idnits 2.17.00 (12 Aug 2021) /tmp/idnits26774/draft-ietf-6lo-fragment-recovery-10.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 (6 February 2020) is 835 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: draft-ietf-6lo-minimal-fragment has been published as RFC 8930 == Outdated reference: A later version (-02) exists of draft-ietf-lwig-6lowpan-virtual-reassembly-01 == Outdated reference: draft-ietf-intarea-frag-fragile has been published as RFC 8900 == Outdated reference: draft-ietf-6tisch-architecture has been published as RFC 9030 Summary: 0 errors (**), 0 flaws (~~), 5 warnings (==), 2 comments (--).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) 6 February 2020 5 Intended status: Standards Track 6 Expires: 9 August 2020 8 6LoWPAN Selective Fragment Recovery 9 draft-ietf-6lo-fragment-recovery-10 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 9 August 2020. 34 Copyright Notice 36 Copyright (c) 2020 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 (https://trustee.ietf.org/ 41 license-info) in effect on the date of publication of this document. 42 Please review these documents carefully, as they describe your rights 43 and restrictions with respect to this document. Code Components 44 extracted from this document must include Simplified BSD License text 45 as described in Section 4.e of the Trust Legal Provisions and are 46 provided without warranty as described in the Simplified BSD License. 48 Table of Contents 50 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 51 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 4 52 2.1. BCP 14 . . . . . . . . . . . . . . . . . . . . . . . . . 4 53 2.2. References . . . . . . . . . . . . . . . . . . . . . . . 4 54 2.3. New Terms . . . . . . . . . . . . . . . . . . . . . . . . 5 55 3. Updating RFC 4944 . . . . . . . . . . . . . . . . . . . . . . 6 56 4. Extending draft-ietf-6lo-minimal-fragment . . . . . . . . . . 6 57 4.1. Slack in the First Fragment . . . . . . . . . . . . . . . 7 58 4.2. Gap between frames . . . . . . . . . . . . . . . . . . . 7 59 4.3. Modifying the First Fragment . . . . . . . . . . . . . . 7 60 5. New Dispatch types and headers . . . . . . . . . . . . . . . 8 61 5.1. Recoverable Fragment Dispatch type and Header . . . . . . 8 62 5.2. RFRAG Acknowledgment Dispatch type and Header . . . . . . 11 63 6. Fragments Recovery . . . . . . . . . . . . . . . . . . . . . 12 64 6.1. Forwarding Fragments . . . . . . . . . . . . . . . . . . 14 65 6.1.1. Receiving the first fragment . . . . . . . . . . . . 15 66 6.1.2. Receiving the next fragments . . . . . . . . . . . . 15 67 6.2. Receiving RFRAG Acknowledgments . . . . . . . . . . . . . 16 68 6.3. Aborting the Transmission of a Fragmented Packet . . . . 16 69 6.4. Applying Recoverable Fragmentation along a Diverse 70 Path . . . . . . . . . . . . . . . . . . . . . . . . . . 17 71 7. Management Considerations . . . . . . . . . . . . . . . . . . 17 72 7.1. Protocol Parameters . . . . . . . . . . . . . . . . . . . 17 73 7.2. Observing the network . . . . . . . . . . . . . . . . . . 19 74 8. Security Considerations . . . . . . . . . . . . . . . . . . . 19 75 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 76 10. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 20 77 11. Normative References . . . . . . . . . . . . . . . . . . . . 20 78 12. Informative References . . . . . . . . . . . . . . . . . . . 21 79 Appendix A. Rationale . . . . . . . . . . . . . . . . . . . . . 24 80 Appendix B. Requirements . . . . . . . . . . . . . . . . . . . . 25 81 Appendix C. Considerations on Flow Control . . . . . . . . . . . 26 82 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 27 84 1. Introduction 86 In most Low Power and Lossy Network (LLN) applications, the bulk of 87 the traffic consists of small chunks of data (on the order of a few 88 bytes to a few tens of bytes) at a time. Given that an IEEE Std. 89 802.15.4 [IEEE.802.15.4] frame can carry a payload of 74 bytes or 90 more, fragmentation is usually not required. However, and though 91 this happens only occasionally, a number of mission critical 92 applications do require the capability to transfer larger chunks of 93 data, for instance to support the firmware upgrade of the LLN nodes 94 or the extraction of logs from LLN nodes. In the former case, the 95 large chunk of data is transferred to the LLN node, whereas in the 96 latter, the large chunk flows away from the LLN node. In both cases, 97 the size can be on the order of 10 kilobytes or more and an end-to- 98 end reliable transport is required. 100 "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944] 101 defines the original 6LoWPAN datagram fragmentation mechanism for 102 LLNs. One critical issue with this original design is that routing 103 an IPv6 [RFC8200] packet across a route-over mesh requires 104 reassembling the full packet at each hop, which may cause latency 105 along a path and an overall buffer bloat in the network. The "6TiSCH 106 Architecture" [I-D.ietf-6tisch-architecture] recommends using a 107 fragment forwarding (FF) technique to alleviate those undesirable 108 effects. "LLN Minimal Fragment Forwarding" 109 [I-D.ietf-6lo-minimal-fragment] specifies the general behavior that 110 all FF techniques including this specification follow, and presents 111 the associated caveats. In particular, the routing information is 112 fully indicated in the first fragment, which is always forwarded 113 first. A state is formed and used to forward all the next fragments 114 along the same path. The datagram_tag is locally significant to the 115 Layer-2 source of the packet and is swapped at each hop. 117 "Virtual reassembly buffers in 6LoWPAN" 118 [I-D.ietf-lwig-6lowpan-virtual-reassembly] (VRB) proposes a FF 119 technique that is compatible with [RFC4944] without the need to 120 define a new protocol. However, adding that capability alone to the 121 local implementation of the original 6LoWPAN fragmentation would not 122 address the inherent fragility of fragmentation (see 123 [I-D.ietf-intarea-frag-fragile]) in particular the issues of 124 resources locked on the receiver and the wasted transmissions due to 125 the loss of a single fragment in a whole datagram. [Kent] compares 126 the unreliable delivery of fragments with a mechanism it calls 127 "selective acknowledgements" that recovers the loss of a fragment 128 individually. The paper illustrates the benefits that can be derived 129 from such a method in figures 1, 2 and 3, on pages 6 and 7. 130 [RFC4944] as no selective recovery and the whole datagram fails when 131 one fragment is not delivered to the destination 6LoWPAN endpoint. 132 Constrained memory resources are blocked on the receiver until the 133 receiver times out, possibly causing the loss of subsequent packets 134 that cannot be received for the lack of buffers. 136 That problem is exacerbated when forwarding fragments over multiple 137 hops since a loss at an intermediate hop will not be discovered by 138 either the source or the destination, and the source will keep on 139 sending fragments, wasting even more resources in the network and 140 possibly contributing to the condition that caused the loss to no 141 avail since the datagram cannot arrive in its entirety. RFC 4944 is 142 also missing signaling to abort a multi-fragment transmission at any 143 time and from either end, and, if the capability to forward fragments 144 is implemented, clean up the related state in the network. It is 145 also lacking flow control capabilities to avoid participating in 146 congestion that may in turn cause the loss of a fragment and 147 potentially the retransmission of the full datagram. 