idnits 2.17.1 draft-ietf-6lo-fragment-recovery-04.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC4944, updated by this document, for RFC5378 checks: 2005-07-13) -- 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 (June 11, 2019) is 1779 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-15) exists of draft-ietf-6lo-minimal-fragment-01 == Outdated reference: A later version (-30) exists of draft-ietf-6tisch-architecture-20 Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 6lo P. Thubert, Ed. 3 Internet-Draft Cisco Systems 4 Updates: 4944 (if approved) June 11, 2019 5 Intended status: Standards Track 6 Expires: December 13, 2019 8 6LoWPAN Selective Fragment Recovery 9 draft-ietf-6lo-fragment-recovery-04 11 Abstract 13 This draft updates RFC 4944 with a simple protocol to recover 14 individual fragments across a route-over mesh network, with a minimal 15 flow control to protect the network against bloat. 17 Status of This Memo 19 This Internet-Draft is submitted in full conformance with the 20 provisions of BCP 78 and BCP 79. 22 Internet-Drafts are working documents of the Internet Engineering 23 Task Force (IETF). Note that other groups may also distribute 24 working documents as Internet-Drafts. The list of current Internet- 25 Drafts is at https://datatracker.ietf.org/drafts/current/. 27 Internet-Drafts are draft documents valid for a maximum of six months 28 and may be updated, replaced, or obsoleted by other documents at any 29 time. It is inappropriate to use Internet-Drafts as reference 30 material or to cite them other than as "work in progress." 32 This Internet-Draft will expire on December 13, 2019. 34 Copyright Notice 36 Copyright (c) 2019 IETF Trust and the persons identified as the 37 document authors. All rights reserved. 39 This document is subject to BCP 78 and the IETF Trust's Legal 40 Provisions Relating to IETF Documents 41 (https://trustee.ietf.org/license-info) in effect on the date of 42 publication of this document. Please review these documents 43 carefully, as they describe your rights and restrictions with respect 44 to this document. Code Components extracted from this document must 45 include Simplified BSD License text as described in Section 4.e of 46 the Trust Legal Provisions and are provided without warranty as 47 described in the Simplified BSD License. 49 Table of Contents 51 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 52 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 4 53 2.1. BCP 14 . . . . . . . . . . . . . . . . . . . . . . . . . 4 54 2.2. References . . . . . . . . . . . . . . . . . . . . . . . 4 55 2.3. 6LoWPAN Acronyms . . . . . . . . . . . . . . . . . . . . 4 56 2.4. Referenced Work . . . . . . . . . . . . . . . . . . . . . 4 57 2.5. New Terms . . . . . . . . . . . . . . . . . . . . . . . . 5 58 3. Updating RFC 4944 . . . . . . . . . . . . . . . . . . . . . . 6 59 4. Updating draft-ietf-6lo-minimal-fragment . . . . . . . . . . 6 60 4.1. Slack in the First Fragment . . . . . . . . . . . . . . . 7 61 4.2. Gap between frames . . . . . . . . . . . . . . . . . . . 7 62 4.3. Modifying the First Fragment . . . . . . . . . . . . . . 7 63 5. New Dispatch types and headers . . . . . . . . . . . . . . . 8 64 5.1. Recoverable Fragment Dispatch type and Header . . . . . . 9 65 5.2. RFRAG Acknowledgment Dispatch type and Header . . . . . . 11 66 6. Fragments Recovery . . . . . . . . . . . . . . . . . . . . . 13 67 6.1. Forwarding Fragments . . . . . . . . . . . . . . . . . . 15 68 6.1.1. Upon the first fragment . . . . . . . . . . . . . . . 15 69 6.1.2. Upon the next fragments . . . . . . . . . . . . . . . 15 70 6.2. Upon the RFRAG Acknowledgments . . . . . . . . . . . . . 16 71 6.3. Aborting the Transmission of a Fragmented Packet . . . . 17 72 7. Management Considerations . . . . . . . . . . . . . . . . . . 17 73 7.1. Protocol Parameters . . . . . . . . . . . . . . . . . . . 17 74 7.2. Observing the network . . . . . . . . . . . . . . . . . . 18 75 8. Security Considerations . . . . . . . . . . . . . . . . . . . 19 76 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 19 77 10. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 19 78 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 19 79 11.1. Normative References . . . . . . . . . . . . . . . . . . 19 80 11.2. Informative References . . . . . . . . . . . . . . . . . 20 81 Appendix A. Rationale . . . . . . . . . . . . . . . . . . . . . 22 82 Appendix B. Requirements . . . . . . . . . . . . . . . . . . . . 24 83 Appendix C. Considerations On Flow Control . . . . . . . . . . . 24 84 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 26 86 1. Introduction 88 In most Low Power and Lossy Network (LLN) applications, the bulk of 89 the traffic consists of small chunks of data (in the order few bytes 90 to a few tens of bytes) at a time. Given that an IEEE Std. 802.15.4 91 [IEEE.802.15.4] frame can carry a payload of 74 bytes or more, 92 fragmentation is usually not required. However, and though this 93 happens only occasionally, a number of mission critical applications 94 do require the capability to transfer larger chunks of data, for 95 instance to support the firmware upgrade of the LLN nodes or the 96 extraction of logs from LLN nodes. In the former case, the large 97 chunk of data is transferred to the LLN node, whereas in the latter, 98 the large chunk flows away from the LLN node. In both cases, the 99 size can be on the order of 10 kilobytes or more and an end-to-end 100 reliable transport is required. 102 "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944] 103 defines the original 6LoWPAN datagram fragmentation mechanism for 104 LLNs. One critical issue with this original design is that routing 105 an IPv6 [RFC8200] packet across a route-over mesh requires to 106 reassemble the full packet at each hop, which may cause latency along 107 a path and an overall buffer bloat in the network. The "6TiSCH 108 Architecture" [I-D.ietf-6tisch-architecture] recommends to use a hop- 109 by-hop fragment forwarding technique to alleviate those undesirable 110 effects. "LLN Minimal Fragment Forwarding" 111 [I-D.ietf-6lo-minimal-fragment] proposes such a technique, in a 112 fashion that is compatible with [RFC4944] without the need to define 113 a new protocol. 115 However, adding that capability alone to the local implementation of 116 the original 6LoWPAN fragmentation would not address the issues of 117 resources locked and wasted transmissions due to the loss of a 118 fragment. [RFC4944] does not define a mechanism to first discover a 119 fragment loss, and then to recover that loss. With RFC 4944, the 120 forwarding of a whole datagram fails when one fragment is not 121 delivered properly to the destination 6LoWPAN endpoint. Constrained 122 memory resources are blocked on the receiver until the receiver times 123 out. 125 That problem is exacerbated when forwarding fragments over multiple 126 hops since a loss at an intermediate hop will not be discovered by 127 either the source or the destination, and the source will keep on 128 sending fragments, wasting even more resources in the network and 129 possibly contributing to the condition that caused the loss to no 130 avail since the datagram cannot arrive in its entirety. RFC 4944 is 131 also missing signaling to abort a multi-fragment transmission at any 132 time and from either end, and, if the capability to forward fragments 133 is implemented, clean up the related state in the network. It is 134 also lacking flow control capabilities to avoid participating to a 135 congestion that may in turn cause the loss of a fragment and 136 potentially the retransmission of the full datagram. 