idnits 2.17.1
draft-ietf-tsvwg-ecn-tunnel-05.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
== The document seems to use 'NOT RECOMMENDED' as an RFC 2119 keyword, but
does not include the phrase in its RFC 2119 key words list.
-- The exact meaning of the all-uppercase expression 'NOT REQUIRED' is not
defined in RFC 2119. If it is intended as a requirements expression, it
should be rewritten using one of the combinations defined in RFC 2119;
otherwise it should not be all-uppercase.
(Using the creation date from RFC3168, updated by this document, for
RFC5378 checks: 2000-11-17)
-- The document seems to lack a disclaimer for pre-RFC5378 work, but may
have content which was first submitted before 10 November 2008. If you
have contacted all the original authors and they are all willing to grant
the BCP78 rights to the IETF Trust, then this is fine, and you can ignore
this comment. If not, you may need to add the pre-RFC5378 disclaimer.
(See the Legal Provisions document at
https://trustee.ietf.org/license-info for more information.)
-- The document date (December 18, 2009) is 5242 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== Outdated reference: A later version (-11) exists of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-00
== Outdated reference: A later version (-02) exists of
draft-ietf-pcn-3-state-encoding-00
== 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 (~~), 6 warnings (==), 8 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 December 18, 2009
5 (if approved)
6 Intended status: Standards Track
7 Expires: June 21, 2010
9 Tunnelling of Explicit Congestion Notification
10 draft-ietf-tsvwg-ecn-tunnel-05
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 (backward compatible).
25 Nonetheless, operators wanting to support two severity levels (e.g.
26 for pre-congestion notification--PCN) can require compliance with
27 this new specification. A thorough analysis of the reasoning for
28 these changes and the implications is included. In the unlikely
29 event that the new rules do not meet a specific need, RFC4774 gives
30 guidance on designing alternate ECN semantics and this document
31 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 June 21, 2010.
56 Copyright Notice
58 Copyright (c) 2009 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 Table of Contents
73 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 8
74 1.1. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . 10
75 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 10
76 3. Summary of Pre-Existing RFCs . . . . . . . . . . . . . . . . . 11
77 3.1. Encapsulation at Tunnel Ingress . . . . . . . . . . . . . 12
78 3.2. Decapsulation at Tunnel Egress . . . . . . . . . . . . . . 13
79 4. New ECN Tunnelling Rules . . . . . . . . . . . . . . . . . . . 14
80 4.1. Default Tunnel Ingress Behaviour . . . . . . . . . . . . . 14
81 4.2. Default Tunnel Egress Behaviour . . . . . . . . . . . . . 15
82 4.3. Encapsulation Modes . . . . . . . . . . . . . . . . . . . 17
83 4.4. Single Mode of Decapsulation . . . . . . . . . . . . . . . 18
84 5. Updates to Earlier RFCs . . . . . . . . . . . . . . . . . . . 19
85 5.1. Changes to RFC4301 ECN processing . . . . . . . . . . . . 19
86 5.2. Changes to RFC3168 ECN processing . . . . . . . . . . . . 20
87 5.3. Motivation for Changes . . . . . . . . . . . . . . . . . . 20
88 5.3.1. Motivation for Changing Encapsulation . . . . . . . . 21
89 5.3.2. Motivation for Changing Decapsulation . . . . . . . . 22
90 6. Backward Compatibility . . . . . . . . . . . . . . . . . . . . 24
91 6.1. Non-Issues Updating Decapsulation . . . . . . . . . . . . 24
92 6.2. Non-Update of RFC4301 IPsec Encapsulation . . . . . . . . 25
93 6.3. Update to RFC3168 Encapsulation . . . . . . . . . . . . . 25
94 7. Design Principles for Alternate ECN Tunnelling Semantics . . . 26
95 8. Security Considerations . . . . . . . . . . . . . . . . . . . 28
96 9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 29
97 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 30
98 11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 30
99 11.1. Normative References . . . . . . . . . . . . . . . . . . . 30
100 11.2. Informative References . . . . . . . . . . . . . . . . . . 31
101 Appendix A. Early ECN Tunnelling RFCs . . . . . . . . . . . . . . 33
102 Appendix B. Design Constraints . . . . . . . . . . . . . . . . . 33
103 B.1. Security Constraints . . . . . . . . . . . . . . . . . . . 33
104 B.2. Control Constraints . . . . . . . . . . . . . . . . . . . 35
105 B.3. Management Constraints . . . . . . . . . . . . . . . . . . 37
106 Appendix C. Contribution to Congestion across a Tunnel . . . . . 37
107 Appendix D. Why Losing ECT(1) on Decapsulation Impedes PCN . . . 38
108 Appendix E. Why Resetting ECN on Encapsulation Impedes PCN . . . 39
109 Appendix F. Compromise on Decap with ECT(1) Inner and ECT(0)
110 Outer . . . . . . . . . . . . . . . . . . . . . . . . 40
111 Appendix G. Open Issues . . . . . . . . . . . . . . . . . . . . . 41
113 Request to the RFC Editor (to be removed on publication):
115 In the RFC index, RFC3168 should be identified as an update to
116 RFC2003. RFC4301 should be identified as an update to RFC3168.
118 Changes from previous drafts (to be removed by the RFC Editor)
120 Full text differences between IETF draft versions are available at
121 , and
122 between earlier individual draft versions at
123
125 From ietf-04 to ietf-05 (current):
127 * Functional changes:
129 + Section 4.2: ECT(1) outer with Not-ECT inner: reverted to
130 forwarding as Not-ECT (as in RFC3168 & RFC4301), rather than
131 dropping.
133 + Altered rationale in bullet 3 of Section 5.3.2 to justify
134 this.
136 + Distinguished alarms for dangerous and invalid combinations
137 and allowed combinations that are valid in some tunnel
138 configurations but dangerous in others to be alarmed at the
139 discretion of the implementer and/or operator.
141 + Altered advice on designing alternate ECN tunnelling
142 semantics to reflect the above changes.
144 * Textual changes:
146 + Changed "Future non-default schemes" to "Alternate ECN
147 Tunnelling Semantics" throughout.
149 + Cut down Appendix D and Appendix E for brevity.
151 + A number of clarifying edits & updated refs.
153 From ietf-03 to ietf-04:
155 * Functional changes: none
157 * Structural changes:
159 + Added "Open Issues" appendix
161 * Textual changes:
163 + Section title: "Changes from Earlier RFCs" -> "Updates to
164 Earlier RFCs"
166 + Emphasised that change on decap to previously unused
167 combinations will propagate PCN encoding.
169 + Acknowledged additional reviewers and updated references
171 From ietf-02 to ietf-03:
173 * Functional changes:
175 + Corrected errors in recap of previous RFCs, which wrongly
176 stated the different decapsulation behaviours of RFC3168 &
177 RFC4301 with a Not-ECT inner header. This also required
178 corrections to the "Changes from Earlier RFCs" and the
179 Motivations for these changes.
181 + Mandated that any future standards action SHOULD NOT use the
182 ECT(0) codepoint as an indication of congestion, without
183 giving strong reasons.
185 + Added optional alarm when decapsulating ECT(1) outer,
186 ECT(0), but noted it would need to be disabled for
187 2-severity level congestion (e.g. PCN).
189 * Structural changes:
191 + Removed Document Roadmap which merely repeated the Contents
192 (previously Section 1.2).
194 + Moved "Changes from Earlier RFCs" (Section 5) before
195 Section 6 on Backward Compatibility and internally organised
196 both by RFC, rather than by ingress then egress.
198 + Moved motivation for changing existing RFCs (Section 5.3) to
199 after the changes are specified.
201 + Moved informative "Design Principles for Future Non-Default
202 Schemes" after all the normative sections.
204 + Added Appendix A on early history of ECN tunnelling RFCs.
206 + Removed specialist appendix on "Relative Placement of
207 Tunnelling and In-Path Load Regulation" (Appendix D in the
208 -02 draft)
210 + Moved and updated specialist text on "Compromise on Decap
211 with ECT(1) Inner and ECT(0) Outer" from Security
212 Considerations to Appendix F
214 * Textual changes:
216 + Simplified vocabulary for non-native-english speakers
218 + Simplified Introduction and defined regularly used terms in
219 an expanded Terminology section.
221 + More clearly distinguished statically configured tunnels
222 from dynamic tunnel endpoint discovery, before explaining
223 operating modes.
225 + Simplified, cut-down and clarified throughout
227 + Updated references.
229 From ietf-01 to ietf-02:
231 * Scope reduced from any encapsulation of an IP packet to solely
232 IP in IP tunnelled encapsulation. Consequently changed title
233 and removed whole section 'Design Guidelines for New
234 Encapsulations of Congestion Notification' (to be included in a
235 future companion informational document).
237 * Included a new normative decapsulation rule for ECT(0) inner
238 and ECT(1) outer that had previously only been outlined in the
239 non-normative appendix 'Comprehensive Decapsulation Rules'.
240 Consequently:
242 + The Introduction has been completely re-written to motivate
243 this change to decapsulation along with the existing change
244 to encapsulation.
246 + The tentative text in the appendix that first proposed this
247 change has been split between normative standards text in
248 Section 4 and Appendix D, which explains specifically why
249 this change would streamline PCN. New text on the logic of
250 the resulting decap rules added.
252 * If inner/outer is Not-ECT/ECT(0), changed decapsulation to
253 propagate Not-ECT rather than drop the packet; and added
254 reasoning.
256 * Considerably restructured:
258 + "Design Constraints" analysis moved to an appendix
259 (Appendix B);
261 + Added Section 3 to summarise relevant existing RFCs;
263 + Structured Section 4 and Section 6 into subsections.
265 + Added tables to sections on old and new rules, for precision
266 and comparison.
268 + Moved Section 7 on Design Principles to the end of the
269 section specifying the new default normative tunnelling
270 behaviour. Rewritten and shifted text on identifiers and
271 in-path load regulators to Appendix B.1 [deleted in revision
272 -03].
274 From ietf-00 to ietf-01:
276 * Identified two additional alarm states in the decapsulation
277 rules (Figure 4) if ECT(X) in outer and inner contradict each
278 other.
280 * Altered Comprehensive Decapsulation Rules (Appendix D) so that
281 ECT(0) in the outer no longer overrides ECT(1) in the inner.
282 Used the term 'Comprehensive' instead of 'Ideal'. And
283 considerably updated the text in this appendix.
285 * Added Appendix D.1 (removed again in a later revision) to weigh
286 up the various ways the Comprehensive Decapsulation Rules might
287 be introduced. This replaces the previous contradictory
288 statements saying complex backwards compatibility interactions
289 would be introduced while also saying there would be no
290 backwards compatibility issues.
292 * Updated references.
294 From briscoe-01 to ietf-00:
296 * Re-wrote Appendix C giving much simpler technique to measure
297 contribution to congestion across a tunnel.