149 This specification provides a method to forward fragments across a 150 multi-hop route-over mesh, and a selective acknowledgment to recover 151 individual fragments between 6LoWPAN endpoints. The method is 152 designed to limit congestion loss in the network and addresses the 153 requirements that are detailed in Appendix B. 155 2. Terminology 157 2.1. BCP 14 159 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 160 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 161 "OPTIONAL" in this document are to be interpreted as described in BCP 162 14 [RFC2119][RFC8174] when, and only when, they appear in all 163 capitals, as shown here. 165 2.2. References 167 In this document, readers will encounter terms and concepts that are 168 discussed in "Problem Statement and Requirements for IPv6 over 169 Low-Power Wireless Personal Area Network (6LoWPAN) Routing" [RFC6606] 171 "LLN Minimal Fragment Forwarding" [I-D.ietf-6lo-minimal-fragment] 172 introduces the generic concept of a Virtual Reassembly Buffer (VRB) 173 and specifies behaviours and caveats that are common to a large 174 family of FF techniques including this, which fully inherits from 175 that specification. 177 Past experience with fragmentation has shown that misassociated or 178 lost fragments can lead to poor network behavior and, occasionally, 179 trouble at the application layer. The reader is encouraged to read 180 "IPv4 Reassembly Errors at High Data Rates" [RFC4963] and follow the 181 references for more information. 183 That experience led to the definition of "Path MTU discovery" 184 [RFC8201] (PMTUD) protocol that limits fragmentation over the 185 Internet. 187 Specifically in the case of UDP, valuable additional information can 188 be found in "UDP Usage Guidelines for Application Designers" 189 [RFC8085]. 191 Readers are expected to be familiar with all the terms and concepts 192 that are discussed in "IPv6 over Low-Power Wireless Personal Area 193 Networks (6LoWPANs): Overview, Assumptions, Problem Statement, and 194 Goals" [RFC4919] and "Transmission of IPv6 Packets over IEEE 802.15.4 195 Networks" [RFC4944]. 197 "The Benefits of Using Explicit Congestion Notification (ECN)" 198 [RFC8087] provides useful information on the potential benefits and 199 pitfalls of using ECN. 201 Quoting the "Multiprotocol Label Switching (MPLS) Architecture" 202 [RFC3031]: with MPLS, 'packets are "labeled" before they are 203 forwarded' along a Label Switched Path (LSP). At subsequent hops, 204 there is no further analysis of the packet's network layer header. 205 Rather, the label is used as an index into a table which specifies 206 the next hop, and a new label". The MPLS technique is leveraged in 207 the present specification to forward fragments that actually do not 208 have a network layer header, since the fragmentation occurs below IP. 210 2.3. New Terms 212 This specification uses the following terms: 214 6LoWPAN endpoints: The LLN nodes in charge of generating or 215 expanding a 6LoWPAN header from/to a full IPv6 packet. The 216 6LoWPAN endpoints are the points where fragmentation and 217 reassembly take place. 219 Compressed Form: This specification uses the generic term Compressed 220 Form to refer to the format of a datagram after the action of 221 [RFC6282] and possibly [RFC8138] for RPL [RFC6550] artifacts. 223 datagram_size: The size of the datagram in its Compressed Form 224 before it is fragmented. The datagram_size is expressed in a unit 225 that depends on the MAC layer technology, by default a byte. 227 datagram_tag: An identifier of a datagram that is locally unique to 228 the Layer-2 sender. Associated with the MAC address of the 229 sender, this becomes a globally unique identifier for the 230 datagram. 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 RFRAG: Recoverable Fragment 238 RFRAG-ACK: Recoverable Fragment Acknowledgement 240 RFRAG Acknowledgment Request: An RFRAG with the Acknowledgement 241 Request flag ('X' flag) set. 243 NULL bitmap: Refers to a bitmap with all bits set to zero. 245 FULL bitmap: Refers to a bitmap with all bits set to one. 247 Forward: The direction of a LSP path, followed by the RFRAG. 249 Reverse: The reverse direction of a LSP path, taken by the RFRAG- 250 ACK. 252 3. Updating RFC 4944 254 This specification updates the fragmentation mechanism that is 255 specified in "Transmission of IPv6 Packets over IEEE 802.15.4 256 Networks" [RFC4944] for use in route-over LLNs by providing a model 257 where fragments can be forwarded end-to-end across a 6LoWPAN LLN, and 258 where fragments that are lost on the way can be recovered 259 individually. A new format for fragments is introduced and new 260 dispatch types are defined in Section 5. 262 [RFC8138] allows modifying the size of a packet en route by removing 263 the consumed hops in a compressed Routing Header. This requires that 264 fragment_offset and datagram_size (see Section 2.3) are also modified 265 en route, which is difficult to do in the uncompressed form. This 266 specification expresses those fields in the Compressed Form and 267 allows modifying them en route (see Section 4.3) easily. 269 Note that consistent with Section 2 of [RFC6282], for the 270 fragmentation mechanism described in Section 5.3 of [RFC4944], any 271 header that cannot fit within the first fragment MUST NOT be 272 compressed when using the fragmentation mechanism described in this 273 specification. 275 4. Extending draft-ietf-6lo-minimal-fragment 277 This specification implements the generic FF technique specified in 278 "LLN Minimal Fragment Forwarding" [I-D.ietf-6lo-minimal-fragment] in 279 a fashion that enables end-to-end recovery of fragments and some 280 degree of flow control. 282 4.1. Slack in the First Fragment 284 [I-D.ietf-6lo-minimal-fragment] allows for refragmenting in 285 intermediate nodes, meaning that some bytes from a given fragment may 286 be left in the VRB to be added to the next fragment. The reason for 287 this happening would be the need for space in the outgoing fragment 288 that was not needed in the incoming fragment, for instance because 289 the 6LoWPAN Header Compression is not as efficient on the outgoing 290 link, e.g., if the Interface ID (IID) of the source IPv6 address is 291 elided by the originator on the first hop because it matches the 292 source MAC address, but cannot be on the next hops because the source 293 MAC address changes. 295 This specification cannot allow this operation since fragments are 296 recovered end-to-end based on a sequence number. This means that the 297 fragments that contain a 6LoWPAN-compressed header MUST have enough 298 slack to enable a less efficient compression in the next hops that 299 still fits in one MAC frame. For instance, if the IID of the source 300 IPv6 address is elided by the originator, then it MUST compute the 301 fragment_size as if the MTU was 8 bytes less. This way, the next hop 302 can restore the source IID to the first fragment without impacting 303 the second fragment. 305 4.2. Gap between frames 307 This specification introduces a concept of an inter-frame gap, which 308 is a configurable interval of time between transmissions to a same 309 next hop. In the case of half duplex interfaces, this inter-frame 310 gap ensures that the next hop has completed processing of the 311 previous frame and is capable of receiving the next one. 313 In the case of a mesh operating at a single frequency with 314 omnidirectional antennas, a larger inter-frame gap is required to 315 protect the frame against hidden terminal collisions with the 316 previous frame of a same flow that is still progressing along a 317 common path. 319 The inter-frame gap is useful even for unfragmented datagrams, but it 320 becomes a necessity for fragments that are typically generated in a 321 fast sequence and are all sent over the exact same path. 323 4.3. Modifying the First Fragment 325 The compression of the Hop Limit, of the source and destination 326 addresses in the IPv6 Header, and of the Routing Header may change en 327 route in a Route-Over mesh LLN. If the size of the first fragment is 328 modified, then the intermediate node MUST adapt the datagram_size to 329 reflect that difference. 