idnits 2.17.1 draft-ietf-tsvwg-ecn-tunnel-08.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** You're using the IETF Trust Provisions' Section 6.b License Notice from 12 Sep 2009 rather than the newer Notice from 28 Dec 2009. (See https://trustee.ietf.org/license-info/) Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- -- The draft header indicates that this document updates RFC4301, but the abstract doesn't seem to directly say this. It does mention RFC4301 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 (Using the creation date from RFC3168, updated by this document, for RFC5378 checks: 2000-11-17) -- The document seems to contain a disclaimer for pre-RFC5378 work, and may have content which was first submitted before 10 November 2008. The disclaimer is necessary when there are original authors that you have been unable to contact, or if some do not wish to grant the BCP78 rights to the IETF Trust. If you are able to get all authors (current and original) to grant those rights, you can and should remove the disclaimer; otherwise, the disclaimer is needed and you can ignore this comment. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (March 03, 2010) is 5166 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-11) exists of draft-ietf-pcn-3-in-1-encoding-01 == Outdated reference: A later version (-02) exists of draft-ietf-pcn-3-state-encoding-01 == Outdated reference: A later version (-02) exists of draft-ietf-pcn-psdm-encoding-00 == Outdated reference: A later version (-12) exists of draft-ietf-pcn-sm-edge-behaviour-01 -- Obsolete informational reference (is this intentional?): RFC 2401 (Obsoleted by RFC 4301) -- Obsolete informational reference (is this intentional?): RFC 2481 (Obsoleted by RFC 3168) -- Obsolete informational reference (is this intentional?): RFC 4306 (Obsoleted by RFC 5996) -- Obsolete informational reference (is this intentional?): RFC 5696 (Obsoleted by RFC 6660) Summary: 1 error (**), 0 flaws (~~), 5 warnings (==), 7 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Transport Area Working Group B. Briscoe 3 Internet-Draft BT 4 Updates: 3168, 4301 March 03, 2010 5 (if approved) 6 Intended status: Standards Track 7 Expires: September 4, 2010 9 Tunnelling of Explicit Congestion Notification 10 draft-ietf-tsvwg-ecn-tunnel-08 12 Abstract 14 This document redefines how the explicit congestion notification 15 (ECN) field of the IP header should be constructed on entry to and 16 exit from any IP in IP tunnel. On encapsulation it updates RFC3168 17 to bring all IP in IP tunnels (v4 or v6) into line with RFC4301 IPsec 18 ECN processing. On decapsulation it updates both RFC3168 and RFC4301 19 to add new behaviours for previously unused combinations of inner and 20 outer header. The new rules ensure the ECN field is correctly 21 propagated across a tunnel whether it is used to signal one or two 22 severity levels of congestion, whereas before only one severity level 23 was supported. Tunnel endpoints can be updated in any order without 24 affecting pre-existing uses of the ECN field, providing backward 25 compatibility. Nonetheless, operators wanting to support two 26 severity levels (e.g. for pre-congestion notification--PCN) can 27 require compliance with this new specification. A thorough analysis 28 of the reasoning for these changes and the implications is included. 29 In the unlikely event that the new rules do not meet a specific need, 30 RFC4774 gives guidance on designing alternate ECN semantics and this 31 document extends that to include tunnelling issues. 33 Status of This Memo 35 This Internet-Draft is submitted to IETF in full conformance with the 36 provisions of BCP 78 and BCP 79. 38 Internet-Drafts are working documents of the Internet Engineering 39 Task Force (IETF), its areas, and its working groups. Note that 40 other groups may also distribute working documents as Internet- 41 Drafts. 43 Internet-Drafts are draft documents valid for a maximum of six months 44 and may be updated, replaced, or obsoleted by other documents at any 45 time. It is inappropriate to use Internet-Drafts as reference 46 material or to cite them other than as "work in progress." 48 The list of current Internet-Drafts can be accessed at 49 http://www.ietf.org/ietf/1id-abstracts.txt. 51 The list of Internet-Draft Shadow Directories can be accessed at 52 http://www.ietf.org/shadow.html. 54 This Internet-Draft will expire on September 4, 2010. 56 Copyright Notice 58 Copyright (c) 2010 IETF Trust and the persons identified as the 59 document authors. All rights reserved. 61 This document is subject to BCP 78 and the IETF Trust's Legal 62 Provisions Relating to IETF Documents 63 (http://trustee.ietf.org/license-info) in effect on the date of 64 publication of this document. Please review these documents 65 carefully, as they describe your rights and restrictions with respect 66 to this document. Code Components extracted from this document must 67 include Simplified BSD License text as described in Section 4.e of 68 the Trust Legal Provisions and are provided without warranty as 69 described in the BSD License. 71 This document may contain material from IETF Documents or IETF 72 Contributions published or made publicly available before November 73 10, 2008. The person(s) controlling the copyright in some of this 74 material may not have granted the IETF Trust the right to allow 75 modifications of such material outside the IETF Standards Process. 76 Without obtaining an adequate license from the person(s) controlling 77 the copyright in such materials, this document may not be modified 78 outside the IETF Standards Process, and derivative works of it may 79 not be created outside the IETF Standards Process, except to format 80 it for publication as an RFC or to translate it into languages other 81 than English. 83 Table of Contents 85 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 9 86 1.1. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . 11 87 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 11 88 3. Summary of Pre-Existing RFCs . . . . . . . . . . . . . . . . . 12 89 3.1. Encapsulation at Tunnel Ingress . . . . . . . . . . . . . 12 90 3.2. Decapsulation at Tunnel Egress . . . . . . . . . . . . . . 13 91 4. New ECN Tunnelling Rules . . . . . . . . . . . . . . . . . . . 14 92 4.1. Default Tunnel Ingress Behaviour . . . . . . . . . . . . . 15 93 4.2. Default Tunnel Egress Behaviour . . . . . . . . . . . . . 15 94 4.3. Encapsulation Modes . . . . . . . . . . . . . . . . . . . 17 95 4.4. Single Mode of Decapsulation . . . . . . . . . . . . . . . 19 96 5. Updates to Earlier RFCs . . . . . . . . . . . . . . . . . . . 20 97 5.1. Changes to RFC4301 ECN processing . . . . . . . . . . . . 20 98 5.2. Changes to RFC3168 ECN processing . . . . . . . . . . . . 20 99 5.3. Motivation for Changes . . . . . . . . . . . . . . . . . . 22 100 5.3.1. Motivation for Changing Encapsulation . . . . . . . . 22 101 5.3.2. Motivation for Changing Decapsulation . . . . . . . . 23 102 6. Backward Compatibility . . . . . . . . . . . . . . . . . . . . 25 103 6.1. Non-Issues Updating Decapsulation . . . . . . . . . . . . 25 104 6.2. Non-Update of RFC4301 IPsec Encapsulation . . . . . . . . 26 105 6.3. Update to RFC3168 Encapsulation . . . . . . . . . . . . . 26 106 7. Design Principles for Alternate ECN Tunnelling Semantics . . . 27 107 8. Security Considerations . . . . . . . . . . . . . . . . . . . 29 108 9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 30 109 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 31 110 11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 31 111 11.1. Normative References . . . . . . . . . . . . . . . . . . . 31 112 11.2. Informative References . . . . . . . . . . . . . . . . . . 32 113 Editorial Comments . . . . . . . . . . . . . . . . . . . . . . . . 114 Appendix A. Early ECN Tunnelling RFCs . . . . . . . . . . . . . . 34 115 Appendix B. Design Constraints . . . . . . . . . . . . . . . . . 35 116 B.1. Security Constraints . . . . . . . . . . . . . . . . . . . 35 117 B.2. Control Constraints . . . . . . . . . . . . . . . . . . . 37 118 B.3. Management Constraints . . . . . . . . . . . . . . . . . . 38 119 Appendix C. Contribution to Congestion across a Tunnel . . . . . 39 120 Appendix D. Why Losing ECT(1) on Decapsulation Impedes PCN 121 (to be removed before publication) . . . . . . . . . 40 122 Appendix E. Why Resetting ECN on Encapsulation Impedes PCN 123 (to be removed before publication) . . . . . . . . . 41 124 Appendix F. Compromise on Decap with ECT(1) Inner and ECT(0) 125 Outer . . . . . . . . . . . . . . . . . . . . . . . . 42 126 Appendix G. Open Issues . . . . . . . . . . . . . . . . . . . . . 43 128 Request to the RFC Editor (to be removed on publication): 130 In the RFC index, RFC3168 should be identified as an update to 131 RFC2003. RFC4301 should be identified as an update to RFC3168. 133 Changes from previous drafts (to be removed by the RFC Editor) 135 Full text differences between IETF draft versions are available at 136 , and 137 between earlier individual draft versions at 138 140 From ietf-06 to ietf-07 (current): 142 * Emphasised that this is the opposite of a fork in the RFC 143 series. 145 * Altered Section 5 to focus on updates to implementations of 146 earlier RFCs, rather than on updates to the text of the RFCs. 148 * Removed potential loop-holes in normative text that 149 implementers might have used to claim compliance without 150 implementing normal mode. Highlighted the deliberate 151 distinction between "MUST implement" and "SHOULD use" normal 152 mode. 154 * Added question for Security Directorate reviewers on whether to 155 mention a corner-case concerning manual keying of IPsec 156 tunnels. 158 * Minor clarifications, updated references and updated acks. 160 * Marked two appendices about PCN motivations for removal before 161 publication. 163 From ietf-05 to ietf-06: 165 * Minor textual clarifications and corrections. 167 From ietf-04 to ietf-05: 169 * Functional changes: 171 + Section 4.2: ECT(1) outer with Not-ECT inner: reverted to 172 forwarding as Not-ECT (as in RFC3168 & RFC4301), rather than 173 dropping. 175 + Altered rationale in bullet 3 of Section 5.3.2 to justify 176 this. 178 + Distinguished alarms for dangerous and invalid combinations 179 and allowed combinations that are valid in some tunnel 180 configurations but dangerous in others to be alarmed at the 181 discretion of the implementer and/or operator. 183 + Altered advice on designing alternate ECN tunnelling 184 semantics to reflect the above changes. 186 * Textual changes: 188 + Changed "Future non-default schemes" to "Alternate ECN 189 Tunnelling Semantics" throughout. 191 + Cut down Appendix D and Appendix E for brevity. 193 + A number of clarifying edits & updated refs. 195 From ietf-03 to ietf-04: 197 * Functional changes: none 199 * Structural changes: 201 + Added "Open Issues" appendix 203 * Textual changes: 205 + Section title: "Changes from Earlier RFCs" -> "Updates to 206 Earlier RFCs" 208 + Emphasised that change on decap to previously unused 209 combinations will propagate PCN encoding. 211 + Acknowledged additional reviewers and updated references 213 From ietf-02 to ietf-03: 215 * Functional changes: 217 + Corrected errors in recap of previous RFCs, which wrongly 218 stated the different decapsulation behaviours of RFC3168 & 219 RFC4301 with a Not-ECT inner header. This also required 220 corrections to the "Changes from Earlier RFCs" and the 221 Motivations for these changes. 223 + Mandated that any future standards action SHOULD NOT use the 224 ECT(0) codepoint as an indication of congestion, without 225 giving strong reasons. 227 + Added optional alarm when decapsulating ECT(1) outer, 228 ECT(0), but noted it would need to be disabled for 229 2-severity level congestion (e.g. PCN). 231 * Structural changes: 233 + Removed Document Roadmap which merely repeated the Contents 234 (previously Section 1.2). 236 + Moved "Changes from Earlier RFCs" (Section 5) before 237 Section 6 on Backward Compatibility and internally organised 238 both by RFC, rather than by ingress then egress. 240 + Moved motivation for changing existing RFCs (Section 5.3) to 241 after the changes are specified. 243 + Moved informative "Design Principles for Future Non-Default 244 Schemes" after all the normative sections. 