idnits 2.17.1 draft-ietf-6lo-fragment-recovery-08.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 (28 November 2019) is 1608 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-15) exists of draft-ietf-6lo-minimal-fragment-04 == Outdated reference: A later version (-02) exists of draft-ietf-lwig-6lowpan-virtual-reassembly-01 == Outdated reference: A later version (-30) exists of draft-ietf-6tisch-architecture-28 Summary: 0 errors (**), 0 flaws (~~), 4 warnings (==), 2 comments (--).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) 28 November 2019 5 Intended status: Standards Track 6 Expires: 31 May 2020 8 6LoWPAN Selective Fragment Recovery 9 draft-ietf-6lo-fragment-recovery-08 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 31 May 2020. 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 (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. Upon the first fragment . . . . . . . . . . . . . . . 14 66 6.1.2. Upon the next fragments . . . . . . . . . . . . . . . 15 67 6.2. Upon the 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 (in the order few bytes 88 to a few tens of bytes) at a time. Given that an IEEE Std. 802.15.4 89 [IEEE.802.15.4] frame can carry a payload of 74 bytes or more, 90 fragmentation is usually not required. However, and though this 91 happens only occasionally, a number of mission critical applications 92 do require the capability to transfer larger chunks of data, for 93 instance to support the firmware upgrade of the LLN nodes or the 94 extraction of logs from LLN nodes. In the former case, the large 95 chunk of data is transferred to the LLN node, whereas in the latter, 96 the large chunk flows away from the LLN node. In both cases, the 97 size can be on the order of 10 kilobytes or more and an end-to-end 98 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 to 104 reassemble the full packet at each hop, which may cause latency along 105 a path and an overall buffer bloat in the network. The "6TiSCH 106 Architecture" [I-D.ietf-6tisch-architecture] recommends to use 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, pages 6 and 7. [RFC4944] 130 as no selective recovery and the whole datagram fails when one 131 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 can not 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 to a 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 application layer. The reader is encouraged to read "IPv4 180 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 fragment is introduced and new 260 dispatch types are defined in Section 5. 262 [RFC8138] allows to modify the size of a packet en-route by removing 263 the consumed hops in a compressed Routing Header. It results that 264 fragment_offset and datagram_size (see Section 2.3) must also be 265 modified en-route, whcih is difficult to do in the uncompressed form. 266 This specification expresses those fields in the Compressed Form and 267 allows to modify them en-route (see Section 4.3) easily. 269 Note that consistently 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 At the time of this writing, [I-D.ietf-6lo-minimal-fragment] allows 285 for refragmenting in intermediate nodes, meaning that some bytes from 286 a given fragment may be left in the VRB to be added to the next 287 fragment. The reason for this to happen would be the need for space 288 in the outgoing fragment that was not needed in the incoming 289 fragment, for instance because the 6LoWPAN Header Compression is not 290 as efficient on the outgoing link, e.g., if the Interface ID (IID) of 291 the source IPv6 address is elided by the originator on the first hop 292 because it matches the source MAC address, but cannot be on the next 293 hops because the source 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 Inter-Frame Gap, which is 308 a configurable interval of time between transmissions to a same next 309 hop. In the case of half duplex interfaces, this InterFrameGap 310 ensures that the next hop has progressed the previous frame and is 311 capable of receiving the next one. 313 In the case of a mesh operating at a single frequency with 314 omnidirectional antennas, a larger InterFrameGap 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 327 en-route in a Route-Over mesh LLN. If the size of the first fragment 328 is modified, then the intermediate node MUST adapt the datagram_size 329 to 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 on the Compressed Form of the 376 packet form as opposed to the uncompressed - native - packet form. 378 The format of the fragment header is shown in Figure 1. It is the 379 same for all fragments. The format has a length and an offset, as 380 well as a sequence field. This would be redundant if the offset was 381 computed as the product of the sequence by the length, but this is 382 not the case. The position of a fragment in the reassembly buffer is 383 neither correlated with the value of the sequence field nor with the 384 order in which the fragments are received. This enables out-of- 385 sequence subfragmenting, e.g., a fragment seq. 5 that is retried end- 386 to-end as smaller fragments seq. 5, 13 and 14 due to a change of MTU 387 along the path between the 6LoWPAN endpoints. 389 1 2 3 390 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 1 391 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 392 |1 1 1 0 1 0 0|E| datagram_tag | 393 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 394 |X| sequence| fragment_size | fragment_offset | 395 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 397 X set == Ack-Request 399 Figure 1: RFRAG Dispatch type and Header 401 There is no requirement on the receiver to check for contiguity of 402 the received fragments, and the sender MUST ensure that when all 403 fragments are acknowledged, then the datagram is fully received. 