138 This specification proposes a method to forward fragments across a 139 multi-hop route-over mesh, and to recover individual fragments 140 between LLN endpoints. The method is designed to limit congestion 141 loss in the network and addresses the requirements that are detailed 142 in Appendix B. 144 2. Terminology 146 2.1. BCP 14 148 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 149 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 150 "OPTIONAL" in this document are to be interpreted as described in BCP 151 14 [RFC2119][RFC8174] when, and only when, they appear in all 152 capitals, as shown here. 154 2.2. References 156 In this document, readers will encounter terms and concepts that are 157 discussed in "Problem Statement and Requirements for IPv6 over Low- 158 Power Wireless Personal Area Network (6LoWPAN) Routing" [RFC6606] 160 2.3. 6LoWPAN Acronyms 162 This document uses the following acronyms: 164 6BBR: 6LoWPAN Backbone Router 166 6LBR: 6LoWPAN Border Router 168 6LN: 6LoWPAN Node 170 6LR: 6LoWPAN Router 172 LLN: Low-Power and Lossy Network 174 2.4. Referenced Work 176 Past experience with fragmentation has shown that misassociated or 177 lost fragments can lead to poor network behavior and, occasionally, 178 trouble at application layer. The reader is encouraged to read "IPv4 179 Reassembly Errors at High Data Rates" [RFC4963] and follow the 180 references for more information. 182 That experience led to the definition of "Path MTU discovery" 183 [RFC8201] (PMTUD) protocol that limits fragmentation over the 184 Internet. 186 Specifically in the case of UDP, valuable additional information can 187 be found in "UDP Usage Guidelines for Application Designers" 188 [RFC8085]. 190 Readers are expected to be familiar with all the terms and concepts 191 that are discussed in "IPv6 over Low-Power Wireless Personal Area 192 Networks (6LoWPANs): Overview, Assumptions, Problem Statement, and 193 Goals" [RFC4919] and "Transmission of IPv6 Packets over IEEE 802.15.4 194 Networks" [RFC4944]. 196 "The Benefits of Using Explicit Congestion Notification (ECN)" 197 [RFC8087] provides useful information on the potential benefits and 198 pitfalls of using ECN. 200 Quoting the "Multiprotocol Label Switching (MPLS) Architecture" 201 [RFC3031]: with MPLS, 'packets are "labeled" before they are 202 forwarded'. At subsequent hops, there is no further analysis of the 203 packet's network layer header. Rather, the label is used as an index 204 into a table which specifies the next hop, and a new label". The 205 MPLS technique is leveraged in the present specification to forward 206 fragments that actually do not have a network layer header, since the 207 fragmentation occurs below IP. 209 "LLN Minimal Fragment Forwarding" [I-D.ietf-6lo-minimal-fragment] 210 introduces the concept of a Virtual Reassembly Buffer (VRB) and an 211 associated technique to forward fragments as they come, using the 212 datagram_tag as a label in a fashion similar to MPLS. This 213 specification reuses that technique with slightly modified controls. 215 2.5. New Terms 217 This specification uses the following terms: 219 6LoWPAN endpoints The LLN nodes in charge of generating or expanding 220 a 6LoWPAN header from/to a full IPv6 packet. The 6LoWPAN 221 endpoints are the points where fragmentation and reassembly take 222 place. 224 Compressed Form This specification uses the generic term Compressed 225 Form to refer to the format of a datagram after the action of 226 [RFC6282] and possibly [RFC8138] for RPL [RFC6550] artifacts. 228 datagram_size: The size of the datagram in its Compressed Form 229 before it is fragmented. The datagram_size is expressed in a unit 230 that depends on the MAC layer technology, by default a byte. 232 fragment_offset: The offset of a particular fragment of a datagram 233 in its Compressed Form. The fragment_offset is expressed in a 234 unit that depends on the MAC layer technology and is by default a 235 byte. 237 datagram_tag: An identifier of a datagram that is locally unique to 238 the Layer-2 sender. Associated with the MAC address of the 239 sender, this becomes a globally unique identifier for the 240 datagram. 242 RFRAG: Recoverable Fragment 244 RFRAG-ACK: Recoverable Fragment Acknowledgement 246 RFRAG Acknowledgment Request: An RFRAG with the Acknowledgement 247 Request flag ('X' flag) set. 249 All 0's: Refers to a bitmap with all bits set to zero. 251 All 1's: Refers to a bitmap with all bits set to one. 253 3. Updating RFC 4944 255 This specification updates the fragmentation mechanism that is 256 specified in "Transmission of IPv6 Packets over IEEE 802.15.4 257 Networks" [RFC4944] for use in route-over LLNs by providing a model 258 where fragments can be forwarded end-to-end across a 6LoWPAN LLN, and 259 where fragments that are lost on the way can be recovered 260 individually. A new format for fragment is introduced and new 261 dispatch types are defined in Section 5. 263 [RFC8138] allows to modify the size of a packet en-route by removing 264 the consumed hops in a compressed Routing Header. It results that 265 fragment_offset and datagram_size (see Section 2.5) must also be 266 modified en-route, whcih is difficult to do in the uncompressed form. 267 This specification expresses those fields in the Compressed Form and 268 allows to modify them en-route (see Section 4.3) easily. 270 Note that consistently with Section 2 of [RFC6282] for the 271 fragmentation mechanism described in Section 5.3 of [RFC4944], any 272 header that cannot fit within the first fragment MUST NOT be 273 compressed when using the fragmentation mechanism described in this 274 specification. 276 4. Updating draft-ietf-6lo-minimal-fragment 278 This specification updates the fragment forwarding mechanism 279 specified in "LLN Minimal Fragment Forwarding" 280 [I-D.ietf-6lo-minimal-fragment] by providing additional operations to 281 improve the management of the Virtual Reassembly Buffer (VRB). 283 4.1. Slack in the First Fragment 285 At the time of this writing, [I-D.ietf-6lo-minimal-fragment] allows 286 for refragmenting in intermediate nodes, meaning that some bytes from 287 a given fragment may be left in the VRB to be added to the next 288 fragment. The reason for this to happen would be the need for space 289 in the outgoing fragment that was not needed in the incoming 290 fragment, for instance because the 6LoWPAN Header Compression is not 291 as efficient on the outgoing link, e.g., if the Interface ID (IID) of 292 the source IPv6 address is elided by the originator on the first hop 293 because it matches the source MAC address, but cannot be on the next 294 hops because the source MAC address changes. 