299 * Added discussion of backward compatibility of the ideal
300 decapsulation scheme in Appendix D
302 * Updated references. Minor corrections & clarifications
303 throughout.
305 From briscoe-00 to briscoe-01:
307 * Related everything conceptually to the uniform and pipe models
308 of RFC2983 on Diffserv Tunnels, and completely removed the
309 dependence of tunnelling behaviour on the presence of any in-
310 path load regulation by using the [1 - Before] [2 - Outer]
311 function placement concepts from RFC2983;
313 * Added specific cases where the existing standards limit new
314 proposals, particularly Appendix E;
316 * Added sub-structure to Introduction (Need for Rationalisation,
317 Roadmap), added new Introductory subsection on "Scope" and
318 improved clarity;
320 * Added Design Guidelines for New Encapsulations of Congestion
321 Notification;
323 * Considerably clarified the Backward Compatibility section
324 (Section 6);
326 * Considerably extended the Security Considerations section
327 (Section 8);
329 * Summarised the primary rationale much better in the
330 conclusions;
332 * Added numerous extra acknowledgements;
334 * Added Appendix E. "Why resetting CE on encapsulation harms
335 PCN", Appendix C. "Contribution to Congestion across a Tunnel"
336 and Appendix D. "Ideal Decapsulation Rules";
338 * Re-wrote Appendix B [deleted in a later revision], explaining
339 how tunnel encapsulation no longer depends on in-path load-
340 regulation (changed title from "In-path Load Regulation" to
341 "Non-Dependence of Tunnelling on In-path Load Regulation"), but
342 explained how an in-path load regulation function must be
343 carefully placed with respect to tunnel encapsulation (in a new
344 sub-section entitled "Dependence of In-Path Load Regulation on
345 Tunnelling").
347 1. Introduction
349 Explicit congestion notification (ECN [RFC3168]) allows a forwarding
350 element to notify the onset of congestion without having to drop
351 packets. Instead it can explicitly mark a proportion of packets in
352 the 2-bit ECN field in the IP header (Table 1 recaps the ECN
353 codepoints).
355 The outer header of an IP packet can encapsulate one or more IP
356 headers for tunnelling. A forwarding element using ECN to signify
357 congestion will only mark the immediately visible outer IP header.
358 When a tunnel decapsulator later removes this outer header, it
359 follows rules to propagate congestion markings by combining the ECN
360 fields of the inner and outer IP header into one outgoing IP header.
362 This document updates those rules for IPsec [RFC4301] and non-IPsec
363 [RFC3168] tunnels to add new behaviours for previously unused
364 combinations of inner and outer header. It also updates the tunnel
365 ingress behaviour of RFC3168 to match that of RFC4301. The updated
366 rules are backward compatible with RFC4301 and RFC3168 when
367 interworking with any other tunnel endpoint complying with any
368 earlier specification.
370 When ECN and its tunnelling was defined in RFC3168, only the minimum
371 necessary changes to the ECN field were propagated through tunnel
372 endpoints--just enough for the basic ECN mechanism to work. This was
373 due to concerns that the ECN field might be toggled to communicate
374 between a secure site and someone on the public Internet--a covert
375 channel. This was because a mutable field like ECN cannot be
376 protected by IPsec's integrity mechanisms--it has to be able to
377 change as it traverses the Internet.
379 Nonetheless, the latest IPsec architecture [RFC4301] considers a
380 bandwidth limit of 2 bits per packet on a covert channel makes it a
381 manageable risk. Therefore, for simplicity, an RFC4301 ingress
382 copies the whole ECN field to encapsulate a packet. It also
383 dispenses with the two modes of RFC3168, one which partially copied
384 the ECN field, and the other which blocked all propagation of ECN
385 changes.
387 Unfortunately, this entirely reasonable sequence of standards actions
388 resulted in a perverse outcome; non-IPsec tunnels (RFC3168) blocked
389 the 2-bit covert channel, while IPsec tunnels (RFC4301) did not--at
390 least not at the ingress. At the egress, both IPsec and non-IPsec
391 tunnels still partially restricted propagation of the full ECN field.
393 The trigger for the changes in this document was the introduction of
394 pre-congestion notification (PCN [RFC5670]) to the IETF standards
395 track. PCN needs the ECN field to be copied at a tunnel ingress and
396 it needs four states of congestion signalling to be propagated at the
397 egress, but pre-existing tunnels only propagate three in the ECN
398 field.
400 This document draws on currently unused (CU) combinations of inner
401 and outer headers to add tunnelling of four-state congestion
402 signalling to RFC3168 and RFC4301. Operators of tunnels who
403 specifically want to support four states can require that all their
404 tunnels comply with this specification. Nonetheless, all tunnel
405 endpoint implementations (RFC4301, RFC3168, RFC2481, RFC2401,
406 RFC2003) can safely be updated to this new specification as part of
407 general code maintenance. This will gradually add support for four
408 congestion states to the Internet. Existing three state schemes will
409 continue to work as before.
411 At the same time as harmonising covert channel constraints, the
412 opportunity has been taken to draw together diverging tunnel
413 specifications into a single consistent behaviour. Then any tunnel
414 can be deployed unilaterally, and it will support the full range of
415 congestion control and management schemes without any modes or
416 configuration. Further, any host or router can expect the ECN field
417 to behave in the same way, whatever type of tunnel might intervene in
418 the path.
420 1.1. Scope
422 This document only concerns wire protocol processing of the ECN field
423 at tunnel endpoints and makes no changes or recommendations
424 concerning algorithms for congestion marking or congestion response.
426 This document specifies common ECN field processing at encapsulation
427 and decapsulation for any IP in IP tunnelling, whether IPsec or non-
428 IPsec tunnels. It applies irrespective of whether IPv4 or IPv6 is
429 used for either of the inner and outer headers. It applies for
430 packets with any destination address type, whether unicast or
431 multicast. It applies as the default for all Diffserv per-hop
432 behaviours (PHBs), unless stated otherwise in the specification of a
433 PHB. It is intended to be a good trade off between somewhat
434 conflicting security, control and management requirements.
436 [RFC2983] is a comprehensive primer on differentiated services and
437 tunnels. Given ECN raises similar issues to differentiated services
438 when interacting with tunnels, useful concepts introduced in RFC2983
439 are used throughout, with brief recaps of the explanations where
440 necessary.
442 2. Terminology
444 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
445 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
446 document are to be interpreted as described in RFC 2119 [RFC2119].
448 Table 1 recaps the names of the ECN codepoints [RFC3168].
450 +------------------+----------------+---------------------------+
451 | Binary codepoint | Codepoint name | Meaning |
452 +------------------+----------------+---------------------------+
453 | 00 | Not-ECT | Not ECN-capable transport |
454 | 01 | ECT(1) | ECN-capable transport |
455 | 10 | ECT(0) | ECN-capable transport |
456 | 11 | CE | Congestion experienced |
457 +------------------+----------------+---------------------------+
459 Table 1: Recap of Codepoints of the ECN Field [RFC3168] in the IP
460 Header
462 Further terminology used within this document:
464 Encapsulator: The tunnel endpoint function that adds an outer IP
465 header to tunnel a packet (also termed the 'ingress tunnel
466 endpoint' or just the 'ingress' where the context is clear).
468 Decapsulator: The tunnel endpoint function that removes an outer IP
469 header from a tunnelled packet (also termed the 'egress tunnel
470 endpoint' or just the 'egress' where the context is clear).
472 Incoming header: The header of an arriving packet before
473 encapsulation.
475 Outer header: The header added to encapsulate a tunnelled packet.
477 Inner header: The header encapsulated by the outer header.
479 Outgoing header: The header constructed by the decapsulator using
480 logic that combines the fields in the outer and inner headers.
482 Copying ECN: On encapsulation, setting the ECN field of the new
483 outer header to be a copy of the ECN field in the incoming header.
485 Zeroing ECN: On encapsulation, clearing the ECN field of the new
486 outer header to Not-ECT ("00").
488 Resetting ECN: On encapsulation, setting the ECN field of the new
489 outer header to be a copy of the ECN field in the incoming header
490 except the outer ECN field is set to the ECT(0) codepoint if the
491 incoming ECN field is CE ("11").
493 3. Summary of Pre-Existing RFCs
495 This section is informative not normative, as it recaps pre-existing
496 RFCs. Earlier relevant RFCs that were either experimental or
497 incomplete with respect to ECN tunnelling (RFC2481, RFC2401 and
498 RFC2003) are briefly outlined in Appendix A. The question of whether
499 tunnel implementations used in the Internet comply with any of these
500 RFCs is not discussed.
502 3.1. Encapsulation at Tunnel Ingress
504 At the encapsulator, the controversy has been over whether to
505 propagate information about congestion experienced on the path so far
506 into the outer header of the tunnel.
508 Specifically, RFC3168 says that, if a tunnel fully supports ECN
509 (termed a 'full-functionality' ECN tunnel in [RFC3168]), the
510 encapsulator must not copy a CE marking from the inner header into
511 the outer header that it creates. Instead the encapsulator must set
512 the outer header to ECT(0) if the ECN field is marked CE in the
513 arriving IP header. We term this 'resetting' a CE codepoint.
515 However, the new IPsec architecture in [RFC4301] reverses this rule,
516 stating that the encapsulator must simply copy the ECN field from the
517 incoming header to the outer header.
519 RFC3168 also provided a Limited Functionality mode that turns off ECN
520 processing over the scope of the tunnel by setting the outer header
521 to Not-ECT ("00"). Then such packets will be dropped to indicate
522 congestion rather than marked with ECN. This is necessary for the
523 ingress to interwork with legacy decapsulators ([RFC2481], [RFC2401]
524 and [RFC2003]) that do not propagate ECN markings added to the outer
525 header. Otherwise such legacy decapsulators would throw away
526 congestion notifications before they reached the transport layer.
528 Neither Limited Functionality mode nor Full Functionality mode are
529 used by an RFC4301 IPsec encapsulator, which simply copies the
530 incoming ECN field into the outer header. An earlier key-exchange
531 phase ensures an RFC4301 ingress will not have to interwork with a
532 legacy egress that does not support ECN.
534 These pre-existing behaviours are summarised in Figure 1.
536 +-----------------+-----------------------------------------------+
537 | Incoming Header | Outgoing Outer Header |
538 | (also equal to +---------------+---------------+---------------+
539 | Outgoing Inner | RFC3168 ECN | RFC3168 ECN | RFC4301 IPsec |
540 | Header) | Limited | Full | |
541 | | Functionality | Functionality | |
542 +-----------------+---------------+---------------+---------------+
543 | Not-ECT | Not-ECT | Not-ECT | Not-ECT |
544 | ECT(0) | Not-ECT | ECT(0) | ECT(0) |
545 | ECT(1) | Not-ECT | ECT(1) | ECT(1) |
546 | CE | Not-ECT | ECT(0) | CE |
547 +-----------------+---------------+---------------+---------------+
549 Figure 1: IP in IP Encapsulation: Recap of Pre-existing Behaviours
551 3.2. Decapsulation at Tunnel Egress
553 RFC3168 and RFC4301 specify the decapsulation behaviour summarised in
554 Figure 2. The ECN field in the outgoing header is set to the
555 codepoint at the intersection of the appropriate incoming inner
556 header (row) and incoming outer header (column).
