idnits 2.17.1 draft-ietf-mptcp-rfc6824bis-09.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 abstract seems to contain references ([RFC6824]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. -- 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 date (July 28, 2017) is 2465 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 492, but not defined == Missing Reference: 'Checksum' is mentioned on line 493, but not defined == Missing Reference: 'Data FIN' is mentioned on line 521, but not defined == Missing Reference: 'DFIN' is mentioned on line 3345, 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 1323 (Obsoleted by RFC 7323) -- Obsolete informational reference (is this intentional?): RFC 5226 (Obsoleted by RFC 8126) -- 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: 3 errors (**), 0 flaws (~~), 5 warnings (==), 7 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: January 29, 2018 M. Handley 7 U. College London 8 O. Bonaventure 9 U. catholique de Louvain 10 C. Paasch 11 Apple, Inc. 12 July 28, 2017 14 TCP Extensions for Multipath Operation with Multiple Addresses 15 draft-ietf-mptcp-rfc6824bis-09 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 [RFC6824] through clarifications and 35 modifications 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 January 29, 2018. 54 Copyright Notice 56 Copyright (c) 2017 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 . . . . . . . . . . . . . . . . . . . . . . 6 76 1.5. Requirements Language . . . . . . . . . . . . . . . . . . 8 77 2. Operation Overview . . . . . . . . . . . . . . . . . . . . . 8 78 2.1. Initiating an MPTCP Connection . . . . . . . . . . . . . 8 79 2.2. Associating a New Subflow with an Existing MPTCP 80 Connection . . . . . . . . . . . . . . . . . . . . . . . 9 81 2.3. Informing the Other Host about Another Potential Address 9 82 2.4. Data Transfer Using MPTCP . . . . . . . . . . . . . . . . 10 83 2.5. Requesting a Change in a Path's Priority . . . . . . . . 11 84 2.6. Closing an MPTCP Connection . . . . . . . . . . . . . . . 11 85 2.7. Notable Features . . . . . . . . . . . . . . . . . . . . 12 86 3. MPTCP Protocol . . . . . . . . . . . . . . . . . . . . . . . 12 87 3.1. Connection Initiation . . . . . . . . . . . . . . . . . . 13 88 3.2. Starting a New Subflow . . . . . . . . . . . . . . . . . 20 89 3.3. General MPTCP Operation . . . . . . . . . . . . . . . . . 25 90 3.3.1. Data Sequence Mapping . . . . . . . . . . . . . . . . 27 91 3.3.2. Data Acknowledgments . . . . . . . . . . . . . . . . 30 92 3.3.3. Closing a Connection . . . . . . . . . . . . . . . . 31 93 3.3.4. Receiver Considerations . . . . . . . . . . . . . . . 32 94 3.3.5. Sender Considerations . . . . . . . . . . . . . . . . 33 95 3.3.6. Reliability and Retransmissions . . . . . . . . . . . 34 96 3.3.7. Congestion Control Considerations . . . . . . . . . . 35 97 3.3.8. Subflow Policy . . . . . . . . . . . . . . . . . . . 36 98 3.4. Address Knowledge Exchange (Path Management) . . . . . . 37 99 3.4.1. Address Advertisement . . . . . . . . . . . . . . . . 38 100 3.4.2. Remove Address . . . . . . . . . . . . . . . . . . . 42 101 3.5. Fast Close . . . . . . . . . . . . . . . . . . . . . . . 43 102 3.6. Subflow Reset . . . . . . . . . . . . . . . . . . . . . . 44 103 3.7. MPTCP Experimental Option . . . . . . . . . . . . . . . . 46 104 3.8. Fallback . . . . . . . . . . . . . . . . . . . . . . . . 47 105 3.9. Error Handling . . . . . . . . . . . . . . . . . . . . . 51 106 3.10. Heuristics . . . . . . . . . . . . . . . . . . . . . . . 51 107 3.10.1. Port Usage . . . . . . . . . . . . . . . . . . . . . 52 108 3.10.2. Delayed Subflow Start and Subflow Symmetry . . . . . 52 109 3.10.3. Failure Handling . . . . . . . . . . . . . . . . . . 53 110 4. Semantic Issues . . . . . . . . . . . . . . . . . . . . . . . 54 111 5. Security Considerations . . . . . . . . . . . . . . . . . . . 55 112 6. Interactions with Middleboxes . . . . . . . . . . . . . . . . 58 113 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 61 114 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 61 115 8.1. MPTCP Option Subtypes . . . . . . . . . . . . . . . . . . 62 116 8.2. MPTCP Handshake Algorithms . . . . . . . . . . . . . . . 63 117 8.3. MP_TCPRST Reason Codes . . . . . . . . . . . . . . . . . 63 118 8.4. Experimental option registry . . . . . . . . . . . . . . 64 119 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 64 120 9.1. Normative References . . . . . . . . . . . . . . . . . . 64 121 9.2. Informative References . . . . . . . . . . . . . . . . . 65 122 Appendix A. Notes on Use of TCP Options . . . . . . . . . . . . 68 123 Appendix B. Control Blocks . . . . . . . . . . . . . . . . . . . 69 124 B.1. MPTCP Control Block . . . . . . . . . . . . . . . . . . . 70 125 B.1.1. Authentication and Metadata . . . . . . . . . . . . . 70 126 B.1.2. Sending Side . . . . . . . . . . . . . . . . . . . . 70 127 B.1.3. Receiving Side . . . . . . . . . . . . . . . . . . . 70 128 B.2. TCP Control Blocks . . . . . . . . . . . . . . . . . . . 71 129 B.2.1. Sending Side . . . . . . . . . . . . . . . . . . . . 71 130 B.2.2. Receiving Side . . . . . . . . . . . . . . . . . . . 71 131 Appendix C. Finite State Machine . . . . . . . . . . . . . . . . 71 132 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 72 134 1. Introduction 136 Multipath TCP (MPTCP) is a set of extensions to regular TCP [RFC0793] 137 to provide a Multipath TCP [RFC6182] service, which enables a 138 transport connection to operate across multiple paths simultaneously. 139 This document presents the protocol changes required to add multipath 140 capability to TCP; specifically, those for signaling and setting up 141 multiple paths ("subflows"), managing these subflows, reassembly of 142 data, and termination of sessions. This is not the only information 143 required to create a Multipath TCP implementation, however. This 144 document is complemented by three others: 146 o Architecture [RFC6182], which explains the motivations behind 147 Multipath TCP, contains a discussion of high-level design 148 decisions on which this design is based, and an explanation of a 149 functional separation through which an extensible MPTCP 150 implementation can be developed. 152 o Congestion control [RFC6356] presents a safe congestion control 153 algorithm for coupling the behavior of the multiple paths in order 154 to "do no harm" to other network users. 156 o Application considerations [RFC6897] discusses what impact MPTCP 157 will have on applications, what applications will want to do with 158 MPTCP, and as a consequence of these factors, what API extensions 159 an MPTCP implementation should present. 161 This document is an update to, and obsoletes, the v0 specification of 162 Multipath TCP [RFC6824]. This document specifies MPTCP v1, which is 163 not backward compatible with MPTCP v0. This document additionally 164 defines version negotiation procedures for implementations that 165 support both versions. 167 1.1. Design Assumptions 169 In order to limit the potentially huge design space, the working 170 group imposed two key constraints on the Multipath TCP design 171 presented in this document: 173 o It must be backwards-compatible with current, regular TCP, to 174 increase its chances of deployment. 176 o It can be assumed that one or both hosts are multihomed and 177 multiaddressed. 179 To simplify the design, we assume that the presence of multiple 180 addresses at a host is sufficient to indicate the existence of 181 multiple paths. These paths need not be entirely disjoint: they may 182 share one or many routers between them. Even in such a situation, 183 making use of multiple paths is beneficial, improving resource 184 utilization and resilience to a subset of node failures. The 185 congestion control algorithms defined in [RFC6356] ensure this does 186 not act detrimentally. Furthermore, there may be some scenarios 187 where different TCP ports on a single host can provide disjoint paths 188 (such as through certain Equal-Cost Multipath (ECMP) implementations 189 [RFC2992]), and so the MPTCP design also supports the use of ports in 190 path identifiers. 192 There are three aspects to the backwards-compatibility listed above 193 (discussed in more detail in [RFC6182]): 195 External Constraints: The protocol must function through the vast 196 majority of existing middleboxes such as NATs, firewalls, and 197 proxies, and as such must resemble existing TCP as far as possible 198 on the wire. Furthermore, the protocol must not assume the 199 segments it sends on the wire arrive unmodified at the 200 destination: they may be split or coalesced; TCP options may be 201 removed or duplicated. 203 Application Constraints: The protocol must be usable with no change 204 to existing applications that use the common TCP API (although it 205 is reasonable that not all features would be available to such 206 legacy applications). Furthermore, the protocol must provide the 207 same service model as regular TCP to the application. 209 Fallback: The protocol should be able to fall back to standard TCP 210 with no interference from the user, to be able to communicate with 211 legacy hosts. 213 The complementary application considerations document [RFC6897] 214 discusses the necessary features of an API to provide backwards- 215 compatibility, as well as API extensions to convey the behavior of 216 MPTCP at a level of control and information equivalent to that 217 available with regular, single-path TCP. 219 Further discussion of the design constraints and associated design 220 decisions are given in the MPTCP Architecture document [RFC6182] and 221 in [howhard]. 223 1.2. Multipath TCP in the Networking Stack 225 MPTCP operates at the transport layer and aims to be transparent to 226 both higher and lower layers. It is a set of additional features on 227 top of standard TCP; Figure 1 illustrates this layering. MPTCP is 228 designed to be usable by legacy applications with no changes; 229 detailed discussion of its interactions with applications is given in 230 [RFC6897]. 232 +-------------------------------+ 233 | Application | 234 +---------------+ +-------------------------------+ 235 | Application | | MPTCP | 236 +---------------+ + - - - - - - - + - - - - - - - + 237 | TCP | | Subflow (TCP) | Subflow (TCP) | 238 +---------------+ +-------------------------------+ 239 | IP | | IP | IP | 240 +---------------+ +-------------------------------+ 242 Figure 1: Comparison of Standard TCP and MPTCP Protocol Stacks 244 1.3. Terminology 246 This document makes use of a number of terms that are either MPTCP- 247 specific or have defined meaning in the context of MPTCP, as follows: 249 Path: A sequence of links between a sender and a receiver, defined 250 in this context by a 4-tuple of source and destination address/ 251 port pairs. 253 Subflow: A flow of TCP segments operating over an individual path, 254 which forms part of a larger MPTCP connection. A subflow is 255 started and terminated similar to a regular TCP connection. 257 (MPTCP) Connection: A set of one or more subflows, over which an 258 application can communicate between two hosts. There is a one-to- 259 one mapping between a connection and an application socket. 261 Data-level: The payload data is nominally transferred over a 262 connection, which in turn is transported over subflows. Thus, the 263 term "data-level" is synonymous with "connection level", in 264 contrast to "subflow-level", which refers to properties of an 265 individual subflow. 267 Token: A locally unique identifier given to a multipath connection 268 by a host. May also be referred to as a "Connection ID". 270 Host: An end host operating an MPTCP implementation, and either 271 initiating or accepting an MPTCP connection. 273 In addition to these terms, note that MPTCP's interpretation of, and 274 effect on, regular single-path TCP semantics are discussed in 275 Section 4. 277 1.4. MPTCP Concept 279 This section provides a high-level summary of normal operation of 280 MPTCP, and is illustrated by the scenario shown in Figure 2. A 281 detailed description of operation is given in Section 3. 283 o To a non-MPTCP-aware application, MPTCP will behave the same as 284 normal TCP. Extended APIs could provide additional control to 285 MPTCP-aware applications [RFC6897]. An application begins by 286 opening a TCP socket in the normal way. MPTCP signaling and 287 operation are handled by the MPTCP implementation. 289 o An MPTCP connection begins similarly to a regular TCP connection. 290 This is illustrated in Figure 2 where an MPTCP connection is 291 established between addresses A1 and B1 on Hosts A and B, 292 respectively. 294 o If extra paths are available, additional TCP sessions (termed 295 MPTCP "subflows") are created on these paths, and are combined 296 with the existing session, which continues to appear as a single 297 connection to the applications at both ends. The creation of the 298 additional TCP session is illustrated between Address A2 on Host A 299 and Address B1 on Host B. 301 o MPTCP identifies multiple paths by the presence of multiple 302 addresses at hosts. Combinations of these multiple addresses 303 equate to the additional paths. In the example, other potential 304 paths that could be set up are A1<->B2 and A2<->B2. Although this 305 additional session is shown as being initiated from A2, it could 306 equally have been initiated from B1. 308 o The discovery and setup of additional subflows will be achieved 309 through a path management method; this document describes a 310 mechanism by which a host can initiate new subflows by using its 311 own additional addresses, or by signaling its available addresses 312 to the other host. 314 o MPTCP adds connection-level sequence numbers to allow the 315 reassembly of segments arriving on multiple subflows with 316 differing network delays. 318 o Subflows are terminated as regular TCP connections, with a four- 319 way FIN handshake. The MPTCP connection is terminated by a 320 connection-level FIN. 322 Host A Host B 323 ------------------------ ------------------------ 324 Address A1 Address A2 Address B1 Address B2 325 ---------- ---------- ---------- ---------- 326 | | | | 327 | (initial connection setup) | | 328 |----------------------------------->| | 329 |<-----------------------------------| | 330 | | | | 331 | (additional subflow setup) | 332 | |--------------------->| | 333 | |<---------------------| | 334 | | | | 335 | | | | 337 Figure 2: Example MPTCP Usage Scenario 339 1.5. Requirements Language 341 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 342 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 343 document are to be interpreted as described in RFC 2119 [RFC2119]. 345 2. Operation Overview 347 This section presents a single description of common MPTCP operation, 348 with reference to the protocol operation. This is a high-level 349 overview of the key functions; the full specification follows in 350 Section 3. Extensibility and negotiated features are not discussed 351 here. Considerable reference is made to symbolic names of MPTCP 352 options throughout this section -- these are subtypes of the IANA- 353 assigned MPTCP option (see Section 8), and their formats are defined 354 in the detailed protocol specification that follows in Section 3. 356 A Multipath TCP connection provides a bidirectional bytestream 357 between two hosts communicating like normal TCP and, thus, does not 358 require any change to the applications. However, Multipath TCP 359 enables the hosts to use different paths with different IP addresses 360 to exchange packets belonging to the MPTCP connection. A Multipath 361 TCP connection appears like a normal TCP connection to an 362 application. However, to the network layer, each MPTCP subflow looks 363 like a regular TCP flow whose segments carry a new TCP option type. 364 Multipath TCP manages the creation, removal, and utilization of these 365 subflows to send data. The number of subflows that are managed 366 within a Multipath TCP connection is not fixed and it can fluctuate 367 during the lifetime of the Multipath TCP connection. 369 All MPTCP operations are signaled with a TCP option -- a single 370 numerical type for MPTCP, with "sub-types" for each MPTCP message. 371 What follows is a summary of the purpose and rationale of these 372 messages. 374 2.1. Initiating an MPTCP Connection 376 This is the same signaling as for initiating a normal TCP connection, 377 but the SYN, SYN/ACK, and initial ACK packets also carry the 378 MP_CAPABLE option. This option is variable length and serves 379 multiple purposes. Firstly, it verifies whether the remote host 380 supports Multipath TCP; secondly, this option allows the hosts to 381 exchange some information to authenticate the establishment of 382 additional subflows. Further details are given in Section 3.1. 384 Host A Host B 385 ------ ------ 386 MP_CAPABLE -> 387 [flags] 388 <- MP_CAPABLE 389 [B's key, flags] 390 ACK + MP_CAPABLE (+ data) -> 391 [A's key, B's key, flags, (data-level details)] 393 2.2. Associating a New Subflow with an Existing MPTCP Connection 395 The exchange of keys in the MP_CAPABLE handshake provides material 396 that can be used to authenticate the endpoints when new subflows will 397 be set up. Additional subflows begin in the same way as initiating a 398 normal TCP connection, but the SYN, SYN/ACK, and ACK packets also 399 carry the MP_JOIN option. 401 Host A initiates a new subflow between one of its addresses and one 402 of Host B's addresses. The token -- generated from the key -- is 403 used to identify which MPTCP connection it is joining, and the HMAC 404 is used for authentication. The Hash-based Message Authentication 405 Code (HMAC) uses the keys exchanged in the MP_CAPABLE handshake, and 406 the random numbers (nonces) exchanged in these MP_JOIN options. 407 MP_JOIN also contains flags and an Address ID that can be used to 408 refer to the source address without the sender needing to know if it 409 has been changed by a NAT. Further details are in Section 3.2. 411 Host A Host B 412 ------ ------ 413 MP_JOIN -> 414 [B's token, A's nonce, 415 A's Address ID, flags] 416 <- MP_JOIN 417 [B's HMAC, B's nonce, 418 B's Address ID, flags] 419 ACK + MP_JOIN -> 420 [A's HMAC] 422 <- ACK 424 2.3. Informing the Other Host about Another Potential Address 426 The set of IP addresses associated to a multihomed host may change 427 during the lifetime of an MPTCP connection. MPTCP supports the 428 addition and removal of addresses on a host both implicitly and 429 explicitly. If Host A has established a subflow starting at address/ 430 port pair IP#-A1 and wants to open a second subflow starting at 431 address/port pair IP#-A2, it simply initiates the establishment of 432 the subflow as explained above. The remote host will then be 433 implicitly informed about the new address. 435 In some circumstances, a host may want to advertise to the remote 436 host the availability of an address without establishing a new 437 subflow, for example, when a NAT prevents setup in one direction. In 438 the example below, Host A informs Host B about its alternative IP 439 address/port pair (IP#-A2). Host B may later send an MP_JOIN to this 440 new address. This option contains a HMAC to authenticate the address 441 as having been sent from the originator of the connection. Further 442 details are in Section 3.4.1. 444 Host A Host B 445 ------ ------ 446 ADD_ADDR -> 447 [IP#-A2, 448 IP#-A2's Address ID, 449 HMAC of IP#-A2] 451 There is a corresponding signal for address removal, making use of 452 the Address ID that is signaled in the add address handshake. 453 Further details in Section 3.4.2. 455 Host A Host B 456 ------ ------ 457 REMOVE_ADDR -> 458 [IP#-A2's Address ID] 460 2.4. Data Transfer Using MPTCP 462 To ensure reliable, in-order delivery of data over subflows that may 463 appear and disappear at any time, MPTCP uses a 64-bit data sequence 464 number (DSN) to number all data sent over the MPTCP connection. Each 465 subflow has its own 32-bit sequence number space, utilising the 466 regular TCP sequence number header, and an MPTCP option maps the 467 subflow sequence space to the data sequence space. In this way, data 468 can be retransmitted on different subflows (mapped to the same DSN) 469 in the event of failure. 471 The "Data Sequence Signal" carries the "Data Sequence Mapping". The 472 data sequence mapping consists of the subflow sequence number, data 473 sequence number, and length for which this mapping is valid. This 474 option can also carry a connection-level acknowledgment (the "Data 475 ACK") for the received DSN. 477 With MPTCP, all subflows share the same receive buffer and advertise 478 the same receive window. There are two levels of acknowledgment in 479 MPTCP. Regular TCP acknowledgments are used on each subflow to 480 acknowledge the reception of the segments sent over the subflow 481 independently of their DSN. In addition, there are connection-level 482 acknowledgments for the data sequence space. These acknowledgments 483 track the advancement of the bytestream and slide the receiving 484 window. 486 Further details are in Section 3.3. 