404 This may be useful in particular in the case where the MTU changes 405 and a fragment sequence is retried with a smaller fragment_size, the 406 remainder of the original fragment being retried with new sequence 407 values. 409 The first fragment is recognized by a sequence of 0; it carries its 410 fragment_size and the datagram_size of the compressed packet before 411 it is fragmented, whereas the other fragments carry their 412 fragment_size and fragment_offset. The last fragment for a datagram 413 is recognized when its fragment_offset and its fragment_size add up 414 to the datagram_size. 416 Recoverable Fragments are sequenced and a bitmap is used in the RFRAG 417 Acknowledgment to indicate the received fragments by setting the 418 individual bits that correspond to their sequence. 420 X: 1 bit; Ack-Request: when set, the sender requires an RFRAG 421 Acknowledgment from the receiver. 423 E: 1 bit; Explicit Congestion Notification; the "E" flag is reset by 424 the source of the fragment and set by intermediate routers to 425 signal that this fragment experienced congestion along its path. 427 Fragment_size: 10 bit unsigned integer; the size of this fragment in 428 a unit that depends on the MAC layer technology. Unless 429 overridden by a more specific specification, that unit is the 430 octet which allows fragments up to 512 bytes. 432 datagram_tag: 8 bits; an identifier of the datagram that is locally 433 unique to the sender. 435 Sequence: 5 bit unsigned integer; the sequence number of the 436 fragment in the acknowledgement bitmap. Fragments are numbered 437 [0..N] where N is in [0..31]. A Sequence of 0 indicates the first 438 fragment in a datagram, but non-zero values are not indicative of 439 the position in the reassembly buffer. 441 Fragment_offset: 16 bit unsigned integer. 443 When the Fragment_offset is set to a non-0 value, its semantics 444 depend on the value of the Sequence field as follows: 446 * For a first fragment (i.e. with a Sequence of 0), this field 447 indicates the datagram_size of the compressed datagram, to help 448 the receiver allocate an adapted buffer for the reception and 449 reassembly operations. The fragment may be stored for local 450 reassembly. Alternatively, it may be routed based on the 451 destination IPv6 address. In that case, a VRB state must be 452 installed as described in Section 6.1.1. 453 * When the Sequence is not 0, this field indicates the offset of 454 the fragment in the Compressed Form of the datagram. The 455 fragment may be added to a local reassembly buffer or forwarded 456 based on an existing VRB as described in Section 6.1.2. 458 A Fragment_offset that is set to a value of 0 indicates an abort 459 condition and all state regarding the datagram should be cleaned 460 up once the processing of the fragment is complete; the processing 461 of the fragment depends on whether there is a VRB already 462 established for this datagram, and the next hop is still 463 reachable: 465 * if a VRB already exists and is not broken, the fragment is to 466 be forwarded along the associated Label Switched Path (LSP) as 467 described in Section 6.1.2, but regardless of the value of the 468 Sequence field; 469 * else, if the Sequence is 0, then the fragment is to be routed 470 as described in Section 6.1.1 but no state is conserved 471 afterwards. In that case, the session if it exists is aborted 472 and the packet is also forwarded in an attempt to clean up the 473 next hops as along the path indicated by the IPv6 header 474 (possibly including a routing header). 476 If the fragment cannot be forwarded or routed, then an abort 477 RFRAG-ACK is sent back to the source as described in 478 Section 6.1.2. 480 5.2. RFRAG Acknowledgment Dispatch type and Header 482 This specification also defines a 4-octet RFRAG Acknowledgment bitmap 483 that is used by the reassembling endpoint to confirm selectively the 484 reception of individual fragments. A given offset in the bitmap maps 485 one to one with a given sequence number and indicates which fragment 486 is acknowledged as follows: 488 1 2 3 489 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 1 490 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 491 | RFRAG Acknowledgment Bitmap | 492 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 493 ^ ^ 494 | | bitmap indicating whether: 495 | +----- Fragment with sequence 9 was received 496 +----------------------- Fragment with sequence 0 was received 498 Figure 2: RFRAG Acknowledgment bitmap encoding 500 Figure 3 shows an example Acknowledgment bitmap which indicates that 501 all fragments from sequence 0 to 20 were received, except for 502 fragments 1, 2 and 16 that were lost and must be retried. 504 1 2 3 505 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 1 506 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 507 |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| 508 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 510 Figure 3: Example RFRAG Acknowledgment Bitmap 512 The RFRAG Acknowledgment Bitmap is included in a RFRAG Acknowledgment 513 header, as follows: 515 1 2 3 516 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 1 517 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 518 |1 1 1 0 1 0 1|E| datagram_tag | 519 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 520 | RFRAG Acknowledgment Bitmap (32 bits) | 521 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 522 Figure 4: RFRAG Acknowledgment Dispatch type and Header 524 E: 1 bit; Explicit Congestion Notification Echo 526 When set, the sender indicates that at least one of the 527 acknowledged fragments was received with an Explicit Congestion 528 Notification, indicating that the path followed by the fragments 529 is subject to congestion. More in Appendix C. 531 RFRAG Acknowledgment Bitmap: An RFRAG Acknowledgment Bitmap, whereby 532 setting the bit at offset x indicates that fragment x was 533 received, as shown in Figure 2. A NULL bitmap that indicates that 534 the fragmentation process is aborted. A FULL bitmap that 535 indicates that the fragmentation process is complete, all 536 fragments were received at the reassembly endpoint. 