331 The intermediate node MUST also save the difference of datagram_size 332 of the first fragment in the VRB and add it to the datagram_size and 333 to the fragment_offset of all the subsequent fragments for that 334 datagram. 336 5. New Dispatch types and headers 338 This specification enables the 6LoWPAN fragmentation sublayer to 339 provide an MTU up to 2048 bytes to the upper layer, which can be the 340 6LoWPAN Header Compression sublayer that is defined in the 341 "Compression Format for IPv6 Datagrams" [RFC6282] specification. In 342 order to achieve this, this specification enables the fragmentation 343 and the reliable transmission of fragments over a multihop 6LoWPAN 344 mesh network. 346 This specification provides a technique that is derived from MPLS to 347 forward individual fragments across a 6LoWPAN route-over mesh without 348 reassembly at each hop. The datagram_tag is used as a label; it is 349 locally unique to the node that owns the source MAC address of the 350 fragment, so together the MAC address and the label can identify the 351 fragment globally. A node may build the datagram_tag in its own 352 locally-significant way, as long as the chosen datagram_tag stays 353 unique to the particular datagram for the lifetime of that datagram. 354 It results that the label does not need to be globally unique but 355 also that it must be swapped at each hop as the source MAC address 356 changes. 358 This specification extends RFC 4944 [RFC4944] with 2 new Dispatch 359 types, for Recoverable Fragment (RFRAG) and for the RFRAG 360 Acknowledgment back. The new 6LoWPAN Dispatch types are taken from 361 Page 0 [RFC8025] as indicated in Table 1 in Section 9. 363 In the following sections, a "datagram_tag" extends the semantics 364 defined in [RFC4944] Section 5.3."Fragmentation Type and Header". 365 The datagram_tag is a locally unique identifier for the datagram from 366 the perspective of the sender. This means that the datagram_tag 367 identifies a datagram uniquely in the network when associated with 368 the source of the datagram. As the datagram gets forwarded, the 369 source changes and the datagram_tag must be swapped as detailed in 370 [I-D.ietf-6lo-minimal-fragment]. 372 5.1. Recoverable Fragment Dispatch type and Header 374 In this specification, if the packet is compressed then the size and 375 offset of the fragments are expressed with respect to the Compressed 376 Form of the packet form as opposed to the uncompressed (native) 377 packet form. 379 The format of the fragment header is shown in Figure 1. It is the 380 same for all fragments. The format has a length and an offset, as 381 well as a sequence field. This would be redundant if the offset was 382 computed as the product of the sequence by the length, but this is 383 not the case. The position of a fragment in the reassembly buffer is 384 neither correlated with the value of the sequence field nor with the 385 order in which the fragments are received. This enables out-of- 386 sequence subfragmenting, e.g., a fragment seq. 5 that is retried end- 387 to-end as smaller fragments seq. 5, 13 and 14 due to a change of MTU 388 along the path between the 6LoWPAN endpoints. 390 1 2 3 391 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 1 392 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 393 |1 1 1 0 1 0 0|E| datagram_tag | 394 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 395 |X| sequence| fragment_size | fragment_offset | 396 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 398 X set == Ack-Request 400 Figure 1: RFRAG Dispatch type and Header 402 There is no requirement on the receiver to check for contiguity of 403 the received fragments, and the sender MUST ensure that when all 404 fragments are acknowledged, then the datagram is fully received. 405 This may be useful in particular in the case where the MTU changes 406 and a fragment sequence is retried with a smaller fragment_size, the 407 remainder of the original fragment being retried with new sequence 408 values. 410 The first fragment is recognized by a sequence of 0; it carries its 411 fragment_size and the datagram_size of the compressed packet before 412 it is fragmented, whereas the other fragments carry their 413 fragment_size and fragment_offset. The last fragment for a datagram 414 is recognized when its fragment_offset and its fragment_size add up 415 to the datagram_size. 417 Recoverable Fragments are sequenced and a bitmap is used in the RFRAG 418 Acknowledgment to indicate the received fragments by setting the 419 individual bits that correspond to their sequence. 421 X: 1 bit; Ack-Request: when set, the sender requires an RFRAG 422 Acknowledgment from the receiver. 424 E: 1 bit; Explicit Congestion Notification; the "E" flag is reset by 425 the source of the fragment and set by intermediate routers to 426 signal that this fragment experienced congestion along its path. 428 Fragment_size: 10-bit unsigned integer; the size of this fragment in 429 a unit that depends on the MAC layer technology. Unless 430 overridden by a more specific specification, that unit is the 431 octet, which allows fragments up to 512 bytes. 433 datagram_tag: 8 bits; an identifier of the datagram that is locally 434 unique to the sender. 436 Sequence: 5-bit unsigned integer; the sequence number of the 437 fragment in the acknowledgement bitmap. Fragments are numbered 438 [0..N] where N is in [0..31]. A Sequence of 0 indicates the first 439 fragment in a datagram, but non-zero values are not indicative of 440 the position in the reassembly buffer. 442 Fragment_offset: 16-bit unsigned integer. 444 When the Fragment_offset is set to a non-0 value, its semantics 445 depend on the value of the Sequence field as follows: 447 * For a first fragment (i.e., with a Sequence of 0), this field 448 indicates the datagram_size of the compressed datagram, to help 449 the receiver allocate an adapted buffer for the reception and 450 reassembly operations. The fragment may be stored for local 451 reassembly. Alternatively, it may be routed based on the 452 destination IPv6 address. In that case, a VRB state must be 453 installed as described in Section 6.1.1. 454 * When the Sequence is not 0, this field indicates the offset of 455 the fragment in the Compressed Form of the datagram. The 456 fragment may be added to a local reassembly buffer or forwarded 457 based on an existing VRB as described in Section 6.1.2. 459 A Fragment_offset that is set to a value of 0 indicates an abort 460 condition and all state regarding the datagram should be cleaned 461 up once the processing of the fragment is complete; the processing 462 of the fragment depends on whether there is a VRB already 463 established for this datagram, and the next hop is still 464 reachable: 466 * if a VRB already exists and is not broken, the fragment is to 467 be forwarded along the associated Label Switched Path (LSP) as 468 described in Section 6.1.2, but regardless of the value of the 469 Sequence field; 470 * else, if the Sequence is 0, then the fragment is to be routed 471 as described in Section 6.1.1, but no state is conserved 472 afterwards. In that case, the session if it exists is aborted 473 and the packet is also forwarded in an attempt to clean up the 474 next hops along the path indicated by the IPv6 header (possibly 475 including a routing header). 