296 This specification cannot allow this operation since fragments are 297 recovered end-to-end based on a sequence number. This means that the 298 fragments that contain a 6LoWPAN-compressed header MUST have enough 299 slack to enable a less efficient compression in the next hops that 300 still fits in one MAC frame. For instance, if the IID of the source 301 IPv6 address is elided by the originator, then it MUST compute the 302 fragment_size as if the MTU was 8 bytes less. This way, the next hop 303 can restore the source IID to the first fragment without impacting 304 the second fragment. 306 4.2. Gap between frames 308 This specification introduces a concept of Inter-Frame Gap, which is 309 a configurable interval of time between transmissions to a same next 310 hop. In the case of half duplex interfaces, this InterFrameGap 311 ensures that the next hop has progressed the previous frame and is 312 capable of receiving the next one. 314 In the case of a mesh operating at a single frequency with 315 omnidirectional antennas, a larger InterFrameGap is required to 316 protect the frame against hidden terminal collisions with the 317 previous frame of a same flow that is still progressing along a 318 common path. 320 The Inter-Frame Gap is useful even for unfragmented datagrams, but it 321 becomes a necessity for fragments that are typically generated in a 322 fast sequence and are all sent over the exact same path. 324 4.3. Modifying the First Fragment 326 The compression of the Hop Limit, of the source and destination 327 addresses in the IPv6 Header, and of the Routing Header, may change 328 en-route in a Route-Over mesh LLN. If the size of the first fragment 329 is modified, then the intermediate node MUST adapt the datagram_size 330 to reflect that difference. 332 The intermediate node MUST also save the difference of datagram_size 333 of the first fragment in the VRB and add it to the datagram_size and 334 to the fragment_offset of all the subsequent fragments for that 335 datagram. 337 5. New Dispatch types and headers 339 This specification enables the 6LoWPAN fragmentation sublayer to 340 provide an MTU up to 2048 bytes to the upper layer, which can be the 341 6LoWPAN Header Compression sublayer that is defined in the 342 "Compression Format for IPv6 Datagrams" [RFC6282] specification. In 343 order to achieve this, this specification enables the fragmentation 344 and the reliable transmission of fragments over a multihop 6LoWPAN 345 mesh network. 347 This specification provides a technique that is derived from MPLS to 348 forward individual fragments across a 6LoWPAN route-over mesh without 349 reassembly at each hop. The datagram_tag is used as a label; it is 350 locally unique to the node that owns the source MAC address of the 351 fragment, so together the MAC address and the label can identify the 352 fragment globally. A node may build the datagram_tag in its own 353 locally-significant way, as long as the chosen datagram_tag stays 354 unique to the particular datagram for the lifetime of that datagram. 355 It results that the label does not need to be globally unique but 356 also that it must be swapped at each hop as the source MAC address 357 changes. 359 This specification extends RFC 4944 [RFC4944] with 2 new Dispatch 360 types, for Recoverable Fragment (RFRAG) and for the RFRAG 361 Acknowledgment back. 363 (to be confirmed by IANA) The new 6LoWPAN Dispatch types use the 364 Value Bit Pattern of 11 1010xx from Page 0 [RFC8025], as follows: 366 Pattern Header Type 367 +------------+------------------------------------------+ 368 | 11 10100x | RFRAG - Recoverable Fragment | 369 | 11 10101x | RFRAG-ACK - RFRAG Acknowledgment | 370 +------------+------------------------------------------+ 372 Figure 1: Additional Dispatch Value Bit Patterns 374 In the following sections, a "datagram_tag" extends the semantics 375 defined in [RFC4944] Section 5.3."Fragmentation Type and Header". 376 The datagram_tag is a locally unique identifier for the datagram from 377 the perspective of the sender. This means that the datagram_tag 378 identifies a datagram uniquely in the network when associated with 379 the source of the datagram. As the datagram gets forwarded, the 380 source changes and the datagram_tag must be swapped as detailed in 381 [I-D.ietf-6lo-minimal-fragment]. 383 5.1. Recoverable Fragment Dispatch type and Header 385 In this specification, if the packet is compressed then the size and 386 offset of the fragments are expressed on the Compressed Form of the 387 packet form as opposed to the uncompressed - native - packet form. 389 The format of the fragment header is shown in Figure 2. It is the 390 same for all fragments. The format has a length and an offset, as 391 well as a sequence field. This would be redundant if the offset was 392 computed as the product of the sequence by the length, but this is 393 not the case. The position of a fragment in the reassembly buffer is 394 neither correlated with the value of the sequence field nor with the 395 order in which the fragments are received. This enables out-of- 396 sequence and overlapping fragments, e.g., a fragment 5 that is 397 retried as smaller fragments 5, 13 and 14 due to a change of MTU. 399 There is no requirement on the receiver to check for contiguity of 400 the received fragments, and the sender MUST ensure that when all 401 fragments are acknowledged, then the datagram is fully received. 402 This may be useful in particular in the case where the MTU changes 403 and a fragment sequence is retried with a smaller fragment_size, the 404 remainder of the original fragment being retried with new sequence 405 values. 407 The first fragment is recognized by a sequence of 0; it carries its 408 fragment_size and the datagram_size of the compressed packet before 409 it is fragmented, whereas the other fragments carry their 410 fragment_size and fragment_offset. The last fragment for a datagram 411 is recognized when its fragment_offset and its fragment_size add up 412 to the datagram_size. 414 Recoverable Fragments are sequenced and a bitmap is used in the RFRAG 415 Acknowledgment to indicate the received fragments by setting the 416 individual bits that correspond to their sequence. 418 1 2 3 419 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 420 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 421 |1 1 1 0 1 0 0|E| datagram_tag | 422 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 423 |X| sequence| fragment_size | fragment_offset | 424 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 426 X set == Ack-Request 428 Figure 2: RFRAG Dispatch type and Header 430 X: 1 bit; Ack-Request: when set, the sender requires an RFRAG 431 Acknowledgment from the receiver. 