557 +---------+------------------------------------------------+
558 |Incoming | Incoming Outer Header |
559 | Inner +---------+------------+------------+------------+
560 | Header | Not-ECT | ECT(0) | ECT(1) | CE |
561 +---------+---------+------------+------------+------------+
562 RFC3168->| Not-ECT | Not-ECT |Not-ECT |Not-ECT | drop |
563 RFC4301->| Not-ECT | Not-ECT |Not-ECT |Not-ECT |Not-ECT |
564 | ECT(0) | ECT(0) | ECT(0) | ECT(0) | CE |
565 | ECT(1) | ECT(1) | ECT(1) | ECT(1) | CE |
566 | CE | CE | CE | CE | CE |
567 +---------+---------+------------+------------+------------+
568 | Outgoing Header |
569 +------------------------------------------------+
571 Figure 2: IP in IP Decapsulation; Recap of Pre-existing Behaviour
573 The behaviour in the table derives from the logic given in RFC3168
574 and RFC4301, briefly recapped as follows:
576 o On decapsulation, if the inner ECN field is Not-ECT the outer is
577 discarded. RFC3168 (but not RFC4301) also specified that the
578 decapsulator must drop a packet with a Not-ECT inner and CE in the
579 outer.
581 o In all other cases, if the outer is CE, the outgoing ECN field is
582 set to CE, but otherwise the outer is ignored and the inner is
583 used for the outgoing ECN field.
585 RFC3168 also made it an auditable event for an IPsec tunnel "if the
586 ECN Field is changed inappropriately within an IPsec tunnel...".
587 Inappropriate changes were not specifically enumerated. RFC4301 did
588 not mention inappropriate ECN changes.
590 4. New ECN Tunnelling Rules
592 The standards actions below in Section 4.1 (ingress encapsulation)
593 and Section 4.2 (egress decapsulation) define new default ECN tunnel
594 processing rules for any IP packet (v4 or v6) with any Diffserv
595 codepoint.
597 If unavoidable, an alternate congestion encapsulation behaviour can
598 be introduced as part of the definition of an alternate congestion
599 marking scheme used by a specific Diffserv PHB (see S.5 of [RFC3168]
600 and [RFC4774]). When designing such new encapsulation schemes, the
601 principles in Section 7 should be followed. However, alternate ECN
602 tunnelling schemes are NOT RECOMMENDED as the deployment burden of
603 handling exceptional PHBs in implementations of all affected tunnels
604 should not be underestimated. There is no requirement for a PHB
605 definition to state anything about ECN tunnelling behaviour if the
606 default behaviour in the present specification is sufficient.
608 4.1. Default Tunnel Ingress Behaviour
610 Two modes of encapsulation are defined here; `normal mode' and
611 `compatibility mode', which is for backward compatibility with tunnel
612 decapsulators that do not understand ECN. Section 4.3 explains why
613 two modes are necessary and specifies the circumstances in which it
614 is sufficient to solely implement normal mode. Note that these are
615 modes of the ingress tunnel endpoint only, not the whole tunnel.
617 Whatever the mode, an encapsulator forwards the inner header without
618 changing the ECN field.
620 In normal mode an encapsulator compliant with this specification MUST
621 construct the outer encapsulating IP header by copying the 2-bit ECN
622 field of the incoming IP header. In compatibility mode it clears the
623 ECN field in the outer header to the Not-ECT codepoint. These rules
624 are tabulated for convenience in Figure 3.
626 +-----------------+-------------------------------+
627 | Incoming Header | Outgoing Outer Header |
628 | (also equal to +---------------+---------------+
629 | Outgoing Inner | Compatibility | Normal |
630 | Header) | Mode | Mode |
631 +-----------------+---------------+---------------+
632 | Not-ECT | Not-ECT | Not-ECT |
633 | ECT(0) | Not-ECT | ECT(0) |
634 | ECT(1) | Not-ECT | ECT(1) |
635 | CE | Not-ECT | CE |
636 +-----------------+---------------+---------------+
638 Figure 3: New IP in IP Encapsulation Behaviours
640 An ingress in compatibility mode encapsulates packets identically to
641 an ingress in RFC3168's limited functionality mode. An ingress in
642 normal mode encapsulates packets identically to an RFC4301 IPsec
643 ingress.
645 4.2. Default Tunnel Egress Behaviour
647 To decapsulate the inner header at the tunnel egress, a compliant
648 tunnel egress MUST set the outgoing ECN field to the codepoint at the
649 intersection of the appropriate incoming inner header (row) and outer
650 header (column) in Figure 4 (the IPv4 header checksum also changes
651 whenever the ECN field is changed). There is no need for more than
652 one mode of decapsulation, as these rules cater for all known
653 requirements.
654 +---------+------------------------------------------------+
655 |Incoming | Incoming Outer Header |
656 | Inner +---------+------------+------------+------------+
657 | Header | Not-ECT | ECT(0) | ECT(1) | CE |
658 +---------+---------+------------+------------+------------+
659 | Not-ECT | Not-ECT |Not-ECT(!!!)|Not-ECT(!!!)| drop(!!!)|
660 | ECT(0) | ECT(0) | ECT(0) | ECT(1) | CE |
661 | ECT(1) | ECT(1) | ECT(1) (!) | ECT(1) | CE |
662 | CE | CE | CE | CE(!!!)| CE |
663 +---------+---------+------------+------------+------------+
664 | Outgoing Header |
665 +------------------------------------------------+
666 Unexpected combinations are indicated by '(!!!)'
668 Figure 4: New IP in IP Decapsulation Behaviour
670 This table for decapsulation behaviour is derived from the following
671 logic:
673 o If the inner ECN field is Not-ECT the decapsulator MUST NOT
674 propagate any other ECN codepoint onwards. This is because the
675 inner Not-ECT marking is set by transports that use drop as an
676 indication of congestion and would not understand or respond to
677 any other ECN codepoint [RFC4774]. In addition:
679 * If the inner ECN field is Not-ECT and the outer ECN field is CE
680 the decapsulator MUST drop the packet.
682 * If the inner ECN field is Not-ECT and the outer ECN field is
683 Not-ECT, ECT(0) or ECT(1) the decapsulator MUST forward the
684 outgoing packet with the ECN field cleared to Not-ECT.
686 o In all other cases where the inner supports ECN, the outgoing ECN
687 field is set to the more severe marking of the outer and inner ECN
688 fields, where the ranking of severity from highest to lowest is
689 CE, ECT(1), ECT(0), Not-ECT. This in no way precludes cases where
690 ECT(1) and ECT(0) have the same severity;
692 o Certain combinations of inner and outer ECN fields cannot result
693 from any currently used transition in any current or previous ECN
694 tunneling specification. These cases are indicated in Figure 4 by
695 '(!!!)' or '(!)', where '(!!!)' means the combination is both
696 invalid and always potentially dangerous, while '(!)' means it is
697 invalid and possibly dangerous. In these cases, particularly the
698 more dangerous ones, the decapsulator SHOULD log the event and MAY
699 also raise an alarm. Just because the highlighted combinations
700 are always invalid, does not mean that all the other combinations
701 are always valid. Some are only valid if they have arrived from a
702 particular type of legacy ingress, and dangerous otherwise.
703 Therefore an implementation MAY allow an operator to configure
704 logging and alarms for such additional header combinations known
705 to be dangerous or invalid for the particular configuration of
706 tunnel endpoints deployed at run-time.
708 Alarms should be rate-limited so that the illegal combinations
709 will not amplify into a flood of alarm messages. It MUST be
710 possible to suppress alarms or logging, e.g. if it becomes
711 apparent that a combination that previously was not used has
712 started to be used for legitimate purposes such as a new standards
713 action.
715 The above logic allows for ECT(0) and ECT(1) to both represent the
716 same severity of congestion marking (e.g. "not congestion marked").
717 But it also allows future schemes to be defined where ECT(1) is a
718 more severe marking than ECT(0), in particular enabling the simplest
719 possible encoding for PCN [I-D.ietf-pcn-3-in-1-encoding]. This
720 approach is discussed in Appendix D and in the discussion of the ECN
721 nonce [RFC3540] in Section 8, which in turn refers to Appendix F.
723 4.3. Encapsulation Modes
725 Section 4.1 introduces two encapsulation modes, normal mode and
726 compatibility mode, defining their encapsulation behaviour (i.e.
727 header copying or zeroing respectively). Note that these are modes
728 of the ingress tunnel endpoint only, not the tunnel as a whole.
730 A tunnel ingress MUST at least implement `normal mode' and, if it
731 might be used with legacy tunnel egress nodes (RFC2003, RFC2401 or
732 RFC2481 or the limited functionality mode of RFC3168), it MUST also
733 implement `compatibility mode' for backward compatibility with tunnel
734 egresses that do not propagate explicit congestion notifications
735 [RFC4774]. If the egress does support propagation of ECN (full
736 functionality mode of RFC3168 or RFC4301 or the present
737 specification), the ingress SHOULD use normal mode, in order to
738 support ECN where possible.
740 We can categorise the way that an ingress tunnel endpoint is paired
741 with an egress as either:
743 static: those paired together by prior configuration or;
745 dynamically discovered: those paired together by some form of tunnel
746 endpoint discovery, typically driven by the path taken by arriving
747 packets.
749 Static: Some implementations of encapsulator might be constrained to
750 be statically deployed, and constrained to never be paired with a
751 legacy decapsulator (RFC2003, RFC2401 or RFC2481 or the limited
752 functionality mode of RFC3168). In such a case, only normal mode
753 needs to be implemented.
755 For instance, RFC4301-compatible IPsec tunnel endpoints invariably
756 use IKEv2 [RFC4306] for key exchange, which was introduced alongside
757 RFC4301. Therefore both endpoints of an RFC4301 tunnel can be sure
758 that the other end is RFC4301-compatible, because the tunnel is only
759 formed after IKEv2 key management has completed, at which point both
760 ends will be RFC4301-compliant by definition. Further, an RFC4301
761 encapsulator behaves identically to the normal mode of the present
762 specification and does not need to implement compatibility mode as it
763 will never interact with legacy ECN tunnels.
