idnits 2.17.1 draft-ietf-tsvwg-datagram-plpmtud-16.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 : ---------------------------------------------------------------------------- -- The abstract seems to indicate that this document updates RFC8201, but the header doesn't have an 'Updates:' line to match this. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC4821, updated by this document, for RFC5378 checks: 2003-10-21) -- 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 (9 March 2020) is 1508 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-34) exists of draft-ietf-quic-transport-27 ** Obsolete normative reference: RFC 4960 (Obsoleted by RFC 9260) == Outdated reference: A later version (-13) exists of draft-ietf-intarea-tunnels-10 Summary: 1 error (**), 0 flaws (~~), 3 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Internet Engineering Task Force G. Fairhurst 3 Internet-Draft T. Jones 4 Updates: 4821, 4960, 6951, 8085, 8261 (if University of Aberdeen 5 approved) M. Tuexen 6 Intended status: Standards Track I. Ruengeler 7 Expires: 10 September 2020 T. Voelker 8 Muenster University of Applied Sciences 9 9 March 2020 11 Packetization Layer Path MTU Discovery for Datagram Transports 12 draft-ietf-tsvwg-datagram-plpmtud-16 14 Abstract 16 This document describes a robust method for Path MTU Discovery 17 (PMTUD) for datagram Packetization Layers (PLs). It describes an 18 extension to RFC 1191 and RFC 8201, which specifies ICMP-based Path 19 MTU Discovery for IPv4 and IPv6. The method allows a PL, or a 20 datagram application that uses a PL, to discover whether a network 21 path can support the current size of datagram. This can be used to 22 detect and reduce the message size when a sender encounters a packet 23 black hole (where packets are discarded). The method can probe a 24 network path with progressively larger packets to discover whether 25 the maximum packet size can be increased. This allows a sender to 26 determine an appropriate packet size, providing functionality for 27 datagram transports that is equivalent to the Packetization Layer 28 PMTUD specification for TCP, specified in RFC 4821. 30 The document updates RFC 4821 to specify the method for datagram PLs, 31 and updates RFC 8085 as the method to use in place of RFC 4821 with 32 UDP datagrams. Section 7.3 of RFC4960 recommends an endpoint apply 33 the techniques in RFC 4821 on a per-destination-address basis. RFC 34 4960, RFC 6951 and RFC 8261 are updated to recommend that SCTP, SCTP 35 encapsulated in UDP and SCTP encapsulated in DTLS use the method 36 specified in this document instead of the method in RFC 4821. 38 The document also provides implementation notes for incorporating 39 Datagram PMTUD into IETF datagram transports or applications that use 40 datagram transports. 42 When published, this specification updates RFC 4960, RFC 4821, RFC 43 8085 and RFC 8261. 45 Status of This Memo 47 This Internet-Draft is submitted in full conformance with the 48 provisions of BCP 78 and BCP 79. 50 Internet-Drafts are working documents of the Internet Engineering 51 Task Force (IETF). Note that other groups may also distribute 52 working documents as Internet-Drafts. The list of current Internet- 53 Drafts is at https://datatracker.ietf.org/drafts/current/. 55 Internet-Drafts are draft documents valid for a maximum of six months 56 and may be updated, replaced, or obsoleted by other documents at any 57 time. It is inappropriate to use Internet-Drafts as reference 58 material or to cite them other than as "work in progress." 60 This Internet-Draft will expire on 10 September 2020. 62 Copyright Notice 64 Copyright (c) 2020 IETF Trust and the persons identified as the 65 document authors. All rights reserved. 67 This document is subject to BCP 78 and the IETF Trust's Legal 68 Provisions Relating to IETF Documents (https://trustee.ietf.org/ 69 license-info) in effect on the date of publication of this document. 70 Please review these documents carefully, as they describe your rights 71 and restrictions with respect to this document. Code Components 72 extracted from this document must include Simplified BSD License text 73 as described in Section 4.e of the Trust Legal Provisions and are 74 provided without warranty as described in the Simplified BSD License. 76 Table of Contents 78 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4 79 1.1. Classical Path MTU Discovery . . . . . . . . . . . . . . 4 80 1.2. Packetization Layer Path MTU Discovery . . . . . . . . . 6 81 1.3. Path MTU Discovery for Datagram Services . . . . . . . . 7 82 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 8 83 3. Features Required to Provide Datagram PLPMTUD . . . . . . . . 10 84 4. DPLPMTUD Mechanisms . . . . . . . . . . . . . . . . . . . . . 13 85 4.1. PLPMTU Probe Packets . . . . . . . . . . . . . . . . . . 13 86 4.2. Confirmation of Probed Packet Size . . . . . . . . . . . 14 87 4.3. Black Hole Detection . . . . . . . . . . . . . . . . . . 15 88 4.4. The Maximum Packet Size (MPS) . . . . . . . . . . . . . . 16 89 4.5. Disabling the Effect of PMTUD . . . . . . . . . . . . . . 17 90 4.6. Response to PTB Messages . . . . . . . . . . . . . . . . 17 91 4.6.1. Validation of PTB Messages . . . . . . . . . . . . . 17 92 4.6.2. Use of PTB Messages . . . . . . . . . . . . . . . . . 18 93 5. Datagram Packetization Layer PMTUD . . . . . . . . . . . . . 19 94 5.1. DPLPMTUD Components . . . . . . . . . . . . . . . . . . . 20 95 5.1.1. Timers . . . . . . . . . . . . . . . . . . . . . . . 20 96 5.1.2. Constants . . . . . . . . . . . . . . . . . . . . . . 21 97 5.1.3. Variables . . . . . . . . . . . . . . . . . . . . . . 22 98 5.1.4. Overview of DPLPMTUD Phases . . . . . . . . . . . . . 23 99 5.2. State Machine . . . . . . . . . . . . . . . . . . . . . . 24 100 5.3. Search to Increase the PLPMTU . . . . . . . . . . . . . . 27 101 5.3.1. Probing for a larger PLPMTU . . . . . . . . . . . . . 27 102 5.3.2. Selection of Probe Sizes . . . . . . . . . . . . . . 28 103 5.3.3. Resilience to Inconsistent Path Information . . . . . 28 104 5.4. Robustness to Inconsistent Paths . . . . . . . . . . . . 29 105 6. Specification of Protocol-Specific Methods . . . . . . . . . 29 106 6.1. Application support for DPLPMTUD with UDP or UDP-Lite . . 29 107 6.1.1. Application Request . . . . . . . . . . . . . . . . . 30 108 6.1.2. Application Response . . . . . . . . . . . . . . . . 30 109 6.1.3. Sending Application Probe Packets . . . . . . . . . . 30 110 6.1.4. Initial Connectivity . . . . . . . . . . . . . . . . 30 111 6.1.5. Validating the Path . . . . . . . . . . . . . . . . . 30 112 6.1.6. Handling of PTB Messages . . . . . . . . . . . . . . 30 113 6.2. DPLPMTUD for SCTP . . . . . . . . . . . . . . . . . . . . 31 114 6.2.1. SCTP/IPv4 and SCTP/IPv6 . . . . . . . . . . . . . . . 31 115 6.2.1.1. Initial Connectivity . . . . . . . . . . . . . . 31 116 6.2.1.2. Sending SCTP Probe Packets . . . . . . . . . . . 31 117 6.2.1.3. Validating the Path with SCTP . . . . . . . . . . 32 118 6.2.1.4. PTB Message Handling by SCTP . . . . . . . . . . 32 119 6.2.2. DPLPMTUD for SCTP/UDP . . . . . . . . . . . . . . . . 32 120 6.2.2.1. Initial Connectivity . . . . . . . . . . . . . . 32 121 6.2.2.2. Sending SCTP/UDP Probe Packets . . . . . . . . . 33 122 6.2.2.3. Validating the Path with SCTP/UDP . . . . . . . . 33 123 6.2.2.4. Handling of PTB Messages by SCTP/UDP . . . . . . 33 124 6.2.3. DPLPMTUD for SCTP/DTLS . . . . . . . . . . . . . . . 33 125 6.2.3.1. Initial Connectivity . . . . . . . . . . . . . . 33 126 6.2.3.2. Sending SCTP/DTLS Probe Packets . . . . . . . . . 33 127 6.2.3.3. Validating the Path with SCTP/DTLS . . . . . . . 33 128 6.2.3.4. Handling of PTB Messages by SCTP/DTLS . . . . . . 34 129 6.3. DPLPMTUD for QUIC . . . . . . . . . . . . . . . . . . . . 34 130 6.3.1. Initial Connectivity . . . . . . . . . . . . . . . . 34 131 6.3.2. Sending QUIC Probe Packets . . . . . . . . . . . . . 34 132 6.3.3. Validating the Path with QUIC . . . . . . . . . . . . 35 133 6.3.4. Handling of PTB Messages by QUIC . . . . . . . . . . 35 134 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 35 135 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 35 136 9. Security Considerations . . . . . . . . . . . . . . . . . . . 35 137 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 37 138 10.1. Normative References . . . . . . . . . . . . . . . . . . 37 139 10.2. Informative References . . . . . . . . . . . . . . . . . 38 140 Appendix A. Revision Notes . . . . . . . . . . . . . . . . . . . 