433 E: 1 bit; Explicit Congestion Notification; the "E" flag is reset by 434 the source of the fragment and set by intermediate routers to 435 signal that this fragment experienced congestion along its path. 437 Fragment_size: 10 bit unsigned integer; the size of this fragment in 438 a unit that depends on the MAC layer technology. By default, that 439 unit is the octet which allows fragments up to 512 bytes. For 440 IEEE Std. 802.15.4, the unit is octet, and the maximum fragment 441 size, when it is constrained by the maximum frame size of 128 442 octet minus the overheads of the MAC and Fragment Headers, is not 443 limited by this encoding. 445 datagram_tag: 16 bits; an identifier of the datagram that is locally 446 unique to the sender. 448 Sequence: 5 bit unsigned integer; the sequence number of the 449 fragment in the acknowledgement bitmap. Fragments are numbered 450 [0..N] where N is in [0..31]. A Sequence of 0 indicates the first 451 fragment in a datagram, but non-zero values are not indicative of 452 the position in the reassembly buffer. 454 Fragment_offset: 16 bit unsigned integer; 456 * When the Fragment_offset is set to a non-0 value, its semantics 457 depend on the value of the Sequence field. 459 + For a first fragment (i.e. with a Sequence of 0), this field 460 indicates the datagram_size of the compressed datagram, to 461 help the receiver allocate an adapted buffer for the 462 reception and reassembly operations. The fragment may be 463 stored for local reassembly. Alternatively, it may be 464 routed based on the destination IPv6 address. In that case, 465 a VRB state must be installed as described in Section 6.1.1. 467 + When the Sequence is not 0, this field indicates the offset 468 of the fragment in the Compressed Form of the datagram. The 469 fragment may be added to a local reassembly buffer or 470 forwarded based on an existing VRB as described in 471 Section 6.1.2. 473 * A Fragment_offset that is set to a value of 0 indicates an 474 abort condition and all state regarding the datagram should be 475 cleaned up once the processing of the fragment is complete; the 476 processing of the fragment depends on whether there is a VRB 477 already established for this datagram, and the next hop is 478 still reachable: 480 + if a VRB already exists and is not broken, the fragment is 481 to be forwarded along the associated Label Switched Path 482 (LSP) as described in Section 6.1.2, but regardless of the 483 value of the Sequence field; 485 + else, if the Sequence is 0, then the fragment is to be 486 routed as described in Section 6.1.1 but no state is 487 conserved afterwards. In that case, the session if it 488 exists is aborted and the packet is also forwarded in an 489 attempt to clean up the next hops as along the path 490 indicated by the IPv6 header (possibly including a routing 491 header). 493 If the fragment cannot be forwarded or routed, then an abort 494 RFRAG-ACK is sent back to the source as described in 495 Section 6.1.2. 497 5.2. RFRAG Acknowledgment Dispatch type and Header 499 This specification also defines a 4-octet RFRAG Acknowledgment bitmap 500 that is used by the reassembling end point to confirm selectively the 501 reception of individual fragments. A given offset in the bitmap maps 502 one to one with a given sequence number. 504 The offset of the bit in the bitmap indicates which fragment is 505 acknowledged as follows: 507 1 2 3 508 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 509 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 510 | RFRAG Acknowledgment Bitmap | 511 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 512 ^ ^ 513 | | bitmap indicating whether: 514 | +----- Fragment with sequence 9 was received 515 +----------------------- Fragment with sequence 0 was received 517 Figure 3: RFRAG Acknowledgment bitmap encoding 519 Figure 4 shows an example Acknowledgment bitmap which indicates that 520 all fragments from sequence 0 to 20 were received, except for 521 fragments 1, 2 and 16 that were lost and must be retried. 523 1 2 3 524 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 525 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 526 |1|0|0|1|1|1|1|1|1|1|1|1|1|1|1|1|0|1|1|1|1|0|0|0|0|0|0|0|0|0|0|0| 527 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 529 Figure 4: Example RFRAG Acknowledgment Bitmap 531 The RFRAG Acknowledgment Bitmap is included in a RFRAG Acknowledgment 532 header, as follows: 534 1 2 3 535 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 536 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 537 |1 1 1 0 1 0 1|E| datagram_tag | 538 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 539 | RFRAG Acknowledgment Bitmap (32 bits) | 540 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 542 Figure 5: RFRAG Acknowledgment Dispatch type and Header 544 E: 1 bit; Explicit Congestion Notification Echo 546 When set, the sender indicates that at least one of the 547 acknowledged fragments was received with an Explicit Congestion 548 Notification, indicating that the path followed by the fragments 549 is subject to congestion. More in Appendix C. 551 RFRAG Acknowledgment Bitmap 552 An RFRAG Acknowledgment Bitmap, whereby setting the bit at offset 553 x indicates that fragment x was received, as shown in Figure 3. 554 All 0's is a NULL bitmap that indicates that the fragmentation 555 process is aborted. All 1's is a FULL bitmap that indicates that 556 the fragmentation process is complete, all fragments were received 557 at the reassembly end point. 559 6. Fragments Recovery 561 The Recoverable Fragment header RFRAG is used to transport a fragment 562 and optionally request an RFRAG Acknowledgment that will confirm the 563 good reception of one or more fragments. An RFRAG Acknowledgment is 564 carried as a standalone fragment header (i.e. with no 6LoWPAN 565 payload) in a message that is propagated back to the 6LoWPAN endpoint 566 that was the originator of the fragments. To achieve this, each hop 567 that performed an MPLS-like operation on fragments reverses that 568 operation for the RFRAG_ACK by sending a frame from the next hop to 569 the previous hop as known by its MAC address in the VRB. The 570 datagram_tag in the RFRAG_ACK is unique to the receiver and is enough 571 information for an intermediate hop to locate the VRB that contains 572 the datagram_tag used by the previous hop and the Layer-2 information 573 associated to it (interface and MAC address). 575 The 6LoWPAN endpoint that fragments the packets at 6LoWPAN level (the 576 sender) also controls the amount of acknowledgments by setting the 577 Ack-Request flag in the RFRAG packets. The sender may set the Ack- 578 Request flag on any fragment to perform congestion control by 579 limiting the number of outstanding fragments, which are the fragments 580 that have been sent but for which reception or loss was not 581 positively confirmed by the reassembling endpoint. The maximum 582 number of outstanding fragments is the Window-Size. It is 583 configurable and may vary in case of ECN notification. When the 584 6LoWPAN endpoint that reassembles the packets at 6LoWPAN level (the 585 receiver) receives a fragment with the Ack-Request flag set, it MUST 586 send an RFRAG Acknowledgment back to the originator to confirm 587 reception of all the fragments it has received so far. 589 The Ack-Request ('X') set in an RFRAG marks the end of a window. 