246 + Added Appendix A on early history of ECN tunnelling RFCs. 248 + Removed specialist appendix on "Relative Placement of 249 Tunnelling and In-Path Load Regulation" (Appendix D in the 250 -02 draft) 252 + Moved and updated specialist text on "Compromise on Decap 253 with ECT(1) Inner and ECT(0) Outer" from Security 254 Considerations to Appendix F 256 * Textual changes: 258 + Simplified vocabulary for non-native-english speakers 260 + Simplified Introduction and defined regularly used terms in 261 an expanded Terminology section. 263 + More clearly distinguished statically configured tunnels 264 from dynamic tunnel endpoint discovery, before explaining 265 operating modes. 267 + Simplified, cut-down and clarified throughout 269 + Updated references. 271 From ietf-01 to ietf-02: 273 * Scope reduced from any encapsulation of an IP packet to solely 274 IP in IP tunnelled encapsulation. Consequently changed title 275 and removed whole section 'Design Guidelines for New 276 Encapsulations of Congestion Notification' (to be included in a 277 future companion informational document). 279 * Included a new normative decapsulation rule for ECT(0) inner 280 and ECT(1) outer that had previously only been outlined in the 281 non-normative appendix 'Comprehensive Decapsulation Rules'. 282 Consequently: 284 + The Introduction has been completely re-written to motivate 285 this change to decapsulation along with the existing change 286 to encapsulation. 288 + The tentative text in the appendix that first proposed this 289 change has been split between normative standards text in 290 Section 4 and Appendix D, which explains specifically why 291 this change would streamline PCN. New text on the logic of 292 the resulting decap rules added. 294 * If inner/outer is Not-ECT/ECT(0), changed decapsulation to 295 propagate Not-ECT rather than drop the packet; and added 296 reasoning. 298 * Considerably restructured: 300 + "Design Constraints" analysis moved to an appendix 301 (Appendix B); 303 + Added Section 3 to summarise relevant existing RFCs; 305 + Structured Section 4 and Section 6 into subsections. 307 + Added tables to sections on old and new rules, for precision 308 and comparison. 310 + Moved Section 7 on Design Principles to the end of the 311 section specifying the new default normative tunnelling 312 behaviour. Rewritten and shifted text on identifiers and 313 in-path load regulators to Appendix B.1 [deleted in revision 314 -03]. 316 From ietf-00 to ietf-01: 318 * Identified two additional alarm states in the decapsulation 319 rules (Figure 4) if ECT(X) in outer and inner contradict each 320 other. 322 * Altered Comprehensive Decapsulation Rules (Appendix D) so that 323 ECT(0) in the outer no longer overrides ECT(1) in the inner. 324 Used the term 'Comprehensive' instead of 'Ideal'. And 325 considerably updated the text in this appendix. 327 * Added Appendix D.1 (removed again in a later revision) to weigh 328 up the various ways the Comprehensive Decapsulation Rules might 329 be introduced. This replaces the previous contradictory 330 statements saying complex backwards compatibility interactions 331 would be introduced while also saying there would be no 332 backwards compatibility issues. 334 * Updated references. 336 From briscoe-01 to ietf-00: 338 * Re-wrote Appendix C giving much simpler technique to measure 339 contribution to congestion across a tunnel. 341 * Added discussion of backward compatibility of the ideal 342 decapsulation scheme in Appendix D 344 * Updated references. Minor corrections & clarifications 345 throughout. 347 From briscoe-00 to briscoe-01: 349 * Related everything conceptually to the uniform and pipe models 350 of RFC2983 on Diffserv Tunnels, and completely removed the 351 dependence of tunnelling behaviour on the presence of any in- 352 path load regulation by using the [1 - Before] [2 - Outer] 353 function placement concepts from RFC2983; 355 * Added specific cases where the existing standards limit new 356 proposals, particularly Appendix E; 358 * Added sub-structure to Introduction (Need for Rationalisation, 359 Roadmap), added new Introductory subsection on "Scope" and 360 improved clarity; 362 * Added Design Guidelines for New Encapsulations of Congestion 363 Notification; 365 * Considerably clarified the Backward Compatibility section 366 (Section 6); 368 * Considerably extended the Security Considerations section 369 (Section 8); 371 * Summarised the primary rationale much better in the 372 conclusions; 374 * Added numerous extra acknowledgements; 376 * Added Appendix E. "Why resetting CE on encapsulation harms 377 PCN", Appendix C. "Contribution to Congestion across a Tunnel" 378 and Appendix D. "Ideal Decapsulation Rules"; 380 * Re-wrote Appendix B [deleted in a later revision], explaining 381 how tunnel encapsulation no longer depends on in-path load- 382 regulation (changed title from "In-path Load Regulation" to 383 "Non-Dependence of Tunnelling on In-path Load Regulation"), but 384 explained how an in-path load regulation function must be 385 carefully placed with respect to tunnel encapsulation (in a new 386 sub-section entitled "Dependence of In-Path Load Regulation on 387 Tunnelling"). 389 1. Introduction 391 Explicit congestion notification (ECN [RFC3168]) allows a forwarding 392 element (e.g. a router) to notify the onset of congestion without 393 having to drop packets. Instead it can explicitly mark a proportion 394 of packets in the 2-bit ECN field in the IP header (Table 1 recaps 395 the ECN codepoints). 397 The outer header of an IP packet can encapsulate one or more IP 398 headers for tunnelling. A forwarding element using ECN to signify 399 congestion will only mark the immediately visible outer IP header. 400 When a tunnel decapsulator later removes this outer header, it 401 follows rules to propagate congestion markings by combining the ECN 402 fields of the inner and outer IP header into one outgoing IP header. 404 This document updates those rules for IPsec [RFC4301] and non-IPsec 405 [RFC3168] tunnels to add new behaviours for previously unused 406 combinations of inner and outer header. It also updates the tunnel 407 ingress behaviour of RFC3168 to match that of RFC4301. The updated 408 rules are backward compatible with RFC4301 and RFC3168 when 409 interworking with any other tunnel endpoint complying with any 410 earlier specification. 412 When ECN and its tunnelling was defined in RFC3168, only the minimum 413 necessary changes to the ECN field were propagated through tunnel 414 endpoints--just enough for the basic ECN mechanism to work. This was 415 due to concerns that the ECN field might be toggled to communicate 416 between a secure site and someone on the public Internet--a covert 417 channel. This was because a mutable field like ECN cannot be 418 protected by IPsec's integrity mechanisms--it has to be able to 419 change as it traverses the Internet. 421 Nonetheless, the latest IPsec architecture [RFC4301] considered a 422 bandwidth limit of 2 bits per packet on a covert channel made it a 423 manageable risk. Therefore, for simplicity, an RFC4301 ingress 424 copied the whole ECN field to encapsulate a packet. It dispensed 425 with the two modes of RFC3168, one which partially copied the ECN 426 field, and the other which blocked all propagation of ECN changes. 428 Unfortunately, this entirely reasonable sequence of standards actions 429 resulted in a perverse outcome; non-IPsec tunnels (RFC3168) blocked 430 the 2-bit covert channel, while IPsec tunnels (RFC4301) did not--at 431 least not at the ingress. At the egress, both IPsec and non-IPsec 432 tunnels still partially restricted propagation of the full ECN field. 434 The trigger for the changes in this document was the introduction of 435 pre-congestion notification (PCN [RFC5670]) to the IETF standards 436 track. PCN needs the ECN field to be copied at a tunnel ingress and 437 it needs four states of congestion signalling to be propagated at the 438 egress, but pre-existing tunnels only propagate three in the ECN 439 field. 441 This document draws on currently unused (CU) combinations of inner 442 and outer headers to add tunnelling of four-state congestion 443 signalling to RFC3168 and RFC4301. Operators of tunnels who 444 specifically want to support four states can require that all their 445 tunnels comply with this specification. However, this is not a fork 446 in the RFC series. It is an update that can be deployed first by 447 those that need it, and subsequently by all tunnel endpoint 448 implementations (RFC4301, RFC3168, RFC2481, RFC2401, RFC2003), which 449 can safely be updated to this new specification as part of general 450 code maintenance. This will gradually add support for four 451 congestion states to the Internet. Existing three state schemes will 452 continue to work as before. 454 In fact, this document is the opposite of a fork. At the same time 455 as supporting a fourth state, the opportunity has been taken to draw 456 together divergent ECN tunnelling specifications into a single 457 consistent behaviour, harmonising differences such as perverse covert 458 channel treatment. Then any tunnel can be deployed unilaterally, and 459 it will support the full range of congestion control and management 460 schemes without any modes or configuration. Further, any host or 461 router can expect the ECN field to behave in the same way, whatever 462 type of tunnel might intervene in the path. 464 1.1. Scope 466 This document only concerns wire protocol processing of the ECN field 467 at tunnel endpoints and makes no changes or recommendations 468 concerning algorithms for congestion marking or congestion response. 470 This document specifies common ECN field processing at encapsulation 471 and decapsulation for any IP in IP tunnelling, whether IPsec or non- 472 IPsec tunnels. It applies irrespective of whether IPv4 or IPv6 is 473 used for either of the inner and outer headers. It applies for 474 packets with any destination address type, whether unicast or 475 multicast. It applies as the default for all Diffserv per-hop 476 behaviours (PHBs), unless stated otherwise in the specification of a 477 PHB (but Section 4 strongly deprecates such exceptions). It is 478 intended to be a good trade off between somewhat conflicting 479 security, control and management requirements. 481 [RFC2983] is a comprehensive primer on differentiated services and 482 tunnels. Given ECN raises similar issues to differentiated services 483 when interacting with tunnels, useful concepts introduced in RFC2983 484 are used throughout, with brief recaps of the explanations where 485 necessary. 487 2. Terminology 489 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 490 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 491 document are to be interpreted as described in RFC 2119 [RFC2119]. 493 Table 1 recaps the names of the ECN codepoints [RFC3168]. 495 +------------------+----------------+---------------------------+ 496 | Binary codepoint | Codepoint name | Meaning | 497 +------------------+----------------+---------------------------+ 498 | 00 | Not-ECT | Not ECN-capable transport | 499 | 01 | ECT(1) | ECN-capable transport | 500 | 10 | ECT(0) | ECN-capable transport | 501 | 11 | CE | Congestion experienced | 502 +------------------+----------------+---------------------------+ 504 Table 1: Recap of Codepoints of the ECN Field [RFC3168] in the IP 505 Header 507 Further terminology used within this document: 509 Encapsulator: The tunnel endpoint function that adds an outer IP 510 header to tunnel a packet (also termed the 'ingress tunnel 511 endpoint' or just the 'ingress' where the context is clear). 513 Decapsulator: The tunnel endpoint function that removes an outer IP 514 header from a tunnelled packet (also termed the 'egress tunnel 515 endpoint' or just the 'egress' where the context is clear). 517 Incoming header: The header of an arriving packet before 518 encapsulation. 520 Outer header: The header added to encapsulate a tunnelled packet. 522 Inner header: The header encapsulated by the outer header. 524 Outgoing header: The header constructed by the decapsulator using 525 logic that combines the fields in the outer and inner headers. 527 Copying ECN: On encapsulation, setting the ECN field of the new 528 outer header to be a copy of the ECN field in the incoming header. 530 Zeroing ECN: On encapsulation, clearing the ECN field of the new 531 outer header to Not-ECT ("00"). 