 2877 following ways: 2879 o TCP options may be removed, or packets with unknown options 2880 dropped, by many classes of middleboxes. It is intended that the 2881 initial SYN exchange, with a TCP option, will be sufficient to 2882 identify the path capabilities. If such a packet does not get 2883 through, MPTCP will end up falling back to regular TCP. 2885 o Segmentation/Coalescing (e.g., TCP segmentation offloading) might 2886 copy options between packets and might strip some options. 2887 MPTCP's data sequence mapping includes the relative subflow 2888 sequence number instead of using the sequence number in the 2889 segment. In this way, the mapping is independent of the packets 2890 that carry it. 2892 o The receive window may be shrunk by some middleboxes at the 2893 subflow level. MPTCP will use the maximum window at data level, 2894 but will also obey subflow-specific windows. 2896 7. Acknowledgments 2898 The authors gratefully acknowledge significant input into this 2899 document from Sebastien Barre and Andrew McDonald. 2901 The authors also wish to acknowledge reviews and contributions from 2902 Iljitsch van Beijnum, Lars Eggert, Marcelo Bagnulo, Robert Hancock, 2903 Pasi Sarolahti, Toby Moncaster, Philip Eardley, Sergio Lembo, 2904 Lawrence Conroy, Yoshifumi Nishida, Bob Briscoe, Stein Gjessing, 2905 Andrew McGregor, Georg Hampel, Anumita Biswas, Wes Eddy, Alexey 2906 Melnikov, Francis Dupont, Adrian Farrel, Barry Leiba, Robert Sparks, 2907 Sean Turner, Stephen Farrell, Martin Stiemerling, Gregory Detal, 2908 Fabien Duchene, Xavier de Foy, Rahul Jadhav, and Klemens Schragel. 2910 8. IANA Considerations 2912 This document obsoletes [RFC6824] and as such IANA is requested to 2913 update the TCP option space registry to point to this document for 2914 Multipath TCP, as follows: 2916 +------+--------+-----------------------+---------------+ 2917 | Kind | Length | Meaning | Reference | 2918 +------+--------+-----------------------+---------------+ 2919 | 30 | N | Multipath TCP (MPTCP) | This document | 2920 +------+--------+-----------------------+---------------+ 2922 Table 1: TCP Option Kind Numbers 2924 8.1. MPTCP Option Subtypes 2926 The 4-bit MPTCP subtype sub-registry ("MPTCP Option Subtypes" under 2927 the "Transmission Control Protocol (TCP) Parameters" registry) was 2928 defined in [RFC6824]. This document defines one additional subtype 2929 (ADD_ADDR) and updates the references to this document for all sub- 2930 types except ADD_ADDR, which is deprecated. The updates are listed 2931 in the following table. 2933 +-------+-----------------+-------------------------+---------------+ 2934 | Value | Symbol | Name | Reference | 2935 +-------+-----------------+-------------------------+---------------+ 2936 | 0x0 | MP_CAPABLE | Multipath Capable | This | 2937 | | | | document, | 2938 | | | | Section 3.1 | 2939 | 0x1 | MP_JOIN | Join Connection | This | 2940 | | | | document, | 2941 | | | | Section 3.2 | 2942 | 0x2 | DSS | Data Sequence Signal | This | 2943 | | | (Data ACK and data | document, | 2944 | | | sequence mapping) | Section 3.3 | 2945 | 0x3 | ADD_ADDR | Add Address | This | 2946 | | | | document, | 2947 | | | | Section 3.4.1 | 2948 | 0x4 | REMOVE_ADDR | Remove Address | This | 2949 | | | | document, | 2950 | | | | Section 3.4.2 | 2951 | 0x5 | MP_PRIO | Change Subflow Priority | This | 2952 | | | | document, | 2953 | | | | Section 3.3.8 | 2954 | 0x6 | MP_FAIL | Fallback | This | 2955 | | | | document, | 2956 | | | | Section 3.7 | 2957 | 0x7 | MP_FASTCLOSE | Fast Close | This | 2958 | | | | document, | 2959 | | | | Section 3.5 | 2960 | 0x8 | MP_TCPRST | Subflow Reset | This | 2961 | | | | document, | 2962 | | | | Section 3.6 | 2963 | 0xf | MP_EXPERIMENTAL | Reserved for private | | 2964 | | | experiments | | 2965 +-------+-----------------+-------------------------+---------------+ 2967 Table 2: MPTCP Option Subtypes 2969 Values 0x9 through 0xe are currently unassigned. Option 0xf is 2970 reserved for use by private experiments. Its use may be formalized 2971 in a future specification. 