765 Dynamic Discovery: This specification does not require or recommend
766 dynamic discovery and it does not define how dynamic negotiation
767 might be done, but it recognises that proprietary tunnel endpoint
768 discovery protocols exist. It therefore sets down some constraints
769 on discovery protocols to ensure safe interworking.
771 If dynamic tunnel endpoint discovery might pair an ingress with a
772 legacy egress (RFC2003, RFC2401 or RFC2481 or the limited
773 functionality mode of RFC3168), the ingress MUST implement both
774 normal and compatibility mode. If the tunnel discovery process is
775 arranged to only ever find a tunnel egress that propagates ECN
776 (RFC3168 full functionality mode, RFC4301 or this present
777 specification), then a tunnel ingress can be complaint with the
778 present specification without implementing compatibility mode.
780 If a compliant tunnel ingress is discovering an egress, it MUST send
781 packets in compatibility mode in case the egress it discovers is a
782 legacy egress. If, through the discovery protocol, the egress
783 indicates that it is compliant with the present specification, with
784 RFC4301 or with RFC3168 full functionality mode, the ingress can
785 switch itself into normal mode. If the egress denies compliance with
786 any of these or returns an error that implies it does not understand
787 a request to work to any of these ECN specifications, the tunnel
788 ingress MUST remain in compatibility mode.
790 An ingress cannot claim compliance with this specification simply by
791 disabling ECN processing across the tunnel (i.e. only implementing
792 compatibility mode). It is true that such a tunnel ingress is at
793 least safe with the ECN behaviour of any egress it may encounter, but
794 it does not meet the aim of introducing ECN support to tunnels.
796 Implementation note: if a compliant node is the ingress for multiple
797 tunnels, a mode setting will need to be stored for each tunnel
798 ingress. However, if a node is the egress for multiple tunnels, none
799 of the tunnels will need to store a mode setting, because a compliant
800 egress can only be in one mode.
802 4.4. Single Mode of Decapsulation
804 A compliant decapsulator only has one mode of operation. However, if
805 a complaint egress is implemented to be dynamically discoverable, it
806 may need to respond to discovery requests from various types of
807 legacy tunnel ingress. This specification does not define how
808 dynamic negotiation might be done by (proprietary) discovery
809 protocols, but it sets down some constraints to ensure safe
810 interworking.
812 Through the discovery protocol, a tunnel ingress compliant with the
813 present specification might ask if the egress is compliant with the
814 present specification, with RFC4301 or with RFC3168 full
815 functionality mode. Or an RFC3168 tunnel ingress might try to
816 negotiate to use limited functionality or full functionality mode
818 [RFC3168]. In all these cases, a decapsulating tunnel egress
819 compliant with this specification MUST agree to any of these
820 requests, since it will behave identically in all these cases.
822 If no ECN-related mode is requested, a compliant tunnel egress MUST
823 continue without raising any error or warning as its egress behaviour
824 is compatible with all the legacy ingress behaviours that do not
825 negotiate capabilities.
827 For 'forward compatibility', a compliant tunnel egress SHOULD raise a
828 warning alarm about any requests to enter modes it does not
829 recognise, but it SHOULD continue operating.
831 5. Updates to Earlier RFCs
833 5.1. Changes to RFC4301 ECN processing
835 Ingress: An RFC4301 IPsec encapsulator is not changed at all by the
836 present specification
838 Egress: The new decapsulation behaviour in Figure 4 updates RFC4301.
839 However, it solely updates combinations of inner and outer that
840 would never result from any protocol defined in the RFC series so
841 far, even though they were catered for in RFC4301 for
842 completeness. Therefore, the present specification adds new
843 behaviours to RFC4301 decapsulation without altering existing
844 behaviours. The following specific updates have been made:
846 * The outer, not the inner, is propagated when the outer is
847 ECT(1) and the inner is ECT(0);
849 * A packet with Not-ECT in the inner and an outer of CE is
850 dropped rather than forwarded as Not-ECT;
852 * Certain combinations of inner and outer ECN field have been
853 identified as currently unused. These can trigger logging
854 and/or raise alarms.
856 Modes: RFC4301 does not need modes and is not updated by the modes
857 in the present specification. The normal mode of encapsulation is
858 unchanged from RFC4301 encapsulation and an RFC4301 IPsec ingress
859 will never need compatibility mode as explained in Section 4.3
860 (except in one corner-case described below).
861 One corner case can exist where an RFC4301 ingress does not use
862 IKEv2, but uses manual keying instead. Then an RFC4301 ingress
863 could conceivably be configured to tunnel to an egress with
864 limited functionality ECN handling. Strictly, for this corner-
865 case, the requirement to use compatibility mode in this
866 specification updates RFC4301. However, this is such a remote
867 possibility that RFC4301 IPsec implementations are NOT REQUIRED to
868 implement compatibility mode.
870 5.2. Changes to RFC3168 ECN processing
872 Ingress: On encapsulation, the new rule in Figure 3 that a normal
873 mode tunnel ingress copies any ECN field into the outer header
874 updates the ingress behaviour of RFC3168. Nonetheless, the new
875 compatibility mode is identical to the limited functionality mode
876 of RFC3168.
878 Egress: The new decapsulation behaviour in Figure 4 updates RFC3168.
879 However, the present specification solely updates combinations of
880 inner and outer that would never result from any protocol defined
881 in the RFC series so far, even though they were catered for in
882 RFC4301 for completeness. Therefore, the present specification
883 adds new behaviours to RFC3168 decapsulation without altering
884 existing behaviours. The following specific updates have been
885 made:
887 * The outer, not the inner, is propagated when the outer is
888 ECT(1) and the inner is ECT(0);
890 * Certain combinations of inner and outer ECN field have been
891 identified as currently unused. These can trigger logging
892 and/or raise alarms.
894 Modes: RFC3168 defines a (required) limited functionality mode and
895 an (optional) full functionality mode for a tunnel. In RFC3168,
896 modes applied to both ends of the tunnel, while in the present
897 specification, modes are only used at the ingress--a single egress
898 behaviour covers all cases. The normal mode of encapsulation
899 updates the encapsulation behaviour of the full functionality mode
900 of RFC3168. The compatibility mode of encapsulation is identical
901 to the encapsulation behaviour of the limited functionality mode
902 of RFC3168. The constraints on how tunnel discovery protocols set
903 modes in Section 4.3 and Section 4.4 are an update to RFC3168.
905 5.3. Motivation for Changes
907 An overriding goal is to ensure the same ECN signals can mean the
908 same thing whatever tunnels happen to encapsulate an IP packet flow.
909 This removes gratuitous inconsistency, which otherwise constrains the
910 available design space and makes it harder to design networks and new
911 protocols that work predictably.
913 5.3.1. Motivation for Changing Encapsulation
915 The normal mode in Section 4 updates RFC3168 to make all IP in IP
916 encapsulation of the ECN field consistent--consistent with the way
917 both RFC4301 IPsec [RFC4301] and IP in MPLS or MPLS in MPLS
918 encapsulation [RFC5129] construct the ECN field.
920 Compatibility mode has also been defined so a non-RFC4301 ingress can
921 still switch to using drop across a tunnel for backwards
922 compatibility with legacy decapsulators that do not propagate ECN
923 correctly.
925 The trigger that motivated this update to RFC3168 encapsulation was a
926 standards track proposal for pre-congestion notification (PCN
927 [RFC5670]). PCN excess rate marking only works correctly if the ECN
928 field is copied on encapsulation (as in RFC4301 and RFC5129); it does
929 not work if ECN is reset (as in RFC3168). This is because PCN excess
930 rate marking depends on the outer header revealing any congestion
931 experienced so far on the whole path, not just since the last tunnel
932 ingress (see Appendix E for a full explanation).
934 PCN allows a network operator to add flow admission and termination
935 for inelastic traffic at the edges of a Diffserv domain, but without
936 any per-flow mechanisms in the interior and without the generous
937 provisioning typical of Diffserv, aiming to significantly reduce
938 costs. The PCN architecture [RFC5559] states that RFC3168 IP in IP
939 tunnelling of the ECN field cannot be used for any tunnel ingress in
940 a PCN domain. Prior to the present specification, this left a stark
941 choice between not being able to use PCN for inelastic traffic
942 control or not being able to use the many tunnels already deployed
943 for Mobile IP, VPNs and so forth.
945 The present specification provides a clean solution to this problem,
946 so that network operators who want to use both PCN and tunnels can
947 specify that every tunnel ingress in a PCN region must comply with
948 this latest specification.
950 Rather than allow tunnel specifications to fragment further into one
951 for PCN, one for IPsec and one for other tunnels, the opportunity has
952 been taken to consolidate the diverging specifications back into a
953 single tunnelling behaviour. Resetting ECN was originally motivated
954 by a covert channel concern that has been deliberately set aside in
955 RFC4301 IPsec. Therefore the reset behaviour of RFC3168 is an
956 anomaly that we do not need to keep. Copying ECN on encapsulation is
957 anyway simpler than resetting. So, as more tunnel endpoints comply
958 with this single consistent specification, encapsulation will be
959 simpler as well as more predictable.
961 Appendix B assesses whether copying rather than resetting CE on
962 ingress will cause any unintended side-effects, from the three
963 perspectives of security, control and management. In summary this
964 analysis finds that:
966 o From the control perspective either copying or resetting works for
967 existing arrangements, but copying has more potential for
968 simplifying control and resetting breaks at least one proposal
969 already on the standards track.
971 o From the management and monitoring perspective copying is
972 preferable.
974 o From the traffic security perspective (enforcing congestion
975 control, mitigating denial of service etc) copying is preferable.
977 o From the information security perspective resetting is preferable,
978 but the IETF Security Area now considers copying acceptable given
979 the bandwidth of a 2-bit covert channel can be managed.
981 Therefore there are two points against resetting CE on ingress while
982 copying CE causes no significant harm.
984 5.3.2. Motivation for Changing Decapsulation
986 The specification for decapsulation in Section 4 fixes three problems
987 with the pre-existing behaviours of both RFC3168 and RFC4301:
989 1. The pre-existing rules prevented the introduction of alternate
990 ECN semantics to signal more than one severity level of
991 congestion [RFC4774], [RFC5559]. The four states of the 2-bit
992 ECN field provide room for signalling two severity levels in
993 addition to not-congested and not-ECN-capable states. But, the
994 pre-existing rules assumed that two of the states (ECT(0) and
995 ECT(1)) are always equivalent. This unnecessarily restricts the
996 use of one of four codepoints (half a bit) in the IP (v4 & v6)
997 header. The new rules are designed to work in either case;
998 whether ECT(1) is more severe than or equivalent to ECT(0).
1000 As explained in Appendix B.1, the original reason for not
1001 forwarding the outer ECT codepoints was to limit the covert
1002 channel across a decapsulator to 1 bit per packet. However, now
1003 that the IETF Security Area has deemed that a 2-bit covert
1004 channel through an encapsulator is a manageable risk, the same
1005 should be true for a decapsulator.