538 6. Fragments Recovery 540 The Recoverable Fragment header RFRAG is used to transport a fragment 541 and optionally request an RFRAG Acknowledgment that will confirm the 542 good reception of one or more fragments. An RFRAG Acknowledgment is 543 carried as a standalone fragment header (i.e. with no 6LoWPAN 544 payload) in a message that is propagated back to the 6LoWPAN endpoint 545 that was the originator of the fragments. To achieve this, each hop 546 that performed an MPLS-like operation on fragments reverses that 547 operation for the RFRAG_ACK by sending a frame from the next hop to 548 the previous hop as known by its MAC address in the VRB. The 549 datagram_tag in the RFRAG_ACK is unique to the receiver and is enough 550 information for an intermediate hop to locate the VRB that contains 551 the datagram_tag used by the previous hop and the Layer-2 information 552 associated to it (interface and MAC address). 554 The 6LoWPAN endpoint that fragments the packets at 6LoWPAN level (the 555 sender) also controls the amount of acknowledgments by setting the 556 Ack-Request flag in the RFRAG packets. The sender may set the Ack- 557 Request flag on any fragment to perform congestion control by 558 limiting the number of outstanding fragments, which are the fragments 559 that have been sent but for which reception or loss was not 560 positively confirmed by the reassembling endpoint. The maximum 561 number of outstanding fragments is the Window-Size. It is 562 configurable and may vary in case of ECN notification. When the 563 6LoWPAN endpoint that reassembles the packets at 6LoWPAN level (the 564 receiver) receives a fragment with the Ack-Request flag set, it MUST 565 send an RFRAG Acknowledgment back to the originator to confirm 566 reception of all the fragments it has received so far. 568 The Ack-Request ('X') set in an RFRAG marks the end of a window. 569 This flag MUST be set on the last fragment if the sender wishes to 570 protect the datagram, and it MAY be set in any intermediate fragment 571 for the purpose of flow control. This ARQ process MUST be protected 572 by a timer, and the fragment that carries the 'X' flag MAY be retried 573 upon time out a configurable amount of times (see Section 7.1). Upon 574 exhaustion of the retries the sender may either abort the 575 transmission of the datagram or retry the datagram from the first 576 fragment with an 'X' flag set in order to reestablish a path and 577 discover which fragments were received over the old path in the 578 acknowledgment bitmap. When the sender of the fragment knows that an 579 underlying link-layer mechanism protects the fragments, it may 580 refrain from using the RFRAG Acknowledgment mechanism, and never set 581 the Ack-Request bit. 583 The RFRAG Acknowledgment can optionally carry an ECN indication for 584 flow control (see Appendix C). The receiver of a fragment with the 585 'E' (ECN) flag set MUST echo that information by setting the 'E' 586 (ECN) flag in the next RFRAG Acknowledgment. 588 In order to protect the datagram, the sender transfers a controlled 589 number of fragments and flags the last fragment of a window with an 590 RFRAG Acknowledgment Request. The receiver MUST acknowledge a 591 fragment with the acknowledgment request bit set. If any fragment 592 immediately preceeding an acknowledgment request is still missing, 593 the receiver MAY intentionally delay its acknowledgment to allow in- 594 transit fragments to arrive. Because it might defeat the round trip 595 delay computation, delaying the acknowledgment should be configurable 596 and not enabled by default. 598 The receiver MAY issue unsolicited acknowledgments. An unsolicited 599 acknowledgment signals to the sender endpoint that it can resume 600 sending if it had reached its maximum number of outstanding 601 fragments. Another use is to inform that the reassembling endpoint 602 aborted the process of an individual datagram. 604 When all the fragments are received, the receiving endpoint 605 reconstructs the packet, passes it to the upper layer, sends a RFRAG 606 Acknowledgment on the reverse path with a FULL bitmap, and arms a 607 short timer to absorb fragments that are still in flight for that 608 datagram without creating a new state and abort the communication if 609 it keeps going on beyond a reasonable time. 611 Note that acknowledgments might consume precious resources so the use 612 of unsolicited acknowledgments should be configurable and not enabled 613 by default. 615 An observation is that streamlining forwarding of fragments generally 616 reduces the latency over the LLN mesh, providing room for retries 617 within existing upper-layer reliability mechanisms. The sender 618 protects the transmission over the LLN mesh with a retry timer that 619 is computed according to the method detailed in [RFC6298]. It is 620 expected that the upper layer retries obey the recommendations in 621 "UDP Usage Guidelines" [RFC8085], in which case a single round of 622 fragment recovery should fit within the upper layer recovery timers. 624 Fragments are sent in a round robin fashion: the sender sends all the 625 fragments for a first time before it retries any lost fragment; lost 626 fragments are retried in sequence, oldest first. This mechanism 627 enables the receiver to acknowledge fragments that were delayed in 628 the network before they are retried. 630 When a single frequency is used by contiguous hops, the sender should 631 wait a reasonable amount of time between fragments so as to let a 632 fragment progress a few hops and avoid hidden terminal issues. This 633 precaution is not required on channel hopping technologies such as 634 Time Slotted Channel Hopping (TSCH) [RFC6554], where nodes that 635 communicate at Layer-2 are scheduled to send and receive 636 respectively, and different hops operate on different channels. 