2973 8.2. MPTCP Handshake Algorithms 2975 IANA has created another sub-registry, "MPTCP Handshake Algorithms" 2976 under the "Transmission Control Protocol (TCP) Parameters" registry, 2977 based on the flags in MP_CAPABLE (Section 3.1). IANA is requested to 2978 update the references of this table to this document, as follows: 2980 +-------+----------------------------------------+------------------+ 2981 | Flag | Meaning | Reference | 2982 | Bit | | | 2983 +-------+----------------------------------------+------------------+ 2984 | A | Checksum required | This document, | 2985 | | | Section 3.1 | 2986 | B | Extensibility | This document, | 2987 | | | Section 3.1 | 2988 | C | Do not attempt to establish new | This document, | 2989 | | subflows to the source address. | Section 3.1 | 2990 | D-G | Unassigned | | 2991 | H | HMAC-SHA256 | This document, | 2992 | | | Section 3.2 | 2993 +-------+----------------------------------------+------------------+ 2995 Table 3: MPTCP Handshake Algorithms 2997 Note that the meanings of bits D through H can be dependent upon bit 2998 B, depending on how Extensibility is defined in future 2999 specifications; see Section 3.1 for more information. 3001 Future assignments in this registry are also to be defined by 3002 Standards Action as defined by [RFC8126]. Assignments consist of the 3003 value of the flags, a symbolic name for the algorithm, and a 3004 reference to its specification. 3006 8.3. MP_TCPRST Reason Codes 3008 IANA is requested to create a further sub-registry, "MP_TCPRST Reason 3009 Codes" under the "Transmission Control Protocol (TCP) Parameters" 3010 registry, based on the reason code in MP_TCPRST (Section 3.6): 3012 +------+-----------------------------+----------------------------+ 3013 | Code | Meaning | Reference | 3014 +------+-----------------------------+----------------------------+ 3015 | 0x00 | Unspecified TCP error | This document, Section 3.6 | 3016 | 0x01 | MPTCP specific error | This document, Section 3.6 | 3017 | 0x02 | Lack of resources | This document, Section 3.6 | 3018 | 0x03 | Administratively prohibited | This document, Section 3.6 | 3019 | 0x04 | Too much outstanding data | This document, Section 3.6 | 3020 | 0x05 | Unacceptable performance | This document, Section 3.6 | 3021 | 0x06 | Middlebox interference | This document, Section 3.6 | 3022 +------+-----------------------------+----------------------------+ 3024 Table 4: MPTCP MP_TCPRST Reason Codes 3026 9. References 3028 9.1. Normative References 3030 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, 3031 RFC 793, DOI 10.17487/RFC0793, September 1981, 3032 . 3034 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 3035 Requirement Levels", BCP 14, RFC 2119, 3036 DOI 10.17487/RFC2119, March 1997, . 3039 [RFC6182] Ford, A., Raiciu, C., Handley, M., Barre, S., and J. 3040 Iyengar, "Architectural Guidelines for Multipath TCP 3041 Development", RFC 6182, DOI 10.17487/RFC6182, March 2011, 3042 . 3044 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 3045 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 3046 May 2017, . 3048 [SHS] National Institute of Science and Technology, "Secure Hash 3049 Standard", Federal Information Processing Standard 3050 (FIPS) 180-4, August 2015, 3051 . 3054 9.2. Informative References 3056 [deployments] 3057 Bonaventure, O. and S. Seo, "Multipath TCP Deployments", 3058 IETF Journal 2016, November 2016, 3059 . 3061 [howhard] Raiciu, C., Paasch, C., Barre, S., Ford, A., Honda, M., 3062 Duchene, F., Bonaventure, O., and M. Handley, "How Hard 3063 Can It Be? Designing and Implementing a Deployable 3064 Multipath TCP", Usenix Symposium on Networked Systems 3065 Design and Implementation 2012, 2012, 3066 . 3069 [norm] Handley, M., Paxson, V., and C. Kreibich, "Network 3070 Intrusion Detection: Evasion, Traffic Normalization, and 3071 End-to-End Protocol Semantics", Usenix Security 2001, 3072 2001, 3073 . 3076 [RFC1122] Braden, R., Ed., "Requirements for Internet Hosts - 3077 Communication Layers", STD 3, RFC 1122, 3078 DOI 10.17487/RFC1122, October 1989, . 3081 [RFC1918] Rekhter, Y., Moskowitz, B., Karrenberg, D., de Groot, G., 3082 and E. Lear, "Address Allocation for Private Internets", 3083 BCP 5, RFC 1918, DOI 10.17487/RFC1918, February 1996, 3084 . 