488 Host A Host B 489 ------ ------ 490 DATA_SEQUENCE_SIGNAL -> 491 [Data Sequence Mapping] 492 [Data ACK] 493 [Checksum] 495 2.5. Requesting a Change in a Path's Priority 497 Hosts can indicate at initial subflow setup whether they wish the 498 subflow to be used as a regular or backup path -- a backup path only 499 being used if there are no regular paths available. During a 500 connection, Host A can request a change in the priority of a subflow 501 through the MP_PRIO signal to Host B. Further details are in 502 Section 3.3.8. 504 Host A Host B 505 ------ ------ 506 MP_PRIO -> 508 2.6. Closing an MPTCP Connection 510 When Host A wants to inform Host B that it has no more data to send, 511 it signals this "Data FIN" as part of the Data Sequence Signal (see 512 above). It has the same semantics and behavior as a regular TCP FIN, 513 but at the connection level. Once all the data on the MPTCP 514 connection has been successfully received, then this message is 515 acknowledged at the connection level with a DATA_ACK. Further 516 details are in Section 3.3.3. 518 Host A Host B 519 ------ ------ 520 DATA_SEQUENCE_SIGNAL -> 521 [Data FIN] 523 <- (MPTCP DATA_ACK) 525 2.7. Notable Features 527 It is worth highlighting that MPTCP's signaling has been designed 528 with several key requirements in mind: 530 o To cope with NATs on the path, addresses are referred to by 531 Address IDs, in case the IP packet's source address gets changed 532 by a NAT. Setting up a new TCP flow is not possible if the 533 passive opener is behind a NAT; to allow subflows to be created 534 when either end is behind a NAT, MPTCP uses the ADD_ADDR message. 536 o MPTCP falls back to ordinary TCP if MPTCP operation is not 537 possible, for example, if one host is not MPTCP capable or if a 538 middlebox alters the payload. 540 o To meet the threats identified in [RFC6181], the following steps 541 are taken: keys are sent in the clear in the MP_CAPABLE messages; 542 MP_JOIN messages are secured with HMAC-SHA256 ([RFC2104], [SHS]) 543 using those keys; and standard TCP validity checks are made on the 544 other messages (ensuring sequence numbers are in-window 545 [RFC5961]). 547 3. MPTCP Protocol 549 This section describes the operation of the MPTCP protocol, and is 550 subdivided into sections for each key part of the protocol operation. 552 All MPTCP operations are signaled using optional TCP header fields. 553 A single TCP option number ("Kind") has been assigned by IANA for 554 MPTCP (see Section 8), and then individual messages will be 555 determined by a "subtype", the values of which are also stored in an 556 IANA registry (and are also listed in Section 8). 558 Throughout this document, when reference is made to an MPTCP option 559 by symbolic name, such as "MP_CAPABLE", this refers to a TCP option 560 with the single MPTCP option type, and with the subtype value of the 561 symbolic name as defined in Section 8. This subtype is a 4-bit field 562 -- the first 4 bits of the option payload, as shown in Figure 3. The 563 MPTCP messages are defined in the following sections. 565 1 2 3 566 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 1 567 +---------------+---------------+-------+-----------------------+ 568 | Kind | Length |Subtype| | 569 +---------------+---------------+-------+ | 570 | Subtype-specific data | 571 | (variable length) | 572 +---------------------------------------------------------------+ 574 Figure 3: MPTCP Option Format 576 Those MPTCP options associated with subflow initiation are used on 577 packets with the SYN flag set. Additionally, there is one MPTCP 578 option for signaling metadata to ensure segmented data can be 579 recombined for delivery to the application. 581 The remaining options, however, are signals that do not need to be on 582 a specific packet, such as those for signaling additional addresses. 583 Whilst an implementation may desire to send MPTCP options as soon as 584 possible, it may not be possible to combine all desired options (both 585 those for MPTCP and for regular TCP, such as SACK (selective 586 acknowledgment) [RFC2018]) on a single packet. Therefore, an 587 implementation may choose to send duplicate ACKs containing the 588 additional signaling information. This changes the semantics of a 589 duplicate ACK; these are usually only sent as a signal of a lost 590 segment [RFC5681] in regular TCP. Therefore, an MPTCP implementation 591 receiving a duplicate ACK that contains an MPTCP option MUST NOT 592 treat it as a signal of congestion. Additionally, an MPTCP 593 implementation SHOULD NOT send more than two duplicate ACKs in a row 594 for the purposes of sending MPTCP options alone, in order to ensure 595 no middleboxes misinterpret this as a sign of congestion. 597 Furthermore, standard TCP validity checks (such as ensuring the 598 sequence number and acknowledgment number are within window) MUST be 599 undertaken before processing any MPTCP signals, as described in 600 [RFC5961], and initial subfow sequence numbers SHOULD be generated 601 according to the recommendations in [RFC6528]. 603 3.1. Connection Initiation 605 Connection initiation begins with a SYN, SYN/ACK, ACK exchange on a 606 single path. Each packet contains the Multipath Capable (MP_CAPABLE) 607 MPTCP option (Figure 4). This option declares its sender is capable 608 of performing Multipath TCP and wishes to do so on this particular 609 connection. 611 The MP_CAPABLE exchange in this specification (v1) is different to 612 that specified in v0 [RFC6824]. If a host supports multiple versions 613 of MPTCP, the sender of the MP_CAPABLE option SHOULD signal the 614 highest version number it supports. The passive opener, on receipt 615 of this, will signal the version number it wishes to use, which MUST 616 be equal to or lower than the version number indicated in the initial 617 MP_CAPABLE. Given the SYN exchange is different between v1 and v0 618 the exchange cannot be immediately downgraded, and therefore if the 619 far end has requested a lower version then the initiator SHOULD 620 respond with an ACK without any MP_CAPABLE option, to fall back to 621 regular TCP. If the initiator supports the requsted version, on 622 future connections to the target host, the initiator MAY cache the 623 version preference. Alternatively, the initiator MAY close the 624 connection with a TCP RST and immediately re-establish with the 625 requested version of MPTCP. 627 The MP_CAPABLE option is variable-length, with different fields 628 included depending on which packet the option is used on. The full 629 MP_CAPABLE option is shown in Figure 4. 631 1 2 3 632 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 1 633 +---------------+---------------+-------+-------+---------------+ 634 | Kind | Length |Subtype|Version|A|B|C|D|E|F|G|H| 635 +---------------+---------------+-------+-------+---------------+ 636 | Option Sender's Key (64 bits) | 637 | (if option Length > 4) | 638 | | 639 +---------------------------------------------------------------+ 640 | Option Receiver's Key (64 bits) | 641 | (if option Length > 12) | 642 | | 643 +-------------------------------+-------------------------------+ 644 | Data-Level Length (16 bits) | Checksum (16 bits, optional) | 645 +-------------------------------+-------------------------------+ 647 Figure 4: Multipath Capable (MP_CAPABLE) Option 649 The MP_CAPABLE option is carried on the SYN, SYN/ACK, and ACK packets 650 that start the first subflow of an MPTCP connection, as well as the 651 first packet that carries data, if the initiator wishs to send first. 652 The data carried by each option is as follows, where A = initiator 653 and B = listener. 655 o SYN (A->B): only the first four octets (Length = 4). 657 o SYN/ACK (B->A): B's Key for this connection (Length = 12). 659 o ACK (no data) (A->B): A's Key followed by B's Key (Length = 20). 661 o ACK (with first data) (A->B): A's Key followed by B's Key followed 662 by Data-Level Length, and optional Checksum (Length = 22 or 24). 664 The contents of the option is determined by the SYN and ACK flags of 665 the packet, along with the option's length field. For the diagram 666 shown in Figure 4, "sender" and "receiver" refer to the sender or 667 receiver of the TCP packet (which can be either host). 669 The initial SYN, containing just the MP_CAPABLE header, is used to 670 define the version of MPTCP beign requested, as well as exchanging 671 flags to negotiate connection features, described later. 673 This option is used to declare the 64-bit keys that the end hosts 674 have generated for this MPTCP connection. This key is used to 675 authenticate the addition of future subflows to this connection. 676 This is the only time the key will be sent in clear on the wire 677 (unless "fast close", Section 3.5, is used); all future subflows will 678 identify the connection using a 32-bit "token". This token is a 679 cryptographic hash of this key. The algorithm for this process is 680 dependent on the authentication algorithm selected; the method of 681 selection is defined later in this section. 683 Upon reception of the initial SYN-segment, a stateful server 684 generates a random key and replies with a SYN/ACK. The key's method 685 of generation is implementation specific. The key MUST be hard to 686 guess, and it MUST be unique for the sending host at any one time. 687 Recommendations for generating random numbers for use in keys are 688 given in [RFC4086]. Connections will be indexed at each host by the 689 token (a one-way hash of the key). Therefore, an implementation will 690 require a mapping from each token to the corresponding connection, 691 and in turn to the keys for the connection. 693 There is a risk that two different keys will hash to the same token. 694 The risk of hash collisions is usually small, unless the host is 695 handling many tens of thousands of connections. Therefore, an 696 implementation SHOULD check its list of connection tokens to ensure 697 there is not a collision before sending its key, and if there is, 698 then it should generate a new key. This would, however, be costly 699 for a server with thousands of connections. The subflow handshake 700 mechanism (Section 3.2) will ensure that new subflows only join the 701 correct connection, however, through the cryptographic handshake, as 702 well as checking the connection tokens in both directions, and 703 ensuring sequence numbers are in-window. So in the worst case if 704 there was a token collision, the new subflow would not succeed, but 705 the MPTCP connection would continue to provide a regular TCP service. 707 Since key generation is implementation-specific, there is no 708 requirement that they be simply random numbers. An implemention is 709 free to exchange cryptographic material out-of-band and generate 710 these keys from this, in order to provide additional mechanisms by 711 which to verify the identity of the communicating entities. For 712 example, an implementation could choose to link its MPTCP keys to 713 those used in higher-layer TLS or SSH connections. 715 If the server behaves in a stateless manner, it has to generate its 716 own key in a verifiable fashion. This verifiable way of generating 717 the key can be done by using a hash of the 4-tuple, sequence number 718 and a local secret (similar to what is done for the TCP-sequence 719 number [RFC4987]). It will thus be able to verify whether it is 720 indeed the originator of the key echoed back in the later MP_CAPABLE 721 option. As for a stateful server, the tokens SHOULD be checked for 722 uniqueness, however if uniqueness is not met, and there is no way to 723 generate an alternative verifiable key, then the connection MUST fall 724 back to using regular TCP by not sending a MP_CAPABLE in the SYN/ACK. 726 The ACK carries both A's key and B's key. This is the first time 727 that A's key is seen on the wire, although it is expected that A will 728 have generated a key locally before the initial SYN. The echoing of 729 B's key allows B to operate statelessly, as described above. 730 Therefore, A's key must be delivered reliably to B, and in order to 731 do this, the transmission of this packet must be made reliable. 733 If B has data to send first, then the reliable delivery of the ACK 734 can be inferred by the receipt of this data with a MPTCP Data 735 Sequence Signal (DSS) option (Section 3.3). If, however, A wishes to 736 send data first, it would not know whether the ACK has successfully 737 been received, and thus whether the MPTCP is successfully 738 established. Therefore, on the first data A has to send (if it has 739 not received any data from B), it MUST also include a MP_CAPABLE 740 option, with additional data parameters (the Data-Level Length and 741 optional Checksum as shown in Figure 4). This packet may be the 742 third ACK if data is ready to be sent by the application, or may be a 743 later packet if the application only later has data to send. This 744 MP_CAPABLE option is in place of the DSS, and simply specifies the 745 data-level length of the payload, and the checksum (if the use of 746 checksums is negotiated). This is the minimal data required to 747 establish a MPTCP connection - it allows validation of the payload, 748 and given it is the first data, the Initial Data Sequence Number 749 (IDSN) is also known (as it is generated from the key, as described 750 below). Conveying the keys on the first data packet allows the TCP 751 reliability mechanisms to ensure the packet is successfully 752 delivered. The receiver will acknowledge this data a the connection 753 level with a Data ACK, as if a DSS option has been received. 755 There could be situations where both A and B attempt to transmit 756 initial data at the same time. For example, if A did not initially 757 have data to send, but then needed to transmit data before it had 758 received anything from B, it would use a MP_CAPABLE option with data 759 parameters (since it would not know if the MP_CAPABLE on the ACK was 760 received). In such a situation, B may also have transmitted data 761 with a DSS option, but it had not yet been received at A. Therefore, 762 B has received data with a MP_CAPABLE mapping after it has sent data 763 with a DSS option. To ensure these situations can be handled, it 764 follows that the data parameters in a MP_CAPABLE are semantically 765 equivalent to those in a DSS option and can be used interchangeably. 766 Similar situations could occur when the MP_CAPABLE with data is lost 767 and retransmitted. Furthermore, in the case of TCP Segmentation 768 Offloading, the MP_CAPABLE with data parameters may be duplicated 769 across multiple packets, and implementations must also be able to 770 cope with duplicate MP_CAPABLE mappings as well as duplicate DSS 771 mappings. 773 Additionally, the MP_CAPABLE exchange allows the safe passage of 774 MPTCP options on SYN packets to be determined. If any of these 775 options are dropped, MPTCP will gracefully fall back to regular 776 single-path TCP, as documented in Section 3.8. Note that new 777 subflows MUST NOT be established (using the process documented in 778 Section 3.2) until a Data Sequence Signal (DSS) option has been 779 successfully received across the path (as documented in Section 3.3). 781 The first 4 bits of the first octet in the MP_CAPABLE option 782 (Figure 4) define the MPTCP option subtype (see Section 8; for 783 MP_CAPABLE, this is 0), and the remaining 4 bits of this octet 784 specify the MPTCP version in use (for this specification, this is 1). 786 The second octet is reserved for flags, allocated as follows: 788 A: The leftmost bit, labeled "A", SHOULD be set to 1 to indicate 789 "Checksum Required", unless the system administrator has decided 790 that checksums are not required (for example, if the environment 791 is controlled and no middleboxes exist that might adjust the 792 payload). 794 B: The second bit, labeled "B", is an extensibility flag, and MUST be 795 set to 0 for current implementations. This will be used for an 796 extensibility mechanism in a future specification, and the impact 797 of this flag will be defined at a later date. If receiving a 798 message with the 'B' flag set to 1, and this is not understood, 799 then this SYN MUST be silently ignored; the sender is expected to 800 retry with a format compatible with this legacy specification. 801 Note that the length of the MP_CAPABLE option, and the meanings of 802 bits "C" through "H", may be altered by setting B=1. 804 C: The third bit, labeled "C", is set to "1" to indicate that the 805 sender of this option will not accept additional MPTCP subflows to 806 the source address and port, and therefore the receiver MUST NOT 807 try to open any additional subflows towards this address and port. 808 This is an efficiency improvement for situations where the sender 809 knows a restriction is in place, for example if the sender is 810 behind a strict NAT, or operating behind a legacy Layer 4 load 811 balancer. 813 D through H: The remaining bits, labeled "D" through "H", are used 814 for crypto algorithm negotiation. Currently only the rightmost 815 bit, labeled "H", is assigned. Bit "H" indicates the use of HMAC- 816 SHA256 (as defined in Section 3.2). An implementation that only 817 supports this method MUST set bit "H" to 1, and bits "D" through 818 "G" to 0. 820 A crypto algorithm MUST be specified. If flag bits D through H are 821 all 0, the MP_CAPABLE option MUST be treated as invalid and ignored 822 (that is, it must be treated as a regular TCP handshake). 824 The selection of the authentication algorithm also impacts the 825 algorithm used to generate the token and the Initial Data Sequence 826 Number (IDSN). In this specification, with only the SHA-256 827 algorithm (bit "H") specified and selected, the token MUST be a 828 truncated (most significant 32 bits) SHA-256 hash ([SHS], [RFC6234]) 829 of the key. A different, 64-bit truncation (the least significant 64 830 bits) of the SHA-256 hash of the key MUST be used as the IDSN. Note 831 that the key MUST be hashed in network byte order. Also note that 832 the "least significant" bits MUST be the rightmost bits of the 833 SHA-256 digest, as per [SHS]. Future specifications of the use of 834 the crypto bits may choose to specify different algorithms for token 835 and IDSN generation. 837 Both the crypto and checksum bits negotiate capabilities in similar 838 ways. For the Checksum Required bit (labeled "A"), if either host 839 requires the use of checksums, checksums MUST be used. In other 840 words, the only way for checksums not to be used is if both hosts in 841 their SYNs set A=0. This decision is confirmed by the setting of the 842 "A" bit in the third packet (the ACK) of the handshake. For example, 843 if the initiator sets A=0 in the SYN, but the responder sets A=1 in 844 the SYN/ACK, checksums MUST be used in both directions, and the 845 initiator will set A=1 in the ACK. The decision whether to use 846 checksums will be stored by an implementation in a per-connection 847 binary state variable. If A=1 is received by a host that does not 848 want to use checksums, it MUST fall back to regular TCP by ignoring 849 the MP_CAPABLE option as if it was invalid. 851 For crypto negotiation, the responder has the choice. The initiator 852 creates a proposal setting a bit for each algorithm it supports to 1 853 (in this version of the specification, there is only one proposal, so 854 bit "H" will be always set to 1). The responder responds with only 1 855 bit set -- this is the chosen algorithm. The rationale for this 856 behavior is that the responder will typically be a server with 857 potentially many thousands of connections, so it may wish to choose 858 an algorithm with minimal computational complexity, depending on the 859 load. If a responder does not support (or does not want to support) 860 any of the initiator's proposals, it can respond without an 861 MP_CAPABLE option, thus forcing a fallback to regular TCP. 863 The MP_CAPABLE option is only used in the first subflow of a 864 connection, in order to identify the connection; all following 865 subflows will use the "Join" option (see Section 3.2) to join the 866 existing connection. 868 If a SYN contains an MP_CAPABLE option but the SYN/ACK does not, it 869 is assumed that the passive opener is not multipath capable; thus, 870 the MPTCP session MUST operate as a regular, single-path TCP. If a 871 SYN does not contain a MP_CAPABLE option, the SYN/ACK MUST NOT 872 contain one in response. If the third packet (the ACK) does not 873 contain the MP_CAPABLE option, then the session MUST fall back to 874 operating as a regular, single-path TCP. This is to maintain 875 compatibility with middleboxes on the path that drop some or all TCP 876 options. Note that an implementation MAY choose to attempt sending 877 MPTCP options more than one time before making this decision to 878 operate as regular TCP (see Section 3.10). 880 If the SYN packets are unacknowledged, it is up to local policy to 881 decide how to respond. It is expected that a sender will eventually 882 fall back to single-path TCP (i.e., without the MP_CAPABLE option) in 883 order to work around middleboxes that may drop packets with unknown 884 options; however, the number of multipath-capable attempts that are 885 made first will be up to local policy. It is possible that MPTCP and 886 non-MPTCP SYNs could get reordered in the network. Therefore, the 887 final state is inferred from the presence or absence of the 888 MP_CAPABLE option in the third packet of the TCP handshake. If this 889 option is not present, the connection SHOULD fall back to regular 890 TCP, as documented in Section 3.8. 