638 6.1. Forwarding Fragments 640 It is assumed that the first Fragment is large enough to carry the 641 IPv6 header and make routing decisions. If that is not so, then this 642 specification MUST NOT be used. 644 This specification extends the Virtual Reassembly Buffer (VRB) 645 technique to forward fragments with no intermediate reconstruction of 646 the entire packet. It inherits operations like datagram_tag 647 Switching and using a timer to clean the VRB when the traffic dries 648 up. In more details, the first fragment carries the IP header and it 649 is routed all the way from the fragmenting endpoint to the 650 reassembling endpoint. Upon the first fragment, the routers along 651 the path install a label-switched path (LSP), and the following 652 fragments are label-switched along that path. As a consequence, the 653 next fragments can only follow the path that was set up by the first 654 fragment and cannot follow an alternate route. The datagram_tag is 655 used to carry the label, that is swapped at each hop. All fragments 656 follow the same path and fragments are delivered in the order at 657 which they are sent. 659 6.1.1. Upon the first fragment 661 In Route-Over mode, the source and destination MAC addresses in a 662 frame change at each hop. The label that is formed and placed in the 663 datagram_tag is associated to the source MAC and only valid (and 664 unique) for that source MAC. Upon a first fragment (i.e. with a 665 sequence of zero), an intermediate router creates a VRB and the 666 associated LSP state for the tuple (source MAC address, datagram_tag) 667 and the fragment is forwarded along the IPv6 route that matches the 668 destination IPv6 address in the IPv6 header as prescribed by 669 [I-D.ietf-6lo-minimal-fragment], whereas the receiving endpoint 670 allocates a reassembly buffer. 672 The LSP state enables to match the (previous MAC address, 673 datagram_tag) in an incoming fragment to the tuple (next MAC address, 674 swapped datagram_tag) used in the forwarded fragment and points at 675 the VRB. In addition, the router also forms a Reverse LSP state 676 indexed by the MAC address of the next hop and the swapped 677 datagram_tag. This reverse LSP state also points at the VRB and 678 enables to match the (next MAC address, swapped_datagram_tag) found 679 in an RFRAG Acknowledgment to the tuple (previous MAC address, 680 datagram_tag) used when forwarding a Fragment Acknowledgment (RFRAG- 681 ACK) back to the sender endpoint. 683 The first fragment may be received a second time, indicating that it 684 did not reach the destination and was retried. In that case, it 685 SHOULD follow the same path as the first occurrence. It is up to 686 sending endpoint to abort a transmission and then retry it from 687 scratch, which may build an entirely new path. 689 6.1.2. Upon the next fragments 691 Upon a next fragment (i.e. with a non-zero sequence), an intermediate 692 router looks up a LSP indexed by the tuple (MAC address, 693 datagram_tag) found in the fragment. If it is found, the router 694 forwards the fragment using the associated VRB as prescribed by 695 [I-D.ietf-6lo-minimal-fragment]. 697 if the VRB for the tuple is not found, the router builds an RFRAG-ACK 698 to abort the transmission of the packet. The resulting message has 699 the following information: 701 * The source and destination MAC addresses are swapped from those 702 found in the fragment 703 * The datagram_tag set to the datagram_tag found in the fragment 704 * A NULL bitmap is used to signal the abort condition 706 At this point the router is all set and can send the RFRAG-ACK back 707 to the previous router. The RFRAG-ACK should normally be forwarded 708 all the way to the source using the reverse LSP state in the VRBs in 709 the intermediate routers as described in the next section. 711 [I-D.ietf-6lo-minimal-fragment] indicates that the receiving endpoint 712 stores "the actual packet data from the fragments received so far, in 713 a form that makes it possible to detect when the whole packet has 714 been received and can be processed or forwarded". How this is 715 computed in implementation specific but relies on receiving all the 716 bytes up to the datagram_size indicated in the first fragment. An 717 implementation may receive overlapping fragments as the result of 718 retries after an MTU change. 720 6.2. Upon the RFRAG Acknowledgments 722 Upon an RFRAG-ACK, the router looks up a Reverse LSP indexed by the 723 tuple (MAC address, datagram_tag), which are respectively the source 724 MAC address of the received frame and the received datagram_tag. If 725 it is found, the router forwards the fragment using the associated 726 VRB as prescribed by [I-D.ietf-6lo-minimal-fragment], but using the 727 Reverse LSP so that the RFRAG-ACK flows back to the sender endpoint. 729 If the Reverse LSP is not found, the router MUST silently drop the 730 RFRAG-ACK message. 732 Either way, if the RFRAG-ACK indicates that the fragment was entirely 733 received (FULL bitmap), it arms a short timer, and upon timeout, the 734 VRB and all the associated state are destroyed. Until the timer 735 elapses, fragments of that datagram may still be received, e.g. if 736 the RFRAG-ACK was lost on the way back and the source retried the 737 last fragment. In that case, the router forwards the fragment 738 according to the state in the VRB. 740 This specification does not provide a method to discover the number 741 of hops or the minimal value of MTU along those hops. But should the 742 minimal MTU decrease, it is possible to retry a long fragment (say 743 sequence of 5) with first a shorter fragment of the same sequence (5 744 again) and then one or more other fragments with a sequence that was 745 not used before (e.g., 13 and 14). Note that Path MTU Discovery is 746 out of scope for this document. 