477 If the fragment cannot be forwarded or routed, then an abort 478 RFRAG-ACK is sent back to the source as described in 479 Section 6.1.2. 481 5.2. RFRAG Acknowledgment Dispatch type and Header 483 This specification also defines a 4-octet RFRAG Acknowledgment bitmap 484 that is used by the reassembling endpoint to confirm selectively the 485 reception of individual fragments. A given offset in the bitmap maps 486 one-to-one with a given sequence number and indicates which fragment 487 is acknowledged as follows: 489 1 2 3 490 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 1 491 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 492 | RFRAG Acknowledgment Bitmap | 493 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 494 ^ ^ 495 | | bitmap indicating whether: 496 | +----- Fragment with sequence 9 was received 497 +----------------------- Fragment with sequence 0 was received 499 Figure 2: RFRAG Acknowledgment Bitmap Encoding 501 Figure 3 shows an example Acknowledgment bitmap which indicates that 502 all fragments from sequence 0 to 20 were received, except for 503 fragments 1, 2 and 16 were lost and must be retried. 505 1 2 3 506 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 1 507 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 508 |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| 509 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 511 Figure 3: Example RFRAG Acknowledgment Bitmap 513 The RFRAG Acknowledgment Bitmap is included in an RFRAG 514 Acknowledgment header, as follows: 516 1 2 3 517 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 1 518 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 519 |1 1 1 0 1 0 1|E| datagram_tag | 520 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 521 | RFRAG Acknowledgment Bitmap (32 bits) | 522 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 523 Figure 4: RFRAG Acknowledgment Dispatch type and Header 525 E: 1 bit; Explicit Congestion Notification Echo 527 When set, the sender indicates that at least one of the 528 acknowledged fragments was received with an Explicit Congestion 529 Notification, indicating that the path followed by the fragments 530 is subject to congestion. More in Appendix C. 532 RFRAG Acknowledgment Bitmap: An RFRAG Acknowledgment Bitmap, whereby 533 setting the bit at offset x indicates that fragment x was 534 received, as shown in Figure 2. A NULL bitmap that indicates that 535 the fragmentation process is aborted. A FULL bitmap that 536 indicates that the fragmentation process is complete, all 537 fragments were received at the reassembly endpoint. 539 6. Fragments Recovery 541 The Recoverable Fragment header RFRAG is used to transport a fragment 542 and optionally request an RFRAG Acknowledgment that will confirm the 543 good reception of one or more fragments. An RFRAG Acknowledgment is 544 carried as a standalone fragment header (i.e., with no 6LoWPAN 545 payload) in a message that is propagated back to the 6LoWPAN endpoint 546 that was the originator of the fragments. To achieve this, each hop 547 that performed an MPLS-like operation on fragments reverses that 548 operation for the RFRAG_ACK by sending a frame from the next hop to 549 the previous hop as known by its MAC address in the VRB. The 550 datagram_tag in the RFRAG_ACK is unique to the receiver and is enough 551 information for an intermediate hop to locate the VRB that contains 552 the datagram_tag used by the previous hop and the Layer-2 information 553 associated to it (interface and MAC address). 555 The 6LoWPAN endpoint that fragments the packets at the 6LoWPAN level 556 (the sender) also controls the amount of acknowledgments by setting 557 the Ack-Request flag in the RFRAG packets. The sender may set the 558 Ack-Request flag on any fragment to perform congestion control by 559 limiting the number of outstanding fragments, which are the fragments 560 that have been sent but for which reception or loss was not 561 positively confirmed by the reassembling endpoint. The maximum 562 number of outstanding fragments is the Window-Size. It is 563 configurable and may vary in case of ECN notification. When the 564 6LoWPAN endpoint that reassembles the packets at the 6LoWPAN level 565 (the receiver) receives a fragment with the Ack-Request flag set, it 566 MUST send an RFRAG Acknowledgment back to the originator to confirm 567 reception of all the fragments it has received so far. 569 The Ack-Request ('X') set in an RFRAG marks the end of a window. 570 This flag MUST be set on the last fragment if the sender wishes to 571 protect the datagram, and it MAY be set in any intermediate fragment 572 for the purpose of flow control. 574 This automatic repeat request (ARQ) process MUST be protected by a 575 timer, and the fragment that carries the 'X' flag MAY be retried upon 576 a time out for a configurable number of times (see Section 7.1). 577 Upon exhaustion of the retries the sender may either abort the 578 transmission of the datagram or retry the datagram from the first 579 fragment with an 'X' flag set in order to reestablish a path and 580 discover which fragments were received over the old path in the 581 acknowledgment bitmap. When the sender of the fragment knows that an 582 underlying link-layer mechanism protects the fragments, it may 583 refrain from using the RFRAG Acknowledgment mechanism, and never set 584 the Ack-Request bit. 586 The receiver MAY issue unsolicited acknowledgments. An unsolicited 587 acknowledgment signals to the sender endpoint that it can resume 588 sending if it had reached its maximum number of outstanding 589 fragments. Another use is to inform the sender that the reassembling 590 endpoint aborted the processing of an individual datagram. 592 The RFRAG Acknowledgment can optionally carry an ECN indication for 593 flow control (see Appendix C). The receiver of a fragment with the 594 'E' (ECN) flag set MUST echo that information by setting the 'E' 595 (ECN) flag in the next RFRAG Acknowledgment. 597 In order to protect the datagram, the sender transfers a controlled 598 number of fragments and flags the last fragment of a window with an 599 RFRAG Acknowledgment Request. The receiver MUST acknowledge a 600 fragment with the acknowledgment request bit set. If any fragment 601 immediately preceding an acknowledgment request is still missing, the 602 receiver MAY intentionally delay its acknowledgment to allow in- 603 transit fragments to arrive. Because it might defeat the round-trip 604 delay computation, delaying the acknowledgment should be configurable 605 and not enabled by default. 607 When all the fragments are received, the receiving endpoint 608 reconstructs the packet, passes it to the upper layer, sends an RFRAG 609 Acknowledgment on the reverse path with a FULL bitmap, and arms a 610 short timer, e.g., in the order of an average round-trip delay in the 611 network. As the timer runs, the receiving endpoint absorbs the 612 fragments that were still in flight for that datagram without 613 creating a new state. The receiving endpoint abort the communication 614 if it keeps going on beyond the duration of the timer. 616 Note that acknowledgments might consume precious resources so the use 617 of unsolicited acknowledgments should be configurable and not enabled 618 by default. 620 An observation is that streamlining forwarding of fragments generally 621 reduces the latency over the LLN mesh, providing room for retries 622 within existing upper-layer reliability mechanisms. The sender 623 protects the transmission over the LLN mesh with a retry timer that 624 is computed according to the method detailed in [RFC6298]. It is 625 expected that the upper layer retries obey the recommendations in 626 "UDP Usage Guidelines" [RFC8085], in which case a single round of 627 fragment recovery should fit within the upper layer recovery timers. 