590 This flag SHOULD be set on the last fragment to protect the datagram, 591 and it MAY be set in any intermediate fragment for the purpose of 592 flow control. This ARQ process MUST be protected by a timer, and the 593 fragment that carries the 'X' flag MAY be retried upon time out a 594 configurable amount of times (see Section 7.1). Upon exhaustion of 595 the retries the sender may either abort the transmission of the 596 datagram or retry the datagram from the first fragment with an 'X' 597 flag set in order to reestablish a path and discover which fragments 598 were received over the old path in the acknowledgment bitmap. When 599 the sender of the fragment knows that an underlying link-layer 600 mechanism protects the fragments, it may refrain from using the RFRAG 601 Acknowledgment mechanism, and never set the Ack-Request bit. 603 The RFRAG Acknowledgment can optionally carry an ECN indication for 604 flow control (see Appendix C). The receiver of a fragment with the 605 'E' (ECN) flag set MUST echo that information by setting the 'E' 606 (ECN) flag in the next RFRAG Acknowledgment. 608 The sender transfers a controlled number of fragments and MAY flag 609 the last fragment of a window with an RFRAG Acknowledgment Request. 610 The receiver MUST acknowledge a fragment with the acknowledgment 611 request bit set. If any fragment immediately preceding an 612 acknowledgment request is still missing, the receiver MAY 613 intentionally delay its acknowledgment to allow in-transit fragments 614 to arrive. Because it might defeat the round trip delay computation, 615 delaying the acknowledgment should be configurable and not enabled by 616 default. 618 The receiver MAY issue unsolicited acknowledgments. An unsolicited 619 acknowledgment signals to the sender endpoint that it can resume 620 sending if it had reached its maximum number of outstanding 621 fragments. Another use is to inform that the reassembling endpoint 622 aborted the process of an individual datagram. Note that 623 acknowledgments might consume precious resources so the use of 624 unsolicited acknowledgments should be configurable and not enabled by 625 default. 627 An observation is that streamlining forwarding of fragments generally 628 reduces the latency over the LLN mesh, providing room for retries 629 within existing upper-layer reliability mechanisms. The sender 630 protects the transmission over the LLN mesh with a retry timer that 631 is computed according to the method detailed in [RFC6298]. It is 632 expected that the upper layer retries obey the recommendations in 633 "UDP Usage Guidelines" [RFC8085], in which case a single round of 634 fragment recovery should fit within the upper layer recovery timers. 636 Fragments are sent in a round robin fashion: the sender sends all the 637 fragments for a first time before it retries any lost fragment; lost 638 fragments are retried in sequence, oldest first. This mechanism 639 enables the receiver to acknowledge fragments that were delayed in 640 the network before they are retried. 642 When a single frequency is used by contiguous hops, the sender should 643 wait a reasonable amount of time between fragments so as to let a 644 fragment progress a few hops and avoid hidden terminal issues. This 645 precaution is not required on channel hopping technologies such as 646 Time Slotted Channel Hopping (TSCH) [RFC6554] 648 6.1. Forwarding Fragments 650 It is assumed that the first Fragment is large enough to carry the 651 IPv6 header and make routing decisions. If that is not so, then this 652 specification MUST NOT be used. 654 This specification extends the Virtual Reassembly Buffer (VRB) 655 technique to forward fragments with no intermediate reconstruction of 656 the entire packet. It inherits operations like datagram_tag 657 Switching and using a timer to clean the VRB when the traffic dries 658 up. In more details, the first fragment carries the IP header and it 659 is routed all the way from the fragmenting end point to the 660 reassembling end point. Upon the first fragment, the routers along 661 the path install a label-switched path (LSP), and the following 662 fragments are label-switched along that path. As a consequence, the 663 next fragments can only follow the path that was set up by the first 664 fragment and cannot follow an alternate route. The datagram_tag is 665 used to carry the label, that is swapped at each hop. All fragments 666 follow the same path and fragments are delivered in the order at 667 which they are sent. 669 6.1.1. Upon the first fragment 671 In Route-Over mode, the source and destination MAC addressed in a 672 frame change at each hop. The label that is formed and placed in the 673 datagram_tag is associated to the source MAC and only valid (and 674 unique) for that source MAC. Upon a first fragment (i.e. with a 675 sequence of zero), a VRB and the associated LSP state are created for 676 the tuple (source MAC address, datagram_tag) and the fragment is 677 forwarded along the IPv6 route that matches the destination IPv6 678 address in the IPv6 header as prescribed by 679 [I-D.ietf-6lo-minimal-fragment]. The LSP state enables to match the 680 (previous MAC address, datagram_tag) in an incoming fragment to the 681 tuple (next MAC address, swapped datagram_tag) used in the forwarded 682 fragment and points at the VRB. In addition, the router also forms a 683 Reverse LSP state indexed by the MAC address of the next hop and the 684 swapped datagram_tag. This reverse LSP state also points at the VRB 685 and enables to match the (next MAC address, swapped_datagram_tag) 686 found in an RFRAG Acknowledgment to the tuple (previous MAC address, 687 datagram_tag) used when forwarding a Fragment Acknowledgment (RFRAG- 688 ACK) back to the sender endpoint. 690 6.1.2. Upon the next fragments 692 Upon a next fragment (i.e. with a non-zero sequence), the router 693 looks up a LSP indexed by the tuple (MAC address, datagram_tag) found 694 in the fragment. If it is found, the router forwards the fragment 695 using the associated VRB as prescribed by 696 [I-D.ietf-6lo-minimal-fragment]. 698 if the VRB for the tuple is not found, the router builds an RFRAG-ACK 699 to abort the transmission of the packet. The resulting message has 700 the following information: 702 o The source and destination MAC addresses are swapped from those 703 found in the fragment 705 o The datagram_tag set to the datagram_tag found in the fragment 707 o A NULL bitmap is used to signal the abort condition 709 At this point the router is all set and can send the RFRAG-ACK back 710 to the previous router. The RFRAG-ACK should normally be forwarded 711 all the way to the source using the reverse LSP state in the VRBs in 712 the intermediate routers as described in the next section. 