1007 As well as being useful for general future-proofing, this problem
1008 is immediately pressing for standardisation of pre-congestion
1009 notification (PCN), which uses two severity levels of congestion.
1010 If a congested queue used ECT(1) in the outer header to signal
1011 more severe congestion than ECT(0), the pre-existing
1012 decapsulation rules would have thrown away this congestion
1013 signal, preventing tunnelled traffic from ever knowing that it
1014 should reduce its load.
1016 The PCN working group has had to consider a number of wasteful or
1017 convoluted work-rounds to this problem (see Appendix D). But by
1018 far the simplest approach is just to remove the covert channel
1019 blockages from tunnelling behaviour--now deemed unnecessary
1020 anyway. Then network operators that want to support two
1021 congestion severity-levels for PCN can specify that every tunnel
1022 egress in a PCN region must comply with this latest
1023 specification.
1025 Not only does this make two congestion severity-levels available
1026 for PCN standardisation, but also for other potential uses of the
1027 extra ECN codepoint (e.g. [VCP]).
1029 2. Cases are documented where a middlebox (e.g. a firewall) drops
1030 packets with header values that were currently unused (CU) when
1031 the box was deployed, often on the grounds that anything
1032 unexpected might be an attack. This tends to bar future use of
1033 CU values. The new decapsulation rules specify optional logging
1034 and/or alarms for specific combinations of inner and outer header
1035 that are currently unused. The aim is to give implementers a
1036 recourse other than drop if they are concerned about the security
1037 of CU values. It recognises legitimate security concerns about
1038 CU values but still eases their future use. If the alarms are
1039 interpreted as an attack (e.g. by a management system) the
1040 offending packets can be dropped. But alarms can be turned off
1041 if these combinations come into regular use (e.g. a through a
1042 future standards action).
1044 3. While reviewing currently unused combinations of inner and outer,
1045 the opportunity was taken to define a single consistent behaviour
1046 for the three cases with a Not-ECT inner header but a different
1047 outer. RFC3168 and RFC4301 had diverged in this respect. None
1048 of these combinations should result from Internet protocols in
1049 the RFC series, but future standards actions might put any or all
1050 of them to good use. Therefore it was decided that a
1051 decapsulator must forward a Not-ECT inner unchanged, even if the
1052 arriving outer was ECT(0) or ECT(1). But for safety it should
1053 drop the Not-ECT inner if the arriving outer was CE. Then, if
1054 some unfortunate misconfiguration resulted in a congested router
1055 marking CE on a packet that was originally Not-ECT, drop would be
1056 the only appropriate signal for the egress to propagate--the only
1057 signal a non-ECN-capable transport (Not-ECT) would understand.
1059 ECT(1) is being proposed as an intermediate level of congestion
1060 in a scheme progressing through the IETF
1061 [I-D.ietf-pcn-3-in-1-encoding]. But it was decided that it would
1062 still be safe to mandate forwarding as Not-ECT for a Not-ECT
1063 inner with an ECT(1) outer, thus keeping this combination
1064 available for future use. The rationale was as follows: if any
1065 misconfiguration led to ECT(1) congestion signals with a Not-ECT
1066 inner, it would be safe for the egress to suppress these signals.
1067 This is because the congestion would then escalate to CE marking,
1068 which the egress would drop, thus avoiding any risk of congestion
1069 collapse.
1071 Problems 2 & 3 alone would not warrant a change to decapsulation, but
1072 it was decided they are worth fixing and making consistent at the
1073 same time as decapsulation code is changed to fix problem 1 (two
1074 congestion severity-levels).
1076 6. Backward Compatibility
1078 A tunnel endpoint compliant with the present specification is
1079 backward compatible when paired with any tunnel endpoint compliant
1080 with any previous tunnelling RFC, whether RFC4301, RFC3168 (see
1081 Section 3) or the earlier RFCs summarised in Appendix A (RFC2481,
1082 RFC2401 and RFC2003). Each case is enumerated below.
1084 6.1. Non-Issues Updating Decapsulation
1086 At the egress, this specification only augments the per-packet
1087 calculation of the ECN field (RFC3168 and RFC4301) for combinations
1088 of inner and outer headers that have so far not been used in any IETF
1089 protocols.
1091 Therefore, all other things being equal, if an RFC4301 IPsec egress
1092 is updated to comply with the new rules, it will still interwork with
1093 any RFC4301 compliant ingress and the packet outputs will be
1094 identical to those it would have output before (fully backward
1095 compatible).
1097 And, all other things being equal, if an RFC3168 egress is updated to
1098 comply with the same new rules, it will still interwork with any
1099 ingress complying with any previous specification (both modes of
1100 RFC3168, both modes of RFC2481, RFC2401 and RFC2003) and the packet
1101 outputs will be identical to those it would have output before (fully
1102 backward compatible).
1104 A compliant tunnel egress merely needs to implement the one behaviour
1105 in Section 4 with no additional mode or option configuration at the
1106 ingress or egress nor any additional negotiation with the ingress.
1107 The new decapsulation rules have been defined in such a way that
1108 congestion control will still work safely if any of the earlier
1109 versions of ECN processing are used unilaterally at the encapsulating
1110 ingress of the tunnel (any of RFC2003, RFC2401, either mode of
1111 RFC2481, either mode of RFC3168, RFC4301 and this present
1112 specification).
1114 6.2. Non-Update of RFC4301 IPsec Encapsulation
1116 An RFC4301 IPsec ingress can comply with this new specification
1117 without any update and it has no need for any new modes, options or
1118 configuration. So, all other things being equal, it will continue to
1119 interwork identically with any egress it worked with before (fully
1120 backward compatible).
1122 6.3. Update to RFC3168 Encapsulation
1124 The encapsulation behaviour of the new normal mode copies the ECN
1125 field whereas RFC3168 full functionality mode reset it. However, all
1126 other things being equal, if RFC3168 ingress is updated to the
1127 present specification, the outgoing packets from any tunnel egress
1128 will still be unchanged. This is because all variants of tunnelling
1129 at either end (RFC4301, both modes of RFC3168, both modes of RFC2481,
1130 RFC2401, RFC2003 and the present specification) have always
1131 propagated an incoming CE marking through the inner header and onward
1132 into the outgoing header, whether the outer header is reset or
1133 copied. Therefore, If the tunnel is considered as a black box, the
1134 packets output from any egress will be identical with or without an
1135 update to the ingress. Nonetheless, if packets are observed within
1136 the black box (between the tunnel endpoints), CE markings copied by
1137 the updated ingress will be visible within the black box, whereas
1138 they would not have been before. Therefore, the update to
1139 encapsulation can be termed 'black-box backwards compatible' (i.e.
1140 identical unless you look inside the tunnel).
1142 This specification introduces no new backward compatibility issues
1143 when a compliant ingress talks with a legacy egress, but it has to
1144 provide similar safeguards to those already defined in RFC3168.
1145 RFC3168 laid down rules to ensure that an RFC3168 ingress turns off
1146 ECN (limited functionality mode) if it is paired with a legacy egress
1147 (RFC 2481, RFC2401 or RFC2003), which would not propagate ECN
1148 correctly. The present specification carries forward those rules
1149 (Section 4.3). It uses compatibility mode whenever RFC3168 would
1150 have used limited functionality mode, and their per-packet behaviours
1151 are identical. Therefore, all other things being equal, an ingress
1152 using the new rules will interwork with any legacy tunnel egress in
1153 exactly the same way as an RFC3168 ingress (still black-box backward
1154 compatible).
1156 7. Design Principles for Alternate ECN Tunnelling Semantics
1158 This section is informative not normative.
1160 S.5 of RFC3168 permits the Diffserv codepoint (DSCP)[RFC2474] to
1161 'switch in' alternative behaviours for marking the ECN field, just as
1162 it switches in different per-hop behaviours (PHBs) for scheduling.
1163 [RFC4774] gives best current practice for designing such alternative
1164 ECN semantics and very briefly mentions in section 5.4 that
1165 tunnelling should be considered. The guidance below extends RFC4774,
1166 giving additional guidance on designing any alternate ECN semantics
1167 that would also require alternate tunnelling semantics.
1169 The overriding guidance is: "Avoid designing alternate ECN tunnelling
1170 semantics, if at all possible." If a scheme requires tunnels to
1171 implement special processing of the ECN field for certain DSCPs, it
1172 will be hard to guarantee that every implementer of every tunnel will
1173 have added the required exception or that operators will have
1174 ubiquitously deployed the required updates. It is unlikely a single
1175 authority is even aware of all the tunnels in a network, which may
1176 include tunnels set up by applications between endpoints, or
1177 dynamically created in the network. Therefore it is highly likely
1178 that some tunnels within a network or on hosts connected to it will
1179 not implement the required special case.
1181 That said, if a non-default scheme for tunnelling the ECN field is
1182 really required, the following guidelines may prove useful in its
1183 design:
1185 On encapsulation in any new scheme:
1187 1. The ECN field of the outer header should be cleared to Not-ECT
1188 ("00") unless it is guaranteed that the corresponding tunnel
1189 egress will correctly propagate congestion markings introduced
1190 across the tunnel in the outer header.
1192 2. If it has established that ECN will be correctly propagated,
1193 an encapsulator should also copy incoming congestion
1194 notification into the outer header. The general principle
1195 here is that the outer header should reflect congestion
1196 accumulated along the whole upstream path, not just since the
1197 tunnel ingress (Appendix B.3 on management and monitoring
1198 explains).
1200 In some circumstances (e.g. pseudowires, PCN), the whole path
1201 is divided into segments, each with its own congestion
1202 notification and feedback loop. In these cases, the function
1203 that regulates load at the start of each segment will need to
1204 reset congestion notification for its segment. Often the
1205 point where congestion notification is reset will also be
1206 located at the start of a tunnel. However, the resetting
1207 function should be thought of as being applied to packets
1208 after the encapsulation function--two logically separate
1209 functions even though they might run on the same physical box.
1210 Then the code module doing encapsulation can keep to the
1211 copying rule and the load regulator module can reset
1212 congestion, without any code in either module being
1213 conditional on whether the other is there.
1215 On decapsulation in any new scheme:
1217 1. If the arriving inner header is Not-ECT it implies the
1218 transport will not understand other ECN codepoints. If the
1219 outer header carries an explicit congestion marking, the
1220 alternate scheme will probably need to drop the packet--the
1221 only indication of congestion the transport will understand.
1222 If the outer carries any other ECN codepoint that does not
1223 indicate congestion, the alternate scheme can forward the
1224 packet, but probably only as Not-ECT.
1226 2. If the arriving inner header is other than Not-ECT, the ECN
1227 field that the alternate decapsulation scheme forwards should
1228 reflect the more severe congestion marking of the arriving
1229 inner and outer headers.