629 Fragments are sent in a round-robin fashion: the sender sends all the 630 fragments for a first time before it retries any lost fragment; lost 631 fragments are retried in sequence, oldest first. This mechanism 632 enables the receiver to acknowledge fragments that were delayed in 633 the network before they are retried. 635 When a single frequency is used by contiguous hops, the sender should 636 insert a delay between fragments of a same datagram that covers 637 multiple transmissions so as to let a fragment progress a few hops 638 and avoid hidden terminal issues. This precaution is not required on 639 channel hopping technologies such as Time Slotted Channel Hopping 640 (TSCH) [RFC6554], where nodes that communicate at Layer-2 are 641 scheduled to send and receive respectively, and different hops 642 operate on different channels. 644 6.1. Forwarding Fragments 646 It is assumed that the first fragment is large enough to carry the 647 IPv6 header and make routing decisions. If that is not so, then this 648 specification MUST NOT be used. 650 This specification extends the Virtual Reassembly Buffer (VRB) 651 technique to forward fragments with no intermediate reconstruction of 652 the entire packet. It inherits operations like datagram_tag 653 switching and using a timer to clean the VRB when the traffic dries 654 up. The first fragment carries the IP header and it is routed all 655 the way from the fragmenting endpoint to the reassembling endpoint. 656 Upon receiving the first fragment, the routers along the path install 657 a label-switched path (LSP), and the following fragments are label- 658 switched along that path. As a consequence, the next fragments can 659 only follow the path that was set up by the first fragment and cannot 660 follow an alternate route. The datagram_tag is used to carry the 661 label, which is swapped in each hop. All fragments follow the same 662 path and fragments are delivered in the order at which they are sent. 664 6.1.1. Receiving the first fragment 666 In Route-Over mode, the source and destination MAC addresses in a 667 frame change at each hop. The label that is formed and placed in the 668 datagram_tag is associated with the source MAC address and only valid 669 (and unique) for that source MAC address. Upon a first fragment 670 (i.e., with a sequence of zero), an intermediate router creates a VRB 671 and the associated LSP state for the tuple (source MAC address, 672 datagram_tag) and the fragment is forwarded along the IPv6 route that 673 matches the destination IPv6 address in the IPv6 header as prescribed 674 by [I-D.ietf-6lo-minimal-fragment], where the receiving endpoint 675 allocates a reassembly buffer. 677 The LSP state enables to match the (previous MAC address, 678 datagram_tag) in an incoming fragment to the tuple (next MAC address, 679 swapped datagram_tag) used in the forwarded fragment and points at 680 the VRB. In addition, the router also forms a reverse LSP state 681 indexed by the MAC address of the next hop and the swapped 682 datagram_tag. This reverse LSP state also points at the VRB and 683 enables matching the (next MAC address, swapped_datagram_tag) found 684 in an RFRAG Acknowledgment to the tuple (previous MAC address, 685 datagram_tag) used when forwarding a Fragment Acknowledgment (RFRAG- 686 ACK) back to the sender endpoint. 688 The first fragment may be received a second time, indicating that it 689 did not reach the destination and was retried. In that case, it 690 SHOULD follow the same path as the first occurrence. It is up to 691 sending endpoint to determine whether to abort a transmission and 692 then retry it from scratch, which may build an entirely new path. 694 6.1.2. Receiving the next fragments 696 Upon receiving a next fragment (i.e., with a non-zero sequence), an 697 intermediate router looks up a LSP indexed by the tuple (MAC address, 698 datagram_tag) found in the fragment. If it is found, the router 699 forwards the fragment using the associated VRB as prescribed by 700 [I-D.ietf-6lo-minimal-fragment]. 702 If the VRB for the tuple is not found, the router builds an RFRAG-ACK 703 to abort the transmission of the packet. The resulting message has 704 the following information: 706 * The source and destination MAC addresses are swapped from those 707 found in the fragment 708 * The datagram_tag is set to the datagram_tag found in the fragment 709 * A NULL bitmap is used to signal the abort condition 710 At this point the router is all set and can send the RFRAG-ACK back 711 to the previous router. The RFRAG-ACK should normally be forwarded 712 all the way to the source using the reverse LSP state in the VRBs in 713 the intermediate routers as described in the next section. 715 [I-D.ietf-6lo-minimal-fragment] indicates that the receiving endpoint 716 stores "the actual packet data from the fragments received so far, in 717 a form that makes it possible to detect when the whole packet has 718 been received and can be processed or forwarded". How this is 719 computed in implementation specific but relies on receiving all the 720 bytes up to the datagram_size indicated in the first fragment. An 721 implementation may receive overlapping fragments as the result of 722 retries after an MTU change. 724 6.2. Receiving RFRAG Acknowledgments 726 Upon receipt of an RFRAG-ACK, the router looks up a reverse LSP 727 indexed by the tuple (MAC address, datagram_tag), which are 728 respectively the source MAC address of the received frame and the 729 received datagram_tag. If it is found, the router forwards the 730 fragment using the associated VRB as prescribed by 731 [I-D.ietf-6lo-minimal-fragment], but using the reverse LSP so that 732 the RFRAG-ACK flows back to the sender endpoint. 734 If the reverse LSP is not found, the router MUST silently drop the 735 RFRAG-ACK message. 737 Either way, if the RFRAG-ACK indicates that the fragment was entirely 738 received (FULL bitmap), it arms a short timer, and upon timeout, the 739 VRB and all the associated state are destroyed. Until the timer 740 elapses, fragments of that datagram may still be received, e.g. if 741 the RFRAG-ACK was lost on the way back and the source retried the 742 last fragment. In that case, the router forwards the fragment 743 according to the state in the VRB. 745 This specification does not provide a method to discover the number 746 of hops or the minimal value of MTU along those hops. But should the 747 minimal MTU decrease, it is possible to retry a long fragment (say 748 sequence of 5) with first a shorter fragment of the same sequence (5 749 again) and then one or more other fragments with a sequence that was 750 not used before (e.g., 13 and 14). Note that Path MTU Discovery is 751 out of scope for this document. 753 6.3. Aborting the Transmission of a Fragmented Packet 755 A reset is signaled on the forward path with a pseudo fragment that 756 has the fragment_offset, sequence, and fragment_size all set to 0, 757 and no data. 759 When the sender or a router on the way decides that a packet should 760 be dropped and the fragmentation process aborted, it generates a 761 reset pseudo fragment and forwards it down the fragment path. 763 Each router next along the path the way forwards the pseudo fragment 764 based on the VRB state. If an acknowledgment is not requested, the 765 VRB and all associated state are destroyed. 767 Upon reception of the pseudo fragment, the receiver cleans up all 768 resources for the packet associated with the datagram_tag. If an 769 acknowledgment is requested, the receiver responds with a NULL 770 bitmap. 