714 6.2. Upon the RFRAG Acknowledgments 716 Upon an RFRAG-ACK, the router looks up a Reverse LSP indexed by the 717 tuple (MAC address, datagram_tag), which are respectively the source 718 MAC address of the received frame and the received datagram_tag. If 719 it is found, the router forwards the fragment using the associated 720 VRB as prescribed by [I-D.ietf-6lo-minimal-fragment], but using the 721 Reverse LSP so that the RFRAG-ACK flows back to the sender endpoint. 723 If the Reverse LSP is not found, the router MUST silently drop the 724 RFRAG-ACK message. 726 Either way, if the RFRAG-ACK indicates that the fragment was entirely 727 received (FULL bitmap), it arms a short timer, and upon timeout, the 728 VRB and all the associated state are destroyed. Until the timer 729 elapses, fragments of that datagram may still be received, e.g. if 730 the RFRAG-ACK was lost on the way back and the source retried the 731 last fragment. In that case, the router forwards the fragment 732 according to the state in the VRB. 734 This specification does not provide a method to discover the number 735 of hops or the minimal value of MTU along those hops. But should the 736 minimal MTU decrease, it is possible to retry a long fragment (say 737 sequence of 5) with first a shorter fragment of the same sequence (5 738 again) and then one or more other fragments with a sequence that was 739 not used before (e.g., 13 and 14). Note that Path MTU Discovery is 740 out of scope for this document. 742 6.3. Aborting the Transmission of a Fragmented Packet 744 A reset is signaled on the forward path with a pseudo fragment that 745 has the fragment_offset, sequence and fragment_size all set to 0, and 746 no data. 748 When the sender or a router on the way decides that a packet should 749 be dropped and the fragmentation process aborted, it generates a 750 reset pseudo fragment and forwards it down the fragment path. 752 Each router next along the path the way forwards the pseudo fragment 753 based on the VRB state. If an acknowledgment is not requested, the 754 VRB and all associated state are destroyed. 756 Upon reception of the pseudo fragment, the receiver cleans up all 757 resources for the packet associated to the datagram_tag. If an 758 acknowledgment is requested, the receiver responds with a NULL 759 bitmap. 761 The other way around, the receiver might need to abort the process of 762 a fragmented packet for internal reasons, for instance if it is out 763 of reassembly buffers, or considers that this packet is already fully 764 reassembled and passed to the upper layer. In that case, the 765 receiver SHOULD indicate so to the sender with a NULL bitmap in a 766 RFRAG Acknowledgment. Upon an acknowledgment with a NULL bitmap, the 767 sender endpoint MUST abort the transmission of the fragmented 768 datagram. 770 7. Management Considerations 772 7.1. Protocol Parameters 774 There is no particular configuration on the receiver, as echoing ECN 775 is always on. The configuration only applies to the sender, which is 776 in control of the transmission. The management system SHOULD be 777 capable of providing the parameters below: 779 MinFragmentSize: The MinFragmentSize is the minimum value for the 780 Fragment_Size. 782 OptFragmentSize: The MinFragmentSize is the value for the 783 Fragment_Size that the sender should use to start with. 785 MaxFragmentSize: The MaxFragmentSize is the maximum value for the 786 Fragment_Size. It MUST be lower than the minimum MTU along the 787 path. A large value augments the chances of buffer bloat and 788 transmission loss. The value MUST be less than 512 if the unit 789 that is defined for the PHY layer is the octet. 791 UseECN: Indicates whether the sender should react to ECN. When the 792 sender reacts to ECN the Window_Size will vary between 793 MinWindowSize and MaxWindowSize. 795 MinWindowSize: The minimum value of Window_Size that the sender can 796 use. 798 OptWindowSize: The OptWindowSize is the value for the Window_Size 799 that the sender should use to start with. 801 MaxWindowSize: The maximum value of Window_Size that the sender can 802 use. The value MUSt be less than 32. 804 InterFrameGap: Indicates a minimum amount of time between 805 transmissions. All packets to a same destination, and in 806 particular fragments, may be subject to receive while 807 transmitting and hidden terminal collisions with the next or 808 the previous transmission as the fragments progress along a 809 same path. The InterFrameGap protects the propagation of one 810 transmission before the next one is triggered and creates a 811 duty cycle that controls the ratio of air time and memory in 812 intermediate nodes that a particular datagram will use. 814 MinARQTimeOut: The maximum amount of time a node should wait for an 815 RFRAG Acknowledgment before it takes a next action. 817 OptARQTimeOut: The starting point of the value of the amount that a 818 sender should wait for an RFRAG Acknowledgment before it takes 819 a next action. 821 MaxARQTimeOut: The maximum amount of time a node should wait for an 822 RFRAG Acknowledgment before it takes a next action. 824 MaxFragRetries: The maximum number of retries for a particular 825 Fragment. 827 MaxDatagramRetries: The maximum number of retries from scratch for a 828 particular Datagram. 830 7.2. Observing the network 832 The management system should monitor the amount of retries and of ECN 833 settings that can be observed from the perspective of the both the 834 sender and the receiver, and may tune the optimum size of 835 Fragment_Size and of the Window_Size, OptWindowSize and OptWindowSize 836 respectively, at the sender. The values should be bounded by the 837 expected number of hops and reduced beyond that when the number of 838 datagrams that can traverse an intermediate point may exceed its 839 capacity and cause a congestion loss. The InterFrameGap is another 840 tool that can be used to increase the spacing between fragments of a 841 same datagram and reduce the ratio of time when a particular 842 intermediate node holds a fragment of that datagram. 844 8. Security Considerations 846 The considerations in the Security section of [I-D.ietf-core-cocoa] 847 apply equally to this specification. 849 The process of recovering fragments does not appear to create any 850 opening for new threat compared to "Transmission of IPv6 Packets over 851 IEEE 802.15.4 Networks" [RFC4944]. 853 9. IANA Considerations 855 Need extensions for formats defined in "Transmission of IPv6 Packets 856 over IEEE 802.15.4 Networks" [RFC4944]. 858 10. Acknowledgments 860 The author wishes to thank Michel Veillette, Dario Tedeschi, Laurent 861 Toutain, Carles Gomez Montenegro, Thomas Watteyne and Michael 862 Richardson for in-depth reviews and comments. Also many thanks to 863 Jonathan Hui, Jay Werb, Christos Polyzois, Soumitri Kolavennu, Pat 864 Kinney, Margaret Wasserman, Richard Kelsey, Carsten Bormann and Harry 865 Courtice for their various contributions. 