3086 [RFC2018] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP 3087 Selective Acknowledgment Options", RFC 2018, 3088 DOI 10.17487/RFC2018, October 1996, . 3091 [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, "HMAC: Keyed- 3092 Hashing for Message Authentication", RFC 2104, 3093 DOI 10.17487/RFC2104, February 1997, . 3096 [RFC2979] Freed, N., "Behavior of and Requirements for Internet 3097 Firewalls", RFC 2979, DOI 10.17487/RFC2979, October 2000, 3098 . 3100 [RFC2992] Hopps, C., "Analysis of an Equal-Cost Multi-Path 3101 Algorithm", RFC 2992, DOI 10.17487/RFC2992, November 2000, 3102 . 3104 [RFC3022] Srisuresh, P. and K. Egevang, "Traditional IP Network 3105 Address Translator (Traditional NAT)", RFC 3022, 3106 DOI 10.17487/RFC3022, January 2001, . 3109 [RFC3135] Border, J., Kojo, M., Griner, J., Montenegro, G., and Z. 3110 Shelby, "Performance Enhancing Proxies Intended to 3111 Mitigate Link-Related Degradations", RFC 3135, 3112 DOI 10.17487/RFC3135, June 2001, . 3115 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 3116 of Explicit Congestion Notification (ECN) to IP", 3117 RFC 3168, DOI 10.17487/RFC3168, September 2001, 3118 . 3120 [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker, 3121 "Randomness Requirements for Security", BCP 106, RFC 4086, 3122 DOI 10.17487/RFC4086, June 2005, . 3125 [RFC4987] Eddy, W., "TCP SYN Flooding Attacks and Common 3126 Mitigations", RFC 4987, DOI 10.17487/RFC4987, August 2007, 3127 . 3129 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 3130 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 3131 . 3133 [RFC5961] Ramaiah, A., Stewart, R., and M. Dalal, "Improving TCP's 3134 Robustness to Blind In-Window Attacks", RFC 5961, 3135 DOI 10.17487/RFC5961, August 2010, . 3138 [RFC6181] Bagnulo, M., "Threat Analysis for TCP Extensions for 3139 Multipath Operation with Multiple Addresses", RFC 6181, 3140 DOI 10.17487/RFC6181, March 2011, . 3143 [RFC6234] Eastlake 3rd, D. and T. Hansen, "US Secure Hash Algorithms 3144 (SHA and SHA-based HMAC and HKDF)", RFC 6234, 3145 DOI 10.17487/RFC6234, May 2011, . 3148 [RFC6356] Raiciu, C., Handley, M., and D. Wischik, "Coupled 3149 Congestion Control for Multipath Transport Protocols", 3150 RFC 6356, DOI 10.17487/RFC6356, October 2011, 3151 . 3153 [RFC6528] Gont, F. and S. Bellovin, "Defending against Sequence 3154 Number Attacks", RFC 6528, DOI 10.17487/RFC6528, February 3155 2012, . 3157 [RFC6824] Ford, A., Raiciu, C., Handley, M., and O. Bonaventure, 3158 "TCP Extensions for Multipath Operation with Multiple 3159 Addresses", RFC 6824, DOI 10.17487/RFC6824, January 2013, 3160 . 3162 [RFC6897] Scharf, M. and A. Ford, "Multipath TCP (MPTCP) Application 3163 Interface Considerations", RFC 6897, DOI 10.17487/RFC6897, 3164 March 2013, . 3166 [RFC7323] Borman, D., Braden, B., Jacobson, V., and R. 3167 Scheffenegger, Ed., "TCP Extensions for High Performance", 3168 RFC 7323, DOI 10.17487/RFC7323, September 2014, 3169 . 3171 [RFC7413] Cheng, Y., Chu, J., Radhakrishnan, S., and A. Jain, "TCP 3172 Fast Open", RFC 7413, DOI 10.17487/RFC7413, December 2014, 3173 . 3175 [RFC7430] Bagnulo, M., Paasch, C., Gont, F., Bonaventure, O., and C. 3176 Raiciu, "Analysis of Residual Threats and Possible Fixes 3177 for Multipath TCP (MPTCP)", RFC 7430, 3178 DOI 10.17487/RFC7430, July 2015, . 3181 [RFC8041] Bonaventure, O., Paasch, C., and G. Detal, "Use Cases and 3182 Operational Experience with Multipath TCP", RFC 8041, 3183 DOI 10.17487/RFC8041, January 2017, . 3186 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for 3187 Writing an IANA Considerations Section in RFCs", BCP 26, 3188 RFC 8126, DOI 10.17487/RFC8126, June 2017, 3189 . 3191 [TCPLO] Ramaiah, A., "TCP option space extension", Work 3192 in Progress, March 2012. 3194 Appendix A. Notes on Use of TCP Options 3196 The TCP option space is limited due to the length of the Data Offset 3197 field in the TCP header (4 bits), which defines the TCP header length 3198 in 32-bit words. With the standard TCP header being 20 bytes, this 3199 leaves a maximum of 40 bytes for options, and many of these may 3200 already be used by options such as timestamp and SACK. 3202 We have performed a brief study on the commonly used TCP options in 3203 SYN, data, and pure ACK packets, and found that there is enough room 3204 to fit all the options we propose using in this document. 