772 The other way around, the receiver might need to abort the process of 773 a fragmented packet for internal reasons, for instance if it is out 774 of reassembly buffers, already uses all 256 possible values of the 775 datagram_tag, or if it keeps receiving fragments beyond a reasonable 776 time while it considers that this packet is already fully reassembled 777 and was passed to the upper layer. In that case, the receiver SHOULD 778 indicate so to the sender with a NULL bitmap in an RFRAG 779 Acknowledgment. The RFRAG Acknowledgment is forwarded all the way 780 back to the source of the packet and cleans up all resources on the 781 way. Upon an acknowledgment with a NULL bitmap, the sender endpoint 782 MUST abort the transmission of the fragmented datagram with one 783 exception: In the particular case of the first fragment, it MAY 784 decide to retry via an alternate next hop instead. 786 6.4. Applying Recoverable Fragmentation along a Diverse Path 788 The text above can be read with the assumption of a serial path 789 between a source and a destination. Section 4.5.3 of the "6TiSCH 790 Architecture" [I-D.ietf-6tisch-architecture] defines the concept of a 791 Track that can be a complex path between a source and a destination 792 with Packet ARQ, Replication, Elimination and Overhearing (PAREO) 793 along the Track. This specification can be used along any subset of 794 the complex Track where the first fragment is flooded. The last 795 RFRAG Acknowledgment is flooded on that same subset in the reverse 796 direction. Intermediate RFRAG Acknowledgments can be flooded on any 797 sub-subset of that reverse subset that reach back to the source. 799 7. Management Considerations 801 7.1. Protocol Parameters 803 There is no particular configuration on the receiver, as echoing ECN 804 is always on. The configuration only applies to the sender, which is 805 in control of the transmission. The management system SHOULD be 806 capable of providing the parameters below: 808 MinFragmentSize: The MinFragmentSize is the minimum value for the 809 Fragment_Size. 811 OptFragmentSize: The MinFragmentSize is the value for the 812 Fragment_Size that the sender should use to start with. 814 MaxFragmentSize: The MaxFragmentSize is the maximum value for the 815 Fragment_Size. It MUST be lower than the minimum MTU along the 816 path. A large value augments the chances of buffer bloat and 817 transmission loss. The value MUST be less than 512 if the unit 818 that is defined for the PHY layer is the octet. 820 UseECN: Indicates whether the sender should react to ECN. When the 821 sender reacts to ECN the Window_Size will vary between 822 MinWindowSize and MaxWindowSize. 824 MinWindowSize: The minimum value of Window_Size that the sender can 825 use. 827 OptWindowSize: The OptWindowSize is the value for the Window_Size 828 that the sender should use to start with. 830 MaxWindowSize: The maximum value of Window_Size that the sender can 831 use. The value MUSt be less than 32. 833 inter-frame gap: Indicates a minimum amount of time between 834 transmissions. All packets to a same destination, and in 835 particular fragments, may be subject to receive while transmitting 836 and hidden terminal collisions with the next or the previous 837 transmission as the fragments progress along a same path. The 838 inter-frame gap protects the propagation of one transmission 839 before the next one is triggered and creates a duty cycle that 840 controls the ratio of air time and memory in intermediate nodes 841 that a particular datagram will use. 843 MinARQTimeOut: The maximum amount of time a node should wait for an 844 RFRAG Acknowledgment before it takes a next action. 846 OptARQTimeOut: The starting point of the value of the amount of time 847 that a sender should wait for an RFRAG Acknowledgment before it 848 takes a next action. 850 MaxARQTimeOut: The maximum amount of time a node should wait for an 851 RFRAG Acknowledgment before it takes a next action. 853 MaxFragRetries: The maximum number of retries for a particular 854 fragment. 856 MaxDatagramRetries: The maximum number of retries from scratch for a 857 particular datagram. 859 7.2. Observing the network 861 The management system should monitor the amount of retries and of ECN 862 settings that can be observed from the perspective of both the sender 863 and the receiver, and may tune the optimum size of Fragment_Size and 864 of the Window_Size, OptDatagramSize and OptWindowSize respectively, 865 at the sender. The values should be bounded by the expected number 866 of hops and reduced beyond that when the number of datagrams that can 867 traverse an intermediate point may exceed its capacity and cause a 868 congestion loss. The inter-frame gap is another tool that can be 869 used to increase the spacing between fragments of the same datagram 870 and reduce the ratio of time when a particular intermediate node 871 holds a fragment of that datagram. 873 8. Security Considerations 875 The considerations in the Security sections of [I-D.ietf-core-cocoa] 876 and [I-D.ietf-6lo-minimal-fragment] apply equally to this 877 specification. 879 The process of recovering fragments does not appear to create any 880 opening for new threat compared to "Transmission of IPv6 Packets over 881 IEEE 802.15.4 Networks" [RFC4944] beyond the change of size of the 882 datagram_tag. By being reduced to 8 bits, the tag will wrap faster 883 than with [RFC4944]. But for a constrained network where a node is 884 expected to be able to hold only one or a few large packets in 885 memory, 256 is still a large number. Also, the acknowledgement 886 mechanism allows cleaning up the state rapidly once the packet is 887 fully transmitted or aborted. 889 The abstract Virtual Recovery Buffer inherited from 890 [I-D.ietf-6lo-minimal-fragment] may be used to perform a Denial-of- 891 Service (DoS) attack against the intermediate Routers since the 892 routers need to maintain a state per flow. The particular VRB 893 implementation technique described in 894 [I-D.ietf-lwig-6lowpan-virtual-reassembly] allows realigning which 895 data goes in which fragment, which causes the intermediate node to 896 store a portion of the data, which adds an attack vector that is not 897 present with this specification. With this specification, the data 898 that is transported in each fragment is conserved and the state to 899 keep does not include any data that would not fit in the previous 900 fragment. 902 9. IANA Considerations 904 This document allocates 2 patterns for a total of 4 dispatch values 905 in Page 0 for recoverable fragments from the "Dispatch Type Field" 906 registry that was created by "Transmission of IPv6 Packets over IEEE 907 802.15.4 Networks" [RFC4944] and reformatted by "6LoWPAN Paging 908 Dispatch" [RFC8025]. 910 The suggested patterns (to be confirmed by IANA) are indicated in 911 Table 1. 913 +-------------+------+----------------------------------+-----------+ 914 | Bit Pattern | Page | Header Type | Reference | 915 +=============+======+==================================+===========+ 916 | 11 10100x | 0 | RFRAG - Recoverable Fragment | THIS RFC | 917 +-------------+------+----------------------------------+-----------+ 918 | 11 10100x | 1-14 | Unassigned | | 919 +-------------+------+----------------------------------+-----------+ 920 | 11 10100x | 15 | Reserved for Experimental Use | RFC 8025 | 921 +-------------+------+----------------------------------+-----------+ 922 | 11 10101x | 0 | RFRAG-ACK - RFRAG | THIS RFC | 923 | | | Acknowledgment | | 924 +-------------+------+----------------------------------+-----------+ 925 | 11 10101x | 1-14 | Unassigned | | 926 +-------------+------+----------------------------------+-----------+ 927 | 11 10101x | 15 | Reserved for Experimental Use | RFC 8025 | 928 +-------------+------+----------------------------------+-----------+ 930 Table 1: Additional Dispatch Value Bit Patterns 932 10. Acknowledgments 934 The author wishes to thank Michel Veillette, Dario Tedeschi, Laurent 935 Toutain, Carles Gomez Montenegro, Thomas Watteyne, and Michael 936 Richardson for in-depth reviews and comments. Also many thanks to 937 Peter Yee and Erik Nordmark for their careful reviews and for helping 938 through the IESG review process, and to Jonathan Hui, Jay Werb, 939 Christos Polyzois, Soumitri Kolavennu, Pat Kinney, Margaret 940 Wasserman, Richard Kelsey, Carsten Bormann, and Harry Courtice for 941 their various contributions. 