867 11. References 869 11.1. Normative References 871 [I-D.ietf-6lo-minimal-fragment] 872 Watteyne, T., Bormann, C., and P. Thubert, "LLN Minimal 873 Fragment Forwarding", draft-ietf-6lo-minimal-fragment-01 874 (work in progress), March 2019. 876 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 877 Requirement Levels", BCP 14, RFC 2119, 878 DOI 10.17487/RFC2119, March 1997, 879 . 881 [RFC4944] Montenegro, G., Kushalnagar, N., Hui, J., and D. Culler, 882 "Transmission of IPv6 Packets over IEEE 802.15.4 883 Networks", RFC 4944, DOI 10.17487/RFC4944, September 2007, 884 . 886 [RFC6282] Hui, J., Ed. and P. Thubert, "Compression Format for IPv6 887 Datagrams over IEEE 802.15.4-Based Networks", RFC 6282, 888 DOI 10.17487/RFC6282, September 2011, 889 . 891 [RFC6554] Hui, J., Vasseur, JP., Culler, D., and V. Manral, "An IPv6 892 Routing Header for Source Routes with the Routing Protocol 893 for Low-Power and Lossy Networks (RPL)", RFC 6554, 894 DOI 10.17487/RFC6554, March 2012, 895 . 897 [RFC8025] Thubert, P., Ed. and R. Cragie, "IPv6 over Low-Power 898 Wireless Personal Area Network (6LoWPAN) Paging Dispatch", 899 RFC 8025, DOI 10.17487/RFC8025, November 2016, 900 . 902 [RFC8138] Thubert, P., Ed., Bormann, C., Toutain, L., and R. Cragie, 903 "IPv6 over Low-Power Wireless Personal Area Network 904 (6LoWPAN) Routing Header", RFC 8138, DOI 10.17487/RFC8138, 905 April 2017, . 907 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 908 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 909 May 2017, . 911 11.2. Informative References 913 [I-D.ietf-6tisch-architecture] 914 Thubert, P., "An Architecture for IPv6 over the TSCH mode 915 of IEEE 802.15.4", draft-ietf-6tisch-architecture-20 (work 916 in progress), March 2019. 918 [I-D.ietf-core-cocoa] 919 Bormann, C., Betzler, A., Gomez, C., and I. Demirkol, 920 "CoAP Simple Congestion Control/Advanced", draft-ietf- 921 core-cocoa-03 (work in progress), February 2018. 923 [IEEE.802.15.4] 924 IEEE, "IEEE Standard for Low-Rate Wireless Networks", 925 IEEE Standard 802.15.4, DOI 10.1109/IEEE 926 P802.15.4-REVd/D01, 927 . 929 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, 930 RFC 2914, DOI 10.17487/RFC2914, September 2000, 931 . 933 [RFC3031] Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol 934 Label Switching Architecture", RFC 3031, 935 DOI 10.17487/RFC3031, January 2001, 936 . 938 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 939 of Explicit Congestion Notification (ECN) to IP", 940 RFC 3168, DOI 10.17487/RFC3168, September 2001, 941 . 943 [RFC4919] Kushalnagar, N., Montenegro, G., and C. Schumacher, "IPv6 944 over Low-Power Wireless Personal Area Networks (6LoWPANs): 945 Overview, Assumptions, Problem Statement, and Goals", 946 RFC 4919, DOI 10.17487/RFC4919, August 2007, 947 . 949 [RFC4963] Heffner, J., Mathis, M., and B. Chandler, "IPv4 Reassembly 950 Errors at High Data Rates", RFC 4963, 951 DOI 10.17487/RFC4963, July 2007, 952 . 954 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 955 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 956 . 958 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 959 "Computing TCP's Retransmission Timer", RFC 6298, 960 DOI 10.17487/RFC6298, June 2011, 961 . 963 [RFC6550] Winter, T., Ed., Thubert, P., Ed., Brandt, A., Hui, J., 964 Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, 965 JP., and R. Alexander, "RPL: IPv6 Routing Protocol for 966 Low-Power and Lossy Networks", RFC 6550, 967 DOI 10.17487/RFC6550, March 2012, 968 . 970 [RFC6606] Kim, E., Kaspar, D., Gomez, C., and C. Bormann, "Problem 971 Statement and Requirements for IPv6 over Low-Power 972 Wireless Personal Area Network (6LoWPAN) Routing", 973 RFC 6606, DOI 10.17487/RFC6606, May 2012, 974 . 976 [RFC7554] Watteyne, T., Ed., Palattella, M., and L. Grieco, "Using 977 IEEE 802.15.4e Time-Slotted Channel Hopping (TSCH) in the 978 Internet of Things (IoT): Problem Statement", RFC 7554, 979 DOI 10.17487/RFC7554, May 2015, 980 . 982 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 983 Recommendations Regarding Active Queue Management", 984 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 985 . 987 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 988 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 989 March 2017, . 991 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 992 Explicit Congestion Notification (ECN)", RFC 8087, 993 DOI 10.17487/RFC8087, March 2017, 994 . 996 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 997 (IPv6) Specification", STD 86, RFC 8200, 998 DOI 10.17487/RFC8200, July 2017, 999 . 1001 [RFC8201] McCann, J., Deering, S., Mogul, J., and R. Hinden, Ed., 1002 "Path MTU Discovery for IP version 6", STD 87, RFC 8201, 1003 DOI 10.17487/RFC8201, July 2017, 1004 . 1006 Appendix A. Rationale 1008 There are a number of uses for large packets in Wireless Sensor 1009 Networks. Such usages may not be the most typical or represent the 1010 largest amount of traffic over the LLN; however, the associated 1011 functionality can be critical enough to justify extra care for 1012 ensuring effective transport of large packets across the LLN. 1014 The list of those usages includes: 1016 Towards the LLN node: 1018 Firmware update: For example, a new version of the LLN node 1019 software is downloaded from a system manager over unicast or 1020 multicast services. Such a reflashing operation typically 1021 involves updating a large number of similar LLN nodes over a 1022 relatively short period of time. 1024 Packages of Commands: A number of commands or a full 1025 configuration can be packaged as a single message to ensure 1026 consistency and enable atomic execution or complete roll back. 1027 Until such commands are fully received and interpreted, the 1028 intended operation will not take effect. 1030 From the LLN node: 1032 Waveform captures: A number of consecutive samples are measured 1033 at a high rate for a short time and then transferred from a 1034 sensor to a gateway or an edge server as a single large report. 1036 Data logs: LLN nodes may generate large logs of sampled data for 1037 later extraction. LLN nodes may also generate system logs to 1038 assist in diagnosing problems on the node or network. 1040 Large data packets: Rich data types might require more than one 1041 fragment. 1043 Uncontrolled firmware download or waveform upload can easily result 1044 in a massive increase of the traffic and saturate the network. 1046 When a fragment is lost in transmission, the lack of recovery in the 1047 original fragmentation system of RFC 4944 implies that all fragments 1048 would need to be resent, further contributing to the congestion that 1049 caused the initial loss, and potentially leading to congestion 1050 collapse. 1052 This saturation may lead to excessive radio interference, or random 1053 early discard (leaky bucket) in relaying nodes. Additional queuing 1054 and memory congestion may result while waiting for a low power next 1055 hop to emerge from its sleeping state. 