748 6.3. Aborting the Transmission of a Fragmented Packet 750 A reset is signaled on the forward path with a pseudo fragment that 751 has the fragment_offset, sequence and fragment_size all set to 0, and 752 no data. 754 When the sender or a router on the way decides that a packet should 755 be dropped and the fragmentation process aborted, it generates a 756 reset pseudo fragment and forwards it down the fragment path. 758 Each router next along the path the way forwards the pseudo fragment 759 based on the VRB state. If an acknowledgment is not requested, the 760 VRB and all associated state are destroyed. 762 Upon reception of the pseudo fragment, the receiver cleans up all 763 resources for the packet associated to the datagram_tag. If an 764 acknowledgment is requested, the receiver responds with a NULL 765 bitmap. 767 The other way around, the receiver might need to abort the process of 768 a fragmented packet for internal reasons, for instance if it is out 769 of reassembly buffers, already uses all 256 possible values of the 770 datagram_tag, or if it keeps receiving fragments beyond a reasonable 771 time while it considers that this packet is already fully reassembled 772 and was passed to the upper layer. In that case, the receiver SHOULD 773 indicate so to the sender with a NULL bitmap in a RFRAG 774 Acknowledgment. The RFRAG Acknowledgment is frowarded all the way 775 back to the source of the packet and cleans up all resources on the 776 way. Upon an acknowledgment with a NULL bitmap, the sender endpoint 777 MUST abort the transmission of the fragmented datagram with one 778 exception: In the particular case of the first fragment, it MAY 779 decide to retry via an alternate next hop instead. 781 6.4. Applying Recoverable Fragmentation along a Diverse Path 783 The text above can be read with the assumption of a serial path 784 between a source and a destination. Section 4.5.3 of the "6TiSCH 785 Architecture" [I-D.ietf-6tisch-architecture] defines the concept of a 786 Track that can be a complex path between a source and a destination 787 with Packet ARQ, Replication, Elimination and Overhearing (PAREO) 788 along the Track. This specification can be used along any subset of 789 the complex Track where the first fragment is flooded. The last 790 RFRAG Acknowledgment is flooded on that same subset in the reverse 791 direction. Intermediate RFRAG Acknowledgments can be flooded on any 792 sub-subset of that reverse subset that reach back to the source. 794 7. Management Considerations 796 7.1. Protocol Parameters 798 There is no particular configuration on the receiver, as echoing ECN 799 is always on. The configuration only applies to the sender, which is 800 in control of the transmission. The management system SHOULD be 801 capable of providing the parameters below: 803 MinFragmentSize: The MinFragmentSize is the minimum value for the 804 Fragment_Size. 806 OptFragmentSize: The MinFragmentSize is the value for the 807 Fragment_Size that the sender should use to start with. 809 MaxFragmentSize: The MaxFragmentSize is the maximum value for the 810 Fragment_Size. It MUST be lower than the minimum MTU along the 811 path. A large value augments the chances of buffer bloat and 812 transmission loss. The value MUST be less than 512 if the unit 813 that is defined for the PHY layer is the octet. 815 UseECN: Indicates whether the sender should react to ECN. When the 816 sender reacts to ECN the Window_Size will vary between 817 MinWindowSize and MaxWindowSize. 819 MinWindowSize: The minimum value of Window_Size that the sender can 820 use. 822 OptWindowSize: The OptWindowSize is the value for the Window_Size 823 that the sender should use to start with. 825 MaxWindowSize: The maximum value of Window_Size that the sender can 826 use. The value MUSt be less than 32. 828 InterFrameGap: Indicates a minimum amount of time between 829 transmissions. All packets to a same destination, and in 830 particular fragments, may be subject to receive while transmitting 831 and hidden terminal collisions with the next or the previous 832 transmission as the fragments progress along a same path. The 833 InterFrameGap protects the propagation of one transmission before 834 the next one is triggered and creates a duty cycle that controls 835 the ratio of air time and memory in intermediate nodes that a 836 particular datagram will use. 838 MinARQTimeOut: The maximum amount of time a node should wait for an 839 RFRAG Acknowledgment before it takes a next action. 841 OptARQTimeOut: The starting point of the value of the amount that a 842 sender should wait for an RFRAG Acknowledgment before it takes a 843 next action. 845 MaxARQTimeOut: The maximum amount of time a node should wait for an 846 RFRAG Acknowledgment before it takes a next action. 848 MaxFragRetries: The maximum number of retries for a particular 849 Fragment. 851 MaxDatagramRetries: The maximum number of retries from scratch for a 852 particular Datagram. 854 7.2. Observing the network 856 The management system should monitor the amount of retries and of ECN 857 settings that can be observed from the perspective of the both the 858 sender and the receiver, and may tune the optimum size of 859 Fragment_Size and of the Window_Size, OptWindowSize and OptWindowSize 860 respectively, at the sender. The values should be bounded by the 861 expected number of hops and reduced beyond that when the number of 862 datagrams that can traverse an intermediate point may exceed its 863 capacity and cause a congestion loss. The InterFrameGap is another 864 tool that can be used to increase the spacing between fragments of a 865 same datagram and reduce the ratio of time when a particular 866 intermediate node holds a fragment of that datagram. 