3206 SYN packets typically include Maximum Segment Size (MSS) (4 bytes), 3207 window scale (3 bytes), SACK permitted (2 bytes), and timestamp (10 3208 bytes) options. Together these sum to 19 bytes. Some operating 3209 systems appear to pad each option up to a word boundary, thus using 3210 24 bytes (a brief survey suggests Windows XP and Mac OS X do this, 3211 whereas Linux does not). Optimistically, therefore, we have 21 bytes 3212 spare, or 16 if it has to be word-aligned. In either case, however, 3213 the SYN versions of Multipath Capable (12 bytes) and Join (12 or 16 3214 bytes) options will fit in this remaining space. 3216 Note that due to the use of a 64-bit data-level sequence space, it is 3217 feasible that MPTCP will not require the timestamp option for 3218 protection against wrapped sequence numbers (PAWS [RFC7323]), since 3219 the data-level sequence space has far less chance of wrapping. 3220 Confirmation of the validity of this optimisation is for further 3221 study. 3223 TCP data packets typically carry timestamp options in every packet, 3224 taking 10 bytes (or 12 with padding). That leaves 30 bytes (or 28, 3225 if word-aligned). The Data Sequence Signal (DSS) option varies in 3226 length depending on whether the data sequence mapping and DATA_ACK 3227 are included, and whether the sequence numbers in use are 4 or 8 3228 octets. The maximum size of the DSS option is 28 bytes, so even that 3229 will fit in the available space. But unless a connection is both 3230 bidirectional and high-bandwidth, it is unlikely that all that option 3231 space will be required on each DSS option. 3233 Within the DSS option, it is not necessary to include the data 3234 sequence mapping and DATA_ACK in each packet, and in many cases it 3235 may be possible to alternate their presence (so long as the mapping 3236 covers the data being sent in the following packet). It would also 3237 be possible to alternate between 4- and 8-byte sequence numbers in 3238 each option. 3240 On subflow and connection setup, an MPTCP option is also set on the 3241 third packet (an ACK). These are 20 bytes (for Multipath Capable) 3242 and 24 bytes (for Join), both of which will fit in the available 3243 option space. 3245 Pure ACKs in TCP typically contain only timestamps (10 bytes). Here, 3246 Multipath TCP typically needs to encode only the DATA_ACK (maximum of 3247 12 bytes). Occasionally, ACKs will contain SACK information. 3248 Depending on the number of lost packets, SACK may utilize the entire 3249 option space. If a DATA_ACK had to be included, then it is probably 3250 necessary to reduce the number of SACK blocks to accommodate the 3251 DATA_ACK. However, the presence of the DATA_ACK is unlikely to be 3252 necessary in a case where SACK is in use, since until at least some 3253 of the SACK blocks have been retransmitted, the cumulative data-level 3254 ACK will not be moving forward (or if it does, due to retransmissions 3255 on another path, then that path can also be used to transmit the new 3256 DATA_ACK). 3258 The ADD_ADDR option can be between 16 and 30 bytes, depending on 3259 whether IPv4 or IPv6 is used, and whether or not the port number is 3260 present. It is unlikely that such signaling would fit in a data 3261 packet (although if there is space, it is fine to include it). It is 3262 recommended to use duplicate ACKs with no other payload or options in 3263 order to transmit these rare signals. Note this is the reason for 3264 mandating that duplicate ACKs with MPTCP options are not taken as a 3265 signal of congestion. 3267 Appendix B. TCP Fast Open and MPTCP 3269 TCP Fast Open (TFO) is an experimental TCP extension, described in 3270 [RFC7413], which has been introduced to allow sending data one RTT 3271 earlier than with regular TCP. This is considered a valuable gain as 3272 very short connections are very common, especially for HTTP request/ 3273 response schemes. It achieves this by sending the SYN-segment 3274 together with the application's data and allowing the listener to 3275 reply immediately with data after the SYN/ACK. [RFC7413] secures 3276 this mechanism, by using a new TCP option that includes a cookie 3277 which is negotiated in a preceding connection. 