533 Resetting ECN: On encapsulation, setting the ECN field of the new 534 outer header to be a copy of the ECN field in the incoming header 535 except the outer ECN field is set to the ECT(0) codepoint if the 536 incoming ECN field is CE. 538 3. Summary of Pre-Existing RFCs 540 This section is informative not normative, as it recaps pre-existing 541 RFCs. Earlier relevant RFCs that were either experimental or 542 incomplete with respect to ECN tunnelling (RFC2481, RFC2401 and 543 RFC2003) are briefly outlined in Appendix A. The question of whether 544 tunnel implementations used in the Internet comply with any of these 545 RFCs is not discussed. 547 3.1. Encapsulation at Tunnel Ingress 549 At the encapsulator, the controversy has been over whether to 550 propagate information about congestion experienced on the path so far 551 into the outer header of the tunnel. 553 Specifically, RFC3168 says that, if a tunnel fully supports ECN 554 (termed a 'full-functionality' ECN tunnel in [RFC3168]), the 555 encapsulator must not copy a CE marking from the inner header into 556 the outer header that it creates. Instead the encapsulator must set 557 the outer header to ECT(0) if the ECN field is marked CE in the 558 arriving IP header. We term this 'resetting' a CE codepoint. 560 However, the new IPsec architecture in [RFC4301] reverses this rule, 561 stating that the encapsulator must simply copy the ECN field from the 562 incoming header to the outer header. 564 RFC3168 also provided a Limited Functionality mode that turns off ECN 565 processing over the scope of the tunnel by setting the outer header 566 to Not-ECT ("00"). Then such packets will be dropped to indicate 567 congestion rather than marked with ECN. This is necessary for the 568 ingress to interwork with legacy decapsulators ([RFC2481], [RFC2401] 569 and [RFC2003]) that do not propagate ECN markings added to the outer 570 header. Otherwise such legacy decapsulators would throw away 571 congestion notifications before they reached the transport layer. 573 Neither Limited Functionality mode nor Full Functionality mode are 574 used by an RFC4301 IPsec encapsulator, which simply copies the 575 incoming ECN field into the outer header. An earlier key-exchange 576 phase ensures an RFC4301 ingress will not have to interwork with a 577 legacy egress that does not support ECN. 579 These pre-existing behaviours are summarised in Figure 1. 580 +-----------------+-----------------------------------------------+ 581 | Incoming Header | Outgoing Outer Header | 582 | (also equal to +---------------+---------------+---------------+ 583 | Outgoing Inner | RFC3168 ECN | RFC3168 ECN | RFC4301 IPsec | 584 | Header) | Limited | Full | | 585 | | Functionality | Functionality | | 586 +-----------------+---------------+---------------+---------------+ 587 | Not-ECT | Not-ECT | Not-ECT | Not-ECT | 588 | ECT(0) | Not-ECT | ECT(0) | ECT(0) | 589 | ECT(1) | Not-ECT | ECT(1) | ECT(1) | 590 | CE | Not-ECT | ECT(0) | CE | 591 +-----------------+---------------+---------------+---------------+ 593 Figure 1: IP in IP Encapsulation: Recap of Pre-existing Behaviours 595 3.2. Decapsulation at Tunnel Egress 597 RFC3168 and RFC4301 specify the decapsulation behaviour summarised in 598 Figure 2. The ECN field in the outgoing header is set to the 599 codepoint at the intersection of the appropriate incoming inner 600 header (row) and incoming outer header (column). 602 +---------+------------------------------------------------+ 603 |Incoming | Incoming Outer Header | 604 | Inner +---------+------------+------------+------------+ 605 | Header | Not-ECT | ECT(0) | ECT(1) | CE | 606 +---------+---------+------------+------------+------------+ 607 RFC3168->| Not-ECT | Not-ECT |Not-ECT |Not-ECT | drop | 608 RFC4301->| Not-ECT | Not-ECT |Not-ECT |Not-ECT |Not-ECT | 609 | ECT(0) | ECT(0) | ECT(0) | ECT(0) | CE | 610 | ECT(1) | ECT(1) | ECT(1) | ECT(1) | CE | 611 | CE | CE | CE | CE | CE | 612 +---------+---------+------------+------------+------------+ 614 In pre-existing RFCs, the ECN field in the outgoing header was set to 615 the codepoint at the intersection of the appropriate incoming inner 616 header (row) and incoming outer header (column). 618 Figure 2: IP in IP Decapsulation; Recap of Pre-existing Behaviour 620 The behaviour in the table derives from the logic given in RFC3168 621 and RFC4301, briefly recapped as follows: 623 o On decapsulation, if the inner ECN field is Not-ECT the outer is 624 ignored. RFC3168 (but not RFC4301) also specified that the 625 decapsulator must drop a packet with a Not-ECT inner and CE in the 626 outer. 628 o In all other cases, if the outer is CE, the outgoing ECN field is 629 set to CE, but otherwise the outer is ignored and the inner is 630 used for the outgoing ECN field. 632 Section 9.2.2 of RFC3168 also made it an auditable event for an IPsec 633 tunnel "if the ECN Field is changed inappropriately within an IPsec 634 tunnel...". Inappropriate changes were not specifically enumerated. 635 RFC4301 did not mention inappropriate ECN changes. 637 4. New ECN Tunnelling Rules 639 The standards actions below in Section 4.1 (ingress encapsulation) 640 and Section 4.2 (egress decapsulation) define new default ECN tunnel 641 processing rules for any IP packet (v4 or v6) with any Diffserv 642 codepoint. 644 If these defaults do not meet a particular requirement, an alternate 645 ECN tunnelling scheme can be introduced as part of the definition of 646 an alternate congestion marking scheme used by a specific Diffserv 647 PHB (see S.5 of [RFC3168] and [RFC4774]). When designing such 648 alternate ECN tunnelling schemes, the principles in Section 7 should 649 be followed. However, alternate ECN tunnelling schemes SHOULD be 650 avoided whenever possible as the deployment burden of handling 651 exceptional PHBs in implementations of all affected tunnels should 652 not be underestimated. There is no requirement for a PHB definition 653 to state anything about ECN tunnelling behaviour if the default 654 behaviour in the present specification is sufficient. 656 4.1. Default Tunnel Ingress Behaviour 658 Two modes of encapsulation are defined here; a REQUIRED `normal mode' 659 and a `compatibility mode', which is for backward compatibility with 660 tunnel decapsulators that do not understand ECN. Note that these are 661 modes of the ingress tunnel endpoint only, not the whole tunnel. 662 Section 4.3 explains why two modes are necessary and specifies the 663 circumstances in which it is sufficient to solely implement normal 664 mode. 666 Whatever the mode, an encapsulator forwards the inner header without 667 changing the ECN field. 669 In normal mode an encapsulator compliant with this specification MUST 670 construct the outer encapsulating IP header by copying the 2-bit ECN 671 field of the incoming IP header. In compatibility mode it clears the 672 ECN field in the outer header to the Not-ECT codepoint (the IPv4 673 header checksum also changes whenever the ECN field is changed). 674 These rules are tabulated for convenience in Figure 3. 675 +-----------------+-------------------------------+ 676 | Incoming Header | Outgoing Outer Header | 677 | (also equal to +---------------+---------------+ 678 | Outgoing Inner | Compatibility | Normal | 679 | Header) | Mode | Mode | 680 +-----------------+---------------+---------------+ 681 | Not-ECT | Not-ECT | Not-ECT | 682 | ECT(0) | Not-ECT | ECT(0) | 683 | ECT(1) | Not-ECT | ECT(1) | 684 | CE | Not-ECT | CE | 685 +-----------------+---------------+---------------+ 687 Figure 3: New IP in IP Encapsulation Behaviours 689 4.2. Default Tunnel Egress Behaviour 691 To decapsulate the inner header at the tunnel egress, a compliant 692 tunnel egress MUST set the outgoing ECN field to the codepoint at the 693 intersection of the appropriate incoming inner header (row) and outer 694 header (column) in Figure 4 (the IPv4 header checksum also changes 695 whenever the ECN field is changed). There is no need for more than 696 one mode of decapsulation, as these rules cater for all known 697 requirements. 698 +---------+------------------------------------------------+ 699 |Incoming | Incoming Outer Header | 700 | Inner +---------+------------+------------+------------+ 701 | Header | Not-ECT | ECT(0) | ECT(1) | CE | 702 +---------+---------+------------+------------+------------+ 703 | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| drop(!!!)| 704 | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE | 705 | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE | 706 | CE | CE | CE | CE(!!!)| CE | 707 +---------+---------+------------+------------+------------+ 709 The ECN field in the outgoing header is set to the codepoint at the 710 intersection of the appropriate incoming inner header (row) and 711 incoming outer header (column). Currently unused combinations are 712 indicated by '(!!!)' or '(!)' 714 Figure 4: New IP in IP Decapsulation Behaviour 716 This table for decapsulation behaviour is derived from the following 717 logic: 719 o If the inner ECN field is Not-ECT the decapsulator MUST NOT 720 propagate any other ECN codepoint onwards. This is because the 721 inner Not-ECT marking is set by transports that use drop as an 722 indication of congestion and would not understand or respond to 723 any other ECN codepoint [RFC4774]. Specifically: 725 * If the inner ECN field is Not-ECT and the outer ECN field is CE 726 the decapsulator MUST drop the packet. 728 * If the inner ECN field is Not-ECT and the outer ECN field is 729 Not-ECT, ECT(0) or ECT(1) the decapsulator MUST forward the 730 outgoing packet with the ECN field cleared to Not-ECT. 732 o In all other cases where the inner supports ECN, the decapsulator 733 MUST set the outgoing ECN field to the more severe marking of the 734 outer and inner ECN fields, where the ranking of severity from 735 highest to lowest is CE, ECT(1), ECT(0), Not-ECT. This in no way 736 precludes cases where ECT(1) and ECT(0) have the same severity; 738 o Certain combinations of inner and outer ECN fields cannot result 739 from any transition in any current or previous ECN tunneling 740 specification. These currently unused (CU) combinations are 741 indicated in Figure 4 by '(!!!)' or '(!)', where '(!!!)' means the 742 combination is CU and always potentially dangerous, while '(!)' 743 means it is CU and possibly dangerous. In these cases, 744 particularly the more dangerous ones, the decapsulator SHOULD log 745 the event and MAY also raise an alarm. 747 Just because the highlighted combinations are currently unused, 748 does not mean that all the other combinations are always valid. 749 Some are only valid if they have arrived from a particular type of 750 legacy ingress, and dangerous otherwise. Therefore an 751 implementation MAY allow an operator to configure logging and 752 alarms for such additional header combinations known to be 753 dangerous or CU for the particular configuration of tunnel 754 endpoints deployed at run-time. 756 Alarms SHOULD be rate-limited so that the anomalous combinations 757 will not amplify into a flood of alarm messages. It MUST be 758 possible to suppress alarms or logging, e.g. if it becomes 759 apparent that a combination that previously was not used has 760 started to be used for legitimate purposes such as a new standards 761 action. 763 The above logic allows for ECT(0) and ECT(1) to both represent the 764 same severity of congestion marking (e.g. "not congestion marked"). 765 But it also allows future schemes to be defined where ECT(1) is a 766 more severe marking than ECT(0), in particular enabling the simplest 767 possible encoding for PCN [I-D.ietf-pcn-3-in-1-encoding]. Before the 768 present specification was written, the PCN working-group had proposed 769 a number of work-rounds to the problem of a tunnel egress not 770 propagating two severity levels of congestion. Without wishing to 771 disparage the ingenuity of these work-rounds, none were chosen for 772 the standards track because they were either somewhat wasteful, 773 imprecise or complicated [Note_PCN_egress]. Treating ECT(1) as 774 either the same as ECT(0) or as a higher severity level is explained 775 in the discussion of the ECN nonce [RFC3540] in Section 8, which in 776 turn refers to Appendix F. 778 4.3. Encapsulation Modes 780 Section 4.1 introduces two encapsulation modes, normal mode and 781 compatibility mode, defining their encapsulation behaviour (i.e. 782 header copying or zeroing respectively). Note that these are modes 783 of the ingress tunnel endpoint only, not the tunnel as a whole. 785 To comply with this specification, a tunnel ingress MUST at least 786 implement `normal mode'. Unless it will never be used with legacy 787 tunnel egress nodes (RFC2003, RFC2401 or RFC2481 or the limited 788 functionality mode of RFC3168), an ingress MUST also implement 789 `compatibility mode' for backward compatibility with tunnel egresses 790 that do not propagate explicit congestion notifications [RFC4774]. 