868 8. Security Considerations 870 The considerations in the Security sections of [I-D.ietf-core-cocoa] 871 and [I-D.ietf-6lo-minimal-fragment] apply equally to this 872 specification. 874 The process of recovering fragments does not appear to create any 875 opening for new threat compared to "Transmission of IPv6 Packets over 876 IEEE 802.15.4 Networks" [RFC4944] beyond the change of size of the 877 datagram_tag. By reducing to 8 bits, the tag will wrap faster than 878 with [RFC4944]. But for a constrained network where a node is 879 expected to be able to hold only one or a few large packets in 880 memory, 256 is still a large number. Also, the acknowledgement 881 mechanism allows ot clean up the state rapidly once the packet is 882 fully transmitted or aborted. 884 The abstract Virtual Recovery Buffer inherited from 885 [I-D.ietf-6lo-minimal-fragment] may be used to perform a Denial-of- 886 Service (DoS) attack against the intermediate Routers since the 887 routers need to maintain a state per flow. The particular VRB 888 implementation technique described in 889 [I-D.ietf-lwig-6lowpan-virtual-reassembly] allows to realign which 890 data goes in which fragment which causes the intermediate node to 891 store a portion of the data, which adds an attack vector that is not 892 present with this specification. With this specification, the data 893 that is transported in each fragment is conserved and the state to 894 keep does not include any data that would not fit in the previous 895 fragment. 897 9. IANA Considerations 899 This document allocates 4 values in Page 0 for recoverable fragments 900 from the "Dispatch Type Field" registry that was created by 901 "Transmission of IPv6 Packets over IEEE 802.15.4 Networks" [RFC4944] 902 and reformatted by "6LoWPAN Paging Dispatch" [RFC8025]. 904 The suggested values (to be confirmed by IANA) are indicated in 905 Table 1. 907 +-------------+------+----------------------------------+-----------+ 908 | Bit Pattern | Page | Header Type | Reference | 909 +=============+======+==================================+===========+ 910 | 11 10100x | 0 | RFRAG - Recoverable Fragment | THIS RFC | 911 +-------------+------+----------------------------------+-----------+ 912 | 11 10101x | 0 | RFRAG-ACK - RFRAG | THIS RFC | 913 | | | Acknowledgment | | 914 +-------------+------+----------------------------------+-----------+ 916 Table 1: Additional Dispatch Value Bit Patterns 918 10. Acknowledgments 920 The author wishes to thank Michel Veillette, Dario Tedeschi, Laurent 921 Toutain, Carles Gomez Montenegro, Thomas Watteyne and Michael 922 Richardson for in-depth reviews and comments. Also many thanks to 923 Jonathan Hui, Jay Werb, Christos Polyzois, Soumitri Kolavennu, Pat 924 Kinney, Margaret Wasserman, Richard Kelsey, Carsten Bormann and Harry 925 Courtice for their various contributions. 927 11. Normative References 929 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 930 Requirement Levels", BCP 14, RFC 2119, 931 DOI 10.17487/RFC2119, March 1997, 932 . 934 [RFC4944] Montenegro, G., Kushalnagar, N., Hui, J., and D. Culler, 935 "Transmission of IPv6 Packets over IEEE 802.15.4 936 Networks", RFC 4944, DOI 10.17487/RFC4944, September 2007, 937 . 939 [RFC6282] Hui, J., Ed. and P. Thubert, "Compression Format for IPv6 940 Datagrams over IEEE 802.15.4-Based Networks", RFC 6282, 941 DOI 10.17487/RFC6282, September 2011, 942 . 944 [RFC6554] Hui, J., Vasseur, JP., Culler, D., and V. Manral, "An IPv6 945 Routing Header for Source Routes with the Routing Protocol 946 for Low-Power and Lossy Networks (RPL)", RFC 6554, 947 DOI 10.17487/RFC6554, March 2012, 948 . 950 [RFC8025] Thubert, P., Ed. and R. Cragie, "IPv6 over Low-Power 951 Wireless Personal Area Network (6LoWPAN) Paging Dispatch", 952 RFC 8025, DOI 10.17487/RFC8025, November 2016, 953 . 955 [RFC8138] Thubert, P., Ed., Bormann, C., Toutain, L., and R. Cragie, 956 "IPv6 over Low-Power Wireless Personal Area Network 957 (6LoWPAN) Routing Header", RFC 8138, DOI 10.17487/RFC8138, 958 April 2017, . 960 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 961 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 962 May 2017, . 964 [I-D.ietf-6lo-minimal-fragment] 965 Watteyne, T., Bormann, C., and P. Thubert, "6LoWPAN 966 Fragment Forwarding", Work in Progress, Internet-Draft, 967 draft-ietf-6lo-minimal-fragment-04, 2 September 2019, 968 . 971 12. Informative References 973 [RFC8201] McCann, J., Deering, S., Mogul, J., and R. Hinden, Ed., 974 "Path MTU Discovery for IP version 6", STD 87, RFC 8201, 975 DOI 10.17487/RFC8201, July 2017, 976 . 978 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 979 Recommendations Regarding Active Queue Management", 980 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 981 . 983 [RFC3031] Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol 984 Label Switching Architecture", RFC 3031, 985 DOI 10.17487/RFC3031, January 2001, 986 . 988 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 989 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 990 . 992 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, 993 RFC 2914, DOI 10.17487/RFC2914, September 2000, 994 . 996 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 997 of Explicit Congestion Notification (ECN) to IP", 998 RFC 3168, DOI 10.17487/RFC3168, September 2001, 999 . 1001 [RFC4919] Kushalnagar, N., Montenegro, G., and C. Schumacher, "IPv6 1002 over Low-Power Wireless Personal Area Networks (6LoWPANs): 1003 Overview, Assumptions, Problem Statement, and Goals", 1004 RFC 4919, DOI 10.17487/RFC4919, August 2007, 1005 . 1007 [RFC4963] Heffner, J., Mathis, M., and B. Chandler, "IPv4 Reassembly 1008 Errors at High Data Rates", RFC 4963, 1009 DOI 10.17487/RFC4963, July 2007, 1010 . 1012 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 1013 "Computing TCP's Retransmission Timer", RFC 6298, 1014 DOI 10.17487/RFC6298, June 2011, 1015 . 