1231 3. Any alternate scheme MUST define a behaviour for all
1232 combinations of inner and outer headers, even those that would
1233 not be expected to result from standards known at the time or
1234 from the expected behaviour of the tunnel ingress paired with
1235 the egress at run-time. Consideration should be given to
1236 logging such unexpected combinations and raising an alarm,
1237 particularly if there is a danger that the invalid combination
1238 implies congestion signals are not being propagated correctly.
1239 The presence of currently unused combinations may represent an
1240 attack, but the new scheme should try to define a way to
1241 forward such packets, at least if a safe outgoing codepoint
1242 can be defined. Raising an alarm to warn of the possibility
1243 of an attack is a preferable approach to dropping that ensures
1244 these combinations can be usable in future standards actions.
1246 IANA Considerations (to be removed on publication):
1248 This memo includes no request to IANA.
1250 8. Security Considerations
1252 Appendix B.1 discusses the security constraints imposed on ECN tunnel
1253 processing. The new rules for ECN tunnel processing (Section 4)
1254 trade-off between information security (covert channels) and
1255 congestion monitoring & control. In fact, ensuring congestion
1256 markings are not lost is itself another aspect of security, because
1257 if we allowed congestion notification to be lost, any attempt to
1258 enforce a response to congestion would be much harder.
1260 Specialist security issues:
1262 Tunnels intersecting Diffserv regions with alternate ECN semantics:
1263 If alternate congestion notification semantics are defined for a
1264 certain Diffserv PHB, the scope of the alternate semantics might
1265 typically be bounded by the limits of a Diffserv region or
1266 regions, as envisaged in [RFC4774] (e.g. the pre-congestion
1267 notification architecture [RFC5559]). The inner headers in
1268 tunnels crossing the boundary of such a Diffserv region but ending
1269 within the region can potentially leak the external congestion
1270 notification semantics into the region, or leak the internal
1271 semantics out of the region. [RFC2983] discusses the need for
1272 Diffserv traffic conditioning to be applied at these tunnel
1273 endpoints as if they are at the edge of the Diffserv region.
1274 Similar concerns apply to any processing or propagation of the ECN
1275 field at the edges of a Diffserv region with alternate ECN
1276 semantics. Such edge processing must also be applied at the
1277 endpoints of tunnels with one end inside and the other outside the
1278 domain. [RFC5559] gives specific advice on this for the PCN case,
1279 but other definitions of alternate semantics will need to discuss
1280 the specific security implications in each case.
1282 ECN nonce tunnel coverage: The new decapsulation rules improve the
1283 coverage of the ECN nonce [RFC3540] relative to the previous rules
1284 in RFC3168 and RFC4301. However, nonce coverage is still not
1285 perfect, as this would have led to a safety problem in another
1286 case. Both are corner-cases, so discussion of the compromise
1287 between them is deferred to Appendix F.
1289 Covert channel not turned off: A legacy (RFC3168) tunnel ingress
1290 could ask an RFC3168 egress to turn off ECN processing as well as
1291 itself turning off ECN. An egress compliant with the present
1292 specification will agree to such a request from a legacy ingress,
1293 but it relies on the ingress solely sending Not-ECT in the outer.
1295 If the egress receives other ECN codepoints in the outer it will
1296 process them as normal, so it will actually still copy congestion
1297 markings from the outer to the outgoing header. Referring for
1298 example to Figure 5 (Appendix B.1), although the tunnel ingress
1299 'I' will set all ECN fields in outer headers to Not-ECT, 'M' could
1300 still toggle CE or ECT(1) on and off to communicate covertly with
1301 'B', because we have specified that 'E' only has one mode
1302 regardless of what mode it says it has negotiated. We could have
1303 specified that 'E' should have a limited functionality mode and
1304 check for such behaviour. But we decided not to add the extra
1305 complexity of two modes on a compliant tunnel egress merely to
1306 cater for an historic security concern that is now considered
1307 manageable.
1309 9. Conclusions
1311 This document uses previously unused combinations of inner and outer
1312 header to augment the rules for calculating the ECN field when
1313 decapsulating IP packets at the egress of IPsec (RFC4301) and non-
1314 IPsec (RFC3168) tunnels. In this way it allows tunnels to propagate
1315 an extra level of congestion severity.
1317 This document also updates the ingress tunnelling encapsulation of
1318 RFC3168 ECN to bring all IP in IP tunnels into line with the new
1319 behaviour in the IPsec architecture of RFC4301, which copies rather
1320 than resets the ECN field when creating outer headers.
1322 The need for both these updated behaviours was triggered by the
1323 introduction of pre-congestion notification (PCN) onto the IETF
1324 standards track. Operators wanting to support PCN or other alternate
1325 ECN schemes that use an extra severity level can require that their
1326 tunnels comply with the present specification. Nonetheless, as part
1327 of general code maintenance, any tunnel can safely be updated to
1328 comply with this specification, because it is backward compatible
1329 with all previous tunnelling behaviours which will continue to work
1330 as before--just using one severity level.
1332 The new rules propagate changes to the ECN field across tunnel end-
1333 points that previously blocked them to restrict the bandwidth of a
1334 potential covert channel. Limiting the channel's bandwidth to 2 bits
1335 per packet is now considered sufficient.
1337 At the same time as removing these legacy constraints, the
1338 opportunity has been taken to draw together diverging tunnel
1339 specifications into a single consistent behaviour. Then any tunnel
1340 can be deployed unilaterally, and it will support the full range of
1341 congestion control and management schemes without any modes or
1342 configuration. Further, any host or router can expect the ECN field
1343 to behave in the same way, whatever type of tunnel might intervene in
1344 the path. This new certainty could enable new uses of the ECN field
1345 that would otherwise be confounded by ambiguity.
1347 10. Acknowledgements
1349 Thanks to Anil Agawaal for pointing out a case where it's safe for a
1350 tunnel decapsulator to forward a combination of headers it does not
1351 understand. Thanks to David Black for explaining a better way to
1352 think about function placement. Also thanks to Arnaud Jacquet for
1353 the idea for Appendix C. Thanks to Michael Menth, Bruce Davie, Toby
1354 Moncaster, Gorry Fairhurst, Sally Floyd, Alfred Hoenes, Gabriele
1355 Corliano, Ingemar Johansson, David Black and Phil Eardley for their
1356 thoughts and careful review comments.
1358 Bob Briscoe is partly funded by Trilogy, a research project (ICT-
1359 216372) supported by the European Community under its Seventh
1360 Framework Programme. The views expressed here are those of the
1361 author only.
1363 Comments Solicited (to be removed by the RFC Editor):
1365 Comments and questions are encouraged and very welcome. They can be
1366 addressed to the IETF Transport Area working group mailing list
1367 , and/or to the authors.
1369 11. References
1371 11.1. Normative References
1373 [RFC2003] Perkins, C., "IP Encapsulation
1374 within IP", RFC 2003, October 1996.
1376 [RFC2119] Bradner, S., "Key words for use in
1377 RFCs to Indicate Requirement
1378 Levels", BCP 14, RFC 2119,
1379 March 1997.
1381 [RFC3168] Ramakrishnan, K., Floyd, S., and D.
1382 Black, "The Addition of Explicit
1383 Congestion Notification (ECN) to
1384 IP", RFC 3168, September 2001.
1386 [RFC4301] Kent, S. and K. Seo, "Security
1387 Architecture for the Internet
1388 Protocol", RFC 4301, December 2005.
1390 11.2. Informative References
1392 [I-D.ietf-pcn-3-in-1-encoding] Briscoe, B. and T. Moncaster, "PCN
1393 3-State Encoding Extension in a
1394 single DSCP",
1395 draft-ietf-pcn-3-in-1-encoding-00
1396 (work in progress), July 2009.
1398 [I-D.ietf-pcn-3-state-encoding] Moncaster, T., Briscoe, B., and M.
1399 Menth, "A PCN encoding using 2
1400 DSCPs to provide 3 or more states",
1401 draft-ietf-pcn-3-state-encoding-00
1402 (work in progress), April 2009.
1404 [I-D.ietf-pcn-psdm-encoding] Menth, M., Babiarz, J., Moncaster,
1405 T., and B. Briscoe, "PCN Encoding
1406 for Packet-Specific Dual Marking
1407 (PSDM)",
1408 draft-ietf-pcn-psdm-encoding-00
1409 (work in progress), June 2009.
1411 [I-D.ietf-pcn-sm-edge-behaviour] Charny, A., Karagiannis, G., Menth,
1412 M., and T. Taylor, "PCN Boundary
1413 Node Behaviour for the Single
1414 Marking (SM) Mode of Operation",
1415 draft-ietf-pcn-sm-edge-behaviour-01
1416 (work in progress), October 2009.
1418 [I-D.satoh-pcn-st-marking] Satoh, D., Ueno, H., Maeda, Y., and
1419 O. Phanachet, "Single PCN Threshold
1420 Marking by using PCN baseline
1421 encoding for both admission and
1422 termination controls",
1423 draft-satoh-pcn-st-marking-02 (work
1424 in progress), September 2009.
1426 [RFC2401] Kent, S. and R. Atkinson, "Security
1427 Architecture for the Internet
1428 Protocol", RFC 2401, November 1998.
1430 [RFC2474] Nichols, K., Blake, S., Baker, F.,
1431 and D. Black, "Definition of the
1432 Differentiated Services Field (DS
1433 Field) in the IPv4 and IPv6
1434 Headers", RFC 2474, December 1998.
1436 [RFC2481] Ramakrishnan, K. and S. Floyd, "A
1437 Proposal to add Explicit Congestion
1438 Notification (ECN) to IP",
1439 RFC 2481, January 1999.
1441 [RFC2983] Black, D., "Differentiated Services
1442 and Tunnels", RFC 2983,
1443 October 2000.
1445 [RFC3540] Spring, N., Wetherall, D., and D.
1446 Ely, "Robust Explicit Congestion
1447 Notification (ECN) Signaling with
1448 Nonces", RFC 3540, June 2003.
1450 [RFC4306] Kaufman, C., "Internet Key Exchange
1451 (IKEv2) Protocol", RFC 4306,
1452 December 2005.
1454 [RFC4774] Floyd, S., "Specifying Alternate
1455 Semantics for the Explicit
1456 Congestion Notification (ECN)
1457 Field", BCP 124, RFC 4774,
1458 November 2006.
1460 [RFC5129] Davie, B., Briscoe, B., and J. Tay,
1461 "Explicit Congestion Marking in
1462 MPLS", RFC 5129, January 2008.
1464 [RFC5559] Eardley, P., "Pre-Congestion
1465 Notification (PCN) Architecture",
1466 RFC 5559, June 2009.
1468 [RFC5670] Eardley, P., "Metering and Marking
1469 Behaviour of PCN-Nodes", RFC 5670,
1470 November 2009.
1472 [RFC5696] Moncaster, T., Briscoe, B., and M.