1057 Considering that RFC 4944 defines an MTU is 1280 bytes and that in 1058 most incarnations (but 802.15.4g) a IEEE Std. 802.15.4 frame can 1059 limit the MAC payload to as few as 74 bytes, a packet might be 1060 fragmented into at least 18 fragments at the 6LoWPAN shim layer. 1061 Taking into account the worst-case header overhead for 6LoWPAN 1062 Fragmentation and Mesh Addressing headers will increase the number of 1063 required fragments to around 32. This level of fragmentation is much 1064 higher than that traditionally experienced over the Internet with 1065 IPv4 fragments. At the same time, the use of radios increases the 1066 probability of transmission loss and Mesh-Under techniques compound 1067 that risk over multiple hops. 1069 Mechanisms such as TCP or application-layer segmentation could be 1070 used to support end-to-end reliable transport. One option to support 1071 bulk data transfer over a frame-size-constrained LLN is to set the 1072 Maximum Segment Size to fit within the link maximum frame size. 1073 Doing so, however, can add significant header overhead to each 1074 802.15.4 frame. In addition, deploying such a mechanism requires 1075 that the end-to-end transport is aware of the delivery properties of 1076 the underlying LLN, which is a layer violation, and difficult to 1077 achieve from the far end of the IPv6 network. 1079 Appendix B. Requirements 1081 For one-hop communications, a number of Low Power and Lossy Network 1082 (LLN) link-layers propose a local acknowledgment mechanism that is 1083 enough to detect and recover the loss of fragments. In a multihop 1084 environment, an end-to-end fragment recovery mechanism might be a 1085 good complement to a hop-by-hop MAC level recovery. This draft 1086 introduces a simple protocol to recover individual fragments between 1087 6LoWPAN endpoints that may be multiple hops away. The method 1088 addresses the following requirements of a LLN: 1090 Number of fragments 1092 The recovery mechanism must support highly fragmented packets, 1093 with a maximum of 32 fragments per packet. 1095 Minimum acknowledgment overhead 1097 Because the radio is half duplex, and because of silent time spent 1098 in the various medium access mechanisms, an acknowledgment 1099 consumes roughly as many resources as data fragment. 1101 The new end-to-end fragment recovery mechanism should be able to 1102 acknowledge multiple fragments in a single message and not require 1103 an acknowledgment at all if fragments are already protected at a 1104 lower layer. 1106 Controlled latency 1108 The recovery mechanism must succeed or give up within the time 1109 boundary imposed by the recovery process of the Upper Layer 1110 Protocols. 1112 Optional congestion control 1114 The aggregation of multiple concurrent flows may lead to the 1115 saturation of the radio network and congestion collapse. 1117 The recovery mechanism should provide means for controlling the 1118 number of fragments in transit over the LLN. 1120 Appendix C. Considerations On Flow Control 1122 Considering that a multi-hop LLN can be a very sensitive environment 1123 due to the limited queuing capabilities of a large population of its 1124 nodes, this draft recommends a simple and conservative approach to 1125 Congestion Control, based on TCP congestion avoidance. 1127 Congestion on the forward path is assumed in case of packet loss, and 1128 packet loss is assumed upon time out. The draft allows to control 1129 the number of outstanding fragments, that have been transmitted but 1130 for which an acknowledgment was not received yet. It must be noted 1131 that the number of outstanding fragments should not exceed the number 1132 of hops in the network, but the way to figure the number of hops is 1133 out of scope for this document. 1135 Congestion on the forward path can also be indicated by an Explicit 1136 Congestion Notification (ECN) mechanism. Though whether and how ECN 1137 [RFC3168] is carried out over the LoWPAN is out of scope, this draft 1138 provides a way for the destination endpoint to echo an ECN indication 1139 back to the source endpoint in an acknowledgment message as 1140 represented in Figure 5 in Section 5.2. 1142 It must be noted that congestion and collision are different topics. 1143 In particular, when a mesh operates on a same channel over multiple 1144 hops, then the forwarding of a fragment over a certain hop may 1145 collide with the forwarding of a next fragment that is following over 1146 a previous hop but in a same interference domain. This draft enables 1147 an end-to-end flow control, but leaves it to the sender stack to pace 1148 individual fragments within a transmit window, so that a given 1149 fragment is sent only when the previous fragment has had a chance to 1150 progress beyond the interference domain of this hop. In the case of 1151 6TiSCH [I-D.ietf-6tisch-architecture], which operates over the 1152 TimeSlotted Channel Hopping [RFC7554] (TSCH) mode of operation of 1153 IEEE802.14.5, a fragment is forwarded over a different channel at a 1154 different time and it makes full sense to transmit the next fragment 1155 as soon as the previous fragment has had its chance to be forwarded 1156 at the next hop. 1158 From the standpoint of a source 6LoWPAN endpoint, an outstanding 1159 fragment is a fragment that was sent but for which no explicit 1160 acknowledgment was received yet. This means that the fragment might 1161 be on the way, received but not yet acknowledged, or the 1162 acknowledgment might be on the way back. It is also possible that 1163 either the fragment or the acknowledgment was lost on the way. 1165 From the sender standpoint, all outstanding fragments might still be 1166 in the network and contribute to its congestion. There is an 1167 assumption, though, that after a certain amount of time, a frame is 1168 either received or lost, so it is not causing congestion anymore. 1169 This amount of time can be estimated based on the round trip delay 1170 between the 6LoWPAN endpoints. The method detailed in [RFC6298] is 1171 recommended for that computation. 1173 The reader is encouraged to read through "Congestion Control 1174 Principles" [RFC2914]. Additionally [RFC7567] and [RFC5681] provide 1175 deeper information on why this mechanism is needed and how TCP 1176 handles Congestion Control. Basically, the goal here is to manage 1177 the amount of fragments present in the network; this is achieved by 1178 to reducing the number of outstanding fragments over a congested path 1179 by throttling the sources. 1181 Section 6 describes how the sender decides how many fragments are 1182 (re)sent before an acknowledgment is required, and how the sender 1183 adapts that number to the network conditions. 1185 Author's Address 1187 Pascal Thubert (editor) 1188 Cisco Systems, Inc 1189 Building D 1190 45 Allee des Ormes - BP1200 1191 MOUGINS - Sophia Antipolis 06254 1192 FRANCE 1194 Phone: +33 497 23 26 34 1195 Email: pthubert@cisco.com