892 The initial data sequence number on an MPTCP connection is generated 893 from the key. The algorithm for IDSN generation is also determined 894 from the negotiated authentication algorithm. In this specification, 895 with only the SHA-256 algorithm specified and selected, the IDSN of a 896 host MUST be the least significant 64 bits of the SHA-256 hash of its 897 key, i.e., IDSN-A = Hash(Key-A) and IDSN-B = Hash(Key-B). This 898 deterministic generation of the IDSN allows a receiver to ensure that 899 there are no gaps in sequence space at the start of the connection. 900 The SYN with MP_CAPABLE occupies the first octet of data sequence 901 space, although this does not need to be acknowledged at the 902 connection level until the first data is sent (see Section 3.3). 904 3.2. Starting a New Subflow 906 Once an MPTCP connection has begun with the MP_CAPABLE exchange, 907 further subflows can be added to the connection. Hosts have 908 knowledge of their own address(es), and can become aware of the other 909 host's addresses through signaling exchanges as described in 910 Section 3.4. Using this knowledge, a host can initiate a new subflow 911 over a currently unused pair of addresses. It is permitted for 912 either host in a connection to initiate the creation of a new 913 subflow, but it is expected that this will normally be the original 914 connection initiator (see Section 3.10 for heuristics). 916 A new subflow is started as a normal TCP SYN/ACK exchange. The Join 917 Connection (MP_JOIN) MPTCP option is used to identify the connection 918 to be joined by the new subflow. It uses keying material that was 919 exchanged in the initial MP_CAPABLE handshake (Section 3.1), and that 920 handshake also negotiates the crypto algorithm in use for the MP_JOIN 921 handshake. 923 This section specifies the behavior of MP_JOIN using the HMAC-SHA256 924 algorithm. An MP_JOIN option is present in the SYN, SYN/ACK, and ACK 925 of the three-way handshake, although in each case with a different 926 format. 928 In the first MP_JOIN on the SYN packet, illustrated in Figure 5, the 929 initiator sends a token, random number, and address ID. 931 The token is used to identify the MPTCP connection and is a 932 cryptographic hash of the receiver's key, as exchanged in the initial 933 MP_CAPABLE handshake (Section 3.1). In this specification, the 934 tokens presented in this option are generated by the SHA-256 ([SHS], 935 [RFC6234]) algorithm, truncated to the most significant 32 bits. The 936 token included in the MP_JOIN option is the token that the receiver 937 of the packet uses to identify this connection; i.e., Host A will 938 send Token-B (which is generated from Key-B). Note that the hash 939 generation algorithm can be overridden by the choice of cryptographic 940 handshake algorithm, as defined in Section 3.1. 942 The MP_JOIN SYN sends not only the token (which is static for a 943 connection) but also random numbers (nonces) that are used to prevent 944 replay attacks on the authentication method. Recommendations for the 945 generation of random numbers for this purpose are given in [RFC4086]. 947 The MP_JOIN option includes an "Address ID". This is an identifier 948 that only has significance within a single connection, where it 949 identifies the source address of this packet, even if the IP header 950 has been changed in transit by a middlebox. The Address ID allows 951 address removal (Section 3.4.2) without needing to know what the 952 source address at the receiver is, thus allowing address removal 953 through NATs. The Address ID also allows correlation between new 954 subflow setup attempts and address signaling (Section 3.4.1), to 955 prevent setting up duplicate subflows on the same path, if an MP_JOIN 956 and ADD_ADDR are sent at the same time. 958 The Address IDs of the subflow used in the initial SYN exchange of 959 the first subflow in the connection are implicit, and have the value 960 zero. A host MUST store the mappings between Address IDs and 961 addresses both for itself and the remote host. An implementation 962 will also need to know which local and remote Address IDs are 963 associated with which established subflows, for when addresses are 964 removed from a local or remote host. 966 The MP_JOIN option on packets with the SYN flag set also includes 4 967 bits of flags, 3 of which are currently reserved and MUST be set to 968 zero by the sender. The final bit, labeled "B", indicates whether 969 the sender of this option wishes this subflow to be used as a backup 970 path (B=1) in the event of failure of other paths, or whether it 971 wants it to be used as part of the connection immediately. By 972 setting B=1, the sender of the option is requesting the other host to 973 only send data on this subflow if there are no available subflows 974 where B=0. Subflow policy is discussed in more detail in 975 Section 3.3.8. 977 1 2 3 978 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 1 979 +---------------+---------------+-------+-----+-+---------------+ 980 | Kind | Length = 12 |Subtype| |B| Address ID | 981 +---------------+---------------+-------+-----+-+---------------+ 982 | Receiver's Token (32 bits) | 983 +---------------------------------------------------------------+ 984 | Sender's Random Number (32 bits) | 985 +---------------------------------------------------------------+ 987 Figure 5: Join Connection (MP_JOIN) Option (for Initial SYN) 989 When receiving a SYN with an MP_JOIN option that contains a valid 990 token for an existing MPTCP connection, the recipient SHOULD respond 991 with a SYN/ACK also containing an MP_JOIN option containing a random 992 number and a truncated (leftmost 64 bits) Hash-based Message 993 Authentication Code (HMAC). This version of the option is shown in 994 Figure 6. If the token is unknown, or the host wants to refuse 995 subflow establishment (for example, due to a limit on the number of 996 subflows it will permit), the receiver will send back a reset (RST) 997 signal, analogous to an unknown port in TCP, containing a MP_TCPRST 998 option (Section 3.6) with an appropriate reason code. Although 999 calculating an HMAC requires cryptographic operations, it is believed 1000 that the 32-bit token in the MP_JOIN SYN gives sufficient protection 1001 against blind state exhaustion attacks; therefore, there is no need 1002 to provide mechanisms to allow a responder to operate statelessly at 1003 the MP_JOIN stage. 1005 An HMAC is sent by both hosts -- by the initiator (Host A) in the 1006 third packet (the ACK) and by the responder (Host B) in the second 1007 packet (the SYN/ACK). Doing the HMAC exchange at this stage allows 1008 both hosts to have first exchanged random data (in the first two SYN 1009 packets) that is used as the "message". This specification defines 1010 that HMAC as defined in [RFC2104] is used, along with the SHA-256 1011 hash algorithm [SHS] (potentially implemented as in [RFC6234]), thus 1012 generating a 160-bit / 20-octet HMAC. Due to option space 1013 limitations, the HMAC included in the SYN/ACK is truncated to the 1014 leftmost 64 bits, but this is acceptable since random numbers are 1015 used; thus, an attacker only has one chance to guess the HMAC 1016 correctly (if the HMAC is incorrect, the TCP connection is closed, so 1017 a new MP_JOIN negotiation with a new random number is required). 1019 The initiator's authentication information is sent in its first ACK 1020 (the third packet of the handshake), as shown in Figure 7. This data 1021 needs to be sent reliably, since it is the only time this HMAC is 1022 sent; therefore, receipt of this packet MUST trigger a regular TCP 1023 ACK in response, and the packet MUST be retransmitted if this ACK is 1024 not received. In other words, sending the ACK/MP_JOIN packet places 1025 the subflow in the PRE_ESTABLISHED state, and it moves to the 1026 ESTABLISHED state only on receipt of an ACK from the receiver. It is 1027 not permitted to send data while in the PRE_ESTABLISHED state. The 1028 reserved bits in this option MUST be set to zero by the sender. 1030 The key for the HMAC algorithm, in the case of the message 1031 transmitted by Host A, will be Key-A followed by Key-B, and in the 1032 case of Host B, Key-B followed by Key-A. These are the keys that 1033 were exchanged in the original MP_CAPABLE handshake. The "message" 1034 for the HMAC algorithm in each case is the concatenations of random 1035 number for each host (denoted by R): for Host A, R-A followed by R-B; 1036 and for Host B, R-B followed by R-A. 1038 1 2 3 1039 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 1040 +---------------+---------------+-------+-----+-+---------------+ 1041 | Kind | Length = 16 |Subtype| |B| Address ID | 1042 +---------------+---------------+-------+-----+-+---------------+ 1043 | | 1044 | Sender's Truncated HMAC (64 bits) | 1045 | | 1046 +---------------------------------------------------------------+ 1047 | Sender's Random Number (32 bits) | 1048 +---------------------------------------------------------------+ 1050 Figure 6: Join Connection (MP_JOIN) Option (for Responding SYN/ACK) 1052 1 2 3 1053 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 1054 +---------------+---------------+-------+-----------------------+ 1055 | Kind | Length = 24 |Subtype| (reserved) | 1056 +---------------+---------------+-------+-----------------------+ 1057 | | 1058 | | 1059 | Sender's HMAC (160 bits) | 1060 | | 1061 | | 1062 +---------------------------------------------------------------+ 1064 Figure 7: Join Connection (MP_JOIN) Option (for Third ACK) 1066 These various MPTCP options fit together to enable authenticated 1067 subflow setup as illustrated in Figure 8. 1069 Host A Host B 1070 ------------------------ ---------- 1071 Address A1 Address A2 Address B1 1072 ---------- ---------- ---------- 1073 | | | 1074 | SYN + MP_CAPABLE(Key-A) | 1075 |--------------------------------------------->| 1076 |<---------------------------------------------| 1077 | SYN/ACK + MP_CAPABLE(Key-B) | 1078 | | | 1079 | ACK + MP_CAPABLE(Key-A, Key-B) | 1080 |--------------------------------------------->| 1081 | | | 1082 | | SYN + MP_JOIN(Token-B, R-A) | 1083 | |------------------------------->| 1084 | |<-------------------------------| 1085 | | SYN/ACK + MP_JOIN(HMAC-B, R-B) | 1086 | | | 1087 | | ACK + MP_JOIN(HMAC-A) | 1088 | |------------------------------->| 1089 | |<-------------------------------| 1090 | | ACK | 1092 HMAC-A = HMAC(Key=(Key-A+Key-B), Msg=(R-A+R-B)) 1093 HMAC-B = HMAC(Key=(Key-B+Key-A), Msg=(R-B+R-A)) 1095 Figure 8: Example Use of MPTCP Authentication 1097 If the token received at Host B is unknown or local policy prohibits 1098 the acceptance of the new subflow, the recipient MUST respond with a 1099 TCP RST for the subflow, with a MP_TCPRST option (Section 3.6) with 1100 an appropriate reason code. 1102 If the token is accepted at Host B, but the HMAC returned to Host A 1103 does not match the one expected, Host A MUST close the subflow with a 1104 TCP RST. In this, and all following cases of sending a RST in this 1105 section, the sender SHOULD send a MP_TCPRST option (Section 3.6) on 1106 this RST packet with the reason code for a "MPTCP specific error". 1108 If Host B does not receive the expected HMAC, or the MP_JOIN option 1109 is missing from the ACK, it MUST close the subflow with a TCP RST 1110 with a MP_TCPRST (Section 3.6) option with the reason code for "MPTCP 1111 specific error". 1113 If the HMACs are verified as correct, then both hosts have 1114 authenticated each other as being the same peers as existed at the 1115 start of the connection, and they have agreed of which connection 1116 this subflow will become a part. 1118 If the SYN/ACK as received at Host A does not have an MP_JOIN option, 1119 Host A MUST close the subflow with a TCP RST with a MP_TCPRST 1120 (Section 3.6) option with the reason code for "MPTCP specific error". 1122 This covers all cases of the loss of an MP_JOIN. In more detail, if 1123 MP_JOIN is stripped from the SYN on the path from A to B, and Host B 1124 does not have a passive opener on the relevant port, it will respond 1125 with a RST in the normal way. If in response to a SYN with an 1126 MP_JOIN option, a SYN/ACK is received without the MP_JOIN option 1127 (either since it was stripped on the return path, or it was stripped 1128 on the outgoing path but the passive opener on Host B responded as if 1129 it were a new regular TCP session), then the subflow is unusable and 1130 Host A MUST close it with a RST. 1132 Note that additional subflows can be created between any pair of 1133 ports (but see Section 3.10 for heuristics); no explicit application- 1134 level accept calls or bind calls are required to open additional 1135 subflows. To associate a new subflow with an existing connection, 1136 the token supplied in the subflow's SYN exchange is used for 1137 demultiplexing. This then binds the 5-tuple of the TCP subflow to 1138 the local token of the connection. A consequence is that it is 1139 possible to allow any port pairs to be used for a connection. 1141 Demultiplexing subflow SYNs MUST be done using the token; this is 1142 unlike traditional TCP, where the destination port is used for 1143 demultiplexing SYN packets. Once a subflow is set up, demultiplexing 1144 packets is done using the 5-tuple, as in traditional TCP. The 1145 5-tuples will be mapped to the local connection identifier (token). 1146 Note that Host A will know its local token for the subflow even 1147 though it is not sent on the wire -- only the responder's token is 1148 sent. 1150 3.3. General MPTCP Operation 1152 This section discusses operation of MPTCP for data transfer. At a 1153 high level, an MPTCP implementation will take one input data stream 1154 from an application, and split it into one or more subflows, with 1155 sufficient control information to allow it to be reassembled and 1156 delivered reliably and in order to the recipient application. The 1157 following subsections define this behavior in detail. 1159 The data sequence mapping and the Data ACK are signaled in the Data 1160 Sequence Signal (DSS) option (Figure 9). Either or both can be 1161 signaled in one DSS, depending on the flags set. The data sequence 1162 mapping defines how the sequence space on the subflow maps to the 1163 connection level, and the Data ACK acknowledges receipt of data at 1164 the connection level. These functions are described in more detail 1165 in the following two subsections. 1167 1 2 3 1168 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 1169 +---------------+---------------+-------+----------------------+ 1170 | Kind | Length |Subtype| (reserved) |F|m|M|a|A| 1171 +---------------+---------------+-------+----------------------+ 1172 | Data ACK (4 or 8 octets, depending on flags) | 1173 +--------------------------------------------------------------+ 1174 | Data sequence number (4 or 8 octets, depending on flags) | 1175 +--------------------------------------------------------------+ 1176 | Subflow Sequence Number (4 octets) | 1177 +-------------------------------+------------------------------+ 1178 | Data-Level Length (2 octets) | Checksum (2 octets) | 1179 +-------------------------------+------------------------------+ 1181 Figure 9: Data Sequence Signal (DSS) Option 1183 The flags, when set, define the contents of this option, as follows: 1185 o A = Data ACK present 1187 o a = Data ACK is 8 octets (if not set, Data ACK is 4 octets) 1189 o M = Data Sequence Number (DSN), Subflow Sequence Number (SSN), 1190 Data-Level Length, and Checksum present 1192 o m = Data sequence number is 8 octets (if not set, DSN is 4 octets) 1194 The flags 'a' and 'm' only have meaning if the corresponding 'A' or 1195 'M' flags are set; otherwise, they will be ignored. The maximum 1196 length of this option, with all flags set, is 28 octets. 1198 The 'F' flag indicates "DATA_FIN". If present, this means that this 1199 mapping covers the final data from the sender. This is the 1200 connection-level equivalent to the FIN flag in single-path TCP. A 1201 connection is not closed unless there has been a DATA_FIN exchange or 1202 a timeout. The purpose of the DATA_FIN and the interactions between 1203 this flag, the subflow-level FIN flag, and the data sequence mapping 1204 are described in Section 3.3.3. The remaining reserved bits MUST be 1205 set to zero by an implementation of this specification. 1207 Note that the checksum is only present in this option if the use of 1208 MPTCP checksumming has been negotiated at the MP_CAPABLE handshake 1209 (see Section 3.1). The presence of the checksum can be inferred from 1210 the length of the option. If a checksum is present, but its use had 1211 not been negotiated in the MP_CAPABLE handshake, the checksum field 1212 MUST be ignored. If a checksum is not present when its use has been 1213 negotiated, the receiver MUST close the subflow with a RST as it is 1214 considered broken. This RST SHOULD be accompanied with a MP_TCPRST 1215 option (Section 3.6) with the reason code for a "MPTCP specific 1216 error". 1218 3.3.1. Data Sequence Mapping 1220 The data stream as a whole can be reassembled through the use of the 1221 data sequence mapping components of the DSS option (Figure 9), which 1222 define the mapping from the subflow sequence number to the data 1223 sequence number. This is used by the receiver to ensure in-order 1224 delivery to the application layer. Meanwhile, the subflow-level 1225 sequence numbers (i.e., the regular sequence numbers in the TCP 1226 header) have subflow-only relevance. It is expected (but not 1227 mandated) that SACK [RFC2018] is used at the subflow level to improve 1228 efficiency. 1230 The data sequence mapping specifies a mapping from subflow sequence 1231 space to data sequence space. This is expressed in terms of starting 1232 sequence numbers for the subflow and the data level, and a length of 1233 bytes for which this mapping is valid. This explicit mapping for a 1234 range of data was chosen rather than per-packet signaling to assist 1235 with compatibility with situations where TCP/IP segmentation or 1236 coalescing is undertaken separately from the stack that is generating 1237 the data flow (e.g., through the use of TCP segmentation offloading 1238 on network interface cards, or by middleboxes such as performance 1239 enhancing proxies). It also allows a single mapping to cover many 1240 packets, which may be useful in bulk transfer situations. 1242 A mapping is fixed, in that the subflow sequence number is bound to 1243 the data sequence number after the mapping has been processed. A 1244 sender MUST NOT change this mapping after it has been declared; 1245 however, the same data sequence number can be mapped to by different 1246 subflows for retransmission purposes (see Section 3.3.6). This would 1247 also permit the same data to be sent simultaneously on multiple 1248 subflows for resilience or efficiency purposes, especially in the 1249 case of lossy links. Although the detailed specification of such 1250 operation is outside the scope of this document, an implementation 1251 SHOULD treat the first data that is received at a subflow for the 1252 data sequence space as that which should be delivered to the 1253 application, and any later data for that sequence space ignored. 1255 The data sequence number is specified as an absolute value, whereas 1256 the subflow sequence numbering is relative (the SYN at the start of 1257 the subflow has relative subflow sequence number 0). This is to 1258 allow middleboxes to change the initial sequence number of a subflow, 1259 such as firewalls that undertake ISN randomization. 1261 The data sequence mapping also contains a checksum of the data that 1262 this mapping covers, if use of checksums has been negotiated at the 1263 MP_CAPABLE exchange. Checksums are used to detect if the payload has 1264 been adjusted in any way by a non-MPTCP-aware middlebox. If this 1265 checksum fails, it will trigger a failure of the subflow, or a 1266 fallback to regular TCP, as documented in Section 3.8, since MPTCP 1267 can no longer reliably know the subflow sequence space at the 1268 receiver to build data sequence mappings. 1270 The checksum algorithm used is the standard TCP checksum [RFC0793], 1271 operating over the data covered by this mapping, along with a pseudo- 1272 header as shown in Figure 10. 1274 1 2 3 1275 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 1276 +--------------------------------------------------------------+ 1277 | | 1278 | Data Sequence Number (8 octets) | 1279 | | 1280 +--------------------------------------------------------------+ 1281 | Subflow Sequence Number (4 octets) | 1282 +-------------------------------+------------------------------+ 1283 | Data-Level Length (2 octets) | Zeros (2 octets) | 1284 +-------------------------------+------------------------------+ 1286 Figure 10: Pseudo-Header for DSS Checksum 1288 Note that the data sequence number used in the pseudo-header is 1289 always the 64-bit value, irrespective of what length is used in the 1290 DSS option itself. The standard TCP checksum algorithm has been 1291 chosen since it will be calculated anyway for the TCP subflow, and if 1292 calculated first over the data before adding the pseudo-headers, it 1293 only needs to be calculated once. Furthermore, since the TCP 1294 checksum is additive, the checksum for a DSN_MAP can be constructed 1295 by simply adding together the checksums for the data of each 1296 constituent TCP segment, and adding the checksum for the DSS pseudo- 1297 header. 