3279 When using TCP Fast Open in conjunction with MPTCP, there are two key 3280 points to take into account, detailed hereafter. 3282 B.1. TFO cookie request with MPTCP 3284 When a TFO initiator first connects to a listener, it cannot 3285 immediately include data in the SYN for security reasons [RFC7413]. 3286 Instead, it requests a cookie that will be used in subsequent 3287 connections. This is done with the TCP cookie request/response 3288 options, of respectively 2 bytes and 6-18 bytes (depending on the 3289 chosen cookie length). 3291 TFO and MPTCP can be combined provided that the total length of all 3292 the options does not exceed the maximum 40 bytes possible in TCP: 3294 o In the SYN: MPTCP uses a 4-bytes long MP_CAPABLE option. The 3295 MPTCP and TFO options sum up to 6 bytes. With typical TCP-options 3296 using up to 19 bytes in the SYN (24 bytes if options are padded at 3297 a word boundary), there is enough space to combine the MP_CAPABLE 3298 with the TFO Cookie Request. 3300 o In the SYN+ACK: MPTCP uses a 12-bytes long MP_CAPABLE option, but 3301 now TFO can be as long as 18 bytes. Since the maximum option 3302 length may be exceeded, it is up to the listener to solve this by 3303 using a shorter cookie. As an example, if we consider that 19 3304 bytes are used for classical TCP options, the maximum possible 3305 cookie length would be of 7 bytes. Note that the same limitation 3306 applies to subsequent connections, for the SYN packet (because the 3307 initiator then echoes back the cookie to the listener). Finally, 3308 if the security impact of reducing the cookie size is not deemed 3309 acceptable, the listener can reduce the amount of other TCP- 3310 options by omitting the TCP timestamps (as outlined in 3311 Appendix A). 3313 B.2. Data sequence mapping under TFO 3315 MPTCP uses, in the TCP establishment phase, a key exchange that is 3316 used to generate the Initial Data Sequence Numbers (IDSNs). In 3317 particular, the SYN with MP_CAPABLE occupies the first octet of the 3318 data sequence space. With TFO, one way to handle the data sent 3319 together with the SYN would be to consider an implicit DSS mapping 3320 that covers that SYN segment (since there is not enough space in the 3321 SYN to include a DSS option). The problem with that approach is that 3322 if a middlebox modifies the TFO data, this will not be noticed by 3323 MPTCP because of the absence of a DSS-checksum. For example, a TCP 3324 (but not MPTCP)-aware middlebox could insert bytes at the beginning 3325 of the stream and adapt the TCP checksum and sequence numbers 3326 accordingly. With an implicit mapping, this would give to initiator 3327 and listener a different view on the DSS-mapping, with no way to 3328 detect this inconsistency as the DSS checksum is not present. 3330 To solve this, the TFO data must not be considered part of the Data 3331 Sequence Number space: the SYN with MP_CAPABLE still occupies the 3332 first octet of data sequence space, but then the first non-TFO data 3333 byte occupies the second octet. This guarantees that, if the use of 3334 DSS-checksum is negotiated, all data in the data sequence number 3335 space is checksummed. We also note that this does not entail a loss 3336 of functionality, because TFO-data is always only sent on the initial 3337 subflow before any attempt to create additional subflows. 3339 B.3. Connection establishment examples 3341 The following shows a few examples of possible TFO+MPTCP 3342 establishment scenarios. 3344 Before an initiator can send data together with the SYN, it must 3345 request a cookie to the listener, as shown in Figure Figure 18. This 3346 is done by simply combining the TFO and MPTCP options. 3348 initiator listener 3349 | | 3350 | S Seq=0(Length=0) , | 3351 | -----------------------------------------------------------> | 3352 | | 3353 | S. 0(0) ack 1 , | 3354 | <----------------------------------------------------------- | 3355 | | 3356 | . 0(0) ack 1 | 3357 | -----------------------------------------------------------> | 3358 | | 3360 Figure 18: Cookie request - sequence number and length are annotated 3361 as Seq(Length) and used hereafter in the figures. 