792 We can categorise the way that an ingress tunnel endpoint is paired 793 with an egress as either static or dynamically discovered: 795 Static: Tunnel endpoints paired together by prior configuration. 797 Some implementations of encapsulator might always be statically 798 deployed, and constrained to never be paired with a legacy 799 decapsulator (RFC2003, RFC2401 or RFC2481 or the limited 800 functionality mode of RFC3168). In such a case, only normal mode 801 needs to be implemented. 803 For instance, RFC4301-compatible IPsec tunnel endpoints invariably 804 use IKEv2 [RFC4306] for key exchange, which was introduced 805 alongside RFC4301. Therefore both endpoints of an RFC4301 tunnel 806 can be sure that the other end is RFC4301-compatible, because the 807 tunnel is only formed after IKEv2 key management has completed, at 808 which point both ends will be RFC4301-compliant by definition. 809 Therefore an IPsec tunnel ingress does not need compatibility 810 mode, as it will never interact with legacy ECN tunnels. To 811 comply with the present specification, it only needs to implement 812 the required normal mode, which is identical to the pre-existing 813 RFC4301 behaviour. 815 Dynamic Discovery: Tunnel endpoints paired together by some form of 816 tunnel endpoint discovery, typically finding an egress on the path 817 taken by the first packet. 819 This specification does not require or recommend dynamic discovery 820 and it does not define how dynamic negotiation might be done, but 821 it recognises that proprietary tunnel endpoint discovery protocols 822 exist. It therefore sets down some constraints on discovery 823 protocols to ensure safe interworking. 825 If dynamic tunnel endpoint discovery might pair an ingress with a 826 legacy egress (RFC2003, RFC2401 or RFC2481 or the limited 827 functionality mode of RFC3168), the ingress MUST implement both 828 normal and compatibility mode. If the tunnel discovery process is 829 arranged to only ever find a tunnel egress that propagates ECN 830 (RFC3168 full functionality mode, RFC4301 or this present 831 specification), then a tunnel ingress can be complaint with the 832 present specification without implementing compatibility mode. 834 While a compliant tunnel ingress is discovering an egress, it MUST 835 send packets in compatibility mode in case the egress it discovers 836 is a legacy egress. If, through the discovery protocol, the 837 egress indicates that it is compliant with the present 838 specification, with RFC4301 or with RFC3168 full functionality 839 mode, the ingress can switch itself into normal mode. If the 840 egress denies compliance with any of these or returns an error 841 that implies it does not understand a request to work to any of 842 these ECN specifications, the tunnel ingress MUST remain in 843 compatibility mode. 845 If an ingress claims compliance with this specification it MUST NOT 846 permanently disable ECN processing across the tunnel (i.e. only using 847 compatibility mode). It is true that such a tunnel ingress is at 848 least safe with the ECN behaviour of any egress it may encounter, but 849 it does not meet the central aim of this specification: introducing 850 ECN support to tunnels. 852 Instead, if the ingress knows that the egress does support 853 propagation of ECN (full functionality mode of RFC3168 or RFC4301 or 854 the present specification), it SHOULD use normal mode, in order to 855 support ECN where possible. Note that this section started by saying 856 an ingress "MUST implement "normal mode, while it has just said an 857 ingress "SHOULD use" normal mode. This distinction is deliberate, to 858 allow the mode to be turned off in exceptional circumstances but to 859 ensure all implementations make normal mode available. 861 Implementation note: If a compliant node is the ingress for multiple 862 tunnels, a mode setting will need to be stored for each tunnel 863 ingress. However, if a node is the egress for multiple tunnels, 864 none of the tunnels will need to store a mode setting, because a 865 compliant egress only needs one mode. 867 4.4. Single Mode of Decapsulation 869 A compliant decapsulator only needs one mode of operation. However, 870 if a complaint egress is implemented to be dynamically discoverable, 871 it may need to respond to discovery requests from various types of 872 legacy tunnel ingress. This specification does not define how 873 dynamic negotiation might be done by (proprietary) discovery 874 protocols, but it sets down some constraints to ensure safe 875 interworking. 877 Through the discovery protocol, a tunnel ingress compliant with the 878 present specification might ask if the egress is compliant with the 879 present specification, with RFC4301 or with RFC3168 full 880 functionality mode. Or an RFC3168 tunnel ingress might try to 881 negotiate to use limited functionality or full functionality mode 882 [RFC3168]. In all these cases, a decapsulating tunnel egress 883 compliant with this specification MUST agree to any of these 884 requests, since it will behave identically in all these cases. 886 If no ECN-related mode is requested, a compliant tunnel egress MUST 887 continue without raising any error or warning, because its egress 888 behaviour is compatible with all the legacy ingress behaviours that 889 do not negotiate capabilities. 891 A compliant tunnel egress SHOULD raise a warning alarm about any 892 requests to enter modes it does not recognise but, for 'forward 893 compatibility' with standards actions possibly defined after it was 894 implemented, it SHOULD continue operating. 896 5. Updates to Earlier RFCs 898 5.1. Changes to RFC4301 ECN processing 900 Ingress: An RFC4301 IPsec encapsulator is not changed at all by the 901 present specification. It uses the normal mode of the present 902 specification, which defines packet encapsulation identically to 903 RFC4301. 905 Egress: An RFC4301 egress will need to be updated to the new 906 decapsulation behaviour in Figure 4, in order to comply with the 907 present specification. However, the changes are backward 908 compatible; combinations of inner and outer that result from any 909 protocol defined in the RFC series so far are unaffected. Only 910 combinations that have never been used have been changed, 911 effectively adding new behaviours to RFC4301 decapsulation without 912 altering existing behaviours. The following specific updates have 913 been made: 915 * The outer, not the inner, is propagated when the outer is 916 ECT(1) and the inner is ECT(0); 918 * A packet with Not-ECT in the inner and an outer of CE is 919 dropped rather than forwarded as Not-ECT; 921 * Certain combinations of inner and outer ECN field have been 922 identified as currently unused. These can trigger logging 923 and/or raise alarms. 925 Modes: RFC4301 tunnel endpoints do not need modes and are not 926 updated by the modes in the present specification. Effectively an 927 RFC4301 IPsec ingress solely uses the REQUIRED normal mode of 928 encapsulation, which is unchanged from RFC4301 encapsulation. It 929 will never [Note_Manual_Keying] need the OPTIONAL compatibility 930 mode as explained in Section 4.3. 932 5.2. Changes to RFC3168 ECN processing 933 Ingress: On encapsulation, the new rule in Figure 3 that a normal 934 mode tunnel ingress copies any ECN field into the outer header 935 updates the full functionality behaviour of an RFC3168 ingress. 936 Nonetheless, the new compatibility mode encapsulates packets 937 identically to the limited functionality mode of an RFC3168 938 ingress. 940 Egress: An RFC3168 egress will need to be updated to the new 941 decapsulation behaviour in Figure 4, in order to comply with the 942 present specification. However, the changes are backward 943 compatible; combinations of inner and outer that result from any 944 protocol defined in the RFC series so far are unaffected. Only 945 combinations that have never been used have been changed, 946 effectively adding new behaviours to RFC3168 decapsulation without 947 altering existing behaviours. The following specific updates have 948 been made: 950 * The outer, not the inner, is propagated when the outer is 951 ECT(1) and the inner is ECT(0); 953 * Certain combinations of inner and outer ECN field have been 954 identified as currently unused. These can trigger logging 955 and/or raise alarms. 957 Modes: An RFC3168 ingress will need to be updated if it is to comply 958 with the present specification, whether or not it implemented the 959 optional full functionality mode of RFC3168. 961 RFC3168 defined a (required) limited functionality mode and an 962 (optional) full functionality mode for a tunnel. In RFC3168, 963 modes applied to both ends of the tunnel, while in the present 964 specification, modes are only used at the ingress--a single egress 965 behaviour covers all cases. 967 The normal mode of encapsulation is an update to the encapsulation 968 behaviour of the full functionality mode of an RFC3168 ingress. 969 The compatibility mode of encapsulation is identical to the 970 encapsulation behaviour of the limited functionality mode of an 971 RFC3168 ingress, except it is optional. 973 The constraints on how tunnel discovery protocols set modes in 974 Section 4.3 and Section 4.4 are an update to RFC3168, but they are 975 unlikely to require code changes as they document safe practice. 977 5.3. Motivation for Changes 979 An overriding goal is to ensure the same ECN signals can mean the 980 same thing whatever tunnels happen to encapsulate an IP packet flow. 981 This removes gratuitous inconsistency, which otherwise constrains the 982 available design space and makes it harder to design networks and new 983 protocols that work predictably. 985 5.3.1. Motivation for Changing Encapsulation 987 The normal mode in Section 4 updates RFC3168 to make all IP in IP 988 encapsulation of the ECN field consistent--consistent with the way 989 both RFC4301 IPsec [RFC4301] and IP in MPLS or MPLS in MPLS 990 encapsulation [RFC5129] construct the ECN field. 992 Compatibility mode has also been defined so that a non-RFC4301 993 ingress can still switch to using drop across a tunnel for backwards 994 compatibility with legacy decapsulators that do not propagate ECN 995 correctly. 997 The trigger that motivated this update to RFC3168 encapsulation was a 998 standards track proposal for pre-congestion notification (PCN 999 [RFC5670]). PCN excess rate marking only works correctly if the ECN 1000 field is copied on encapsulation (as in RFC4301 and RFC5129); it does 1001 not work if ECN is reset (as in RFC3168). This is because PCN excess 1002 rate marking depends on the outer header revealing any congestion 1003 experienced so far on the whole path, not just since the last tunnel 1004 ingress [Note_PCN_ingress]. 1006 PCN allows a network operator to add flow admission and termination 1007 for inelastic traffic at the edges of a Diffserv domain, but without 1008 any per-flow mechanisms in the interior and without the generous 1009 provisioning typical of Diffserv, aiming to significantly reduce 1010 costs. The PCN architecture [RFC5559] states that RFC3168 IP in IP 1011 tunnelling of the ECN field cannot be used for any tunnel ingress in 1012 a PCN domain. Prior to the present specification, this left a stark 1013 choice between not being able to use PCN for inelastic traffic 1014 control or not being able to use the many tunnels already deployed 1015 for Mobile IP, VPNs and so forth. 1017 The present specification provides a clean solution to this problem, 1018 so that network operators who want to use both PCN and tunnels can 1019 specify that every tunnel ingress in a PCN region must comply with 1020 this latest specification. 1022 Rather than allow tunnel specifications to fragment further into one 1023 for PCN, one for IPsec and one for other tunnels, the opportunity has 1024 been taken to consolidate the diverging specifications back into a 1025 single tunnelling behaviour. Resetting ECN was originally motivated 1026 by a covert channel concern that has been deliberately set aside in 1027 RFC4301 IPsec. Therefore the reset behaviour of RFC3168 is an 1028 anomaly that we do not need to keep. Copying ECN on encapsulation is 1029 anyway simpler than resetting. So, as more tunnel endpoints comply 1030 with this single consistent specification, encapsulation will be 1031 simpler as well as more predictable. 