39 141 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 43 143 1. Introduction 145 The IETF has specified datagram transport using UDP, SCTP, and DCCP, 146 as well as protocols layered on top of these transports (e.g., SCTP/ 147 UDP, DCCP/UDP, QUIC/UDP), and direct datagram transport over the IP 148 network layer. This document describes a robust method for Path MTU 149 Discovery (PMTUD) that can be used with these transport protocols (or 150 the applications that use their transport service) to discover an 151 appropriate size of packet to use across an Internet path. 153 1.1. Classical Path MTU Discovery 155 Classical Path Maximum Transmission Unit Discovery (PMTUD) can be 156 used with any transport that is able to process ICMP Packet Too Big 157 (PTB) messages (e.g., [RFC1191] and [RFC8201]). In this document, 158 the term PTB message is applied to both IPv4 ICMP Unreachable 159 messages (type 3) that carry the error Fragmentation Needed (Type 3, 160 Code 4) [RFC0792] and ICMPv6 Packet Too Big messages (Type 2) 161 [RFC4443]. When a sender receives a PTB message, it reduces the 162 effective MTU to the value reported as the Link MTU in the PTB 163 message. A method from time-to-time increases the packet size in 164 attempt to discover an increase in the supported PMTU. The packets 165 sent with a size larger than the current effective PMTU are known as 166 probe packets. 168 Packets not intended as probe packets are either fragmented to the 169 current effective PMTU, or the attempt to send fails with an error 170 code. Applications can be provided with a primitive to let them read 171 the Maximum Packet Size (MPS), derived from the current effective 172 PMTU. 174 Classical PMTUD is subject to protocol failures. One failure arises 175 when traffic using a packet size larger than the actual PMTU is 176 black-holed (all datagrams sent with this size, or larger, are 177 discarded). This could arise when the PTB messages are not delivered 178 back to the sender for some reason (see for example [RFC2923]). 180 Examples where PTB messages are not delivered include: 182 * The generation of ICMP messages is usually rate limited. This 183 could result in no PTB messages being generated to the sender (see 184 section 2.4 of [RFC4443]) 186 * ICMP messages can be filtered by middleboxes (including firewalls) 187 [RFC4890]. A stateful firewall could be configured with a policy 188 to block incoming ICMP messages, which would prevent reception of 189 PTB messages to a sending endpoint behind this firewall. 191 * When the router issuing the ICMP message drops a tunneled packet, 192 the resulting ICMP message will be directed to the tunnel ingress. 193 This tunnel endpoint is responsible for forwarding the ICMP 194 message and also processing the quoted packet within the payload 195 field to remove the effect of the tunnel, and return a correctly 196 formatted ICMP message to the sender [I-D.ietf-intarea-tunnels]. 197 Failure to do this prevents the PTB message reaching the original 198 sender. 200 * Asymmetry in forwarding can result in there being no return route 201 to the original sender, which would prevent an ICMP message being 202 delivered to the sender. This issue can also arise when policy- 203 based routing is used, Equal Cost Multipath (ECMP) routing is 204 used, or a middlebox acts as an application load balancer. An 205 example is where the path towards the server is chosen by ECMP 206 routing depending on bytes in the IP payload. In this case, when 207 a packet sent by the server encounters a problem after the ECMP 208 router, then any resulting ICMP message also needs to be directed 209 by the ECMP router towards the original sender. 211 * There are additional cases where the next hop destination fails to 212 receive a packet because of its size. This could be due to 213 misconfiguration of the layer 2 path between nodes, for instance 214 the MTU configured in a layer 2 switch, or misconfiguration of the 215 Maximum Receive Unit (MRU). If a packet is dropped by the link, 216 this will not cause a PTB message to be sent to the original 217 sender. 219 Another failure could result if a node that is not on the network 220 path sends a PTB message that attempts to force a sender to change 221 the effective PMTU [RFC8201]. A sender can protect itself from 222 reacting to such messages by utilising the quoted packet within a PTB 223 message payload to validate that the received PTB message was 224 generated in response to a packet that had actually originated from 225 the sender. However, there are situations where a sender would be 226 unable to provide this validation. Examples where validation of the 227 PTB message is not possible include: 229 * When a router issuing the ICMP message implements RFC792 230 [RFC0792], it is only required to include the first 64 bits of the 231 IP payload of the packet within the quoted payload. There could 232 be insufficient bytes remaining for the sender to interpret the 233 quoted transport information. 235 Note: The recommendation in RFC1812 [RFC1812] is that IPv4 routers 236 return a quoted packet with as much of the original datagram as 237 possible without the length of the ICMP datagram exceeding 576 238 bytes. IPv6 routers include as much of the invoking packet as 239 possible without the ICMPv6 packet exceeding 1280 bytes [RFC4443]. 241 * The use of tunnels/encryption can reduce the size of the quoted 242 packet returned to the original source address, increasing the 243 risk that there could be insufficient bytes remaining for the 244 sender to interpret the quoted transport information. 246 * Even when the PTB message includes sufficient bytes of the quoted 247 packet, the network layer could lack sufficient context to 248 validate the message, because validation depends on information 249 about the active transport flows at an endpoint node (e.g., the 250 socket/address pairs being used, and other protocol header 251 information). 253 * When a packet is encapsulated/tunneled over an encrypted 254 transport, the tunnel/encapsulation ingress might have 255 insufficient context, or computational power, to reconstruct the 256 transport header that would be needed to perform validation. 258 * A Network Addres Translation (NAT) device that translates a packet 259 header, ought to also translate ICMP messages and update the ICMP 260 quoted packet [RFC5508] in that message. If this is not correctly 261 translated then the sender would not be able to associate the 262 message with the PL that originated the packet, and hence this 263 ICMP message cannot be validated. 265 1.2. Packetization Layer Path MTU Discovery 267 The term Packetization Layer (PL) has been introduced to describe the 268 layer that is responsible for placing data blocks into the payload of 269 IP packets and selecting an appropriate MPS. This function is often 270 performed by a transport protocol (e.g., DCCP, RTP, SCTP, QUIC), but 271 can also be performed by other encapsulation methods working above 272 the transport layer. 274 In contrast to PMTUD, Packetization Layer Path MTU Discovery 275 (PLPMTUD) [RFC4821] introduced a method that does not rely upon 276 reception and validation of PTB messages. It is therefore more 277 robust than Classical PMTUD. This has become the recommended 278 approach for implementing discovery of the PMTU [RFC8085]. 280 It uses a general strategy where the PL sends probe packets to search 281 for the largest size of unfragmented datagram that can be sent over a 282 network path. Probe packets are sent to explore using a larger 283 packet size. If a probe packet is successfully delivered (as 284 determined by the PL), then the PLPMTU is raised to the size of the 285 successful probe. If no response is received to a probe packet, the 286 method then reduces the PLPMTU. 288 Datagram PLPMTUD introduces flexibility in implementation. At one 289 extreme, it can be configured to only perform Black Hole Detection 290 and recovery with increased robustness compared to Classical PMTUD. 291 At the other extreme, all PTB processing can be disabled, and PLPMTUD 292 replaces Classical PMTUD. 294 PLPMTUD can also include additional consistency checks without 295 increasing the risk that data is lost when probing to discover the 296 Path MTU. For example, information available at the PL, or higher 297 layers, enables received PTB messages to be validated before being 298 utilized. 300 1.3. Path MTU Discovery for Datagram Services 302 Section 5 of this document presents a set of algorithms for datagram 303 protocols to discover the largest size of unfragmented datagram that 304 can be sent over a network path. The method relies upon features of 305 the PL described in Section 3 and applies to transport protocols 306 operating over IPv4 and IPv6. It does not require cooperation from 307 the lower layers, although it can utilize PTB messages when these 308 received messages are made available to the PL. 310 The message size guidelines in section 3.2 of the UDP Usage 311 Guidelines [RFC8085] state "an application SHOULD either use the Path 312 MTU information provided by the IP layer or implement Path MTU 313 Discovery (PMTUD)", but does not provide a mechanism for discovering 314 the largest size of unfragmented datagram that can be used on a 315 network path. The present document updates RFC 8085 to specify this 316 method in place of PLPMTUD [RFC4821] and provides a mechanism for 317 sharing the discovered largest size as the Maximum Packet Size (MPS) 318 (see Section 4.4). 