943 11. Normative References 945 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 946 Requirement Levels", BCP 14, RFC 2119, 947 DOI 10.17487/RFC2119, March 1997, 948 . 950 [RFC4944] Montenegro, G., Kushalnagar, N., Hui, J., and D. Culler, 951 "Transmission of IPv6 Packets over IEEE 802.15.4 952 Networks", RFC 4944, DOI 10.17487/RFC4944, September 2007, 953 . 955 [RFC6282] Hui, J., Ed. and P. Thubert, "Compression Format for IPv6 956 Datagrams over IEEE 802.15.4-Based Networks", RFC 6282, 957 DOI 10.17487/RFC6282, September 2011, 958 . 960 [RFC6554] Hui, J., Vasseur, JP., Culler, D., and V. Manral, "An IPv6 961 Routing Header for Source Routes with the Routing Protocol 962 for Low-Power and Lossy Networks (RPL)", RFC 6554, 963 DOI 10.17487/RFC6554, March 2012, 964 . 966 [RFC8025] Thubert, P., Ed. and R. Cragie, "IPv6 over Low-Power 967 Wireless Personal Area Network (6LoWPAN) Paging Dispatch", 968 RFC 8025, DOI 10.17487/RFC8025, November 2016, 969 . 971 [RFC8138] Thubert, P., Ed., Bormann, C., Toutain, L., and R. Cragie, 972 "IPv6 over Low-Power Wireless Personal Area Network 973 (6LoWPAN) Routing Header", RFC 8138, DOI 10.17487/RFC8138, 974 April 2017, . 976 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 977 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 978 May 2017, . 980 [I-D.ietf-6lo-minimal-fragment] 981 Watteyne, T., Thubert, P., and C. Bormann, "On Forwarding 982 6LoWPAN Fragments over a Multihop IPv6 Network", Work in 983 Progress, Internet-Draft, draft-ietf-6lo-minimal-fragment- 984 10, 1 February 2020, . 987 12. Informative References 989 [RFC8201] McCann, J., Deering, S., Mogul, J., and R. Hinden, Ed., 990 "Path MTU Discovery for IP version 6", STD 87, RFC 8201, 991 DOI 10.17487/RFC8201, July 2017, 992 . 994 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 995 Recommendations Regarding Active Queue Management", 996 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 997 . 999 [RFC3031] Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol 1000 Label Switching Architecture", RFC 3031, 1001 DOI 10.17487/RFC3031, January 2001, 1002 . 1004 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 1005 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 1006 . 1008 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, 1009 RFC 2914, DOI 10.17487/RFC2914, September 2000, 1010 . 1012 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 1013 of Explicit Congestion Notification (ECN) to IP", 1014 RFC 3168, DOI 10.17487/RFC3168, September 2001, 1015 . 1017 [RFC4919] Kushalnagar, N., Montenegro, G., and C. Schumacher, "IPv6 1018 over Low-Power Wireless Personal Area Networks (6LoWPANs): 1019 Overview, Assumptions, Problem Statement, and Goals", 1020 RFC 4919, DOI 10.17487/RFC4919, August 2007, 1021 . 1023 [RFC4963] Heffner, J., Mathis, M., and B. Chandler, "IPv4 Reassembly 1024 Errors at High Data Rates", RFC 4963, 1025 DOI 10.17487/RFC4963, July 2007, 1026 . 1028 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 1029 "Computing TCP's Retransmission Timer", RFC 6298, 1030 DOI 10.17487/RFC6298, June 2011, 1031 . 1033 [RFC6550] Winter, T., Ed., Thubert, P., Ed., Brandt, A., Hui, J., 1034 Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, 1035 JP., and R. Alexander, "RPL: IPv6 Routing Protocol for 1036 Low-Power and Lossy Networks", RFC 6550, 1037 DOI 10.17487/RFC6550, March 2012, 1038 . 1040 [RFC7554] Watteyne, T., Ed., Palattella, M., and L. Grieco, "Using 1041 IEEE 802.15.4e Time-Slotted Channel Hopping (TSCH) in the 1042 Internet of Things (IoT): Problem Statement", RFC 7554, 1043 DOI 10.17487/RFC7554, May 2015, 1044 . 1046 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1047 (IPv6) Specification", STD 86, RFC 8200, 1048 DOI 10.17487/RFC8200, July 2017, 1049 . 1051 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 1052 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1053 March 2017, . 1055 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 1056 Explicit Congestion Notification (ECN)", RFC 8087, 1057 DOI 10.17487/RFC8087, March 2017, 1058 . 1060 [RFC6606] Kim, E., Kaspar, D., Gomez, C., and C. Bormann, "Problem 1061 Statement and Requirements for IPv6 over Low-Power 1062 Wireless Personal Area Network (6LoWPAN) Routing", 1063 RFC 6606, DOI 10.17487/RFC6606, May 2012, 1064 . 1066 [I-D.ietf-lwig-6lowpan-virtual-reassembly] 1067 Bormann, C. and T. Watteyne, "Virtual reassembly buffers 1068 in 6LoWPAN", Work in Progress, Internet-Draft, draft-ietf- 1069 lwig-6lowpan-virtual-reassembly-01, 11 March 2019, 1070 . 1073 [I-D.ietf-intarea-frag-fragile] 1074 Bonica, R., Baker, F., Huston, G., Hinden, R., Troan, O., 1075 and F. Gont, "IP Fragmentation Considered Fragile", Work 1076 in Progress, Internet-Draft, draft-ietf-intarea-frag- 1077 fragile-17, 30 September 2019, 1078 . 1081 [I-D.ietf-core-cocoa] 1082 Bormann, C., Betzler, A., Gomez, C., and I. Demirkol, 1083 "CoAP Simple Congestion Control/Advanced", Work in 1084 Progress, Internet-Draft, draft-ietf-core-cocoa-03, 21 1085 February 2018, 1086 . 1088 [I-D.ietf-6tisch-architecture] 1089 Thubert, P., "An Architecture for IPv6 over the TSCH mode 1090 of IEEE 802.15.4", Work in Progress, Internet-Draft, 1091 draft-ietf-6tisch-architecture-28, 29 October 2019, 1092 . 1095 [IEEE.802.15.4] 1096 IEEE, "IEEE Standard for Low-Rate Wireless Networks", 1097 IEEE Standard 802.15.4, DOI 10.1109/IEEE 1098 P802.15.4-REVd/D01, 1099 . 1101 [Kent] Kent, C. and J. Mogul, ""Fragmentation Considered 1102 Harmful", In Proc. SIGCOMM '87 Workshop on Frontiers in 1103 Computer Communications Technology", 1104 DOI 10.1145/55483.55524, August 1987, 1105 . 1108 Appendix A. Rationale 1110 There are a number of uses for large packets in Wireless Sensor 1111 Networks. Such usages may not be the most typical or represent the 1112 largest amount of traffic over the LLN; however, the associated 1113 functionality can be critical enough to justify extra care for 1114 ensuring effective transport of large packets across the LLN. 1116 The list of those usages includes: 1118 Towards the LLN node: Firmware update: For example, a new version 1119 of the LLN node software is downloaded from a system manager 1120 over unicast or multicast services. Such a reflashing 1121 operation typically involves updating a large number of similar 1122 LLN nodes over a relatively short period of time. 1124 Packages of Commands: A number of commands or 1125 a full configuration can be packaged as a single message to 1126 ensure consistency and enable atomic execution or complete roll 1127 back. Until such commands are fully received and interpreted, 1128 the intended operation will not take effect. 1130 From the LLN node: Waveform captures: A number of consecutive 1131 samples are measured at a high rate for a short time and then 1132 transferred from a sensor to a gateway or an edge server as a 1133 single large report. 1135 Data logs: LLN nodes may generate large logs of 1136 sampled data for later extraction. LLN nodes may also generate 1137 system logs to assist in diagnosing problems on the node or 1138 network. 