1299 Note that checksumming relies on the TCP subflow containing 1300 contiguous data; therefore, a TCP subflow MUST NOT use the Urgent 1301 Pointer to interrupt an existing mapping. Further note, however, 1302 that if Urgent data is received on a subflow, it SHOULD be mapped to 1303 the data sequence space and delivered to the application analogous to 1304 Urgent data in regular TCP. 1306 To avoid possible deadlock scenarios, subflow-level processing should 1307 be undertaken separately from that at connection level. Therefore, 1308 even if a mapping does not exist from the subflow space to the data- 1309 level space, the data SHOULD still be ACKed at the subflow (if it is 1310 in-window). This data cannot, however, be acknowledged at the data 1311 level (Section 3.3.2) because its data sequence numbers are unknown. 1312 Implementations MAY hold onto such unmapped data for a short while in 1313 the expectation that a mapping will arrive shortly. Such unmapped 1314 data cannot be counted as being within the connection level receive 1315 window because this is relative to the data sequence numbers, so if 1316 the receiver runs out of memory to hold this data, it will have to be 1317 discarded. If a mapping for that subflow-level sequence space does 1318 not arrive within a receive window of data, that subflow SHOULD be 1319 treated as broken, closed with a RST, and any unmapped data silently 1320 discarded. 1322 Data sequence numbers are always 64-bit quantities, and MUST be 1323 maintained as such in implementations. If a connection is 1324 progressing at a slow rate, so protection against wrapped sequence 1325 numbers is not required, then an implementation MAY include just the 1326 lower 32 bits of the data sequence number in the data sequence 1327 mapping and/or Data ACK as an optimization, and an implementation can 1328 make this choice independently for each packet. An implementaton 1329 MUST be able to receive and process both 64-bit or 32-bit sequence 1330 number values, but it is not required that an implementation is able 1331 to send both. 1333 An implementation MUST send the full 64-bit data sequence number if 1334 it is transmitting at a sufficiently high rate that the 32-bit value 1335 could wrap within the Maximum Segment Lifetime (MSL) [RFC1323]. The 1336 lengths of the DSNs used in these values (which may be different) are 1337 declared with flags in the DSS option. Implementations MUST accept a 1338 32-bit DSN and implicitly promote it to a 64-bit quantity by 1339 incrementing the upper 32 bits of sequence number each time the lower 1340 32 bits wrap. A sanity check MUST be implemented to ensure that a 1341 wrap occurs at an expected time (e.g., the sequence number jumps from 1342 a very high number to a very low number) and is not triggered by out- 1343 of-order packets. 1345 As with the standard TCP sequence number, the data sequence number 1346 should not start at zero, but at a random value to make blind session 1347 hijacking harder. This specification requires setting the initial 1348 data sequence number (IDSN) of each host to the least significant 64 1349 bits of the SHA-256 hash of the host's key, as described in 1350 Section 3.1. This is required also in order for the receiver to know 1351 what the expected IDSN is, and thus determine if any initial 1352 connection-level packets are missing; this is particularly relevant 1353 if two subflows start transmitting simultaneously. 1355 A data sequence mapping does not need to be included in every MPTCP 1356 packet, as long as the subflow sequence space in that packet is 1357 covered by a mapping known at the receiver. This can be used to 1358 reduce overhead in cases where the mapping is known in advance; one 1359 such case is when there is a single subflow between the hosts, 1360 another is when segments of data are scheduled in larger than packet- 1361 sized chunks. 1363 An "infinite" mapping can be used to fall back to regular TCP by 1364 mapping the subflow-level data to the connection-level data for the 1365 remainder of the connection (see Section 3.8). This is achieved by 1366 setting the Data-Level Length field of the DSS option to the reserved 1367 value of 0. The checksum, in such a case, will also be set to zero. 1369 3.3.2. Data Acknowledgments 1371 To provide full end-to-end resilience, MPTCP provides a connection- 1372 level acknowledgment, to act as a cumulative ACK for the connection 1373 as a whole. This is the "Data ACK" field of the DSS option 1374 (Figure 9). The Data ACK is analogous to the behavior of the 1375 standard TCP cumulative ACK -- indicating how much data has been 1376 successfully received (with no holes). This is in comparison to the 1377 subflow-level ACK, which acts analogous to TCP SACK, given that there 1378 may still be holes in the data stream at the connection level. The 1379 Data ACK specifies the next data sequence number it expects to 1380 receive. 1382 The Data ACK, as for the DSN, can be sent as the full 64-bit value, 1383 or as the lower 32 bits. If data is received with a 64-bit DSN, it 1384 MUST be acknowledged with a 64-bit Data ACK. If the DSN received is 1385 32 bits, it is valid for the implementation to choose whether to send 1386 a 32-bit or 64-bit Data ACK. 1388 The Data ACK proves that the data, and all required MPTCP signaling, 1389 has been received and accepted by the remote end. One key use of the 1390 Data ACK signal is that it is used to indicate the left edge of the 1391 advertised receive window. As explained in Section 3.3.4, the 1392 receive window is shared by all subflows and is relative to the Data 1393 ACK. Because of this, an implementation MUST NOT use the RCV.WND 1394 field of a TCP segment at the connection level if it does not also 1395 carry a DSS option with a Data ACK field. Furthermore, separating 1396 the connection-level acknowledgments from the subflow level allows 1397 processing to be done separately, and a receiver has the freedom to 1398 drop segments after acknowledgment at the subflow level, for example, 1399 due to memory constraints when many segments arrive out of order. 1401 An MPTCP sender MUST NOT free data from the send buffer until it has 1402 been acknowledged by both a Data ACK received on any subflow and at 1403 the subflow level by all subflows on which the data was sent. The 1404 former condition ensures liveness of the connection and the latter 1405 condition ensures liveness and self-consistence of a subflow when 1406 data needs to be retransmitted. Note, however, that if some data 1407 needs to be retransmitted multiple times over a subflow, there is a 1408 risk of blocking the sending window. In this case, the MPTCP sender 1409 can decide to terminate the subflow that is behaving badly by sending 1410 a RST, using an appropriate MP_TCPRST (Section 3.6) error code. 1412 The Data ACK MAY be included in all segments; however, optimizations 1413 SHOULD be considered in more advanced implementations, where the Data 1414 ACK is present in segments only when the Data ACK value advances, and 1415 this behavior MUST be treated as valid. This behavior ensures the 1416 sender buffer is freed, while reducing overhead when the data 1417 transfer is unidirectional. 1419 3.3.3. Closing a Connection 1421 In regular TCP, a FIN announces the receiver that the sender has no 1422 more data to send. In order to allow subflows to operate 1423 independently and to keep the appearance of TCP over the wire, a FIN 1424 in MPTCP only affects the subflow on which it is sent. This allows 1425 nodes to exercise considerable freedom over which paths are in use at 1426 any one time. The semantics of a FIN remain as for regular TCP; 1427 i.e., it is not until both sides have ACKed each other's FINs that 1428 the subflow is fully closed. 1430 When an application calls close() on a socket, this indicates that it 1431 has no more data to send; for regular TCP, this would result in a FIN 1432 on the connection. For MPTCP, an equivalent mechanism is needed, and 1433 this is referred to as the DATA_FIN. 1435 A DATA_FIN is an indication that the sender has no more data to send, 1436 and as such can be used to verify that all data has been successfully 1437 received. A DATA_FIN, as with the FIN on a regular TCP connection, 1438 is a unidirectional signal. 1440 The DATA_FIN is signaled by setting the 'F' flag in the Data Sequence 1441 Signal option (Figure 9) to 1. A DATA_FIN occupies 1 octet (the 1442 final octet) of the connection-level sequence space. Note that the 1443 DATA_FIN is included in the Data-Level Length, but not at the subflow 1444 level: for example, a segment with DSN 80, and Data-Level Length 11, 1445 with DATA_FIN set, would map 10 octets from the subflow into data 1446 sequence space 80-89, the DATA_FIN is DSN 90; therefore, this segment 1447 including DATA_FIN would be acknowledged with a DATA_ACK of 91. 1449 Note that when the DATA_FIN is not attached to a TCP segment 1450 containing data, the Data Sequence Signal MUST have a subflow 1451 sequence number of 0, a Data-Level Length of 1, and the data sequence 1452 number that corresponds with the DATA_FIN itself. The checksum in 1453 this case will only cover the pseudo-header. 1455 A DATA_FIN has the semantics and behavior as a regular TCP FIN, but 1456 at the connection level. Notably, it is only DATA_ACKed once all 1457 data has been successfully received at the connection level. Note, 1458 therefore, that a DATA_FIN is decoupled from a subflow FIN. It is 1459 only permissible to combine these signals on one subflow if there is 1460 no data outstanding on other subflows. Otherwise, it may be 1461 necessary to retransmit data on different subflows. Essentially, a 1462 host MUST NOT close all functioning subflows unless it is safe to do 1463 so, i.e., until all outstanding data has been DATA_ACKed, or until 1464 the segment with the DATA_FIN flag set is the only outstanding 1465 segment. 1467 Once a DATA_FIN has been acknowledged, all remaining subflows MUST be 1468 closed with standard FIN exchanges. Both hosts SHOULD send FINs on 1469 all subflows, as a courtesy to allow middleboxes to clean up state 1470 even if an individual subflow has failed. It is also encouraged to 1471 reduce the timeouts (Maximum Segment Life) on subflows at end hosts. 1472 In particular, any subflows where there is still outstanding data 1473 queued (which has been retransmitted on other subflows in order to 1474 get the DATA_FIN acknowledged) MAY be closed with a RST with 1475 MP_TCPRST (Section 3.6) error code for "too much outstanding data". 1477 A connection is considered closed once both hosts' DATA_FINs have 1478 been acknowledged by DATA_ACKs. 1480 As specified above, a standard TCP FIN on an individual subflow only 1481 shuts down the subflow on which it was sent. If all subflows have 1482 been closed with a FIN exchange, but no DATA_FIN has been received 1483 and acknowledged, the MPTCP connection is treated as closed only 1484 after a timeout. This implies that an implementation will have 1485 TIME_WAIT states at both the subflow and connection levels (see 1486 Appendix C). This permits "break-before-make" scenarios where 1487 connectivity is lost on all subflows before a new one can be re- 1488 established. 1490 3.3.4. Receiver Considerations 1492 Regular TCP advertises a receive window in each packet, telling the 1493 sender how much data the receiver is willing to accept past the 1494 cumulative ack. The receive window is used to implement flow 1495 control, throttling down fast senders when receivers cannot keep up. 1497 MPTCP also uses a unique receive window, shared between the subflows. 1498 The idea is to allow any subflow to send data as long as the receiver 1499 is willing to accept it. The alternative, maintaining per subflow 1500 receive windows, could end up stalling some subflows while others 1501 would not use up their window. 1503 The receive window is relative to the DATA_ACK. As in TCP, a 1504 receiver MUST NOT shrink the right edge of the receive window (i.e., 1505 DATA_ACK + receive window). The receiver will use the data sequence 1506 number to tell if a packet should be accepted at the connection 1507 level. 1509 When deciding to accept packets at subflow level, regular TCP checks 1510 the sequence number in the packet against the allowed receive window. 1511 With multipath, such a check is done using only the connection-level 1512 window. A sanity check SHOULD be performed at subflow level to 1513 ensure that the subflow and mapped sequence numbers meet the 1514 following test: SSN - SUBFLOW_ACK <= DSN - DATA_ACK, where SSN is the 1515 subflow sequence number of the received packet and SUBFLOW_ACK is the 1516 RCV.NXT (next expected sequence number) of the subflow (with the 1517 equivalent connection-level definitions for DSN and DATA_ACK). 1519 In regular TCP, once a segment is deemed in-window, it is put either 1520 in the in-order receive queue or in the out-of-order queue. In 1521 Multipath TCP, the same happens but at the connection level: a 1522 segment is placed in the connection level in-order or out-of-order 1523 queue if it is in-window at both connection and subflow levels. The 1524 stack still has to remember, for each subflow, which segments were 1525 received successfully so that it can ACK them at subflow level 1526 appropriately. Typically, this will be implemented by keeping per 1527 subflow out-of-order queues (containing only message headers, not the 1528 payloads) and remembering the value of the cumulative ACK. 1530 It is important for implementers to understand how large a receiver 1531 buffer is appropriate. The lower bound for full network utilization 1532 is the maximum bandwidth-delay product of any one of the paths. 1533 However, this might be insufficient when a packet is lost on a slower 1534 subflow and needs to be retransmitted (see Section 3.3.6). A tight 1535 upper bound would be the maximum round-trip time (RTT) of any path 1536 multiplied by the total bandwidth available across all paths. This 1537 permits all subflows to continue at full speed while a packet is 1538 fast-retransmitted on the maximum RTT path. Even this might be 1539 insufficient to maintain full performance in the event of a 1540 retransmit timeout on the maximum RTT path. It is for future study 1541 to determine the relationship between retransmission strategies and 1542 receive buffer sizing. 1544 3.3.5. Sender Considerations 1546 The sender remembers receiver window advertisements from the 1547 receiver. It should only update its local receive window values when 1548 the largest sequence number allowed (i.e., DATA_ACK + receive window) 1549 increases, on the receipt of a DATA_ACK. This is important to allow 1550 using paths with different RTTs, and thus different feedback loops. 1552 MPTCP uses a single receive window across all subflows, and if the 1553 receive window was guaranteed to be unchanged end-to-end, a host 1554 could always read the most recent receive window value. However, 1555 some classes of middleboxes may alter the TCP-level receive window. 1556 Typically, these will shrink the offered window, although for short 1557 periods of time it may be possible for the window to be larger 1558 (however, note that this would not continue for long periods since 1559 ultimately the middlebox must keep up with delivering data to the 1560 receiver). Therefore, if receive window sizes differ on multiple 1561 subflows, when sending data MPTCP SHOULD take the largest of the most 1562 recent window sizes as the one to use in calculations. This rule is 1563 implicit in the requirement not to reduce the right edge of the 1564 window. 1566 The sender MUST also remember the receive windows advertised by each 1567 subflow. The allowed window for subflow i is (ack_i, ack_i + 1568 rcv_wnd_i), where ack_i is the subflow-level cumulative ACK of 1569 subflow i. This ensures data will not be sent to a middlebox unless 1570 there is enough buffering for the data. 1572 Putting the two rules together, we get the following: a sender is 1573 allowed to send data segments with data-level sequence numbers 1574 between (DATA_ACK, DATA_ACK + receive_window). Each of these 1575 segments will be mapped onto subflows, as long as subflow sequence 1576 numbers are in the allowed windows for those subflows. Note that 1577 subflow sequence numbers do not generally affect flow control if the 1578 same receive window is advertised across all subflows. They will 1579 perform flow control for those subflows with a smaller advertised 1580 receive window. 1582 The send buffer MUST, at a minimum, be as big as the receive buffer, 1583 to enable the sender to reach maximum throughput. 1585 3.3.6. Reliability and Retransmissions 1587 The data sequence mapping allows senders to resend data with the same 1588 data sequence number on a different subflow. When doing this, a host 1589 MUST still retransmit the original data on the original subflow, in 1590 order to preserve the subflow integrity (middleboxes could replay old 1591 data, and/or could reject holes in subflows), and a receiver will 1592 ignore these retransmissions. While this is clearly suboptimal, for 1593 compatibility reasons this is sensible behavior. Optimizations could 1594 be negotiated in future versions of this protocol. Note also that 1595 this property would also permit a sender to always send the same 1596 data, with the same data sequence number, on multiple subflows, if it 1597 so desired for reliability reasons. 1599 This protocol specification does not mandate any mechanisms for 1600 handling retransmissions, and much will be dependent upon local 1601 policy (as discussed in Section 3.3.8). One can imagine aggressive 1602 connection-level retransmissions policies where every packet lost at 1603 subflow level is retransmitted on a different subflow (hence, wasting 1604 bandwidth but possibly reducing application-to-application delays), 1605 or conservative retransmission policies where connection-level 1606 retransmits are only used after a few subflow-level retransmission 1607 timeouts occur. 1609 It is envisaged that a standard connection-level retransmission 1610 mechanism would be implemented around a connection-level data queue: 1611 all segments that haven't been DATA_ACKed are stored. A timer is set 1612 when the head of the connection-level is ACKed at subflow level but 1613 its corresponding data is not ACKed at data level. This timer will 1614 guard against failures in retransmission by middleboxes that 1615 proactively ACK data. 1617 The sender MUST keep data in its send buffer as long as the data has 1618 not been acknowledged at both connection level and on all subflows on 1619 which it has been sent. In this way, the sender can always 1620 retransmit the data if needed, on the same subflow or on a different 1621 one. A special case is when a subflow fails: the sender will 1622 typically resend the data on other working subflows after a timeout, 1623 and will keep trying to retransmit the data on the failed subflow 1624 too. The sender will declare the subflow failed after a predefined 1625 upper bound on retransmissions is reached (which MAY be lower than 1626 the usual TCP limits of the Maximum Segment Life), or on the receipt 1627 of an ICMP error, and only then delete the outstanding data segments. 1629 Multiple retransmissions are triggers that will indicate that a 1630 subflow performs badly and could lead to a host resetting the subflow 1631 with a RST. However, additional research is required to understand 1632 the heuristics of how and when to reset underperforming subflows. 1633 For example, a highly asymmetric path may be misdiagnosed as 1634 underperforming. A RST for this purpose SHOULD be accompanied with 1635 an appropriate MP_TCPRST option (Section 3.6). 1637 3.3.7. Congestion Control Considerations 1639 Different subflows in an MPTCP connection have different congestion 1640 windows. To achieve fairness at bottlenecks and resource pooling, it 1641 is necessary to couple the congestion windows in use on each subflow, 1642 in order to push most traffic to uncongested links. One algorithm 1643 for achieving this is presented in [RFC6356]; the algorithm does not 1644 achieve perfect resource pooling but is "safe" in that it is readily 1645 deployable in the current Internet. By this, we mean that it does 1646 not take up more capacity on any one path than if it was a single 1647 path flow using only that route, so this ensures fair coexistence 1648 with single-path TCP at shared bottlenecks. 1650 It is foreseeable that different congestion controllers will be 1651 implemented for MPTCP, each aiming to achieve different properties in 1652 the resource pooling/fairness/stability design space, as well as 1653 those for achieving different properties in quality of service, 1654 reliability, and resilience. 1656 Regardless of the algorithm used, the design of the MPTCP protocol 1657 aims to provide the congestion control implementations sufficient 1658 information to take the right decisions; this information includes, 1659 for each subflow, which packets were lost and when. 1661 3.3.8. Subflow Policy 1663 Within a local MPTCP implementation, a host may use any local policy 1664 it wishes to decide how to share the traffic to be sent over the 1665 available paths. 