1017 [RFC6550] Winter, T., Ed., Thubert, P., Ed., Brandt, A., Hui, J., 1018 Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, 1019 JP., and R. Alexander, "RPL: IPv6 Routing Protocol for 1020 Low-Power and Lossy Networks", RFC 6550, 1021 DOI 10.17487/RFC6550, March 2012, 1022 . 1024 [RFC7554] Watteyne, T., Ed., Palattella, M., and L. Grieco, "Using 1025 IEEE 802.15.4e Time-Slotted Channel Hopping (TSCH) in the 1026 Internet of Things (IoT): Problem Statement", RFC 7554, 1027 DOI 10.17487/RFC7554, May 2015, 1028 . 1030 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1031 (IPv6) Specification", STD 86, RFC 8200, 1032 DOI 10.17487/RFC8200, July 2017, 1033 . 1035 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 1036 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1037 March 2017, . 1039 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 1040 Explicit Congestion Notification (ECN)", RFC 8087, 1041 DOI 10.17487/RFC8087, March 2017, 1042 . 1044 [RFC6606] Kim, E., Kaspar, D., Gomez, C., and C. Bormann, "Problem 1045 Statement and Requirements for IPv6 over Low-Power 1046 Wireless Personal Area Network (6LoWPAN) Routing", 1047 RFC 6606, DOI 10.17487/RFC6606, May 2012, 1048 . 1050 [I-D.ietf-lwig-6lowpan-virtual-reassembly] 1051 Bormann, C. and T. Watteyne, "Virtual reassembly buffers 1052 in 6LoWPAN", Work in Progress, Internet-Draft, draft-ietf- 1053 lwig-6lowpan-virtual-reassembly-01, 11 March 2019, 1054 . 1057 [I-D.ietf-intarea-frag-fragile] 1058 Bonica, R., Baker, F., Huston, G., Hinden, R., Troan, O., 1059 and F. Gont, "IP Fragmentation Considered Fragile", Work 1060 in Progress, Internet-Draft, draft-ietf-intarea-frag- 1061 fragile-17, 30 September 2019, 1062 . 1065 [I-D.ietf-core-cocoa] 1066 Bormann, C., Betzler, A., Gomez, C., and I. Demirkol, 1067 "CoAP Simple Congestion Control/Advanced", Work in 1068 Progress, Internet-Draft, draft-ietf-core-cocoa-03, 21 1069 February 2018, 1070 . 1072 [I-D.ietf-6tisch-architecture] 1073 Thubert, P., "An Architecture for IPv6 over the TSCH mode 1074 of IEEE 802.15.4", Work in Progress, Internet-Draft, 1075 draft-ietf-6tisch-architecture-28, 29 October 2019, 1076 . 1079 [IEEE.802.15.4] 1080 IEEE, "IEEE Standard for Low-Rate Wireless Networks", 1081 IEEE Standard 802.15.4, DOI 10.1109/IEEE 1082 P802.15.4-REVd/D01, 1083 . 1085 [Kent] Kent, C. and J. Mogul, ""Fragmentation Considered 1086 Harmful", In Proc. SIGCOMM '87 Workshop on Frontiers in 1087 Computer Communications Technology", 1088 DOI 10.1145/55483.55524, August 1987, 1089 . 1092 Appendix A. Rationale 1094 There are a number of uses for large packets in Wireless Sensor 1095 Networks. Such usages may not be the most typical or represent the 1096 largest amount of traffic over the LLN; however, the associated 1097 functionality can be critical enough to justify extra care for 1098 ensuring effective transport of large packets across the LLN. 1100 The list of those usages includes: 1102 Towards the LLN node: Firmware update: For example, a new version 1103 of the LLN node software is downloaded from a system manager 1104 over unicast or multicast services. Such a reflashing 1105 operation typically involves updating a large number of similar 1106 LLN nodes over a relatively short period of time. 1108 Packages of Commands: A number of commands or 1109 a full configuration can be packaged as a single message to 1110 ensure consistency and enable atomic execution or complete roll 1111 back. Until such commands are fully received and interpreted, 1112 the intended operation will not take effect. 1114 From the LLN node: Waveform captures: A number of consecutive 1115 samples are measured at a high rate for a short time and then 1116 transferred from a sensor to a gateway or an edge server as a 1117 single large report. 1119 Data logs: LLN nodes may generate large logs of 1120 sampled data for later extraction. LLN nodes may also generate 1121 system logs to assist in diagnosing problems on the node or 1122 network. 1124 Large data packets: Rich data types might 1125 require more than one fragment. 1127 Uncontrolled firmware download or waveform upload can easily result 1128 in a massive increase of the traffic and saturate the network. 1130 When a fragment is lost in transmission, the lack of recovery in the 1131 original fragmentation system of RFC 4944 implies that all fragments 1132 would need to be resent, further contributing to the congestion that 1133 caused the initial loss, and potentially leading to congestion 1134 collapse. 1136 This saturation may lead to excessive radio interference, or random 1137 early discard (leaky bucket) in relaying nodes. Additional queuing 1138 and memory congestion may result while waiting for a low power next 1139 hop to emerge from its sleeping state. 1141 Considering that RFC 4944 defines an MTU is 1280 bytes and that in 1142 most incarnations (but 802.15.4g) a IEEE Std. 802.15.4 frame can 1143 limit the MAC payload to as few as 74 bytes, a packet might be 1144 fragmented into at least 18 fragments at the 6LoWPAN shim layer. 1145 Taking into account the worst-case header overhead for 6LoWPAN 1146 Fragmentation and Mesh Addressing headers will increase the number of 1147 required fragments to around 32. This level of fragmentation is much 1148 higher than that traditionally experienced over the Internet with 1149 IPv4 fragments. At the same time, the use of radios increases the 1150 probability of transmission loss and Mesh-Under techniques compound 1151 that risk over multiple hops. 1153 Mechanisms such as TCP or application-layer segmentation could be 1154 used to support end-to-end reliable transport. One option to support 1155 bulk data transfer over a frame-size-constrained LLN is to set the 1156 Maximum Segment Size to fit within the link maximum frame size. 1157 Doing so, however, can add significant header overhead to each 1158 802.15.4 frame. In addition, deploying such a mechanism requires 1159 that the end-to-end transport is aware of the delivery properties of 1160 the underlying LLN, which is a layer violation, and difficult to 1161 achieve from the far end of the IPv6 network. 