320 Section 10.2 of [RFC4821] recommended a PLPMTUD probing method for 321 the Stream Control Transport Protocol (SCTP). SCTP utilizes probe 322 packets consisting of a minimal sized HEARTBEAT chunk bundled with a 323 PAD chunk as defined in [RFC4820]. However, RFC 4821 did not provide 324 a complete specification. The present document replaces this by 325 providing a complete specification. 327 The Datagram Congestion Control Protocol (DCCP) [RFC4340] requires 328 implementations to support Classical PMTUD and states that a DCCP 329 sender "MUST maintain the MPS allowed for each active DCCP session". 330 It also defines the current congestion control MPS (CCMPS) supported 331 by a network path. This recommends use of PMTUD, and suggests use of 332 control packets (DCCP-Sync) as path probe packets, because they do 333 not risk application data loss. The method defined in this 334 specification can be used with DCCP. 336 Section 6 specifies the method for datagram transports and provides 337 information to enable the implementation of PLPMTUD with other 338 datagram transports and applications that use datagram transports. 340 Section 6 also provides updated recommendations for [RFC6951] and 341 [RFC8261]. 343 2. Terminology 345 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 346 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 347 "OPTIONAL" in this document are to be interpreted as described in BCP 348 14 [RFC2119] [RFC8174] when, and only when, they appear in all 349 capitals, as shown here. 351 Other terminology is directly copied from [RFC4821], and the 352 definitions in [RFC1122]. 354 Actual PMTU: The Actual PMTU is the PMTU of a network path between a 355 sender PL and a destination PL, which the DPLPMTUD algorithm seeks 356 to determine. 358 Black Hole: A Black Hole is encountered when a sender is unaware 359 that packets are not being delivered to the destination end point. 360 Two types of Black Hole are relevant to DPLPMTUD: 362 * Packets encounter a packet Black Hole when packets are not 363 delivered to the destination endpoint (e.g., when the sender 364 transmits packets of a particular size with a previously known 365 effective PMTU and they are discarded by the network). 367 * An ICMP Black Hole is encountered when the sender is unaware 368 that packets are not delivered to the destination endpoint 369 because PTB messages are not received by the originating PL 370 sender. 372 Classical Path MTU Discovery: Classical PMTUD is a process described 373 in [RFC1191] and [RFC8201], in which nodes rely on PTB messages to 374 learn the largest size of unfragmented packet that can be used 375 across a network path. 377 Datagram: A datagram is a transport-layer protocol data unit, 378 transmitted in the payload of an IP packet. 380 Effective PMTU: The Effective PMTU is the current estimated value 381 for PMTU that is used by a PMTUD. This is equivalent to the 382 PLPMTU derived by PLPMTUD. 384 EMTU_S: The Effective MTU for sending (EMTU_S) is defined in 385 [RFC1122] as "the maximum IP datagram size that may be sent, for a 386 particular combination of IP source and destination addresses...". 388 EMTU_R: The Effective MTU for receiving (EMTU_R) is designated in 389 [RFC1122] as the largest datagram size that can be reassembled by 390 EMTU_R (Effective MTU to receive). 392 Link: A Link is a communication facility or medium over which nodes 393 can communicate at the link layer, i.e., a layer below the IP 394 layer. Examples are Ethernet LANs and Internet (or higher) layer 395 and tunnels. 397 Link MTU: The Link Maximum Transmission Unit (MTU) is the size in 398 bytes of the largest IP packet, including the IP header and 399 payload, that can be transmitted over a link. Note that this 400 could more properly be called the IP MTU, to be consistent with 401 how other standards organizations use the acronym. This includes 402 the IP header, but excludes link layer headers and other framing 403 that is not part of IP or the IP payload. Other standards 404 organizations generally define the link MTU to include the link 405 layer headers. This specification continues the requirement in 406 [RFC4821], that states "All links MUST enforce their MTU: links 407 that might non- deterministically deliver packets that are larger 408 than their rated MTU MUST consistently discard such packets." 410 MAX_PMTU: The MAX_PMTU is the largest size of PLPMTU that DPLPMTUD 411 will attempt to use. 413 MPS: The Maximum Packet Size (MPS) is the largest size of 414 application data block that can be sent across a network path by a 415 PL. In DPLPMTUD this quantity is derived from the PLPMTU by 416 taking into consideration the size of the lower protocol layer 417 headers. Probe packets generated by DPLPMTUD can have a size 418 larger than the MPS. 420 MIN_PMTU: The MIN_PMTU is the smallest size of PLPMTU that DPLPMTUD 421 will attempt to use. 423 Packet: A Packet is the IP header plus the IP payload. 425 Packetization Layer (PL): The Packetization Layer (PL) is a layer of 426 the network stack that places data into packets and performs 427 transport protocol functions. Examples of a PL include: TCP, 428 SCTP, SCTP over DTLS or QUIC. 430 Path: The Path is the set of links and routers traversed by a packet 431 between a source node and a destination node by a particular flow. 433 Path MTU (PMTU): The Path MTU (PMTU) is the minimum of the Link MTU 434 of all the links forming a network path between a source node and 435 a destination node. 437 PTB_SIZE: The PTB_SIZE is a value reported in a validated PTB 438 message that indicates next hop link MTU of a router along the 439 path. 441 PLPMTU: The Packetization Layer PMTU is an estimate of the actual 442 PMTU provided by the DPLPMTUD algorithm. 444 PLPMTUD: Packetization Layer Path MTU Discovery (PLPMTUD), the 445 method described in this document for datagram PLs, which is an 446 extension to Classical PMTU Discovery. 448 Probe packet: A probe packet is a datagram sent with a purposely 449 chosen size (typically the current PLPMTU or larger) to detect if 450 packets of this size can be successfully sent end-to-end across 451 the network path. 453 3. Features Required to Provide Datagram PLPMTUD 455 The principles expressed in [RFC4821] apply to the use of the 456 technique with any PL. TCP PLPMTUD has been defined using standard 457 TCP protocol mechanisms. Unlike TCP, datagram PLs require additional 458 mechanisms and considerations to implement PLPMTUD. 460 The requirements for datagram PLPMTUD are: 462 1. PLPMTU: The PLPMTU (specified as the effective PMTU in Section 1 463 of [RFC1191]) is equivalent to the EMTU_S (specified in 464 [RFC1122]). For datagram PLs,] the PLPMTU is managed by 465 DPLPMTUD. A PL MUST NOT send a packet (other than a probe 466 packet) with a size larger than the current PLPMTU at the 467 network layer. 469 2. Probe packets: On request, a DPLPMTUD sender is REQUIRED to be 470 able to transmit a packet larger than the PLMPMTU. This is used 471 to send a probe packet. In IPv4, a probe packet MUST be sent 472 with the Don't Fragment (DF) bit set in the IP header, and 473 without network layer endpoint fragmentation. In IPv6, a probe 474 packet is always sent without source fragmentation (as specified 475 in section 5.4 of [RFC8201]). 477 3. Reception feedback: The destination PL endpoint is REQUIRED to 478 provide a feedback method that indicates to the DPLPMTUD sender 479 when a probe packet has been received by the destination PL 480 endpoint. 482 4. Probe loss recovery: It is RECOMMENDED to use probe packets that 483 do not carry any user data that would require retransmission if 484 lost. Most datagram transports permit this. If a probe packet 485 contains user data requiring retransmission in case of loss, the 486 PL (or layers above) are REQUIRED to arrange any retransmission/ 487 repair of any resulting loss. The PL is REQUIRED to be robust 488 in the case where probe packets are lost due to other reasons 489 (including link transmission error, congestion). 491 5. PMTU parameters: A DPLPMTUD sender is RECOMMENDED to utilise 492 information about the maximum size of packet that can be 493 transmitted by the sender on the local link (e.g., the local 494 Link MTU). It MAY utilize similar information about the 495 receiver when this is supplied (note this could be less than 496 EMTU_R). This avoids implementations trying to send probe 497 packets that can not be transmitted by the local link. Too high 498 of a value could reduce the efficiency of the search algorithm. 499 Some applications also have a maximum transport protocol data 500 unit (PDU) size, in which case there is no benefit from probing 501 for a size larger than this (unless a transport allows 502 multiplexing multiple applications PDUs into the same datagram). 504 6. Processing PTB messages: A DPLPMTUD sender MAY optionally 505 utilize PTB messages received from the network layer to help 506 identify when a network path does not support the current size 507 of probe packet. Any received PTB message MUST be validated 508 before it is used to update the PLPMTU discovery information 509 [RFC8201]. This validation confirms that the PTB message was 510 sent in response to a packet originating by the sender, and 511 needs to be performed before the PLPMTU discovery method reacts 512 to the PTB message. A PTB message MUST NOT be used to increase 513 the PLPMTU [RFC8201], but could trigger a probe to test for a 514 larger PLPMTU. A PTB_SIZE greater than the currently probed 515 MUST be ignored. 