3363 Once this is done, the received cookie can be used for TFO, as shown 3364 in Figure Figure 19. In this example, the initiator first sends 20 3365 bytes in the SYN. The listener immediately replies with 100 bytes 3366 following the SYN-ACK upon which the initiator replies with 20 more 3367 bytes. Note that the last segment in the figure has a TCP sequence 3368 number of 21, while the DSS subflow sequence number is 1 (because the 3369 TFO data is not part of the data sequence number space, as explained 3370 in Section Appendix B.2. 3372 initiator listener 3373 | | 3374 | S 0(20) , | 3375 | -----------------------------------------------------------> | 3376 | | 3377 | S. 0(0) ack 21 | 3378 | <----------------------------------------------------------- | 3379 | | 3380 | . 1(100) ack 21 | 3381 | <----------------------------------------------------------- | 3382 | | 3383 | . 21(0) ack 1 | 3384 | -----------------------------------------------------------> | 3385 | | 3386 | . 21(20) ack 101 | 3387 | -----------------------------------------------------------> | 3388 | | 3390 Figure 19: The listener supports TFO 3392 In Figure Figure 20, the listener does not support TFO. The 3393 initiator detects that no state is created in the listener (as no 3394 data is acked), and now sends the MP_CAPABLE in the third ack, in 3395 order for the listener to build its MPTCP context at then end of the 3396 establishment. Now, the tfo data, retransmitted, becomes part of the 3397 data sequence mapping because it is effectively sent (in fact re- 3398 sent) after the establishment. 3400 initiator listener 3401 | | 3402 | S 0(20) , | 3403 | -----------------------------------------------------------> | 3404 | | 3405 | S. 0(0) ack 1 | 3406 | <----------------------------------------------------------- | 3407 | | 3408 | . 1(0) ack 1 | 3409 | -----------------------------------------------------------> | 3410 | | 3411 | . 1(20) ack 1 | 3412 | -----------------------------------------------------------> | 3413 | | 3414 | . 0(0) ack 21 | 3415 | <----------------------------------------------------------- | 3416 | | 3418 Figure 20: The listener does not support TFO 3420 It is also possible that the listener acknowledges only part of the 3421 TFO data, as illustrated in Figure Figure 21. The initiator will 3422 simply retransmit the missing data together with a DSS-mapping. 3424 initiator listener 3425 | | 3426 | S 0(1000) , | 3427 | -----------------------------------------------------------> | 3428 | | 3429 | S. 0(0) ack 501 | 3430 | <----------------------------------------------------------- | 3431 | | 3432 | . 501(0) ack 1 | 3433 | -----------------------------------------------------------> | 3434 | | 3435 | . 501(500) ack 1 | 3436 | -----------------------------------------------------------> | 3437 | | 3439 Figure 21: Partial data acknowledgement 3441 Appendix C. Control Blocks 3443 Conceptually, an MPTCP connection can be represented as an MPTCP 3444 protocol control block (PCB) that contains several variables that 3445 track the progress and the state of the MPTCP connection and a set of 3446 linked TCP control blocks that correspond to the subflows that have 3447 been established. 3449 RFC 793 [RFC0793] specifies several state variables. Whenever 3450 possible, we reuse the same terminology as RFC 793 to describe the 3451 state variables that are maintained by MPTCP. 3453 C.1. MPTCP Control Block 3455 The MPTCP control block contains the following variable per 3456 connection. 3458 C.1.1. Authentication and Metadata 3460 Local.Token (32 bits): This is the token chosen by the local host on 3461 this MPTCP connection. The token must be unique among all 3462 established MPTCP connections, generated from the local key. 3464 Local.Key (64 bits): This is the key sent by the local host on this 3465 MPTCP connection. 3467 Remote.Token (32 bits): This is the token chosen by the remote host 3468 on this MPTCP connection, generated from the remote key. 3470 Remote.Key (64 bits): This is the key chosen by the remote host on 3471 this MPTCP connection 3473 MPTCP.Checksum (flag): This flag is set to true if at least one of 3474 the hosts has set the A bit in the MP_CAPABLE options exchanged 3475 during connection establishment, and is set to false otherwise. 