1473 Menth, "Baseline Encoding and
1474 Transport of Pre-Congestion
1475 Information", RFC 5696,
1476 November 2009.
1478 [VCP] Xia, Y., Subramanian, L., Stoica,
1479 I., and S. Kalyanaraman, "One more
1480 bit is enough", Proc. SIGCOMM'05,
1481 ACM CCR 35(4)37--48, 2005, .
1485 Appendix A. Early ECN Tunnelling RFCs
1487 IP in IP tunnelling was originally defined in [RFC2003]. On
1488 encapsulation, the incoming header was copied to the outer and on
1489 decapsulation the outer was simply discarded. Initially, IPsec
1490 tunnelling [RFC2401] followed the same behaviour.
1492 When ECN was introduced experimentally in [RFC2481], legacy (RFC2003
1493 or RFC2401) tunnels would have discarded any congestion markings
1494 added to the outer header, so RFC2481 introduced rules for
1495 calculating the outgoing header from a combination of the inner and
1496 outer on decapsulation. RC2481 also introduced a second mode for
1497 IPsec tunnels, which turned off ECN processing(Not-ECT) in the outer
1498 header on encapsulation because an RFC2401 decapsulator would discard
1499 the outer on decapsulation. For RFC2401 IPsec this had the side-
1500 effect of completely blocking the covert channel.
1502 In RFC2481 the ECN field was defined as two separate bits. But when
1503 ECN moved from the experimental to the standards track [RFC3168], the
1504 ECN field was redefined as four codepoints. This required a
1505 different calculation of the ECN field from that used in RFC2481 on
1506 decapsulation. RFC3168 also had two modes; a 'full functionality
1507 mode' that restricted the covert channel as much as possible but
1508 still allowed ECN to be used with IPsec, and another that completely
1509 turned off ECN processing across the tunnel. This 'limited
1510 functionality mode' both offered a way for operators to completely
1511 block the covert channel and allowed an RFC3168 ingress to interwork
1512 with a legacy tunnel egress (RFC2481, RFC2401 or RFC2003).
1514 The present specification includes a similar compatibility mode to
1515 interwork safely with tunnels compliant with any of these three
1516 earlier RFCs. However, unlike RFC3168, it is only a mode of the
1517 ingress, as decapsulation behaviour is the same in either case.
1519 Appendix B. Design Constraints
1521 Tunnel processing of a congestion notification field has to meet
1522 congestion control and management needs without creating new
1523 information security vulnerabilities (if information security is
1524 required). This appendix documents the analysis of the tradeoffs
1525 between these factors that led to the new encapsulation rules in
1526 Section 4.1.
1528 B.1. Security Constraints
1530 Information security can be assured by using various end to end
1531 security solutions (including IPsec in transport mode [RFC4301]), but
1532 a commonly used scenario involves the need to communicate between two
1533 physically protected domains across the public Internet. In this
1534 case there are certain management advantages to using IPsec in tunnel
1535 mode solely across the publicly accessible part of the path. The
1536 path followed by a packet then crosses security 'domains'; the ones
1537 protected by physical or other means before and after the tunnel and
1538 the one protected by an IPsec tunnel across the otherwise unprotected
1539 domain. We will use the scenario in Figure 5 where endpoints 'A' and
1540 'B' communicate through a tunnel. The tunnel ingress 'I' and egress
1541 'E' are within physically protected edge domains, while the tunnel
1542 spans an unprotected internetwork where there may be 'men in the
1543 middle', M.
1545 physically unprotected physically
1546 <-protected domain-><--domain--><-protected domain->
1547 +------------------+ +------------------+
1548 | | M | |
1549 | A-------->I=========>==========>E-------->B |
1550 | | | |
1551 +------------------+ +------------------+
1552 <----IPsec secured---->
1553 tunnel
1555 Figure 5: IPsec Tunnel Scenario
1557 IPsec encryption is typically used to prevent 'M' seeing messages
1558 from 'A' to 'B'. IPsec authentication is used to prevent 'M'
1559 masquerading as the sender of messages from 'A' to 'B' or altering
1560 their contents. But 'I' can also use IPsec tunnel mode to allow 'A'
1561 to communicate with 'B', but impose encryption to prevent 'A' leaking
1562 information to 'M'. Or 'E' can insist that 'I' uses tunnel mode
1563 authentication to prevent 'M' communicating information to 'B'.
1564 Mutable IP header fields such as the ECN field (as well as the TTL/
1565 Hop Limit and DS fields) cannot be included in the cryptographic
1566 calculations of IPsec. Therefore, if 'I' copies these mutable fields
1567 into the outer header that is exposed across the tunnel it will have
1568 allowed a covert channel from 'A' to M that bypasses its encryption
1569 of the inner header. And if 'E' copies these fields from the outer
1570 header to the inner, even if it validates authentication from 'I', it
1571 will have allowed a covert channel from 'M' to 'B'.
1573 ECN at the IP layer is designed to carry information about congestion
1574 from a congested resource towards downstream nodes. Typically a
1575 downstream transport might feed the information back somehow to the
1576 point upstream of the congestion that can regulate the load on the
1577 congested resource, but other actions are possible (see [RFC3168]
1578 S.6). In terms of the above unicast scenario, ECN effectively
1579 intends to create an information channel (for congestion signalling)
1580 from 'M' to 'B' (for 'B' to feed back to 'A'). Therefore the goals
1581 of IPsec and ECN are mutually incompatible, requiring some
1582 compromise.
1584 With respect to the DS or ECN fields, S.5.1.2 of RFC4301 says,
1585 "controls are provided to manage the bandwidth of this [covert]
1586 channel". Using the ECN processing rules of RFC4301, the channel
1587 bandwidth is two bits per datagram from 'A' to 'M' and one bit per
1588 datagram from 'M' to 'A' (because 'E' limits the combinations of the
1589 2-bit ECN field that it will copy). In both cases the covert channel
1590 bandwidth is further reduced by noise from any real congestion
1591 marking. RFC4301 implies that these covert channels are sufficiently
1592 limited to be considered a manageable threat. However, with respect
1593 to the larger (6b) DS field, the same section of RFC4301 says not
1594 copying is the default, but a configuration option can allow copying
1595 "to allow a local administrator to decide whether the covert channel
1596 provided by copying these bits outweighs the benefits of copying".
1597 Of course, an administrator considering copying of the DS field has
1598 to take into account that it could be concatenated with the ECN field
1599 giving an 8b per datagram covert channel.
1601 For tunnelling the 6b Diffserv field two conceptual models have had
1602 to be defined so that administrators can trade off security against
1603 the needs of traffic conditioning [RFC2983]:
1605 The uniform model: where the Diffserv field is preserved end-to-end
1606 by copying into the outer header on encapsulation and copying from
1607 the outer header on decapsulation.
1609 The pipe model: where the outer header is independent of that in the
1610 inner header so it hides the Diffserv field of the inner header
1611 from any interaction with nodes along the tunnel.
1613 However, for ECN, the new IPsec security architecture in RFC4301 only
1614 standardised one tunnelling model equivalent to the uniform model.
1615 It deemed that simplicity was more important than allowing
1616 administrators the option of a tiny increment in security, especially
1617 given not copying congestion indications could seriously harm
1618 everyone's network service.
1620 B.2. Control Constraints
1622 Congestion control requires that any congestion notification marked
1623 into packets by a resource will be able to traverse a feedback loop
1624 back to a function capable of controlling the load on that resource.
1625 To be precise, rather than calling this function the data source, we
1626 will call it the Load Regulator. This will allow us to deal with
1627 exceptional cases where load is not regulated by the data source, but
1628 usually the two terms will be synonymous. Note the term "a function
1629 _capable of_ controlling the load" deliberately includes a source
1630 application that doesn't actually control the load but ought to (e.g.
1631 an application without congestion control that uses UDP).
1633 A--->R--->I=========>M=========>E-------->B
1635 Figure 6: Simple Tunnel Scenario
1637 We now consider a similar tunnelling scenario to the IPsec one just
1638 described, but without the different security domains so we can just
1639 focus on ensuring the control loop and management monitoring can work
1640 (Figure 6). If we want resources in the tunnel to be able to
1641 explicitly notify congestion and the feedback path is from 'B' to
1642 'A', it will certainly be necessary for 'E' to copy any CE marking
1643 from the outer header to the inner header for onward transmission to
1644 'B', otherwise congestion notification from resources like 'M' cannot
1645 be fed back to the Load Regulator ('A'). But it does not seem
1646 necessary for 'I' to copy CE markings from the inner to the outer
1647 header. For instance, if resource 'R' is congested, it can send
1648 congestion information to 'B' using the congestion field in the inner
1649 header without 'I' copying the congestion field into the outer header
1650 and 'E' copying it back to the inner header. 'E' can still write any
1651 additional congestion marking introduced across the tunnel into the
1652 congestion field of the inner header.
1654 It might be useful for the tunnel egress to be able to tell whether
1655 congestion occurred across a tunnel or upstream of it. If outer
1656 header congestion marking was reset by the tunnel ingress ('I'), at
1657 the end of a tunnel ('E') the outer headers would indicate congestion
1658 experienced across the tunnel ('I' to 'E'), while the inner header
1659 would indicate congestion upstream of 'I'. But similar information
1660 can be gleaned even if the tunnel ingress copies the inner to the
1661 outer headers. At the end of the tunnel ('E'), any packet with an
1662 _extra_ mark in the outer header relative to the inner header
1663 indicates congestion across the tunnel ('I' to 'E'), while the inner
1664 header would still indicate congestion upstream of ('I'). Appendix C
1665 gives a simple and precise method for a tunnel egress to infer the
1666 congestion level introduced across a tunnel.
1668 All this shows that 'E' can preserve the control loop irrespective of
1669 whether 'I' copies congestion notification into the outer header or
1670 resets it.
1672 That is the situation for existing control arrangements but, because
1673 copying reveals more information, it would open up possibilities for
1674 better control system designs. For instance, Appendix E describes
1675 how resetting CE marking on encapsulation breaks a proposed
1676 congestion marking scheme on the standards track. It ends up
1677 removing excessive amounts of traffic unnecessarily. Whereas copying
1678 CE markings at ingress leads to the correct control behaviour.
1680 B.3. Management Constraints
1682 As well as control, there are also management constraints.
1683 Specifically, a management system may monitor congestion markings in
1684 passing packets, perhaps at the border between networks as part of a
1685 service level agreement. For instance, monitors at the borders of
1686 autonomous systems may need to measure how much congestion has
1687 accumulated so far along the path, perhaps to determine between them
1688 how much of the congestion is contributed by each domain.