517 7. Probing and congestion control: The decision about when to send 518 a probe packet does not need to be limited by the congestion 519 controller. When not controlled by the congestion controller, 520 the interval between probe packets MUST be at least one RTT. If 521 transmission of probe packets is limited by the congestion 522 controller, this could result in transmission of probe packets 523 being delayed. 525 8. Loss of a probe packet SHOULD NOT be treated as an indication of 526 congestion and SHOULD NOT trigger a congestion control reaction 527 [RFC4821], because this could result in unnecessary reduction of 528 the sending rate. 530 9. An update to the PLPMTU (or MPS) MUST NOT modify the congestion 531 window measured in bytes [RFC4821]. Therefore, an increase in 532 the packet size does not cause an increase the data rate in 533 bytes per second. 535 10. Probing and flow control: Flow control at the PL concerns the 536 end-to-end flow of data using the PL service. This does not 537 apply to DPLPMTU when probe packets use a design that does not 538 carry user data to the remote application. 540 11. Shared PLPMTU state: The PLPMTU value MAY also be stored with 541 the corresponding entry associated with the destination in the 542 IP layer cache, and used by other PL instances. The 543 specification of PLPMTUD [RFC4821] states: "If PLPMTUD updates 544 the MTU for a particular path, all Packetization Layer sessions 545 that share the path representation (as described in Section 5.2 546 of [RFC4821]) SHOULD be notified to make use of the new MTU". 547 Such methods MUST be robust to the wide variety of underlying 548 network forwarding behaviors. Section 5.2 of [RFC8201] provides 549 guidance on the caching of PMTU information and also the 550 relation to IPv6 flow labels. 552 In addition, the following principles are stated for design of a 553 DPLPMTUD method: 555 * Maximum Packet Size (MPS): A PL MAY be designed to segment data 556 blocks larger than the MPS into multiple datagrams. However, not 557 all datagram PLs support segmentation of data blocks. It is 558 RECOMMENDED that methods avoid forcing an application to use an 559 arbitrary small MPS for transmission while the method is searching 560 for the currently supported PLPMTU. A reduced MPS can adversely 561 impact the performance of an application. 563 * To assist applications in choosing a suitable data block size, the 564 PL is RECOMMENDED to provide a primitive that returns the MPS 565 derived from the PLPMTU to the higher layer using the PL. The 566 value of the MPS can change following a change in the path, or 567 loss of probe packets. 569 * Path validation: It is RECOMMENDED that methods are robust to path 570 changes that could have occurred since the path characteristics 571 were last confirmed, and to the possibility of inconsistent path 572 information being received. 574 * Datagram reordering: A method is REQUIRED to be robust to the 575 possibility that a flow encounters reordering, or the traffic 576 (including probe packets) is divided over more than one network 577 path. 579 * Datagram delay and duplication: The feedback mechanism is REQUIRED 580 to be robust to the possibility that packets could be 581 significantly delayed or duplicated along a network path. 583 * When to probe: It is RECOMMENDED that methods determine whether 584 the path has changed since it last measured the path. This can 585 help determine when to probe the path again. 587 4. DPLPMTUD Mechanisms 589 This section lists the protocol mechanisms used in this 590 specification. 592 4.1. PLPMTU Probe Packets 594 The DPLPMTUD method relies upon the PL sender being able to generate 595 probe packets with a specific size. TCP is able to generate these 596 probe packets by choosing to appropriately segment data being sent 597 [RFC4821]. In contrast, a datagram PL that constructs a probe packet 598 has to either request an application to send a data block that is 599 larger than that generated by an application, or to utilize padding 600 functions to extend a datagram beyond the size of the application 601 data block. Protocols that permit exchange of control messages 602 (without an application data block) can generate a probe packet by 603 extending a control message with padding data. The total size of a 604 probe packet includes all headers and padding added to the payload 605 data being sent (e.g., including protocol option fields, security- 606 related fields such as an AEAD tag and TLS record layer padding). 608 A receiver is REQUIRED to be able to distinguish an in-band data 609 block from any added padding. This is needed to ensure that any 610 added padding is not passed on to an application at the receiver. 612 This results in three possible ways that a sender can create a probe 613 packet: 615 Probing using padding data: A probe packet that contains only 616 control information together with any padding, which is needed to 617 be inflated to the size of the probe packet. Since these probe 618 packets do not carry an application-supplied data block, they do 619 not typically require retransmission, although they do still 620 consume network capacity and incur endpoint processing. 622 Probing using application data and padding data: A probe packet that 623 contains a data block supplied by an application that is combined 624 with padding to inflate the length of the datagram to the size of 625 the probe packet. 627 Probing using application data: A probe packet that contains a data 628 block supplied by an application that matches the size of the 629 probe packet. This method requests the application to issue a 630 data block of the desired probe size. 632 A PL that uses a probe packet carrying an application data and needs 633 protection from the loss of this probe packet, could perform 634 transport-layer retransmission/repair of the data block (e.g., by 635 retransmission after loss is detected or by duplicating the data 636 block in a datagram without the padding data). This retransmited 637 data block might possibly need to be sent using a smaller PLPMTU, 638 which could need the PL to to use a smaller packet size to traverse 639 the end-to-end path. (This could utilize endpoint network-layer or a 640 PL that can re-segment the data block into multiple datagrams). 642 DPLPMTUD MAY choose to use only one of these methods to simplify the 643 implementation. 645 Probe messages sent by a PL MUST contain enough information to 646 uniquely identify the probe within Maximum Segment Lifetime, while 647 being robust to reordering and replay of probe response and PTB 648 messages. 650 4.2. Confirmation of Probed Packet Size 652 The PL needs a method to determine (confirm) when probe packets have 653 been successfully received end-to-end across a network path. 655 Transport protocols can include end-to-end methods that detect and 656 report reception of specific datagrams that they send (e.g., DCCP and 657 SCTP provide keep-alive/heartbeat features). When supported, this 658 mechanism MAY also be used by DPLPMTUD to acknowledge reception of a 659 probe packet. 661 A PL that does not acknowledge data reception (e.g., UDP and UDP- 662 Lite) is unable itself to detect when the packets that it sends are 663 discarded because their size is greater than the actual PMTU. These 664 PLs need to rely on an application protocol to detect this loss. 666 Section 6 specifies this function for a set of IETF-specified 667 protocols. 669 4.3. Black Hole Detection 671 Black Hole Detection is triggered by an indication that the network 672 path could be unable to support the current PLPMTU size. 674 There are three ways to detect black holes: 676 * A validated PTB message can be received that indicates a PTB_SIZE 677 less than the current PLPMTU. A DPLPMTUD method MUST NOT rely 678 soley on this method. 680 * A PL can use the DPLPMTUD probing mechanism to periodically 681 generate probe packets of the size of the current PLPMTU (e.g., 682 using the confirmation timer Section 5.1.1). A timer tracks 683 whether acknowledgments are received. Successive loss of probes 684 is an indication that the current path no longer supports the 685 PLPMTU (e.g., when the number of probe packets sent without 686 receiving an acknowledgement, PROBE_COUNT, becomes greater than 687 MAX_PROBES). 689 * A PL can utilise an event that indicates the network path no 690 longer sustains the sender's PLPMTU size. This could use a 691 mechanism implemented within the PL to detect excessive loss of 692 data sent with a specific packet size and then conclude that this 693 excessive loss could be a result of an invalid PLPMTU (as in 694 PLPMTUD for TCP [RFC4821]). 