1033 Appendix B assesses whether copying rather than resetting CE on 1034 ingress will cause any unintended side-effects, from the three 1035 perspectives of security, control and management. In summary this 1036 analysis finds that: 1038 o From the control perspective either copying or resetting works for 1039 existing arrangements, but copying has more potential for 1040 simplifying control and resetting breaks at least one proposal 1041 already on the standards track. 1043 o From the management and monitoring perspective copying is 1044 preferable. 1046 o From the traffic security perspective (enforcing congestion 1047 control, mitigating denial of service etc) copying is preferable. 1049 o From the information security perspective resetting is preferable, 1050 but the IETF Security Area now considers copying acceptable given 1051 the bandwidth of a 2-bit covert channel can be managed. 1053 Therefore there are two points against resetting CE on ingress while 1054 copying CE causes no significant harm. 1056 5.3.2. Motivation for Changing Decapsulation 1058 The specification for decapsulation in Section 4 fixes three problems 1059 with the pre-existing behaviours of both RFC3168 and RFC4301: 1061 1. The pre-existing rules prevented the introduction of alternate 1062 ECN semantics to signal more than one severity level of 1063 congestion [RFC4774], [RFC5559]. The four states of the 2-bit 1064 ECN field provide room for signalling two severity levels in 1065 addition to not-congested and not-ECN-capable states. But, the 1066 pre-existing rules assumed that two of the states (ECT(0) and 1067 ECT(1)) are always equivalent. This unnecessarily restricts the 1068 use of one of four codepoints (half a bit) in the IP (v4 & v6) 1069 header. The new rules are designed to work in either case; 1070 whether ECT(1) is more severe than or equivalent to ECT(0). 1072 As explained in Appendix B.1, the original reason for not 1073 forwarding the outer ECT codepoints was to limit the covert 1074 channel across a decapsulator to 1 bit per packet. However, now 1075 that the IETF Security Area has deemed that a 2-bit covert 1076 channel through an encapsulator is a manageable risk, the same 1077 should be true for a decapsulator. 1079 As well as being useful for general future-proofing, this problem 1080 is immediately pressing for standardisation of pre-congestion 1081 notification (PCN), which uses two severity levels of congestion. 1082 If a congested queue used ECT(1) in the outer header to signal 1083 more severe congestion than ECT(0), the pre-existing 1084 decapsulation rules would have thrown away this congestion 1085 signal, preventing tunnelled traffic from ever knowing that it 1086 should reduce its load. 1088 The PCN working group has had to consider a number of wasteful or 1089 convoluted work-rounds to this problem [Note_PCN_egress]. But by 1090 far the simplest approach is just to remove the covert channel 1091 blockages from tunnelling behaviour--now deemed unnecessary 1092 anyway. Then network operators that want to support two 1093 congestion severity-levels for PCN can specify that every tunnel 1094 egress in a PCN region must comply with this latest 1095 specification. 1097 Not only does this make two congestion severity-levels available 1098 for PCN standardisation, but also for other potential uses of the 1099 extra ECN codepoint (e.g. [VCP]). 1101 2. Cases are documented where a middlebox (e.g. a firewall) drops 1102 packets with header values that were currently unused (CU) when 1103 the box was deployed, often on the grounds that anything 1104 unexpected might be an attack. This tends to bar future use of 1105 CU values. The new decapsulation rules specify optional logging 1106 and/or alarms for specific combinations of inner and outer header 1107 that are currently unused. The aim is to give implementers a 1108 recourse other than drop if they are concerned about the security 1109 of CU values. It recognises legitimate security concerns about 1110 CU values but still eases their future use. If the alarms are 1111 interpreted as an attack (e.g. by a management system) the 1112 offending packets can be dropped. But alarms can be turned off 1113 if these combinations come into regular use (e.g. through a 1114 future standards action). 1116 3. While reviewing currently unused combinations of inner and outer, 1117 the opportunity was taken to define a single consistent behaviour 1118 for the three cases with a Not-ECT inner header but a different 1119 outer. RFC3168 and RFC4301 had diverged in this respect and even 1120 their common behaviours had never been justified. 1122 None of these combinations should result from Internet protocols 1123 in the RFC series, but future standards actions might put any or 1124 all of them to good use. Therefore it was decided that a 1125 decapsulator must forward a Not-ECT inner unchanged when the 1126 arriving outer is ECT(0) or ECT(1). But for safety it must drop 1127 a combination of Not-ECT inner and CE outer. Then, if some 1128 unfortunate misconfiguration resulted in a congested router 1129 marking CE on a packet that was originally Not-ECT, drop would be 1130 the only appropriate signal for the egress to propagate--the only 1131 signal a non-ECN-capable transport (Not-ECT) would understand. 1133 It may seem contradictory that the same argument has not been 1134 applied to the ECT(1) codepoint, given it is being proposed as an 1135 intermediate level of congestion in a scheme progressing through 1136 the IETF [I-D.ietf-pcn-3-in-1-encoding]. Instead, a decapsulator 1137 must forward a Not-ECT inner unchanged when its outer is ECT(1). 1138 The rationale for not dropping this CU combination is to ensure 1139 it will be usable if needed in the future. If any 1140 misconfiguration led to ECT(1) congestion signals with a Not-ECT 1141 inner, it would not be disastrous for the tunnel egress to 1142 suppress them, because the congestion should then escalate to CE 1143 marking, which the egress would drop, thus at least preventing 1144 congestion collapse. 1146 Problems 2 & 3 alone would not warrant a change to decapsulation, but 1147 it was decided they are worth fixing and making consistent at the 1148 same time as decapsulation code is changed to fix problem 1 (two 1149 congestion severity-levels). 1151 6. Backward Compatibility 1153 A tunnel endpoint compliant with the present specification is 1154 backward compatible when paired with any tunnel endpoint compliant 1155 with any previous tunnelling RFC, whether RFC4301, RFC3168 (see 1156 Section 3) or the earlier RFCs summarised in Appendix A (RFC2481, 1157 RFC2401 and RFC2003). Each case is enumerated below. 1159 6.1. Non-Issues Updating Decapsulation 1161 At the egress, this specification only augments the per-packet 1162 calculation of the ECN field (RFC3168 and RFC4301) for combinations 1163 of inner and outer headers that have so far not been used in any IETF 1164 protocols. 1166 Therefore, all other things being equal, if an RFC4301 IPsec egress 1167 is updated to comply with the new rules, it will still interwork with 1168 any RFC4301 compliant ingress and the packet outputs will be 1169 identical to those it would have output before (fully backward 1170 compatible). 1172 And, all other things being equal, if an RFC3168 egress is updated to 1173 comply with the same new rules, it will still interwork with any 1174 ingress complying with any previous specification (both modes of 1175 RFC3168, both modes of RFC2481, RFC2401 and RFC2003) and the packet 1176 outputs will be identical to those it would have output before (fully 1177 backward compatible). 1179 A compliant tunnel egress merely needs to implement the one behaviour 1180 in Section 4 with no additional mode or option configuration at the 1181 ingress or egress nor any additional negotiation with the ingress. 1182 The new decapsulation rules have been defined in such a way that 1183 congestion control will still work safely if any of the earlier 1184 versions of ECN processing are used unilaterally at the encapsulating 1185 ingress of the tunnel (any of RFC2003, RFC2401, either mode of 1186 RFC2481, either mode of RFC3168, RFC4301 and this present 1187 specification). 1189 6.2. Non-Update of RFC4301 IPsec Encapsulation 1191 An RFC4301 IPsec ingress can comply with this new specification 1192 without any update and it has no need for any new modes, options or 1193 configuration. So, all other things being equal, it will continue to 1194 interwork identically with any egress it worked with before (fully 1195 backward compatible). 1197 6.3. Update to RFC3168 Encapsulation 1199 The encapsulation behaviour of the new normal mode copies the ECN 1200 field whereas RFC3168 full functionality mode reset it. However, all 1201 other things being equal, if RFC3168 ingress is updated to the 1202 present specification, the outgoing packets from any tunnel egress 1203 will still be unchanged. This is because all variants of tunnelling 1204 at either end (RFC4301, both modes of RFC3168, both modes of RFC2481, 1205 RFC2401, RFC2003 and the present specification) have always 1206 propagated an incoming CE marking through the inner header and onward 1207 into the outgoing header, whether the outer header is reset or 1208 copied. Therefore, If the tunnel is considered as a black box, the 1209 packets output from any egress will be identical with or without an 1210 update to the ingress. Nonetheless, if packets are observed within 1211 the black box (between the tunnel endpoints), CE markings copied by 1212 the updated ingress will be visible within the black box, whereas 1213 they would not have been before. Therefore, the update to 1214 encapsulation can be termed 'black-box backwards compatible' (i.e. 1215 identical unless you look inside the tunnel). 1217 This specification introduces no new backward compatibility issues 1218 when a compliant ingress talks with a legacy egress, but it has to 1219 provide similar safeguards to those already defined in RFC3168. 1220 RFC3168 laid down rules to ensure that an RFC3168 ingress turns off 1221 ECN (limited functionality mode) if it is paired with a legacy egress 1222 (RFC 2481, RFC2401 or RFC2003), which would not propagate ECN 1223 correctly. The present specification carries forward those rules 1224 (Section 4.3). It uses compatibility mode whenever RFC3168 would 1225 have used limited functionality mode, and their per-packet behaviours 1226 are identical. Therefore, all other things being equal, an ingress 1227 using the new rules will interwork with any legacy tunnel egress in 1228 exactly the same way as an RFC3168 ingress (still black-box backward 1229 compatible). 1231 7. Design Principles for Alternate ECN Tunnelling Semantics 1233 This section is informative not normative. 1235 S.5 of RFC3168 permits the Diffserv codepoint (DSCP)[RFC2474] to 1236 'switch in' alternative behaviours for marking the ECN field, just as 1237 it switches in different per-hop behaviours (PHBs) for scheduling. 1238 [RFC4774] gives best current practice for designing such alternative 1239 ECN semantics and very briefly mentions in section 5.4 that 1240 tunnelling should be considered. The guidance below extends RFC4774, 1241 giving additional guidance on designing any alternate ECN semantics 1242 that would also require alternate tunnelling semantics. 1244 The overriding guidance is: "Avoid designing alternate ECN tunnelling 1245 semantics, if at all possible." If a scheme requires tunnels to 1246 implement special processing of the ECN field for certain DSCPs, it 1247 will be hard to guarantee that every implementer of every tunnel will 1248 have added the required exception or that operators will have 1249 ubiquitously deployed the required updates. It is unlikely a single 1250 authority is even aware of all the tunnels in a network, which may 1251 include tunnels set up by applications between endpoints, or 1252 dynamically created in the network. Therefore it is highly likely 1253 that some tunnels within a network or on hosts connected to it will 1254 not implement the required special case. 1256 That said, if a non-default scheme for tunnelling the ECN field is 1257 really required, the following guidelines may prove useful in its 1258 design: 1260 On encapsulation in any alternate scheme: 1262 1. The ECN field of the outer header should be cleared to Not-ECT 1263 ("00") unless it is guaranteed that the corresponding tunnel 1264 egress will correctly propagate congestion markings introduced 1265 across the tunnel in the outer header. 