1667 In the typical use case, where the goal is to maximize throughput, 1668 all available paths will be used simultaneously for data transfer, 1669 using coupled congestion control as described in [RFC6356]. It is 1670 expected, however, that other use cases will appear. 1672 For instance, a possibility is an 'all-or-nothing' approach, i.e., 1673 have a second path ready for use in the event of failure of the first 1674 path, but alternatives could include entirely saturating one path 1675 before using an additional path (the 'overflow' case). Such choices 1676 would be most likely based on the monetary cost of links, but may 1677 also be based on properties such as the delay or jitter of links, 1678 where stability (of delay or bandwidth) is more important than 1679 throughput. Application requirements such as these are discussed in 1680 detail in [RFC6897]. 1682 The ability to make effective choices at the sender requires full 1683 knowledge of the path "cost", which is unlikely to be the case. It 1684 would be desirable for a receiver to be able to signal their own 1685 preferences for paths, since they will often be the multihomed party, 1686 and may have to pay for metered incoming bandwidth. 1688 Whilst fine-grained control may be the most powerful solution, that 1689 would require some mechanism such as overloading the Explicit 1690 Congestion Notification (ECN) signal [RFC3168], which is undesirable, 1691 and it is felt that there would not be sufficient benefit to justify 1692 an entirely new signal. Therefore, the MP_JOIN option (see 1693 Section 3.2) contains the 'B' bit, which allows a host to indicate to 1694 its peer that this path should be treated as a backup path to use 1695 only in the event of failure of other working subflows (i.e., a 1696 subflow where the receiver has indicated B=1 SHOULD NOT be used to 1697 send data unless there are no usable subflows where B=0). 1699 In the event that the available set of paths changes, a host may wish 1700 to signal a change in priority of subflows to the peer (e.g., a 1701 subflow that was previously set as backup should now take priority 1702 over all remaining subflows). Therefore, the MP_PRIO option, shown 1703 in Figure 11, can be used to change the 'B' flag of the subflow on 1704 which it is sent. 1706 1 2 3 1707 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 1708 +---------------+---------------+-------+-----+-+ 1709 | Kind | Length |Subtype| |B| 1710 +---------------+---------------+-------+-----+-+ 1712 Figure 11: Change Subflow Priority (MP_PRIO) Option 1714 It should be noted that the backup flag is a request from a data 1715 receiver to a data sender only, and the data sender SHOULD adhere to 1716 these requests. A host cannot assume that the data sender will do 1717 so, however, since local policies -- or technical difficulties -- may 1718 override MP_PRIO requests. Note also that this signal applies to a 1719 single direction, and so the sender of this option could choose to 1720 continue using the subflow to send data even if it has signaled B=1 1721 to the other host. 1723 3.4. Address Knowledge Exchange (Path Management) 1725 We use the term "path management" to refer to the exchange of 1726 information about additional paths between hosts, which in this 1727 design is managed by multiple addresses at hosts. For more detail of 1728 the architectural thinking behind this design, see the MPTCP 1729 Architecture document [RFC6182]. 1731 This design makes use of two methods of sharing such information, and 1732 both can be used on a connection. The first is the direct setup of 1733 new subflows, already described in Section 3.2, where the initiator 1734 has an additional address. The second method, described in the 1735 following subsections, signals addresses explicitly to the other host 1736 to allow it to initiate new subflows. The two mechanisms are 1737 complementary: the first is implicit and simple, while the explicit 1738 is more complex but is more robust. Together, the mechanisms allow 1739 addresses to change in flight (and thus support operation through 1740 NATs, since the source address need not be known), and also allow the 1741 signaling of previously unknown addresses, and of addresses belonging 1742 to other address families (e.g., both IPv4 and IPv6). 1744 Here is an example of typical operation of the protocol: 1746 o An MPTCP connection is initially set up between address/port A1 of 1747 Host A and address/port B1 of Host B. If Host A is multihomed and 1748 multiaddressed, it can start an additional subflow from its 1749 address A2 to B1, by sending a SYN with a Join option from A2 to 1750 B1, using B's previously declared token for this connection. 1751 Alternatively, if B is multihomed, it can try to set up a new 1752 subflow from B2 to A1, using A's previously declared token. In 1753 either case, the SYN will be sent to the port already in use for 1754 the original subflow on the receiving host. 1756 o Simultaneously (or after a timeout), an ADD_ADDR option 1757 (Section 3.4.1) is sent on an existing subflow, informing the 1758 receiver of the sender's alternative address(es). The recipient 1759 can use this information to open a new subflow to the sender's 1760 additional address. In our example, A will send ADD_ADDR option 1761 informing B of address/port A2. The mix of using the SYN-based 1762 option and the ADD_ADDR option, including timeouts, is 1763 implementation specific and can be tailored to agree with local 1764 policy. 1766 o If subflow A2-B1 is successfully set up, Host B can use the 1767 Address ID in the Join option to correlate this with the ADD_ADDR 1768 option that will also arrive on an existing subflow; now B knows 1769 not to open A2-B1, ignoring the ADD_ADDR. Otherwise, if B has not 1770 received the A2-B1 MP_JOIN SYN but received the ADD_ADDR, it can 1771 try to initiate a new subflow from one or more of its addresses to 1772 address A2. This permits new sessions to be opened if one host is 1773 behind a NAT. 1775 Other ways of using the two signaling mechanisms are possible; for 1776 instance, signaling addresses in other address families can only be 1777 done explicitly using the Add Address option. 1779 3.4.1. Address Advertisement 1781 The Add Address (ADD_ADDR) MPTCP option announces additional 1782 addresses (and optionally, ports) on which a host can be reached 1783 (Figure 12). This option can be used at any time during a 1784 connection, depending on when the sender wishes to enable multiple 1785 paths and/or when paths become available. As with all MPTCP signals, 1786 the receiver MUST undertake standard TCP validity checks, e.g. 1787 [RFC5961], before acting upon it. 1789 Every address has an Address ID that can be used for uniquely 1790 identifying the address within a connection for address removal. 1791 This is also used to identify MP_JOIN options (see Section 3.2) 1792 relating to the same address, even when address translators are in 1793 use. The Address ID MUST uniquely identify the address to the sender 1794 (within the scope of the connection), but the mechanism for 1795 allocating such IDs is implementation specific. 1797 All address IDs learned via either MP_JOIN or ADD_ADDR SHOULD be 1798 stored by the receiver in a data structure that gathers all the 1799 Address ID to address mappings for a connection (identified by a 1800 token pair). In this way, there is a stored mapping between Address 1801 ID, observed source address, and token pair for future processing of 1802 control information for a connection. Note that an implementation 1803 MAY discard incoming address advertisements at will, for example, for 1804 avoiding the required mapping state, or because advertised addresses 1805 are of no use to it (for example, IPv6 addresses when it has IPv4 1806 only). Therefore, a host MUST treat address advertisements as soft 1807 state, and it MAY choose to refresh advertisements periodically. 1809 This option is shown in Figure 12. The illustration is sized for 1810 IPv4 addresses. For IPv6, the length of the address will be 16 1811 octets (instead of 4). 1813 The 2 octets that specify the TCP port number to use are optional and 1814 their presence can be inferred from the length of the option. 1815 Although it is expected that the majority of use cases will use the 1816 same port pairs as used for the initial subflow (e.g., port 80 1817 remains port 80 on all subflows, as does the ephemeral port at the 1818 client), there may be cases (such as port-based load balancing) where 1819 the explicit specification of a different port is required. If no 1820 port is specified, MPTCP SHOULD attempt to connect to the specified 1821 address on the same port as is already in use by the subflow on which 1822 the ADD_ADDR signal was sent; this is discussed in more detail in 1823 Section 3.10. 1825 The Truncated HMAC present in this Option is the rightmost 64 bits of 1826 an HMAC, negotiated and calculated in the same way as for MP_JOIN as 1827 described in Section 3.2. For this specification of MPTCP, as there 1828 is only one hash algorithm option specified, this will be HMAC as 1829 defined in [RFC2104], using the SHA-256 hash algorithm [SHS], 1830 implemented as in [RFC6234]. In the same way as for MP_JOIN, the key 1831 for the HMAC algorithm, in the case of the message transmitted by 1832 Host A, will be Key-A followed by Key-B, and in the case of Host B, 1833 Key-B followed by Key-A. These are the keys that were exchanged in 1834 the original MP_CAPABLE handshake. The message for the HMAC is the 1835 Address ID, IP Address, and Port which precede the HMAC in the 1836 ADD_ADDR option. If the port is not present in the ADD_ADDR option, 1837 the HMAC message will nevertheless include two octets of value zero. 1838 The rationale for the HMAC is to prevent unauthorized entities from 1839 injecting ADD_ADDR signals in an attempt to hijack a connection. 1841 Note that additionally the presence of this HMAC prevents the address 1842 being changed in flight unless the key is known by an intermediary. 1843 If a host receives an ADD_ADDR option for which it cannot validate 1844 the HMAC, it SHOULD silently ignore the option. 1846 A set of four flags are present after the subtype and before the 1847 Address ID. Only the rightmost bit - labelled 'E' - is assinged 1848 today. The other bits are currently unassigned and MUST be set to 1849 zero by a sender and MUST be ignored by the receiver. 1851 The 'E' bit exists to provide reliability for this option. Because 1852 this option will often be sent on pure ACKs, there is no guarantee of 1853 reliability. Therefore, a receiver receiving a fresh ADD_ADDR option 1854 (where E=0), will send the same option back to the sender, but not 1855 including the HMAC, and with E=1. The lack of this echo can be used 1856 by the initial ADD_ADDR sender to retransmit the ADD_ADDR according 1857 to local policy. 1859 1 2 3 1860 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 1861 +---------------+---------------+-------+-------+---------------+ 1862 | Kind | Length |Subtype|(rsv)|E| Address ID | 1863 +---------------+---------------+-------+-------+---------------+ 1864 | Address (IPv4 - 4 octets / IPv6 - 16 octets) | 1865 +-------------------------------+-------------------------------+ 1866 | Port (2 octets, optional) | | 1867 +-------------------------------+ | 1868 | Truncated HMAC (8 octets, if length > 10 octets) | 1869 | +-------------------------------+ 1870 | | 1871 +-------------------------------+ 1873 Figure 12: Add Address (ADD_ADDR) Option 1875 Due to the proliferation of NATs, it is reasonably likely that one 1876 host may attempt to advertise private addresses [RFC1918]. It is not 1877 desirable to prohibit this, since there may be cases where both hosts 1878 have additional interfaces on the same private network, and a host 1879 MAY want to advertise such addresses. The MP_JOIN handshake to 1880 create a new subflow (Section 3.2) provides mechanisms to minimize 1881 security risks. The MP_JOIN message contains a 32-bit token that 1882 uniquely identifies the connection to the receiving host. If the 1883 token is unknown, the host will return with a RST. In the unlikely 1884 event that the token is known, subflow setup will continue, but the 1885 HMAC exchange must occur for authentication. This will fail, and 1886 will provide sufficient protection against two unconnected hosts 1887 accidentally setting up a new subflow upon the signal of a private 1888 address. Further security considerations around the issue of 1889 ADD_ADDR messages that accidentally misdirect, or maliciously direct, 1890 new MP_JOIN attempts are discussed in Section 5. 1892 Ideally, ADD_ADDR and REMOVE_ADDR options would be sent reliably, and 1893 in order, to the other end. This would ensure that this address 1894 management does not unnecessarily cause an outage in the connection 1895 when remove/add addresses are processed in reverse order, and also to 1896 ensure that all possible paths are used. Note, however, that losing 1897 reliability and ordering will not break the multipath connections, it 1898 will just reduce the opportunity to open multipath paths and to 1899 survive different patterns of path failures. 1901 Therefore, implementing reliability signals for these MPTCP options 1902 is not necessary. In order to minimize the impact of the loss of 1903 these options, however, it is RECOMMENDED that a sender should send 1904 these options on all available subflows. If these options need to be 1905 received in order, an implementation SHOULD only send one ADD_ADDR/ 1906 REMOVE_ADDR option per RTT, to minimize the risk of misordering. 1908 A host can send an ADD_ADDR message with an already assigned Address 1909 ID, but the Address MUST be the same as previously assigned to this 1910 Address ID, and the Port MUST be different from one already in use 1911 for this Address ID. If these conditions are not met, the receiver 1912 SHOULD silently ignore the ADD_ADDR. A host wishing to replace an 1913 existing Address ID MUST first remove the existing one 1914 (Section 3.4.2). 1916 A host that receives an ADD_ADDR but finds a connection set up to 1917 that IP address and port number is unsuccessful SHOULD NOT perform 1918 further connection attempts to this address/port combination for this 1919 connection. A sender that wants to trigger a new incoming connection 1920 attempt on a previously advertised address/port combination can 1921 therefore refresh ADD_ADDR information by sending the option again. 1923 During normal MPTCP operation, it is unlikely that there will be 1924 sufficient TCP option space for ADD_ADDR to be included along with 1925 those for data sequence numbering (Section 3.3.1). Therefore, it is 1926 expected that an MPTCP implementation will send the ADD_ADDR option 1927 on separate ACKs. As discussed earlier, however, an MPTCP 1928 implementation MUST NOT treat duplicate ACKs with any MPTCP option, 1929 with the exception of the DSS option, as indications of congestion 1930 [RFC5681], and an MPTCP implementation SHOULD NOT send more than two 1931 duplicate ACKs in a row for signaling purposes. 1933 3.4.2. Remove Address 1935 If, during the lifetime of an MPTCP connection, a previously 1936 announced address becomes invalid (e.g., if the interface 1937 disappears), the affected host SHOULD announce this so that the peer 1938 can remove subflows related to this address. 1940 This is achieved through the Remove Address (REMOVE_ADDR) option 1941 (Figure 13), which will remove a previously added address (or list of 1942 addresses) from a connection and terminate any subflows currently 1943 using that address. 1945 For security purposes, if a host receives a REMOVE_ADDR option, it 1946 must ensure the affected path(s) are no longer in use before it 1947 instigates closure. The receipt of REMOVE_ADDR SHOULD first trigger 1948 the sending of a TCP keepalive [RFC1122] on the path, and if a 1949 response is received the path SHOULD NOT be removed. Typical TCP 1950 validity tests on the subflow (e.g., ensuring sequence and ACK 1951 numbers are correct) MUST also be undertaken. An implementation can 1952 use indications of these test failures as part of intrusion detection 1953 or error logging. 1955 The sending and receipt (if no keepalive response was received) of 1956 this message SHOULD trigger the sending of RSTs by both hosts on the 1957 affected subflow(s) (if possible), as a courtesy to cleaning up 1958 middlebox state, before cleaning up any local state. 1960 Address removal is undertaken by ID, so as to permit the use of NATs 1961 and other middleboxes that rewrite source addresses. If there is no 1962 address at the requested ID, the receiver will silently ignore the 1963 request. 1965 A subflow that is still functioning MUST be closed with a FIN 1966 exchange as in regular TCP, rather than using this option. For more 1967 information, see Section 3.3.3. 1969 1 2 3 1970 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 1971 +---------------+---------------+-------+-------+---------------+ 1972 | Kind | Length = 3+n |Subtype|(resvd)| Address ID | ... 1973 +---------------+---------------+-------+-------+---------------+ 1974 (followed by n-1 Address IDs, if required) 1976 Figure 13: Remove Address (REMOVE_ADDR) Option 1978 3.5. Fast Close 1980 Regular TCP has the means of sending a reset (RST) signal to abruptly 1981 close a connection. With MPTCP, the RST only has the scope of the 1982 subflow and will only close the concerned subflow but not affect the 1983 remaining subflows. MPTCP's connection will stay alive at the data 1984 level, in order to permit break-before-make handover between 1985 subflows. It is therefore necessary to provide an MPTCP-level 1986 "reset" to allow the abrupt closure of the whole MPTCP connection, 1987 and this is the MP_FASTCLOSE option. 1989 MP_FASTCLOSE is used to indicate to the peer that the connection will 1990 be abruptly closed and no data will be accepted anymore. The reasons 1991 for triggering an MP_FASTCLOSE are implementation specific. Regular 1992 TCP does not allow sending a RST while the connection is in a 1993 synchronized state [RFC0793]. Nevertheless, implementations allow 1994 the sending of a RST in this state, if, for example, the operating 1995 system is running out of resources. In these cases, MPTCP should 1996 send the MP_FASTCLOSE. This option is illustrated in Figure 14. 1998 1 2 3 1999 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 2000 +---------------+---------------+-------+-----------------------+ 2001 | Kind | Length |Subtype| (reserved) | 2002 +---------------+---------------+-------+-----------------------+ 2003 | Option Receiver's Key | 2004 | (64 bits) | 2005 | | 2006 +---------------------------------------------------------------+ 2008 Figure 14: Fast Close (MP_FASTCLOSE) Option 2010 If Host A wants to force the closure of an MPTCP connection, the 2011 MPTCP Fast Close procedure is as follows: 2013 o Host A sends an ACK containing the MP_FASTCLOSE option on one 2014 subflow, containing the key of Host B as declared in the initial 2015 connection handshake. On all the other subflows, Host A sends a 2016 regular TCP RST to close these subflows, and tears them down. 2017 Host A now enters FASTCLOSE_WAIT state. 2019 o Upon receipt of an MP_FASTCLOSE, containing the valid key, Host B 2020 answers on the same subflow with a TCP RST and tears down all 2021 subflows. Host B can now close the whole MPTCP connection (it 2022 transitions directly to CLOSED state). 2024 o As soon as Host A has received the TCP RST on the remaining 2025 subflow, it can close this subflow and tear down the whole 2026 connection (transition from FASTCLOSE_WAIT to CLOSED states). If 2027 Host A receives an MP_FASTCLOSE instead of a TCP RST, both hosts 2028 attempted fast closure simultaneously. Host A should reply with a 2029 TCP RST and tear down the connection. 2031 o If Host A does not receive a TCP RST in reply to its MP_FASTCLOSE 2032 after one retransmission timeout (RTO) (the RTO of the subflow 2033 where the MPTCP_RST has been sent), it SHOULD retransmit the 2034 MP_FASTCLOSE. The number of retransmissions SHOULD be limited to 2035 avoid this connection from being retained for a long time, but 2036 this limit is implementation specific. A RECOMMENDED number is 3. 2037 If no TCP RST is received in response, Host A SHOULD send a TCP 2038 RST itself when it releases state in order to clear any remaining 2039 state at middleboxes. 2041 3.6. Subflow Reset 2043 As discussed in Section 3.5 above, the MP_FASTCLOSE option provides a 2044 connection-level reset roughly analagous to a TCP RST. Regular TCP 2045 RST options remain used to at the subflow-level to indicate the 2046 receiving host has no knowledge of the MPTCP subflow or TCP 2047 connection to which the packet belongs. 2049 However, in MPTCP, there may be many reasons for rejecting the 2050 opening of a subflow, but these semantics cannot be carried in a 2051 standard TCP RST. It would be beneficial for a host to the reasons 2052 why its subflow has been closed with a RST, and thus whether it 2053 should try to re-establish the subflow immediately, later, or never 2054 again. These semantics are carried in the MP_TCPRST option that can 2055 be included on a TCP RST packet. 