696 A PL MAY inhibit sending probe packets when no application data has 697 been sent since the previous probe packet. A PL preferring to use an 698 up-to-data PLPMTU once user data is sent again, MAY choose to 699 continue PLPMTU discovery for each path. However, this could result 700 in additional packets being sent. 702 When the method detects the current PLPMTU is not supported, DPLPMTUD 703 sets a lower PLPMTU, and sets a lower MPS. The PL then confirms that 704 the new PLPMTU can be successfully used across the path. A probe 705 packet could need to have a size less than the size of the data block 706 generated by the application. 708 4.4. The Maximum Packet Size (MPS) 710 The result of probing determines a usable PLPMTU, which is used to 711 set the MPS used by the application. The MPS is smaller than the 712 PLPMTU because it is reduced by the size of PL headers (including the 713 overhead of security-related fields such as an AEAD tag and TLS 714 record layer padding) and any IP options or extensions added to the 715 PL packet. The relationship between the MPS and the PLPMTUD is 716 illustrated in Figure 1. 718 any additional 719 headers .--- MPS -----. 720 | | | 721 v v v 722 +------------------------------+ 723 | IP | ** | PL | protocol data | 724 +------------------------------+ 726 <---------- PLPMTU ------------> 728 Figure 1: Relationship between MPS and PLPMTU 730 A PL is unable to send a packet (other than a probe packet) with a 731 size larger than the current PLPMTU at the network layer. To avoid 732 this, a PL MAY be designed to segment data blocks larger than the MPS 733 into multiple datagrams. 735 DPLPMTUD seeks to avoid IP fragmentation. An attempt to send a data 736 block larger than the MPS will therefore fail if a PL is unable to 737 segment data. To determine the largest data block that can be sent, 738 a PL SHOULD provide applications with a primitive that returns the 739 Maximum Packet Size (MPS), derived from the current PLPMTU. 741 If DPLPMTUD results in a change to the MPS, the application needs to 742 adapt to the new MPS. A particular case can arise when packets have 743 been sent with a size less than the MPS and the PLPMTU was 744 subsequently reduced. If these packets are lost, the PL MAY segment 745 the data using the new MPS. If a PL is unable to re-segment a 746 previously sent datagram (e.g., [RFC4960]), then the sender either 747 discards the datagram or could perform retransmission using network- 748 layer fragmentation to form multiple IP packets not larger than the 749 PLPMTU. For IPv4, the use of endpoint fragmentation by the sender is 750 preferred over clearing the DF-bit in the IPv4 header. Operational 751 experience reveals that IP fragmentation can reduce the reliability 752 of Internet communication [I-D.ietf-intarea-frag-fragile], which may 753 reduce the success of retransmission. 755 4.5. Disabling the Effect of PMTUD 757 A PL implementing this specification MUST suspend network layer 758 processing of outgoing packets that enforces a PMTU 759 [RFC1191][RFC8201] for each flow utilising DPLPMTUD, and instead use 760 DPLPMTUD to control the size of packets that are sent by a flow. 761 This removes the need for the network layer to drop or fragment sent 762 packets that have a size greater than the PMTU. 764 4.6. Response to PTB Messages 766 This method requires the DPLPMTUD sender to validate any received PTB 767 message before using the PTB information. The response to a PTB 768 message depends on the PTB_SIZE indicated in the PTB message, the 769 state of the PLPMTUD state machine, and the IP protocol being used. 771 Section 4.6.1 first describes validation for both IPv4 ICMP 772 Unreachable messages (type 3) and ICMPv6 Packet Too Big messages, 773 both of which are referred to as PTB messages in this document. 775 4.6.1. Validation of PTB Messages 777 This section specifies utilization of PTB messages. 779 * A simple implementation MAY ignore received PTB messages and in 780 this case the PLPMTU is not updated when a PTB message is 781 received. 783 * An implementation that supports PTB messages MUST validate 784 messages before they are further processed. 786 A PL that receives a PTB message from a router or middlebox, performs 787 ICMP validation as specified in Section 5.2 of [RFC8085][RFC8201]. 788 Because DPLPMTUD operates at the PL, the PL needs to check that each 789 received PTB message is received in response to a packet transmitted 790 by the endpoint PL performing DPLPMTUD. 792 The PL MUST check the protocol information in the quoted packet 793 carried in an ICMP PTB message payload to validate the message 794 originated from the sending node. This validation includes 795 determining that the combination of the IP addresses, the protocol, 796 the source port and destination port match those returned in the 797 quoted packet - this is also necessary for the PTB message to be 798 passed to the corresponding PL. 800 The validation SHOULD utilize information that it is not simple for 801 an off-path attacker to determine [RFC8085]. For example, by 802 checking the value of a protocol header field known only to the two 803 PL endpoints. A datagram application that uses well-known source and 804 destination ports ought to also rely on other information to complete 805 this validation. 807 These checks are intended to provide protection from packets that 808 originate from a node that is not on the network path. A PTB message 809 that does not complete the validation MUST NOT be further utilized by 810 the DPLPMTUD method. 812 PTB messages that have been validated MAY be utilized by the DPLPMTUD 813 algorithm, but MUST NOT be used directly to set the PLPMTU. A method 814 that utilizes these PTB messages can improve the speed at the which 815 the algorithm detects an appropriate PLPMTU by triggering an 816 immediate probe for the PTB_SIZE, compared to one that relies solely 817 on probing using a timer-based search algorithm. Section 4.6.2 818 describes this processing. 820 4.6.2. Use of PTB Messages 822 A set of checks are intended to provide protection from a router that 823 reports an unexpected PTB_SIZE. The PL also needs to check that the 824 indicated PTB_SIZE is less than the size used by probe packets and at 825 least the minimum size accepted. 827 This section provides a summary of how PTB messages can be utilized. 828 This processing depends on the PTB_SIZE and the current value of a 829 set of variables: 831 PTB_SIZE < MIN_PMTU 832 * Invalid PTB_SIZE see Section 4.6.1. 834 * PTB message ought to be discarded without further processing 835 (e. g. PLPMTU not modified). 837 * The information could be utilized as an input to trigger 838 enabling a resilience mode. 840 MIN_PMTU < PTB_SIZE < BASE_PMTU 841 * A robust PL MAY enter an error state (see Section 5.2) for an 842 IPv4 path when the PTB_SIZE reported in the PTB message is 843 larger than or equal to 68 bytes [RFC0791] and when this is 844 less than the BASE_PMTU. 846 * A robust PL MAY enter an error state (see Section 5.2) for an 847 IPv6 path when the PTB_SIZE reported in the PTB message is 848 larger than or equal to 1280 bytes [RFC8200] and when this is 849 less than the BASE_PMTU. 851 PTB_SIZE = PLPMTU 852 * Completes the search for a larger PLPMTU. 854 PTB_SIZE > PROBED_SIZE 855 * Inconsistent network signal. 857 * PTB message ought to be discarded without further processing 858 (e. g. PLPMTU not modified). 860 * The information could be utilized as an input to trigger 861 enabling a resilience mode. 863 BASE_PMTU <= PTB_SIZE < PLPMTU 864 * This could be an indication of a black hole. The PLPMTU SHOULD 865 be set to BASE_PMTU (the PLPMTU is reduced to the BASE_PMTU to 866 avoid unnecessary packet loss when a black hole is 867 encountered). 869 * The PL ought to start a search to quickly discover the new 870 PLPMTU. The PTB_SIZE reported in the PTB message can be used 871 to initialize a search algorithm. 873 PLPMTU < PTB_SIZE < PROBED_SIZE 874 * The PLPMTU continues to be valid, but the last PROBED_SIZE 875 searched was larger than the actual PMTU. 877 * The PLPMTU is not updated. 879 * The PL can use the reported PTB_SIZE from the PTB message as 880 the next search point when it resumes the search algorithm. 882 5. Datagram Packetization Layer PMTUD 884 This section specifies Datagram PLPMTUD (DPLPMTUD). The method can 885 be introduced at various points (as indicated with * in the figure 886 below) in the IP protocol stack to discover the PLPMTU so that an 887 application can utilize an appropriate MPS for the current network 888 path. 890 DPLPMTUD SHOULD NOT be used by an upper PL or application if it is 891 already used in a lower layer, DPLPMTUD SHOULD only be performed once 892 between a pair of endpoints. A PL MUST adjust the MPS indicated by 893 DPLPMTUD to account for any additional overhead introduced by the PL. 895 +----------------------+ 896 | Application* | 897 +-+-------+----+----+--+ 898 | | | | 899 +---+--+ +--+--+ | +-+---+ 900 | QUIC*| |UDPO*| | |SCTP*| 901 +---+--+ +--+--+ | +--+--+ 902 | | | | | 903 +-------+--+ | | | 904 | | | | 905 +-+-+--+ | 906 | UDP | | 907 +---+--+ | 908 | | 909 +--------------+-----+-+ 910 | Network Interface | 911 +----------------------+ 913 Figure 2: Examples where DPLPMTUD can be implemented 915 The central idea of DPLPMTUD is probing by a sender. Probe packets 916 are sent to find the maximum size of user message that can be 917 completely transferred across the network path from the sender to the 918 destination. 