3476 If this flag is set, the checksum must be computed in all DSS 3477 options. 3479 C.1.2. Sending Side 3481 SND.UNA (64 bits): This is the data sequence number of the next byte 3482 to be acknowledged, at the MPTCP connection level. This variable 3483 is updated upon reception of a DSS option containing a DATA_ACK. 3485 SND.NXT (64 bits): This is the data sequence number of the next byte 3486 to be sent. SND.NXT is used to determine the value of the DSN in 3487 the DSS option. 3489 SND.WND (32 bits with RFC 7323, 16 bits otherwise): This is the 3490 sending window. MPTCP maintains the sending window at the MPTCP 3491 connection level and the same window is shared by all subflows. 3492 All subflows use the MPTCP connection level SND.WND to compute the 3493 SEQ.WND value that is sent in each transmitted segment. 3495 C.1.3. Receiving Side 3497 RCV.NXT (64 bits): This is the data sequence number of the next byte 3498 that is expected on the MPTCP connection. This state variable is 3499 modified upon reception of in-order data. The value of RCV.NXT is 3500 used to specify the DATA_ACK that is sent in the DSS option on all 3501 subflows. 3503 RCV.WND (32 bits with RFC 7323, 16 bits otherwise): This is the 3504 connection-level receive window, which is the maximum of the 3505 RCV.WND on all the subflows. 3507 C.2. TCP Control Blocks 3509 The MPTCP control block also contains a list of the TCP control 3510 blocks that are associated with the MPTCP connection. 3512 Note that the TCP control block on the TCP subflows does not contain 3513 the RCV.WND and SND.WND state variables as these are maintained at 3514 the MPTCP connection level and not at the subflow level. 3516 Inside each TCP control block, the following state variables are 3517 defined. 3519 C.2.1. Sending Side 3521 SND.UNA (32 bits): This is the sequence number of the next byte to 3522 be acknowledged on the subflow. This variable is updated upon 3523 reception of each TCP acknowledgment on the subflow. 3525 SND.NXT (32 bits): This is the sequence number of the next byte to 3526 be sent on the subflow. SND.NXT is used to set the value of 3527 SEG.SEQ upon transmission of the next segment. 3529 C.2.2. Receiving Side 3531 RCV.NXT (32 bits): This is the sequence number of the next byte that 3532 is expected on the subflow. This state variable is modified upon 3533 reception of in-order segments. The value of RCV.NXT is copied to 3534 the SEG.ACK field of the next segments transmitted on the subflow. 3536 RCV.WND (32 bits with RFC 7323, 16 bits otherwise): This is the 3537 subflow-level receive window that is updated with the window field 3538 from the segments received on this subflow. 3540 Appendix D. Finite State Machine 3542 The diagram in Figure 22 shows the Finite State Machine for 3543 connection-level closure. This illustrates how the DATA_FIN 3544 connection-level signal (indicated in the diagram as the DFIN flag on 3545 a DATA_ACK) interacts with subflow-level FINs, and permits "break- 3546 before-make" handover between subflows. 3548 +---------+ 3549 | M_ESTAB | 3550 +---------+ 3551 M_CLOSE | | rcv DATA_FIN 3552 ------- | | ------- 3553 +---------+ snd DATA_FIN / \ snd DATA_ACK[DFIN] +---------+ 3554 | M_FIN |<----------------- ------------------->| M_CLOSE | 3555 | WAIT-1 |--------------------------- | WAIT | 3556 +---------+ rcv DATA_FIN \ +---------+ 3557 | rcv DATA_ACK[DFIN] ------- | M_CLOSE | 3558 | -------------- snd DATA_ACK | ------- | 3559 | CLOSE all subflows | snd DATA_FIN | 3560 V V V 3561 +-----------+ +-----------+ +-----------+ 3562 |M_FINWAIT-2| | M_CLOSING | | M_LAST-ACK| 3563 +-----------+ +-----------+ +-----------+ 3564 | rcv DATA_ACK[DFIN] | rcv DATA_ACK[DFIN] | 3565 | rcv DATA_FIN -------------- | -------------- | 3566 | ------- CLOSE all subflows | CLOSE all subflows | 3567 | snd DATA_ACK[DFIN] V delete MPTCP PCB V 3568 \ +-----------+ +---------+ 3569 ------------------------>|M_TIME WAIT|----------------->| M_CLOSED| 3570 +-----------+ +---------+ 3571 All subflows in CLOSED 3572 ------------ 3573 delete MPTCP PCB 3575 Figure 22: Finite State Machine for Connection Closure 3577 Appendix E. Changes from RFC6184 3579 This section lists the key technical changes between RFC6824 3580 [RFC6824], specifying MPTCP v0, and this document, which obsoletes 3581 RFC6824 and specifies MPTCP v1. Note that this specification is not 3582 backwards compatible with RFC6824. 