1140 Large data packets: Rich data types might 1141 require more than one fragment. 1143 Uncontrolled firmware download or waveform upload can easily result 1144 in a massive increase of the traffic and saturate the network. 1146 When a fragment is lost in transmission, the lack of recovery in the 1147 original fragmentation system of RFC 4944 implies that all fragments 1148 would need to be resent, further contributing to the congestion that 1149 caused the initial loss, and potentially leading to congestion 1150 collapse. 1152 This saturation may lead to excessive radio interference, or random 1153 early discard (leaky bucket) in relaying nodes. Additional queuing 1154 and memory congestion may result while waiting for a low power next 1155 hop to emerge from its sleeping state. 1157 Considering that RFC 4944 defines an MTU is 1280 bytes and that in 1158 most incarnations (but 802.15.4g) a IEEE Std. 802.15.4 frame can 1159 limit the MAC payload to as few as 74 bytes, a packet might be 1160 fragmented into at least 18 fragments at the 6LoWPAN shim layer. 1161 Taking into account the worst-case header overhead for 6LoWPAN 1162 Fragmentation and Mesh Addressing headers will increase the number of 1163 required fragments to around 32. This level of fragmentation is much 1164 higher than that traditionally experienced over the Internet with 1165 IPv4 fragments. At the same time, the use of radios increases the 1166 probability of transmission loss and Mesh-Under techniques compound 1167 that risk over multiple hops. 1169 Mechanisms such as TCP or application-layer segmentation could be 1170 used to support end-to-end reliable transport. One option to support 1171 bulk data transfer over a frame-size-constrained LLN is to set the 1172 Maximum Segment Size to fit within the link maximum frame size. 1173 Doing so, however, can add significant header overhead to each 1174 802.15.4 frame. In addition, deploying such a mechanism requires 1175 that the end-to-end transport is aware of the delivery properties of 1176 the underlying LLN, which is a layer violation, and difficult to 1177 achieve from the far end of the IPv6 network. 1179 Appendix B. Requirements 1181 For one-hop communications, a number of Low Power and Lossy Network 1182 (LLN) link-layers propose a local acknowledgment mechanism that is 1183 enough to detect and recover the loss of fragments. In a multihop 1184 environment, an end-to-end fragment recovery mechanism might be a 1185 good complement to a hop-by-hop MAC level recovery. This draft 1186 introduces a simple protocol to recover individual fragments between 1187 6LoWPAN endpoints that may be multiple hops away. The method 1188 addresses the following requirements of an LLN: 1190 Number of fragments: The recovery mechanism must support highly 1191 fragmented packets, with a maximum of 32 fragments per packet. 1193 Minimum acknowledgment overhead: Because the radio is half duplex, 1194 and because of silent time spent in the various medium access 1195 mechanisms, an acknowledgment consumes roughly as many resources 1196 as a data fragment. 1198 The new end-to-end fragment recovery mechanism should be able to 1199 acknowledge multiple fragments in a single message and not require 1200 an acknowledgment at all if fragments are already protected at a 1201 lower layer. 1203 Controlled latency: The recovery mechanism must succeed or give up 1204 within the time boundary imposed by the recovery process of the 1205 Upper Layer Protocols. 1207 Optional congestion control: The aggregation of multiple concurrent 1208 flows may lead to the saturation of the radio network and 1209 congestion collapse. 1211 The recovery mechanism should provide means for controlling the 1212 number of fragments in transit over the LLN. 1214 Appendix C. Considerations on Flow Control 1216 Considering that a multi-hop LLN can be a very sensitive environment 1217 due to the limited queuing capabilities of a large population of its 1218 nodes, this draft recommends a simple and conservative approach to 1219 Congestion Control, based on TCP congestion avoidance. 1221 Congestion on the forward path is assumed in case of packet loss, and 1222 packet loss is assumed upon time out. The draft allows controlling 1223 the number of outstanding fragments that have been transmitted but 1224 for which an acknowledgment was not received yet. It must be noted 1225 that the number of outstanding fragments should not exceed the number 1226 of hops in the network, but the way to figure the number of hops is 1227 out of scope for this document. 1229 Congestion on the forward path can also be indicated by an Explicit 1230 Congestion Notification (ECN) mechanism. Though whether and how ECN 1231 [RFC3168] is carried out over the LoWPAN is out of scope, this draft 1232 provides a way for the destination endpoint to echo an ECN indication 1233 back to the source endpoint in an acknowledgment message as 1234 represented in Figure 4 in Section 5.2. 1236 It must be noted that congestion and collision are different topics. 1237 In particular, when a mesh operates on a same channel over multiple 1238 hops, then the forwarding of a fragment over a certain hop may 1239 collide with the forwarding of a next fragment that is following over 1240 a previous hop but in a same interference domain. This draft enables 1241 end-to-end flow control, but leaves it to the sender stack to pace 1242 individual fragments within a transmit window, so that a given 1243 fragment is sent only when the previous fragment has had a chance to 1244 progress beyond the interference domain of this hop. In the case of 1245 6TiSCH [I-D.ietf-6tisch-architecture], which operates over the 1246 TimeSlotted Channel Hopping [RFC7554] (TSCH) mode of operation of 1247 IEEE802.14.5, a fragment is forwarded over a different channel at a 1248 different time and it makes full sense to transmit the next fragment 1249 as soon as the previous fragment has had its chance to be forwarded 1250 at the next hop. 1252 From the standpoint of a source 6LoWPAN endpoint, an outstanding 1253 fragment is a fragment that was sent but for which no explicit 1254 acknowledgment was received yet. This means that the fragment might 1255 be on the way, received but not yet acknowledged, or the 1256 acknowledgment might be on the way back. It is also possible that 1257 either the fragment or the acknowledgment was lost on the way. 1259 From the sender standpoint, all outstanding fragments might still be 1260 in the network and contribute to its congestion. There is an 1261 assumption, though, that after a certain amount of time, a frame is 1262 either received or lost, so it is not causing congestion anymore. 1263 This amount of time can be estimated based on the round-trip delay 1264 between the 6LoWPAN endpoints. The method detailed in [RFC6298] is 1265 recommended for that computation. 1267 The reader is encouraged to read through "Congestion Control 1268 Principles" [RFC2914]. Additionally [RFC7567] and [RFC5681] provide 1269 deeper information on why this mechanism is needed and how TCP 1270 handles Congestion Control. Basically, the goal here is to manage 1271 the amount of fragments present in the network; this is achieved by 1272 to reducing the number of outstanding fragments over a congested path 1273 by throttling the sources. 1275 Section 6 describes how the sender decides how many fragments are 1276 (re)sent before an acknowledgment is required, and how the sender 1277 adapts that number to the network conditions. 1279 Author's Address 1281 Pascal Thubert (editor) 1282 Cisco Systems, Inc 1283 Building D 1284 45 Allee des Ormes - BP1200 1285 06254 MOUGINS - Sophia Antipolis 1286 France 1288 Phone: +33 497 23 26 34 1289 Email: pthubert@cisco.com