920 The following sections identify the components needed for 921 implementation, provides an overview of the phases of operation, and 922 specifies the state machine and search algorithm. 924 5.1. DPLPMTUD Components 926 This section describes the timers, constants, and variables of 927 DPLPMTUD. 929 5.1.1. Timers 931 The method utilizes up to three timers: 933 PROBE_TIMER: The PROBE_TIMER is configured to expire after a period 934 longer than the maximum time to receive an acknowledgment to a 935 probe packet. This value MUST NOT be smaller than 1 second, and 936 SHOULD be larger than 15 seconds. Guidance on selection of the 937 timer value are provided in section 3.1.1 of the UDP Usage 938 Guidelines [RFC8085]. 940 PMTU_RAISE_TIMER: The PMTU_RAISE_TIMER is configured to the period a 941 sender will continue to use the current PLPMTU, after which it re- 942 enters the Search phase. This timer has a period of 600 seconds, 943 as recommended by PLPMTUD [RFC4821]. 945 DPLPMTUD MAY inhibit sending probe packets when no application 946 data has been sent since the previous probe packet. A PL 947 preferring to use an up-to-data PMTU once user data is sent again, 948 can choose to continue PMTU discovery for each path. However, 949 this could result in sending additional packets. 951 CONFIRMATION_TIMER: When an acknowledged PL is used, this timer MUST 952 NOT be used. For other PLs, the CONFIRMATION_TIMER is configured 953 to the period a PL sender waits before confirming the current 954 PLPMTU is still supported. This is less than the PMTU_RAISE_TIMER 955 and used to decrease the PLPMTU (e.g., when a black hole is 956 encountered). Confirmation needs to be frequent enough when data 957 is flowing that the sending PL does not black hole extensive 958 amounts of traffic. Guidance on selection of the timer value are 959 provided in section 3.1.1 of the UDP Usage Guidelines [RFC8085]. 961 DPLPMTUD MAY inhibit sending probe packets when no application 962 data has been sent since the previous probe packet. A PL 963 preferring to use an up-to-data PMTU once user data is sent again, 964 can choose to continue PMTU discovery for each path. However, 965 this could result in sending additional packets. 967 An implementation could implement the various timers using a single 968 timer. 970 5.1.2. Constants 972 The following constants are defined: 974 MAX_PROBES: The MAX_PROBES is the maximum value of the PROBE_COUNT 975 counter (see Section 5.1.3). MAX_PROBES represents the limit for 976 the number of consecutive probe attempts of any size. Search 977 algorithms benefit from a MAX_PROBES value greater than 1 because 978 this can provide robustness to isolated packet loss. The default 979 value of MAX_PROBES is 3. 981 MIN_PMTU: The MIN_PMTU is the smallest allowed probe packet size. 982 For IPv6, this value is 1280 bytes, as specified in [RFC8200]. 983 For IPv4, the minimum value is 68 bytes. 985 Note: An IPv4 router is required to be able to forward a datagram 986 of 68 bytes without further fragmentation. This is the combined 987 size of an IPv4 header and the minimum fragment size of 8 bytes. 988 In addition, receivers are required to be able to reassemble 989 fragmented datagrams at least up to 576 bytes, as stated in 990 section 3.3.3 of [RFC1122]. 992 MAX_PMTU: The MAX_PMTU is the largest size of PLPMTU. This has to 993 be less than or equal to the minimum of the local MTU of the 994 outgoing interface and the destination PMTU for receiving. An 995 application, or PL, MAY choose a smaller MAX_PMTU when there is no 996 need to send packets larger than a specific size. 998 BASE_PMTU: The BASE_PMTU is a configured size expected to work for 999 most paths. The size is equal to or larger than the MIN_PMTU and 1000 smaller than the MAX_PMTU. In the case of IPv6, this value is 1001 1280 bytes [RFC8200]. When using IPv4, a size of 1200 bytes is 1002 RECOMMENDED. 1004 5.1.3. Variables 1006 This method utilizes a set of variables: 1008 PROBED_SIZE: The PROBED_SIZE is the size of the current probe 1009 packet. This is a tentative value for the PLPMTU, which is 1010 awaiting confirmation by an acknowledgment. 1012 PROBE_COUNT: The PROBE_COUNT is a count of the number of successive 1013 unsuccessful probe packets that have been sent. Each time a probe 1014 packet is acknowledged, the value is set to zero. (Some probe 1015 loss is expected while searching, therefore loss of a single probe 1016 is not an indication of a PMTU problem.) 1018 The figure below illustrates the relationship between the packet size 1019 constants and variables at a point of time when the DPLPMTUD 1020 algorithm performs path probing to increase the size of the PLPMTU. 1021 A probe packet has been sent of size PROBED_SIZE. Once this is 1022 acknowledged, the PLPMTU will raise to PROBED_SIZE allowing the 1023 DPLPMTUD algorithm to further increase PROBED_SIZE towards the actual 1024 PMTU. 1026 MIN_PMTU MAX_PMTU 1027 <--------------------------------------------------> 1028 | | | | 1029 v | | v 1030 BASE_PMTU | v Actual PMTU 1031 | PROBED_SIZE 1032 v 1033 PLPMTU 1035 Figure 3: Relationships between packet size constants and variables 1037 5.1.4. Overview of DPLPMTUD Phases 1039 This section provides a high-level informative view of the DPLPMTUD 1040 method, by describing the movement of the method through several 1041 phases of operation. More detail is available in the state machine 1042 Section 5.2. 1044 +------+ 1045 +------->| Base |----------------+ Connectivity 1046 | +------+ | or BASE_PMTU 1047 | | | confirmation failed 1048 | | v 1049 | | Connectivity +-------+ 1050 | | and BASE_PMTU | Error | 1051 | | confirmed +-------+ 1052 | | | Consistent 1053 | v | connectivity 1054 PLPMTU | +--------+ | and BASE_PMTU 1055 confirmation | | Search |<--------------+ confirmed 1056 failed | +--------+ 1057 | ^ | 1058 | | | 1059 | Raise | | Search 1060 | timer | | algorithm 1061 | expired | | completed 1062 | | | 1063 | | v 1064 | +-----------------+ 1065 +---| Search Complete | 1066 +-----------------+ 1068 Figure 4: DPLPMTUD Phases 1070 Base: The Base Phase confirms connectivity to the remote peer using 1071 packets of the BASE_PMTU. This phase is implicit for a 1072 connection-oriented PL (where it can be performed in a PL 1073 connection handshake). A connectionless PL sends an acknowledged 1074 probe packet to confirm that the remote peer is reachable. The 1075 sender also confirms that BASE_PMTU is supported across the 1076 network path. 1078 A PL that does not wish to support a path with a PLPMTU less than 1079 BASE_PMTU can simplify the phase into a single step by performing 1080 the connectivity checks with a probe of the BASE_PMTU size. 1082 Once confirmed, DPLPMTUD enters the Search Phase. If this phase 1083 fails to confirm, DPLPMTUD enters the Error Phase. 1085 Search: The Search Phase utilizes a search algorithm to send probe 1086 packets to seek to increase the PLPMTU. The algorithm concludes 1087 when it has found a suitable PLPMTU, by entering the Search 1088 Complete Phase. 1090 A PL could respond to PTB messages using the PTB to advance or 1091 terminate the search, see Section 4.6. 1093 Search Complete: The Search Complete Phase is entered when the 1094 PLPMTU is supported across the network path. A PL can use a 1095 CONFIRMATION_TIMER to periodically repeat a probe packet for the 1096 current PLPMTU size. If the sender is unable to confirm 1097 reachability (e.g., if the CONFIRMATION_TIMER expires) or the PL 1098 signals a lack of reachability, DPLPMTUD enters the Base phase. 1100 The PMTU_RAISE_TIMER is used to periodically resume the search 1101 phase to discover if the PLPMTU can be raised. Black Hole 1102 Detection causes the sender to enter the Base Phase. 1104 Error: The Error Phase is entered when there is conflicting or 1105 invalid PLPMTU information for the path (e.g. a failure to support 1106 the BASE_PMTU) that cause DPLPMTUD to be unable to progress and 1107 the PLPMTU is lowered. 1109 DPLPMTUD remains in the Error Phase until a consistent view of the 1110 path can be discovered and it has also been confirmed that the 1111 path supports the BASE_PMTU (or DPLPMTUD is suspended). 1113 An implementation that only reduces the PLPMTU to a suitable size 1114 would be sufficient to ensure reliable operation, but can be very 1115 inefficient when the actual PMTU changes or when the method (for 1116 whatever reason) makes a suboptimal choice for the PLPMTU. 1118 A full implementation of DPLPMTUD provides an algorithm enabling the 1119 DPLPMTUD sender to increase the PLPMTU following a change in the 1120 characteristics of the path, such as when a link is reconfigured with 1121 a larger MTU, or when there is a change in the set of links traversed 1122 by an end-to-end flow (e.g., after a routing or path fail-over 1123 decision). 