3584 o The document incorporates lessons learnt from the various 3585 implementations, deployments and experiments gathered in the 3586 documents "Use Cases and Operational Experience with Multipath 3587 TCP" [RFC8041] and the IETF Journal article "Multipath TCP 3588 Deployments" [deployments]. 3590 o Connection initiation, through the exchange of the MP_CAPABLE 3591 MPTCP option, is different from RFC6824. In order to permit 3592 servers to act statelessly, the SYN doesn't include A's key (it is 3593 still sent in the ACK). 3595 o This requires MP_CAPABLE to also be sent reliably on the third 3596 ACK. If safe receipt of the third ACK cannot be inferred, the 3597 MP_CAPABLE option must be repeated on the first data packet. 3599 o In the Flags field of MP_CAPABLE, C is now assigned to mean that 3600 the sender of this option will not accept additional MPTCP 3601 subflows to the source address and port. This is an efficiency 3602 improvement, for example where the sender is behind a strict NAT. 3604 o In the Flags field of MP_CAPABLE, H now indicates the use of HMAC- 3605 SHA256 (rather than HMAC-SHA1). 3607 o Connection initiation also defines the procedure for version 3608 negotiation, for implementations that support both v0 (RFC6824) 3609 and v1 (this document). 3611 o The HMAC-SHA256 (rather than HMAC-SHA1) algorithm is used, as the 3612 algorithm provides better security. It is used to generate the 3613 token in the MP_JOIN and ADD_ADDR messages, and to set the initial 3614 data sequence number. 3616 o A new subflow-level option exists to signal reasons for sending a 3617 RST on a subflow (MP_TCPRST Section 3.6), which can help an 3618 implementation decide whether to attempt later re-connection. 3620 o The MP_PRIO option (Section 3.3.8), which is used to signal a 3621 change of priority for a subflow, no longer includes the AddrID 3622 field. Its purpose was to allow the changed priority to be 3623 applied on a subflow other than the one it was sent on. However, 3624 it has been realised that this could be used by a man-in-the- 3625 middle to divert all traffic on to its own path, and MP_PRIO does 3626 not include a token or other security mechanism. 3628 o The ADD_ADDR option (Section 3.4.1), which is used to inform the 3629 other host about another potential address, is different in 3630 several ways. It now includes an HMAC of the added address, for 3631 enhanced security. In addition, reliability for the ADD_ADDR 3632 option has been added: the IPVer field is replaced with a flag 3633 field, and one flag is assigned (E) which is used as an 'Echo' so 3634 a host can indicate that it has received the option. 3636 o An additional way of performing a Fast Close is described, by 3637 sending a MP_FASTCLOSE option on a RST on all subflows. This 3638 allows the host to tear down the subflows and the connection 3639 immediately. 3641 o In the IANA registry a new MPTCP subtype option, MP_EXPERIMENTAL, 3642 is reserved for private experiments. However, the document 3643 doesn't define how to use the subtype option. 3645 o A new Appendix discusses the usage of both the MPTCP and TCP Fast 3646 Open on the same packet (Appendix B). 3648 Authors' Addresses 3650 Alan Ford 3651 Pexip 3653 EMail: alan.ford@gmail.com 3655 Costin Raiciu 3656 University Politehnica of Bucharest 3657 Splaiul Independentei 313 3658 Bucharest 3659 Romania 3661 EMail: costin.raiciu@cs.pub.ro 3663 Mark Handley 3664 University College London 3665 Gower Street 3666 London WC1E 6BT 3667 UK 3669 EMail: m.handley@cs.ucl.ac.uk 3671 Olivier Bonaventure 3672 Universite catholique de Louvain 3673 Pl. Ste Barbe, 2 3674 Louvain-la-Neuve 1348 3675 Belgium 3677 EMail: olivier.bonaventure@uclouvain.be 3679 Christoph Paasch 3680 Apple, Inc. 3681 Cupertino 3682 US 3684 EMail: cpaasch@apple.com