1690 In this document we define the baseline of congestion marking (or the
1691 Congestion Baseline) as the source of the layer that created (or most
1692 recently reset) the congestion notification field. When monitoring
1693 congestion it would be desirable if the Congestion Baseline did not
1694 depend on whether packets were tunnelled or not. Given some tunnels
1695 cross domain borders (e.g. consider M in Figure 6 is monitoring a
1696 border), it would therefore be desirable for 'I' to copy congestion
1697 accumulated so far into the outer headers, so that it is exposed
1698 across the tunnel.
1700 Appendix C. Contribution to Congestion across a Tunnel
1702 This specification mandates that a tunnel ingress determines the ECN
1703 field of each new outer tunnel header by copying the arriving header.
1704 Concern has been expressed that this will make it difficult for the
1705 tunnel egress to monitor congestion introduced only along a tunnel,
1706 which is easy if the outer ECN field is reset at a tunnel ingress
1707 (RFC3168 full functionality mode). However, in fact copying CE marks
1708 at ingress will still make it easy for the egress to measure
1709 congestion introduced across a tunnel, as illustrated below.
1711 Consider 100 packets measured at the egress. Say it measures that 30
1712 are CE marked in the inner and outer headers and 12 have additional
1713 CE marks in the outer but not the inner. This means packets arriving
1714 at the ingress had already experienced 30% congestion. However, it
1715 does not mean there was 12% congestion across the tunnel. The
1716 correct calculation of congestion across the tunnel is p_t = 12/
1717 (100-30) = 12/70 = 17%. This is easy for the egress to measure. It
1718 is simply the packets with additional CE marking in the outer header
1719 (12) as a proportion of packets not marked in the inner header (70).
1721 Figure 7 illustrates this in a combinatorial probability diagram.
1722 The square represents 100 packets. The 30% division along the bottom
1723 represents marking before the ingress, and the p_t division up the
1724 side represents marking introduced across the tunnel.
1726 ^ outer header marking
1727 |
1728 100% +-----+---------+ The large square
1729 | | | represents 100 packets
1730 | 30 | |
1731 | | | p_t = 12/(100-30)
1732 p_t + +---------+ = 12/70
1733 | | 12 | = 17%
1734 0 +-----+---------+--->
1735 0 30% 100% inner header marking
1737 Figure 7: Tunnel Marking of Packets Already Marked at Ingress
1739 Appendix D. Why Losing ECT(1) on Decapsulation Impedes PCN
1741 Congestion notification with two severity levels is currently on the
1742 IETF's standards track agenda in the Congestion and Pre-Congestion
1743 Notification (PCN) working group. PCN needs all four possible states
1744 of congestion signalling in the 2-bit ECN field to be propagated at
1745 the egress, but pre-existing tunnels only propagate three. The four
1746 PCN states are: not PCN-enabled, not marked and two increasingly
1747 severe levels of congestion marking. The less severe marking means
1748 'stop admitting new traffic' and the more severe marking means
1749 'terminate some existing flows', which may be needed after reroutes
1750 (see [RFC5559] for more details). (Note on terminology: wherever
1751 this document counts four congestion states, the PCN working group
1752 would count this as three PCN states plus a not-PCN-enabled state.)
1754 Figure 2 (Section 3.2) shows that pre-existing decapsulation
1755 behaviour would have discarded any ECT(1) markings in outer headers
1756 if the inner was ECT(0). This prevented the PCN working group from
1757 using ECT(1) -- if a PCN node used ECT(1) to indicate one of the
1758 severity levels of congestion, any later tunnel egress would revert
1759 the marking to ECT(0) as if nothing had happened. Effectively the
1760 decapsulation rules of RFC4301 and RFC3168 waste one ECT codepoint;
1761 they treat the ECT(0) and ECT(1) codepoints as a single codepoint.
1763 A number of work-rounds to this problem were proposed in the PCN w-g;
1764 to add the fourth state another way or avoid needing it. Without
1765 wishing to disparage the ingenuity of these work-rounds, none were
1766 chosen for the standards track because they were either somewhat
1767 wasteful, imprecise or complicated:
1769 o One uses a pair of Diffserv codepoint(s) in place of each PCN DSCP
1770 to encode the extra state [I-D.ietf-pcn-3-state-encoding], using
1771 up the rapidly exhausting DSCP space while leaving an ECN
1772 codepoint unused.
1774 o Another survives tunnelling without an extra DSCP
1775 [I-D.ietf-pcn-psdm-encoding], but it requires the PCN edge
1776 gateways to share the initial state of a packet out of band.
1778 o Another proposes a more involved marking algorithm in forwarding
1779 elements to encode the three congestion notification states using
1780 only two ECN codepoints [I-D.satoh-pcn-st-marking].
1782 o Another takes a different approach; it compromises the precision
1783 of the admission control mechanism in some network scenarios, but
1784 manages to work with just three encoding states and a single
1785 marking algorithm [I-D.ietf-pcn-sm-edge-behaviour].
1787 Rather than require the IETF to bless any of these experimental
1788 encoding work-rounds, the present specification fixes the root cause
1789 of the problem so that operators deploying PCN can simply require
1790 that tunnel end-points within a PCN region should comply with this
1791 new ECN tunnelling specification. On the public Internet it would
1792 not be possible to know whether all tunnels complied with this new
1793 specification, but universal compliance is feasible for PCN, because
1794 it is intended to be deployed in a controlled Diffserv region.
1796 Given the present specification, the PCN w-g could progress a
1797 trivially simple four-state ECN encoding
1798 [I-D.ietf-pcn-3-in-1-encoding]. This would replace the interim
1799 standards track baseline encoding of just three states [RFC5696]
1800 which makes a fourth state available for any of the experimental
1801 alternatives.
1803 Appendix E. Why Resetting ECN on Encapsulation Impedes PCN
1805 The PCN architecture says "...if encapsulation is done within the
1806 PCN-domain: Any PCN-marking is copied into the outer header. Note: A
1807 tunnel will not provide this behaviour if it complies with [RFC3168]
1808 tunnelling in either mode, but it will if it complies with [RFC4301]
1809 IPsec tunnelling. "
1811 The specific issue here concerns PCN excess rate marking [RFC5670].
1812 The purpose of excess rate marking is to provide a bulk mechanism for
1813 interior nodes within a PCN domain to mark traffic that is exceeding
1814 a configured threshold bit-rate, perhaps after an unexpected event
1815 such as a reroute, a link or node failure, or a more widespread
1816 disaster. Reroutes are a common cause of QoS degradation in IP
1817 networks. After reroutes it is common for multiple links in a
1818 network to become stressed at once. Therefore, PCN excess rate
1819 marking has been carefully designed to ensure traffic marked at one
1820 queue will not be counted again for marking at subsequent queues (see
1821 the `Excess traffic meter function' of [RFC5670]).
1823 However, if an RFC3168 tunnel ingress intervenes, it resets the ECN
1824 field in all the outer headers. This will cause excess traffic to be
1825 counted more than once, leading to many flows being removed that did
1826 not need to be removed at all. This is why the an RFC3168 tunnel
1827 ingress cannot be used in a PCN domain.
1829 The ECN reset in RFC3168 is no longer deemed necessary, it is
1830 inconsistent with RFC4301, it is not as simple as RFC4301 and it is
1831 impeding deployment of new protocols like PCN. The present
1832 specification corrects this perverse situation.
1834 Appendix F. Compromise on Decap with ECT(1) Inner and ECT(0) Outer
1836 A packet with an ECT(1) inner and an ECT(0) outer should never arise
1837 from any known IETF protocol. Without giving a reason, RFC3168 and
1838 RFC4301 both say the outer should be ignored when decapsulating such
1839 a packet. This appendix explains why it was decided not to change
1840 this advice.
1842 In summary, ECT(0) always means 'not congested' and ECT(1) may imply
1843 the same [RFC3168] or it may imply a higher severity congestion
1844 signal [RFC4774], [I-D.ietf-pcn-3-in-1-encoding], depending on the
1845 transport in use. Whether they mean the same or not, at the ingress
1846 the outer should have started the same as the inner and only a broken
1847 or compromised router could have changed the outer to ECT(0).
1849 The decapsulator can detect this anomaly. But the question is,
1850 should it correct the anomaly by ignoring the outer, or should it
1851 reveal the anomaly to the end-to-end transport by forwarding the
1852 outer?
1854 On balance, it was decided that the decapsulator should correct the
1855 anomaly, but log the event and optionally raise an alarm. This is
1856 the safe action if ECT(1) is being used as a more severe marking than
1857 ECT(0), because it passes the more severe signal to the transport.
1858 However, it is not a good idea to hide anomalies, which is why an
1859 optional alarm is suggested. It should be noted that this anomaly
1860 may be the result of two changes to the outer: a broken or
1861 compromised router within the tunnel might be erasing congestion
1862 markings introduced earlier in the same tunnel by a congested router.
1863 In this case, the anomaly would be losing congestion signals, which
1864 needs immediate attention.
1866 The original reason for defining ECT(0) and ECT(1) as equivalent was
1867 so that the data source could use the ECN nonce [RFC3540] to detect
1868 if congestion signals were being erased. However, in this case, the
1869 decapsulator does not need a nonce to detect any anomalies introduced
1870 within the tunnel, because it has the inner as a record of the header
1871 at the ingress. Therefore, it was decided that the best compromise
1872 would be to give precedence to solving the safety issue over
1873 revealing the anomaly, because the anomaly could at least be detected
1874 and dealt with internally.
1876 Superficially, the opposite case where the inner and outer carry
1877 different ECT values, but with an ECT(1) outer and ECT(0) inner seems
1878 to require a similar compromise. However, because that case is
1879 reversed, no compromise is necessary; it is best to forward the outer
1880 whether the transport expects the ECT(1) to mean a higher severity
1881 than ECT(0) or the same severity. Forwarding the outer either
1882 preserves a higher value (if it is higher) or it reveals an anomaly
1883 to the transport (if the two ECT codepoints mean the same severity).
1885 Appendix G. Open Issues
1887 The new decapsulation behaviour defined in Section 4.2 adds support
1888 for propagation of 2 severity levels of congestion. However
1889 transports have no way to discover whether there are any legacy
1890 tunnels on their path that will not propagate 2 severity levels. It
1891 would have been nice to add a feature for transports to check path
1892 support, but this remains an open issue that will have to be
1893 addressed in any future standards action to define an end-to-end
1894 scheme that requires 2-severity levels of congestion. PCN avoids
1895 this problem, because it is only for a controlled region, so all
1896 legacy tunnels can be upgraded by the same operator that deploys PCN.
1898 Author's Address
1900 Bob Briscoe
1901 BT
1902 B54/77, Adastral Park
1903 Martlesham Heath
1904 Ipswich IP5 3RE
1905 UK
1907 Phone: +44 1473 645196
1908 EMail: bob.briscoe@bt.com
1909 URI: http://bobbriscoe.net/