1163 Appendix B. Requirements 1165 For one-hop communications, a number of Low Power and Lossy Network 1166 (LLN) link-layers propose a local acknowledgment mechanism that is 1167 enough to detect and recover the loss of fragments. In a multihop 1168 environment, an end-to-end fragment recovery mechanism might be a 1169 good complement to a hop-by-hop MAC level recovery. This draft 1170 introduces a simple protocol to recover individual fragments between 1171 6LoWPAN endpoints that may be multiple hops away. The method 1172 addresses the following requirements of a LLN: 1174 Number of fragments The recovery mechanism must support highly 1175 fragmented packets, with a maximum of 32 fragments per packet. 1177 Minimum acknowledgment overhead Because the radio is half duplex, 1178 and because of silent time spent in the various medium access 1179 mechanisms, an acknowledgment consumes roughly as many resources 1180 as data fragment. 1182 The new end-to-end fragment recovery mechanism should be able to 1183 acknowledge multiple fragments in a single message and not require 1184 an acknowledgment at all if fragments are already protected at a 1185 lower layer. 1187 Controlled latency The recovery mechanism must succeed or give up 1188 within the time boundary imposed by the recovery process of the 1189 Upper Layer Protocols. 1191 Optional congestion control The aggregation of multiple concurrent 1192 flows may lead to the saturation of the radio network and 1193 congestion collapse. 1195 The recovery mechanism should provide means for controlling the 1196 number of fragments in transit over the LLN. 1198 Appendix C. Considerations On Flow Control 1200 Considering that a multi-hop LLN can be a very sensitive environment 1201 due to the limited queuing capabilities of a large population of its 1202 nodes, this draft recommends a simple and conservative approach to 1203 Congestion Control, based on TCP congestion avoidance. 1205 Congestion on the forward path is assumed in case of packet loss, and 1206 packet loss is assumed upon time out. The draft allows to control 1207 the number of outstanding fragments, that have been transmitted but 1208 for which an acknowledgment was not received yet. It must be noted 1209 that the number of outstanding fragments should not exceed the number 1210 of hops in the network, but the way to figure the number of hops is 1211 out of scope for this document. 1213 Congestion on the forward path can also be indicated by an Explicit 1214 Congestion Notification (ECN) mechanism. Though whether and how ECN 1215 [RFC3168] is carried out over the LoWPAN is out of scope, this draft 1216 provides a way for the destination endpoint to echo an ECN indication 1217 back to the source endpoint in an acknowledgment message as 1218 represented in Figure 4 in Section 5.2. 1220 It must be noted that congestion and collision are different topics. 1221 In particular, when a mesh operates on a same channel over multiple 1222 hops, then the forwarding of a fragment over a certain hop may 1223 collide with the forwarding of a next fragment that is following over 1224 a previous hop but in a same interference domain. This draft enables 1225 an end-to-end flow control, but leaves it to the sender stack to pace 1226 individual fragments within a transmit window, so that a given 1227 fragment is sent only when the previous fragment has had a chance to 1228 progress beyond the interference domain of this hop. In the case of 1229 6TiSCH [I-D.ietf-6tisch-architecture], which operates over the 1230 TimeSlotted Channel Hopping [RFC7554] (TSCH) mode of operation of 1231 IEEE802.14.5, a fragment is forwarded over a different channel at a 1232 different time and it makes full sense to transmit the next fragment 1233 as soon as the previous fragment has had its chance to be forwarded 1234 at the next hop. 1236 From the standpoint of a source 6LoWPAN endpoint, an outstanding 1237 fragment is a fragment that was sent but for which no explicit 1238 acknowledgment was received yet. This means that the fragment might 1239 be on the way, received but not yet acknowledged, or the 1240 acknowledgment might be on the way back. It is also possible that 1241 either the fragment or the acknowledgment was lost on the way. 1243 From the sender standpoint, all outstanding fragments might still be 1244 in the network and contribute to its congestion. There is an 1245 assumption, though, that after a certain amount of time, a frame is 1246 either received or lost, so it is not causing congestion anymore. 1247 This amount of time can be estimated based on the round trip delay 1248 between the 6LoWPAN endpoints. The method detailed in [RFC6298] is 1249 recommended for that computation. 1251 The reader is encouraged to read through "Congestion Control 1252 Principles" [RFC2914]. Additionally [RFC7567] and [RFC5681] provide 1253 deeper information on why this mechanism is needed and how TCP 1254 handles Congestion Control. Basically, the goal here is to manage 1255 the amount of fragments present in the network; this is achieved by 1256 to reducing the number of outstanding fragments over a congested path 1257 by throttling the sources. 1259 Section 6 describes how the sender decides how many fragments are 1260 (re)sent before an acknowledgment is required, and how the sender 1261 adapts that number to the network conditions. 1263 Author's Address 1265 Pascal Thubert (editor) 1266 Cisco Systems, Inc 1267 Building D, 45 Allee des Ormes - BP1200 1268 06254 MOUGINS - Sophia Antipolis 1269 France 1271 Phone: +33 497 23 26 34 1272 Email: pthubert@cisco.com