1267 2. If it has established that ECN will be correctly propagated, 1268 an encapsulator should also copy incoming congestion 1269 notification into the outer header. The general principle 1270 here is that the outer header should reflect congestion 1271 accumulated along the whole upstream path, not just since the 1272 tunnel ingress (Appendix B.3 on management and monitoring 1273 explains). 1275 In some circumstances (e.g. pseudowires, PCN), the whole path 1276 is divided into segments, each with its own congestion 1277 notification and feedback loop. In these cases, the function 1278 that regulates load at the start of each segment will need to 1279 reset congestion notification for its segment. Often the 1280 point where congestion notification is reset will also be 1281 located at the start of a tunnel. However, the resetting 1282 function should be thought of as being applied to packets 1283 after the encapsulation function--two logically separate 1284 functions even though they might run on the same physical box. 1285 Then the code module doing encapsulation can keep to the 1286 copying rule and the load regulator module can reset 1287 congestion, without any code in either module being 1288 conditional on whether the other is there. 1290 On decapsulation in any new scheme: 1292 1. If the arriving inner header is Not-ECT it implies the 1293 transport will not understand other ECN codepoints. If the 1294 outer header carries an explicit congestion marking, the 1295 alternate scheme would be expected to drop the packet--the 1296 only indication of congestion the transport will understand. 1297 If the alternate scheme recommends forwarding rather than 1298 dropping such a packet, it must clearly justify this decision. 1299 If the inner is Not-ECT and the outer carries any other ECN 1300 codepoint that does not indicate congestion, the alternate 1301 scheme can forward the packet, but probably only as Not-ECT. 1303 2. If the arriving inner header is other than Not-ECT, the ECN 1304 field that the alternate decapsulation scheme forwards should 1305 reflect the more severe congestion marking of the arriving 1306 inner and outer headers. 1308 3. Any alternate scheme must define a behaviour for all 1309 combinations of inner and outer headers, even those that would 1310 not be expected to result from standards known at the time and 1311 even those that would not be expected from the tunnel ingress 1312 paired with the egress at run-time. Consideration should be 1313 given to logging such unexpected combinations and raising an 1314 alarm, particularly if there is a danger that the invalid 1315 combination implies congestion signals are not being 1316 propagated correctly. The presence of currently unused 1317 combinations may represent an attack, but the new scheme 1318 should try to define a way to forward such packets, at least 1319 if a safe outgoing codepoint can be defined. 1321 Raising an alarm allows a management system to decide whether 1322 the anomaly is indeed an attack, in which case it can decide 1323 to drop such packets. This is a preferable approach to hard- 1324 coded discard of packets that seem anomalous today, but may be 1325 needed tomorrow in future standards actions. 1327 IANA Considerations (to be removed on publication): 1329 This memo includes no request to IANA. 1331 8. Security Considerations 1333 Appendix B.1 discusses the security constraints imposed on ECN tunnel 1334 processing. The new rules for ECN tunnel processing (Section 4) 1335 trade-off between information security (covert channels) and traffic 1336 security (congestion monitoring & control). Ensuring congestion 1337 markings are not lost is itself an aspect of security, because if we 1338 allowed congestion notification to be lost, any attempt to enforce a 1339 response to congestion would be much harder. 1341 Specialist security issues: 1343 Tunnels intersecting Diffserv regions with alternate ECN semantics: 1344 If alternate congestion notification semantics are defined for a 1345 certain Diffserv PHB, the scope of the alternate semantics might 1346 typically be bounded by the limits of a Diffserv region or 1347 regions, as envisaged in [RFC4774] (e.g. the pre-congestion 1348 notification architecture [RFC5559]). The inner headers in 1349 tunnels crossing the boundary of such a Diffserv region but ending 1350 within the region can potentially leak the external congestion 1351 notification semantics into the region, or leak the internal 1352 semantics out of the region. [RFC2983] discusses the need for 1353 Diffserv traffic conditioning to be applied at these tunnel 1354 endpoints as if they are at the edge of the Diffserv region. 1355 Similar concerns apply to any processing or propagation of the ECN 1356 field at the endpoints of tunnels with one end inside and the 1357 other outside the domain. [RFC5559] gives specific advice on this 1358 for the PCN case, but other definitions of alternate semantics 1359 will need to discuss the specific security implications in each 1360 case. 1362 ECN nonce tunnel coverage: The new decapsulation rules improve the 1363 coverage of the ECN nonce [RFC3540] relative to the previous rules 1364 in RFC3168 and RFC4301. However, nonce coverage is still not 1365 perfect, as this would have led to a safety problem in another 1366 case. Both are corner-cases, so discussion of the compromise 1367 between them is deferred to Appendix F. 1369 Covert channel not turned off: A legacy (RFC3168) tunnel ingress 1370 could ask an RFC3168 egress to turn off ECN processing as well as 1371 itself turning off ECN. An egress compliant with the present 1372 specification will agree to such a request from a legacy ingress, 1373 but it relies on the ingress always sending Not-ECT in the outer. 1374 If the egress receives other ECN codepoints in the outer it will 1375 process them as normal, so it will actually still copy congestion 1376 markings from the outer to the outgoing header. Referring for 1377 example to Figure 5 (Appendix B.1), although the tunnel ingress 1378 'I' will set all ECN fields in outer headers to Not-ECT, 'M' could 1379 still toggle CE or ECT(1) on and off to communicate covertly with 1380 'B', because we have specified that 'E' only has one mode 1381 regardless of what mode it says it has negotiated. We could have 1382 specified that 'E' should have a limited functionality mode and 1383 check for such behaviour. But we decided not to add the extra 1384 complexity of two modes on a compliant tunnel egress merely to 1385 cater for an historic security concern that is now considered 1386 manageable. 1388 9. Conclusions 1390 This document allows tunnels to propagate an extra level of 1391 congestion severity. It uses previously unused combinations of inner 1392 and outer header to augment the rules for calculating the ECN field 1393 when decapsulating IP packets at the egress of IPsec (RFC4301) and 1394 non-IPsec (RFC3168) tunnels. 1396 This document also updates the ingress tunnelling encapsulation of 1397 RFC3168 ECN to bring all IP in IP tunnels into line with the new 1398 behaviour in the IPsec architecture of RFC4301, which copies rather 1399 than resets the ECN field when creating outer headers. 1401 The need for both these updated behaviours was triggered by the 1402 introduction of pre-congestion notification (PCN) onto the IETF 1403 standards track. Operators wanting to support PCN or other alternate 1404 ECN schemes that use an extra severity level can require that their 1405 tunnels comply with the present specification. This is not a fork in 1406 the RFC series, it is an update that can be deployed first by those 1407 that need it, and subsequently by all tunnel endpoint implementations 1408 during general code maintenance. It is backward compatible with all 1409 previous tunnelling behaviours, so existing single severity level 1410 schemes will continue to work as before, but support for two severity 1411 levels will gradually be added to the Internet. 1413 The new rules propagate changes to the ECN field across tunnel end- 1414 points that previously blocked them to restrict the bandwidth of a 1415 potential covert channel. Limiting the channel's bandwidth to 2 bits 1416 per packet is now considered sufficient. 1418 At the same time as removing these legacy constraints, the 1419 opportunity has been taken to draw together diverging tunnel 1420 specifications into a single consistent behaviour. Then any tunnel 1421 can be deployed unilaterally, and it will support the full range of 1422 congestion control and management schemes without any modes or 1423 configuration. Further, any host or router can expect the ECN field 1424 to behave in the same way, whatever type of tunnel might intervene in 1425 the path. This new certainty could enable new uses of the ECN field 1426 that would otherwise be confounded by ambiguity. 1428 10. Acknowledgements 1430 Thanks to David Black for his insightful reviews and patient 1431 explanations of better ways to think about function placement and 1432 alarms. Thanks to David and to Anil Agawaal for pointing out cases 1433 where it is safe to forward CU combinations of headers. Also thanks 1434 to Arnaud Jacquet for the idea for Appendix C. Thanks to Gorry 1435 Fairhurst, Teco Boot, Michael Menth, Bruce Davie, Toby Moncaster, 1436 Sally Floyd, Alfred Hoenes, Gabriele Corliano, Ingemar Johansson and 1437 Phil Eardley for their thoughts and careful review comments. 1439 Bob Briscoe is partly funded by Trilogy, a research project (ICT- 1440 216372) supported by the European Community under its Seventh 1441 Framework Programme. The views expressed here are those of the 1442 author only. 1444 Comments Solicited (to be removed by the RFC Editor): 1446 Comments and questions are encouraged and very welcome. They can be 1447 addressed to the IETF Transport Area working group mailing list 1448 , and/or to the authors. 1450 11. References 1452 11.1. Normative References 1454 [RFC2003] Perkins, C., "IP Encapsulation 1455 within IP", RFC 2003, October 1996. 1457 [RFC2119] Bradner, S., "Key words for use in 1458 RFCs to Indicate Requirement 1459 Levels", BCP 14, RFC 2119, 1460 March 1997. 1462 [RFC3168] Ramakrishnan, K., Floyd, S., and D. 1463 Black, "The Addition of Explicit 1464 Congestion Notification (ECN) to 1465 IP", RFC 3168, September 2001. 1467 [RFC4301] Kent, S. and K. Seo, "Security 1468 Architecture for the Internet 1469 Protocol", RFC 4301, December 2005. 1471 11.2. Informative References 1473 [I-D.ietf-pcn-3-in-1-encoding] Briscoe, B. and T. Moncaster, "PCN 1474 3-State Encoding Extension in a 1475 single DSCP", 1476 draft-ietf-pcn-3-in-1-encoding-01 1477 (work in progress), February 2010. 1479 [I-D.ietf-pcn-3-state-encoding] Briscoe, B., Moncaster, T., and M. 1480 Menth, "A PCN encoding using 2 1481 DSCPs to provide 3 or more states", 1482 draft-ietf-pcn-3-state-encoding-01 1483 (work in progress), February 2010. 1485 [I-D.ietf-pcn-psdm-encoding] Menth, M., Babiarz, J., Moncaster, 1486 T., and B. Briscoe, "PCN Encoding 1487 for Packet-Specific Dual Marking 1488 (PSDM)", 1489 draft-ietf-pcn-psdm-encoding-00 1490 (work in progress), June 2009. 1492 [I-D.ietf-pcn-sm-edge-behaviour] Charny, A., Karagiannis, G., Menth, 1493 M., and T. Taylor, "PCN Boundary 1494 Node Behaviour for the Single 1495 Marking (SM) Mode of Operation", 1496 draft-ietf-pcn-sm-edge-behaviour-01 1497 (work in progress), October 2009. 1499 [I-D.satoh-pcn-st-marking] Satoh, D., Ueno, H., Maeda, Y., and 1500 O. Phanachet, "Single PCN Threshold 1501 Marking by using PCN baseline 1502 encoding for both admission and 1503 termination controls", 1504 draft-satoh-pcn-st-marking-02 (work 1505 in progress), September 2009. 1507 [RFC2401] Kent, S. and R. Atkinson, "Security 1508 Architecture for the Internet 1509 Protocol", RFC 2401, November 1998. 1511 [RFC2474] Nichols, K., Blake, S., Baker, F., 1512 and D. Black, "Definition of the 1513 Differentiated Services Field (DS 1514 Field) in the IPv4 and IPv6 1515 Headers", RFC 2474, December 1998. 1517 [RFC2481] Ramakrishnan, K. and S. Floyd, "A 1518 Proposal to add Explicit Congestion 1519 Notification (ECN) to IP", 1520 RFC 2481, January 1999. 1522 [RFC2983] Black, D., "Differentiated Services 1523 and Tunnels", RFC 2983, 1524 October 2000. 1526 [RFC3540] Spring, N., Wetherall, D., and D. 1527 Ely, "Robust Explicit Congestion 1528 Notification (ECN) Signaling with 1529 Nonces", RFC 3540, June 2003. 1531 [RFC4306] Kaufman, C., "Internet Key Exchange 1532 (IKEv2) Protocol", RFC 4306, 1533 December 2005. 