2057 1 2 3 2058 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 2059 +---------------+---------------+-------+-----------------------+ 2060 | Kind | Length |Subtype|U|V|W|T| Reason | 2061 +---------------+---------------+-------+-----------------------+ 2063 Figure 15: TCP RST Reason (MP_TCPRST) Option 2065 The MP_TCPRST option contains a reason code that allows the sender of 2066 the option to provide more information about the reason for the 2067 termination of the subflow. Using 12 bits of option space, the first 2068 four bits are reserved for flags (only one of which is currently 2069 defined), and the remaining octet is used to express a reason code 2070 for this subflow termination, from which a receiver MAY infer 2071 information about the usability of this path. 2073 The "T" flag is used by the sender to indicate whether the error 2074 condition that is reported is Transient (T bit set to 1) or Permanent 2075 (T bit set to 0). If the error condition is considered to be 2076 Transient by the sender of the RST segment, the recipient of this 2077 segment MAY try to reestablish a subflow for this connection over the 2078 failed path. The time at which a receiver may try to re-establish 2079 this is implementation-specific, but SHOULD take into account the 2080 properties of the failure defined by the following reason code. If 2081 the error condition is considered to be permanent, the receiver of 2082 the RST segment SHOULD NOT try to reestablish a subflow for this 2083 connection over this path. The "U", "V" and "W" flags are not 2084 defined by this specification and are reserved for future use. 2086 The "Reason" code is an 8-bit field that indicates the reason for the 2087 termination of the subflow. The following codes are defined in this 2088 document: 2090 o Unspecified error (code 0x0). This is the default error implying 2091 the subflow is not longer available. The receiving host SHOULD 2092 take account of the 'T' bit in deciding whether to re-estbalish 2093 this subflow. The presence of this option shows that the RST was 2094 generated by a MPTCP-aware device. 2096 o MPTCP specific error (code 0x01). An error has been detected in 2097 the processing of MPTCP options. This is the usual reason code to 2098 return in the cases where a RST is being sent to close a subflow 2099 for reasons of an invalid response. 2101 o Lack of resources (code 0x02). This code indicates that the 2102 sending host does not have enough ressources to support the 2103 terminated subflow. 2105 o Administratively prohibited (code 0x03). This code indicates that 2106 the requested subflow is prohibited by the policies of the sending 2107 host. 2109 o Too much outstanding data (code 0x04). This code indicates that 2110 there is an excessive amount of data that need to be transmitted 2111 over the terminated subflow while having already been acknowledged 2112 over one or more other subflows. This may occur if a path has 2113 been unavailable for a short period and it is more efficient to 2114 reset and start again than it is to retransmit the queued data. 2116 o Unacceptable performance (code 0x05). This code indicates that 2117 the performance of this subflow was too low compared to the other 2118 subflows of this Multipath TCP connection. 2120 o Middlebox interference (code 0x06). Middlebox interference has 2121 been detected over this subflow making MPTCP signaling invalid. 2122 For example, this may be sent if the checksum does not validate. 2124 3.7. MPTCP Experimental Option 2126 In order to provide a structured identity and negotiation mechanism 2127 for private experimental MPTCP extensions, the MP_EXPERIMENTAL option 2128 has been reserved. 2130 1 2 3 2131 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 2132 +---------------+---------------+-------+-------+---------------+ 2133 | Kind | Length |Subtype|S|U|rsv| Experiment | 2134 +---------------+---------------+-------+-------+---------------+ 2135 | Id. (16 bits) | Subtype-specific data (variable length) ... 2136 +----------------------------------------------------------- ... 2138 Figure 16: MPTCP Experimental (MP_EXPERIMENTAL) Option 2140 Figure 16 shows the format of the experimental option. The 2141 Experiment identifier is a 16 bits integer that shall be assigned by 2142 using the same procedure as defined in [RFC6994]; a request to IANA 2143 is made in Section 8.4. 2145 The two high order flags that are included in the MPTCP Experimental 2146 option have the following semantics: 2148 o "S" flag (highest order bit) : This is the synchronising bit. 2149 When set to 1, it indicates that the host sending this option 2150 expects a reply from the remote host with an option having the 2151 same experiment identifier, but possibly containing other data. 2153 o "U" flag (second highest order bit) : When set to 1, this flag 2154 indicates that the experimental option was received by the sending 2155 host but it was unable to parse it. 2157 The two low order flags are currently reserved for further use. They 2158 MUST be set to zero when sending and ignored upon reception. 2160 To use the Experimental MPTCP option with a given experiment 2161 identifier over a MPTCP connection, the sending host must first 2162 verify the ability of the remote host to support this particular 2163 Experimental option. For this, it first sends in any valid TCP 2164 segment, including a duplicate acknowledgement, an Experimental MPTCP 2165 option with the "S" flag set. Upon reception of this option, the 2166 receiving host will verify whether it supports it. If yes, it shall 2167 return a TCP segment that contains the experimental option with the 2168 same identifier and the "S" and the "U" flags both set to 1. This 2169 option may contain additional data depending on the semantics of the 2170 extension. If the receiving host does not recognise the experimental 2171 option that it has received, it shall return a TCP segment that 2172 contains the received experimental option with the "S" flag set to 0 2173 and the "U" flag set to 1. 2175 If a host receives an Experimental MPTCP option with the "U" flag set 2176 to 0 which it does not support, or which contains information that 2177 the host cannot parse, it shall return the exact option that it 2178 received with the "U" flag set to 1 to indicate the error to the 2179 remote host. If an invalid option is received with the "U" flag set 2180 to 0, it must be silently discarded. 2182 Future documents specifying new experimental MPTCP options should 2183 specify the extract semantic of the Subtype-specific data and whether 2184 additional validation operations are to be followed at both sides. 2185 It should be noted that data can be included in an experimental 2186 option concurrently with the capability check (S/U). 2188 3.8. Fallback 2190 Sometimes, middleboxes will exist on a path that could prevent the 2191 operation of MPTCP. MPTCP has been designed in order to cope with 2192 many middlebox modifications (see Section 6), but there are still 2193 some cases where a subflow could fail to operate within the MPTCP 2194 requirements. These cases are notably the following: the loss of 2195 MPTCP options on a path and the modification of payload data. If 2196 such an event occurs, it is necessary to "fall back" to the previous, 2197 safe operation. This may be either falling back to regular TCP or 2198 removing a problematic subflow. 2200 At the start of an MPTCP connection (i.e., the first subflow), it is 2201 important to ensure that the path is fully MPTCP capable and the 2202 necessary MPTCP options can reach each host. The handshake as 2203 described in Section 3.1 SHOULD fall back to regular TCP if either of 2204 the SYN messages do not have the MPTCP options: this is the same, and 2205 desired, behavior in the case where a host is not MPTCP capable, or 2206 the path does not support the MPTCP options. When attempting to join 2207 an existing MPTCP connection (Section 3.2), if a path is not MPTCP 2208 capable and the MPTCP options do not get through on the SYNs, the 2209 subflow will be closed according to the MP_JOIN logic. 2211 There is, however, another corner case that should be addressed. 2212 That is one of MPTCP options getting through on the SYN, but not on 2213 regular packets. This can be resolved if the subflow is the first 2214 subflow, and thus all data in flight is contiguous, using the 2215 following rules. 2217 A sender MUST include a DSS option with data sequence mapping in 2218 every segment until one of the sent segments has been acknowledged 2219 with a DSS option containing a Data ACK. Upon reception of the 2220 acknowledgment, the sender has the confirmation that the DSS option 2221 passes in both directions and may choose to send fewer DSS options 2222 than once per segment. 2224 If, however, an ACK is received for data (not just for the SYN) 2225 without a DSS option containing a Data ACK, the sender determines the 2226 path is not MPTCP capable. In the case of this occurring on an 2227 additional subflow (i.e., one started with MP_JOIN), the host MUST 2228 close the subflow with a RST. In the case of the first subflow 2229 (i.e., that started with MP_CAPABLE), it MUST drop out of an MPTCP 2230 mode back to regular TCP. The sender will send one final data 2231 sequence mapping, with the Data-Level Length value of 0 indicating an 2232 infinite mapping (in case the path drops options in one direction 2233 only), and then revert to sending data on the single subflow without 2234 any MPTCP options. 2236 Note that this rule essentially prohibits the sending of data on the 2237 third packet of an MP_CAPABLE or MP_JOIN handshake, since both that 2238 option and a DSS cannot fit in TCP option space. If the initiator is 2239 to send first, another segment must be sent that contains the data 2240 and DSS. Note also that an additional subflow cannot be used until 2241 the initial path has been verified as MPTCP capable. 2243 If a subflow breaks during operation, e.g. if it is re-routed and 2244 MPTCP options are no longer permitted, then once this is detected (by 2245 the subflow-level receive buffer filling up), the subflow SHOULD be 2246 treated as broken and closed with a RST, since no data can be 2247 delivered to the application layer, and no fallback signal can be 2248 reliably sent. This RST SHOULD include the MP_TCPRST option 2249 (Section 3.6) with an appropriate reason code. 2251 These rules should cover all cases where such a failure could happen: 2252 whether it's on the forward or reverse path and whether the server or 2253 the client first sends data. If lost options on data packets occur 2254 on any other subflow apart from the initial subflow, it should be 2255 treated as a standard path failure. The data would not be DATA_ACKed 2256 (since there is no mapping for the data), and the subflow can be 2257 closed with a RST, containing a MP_TCPRST option (Section 3.6) with 2258 an appropriate reason code. 2260 The case described above is a specialized case of fallback, for when 2261 the lack of MPTCP support is detected before any data is acknowledged 2262 at the connection level on a subflow. More generally, fallback 2263 (either closing a subflow, or to regular TCP) can become necessary at 2264 any point during a connection if a non-MPTCP-aware middlebox changes 2265 the data stream. 2267 As described in Section 3.3, each portion of data for which there is 2268 a mapping is protected by a checksum, if checksums have been 2269 negotiated. This mechanism is used to detect if middleboxes have 2270 made any adjustments to the payload (added, removed, or changed 2271 data). A checksum will fail if the data has been changed in any way. 2272 This will also detect if the length of data on the subflow is 2273 increased or decreased, and this means the data sequence mapping is 2274 no longer valid. The sender no longer knows what subflow-level 2275 sequence number the receiver is genuinely operating at (the middlebox 2276 will be faking ACKs in return), and it cannot signal any further 2277 mappings. Furthermore, in addition to the possibility of payload 2278 modifications that are valid at the application layer, there is the 2279 possibility that false positives could be hit across MPTCP segment 2280 boundaries, corrupting the data. Therefore, all data from the start 2281 of the segment that failed the checksum onwards is not trustworthy. 2283 Note that if checksum usage has not been negotiated, this fallback 2284 mechanism cannot be used unless there is some higher or lower layer 2285 signal to inform the MPTCP implementation that the payload has been 2286 tampered with. 2288 When multiple subflows are in use, the data in flight on a subflow 2289 will likely involve data that is not contiguously part of the 2290 connection-level stream, since segments will be spread across the 2291 multiple subflows. Due to the problems identified above, it is not 2292 possible to determine what the adjustment has done to the data 2293 (notably, any changes to the subflow sequence numbering). Therefore, 2294 it is not possible to recover the subflow, and the affected subflow 2295 must be immediately closed with a RST, featuring an MP_FAIL option 2296 (Figure 17), which defines the data sequence number at the start of 2297 the segment (defined by the data sequence mapping) that had the 2298 checksum failure. Note that the MP_FAIL option requires the use of 2299 the full 64-bit sequence number, even if 32-bit sequence numbers are 2300 normally in use in the DSS signals on the path. 2302 1 2 3 2303 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 2304 +---------------+---------------+-------+----------------------+ 2305 | Kind | Length=12 |Subtype| (reserved) | 2306 +---------------+---------------+-------+----------------------+ 2307 | | 2308 | Data Sequence Number (8 octets) | 2309 | | 2310 +--------------------------------------------------------------+ 2312 Figure 17: Fallback (MP_FAIL) Option 2314 The receiver MUST discard all data following the data sequence number 2315 specified. Failed data MUST NOT be DATA_ACKed and so will be 2316 retransmitted on other subflows (Section 3.3.6). 2318 A special case is when there is a single subflow and it fails with a 2319 checksum error. If it is known that all unacknowledged data in 2320 flight is contiguous (which will usually be the case with a single 2321 subflow), an infinite mapping can be applied to the subflow without 2322 the need to close it first, and essentially turn off all further 2323 MPTCP signaling. In this case, if a receiver identifies a checksum 2324 failure when there is only one path, it will send back an MP_FAIL 2325 option on the subflow-level ACK, referring to the data-level sequence 2326 number of the start of the segment on which the checksum error was 2327 detected. The sender will receive this, and if all unacknowledged 2328 data in flight is contiguous, will signal an infinite mapping. This 2329 infinite mapping will be a DSS option (Section 3.3) on the first new 2330 packet, containing a data sequence mapping that acts retroactively, 2331 referring to the start of the subflow sequence number of the most 2332 recent segment that was known to be delivered intact (i.e. was 2333 successfully DATA_ACKed). From that point onwards, data can be 2334 altered by a middlebox without affecting MPTCP, as the data stream is 2335 equivalent to a regular, legacy TCP session. The MP_FAIL signal 2336 affects only one direction of traffic. It is not mandatory for the 2337 reciever of an MP_FAIL to also respond with an MP_FAIL, since the 2338 paths may only be damaged in one direction. However, implementations 2339 MAY choose to send a MP_FAIL in the reverse direction and entirely 2340 revert to a regular TCP session. 2342 In the rare case that the data is not contiguous (which could happen 2343 when there is only one subflow but it is retransmitting data from a 2344 subflow that has recently been uncleanly closed), the receiver MUST 2345 close the subflow with a RST with MP_FAIL. The receiver MUST discard 2346 all data that follows the data sequence number specified. The sender 2347 MAY attempt to create a new subflow belonging to the same connection, 2348 and, if it chooses to do so, SHOULD place the single subflow 2349 immediately in single-path mode by setting an infinite data sequence 2350 mapping. This mapping will begin from the data-level sequence number 2351 that was declared in the MP_FAIL. 2353 After a sender signals an infinite mapping, it MUST only use subflow 2354 ACKs to clear its send buffer. This is because Data ACKs may become 2355 misaligned with the subflow ACKs when middleboxes insert or delete 2356 data. The receive SHOULD stop generating Data ACKs after it receives 2357 an infinite mapping. 2359 When a connection has fallen back, only one subflow can send data; 2360 otherwise, the receiver would not know how to reorder the data. In 2361 practice, this means that all MPTCP subflows will have to be 2362 terminated except one. Once MPTCP falls back to regular TCP, it MUST 2363 NOT revert to MPTCP later in the connection. 2365 It should be emphasized that we are not attempting to prevent the use 2366 of middleboxes that want to adjust the payload. An MPTCP-aware 2367 middlebox could provide such functionality by also rewriting 2368 checksums. 2370 3.9. Error Handling 2372 In addition to the fallback mechanism as described above, the 2373 standard classes of TCP errors may need to be handled in an MPTCP- 2374 specific way. Note that changing semantics -- such as the relevance 2375 of a RST -- are covered in Section 4. Where possible, we do not want 2376 to deviate from regular TCP behavior. 2378 The following list covers possible errors and the appropriate MPTCP 2379 behavior: 2381 o Unknown token in MP_JOIN (or HMAC failure in MP_JOIN ACK, or 2382 missing MP_JOIN in SYN/ACK response): send RST (analogous to TCP's 2383 behavior on an unknown port) 2385 o DSN out of window (during normal operation): drop the data, do not 2386 send Data ACKs 2388 o Remove request for unknown address ID: silently ignore 2390 3.10. Heuristics 2392 There are a number of heuristics that are needed for performance or 2393 deployment but that are not required for protocol correctness. In 2394 this section, we detail such heuristics. Note that discussion of 2395 buffering and certain sender and receiver window behaviors are 2396 presented in Sections 3.3.4 and 3.3.5, as well as retransmission in 2397 Section 3.3.6. 2399 3.10.1. Port Usage 2401 Under typical operation, an MPTCP implementation SHOULD use the same 2402 ports as already in use. In other words, the destination port of a 2403 SYN containing an MP_JOIN option SHOULD be the same as the remote 2404 port of the first subflow in the connection. The local port for such 2405 SYNs SHOULD also be the same as for the first subflow (and as such, 2406 an implementation SHOULD reserve ephemeral ports across all local IP 2407 addresses), although there may be cases where this is infeasible. 2408 This strategy is intended to maximize the probability of the SYN 2409 being permitted by a firewall or NAT at the recipient and to avoid 2410 confusing any network monitoring software. 2412 There may also be cases, however, where the passive opener wishes to 2413 signal to the other host that a specific port should be used, and 2414 this facility is provided in the Add Address option as documented in 2415 Section 3.4.1. It is therefore feasible to allow multiple subflows 2416 between the same two addresses but using different port pairs, and 2417 such a facility could be used to allow load balancing within the 2418 network based on 5-tuples (e.g., some ECMP implementations 2419 [RFC2992]). 2421 3.10.2. Delayed Subflow Start and Subflow Symmetry 2423 Many TCP connections are short-lived and consist only of a few 2424 segments, and so the overheads of using MPTCP outweigh any benefits. 2425 A heuristic is required, therefore, to decide when to start using 2426 additional subflows in an MPTCP connection. We expect that 2427 experience gathered from deployments will provide further guidance on 2428 this, and will be affected by particular application characteristics 2429 (which are likely to change over time). However, a suggested 2430 general-purpose heuristic that an implementation MAY choose to employ 2431 is as follows. Results from experimental deployments are needed in 2432 order to verify the correctness of this proposal. 2434 If a host has data buffered for its peer (which implies that the 2435 application has received a request for data), the host opens one 2436 subflow for each initial window's worth of data that is buffered. 2438 Consideration should also be given to limiting the rate of adding new 2439 subflows, as well as limiting the total number of subflows open for a 2440 particular connection. A host may choose to vary these values based 2441 on its load or knowledge of traffic and path characteristics. 2443 Note that this heuristic alone is probably insufficient. Traffic for 2444 many common applications, such as downloads, is highly asymmetric and 2445 the host that is multihomed may well be the client that will never 2446 fill its buffers, and thus never use MPTCP. Advanced APIs that allow 2447 an application to signal its traffic requirements would aid in these 2448 decisions. 