1125 5.2. State Machine 1127 A state machine for DPLPMTUD is depicted in Figure 5. If multipath 1128 or multihoming is supported, a state machine is needed for each path. 1130 Note: Not all changes are shown to simplify the diagram. 1132 | | 1133 | Start | PL indicates loss 1134 | | of connectivity 1135 v v 1136 +---------------+ +---------------+ 1137 | DISABLED | | ERROR | 1138 +---------------+ PROBE_TIMER expiry: +---------------+ 1139 | PL indicates PROBE_COUNT = MAX_PROBES or ^ | 1140 | connectivity PTB: PTB_SIZE < BASE_PMTU | | 1141 +--------------------+ +---------------+ | 1142 | | | 1143 v | BASE_PMTU Probe | 1144 +---------------+ acked | 1145 | BASE |----------------------+ 1146 +---------------+ | 1147 ^ | ^ ^ | 1148 Black hole detected | | | | Black hole detected | 1149 +--------------------+ | | +--------------------+ | 1150 | +----+ | | 1151 | PROBE_TIMER expiry: | | 1152 | PROBE_COUNT < MAX_PROBES | | 1153 | | | 1154 | PMTU_RAISE_TIMER expiry | | 1155 | +-----------------------------------------+ | | 1156 | | | | | 1157 | | v | v 1158 +---------------+ +---------------+ 1159 |SEARCH_COMPLETE| | SEARCHING | 1160 +---------------+ +---------------+ 1161 | ^ ^ | | ^ 1162 | | | | | | 1163 | | +-----------------------------------------+ | | 1164 | | MAX_PMTU Probe acked or | | 1165 | | PROBE_TIMER expiry: PROBE_COUNT = MAX_PROBES or | | 1166 +----+ PTB: PTB_SIZE = PLPMTU +----+ 1167 CONFIRMATION_TIMER expiry: PROBE_TIMER expiry: 1168 PROBE_COUNT < MAX_PROBES or PROBE_COUNT < MAX_PROBES or 1169 PLPMTU Probe acked Probe acked or PTB: 1170 PLPMTU < PTB_SIZE < PROBED_SIZE 1172 Figure 5: State machine for Datagram PLPMTUD 1174 The following states are defined: 1176 DISABLED: The DISABLED state is the initial state before probing has 1177 started. It is also entered from any other state, when the PL 1178 indicates loss of connectivity. This state is left, once the PL 1179 indicates connectivity to the remote PL. 1181 BASE: The BASE state is used to confirm that the BASE_PMTU size is 1182 supported by the network path and is designed to allow an 1183 application to continue working when there are transient 1184 reductions in the actual PMTU. It also seeks to avoid long 1185 periods when a sender searching for a larger PLPMTU is unaware 1186 that packets are not being delivered due to a packet or ICMP Black 1187 Hole. 1189 On entry, the PROBED_SIZE is set to the BASE_PMTU size and the 1190 PROBE_COUNT is set to zero. 1192 Each time a probe packet is sent, the PROBE_TIMER is started. The 1193 state is exited when the probe packet is acknowledged, and the PL 1194 sender enters the SEARCHING state. 1196 The state is also left when the PROBE_COUNT reaches MAX_PROBES or 1197 a received PTB message is validated. This causes the PL sender to 1198 enter the ERROR state. 1200 SEARCHING: The SEARCHING state is the main probing state. This 1201 state is entered when probing for the BASE_PMTU was successful. 1203 Each time a probe packet is acknowledged, the PROBE_COUNT is set 1204 to zero, the PLPMTU is set to the PROBED_SIZE and then the 1205 PROBED_SIZE is increased using the search algorithm. 1207 When a probe packet is sent and not acknowledged within the period 1208 of the PROBE_TIMER, the PROBE_COUNT is incremented and a new probe 1209 packet is transmitted. 1211 The state is exited to enter SEARCH_COMPLETE when the PROBE_COUNT 1212 reaches MAX_PROBES, a validated PTB is received that corresponds 1213 to the last successfully probed size (PTB_SIZE = PLPMTU), or a 1214 probe of size MAX_PMTU is acknowledged (PLPMTU = MAX_PMTU). 1216 When a black hole is detected in the SEARCHING state, this causes 1217 the PL sender to enter the BASE state. 1219 SEARCH_COMPLETE: The SEARCH_COMPLETE state indicates a successful 1220 end to the SEARCHING state. DPLPMTUD remains in this state until 1221 either the PMTU_RAISE_TIMER expires or a black hole is detected. 1223 When DPLPMTUD uses an unacknowledged PL and is in the 1224 SEARCH_COMPLETE state, a CONFIRMATION_TIMER periodically resets 1225 the PROBE_COUNT and schedules a probe packet with the size of the 1226 PLPMTU. If MAX_PROBES successive PLPMTUD sized probes fail to be 1227 acknowledged the method enters the BASE state. When used with an 1228 acknowledged PL (e.g., SCTP), DPLPMTUD SHOULD NOT continue to 1229 generate PLPMTU probes in this state. 1231 ERROR: The ERROR state represents the case where either the network 1232 path is not known to support a PLPMTU of at least the BASE_PMTU 1233 size or when there is contradictory information about the network 1234 path that would otherwise result in excessive variation in the MPS 1235 signalled to the higher layer. The state implements a method to 1236 mitigate oscillation in the state-event engine. It signals a 1237 conservative value of the MPS to the higher layer by the PL. The 1238 state is exited when packet probes no longer detect the error. 1239 The PL sender then enters the SEARCHING state. 1241 Implementations are permitted to enable endpoint fragmentation if 1242 the DPLPMTUD is unable to validate MIN_PMTU within PROBE_COUNT 1243 probes. If DPLPMTUD is unable to validate MIN_PMTU the 1244 implementation will transition to the DISABLED state. 1246 Note: MIN_PMTU could be identical to BASE_PMTU, simplifying the 1247 actions in this state. 1249 5.3. Search to Increase the PLPMTU 1251 This section describes the algorithms used by DPLPMTUD to search for 1252 a larger PLPMTU. 1254 5.3.1. Probing for a larger PLPMTU 1256 Implementations use a search algorithm across the search range to 1257 determine whether a larger PLPMTU can be supported across a network 1258 path. 1260 The method discovers the search range by confirming the minimum 1261 PLPMTU and then using the probe method to select a PROBED_SIZE less 1262 than or equal to MAX_PMTU. MAX_PMTU is the minimum of the local MTU 1263 and EMTU_R (learned from the remote endpoint). The MAX_PMTU MAY be 1264 reduced by an application that sets a maximum to the size of 1265 datagrams it will send. 1267 The PROBE_COUNT is initialized to zero when the first probe with a 1268 size greater than or equal to PLPMTUD is sent. A timer is used to 1269 trigger the sending of probe packets of size PROBED_SIZE, larger than 1270 the PLPMTU. Each probe packet successfully sent to the remote peer 1271 is confirmed by acknowledgement at the PL, see Section 4.1. 1273 Each time a probe packet is sent to the destination, the PROBE_TIMER 1274 is started. The timer is canceled when the PL receives 1275 acknowledgment that the probe packet has been successfully sent 1276 across the path Section 4.1. This confirms that the PROBED_SIZE is 1277 supported, and the PROBED_SIZE value is then assigned to the PLPMTU. 1278 The search algorithm can continue to send subsequent probe packets of 1279 an increasing size. 1281 If the timer expires before a probe packet is acknowledged, the probe 1282 has failed to confirm the PROBED_SIZE. Each time the PROBE_TIMER 1283 expires, the PROBE_COUNT is incremented, the PROBE_TIMER is 1284 reinitialized, and a new probe of the same size or any other size 1285 (determined by the search algorithm) can be sent. The maximum number 1286 of consecutive failed probes is configured (MAX_PROBES). If the 1287 value of the PROBE_COUNT reaches MAX_PROBES, probing will stop, and 1288 the PL sender enters the SEARCH_COMPLETE state. 1290 5.3.2. Selection of Probe Sizes 1292 The search algorithm determines a minimum useful gain in PLPMTU. It 1293 would not be constructive for a PL sender to attempt to probe for all 1294 sizes. This would incur unnecessary load on the path. 1295 Implementations SHOULD select the set of probe packet sizes to 1296 maximize the gain in PLPMTU from each search step. 1298 Implementations could optimize the search procedure by selecting step 1299 sizes from a table of common PMTU sizes. When selecting the 1300 appropriate next size to search, an implementer ought to also 1301 consider that there can be common sizes of MPS that applications seek 1302 to use, and their could be common sizes of MTU used within the 1303 network. 1305 5.3.3. Resilience to Inconsistent Path Information 1307 A decision to increase the PLPMTU needs to be resilient to the 1308 possibility that information learned about the network path is 1309 inconsistent. A path is inconsistent, when, for example, probe 1310 packets are lost due to other reasons (i.e., not packet size) or due 1311 to frequent path changes. Frequent path changes could occur by 1312 unexpected "flapping" - where some packets from a flow pass along one 1313 path, but other packets follow a different path with different 1314 properties. 1316 A PL sender is able to detect inconsistency from the sequence of 1317 PLPMTU probes that are acknowledged or the sequence of PTB messages 1318 that it receives. When inconsistent path information is detected, a 1319 PL sender could use an alternate search mode that clamps the offered 1320 MPS to a smaller value for a period of time. This avoids unnecessary 1321 loss of packets. 