1535 [RFC4774] Floyd, S., "Specifying Alternate 1536 Semantics for the Explicit 1537 Congestion Notification (ECN) 1538 Field", BCP 124, RFC 4774, 1539 November 2006. 1541 [RFC5129] Davie, B., Briscoe, B., and J. Tay, 1542 "Explicit Congestion Marking in 1543 MPLS", RFC 5129, January 2008. 1545 [RFC5559] Eardley, P., "Pre-Congestion 1546 Notification (PCN) Architecture", 1547 RFC 5559, June 2009. 1549 [RFC5670] Eardley, P., "Metering and Marking 1550 Behaviour of PCN-Nodes", RFC 5670, 1551 November 2009. 1553 [RFC5696] Moncaster, T., Briscoe, B., and M. 1554 Menth, "Baseline Encoding and 1555 Transport of Pre-Congestion 1556 Information", RFC 5696, 1557 November 2009. 1559 [VCP] Xia, Y., Subramanian, L., Stoica, 1560 I., and S. Kalyanaraman, "One more 1561 bit is enough", Proc. SIGCOMM'05, 1562 ACM CCR 35(4)37--48, 2005, . 1566 Editorial Comments 1568 [Note_Manual_Keying] Bob Briscoe: Note (To be removed by the RFC 1569 Editor): One corner case can exist where an 1570 RFC4301 ingress does not use IKEv2, but uses 1571 manual keying instead. Then an RFC4301 ingress 1572 could conceivably be configured to tunnel to an 1573 egress with limited functionality ECN handling. 1574 Strictly, for this corner-case, the requirement 1575 to use compatibility mode in this specification 1576 updates RFC4301. However, this is such a remote 1577 possibility that RFC4301 IPsec implementations 1578 are not required to implement compatibility 1579 mode. It is planned to remove this note after 1580 the review process has completed to avoid 1581 unnecessarily complicating the document with a 1582 largely theoretical corner case. 1584 [Note_PCN_egress] Bob Briscoe: During the review process Appendix 1585 D is provided to expand on this point, but it 1586 will be deleted before publication. 1588 [Note_PCN_ingress] Bob Briscoe: During the review process Appendix 1589 E is provided to expand on this point, but it 1590 will be deleted before publication. 1592 Appendix A. Early ECN Tunnelling RFCs 1594 IP in IP tunnelling was originally defined in [RFC2003]. On 1595 encapsulation, the incoming header was copied to the outer and on 1596 decapsulation the outer was simply discarded. Initially, IPsec 1597 tunnelling [RFC2401] followed the same behaviour. 1599 When ECN was introduced experimentally in [RFC2481], legacy (RFC2003 1600 or RFC2401) tunnels would have discarded any congestion markings 1601 added to the outer header, so RFC2481 introduced rules for 1602 calculating the outgoing header from a combination of the inner and 1603 outer on decapsulation. RC2481 also introduced a second mode for 1604 IPsec tunnels, which turned off ECN processing (Not-ECT) in the outer 1605 header on encapsulation because an RFC2401 decapsulator would discard 1606 the outer on decapsulation. For RFC2401 IPsec this had the side- 1607 effect of completely blocking the covert channel. 1609 In RFC2481 the ECN field was defined as two separate bits. But when 1610 ECN moved from the experimental to the standards track [RFC3168], the 1611 ECN field was redefined as four codepoints. This required a 1612 different calculation of the ECN field from that used in RFC2481 on 1613 decapsulation. RFC3168 also had two modes; a 'full functionality 1614 mode' that restricted the covert channel as much as possible but 1615 still allowed ECN to be used with IPsec, and another that completely 1616 turned off ECN processing across the tunnel. This 'limited 1617 functionality mode' both offered a way for operators to completely 1618 block the covert channel and allowed an RFC3168 ingress to interwork 1619 with a legacy tunnel egress (RFC2481, RFC2401 or RFC2003). 1621 The present specification includes a similar compatibility mode to 1622 interwork safely with tunnels compliant with any of these three 1623 earlier RFCs. However, unlike RFC3168, it is only a mode of the 1624 ingress, as decapsulation behaviour is the same in either case. 1626 Appendix B. Design Constraints 1628 Tunnel processing of a congestion notification field has to meet 1629 congestion control and management needs without creating new 1630 information security vulnerabilities (if information security is 1631 required). This appendix documents the analysis of the tradeoffs 1632 between these factors that led to the new encapsulation rules in 1633 Section 4.1. 1635 B.1. Security Constraints 1637 Information security can be assured by using various end to end 1638 security solutions (including IPsec in transport mode [RFC4301]), but 1639 a commonly used scenario involves the need to communicate between two 1640 physically protected domains across the public Internet. In this 1641 case there are certain management advantages to using IPsec in tunnel 1642 mode solely across the publicly accessible part of the path. The 1643 path followed by a packet then crosses security 'domains'; the ones 1644 protected by physical or other means before and after the tunnel and 1645 the one protected by an IPsec tunnel across the otherwise unprotected 1646 domain. The scenario in Figure 5 will be used where endpoints 'A' 1647 and 'B' communicate through a tunnel. The tunnel ingress 'I' and 1648 egress 'E' are within physically protected edge domains, while the 1649 tunnel spans an unprotected internetwork where there may be 'men in 1650 the middle', M. 1652 physically unprotected physically 1653 <-protected domain-><--domain--><-protected domain-> 1654 +------------------+ +------------------+ 1655 | | M | | 1656 | A-------->I=========>==========>E-------->B | 1657 | | | | 1658 +------------------+ +------------------+ 1659 <----IPsec secured----> 1660 tunnel 1662 Figure 5: IPsec Tunnel Scenario 1664 IPsec encryption is typically used to prevent 'M' seeing messages 1665 from 'A' to 'B'. IPsec authentication is used to prevent 'M' 1666 masquerading as the sender of messages from 'A' to 'B' or altering 1667 their contents. 'I' can use IPsec tunnel mode to allow 'A' to 1668 communicate with 'B', but impose encryption to prevent 'A' leaking 1669 information to 'M'. Or 'E' can insist that 'I' uses tunnel mode 1670 authentication to prevent 'M' communicating information to 'B'. 1672 Mutable IP header fields such as the ECN field (as well as the TTL/ 1673 Hop Limit and DS fields) cannot be included in the cryptographic 1674 calculations of IPsec. Therefore, if 'I' copies these mutable fields 1675 into the outer header that is exposed across the tunnel it will have 1676 allowed a covert channel from 'A' to M that bypasses its encryption 1677 of the inner header. And if 'E' copies these fields from the outer 1678 header to the inner, even if it validates authentication from 'I', it 1679 will have allowed a covert channel from 'M' to 'B'. 1681 ECN at the IP layer is designed to carry information about congestion 1682 from a congested resource towards downstream nodes. Typically a 1683 downstream transport might feed the information back somehow to the 1684 point upstream of the congestion that can regulate the load on the 1685 congested resource, but other actions are possible (see [RFC3168] 1686 S.6). In terms of the above unicast scenario, ECN effectively 1687 intends to create an information channel (for congestion signalling) 1688 from 'M' to 'B' (for 'B' to feed back to 'A'). Therefore the goals 1689 of IPsec and ECN are mutually incompatible, requiring some 1690 compromise. 1692 With respect to using the DS or ECN fields as covert channels, 1693 S.5.1.2 of RFC4301 says, "controls are provided to manage the 1694 bandwidth of this channel". Using the ECN processing rules of 1695 RFC4301, the channel bandwidth is two bits per datagram from 'A' to 1696 'M' and one bit per datagram from 'M' to 'A' (because 'E' limits the 1697 combinations of the 2-bit ECN field that it will copy). In both 1698 cases the covert channel bandwidth is further reduced by noise from 1699 any real congestion marking. RFC4301 implies that these covert 1700 channels are sufficiently limited to be considered a manageable 1701 threat. However, with respect to the larger (6b) DS field, the same 1702 section of RFC4301 says not copying is the default, but a 1703 configuration option can allow copying "to allow a local 1704 administrator to decide whether the covert channel provided by 1705 copying these bits outweighs the benefits of copying". Of course, an 1706 administrator considering copying of the DS field has to take into 1707 account that it could be concatenated with the ECN field giving an 8b 1708 per datagram covert channel. 1710 For tunnelling the 6b Diffserv field two conceptual models have had 1711 to be defined so that administrators can trade off security against 1712 the needs of traffic conditioning [RFC2983]: 1714 The uniform model: where the Diffserv field is preserved end-to-end 1715 by copying into the outer header on encapsulation and copying from 1716 the outer header on decapsulation. 1718 The pipe model: where the outer header is independent of that in the 1719 inner header so it hides the Diffserv field of the inner header 1720 from any interaction with nodes along the tunnel. 1722 However, for ECN, the new IPsec security architecture in RFC4301 only 1723 standardised one tunnelling model equivalent to the uniform model. 1724 It deemed that simplicity was more important than allowing 1725 administrators the option of a tiny increment in security, especially 1726 given not copying congestion indications could seriously harm 1727 everyone's network service. 1729 B.2. Control Constraints 1731 Congestion control requires that any congestion notification marked 1732 into packets by a resource will be able to traverse a feedback loop 1733 back to a function capable of controlling the load on that resource. 1734 To be precise, rather than calling this function the data source, it 1735 will be called the Load Regulator. This allows for exceptional cases 1736 where load is not regulated by the data source, but usually the two 1737 terms will be synonymous. Note the term "a function _capable of_ 1738 controlling the load" deliberately includes a source application that 1739 doesn't actually control the load but ought to (e.g. an application 1740 without congestion control that uses UDP). 1742 A--->R--->I=========>M=========>E-------->B 1744 Figure 6: Simple Tunnel Scenario 1746 A similar tunnelling scenario to the IPsec one just described will 1747 now be considered, but without the different security domains, 1748 because the focus now shifts to whether the control loop and 1749 management monitoring work (Figure 6). If resources in the tunnel 1750 are to be able to explicitly notify congestion and the feedback path 1751 is from 'B' to 'A', it will certainly be necessary for 'E' to copy 1752 any CE marking from the outer header to the inner header for onward 1753 transmission to 'B', otherwise congestion notification from resources 1754 like 'M' cannot be fed back to the Load Regulator ('A'). But it does 1755 not seem necessary for 'I' to copy CE markings from the inner to the 1756 outer header. For instance, if resource 'R' is congested, it can 1757 send congestion information to 'B' using the congestion field in the 1758 inner header without 'I' copying the congestion field into the outer 1759 header and 'E' copying it back to the inner header. 'E' can still 1760 write any additional congestion marking introduced across the tunnel 1761 into the congestion field of the inner header. 1763 All this shows that 'E' can preserve the control loop irrespective of 1764 whether 'I' copies congestion notification into the outer header or 1765 resets it. 1767 That is the situation for existing control arrangements but, because 1768 copying reveals more information, it would open up possibilities for 1769 better control system designs. For instance, resetting CE marking on 1770 encapsulation breaks the standards track PCN congestion marking 1771 scheme [RFC5670]. It ends up removing excessive amounts of traffic 1772 unnecessarily. Whereas copying CE markings at ingress leads to the 1773 correct control behaviour. 1775 B.3. Management Constraints 1777 As well as control, there are also management constraints. 1778 Specifically, a management system may monitor congestion markings in 1779 passing packets, perhaps at the border between networks as part of a 1780 service level agreement. For instance, monitors at the borders of 1781 autonomous systems may need to measure how much congestion has 1782 accumulated so far along the path, perhaps to determine between them 1783 how much of the congestion is contributed by each domain. 