2450 An additional time-based heuristic could be applied, opening 2451 additional subflows after a given period of time has passed. This 2452 would alleviate the above issue, and also provide resilience for low- 2453 bandwidth but long-lived applications. 2455 If the two communicating hosts immediately try to set up subflows 2456 from all available addresses to all available addresses on the other 2457 host, this could end up creating two subflows per path. This is an 2458 inefficient use of resources. 2460 If the the same ports are used on all subflows, as recommended above, 2461 then standard TCP simultaneous open logic should take care of this 2462 situation and only one subflow will be established between the 2463 address pairs. However, this relies on the same ports being used at 2464 both end hosts. If a host does not support TCP simultaneous open, it 2465 is RECOMMENDED that some element of randomization is applied to the 2466 time waited before opening new subflows, so that only one subflow 2467 exists between a given address pair. If, however, hosts signal 2468 additional ports to use (for example, for leveraging ECMP on-path), 2469 this heuristic need not apply. 2471 This section has shown some of the considerations that an implementer 2472 should give when developing MPTCP heuristics, but is not intended to 2473 be prescriptive. 2475 3.10.3. Failure Handling 2477 Requirements for MPTCP's handling of unexpected signals have been 2478 given in Section 3.9. There are other failure cases, however, where 2479 a hosts can choose appropriate behavior. 2481 For example, Section 3.1 suggests that a host SHOULD fall back to 2482 trying regular TCP SYNs after one or more failures of MPTCP SYNs for 2483 a connection. A host may keep a system-wide cache of such 2484 information, so that it can back off from using MPTCP, firstly for 2485 that particular destination host, and eventually on a whole 2486 interface, if MPTCP connections continue failing. 2488 Another failure could occur when the MP_JOIN handshake fails. 2489 Section 3.9 specifies that an incorrect handshake MUST lead to the 2490 subflow being closed with a RST. A host operating an active 2491 intrusion detection system may choose to start blocking MP_JOIN 2492 packets from the source host if multiple failed MP_JOIN attempts are 2493 seen. From the connection initiator's point of view, if an MP_JOIN 2494 fails, it SHOULD NOT attempt to connect to the same IP address and 2495 port during the lifetime of the connection, unless the other host 2496 refreshes the information with another ADD_ADDR option. Note that 2497 the ADD_ADDR option is informational only, and does not guarantee the 2498 other host will attempt a connection. 2500 In addition, an implementation may learn, over a number of 2501 connections, that certain interfaces or destination addresses 2502 consistently fail and may default to not trying to use MPTCP for 2503 these. Behavior could also be learned for particularly badly 2504 performing subflows or subflows that regularly fail during use, in 2505 order to temporarily choose not to use these paths. 2507 4. Semantic Issues 2509 In order to support multipath operation, the semantics of some TCP 2510 components have changed. To aid clarity, this section collects these 2511 semantic changes as a reference. 2513 Sequence number: The (in-header) TCP sequence number is specific to 2514 the subflow. To allow the receiver to reorder application data, 2515 an additional data-level sequence space is used. In this data- 2516 level sequence space, the initial SYN and the final DATA_FIN 2517 occupy 1 octet of sequence space. There is an explicit mapping of 2518 data sequence space to subflow sequence space, which is signaled 2519 through TCP options in data packets. 2521 ACK: The ACK field in the TCP header acknowledges only the subflow 2522 sequence number, not the data-level sequence space. 2523 Implementations SHOULD NOT attempt to infer a data-level 2524 acknowledgment from the subflow ACKs. This separates subflow- and 2525 connection-level processing at an end host. 2527 Duplicate ACK: A duplicate ACK that includes any MPTCP signaling 2528 (with the exception of the DSS option) MUST NOT be treated as a 2529 signal of congestion. To limit the chances of non-MPTCP-aware 2530 entities mistakenly interpreting duplicate ACKs as a signal of 2531 congestion, MPTCP SHOULD NOT send more than two duplicate ACKs 2532 containing (non-DSS) MPTCP signals in a row. 2534 Receive Window: The receive window in the TCP header indicates the 2535 amount of free buffer space for the whole data-level connection 2536 (as opposed to for this subflow) that is available at the 2537 receiver. This is the same semantics as regular TCP, but to 2538 maintain these semantics the receive window must be interpreted at 2539 the sender as relative to the sequence number given in the 2540 DATA_ACK rather than the subflow ACK in the TCP header. In this 2541 way, the original flow control role is preserved. Note that some 2542 middleboxes may change the receive window, and so a host SHOULD 2543 use the maximum value of those recently seen on the constituent 2544 subflows for the connection-level receive window, and also needs 2545 to maintain a subflow-level window for subflow-level processing. 2547 FIN: The FIN flag in the TCP header applies only to the subflow it 2548 is sent on, not to the whole connection. For connection-level FIN 2549 semantics, the DATA_FIN option is used. 2551 RST: The RST flag in the TCP header applies only to the subflow it 2552 is sent on, not to the whole connection. The MP_FASTCLOSE option 2553 provides the fast close functionality of a RST at the MPTCP 2554 connection level. 2556 Address List: Address list management (i.e., knowledge of the local 2557 and remote hosts' lists of available IP addresses) is handled on a 2558 per-connection basis (as opposed to per subflow, per host, or per 2559 pair of communicating hosts). This permits the application of 2560 per-connection local policy. Adding an address to one connection 2561 (either explicitly through an Add Address message, or implicitly 2562 through a Join) has no implication for other connections between 2563 the same pair of hosts. 2565 5-tuple: The 5-tuple (protocol, local address, local port, remote 2566 address, remote port) presented by kernel APIs to the application 2567 layer in a non-multipath-aware application is that of the first 2568 subflow, even if the subflow has since been closed and removed 2569 from the connection. This decision, and other related API issues, 2570 are discussed in more detail in [RFC6897]. 2572 5. Security Considerations 2574 As identified in [RFC6181], the addition of multipath capability to 2575 TCP will bring with it a number of new classes of threat. In order 2576 to prevent these, [RFC6182] presents a set of requirements for a 2577 security solution for MPTCP. The fundamental goal is for the 2578 security of MPTCP to be "no worse" than regular TCP today, and the 2579 key security requirements are: 2581 o Provide a mechanism to confirm that the parties in a subflow 2582 handshake are the same as in the original connection setup. 2584 o Provide verification that the peer can receive traffic at a new 2585 address before using it as part of a connection. 2587 o Provide replay protection, i.e., ensure that a request to add/ 2588 remove a subflow is 'fresh'. 2590 In order to achieve these goals, MPTCP includes a hash-based 2591 handshake algorithm documented in Sections 3.1 and 3.2. 2593 The security of the MPTCP connection hangs on the use of keys that 2594 are shared once at the start of the first subflow, and are never sent 2595 again over the network (unless used in the fast close mechanism, 2596 Section 3.5). To ease demultiplexing while not giving away any 2597 cryptographic material, future subflows use a truncated cryptographic 2598 hash of this key as the connection identification "token". The keys 2599 are concatenated and used as keys for creating Hash-based Message 2600 Authentication Codes (HMACs) used on subflow setup, in order to 2601 verify that the parties in the handshake are the same as in the 2602 original connection setup. It also provides verification that the 2603 peer can receive traffic at this new address. Replay attacks would 2604 still be possible when only keys are used; therefore, the handshakes 2605 use single-use random numbers (nonces) at both ends -- this ensures 2606 the HMAC will never be the same on two handshakes. Guidance on 2607 generating random numbers suitable for use as keys is given in 2608 [RFC4086] and discussed in Section 3.1. 2610 The use of crypto capability bits in the initial connection handshake 2611 to negotiate use of a particular algorithm allows the deployment of 2612 additional crypto mechanisms in the future. Note that this would be 2613 susceptible to bid-down attacks only if the attacker was on-path (and 2614 thus would be able to modify the data anyway). The security 2615 mechanism presented in this document should therefore protect against 2616 all forms of flooding and hijacking attacks discussed in [RFC6181]. 2618 During normal operation, regular TCP protection mechanisms (such as 2619 ensuring sequence numbers are in-window) will provide the same level 2620 of protection against attacks on individual TCP subflows as exists 2621 for regular TCP today. Implementations will introduce additional 2622 buffers compared to regular TCP, to reassemble data at the connection 2623 level. The application of window sizing will minimize the risk of 2624 denial-of-service attacks consuming resources. 2626 As discussed in Section 3.4.1, a host may advertise its private 2627 addresses, but these might point to different hosts in the receiver's 2628 network. The MP_JOIN handshake (Section 3.2) will ensure that this 2629 does not succeed in setting up a subflow to the incorrect host. 2630 However, it could still create unwanted TCP handshake traffic. This 2631 feature of MPTCP could be a target for denial-of-service exploits, 2632 with malicious participants in MPTCP connections encouraging the 2633 recipient to target other hosts in the network. Therefore, 2634 implementations should consider heuristics (Section 3.10) at both the 2635 sender and receiver to reduce the impact of this. 2637 A small security risk could theoretically exist with key reuse, but 2638 in order to accomplish a replay attack, both the sender and receiver 2639 keys, and the sender and receiver random numbers, in the MP_JOIN 2640 handshake (Section 3.2) would have to match. 2642 Whilst this specification defines a "medium" security solution, 2643 meeting the criteria specified at the start of this section and the 2644 threat analysis ([RFC6181]), since attacks only ever get worse, it is 2645 likely that a future Standards Track version of MPTCP would need to 2646 be able to support stronger security. There are several ways the 2647 security of MPTCP could potentially be improved; some of these would 2648 be compatible with MPTCP as defined in this document, whilst others 2649 may not be. For now, the best approach is to get experience with the 2650 current approach, establish what might work, and check that the 2651 threat analysis is still accurate. 2653 Possible ways of improving MPTCP security could include: 2655 o defining a new MPCTP cryptographic algorithm, as negotiated in 2656 MP_CAPABLE. A sub-case could be to include an additional 2657 deployment assumption, such as stateful servers, in order to allow 2658 a more powerful algorithm to be used. 2660 o defining how to secure data transfer with MPTCP, whilst not 2661 changing the signaling part of the protocol. 2663 o defining security that requires more option space, perhaps in 2664 conjunction with a "long options" proposal for extending the TCP 2665 options space (such as those surveyed in [TCPLO]), or perhaps 2666 building on the current approach with a second stage of MPTCP- 2667 option-based security. 2669 o revisiting the working group's decision to exclusively use TCP 2670 options for MPTCP signaling, and instead look at also making use 2671 of the TCP payloads. 2673 MPTCP has been designed with several methods available to indicate a 2674 new security mechanism, including: 2676 o available flags in MP_CAPABLE (Figure 4); 2678 o available subtypes in the MPTCP option (Figure 3); 2680 o the version field in MP_CAPABLE (Figure 4); 2682 6. Interactions with Middleboxes 2684 Multipath TCP was designed to be deployable in the present world. 2685 Its design takes into account "reasonable" existing middlebox 2686 behavior. In this section, we outline a few representative 2687 middlebox-related failure scenarios and show how Multipath TCP 2688 handles them. Next, we list the design decisions multipath has made 2689 to accommodate the different middleboxes. 2691 A primary concern is our use of a new TCP option. Middleboxes should 2692 forward packets with unknown options unchanged, yet there are some 2693 that don't. These we expect will either strip options and pass the 2694 data, drop packets with new options, copy the same option into 2695 multiple segments (e.g., when doing segmentation), or drop options 2696 during segment coalescing. 2698 MPTCP uses a single new TCP option "Kind", and all message types are 2699 defined by "subtype" values (see Section 8). This should reduce the 2700 chances of only some types of MPTCP options being passed, and instead 2701 the key differing characteristics are different paths, and the 2702 presence of the SYN flag. 2704 MPTCP SYN packets on the first subflow of a connection contain the 2705 MP_CAPABLE option (Section 3.1). If this is dropped, MPTCP SHOULD 2706 fall back to regular TCP. If packets with the MP_JOIN option 2707 (Section 3.2) are dropped, the paths will simply not be used. 2709 If a middlebox strips options but otherwise passes the packets 2710 unchanged, MPTCP will behave safely. If an MP_CAPABLE option is 2711 dropped on either the outgoing or the return path, the initiating 2712 host can fall back to regular TCP, as illustrated in Figure 18 and 2713 discussed in Section 3.1. 2715 Subflow SYNs contain the MP_JOIN option. If this option is stripped 2716 on the outgoing path, the SYN will appear to be a regular SYN to Host 2717 B. Depending on whether there is a listening socket on the target 2718 port, Host B will reply either with SYN/ACK or RST (subflow 2719 connection fails). When Host A receives the SYN/ACK it sends a RST 2720 because the SYN/ACK does not contain the MP_JOIN option and its 2721 token. Either way, the subflow setup fails, but otherwise does not 2722 affect the MPTCP connection as a whole. 2724 Host A Host B 2725 | Middlebox M | 2726 | | | 2727 | SYN(MP_CAPABLE) | SYN | 2728 |-------------------|---------------->| 2729 | SYN/ACK | 2730 |<------------------------------------| 2731 a) MP_CAPABLE option stripped on outgoing path 2733 Host A Host B 2734 | SYN(MP_CAPABLE) | 2735 |------------------------------------>| 2736 | Middlebox M | 2737 | | | 2738 | SYN/ACK |SYN/ACK(MP_CAPABLE)| 2739 |<----------------|-------------------| 2740 b) MP_CAPABLE option stripped on return path 2742 Figure 18: Connection Setup with Middleboxes that Strip Options from 2743 Packets 2745 We now examine data flow with MPTCP, assuming the flow is correctly 2746 set up, which implies the options in the SYN packets were allowed 2747 through by the relevant middleboxes. If options are allowed through 2748 and there is no resegmentation or coalescing to TCP segments, 2749 Multipath TCP flows can proceed without problems. 2751 The case when options get stripped on data packets has been discussed 2752 in the Fallback section. If a fraction of options are stripped, 2753 behavior is not deterministic. If some data sequence mappings are 2754 lost, the connection can continue so long as mappings exist for the 2755 subflow-level data (e.g., if multiple maps have been sent that 2756 reinforce each other). If some subflow-level space is left unmapped, 2757 however, the subflow is treated as broken and is closed, through the 2758 process described in Section 3.8. MPTCP should survive with a loss 2759 of some Data ACKs, but performance will degrade as the fraction of 2760 stripped options increases. We do not expect such cases to appear in 2761 practice, though: most middleboxes will either strip all options or 2762 let them all through. 2764 We end this section with a list of middlebox classes, their behavior, 2765 and the elements in the MPTCP design that allow operation through 2766 such middleboxes. Issues surrounding dropping packets with options 2767 or stripping options were discussed above, and are not included here: 2769 o NATs [RFC3022] (Network Address (and Port) Translators) change the 2770 source address (and often source port) of packets. This means 2771 that a host will not know its public-facing address for signaling 2772 in MPTCP. Therefore, MPTCP permits implicit address addition via 2773 the MP_JOIN option, and the handshake mechanism ensures that 2774 connection attempts to private addresses [RFC1918] do not cause 2775 problems. Explicit address removal is undertaken by an Address ID 2776 to allow no knowledge of the source address. 2778 o Performance Enhancing Proxies (PEPs) [RFC3135] might proactively 2779 ACK data to increase performance. MPTCP, however, relies on 2780 accurate congestion control signals from the end host, and non- 2781 MPTCP-aware PEPs will not be able to provide such signals. MPTCP 2782 will, therefore, fall back to single-path TCP, or close the 2783 problematic subflow (see Section 3.8). 2785 o Traffic Normalizers [norm] may not allow holes in sequence 2786 numbers, and may cache packets and retransmit the same data. 2787 MPTCP looks like standard TCP on the wire, and will not retransmit 2788 different data on the same subflow sequence number. In the event 2789 of a retransmission, the same data will be retransmitted on the 2790 original TCP subflow even if it is additionally retransmitted at 2791 the connection level on a different subflow. 2793 o Firewalls [RFC2979] might perform initial sequence number 2794 randomization on TCP connections. MPTCP uses relative sequence 2795 numbers in data sequence mapping to cope with this. Like NATs, 2796 firewalls will not permit many incoming connections, so MPTCP 2797 supports address signaling (ADD_ADDR) so that a multiaddressed 2798 host can invite its peer behind the firewall/NAT to connect out to 2799 its additional interface. 2801 o Intrusion Detection Systems look out for traffic patterns and 2802 content that could threaten a network. Multipath will mean that 2803 such data is potentially spread, so it is more difficult for an 2804 IDS to analyze the whole traffic, and potentially increases the 2805 risk of false positives. However, for an MPTCP-aware IDS, tokens 2806 can be read by such systems to correlate multiple subflows and 2807 reassemble for analysis. 2809 o Application-level middleboxes such as content-aware firewalls may 2810 alter the payload within a subflow, such as rewriting URIs in HTTP 2811 traffic. MPTCP will detect these using the checksum and close the 2812 affected subflow(s), if there are other subflows that can be used. 2813 If all subflows are affected, multipath will fall back to TCP, 2814 allowing such middleboxes to change the payload. MPTCP-aware 2815 middleboxes should be able to adjust the payload and MPTCP 2816 metadata in order not to break the connection. 2818 In addition, all classes of middleboxes may affect TCP traffic in the 2819 following ways: 2821 o TCP options may be removed, or packets with unknown options 2822 dropped, by many classes of middleboxes. It is intended that the 2823 initial SYN exchange, with a TCP option, will be sufficient to 2824 identify the path capabilities. If such a packet does not get 2825 through, MPTCP will end up falling back to regular TCP. 2827 o Segmentation/Coalescing (e.g., TCP segmentation offloading) might 2828 copy options between packets and might strip some options. 2829 MPTCP's data sequence mapping includes the relative subflow 2830 sequence number instead of using the sequence number in the 2831 segment. In this way, the mapping is independent of the packets 2832 that carry it. 2834 o The receive window may be shrunk by some middleboxes at the 2835 subflow level. MPTCP will use the maximum window at data level, 2836 but will also obey subflow-specific windows. 2838 7. Acknowledgments 2840 The authors gratefully acknowledge significant input into this 2841 document from Sebastien Barre and Andrew McDonald. 2843 The authors also wish to acknowledge reviews and contributions from 2844 Iljitsch van Beijnum, Lars Eggert, Marcelo Bagnulo, Robert Hancock, 2845 Pasi Sarolahti, Toby Moncaster, Philip Eardley, Sergio Lembo, 2846 Lawrence Conroy, Yoshifumi Nishida, Bob Briscoe, Stein Gjessing, 2847 Andrew McGregor, Georg Hampel, Anumita Biswas, Wes Eddy, Alexey 2848 Melnikov, Francis Dupont, Adrian Farrel, Barry Leiba, Robert Sparks, 2849 Sean Turner, Stephen Farrell, Martin Stiemerling, Gregory Detal, and 2850 Fabien Duchene. 