1785 In this document the baseline of congestion marking (or the 1786 Congestion Baseline) is defined as the source of the layer that 1787 created (or most recently reset) the congestion notification field. 1788 When monitoring congestion it would be desirable if the Congestion 1789 Baseline did not depend on whether packets were tunnelled or not. 1790 Given some tunnels cross domain borders (e.g. consider M in Figure 6 1791 is monitoring a border), it would therefore be desirable for 'I' to 1792 copy congestion accumulated so far into the outer headers, so that it 1793 is exposed across the tunnel. 1795 For management purposes it might be useful for the tunnel egress to 1796 be able to monitor whether congestion occurred across a tunnel or 1797 upstream of it. Superficially it appears that copying congestion 1798 markings at the ingress would make this difficult, whereas it was 1799 straightforward when an RFC3168 ingress reset them. However, 1800 Appendix C gives a simple and precise method for a tunnel egress to 1801 infer the congestion level introduced across a tunnel. It works 1802 irrespective of whether the ingress copies or resets congestion 1803 markings. 1805 Appendix C. Contribution to Congestion across a Tunnel 1807 This specification mandates that a tunnel ingress determines the ECN 1808 field of each new outer tunnel header by copying the arriving header. 1809 Concern has been expressed that this will make it difficult for the 1810 tunnel egress to monitor congestion introduced only along a tunnel, 1811 which is easy if the outer ECN field is reset at a tunnel ingress 1812 (RFC3168 full functionality mode). However, in fact copying CE marks 1813 at ingress will still make it easy for the egress to measure 1814 congestion introduced across a tunnel, as illustrated below. 1816 Consider 100 packets measured at the egress. Say it measures that 30 1817 are CE marked in the inner and outer headers and 12 have additional 1818 CE marks in the outer but not the inner. This means packets arriving 1819 at the ingress had already experienced 30% congestion. However, it 1820 does not mean there was 12% congestion across the tunnel. The 1821 correct calculation of congestion across the tunnel is p_t = 12/ 1822 (100-30) = 12/70 = 17%. This is easy for the egress to measure. It 1823 is simply the proportion of packets not marked in the inner header 1824 (70) that have a CE marking in the outer header (12). This technique 1825 works whether the ingress copies or resets CE markings, so it can be 1826 used by an egress that is not sure which RFC the ingress complies 1827 with. 1829 Figure 7 illustrates this in a combinatorial probability diagram. 1830 The square represents 100 packets. The 30% division along the bottom 1831 represents marking before the ingress, and the p_t division up the 1832 side represents marking introduced across the tunnel. 1834 ^ outer header marking 1835 | 1836 100% +-----+---------+ The large square 1837 | | | represents 100 packets 1838 | 30 | | 1839 | | | p_t = 12/(100-30) 1840 p_t + +---------+ = 12/70 1841 | | 12 | = 17% 1842 0 +-----+---------+---> 1843 0 30% 100% inner header marking 1845 Figure 7: Tunnel Marking of Packets Already Marked at Ingress 1847 Appendix D. Why Losing ECT(1) on Decapsulation Impedes PCN (to be 1848 removed before publication) 1850 Congestion notification with two severity levels is currently on the 1851 IETF's standards track agenda in the Congestion and Pre-Congestion 1852 Notification (PCN) working group. PCN needs all four possible states 1853 of congestion signalling in the 2-bit ECN field to be propagated at 1854 the egress, but pre-existing tunnels only propagate three. The four 1855 PCN states are: not PCN-enabled, not marked and two increasingly 1856 severe levels of congestion marking. The less severe marking means 1857 'stop admitting new traffic' and the more severe marking means 1858 'terminate some existing flows', which may be needed after reroutes 1859 (see [RFC5559] for more details). (Note on terminology: wherever 1860 this document counts four congestion states, the PCN working group 1861 would count this as three PCN states plus a not-PCN-enabled state.) 1863 Figure 2 (Section 3.2) shows that pre-existing decapsulation 1864 behaviour would have discarded any ECT(1) markings in outer headers 1865 if the inner was ECT(0). This prevented the PCN working group from 1866 using ECT(1) -- if a PCN node used ECT(1) to indicate one of the 1867 severity levels of congestion, any later tunnel egress would revert 1868 the marking to ECT(0) as if nothing had happened. Effectively the 1869 decapsulation rules of RFC4301 and RFC3168 waste one ECT codepoint; 1870 they treat the ECT(0) and ECT(1) codepoints as a single codepoint. 1872 A number of work-rounds to this problem were proposed in the PCN w-g; 1873 to add the fourth state another way or avoid needing it. Without 1874 wishing to disparage the ingenuity of these work-rounds, none were 1875 chosen for the standards track because they were either somewhat 1876 wasteful, imprecise or complicated: 1878 o One uses a pair of Diffserv codepoint(s) in place of each PCN DSCP 1879 to encode the extra state [I-D.ietf-pcn-3-state-encoding], using 1880 up the rapidly exhausting DSCP space while leaving an ECN 1881 codepoint unused. 1883 o Another survives tunnelling without an extra DSCP 1884 [I-D.ietf-pcn-psdm-encoding], but it requires the PCN edge 1885 gateways to share the initial state of a packet out of band. 1887 o Another proposes a more involved marking algorithm in forwarding 1888 elements to encode the three congestion notification states using 1889 only two ECN codepoints [I-D.satoh-pcn-st-marking]. 1891 o Another takes a different approach; it compromises the precision 1892 of the admission control mechanism in some network scenarios, but 1893 manages to work with just three encoding states and a single 1894 marking algorithm [I-D.ietf-pcn-sm-edge-behaviour]. 1896 Rather than require the IETF to bless any of these experimental 1897 encoding work-rounds, the present specification fixes the root cause 1898 of the problem so that operators deploying PCN can simply require 1899 that tunnel end-points within a PCN region should comply with this 1900 new ECN tunnelling specification. On the public Internet it would 1901 not be possible to know whether all tunnels complied with this new 1902 specification, but universal compliance is feasible for PCN, because 1903 it is intended to be deployed in a controlled Diffserv region. 1905 Given the present specification, the PCN w-g could progress a 1906 trivially simple four-state ECN encoding 1907 [I-D.ietf-pcn-3-in-1-encoding]. This would replace the interim 1908 standards track baseline encoding of just three states [RFC5696] 1909 which makes a fourth state available for any of the experimental 1910 alternatives. 1912 Appendix E. Why Resetting ECN on Encapsulation Impedes PCN (to be 1913 removed before publication) 1915 The PCN architecture says "...if encapsulation is done within the 1916 PCN-domain: Any PCN-marking is copied into the outer header. Note: A 1917 tunnel will not provide this behaviour if it complies with [RFC3168] 1918 tunnelling in either mode, but it will if it complies with [RFC4301] 1919 IPsec tunnelling. " 1921 The specific issue here concerns PCN excess rate marking [RFC5670]. 1922 The purpose of excess rate marking is to provide a bulk mechanism for 1923 interior nodes within a PCN domain to mark traffic that is exceeding 1924 a configured threshold bit-rate, perhaps after an unexpected event 1925 such as a reroute, a link or node failure, or a more widespread 1926 disaster. Reroutes are a common cause of QoS degradation in IP 1927 networks. After reroutes it is common for multiple links in a 1928 network to become stressed at once. Therefore, PCN excess rate 1929 marking has been carefully designed to ensure traffic marked at one 1930 queue will not be counted again for marking at subsequent queues (see 1931 the `Excess traffic meter function' of [RFC5670]). 1933 However, if an RFC3168 tunnel ingress intervenes, it resets the ECN 1934 field in all the outer headers. This will cause excess traffic to be 1935 counted more than once, leading to many flows being removed that did 1936 not need to be removed at all. This is why the an RFC3168 tunnel 1937 ingress cannot be used in a PCN domain. 1939 The ECN reset in RFC3168 is no longer deemed necessary, it is 1940 inconsistent with RFC4301, it is not as simple as RFC4301 and it is 1941 impeding deployment of new protocols like PCN. The present 1942 specification corrects this perverse situation. 1944 Appendix F. Compromise on Decap with ECT(1) Inner and ECT(0) Outer 1946 A packet with an ECT(1) inner and an ECT(0) outer should never arise 1947 from any known IETF protocol. Without giving a reason, RFC3168 and 1948 RFC4301 both say the outer should be ignored when decapsulating such 1949 a packet. This appendix explains why it was decided not to change 1950 this advice. 1952 In summary, ECT(0) always means 'not congested' and ECT(1) may imply 1953 the same [RFC3168] or it may imply a higher severity congestion 1954 signal [RFC4774], [I-D.ietf-pcn-3-in-1-encoding], depending on the 1955 transport in use. Whether they mean the same or not, at the ingress 1956 the outer should have started the same as the inner and only a broken 1957 or compromised router could have changed the outer to ECT(0). 1959 The decapsulator can detect this anomaly. But the question is, 1960 should it correct the anomaly by ignoring the outer, or should it 1961 reveal the anomaly to the end-to-end transport by forwarding the 1962 outer? 1964 On balance, it was decided that the decapsulator should correct the 1965 anomaly, but log the event and optionally raise an alarm. This is 1966 the safe action if ECT(1) is being used as a more severe marking than 1967 ECT(0), because it passes the more severe signal to the transport. 1968 However, it is not a good idea to hide anomalies, which is why an 1969 optional alarm is suggested. It should be noted that this anomaly 1970 may be the result of two changes to the outer: a broken or 1971 compromised router within the tunnel might be erasing congestion 1972 markings introduced earlier in the same tunnel by a congested router. 1973 In this case, the anomaly would be losing congestion signals, which 1974 needs immediate attention. 1976 The original reason for defining ECT(0) and ECT(1) as equivalent was 1977 so that the data source could use the ECN nonce [RFC3540] to detect 1978 if congestion signals were being erased. However, in this case, the 1979 decapsulator does not need a nonce to detect any anomalies introduced 1980 within the tunnel, because it has the inner as a record of the header 1981 at the ingress. Therefore, it was decided that the best compromise 1982 would be to give precedence to solving the safety issue over 1983 revealing the anomaly, because the anomaly could at least be detected 1984 and dealt with internally. 1986 Superficially, the opposite case where the inner and outer carry 1987 different ECT values, but with an ECT(1) outer and ECT(0) inner, 1988 seems to require a similar compromise. However, because that case is 1989 reversed, no compromise is necessary; it is best to forward the outer 1990 whether the transport expects the ECT(1) to mean a higher severity 1991 than ECT(0) or the same severity. Forwarding the outer either 1992 preserves a higher value (if it is higher) or it reveals an anomaly 1993 to the transport (if the two ECT codepoints mean the same severity). 1995 Appendix G. Open Issues 1997 The new decapsulation behaviour defined in Section 4.2 adds support 1998 for propagation of 2 severity levels of congestion. However 1999 transports have no way to discover whether there are any legacy 2000 tunnels on their path that will not propagate 2 severity levels. It 2001 would have been nice to add a feature for transports to check path 2002 support, but this remains an open issue that will have to be 2003 addressed in any future standards action to define an end-to-end 2004 scheme that requires 2-severity levels of congestion. PCN avoids 2005 this problem because it is only for a controlled region, so all 2006 legacy tunnels can be upgraded by the same operator that deploys PCN. 2008 Author's Address 2010 Bob Briscoe 2011 BT 2012 B54/77, Adastral Park 2013 Martlesham Heath 2014 Ipswich IP5 3RE 2015 UK 2017 Phone: +44 1473 645196 2018 EMail: bob.briscoe@bt.com 2019 URI: http://bobbriscoe.net/