2852 8. IANA Considerations 2854 This document updates [RFC6824] and as such IANA is requested to 2855 update the TCP option space registry to point to this document for 2856 Multipath TCP, as follows: 2858 +------+--------+-----------------------+---------------+ 2859 | Kind | Length | Meaning | Reference | 2860 +------+--------+-----------------------+---------------+ 2861 | 30 | N | Multipath TCP (MPTCP) | This document | 2862 +------+--------+-----------------------+---------------+ 2864 Table 1: TCP Option Kind Numbers 2866 8.1. MPTCP Option Subtypes 2868 The 4-bit MPTCP subtype sub-registry ("MPTCP Option Subtypes" under 2869 the "Transmission Control Protocol (TCP) Parameters" registry) was 2870 defined in [RFC6824]. This document defines one additional subtype 2871 (ADD_ADDR) and updates the references to this document for all sub- 2872 types except ADD_ADDR, which is deprecated. The updates are listed 2873 in the following table. 2875 +-------+-----------------+-------------------------+---------------+ 2876 | Value | Symbol | Name | Reference | 2877 +-------+-----------------+-------------------------+---------------+ 2878 | 0x0 | MP_CAPABLE | Multipath Capable | This | 2879 | | | | document, | 2880 | | | | Section 3.1 | 2881 | 0x1 | MP_JOIN | Join Connection | This | 2882 | | | | document, | 2883 | | | | Section 3.2 | 2884 | 0x2 | DSS | Data Sequence Signal | This | 2885 | | | (Data ACK and data | document, | 2886 | | | sequence mapping) | Section 3.3 | 2887 | 0x3 | ADD_ADDR | Add Address | This | 2888 | | | | document, | 2889 | | | | Section 3.4.1 | 2890 | 0x4 | REMOVE_ADDR | Remove Address | This | 2891 | | | | document, | 2892 | | | | Section 3.4.2 | 2893 | 0x5 | MP_PRIO | Change Subflow Priority | This | 2894 | | | | document, | 2895 | | | | Section 3.3.8 | 2896 | 0x6 | MP_FAIL | Fallback | This | 2897 | | | | document, | 2898 | | | | Section 3.8 | 2899 | 0x7 | MP_FASTCLOSE | Fast Close | This | 2900 | | | | document, | 2901 | | | | Section 3.5 | 2902 | 0x8 | MP_TCPRST | Subflow Reset | This | 2903 | | | | document, | 2904 | | | | Section 3.6 | 2905 | 0xf | MP_EXPERIMENTAL | MPTCP Experimental | This | 2906 | | | Option | document, | 2907 | | | | Section 3.7 | 2908 +-------+-----------------+-------------------------+---------------+ 2910 Table 2: MPTCP Option Subtypes 2912 Values 0x9 through 0xe are currently unassigned. 2914 8.2. MPTCP Handshake Algorithms 2916 IANA has created another sub-registry, "MPTCP Handshake Algorithms" 2917 under the "Transmission Control Protocol (TCP) Parameters" registry, 2918 based on the flags in MP_CAPABLE (Section 3.1). IANA is requested to 2919 update the references of this table to this document, as follows: 2921 +---------+----------------------------------+----------------------+ 2922 | Flag | Meaning | Reference | 2923 | Bit | | | 2924 +---------+----------------------------------+----------------------+ 2925 | A | Checksum required | This document, | 2926 | | | Section 3.1 | 2927 | B | Extensibility | This document, | 2928 | | | Section 3.1 | 2929 | C | Do not attempt to connect to | This document, | 2930 | | source address | Section 3.1 | 2931 | D-G | Unassigned | | 2932 | H | HMAC-SHA256 | This document, | 2933 | | | Section 3.2 | 2934 +---------+----------------------------------+----------------------+ 2936 Table 3: MPTCP Handshake Algorithms 2938 Note that the meanings of bits D through H can be dependent upon bit 2939 B, depending on how Extensibility is defined in future 2940 specifications; see Section 3.1 for more information. 2942 Future assignments in this registry are also to be defined by 2943 Standards Action as defined by [RFC5226]. Assignments consist of the 2944 value of the flags, a symbolic name for the algorithm, and a 2945 reference to its specification. 2947 8.3. MP_TCPRST Reason Codes 2949 IANA is requested to create a further sub-registry, "MP_TCPRST Reason 2950 Codes" under the "Transmission Control Protocol (TCP) Parameters" 2951 registry, based on the reason code in MP_TCPRST (Section 3.6). The 2952 contents of this sub-registry are to to this document, as follows: 2954 +------+-----------------------------+----------------------------+ 2955 | Code | Meaning | Reference | 2956 +------+-----------------------------+----------------------------+ 2957 | 0x00 | Unspecified TCP error | This document, Section 3.6 | 2958 | 0x01 | MPTCP specific error | This document, Section 3.6 | 2959 | 0x02 | Lack of resources | This document, Section 3.6 | 2960 | 0x03 | Administratively prohibited | This document, Section 3.6 | 2961 | 0x04 | Too much outstanding data | This document, Section 3.6 | 2962 | 0x05 | Unacceptable performance | This document, Section 3.6 | 2963 | 0x06 | Middlebox interference | This document, Section 3.6 | 2964 +------+-----------------------------+----------------------------+ 2966 Table 4: MPTCP MP_TCPRST Reason Codes 2968 8.4. Experimental option registry 2970 Section 3.7 has defined the MP_EXPERIMENTAL option for private, 2971 experimental MPTCP options, and the same considerations as for 2972 [RFC6994] apply. IANA should create a "Multipath TCP Experimental 2973 Option Identifiers (MPTCP ExIDs)" sub-registry. This registry 2974 contains the 16 bits ExIDs and a reference (description, document 2975 pointer, or assignee name and e-mail contact) for each entry. MPTCP 2976 ExIDs are assigned on a First Come, First Served (FCFS) basis 2977 [RFC5226]. 2979 IANA will advise applicants of duplicate entries to select an 2980 alternate value, as per typical FCFS processing. 2982 IANA will record known duplicate uses to assist the community in both 2983 debugging assigned uses as well as correcting unauthorized duplicate 2984 uses. 2986 IANA should impose no requirement on making a registration other than 2987 indicating the desired codepoint and providing a point of contact. A 2988 short description or acronym for the use is desired but should not be 2989 required. 2991 9. References 2993 9.1. Normative References 2995 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, 2996 RFC 793, DOI 10.17487/RFC0793, September 1981, 2997 . 2999 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 3000 Requirement Levels", BCP 14, RFC 2119, 3001 DOI 10.17487/RFC2119, March 1997, 3002 . 3004 [RFC6182] Ford, A., Raiciu, C., Handley, M., Barre, S., and J. 3005 Iyengar, "Architectural Guidelines for Multipath TCP 3006 Development", RFC 6182, DOI 10.17487/RFC6182, March 2011, 3007 . 3009 [SHS] National Institute of Science and Technology, "Secure Hash 3010 Standard", Federal Information Processing Standard 3011 (FIPS) 180-4, August 2015, 3012 . 3015 9.2. Informative References 3017 [howhard] Raiciu, C., Paasch, C., Barre, S., Ford, A., Honda, M., 3018 Duchene, F., Bonaventure, O., and M. Handley, "How Hard 3019 Can It Be? Designing and Implementing a Deployable 3020 Multipath TCP", Usenix Symposium on Networked Systems 3021 Design and Implementation 2012, 2012, 3022 . 3025 [norm] Handley, M., Paxson, V., and C. Kreibich, "Network 3026 Intrusion Detection: Evasion, Traffic Normalization, and 3027 End-to-End Protocol Semantics", Usenix Security 2001, 3028 2001, 3029 . 3032 [RFC1122] Braden, R., Ed., "Requirements for Internet Hosts - 3033 Communication Layers", STD 3, RFC 1122, 3034 DOI 10.17487/RFC1122, October 1989, 3035 . 3037 [RFC1323] Jacobson, V., Braden, R., and D. Borman, "TCP Extensions 3038 for High Performance", RFC 1323, DOI 10.17487/RFC1323, May 3039 1992, . 3041 [RFC1918] Rekhter, Y., Moskowitz, B., Karrenberg, D., de Groot, G., 3042 and E. Lear, "Address Allocation for Private Internets", 3043 BCP 5, RFC 1918, DOI 10.17487/RFC1918, February 1996, 3044 . 3046 [RFC2018] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP 3047 Selective Acknowledgment Options", RFC 2018, 3048 DOI 10.17487/RFC2018, October 1996, 3049 . 3051 [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, "HMAC: Keyed- 3052 Hashing for Message Authentication", RFC 2104, 3053 DOI 10.17487/RFC2104, February 1997, 3054 . 3056 [RFC2979] Freed, N., "Behavior of and Requirements for Internet 3057 Firewalls", RFC 2979, DOI 10.17487/RFC2979, October 2000, 3058 . 3060 [RFC2992] Hopps, C., "Analysis of an Equal-Cost Multi-Path 3061 Algorithm", RFC 2992, DOI 10.17487/RFC2992, November 2000, 3062 . 3064 [RFC3022] Srisuresh, P. and K. Egevang, "Traditional IP Network 3065 Address Translator (Traditional NAT)", RFC 3022, 3066 DOI 10.17487/RFC3022, January 2001, 3067 . 3069 [RFC3135] Border, J., Kojo, M., Griner, J., Montenegro, G., and Z. 3070 Shelby, "Performance Enhancing Proxies Intended to 3071 Mitigate Link-Related Degradations", RFC 3135, 3072 DOI 10.17487/RFC3135, June 2001, 3073 . 3075 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 3076 of Explicit Congestion Notification (ECN) to IP", 3077 RFC 3168, DOI 10.17487/RFC3168, September 2001, 3078 . 3080 [RFC4086] Eastlake 3rd, D., Schiller, J., and S. Crocker, 3081 "Randomness Requirements for Security", BCP 106, RFC 4086, 3082 DOI 10.17487/RFC4086, June 2005, 3083 . 3085 [RFC4987] Eddy, W., "TCP SYN Flooding Attacks and Common 3086 Mitigations", RFC 4987, DOI 10.17487/RFC4987, August 2007, 3087 . 3089 [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an 3090 IANA Considerations Section in RFCs", RFC 5226, 3091 DOI 10.17487/RFC5226, May 2008, 3092 . 3094 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 3095 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 3096 . 3098 [RFC5961] Ramaiah, A., Stewart, R., and M. Dalal, "Improving TCP's 3099 Robustness to Blind In-Window Attacks", RFC 5961, 3100 DOI 10.17487/RFC5961, August 2010, 3101 . 3103 [RFC6181] Bagnulo, M., "Threat Analysis for TCP Extensions for 3104 Multipath Operation with Multiple Addresses", RFC 6181, 3105 DOI 10.17487/RFC6181, March 2011, 3106 . 3108 [RFC6234] Eastlake 3rd, D. and T. Hansen, "US Secure Hash Algorithms 3109 (SHA and SHA-based HMAC and HKDF)", RFC 6234, 3110 DOI 10.17487/RFC6234, May 2011, 3111 . 3113 [RFC6356] Raiciu, C., Handley, M., and D. Wischik, "Coupled 3114 Congestion Control for Multipath Transport Protocols", 3115 RFC 6356, DOI 10.17487/RFC6356, October 2011, 3116 . 3118 [RFC6528] Gont, F. and S. Bellovin, "Defending against Sequence 3119 Number Attacks", RFC 6528, DOI 10.17487/RFC6528, February 3120 2012, . 3122 [RFC6824] Ford, A., Raiciu, C., Handley, M., and O. Bonaventure, 3123 "TCP Extensions for Multipath Operation with Multiple 3124 Addresses", RFC 6824, DOI 10.17487/RFC6824, January 2013, 3125 . 3127 [RFC6897] Scharf, M. and A. Ford, "Multipath TCP (MPTCP) Application 3128 Interface Considerations", RFC 6897, DOI 10.17487/RFC6897, 3129 March 2013, . 3131 [RFC6994] Touch, J., "Shared Use of Experimental TCP Options", 3132 RFC 6994, DOI 10.17487/RFC6994, August 2013, 3133 . 3135 [TCPLO] Ramaiah, A., "TCP option space extension", Work 3136 in Progress, March 2012. 3138 Appendix A. Notes on Use of TCP Options 3140 The TCP option space is limited due to the length of the Data Offset 3141 field in the TCP header (4 bits), which defines the TCP header length 3142 in 32-bit words. With the standard TCP header being 20 bytes, this 3143 leaves a maximum of 40 bytes for options, and many of these may 3144 already be used by options such as timestamp and SACK. 3146 We have performed a brief study on the commonly used TCP options in 3147 SYN, data, and pure ACK packets, and found that there is enough room 3148 to fit all the options we propose using in this document. 3150 SYN packets typically include Maximum Segment Size (MSS) (4 bytes), 3151 window scale (3 bytes), SACK permitted (2 bytes), and timestamp (10 3152 bytes) options. Together these sum to 19 bytes. Some operating 3153 systems appear to pad each option up to a word boundary, thus using 3154 24 bytes (a brief survey suggests Windows XP and Mac OS X do this, 3155 whereas Linux does not). Optimistically, therefore, we have 21 bytes 3156 spare, or 16 if it has to be word-aligned. In either case, however, 3157 the SYN versions of Multipath Capable (12 bytes) and Join (12 or 16 3158 bytes) options will fit in this remaining space. 3160 Note that due to the use of a 64-bit data-level sequence space, it is 3161 feasible that MPTCP will not require the timestamp option for 3162 protection against wrapped sequence numbers (PAWS [RFC1323]), since 3163 the data-level sequence space has far less chance of wrapping. 3164 Confirmation of the validity of this optimisation is for further 3165 study. 3167 TCP data packets typically carry timestamp options in every packet, 3168 taking 10 bytes (or 12 with padding). That leaves 30 bytes (or 28, 3169 if word-aligned). The Data Sequence Signal (DSS) option varies in 3170 length depending on whether the data sequence mapping and DATA_ACK 3171 are included, and whether the sequence numbers in use are 4 or 8 3172 octets. The maximum size of the DSS option is 28 bytes, so even that 3173 will fit in the available space. But unless a connection is both 3174 bidirectional and high-bandwidth, it is unlikely that all that option 3175 space will be required on each DSS option. 3177 Within the DSS option, it is not necessary to include the data 3178 sequence mapping and DATA_ACK in each packet, and in many cases it 3179 may be possible to alternate their presence (so long as the mapping 3180 covers the data being sent in the following packet). It would also 3181 be possible to alternate between 4- and 8-byte sequence numbers in 3182 each option. 3184 On subflow and connection setup, an MPTCP option is also set on the 3185 third packet (an ACK). These are 20 bytes (for Multipath Capable) 3186 and 24 bytes (for Join), both of which will fit in the available 3187 option space. 3189 Pure ACKs in TCP typically contain only timestamps (10 bytes). Here, 3190 Multipath TCP typically needs to encode only the DATA_ACK (maximum of 3191 12 bytes). Occasionally, ACKs will contain SACK information. 3192 Depending on the number of lost packets, SACK may utilize the entire 3193 option space. If a DATA_ACK had to be included, then it is probably 3194 necessary to reduce the number of SACK blocks to accommodate the 3195 DATA_ACK. However, the presence of the DATA_ACK is unlikely to be 3196 necessary in a case where SACK is in use, since until at least some 3197 of the SACK blocks have been retransmitted, the cumulative data-level 3198 ACK will not be moving forward (or if it does, due to retransmissions 3199 on another path, then that path can also be used to transmit the new 3200 DATA_ACK). 3202 The ADD_ADDR option can be between 16 and 30 bytes, depending on 3203 whether IPv4 or IPv6 is used, and whether or not the port number is 3204 present. It is unlikely that such signaling would fit in a data 3205 packet (although if there is space, it is fine to include it). It is 3206 recommended to use duplicate ACKs with no other payload or options in 3207 order to transmit these rare signals. Note this is the reason for 3208 mandating that duplicate ACKs with MPTCP options are not taken as a 3209 signal of congestion. 3211 Finally, there are issues with reliable delivery of options. As 3212 options can also be sent on pure ACKs, these are not reliably sent. 3213 This is not an issue for DATA_ACK due to their cumulative nature, but 3214 may be an issue for ADD_ADDR/REMOVE_ADDR options. Here, it is 3215 recommended to send these options redundantly (whether on multiple 3216 paths or on the same path on a number of ACKs -- but interspersed 3217 with data in order to avoid interpretation as congestion). The cases 3218 where options are stripped by middleboxes are discussed in Section 6. 3220 Appendix B. Control Blocks 3222 Conceptually, an MPTCP connection can be represented as an MPTCP 3223 control block that contains several variables that track the progress 3224 and the state of the MPTCP connection and a set of linked TCP control 3225 blocks that correspond to the subflows that have been established. 3227 RFC 793 [RFC0793] specifies several state variables. Whenever 3228 possible, we reuse the same terminology as RFC 793 to describe the 3229 state variables that are maintained by MPTCP. 3231 B.1. MPTCP Control Block 3233 The MPTCP control block contains the following variable per 3234 connection. 3236 B.1.1. Authentication and Metadata 3238 Local.Token (32 bits): This is the token chosen by the local host on 3239 this MPTCP connection. The token MUST be unique among all 3240 established MPTCP connections, generated from the local key. 3242 Local.Key (64 bits): This is the key sent by the local host on this 3243 MPTCP connection. 3245 Remote.Token (32 bits): This is the token chosen by the remote host 3246 on this MPTCP connection, generated from the remote key. 3248 Remote.Key (64 bits): This is the key chosen by the remote host on 3249 this MPTCP connection 3251 MPTCP.Checksum (flag): This flag is set to true if at least one of 3252 the hosts has set the A bit in the MP_CAPABLE options exchanged 3253 during connection establishment, and is set to false otherwise. 3254 If this flag is set, the checksum must be computed in all DSS 3255 options. 3257 B.1.2. Sending Side 3259 SND.UNA (64 bits): This is the data sequence number of the next byte 3260 to be acknowledged, at the MPTCP connection level. This variable 3261 is updated upon reception of a DSS option containing a DATA_ACK. 3263 SND.NXT (64 bits): This is the data sequence number of the next byte 3264 to be sent. SND.NXT is used to determine the value of the DSN in 3265 the DSS option. 3267 SND.WND (32 bits with RFC 1323, 16 bits otherwise): This is the 3268 sending window. MPTCP maintains the sending window at the MPTCP 3269 connection level and the same window is shared by all subflows. 3270 All subflows use the MPTCP connection level SND.WND to compute the 3271 SEQ.WND value that is sent in each transmitted segment. 3273 B.1.3. Receiving Side 3275 RCV.NXT (64 bits): This is the data sequence number of the next byte 3276 that is expected on the MPTCP connection. This state variable is 3277 modified upon reception of in-order data. The value of RCV.NXT is 3278 used to specify the DATA_ACK that is sent in the DSS option on all 3279 subflows. 3281 RCV.WND (32 bits with RFC 1323, 16 bits otherwise): This is the 3282 connection-level receive window, which is the maximum of the 3283 RCV.WND on all the subflows. 3285 B.2. TCP Control Blocks 3287 The MPTCP control block also contains a list of the TCP control 3288 blocks that are associated to the MPTCP connection. 3290 Note that the TCP control block on the TCP subflows does not contain 3291 the RCV.WND and SND.WND state variables as these are maintained at 3292 the MPTCP connection level and not at the subflow level. 3294 Inside each TCP control block, the following state variables are 3295 defined. 3297 B.2.1. Sending Side 3299 SND.UNA (32 bits): This is the sequence number of the next byte to 3300 be acknowledged on the subflow. This variable is updated upon 3301 reception of each TCP acknowledgment on the subflow. 3303 SND.NXT (32 bits): This is the sequence number of the next byte to 3304 be sent on the subflow. SND.NXT is used to set the value of 3305 SEG.SEQ upon transmission of the next segment. 3307 B.2.2. Receiving Side 3309 RCV.NXT (32 bits): This is the sequence number of the next byte that 3310 is expected on the subflow. This state variable is modified upon 3311 reception of in-order segments. The value of RCV.NXT is copied to 3312 the SEG.ACK field of the next segments transmitted on the subflow. 3314 RCV.WND (32 bits with RFC 1323, 16 bits otherwise): This is the 3315 subflow-level receive window that is updated with the window field 3316 from the segments received on this subflow. 3318 Appendix C. Finite State Machine 3320 The diagram in Figure 19 shows the Finite State Machine for 3321 connection-level closure. This illustrates how the DATA_FIN 3322 connection-level signal (indicated as the DFIN flag on a DATA_ACK) 3323 interacts with subflow-level FINs, and permits "break-before-make" 3324 handover between subflows. 3326 +---------+ 3327 | M_ESTAB | 3328 +---------+ 3329 M_CLOSE | | rcv DATA_FIN 3330 ------- | | ------- 3331 +---------+ snd DATA_FIN / \ snd DATA_ACK[DFIN] +---------+ 3332 | M_FIN |<----------------- ------------------->| M_CLOSE | 3333 | WAIT-1 |--------------------------- | WAIT | 3334 +---------+ rcv DATA_FIN \ +---------+ 3335 | rcv DATA_ACK[DFIN] ------- | M_CLOSE | 3336 | -------------- snd DATA_ACK | ------- | 3337 | CLOSE all subflows | snd DATA_FIN | 3338 V V V 3339 +-----------+ +-----------+ +-----------+ 3340 |M_FINWAIT-2| | M_CLOSING | | M_LAST-ACK| 3341 +-----------+ +-----------+ +-----------+ 3342 | rcv DATA_ACK[DFIN] | rcv DATA_ACK[DFIN] | 3343 | rcv DATA_FIN -------------- | -------------- | 3344 | ------- CLOSE all subflows | CLOSE all subflows | 3345 | snd DATA_ACK[DFIN] V delete MPTCP PCB V 3346 \ +-----------+ +---------+ 3347 ------------------------>|M_TIME WAIT|----------------->| M_CLOSED| 3348 +-----------+ +---------+ 3349 All subflows in CLOSED 3350 ------------ 3351 delete MPTCP PCB 3353 Figure 19: Finite State Machine for Connection Closure 3355 Authors' Addresses 3357 Alan Ford 3358 Pexip 3360 EMail: alan.ford@gmail.com 3362 Costin Raiciu 3363 University Politehnica of Bucharest 3364 Splaiul Independentei 313 3365 Bucharest 3366 Romania 3368 EMail: costin.raiciu@cs.pub.ro 3369 Mark Handley 3370 University College London 3371 Gower Street 3372 London WC1E 6BT 3373 UK 3375 EMail: m.handley@cs.ucl.ac.uk 3377 Olivier Bonaventure 3378 Universite catholique de Louvain 3379 Pl. Ste Barbe, 2 3380 Louvain-la-Neuve 1348 3381 Belgium 3383 EMail: olivier.bonaventure@uclouvain.be 3385 Christoph Paasch 3386 Apple, Inc. 3387 Cupertino 3388 US 3390 EMail: cpaasch@apple.com