1323 5.4. Robustness to Inconsistent Paths 1325 Some paths could be unable to sustain packets of the BASE_PMTU size. 1326 To be robust to these paths an implementation could implement the 1327 Error State. This allows fallback to a smaller than desired PLPMTU, 1328 rather than suffer connectivity failure. This could utilize methods 1329 such as endpoint IP fragmentation to enable the PL sender to 1330 communicate using packets smaller than the BASE_PMTU. 1332 6. Specification of Protocol-Specific Methods 1334 DPLPMTUD requires protocol-specific details to be specified for each 1335 PL that is used. 1337 The first subsection provides guidance on how to implement the 1338 DPLPMTUD method as a part of an application using UDP or UDP-Lite. 1339 The guidance also applies to other datagram services that do not 1340 include a specific transport protocol (such as a tunnel 1341 encapsulation). The following subsections describe how DPLPMTUD can 1342 be implemented as a part of the transport service, allowing 1343 applications using the service to benefit from discovery of the 1344 PLPMTU without themselves needing to implement this method when using 1345 SCTP and QUIC. 1347 6.1. Application support for DPLPMTUD with UDP or UDP-Lite 1349 The current specifications of UDP [RFC0768] and UDP-Lite [RFC3828] do 1350 not define a method in the RFC-series that supports PLPMTUD. In 1351 particular, the UDP transport does not provide the transport features 1352 needed to implement datagram PLPMTUD. 1354 The DPLPMTUD method can be implemented as a part of an application 1355 built directly or indirectly on UDP or UDP-Lite, but relies on 1356 higher-layer protocol features to implement the method [RFC8085]. 1358 Some primitives used by DPLPMTUD might not be available via the 1359 Datagram API (e.g., the ability to access the PLPMTU from the IP 1360 layer cache, or interpret received PTB messages). 1362 In addition, it is desirable that PMTU discovery is not performed by 1363 multiple protocol layers. An application SHOULD avoid using DPLPMTUD 1364 when the underlying transport system provides this capability. To 1365 use common method for managing the PLPMTU has benefits, both in the 1366 ability to share state between different processes and opportunities 1367 to coordinate probing. 1369 6.1.1. Application Request 1371 An application needs an application-layer protocol mechanism (such as 1372 a message acknowledgement method) that solicits a response from a 1373 destination endpoint. The method SHOULD allow the sender to check 1374 the value returned in the response to provide additional protection 1375 from off-path insertion of data [RFC8085], suitable methods include a 1376 parameter known only to the two endpoints, such as a session ID or 1377 initialized sequence number. 1379 6.1.2. Application Response 1381 An application needs an application-layer protocol mechanism to 1382 communicate the response from the destination endpoint. This 1383 response could indicate successful reception of the probe across the 1384 path, but could also indicate that some (or all packets) have failed 1385 to reach the destination. 1387 6.1.3. Sending Application Probe Packets 1389 A probe packet that could carry an application data block, but the 1390 successful transmission of this data is at risk when used for 1391 probing. Some applications might prefer to use a probe packet that 1392 does not carry an application data block to avoid disruption to data 1393 transfer. 1395 6.1.4. Initial Connectivity 1397 An application that does not have other higher-layer information 1398 confirming connectivity with the remote peer SHOULD implement a 1399 connectivity mechanism using acknowledged probe packets before 1400 entering the BASE state. 1402 6.1.5. Validating the Path 1404 An application that does not have other higher-layer information 1405 confirming correct delivery of datagrams SHOULD implement the 1406 CONFIRMATION_TIMER to periodically send probe packets while in the 1407 SEARCH_COMPLETE state. 1409 6.1.6. Handling of PTB Messages 1411 An application that is able and wishes to receive PTB messages MUST 1412 perform ICMP validation as specified in Section 5.2 of [RFC8085]. 1413 This requires that the application to check each received PTB 1414 messages to validate it is received in response to transmitted 1415 traffic and that the reported PTB_SIZE is less than the current 1416 probed size (see Section 4.6.2). A validated PTB message MAY be used 1417 as input to the DPLPMTUD algorithm, but MUST NOT be used directly to 1418 set the PLPMTU. 1420 6.2. DPLPMTUD for SCTP 1422 Section 10.2 of [RFC4821] specified a recommended PLPMTUD probing 1423 method for SCTP and Section 7.3 of [RFC4960] and recommended an 1424 endpoint apply the techniques in RFC4821 on a per-destination-address 1425 basis. The specification for DPLPMTUD continues the practice of 1426 using the PL to discover the PMTU, but updates, RFC4960 with a 1427 recommendation to use the method specified in this document: The 1428 RECOMMENDED method for generating probes is to add a chunk consisting 1429 only of padding to an SCTP message. The PAD chunk defined in 1430 [RFC4820] SHOULD be attached to a minimum length HEARTBEAT (HB) chunk 1431 to build a probe packet. This enables probing without affecting the 1432 transfer of user messages and without being limited by congestion 1433 control or flow control. This is preferred to using DATA chunks 1434 (with padding as required) as path probes. 1436 Section 6.9 of [RFC4960] describes dividing the user messages into 1437 data chunks sent by the PL when using SCTP. This notes that once an 1438 SCTP message has been sent, it cannot be re-segmented. [RFC4960] 1439 describes the method to retransmit data chunks when the MPS has 1440 reduced, and the use of IP fragmentation for this case. 1442 6.2.1. SCTP/IPv4 and SCTP/IPv6 1444 6.2.1.1. Initial Connectivity 1446 The base protocol is specified in [RFC4960]. This provides an 1447 acknowledged PL. A sender can therefore enter the BASE state as soon 1448 as connectivity has been confirmed. 1450 6.2.1.2. Sending SCTP Probe Packets 1452 Probe packets consist of an SCTP common header followed by a 1453 HEARTBEAT chunk and a PAD chunk. The PAD chunk is used to control 1454 the length of the probe packet. The HEARTBEAT chunk is used to 1455 trigger the sending of a HEARTBEAT ACK chunk. The reception of the 1456 HEARTBEAT ACK chunk acknowledges reception of a successful probe. A 1457 successful probe updates the association and path counters, but an 1458 unsuccessful probe is discounted (assumed to be a result of choosing 1459 too large a PLPMTU). 1461 The HEARTBEAT chunk carries a Heartbeat Information parameter which 1462 includes, besides the information suggested in [RFC4960], the probe 1463 size, which is the size of the complete datagram. The size of the 1464 PAD chunk is therefore computed by reducing the probing size by the 1465 IPv4 or IPv6 header size, the SCTP common header, the HEARTBEAT 1466 request and the PAD chunk header. The payload of the PAD chunk 1467 contains arbitrary data. 1469 Probing starts directly after the PL handshake, before data is sent. 1470 Assuming this behavior (i.e., the PMTU is smaller than or equal to 1471 the interface MTU), this process will take a few round trip time 1472 periods, dependent on the number of PMTU probes sent. The Heartbeat 1473 timer can be used to implement the PROBE_TIMER. 1475 6.2.1.3. Validating the Path with SCTP 1477 Since SCTP provides an acknowledged PL, a sender MUST NOT implement 1478 the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state. 1480 6.2.1.4. PTB Message Handling by SCTP 1482 Normal ICMP validation MUST be performed as specified in Appendix C 1483 of [RFC4960]. This requires that the first 8 bytes of the SCTP 1484 common header are quoted in the payload of the PTB message, which can 1485 be the case for ICMPv4 and is normally the case for ICMPv6. 1487 When a PTB message has been validated, the PTB_SIZE reported in the 1488 PTB message SHOULD be used with the DPLPMTUD algorithm, providing 1489 that the reported PTB_SIZE is less than the current probe size (see 1490 Section 4.6). 1492 6.2.2. DPLPMTUD for SCTP/UDP 1494 The UDP encapsulation of SCTP is specified in [RFC6951]. 1496 This specification updates the reference to RFC 4821 in section 5.6 1497 of RFC 6951 to refer to XXXTHISRFCXXX. RFC 6951 is updated by 1498 addition of the following sentence is to be added at the end of 1499 section 5.6: "The RECOMMENDED method for determining the MTU of the 1500 path is specified in XXXTHISRFCXXX". 1502 XXX RFC EDITOR - please replace XXXTHISRFCXXX when published XXX 1504 6.2.2.1. Initial Connectivity 1506 A sender can enter the BASE state as soon as SCTP connectivity has 1507 been confirmed. 1509 6.2.2.2. Sending SCTP/UDP Probe Packets 1511 Packet probing can be performed as specified in Section 6.2.1.2. The 1512 maximum payload is reduced by 8 bytes, which has to be considered 1513 when filling the PAD chunk. 1515 6.2.2.3. Validating the Path with SCTP/UDP 1517 Since SCTP provides an acknowledged PL, a sender MUST NOT implement 1518 the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state. 1520 6.2.2.4. Handling of PTB Messages by SCTP/UDP 1522 ICMP validation MUST be performed for PTB messages as specified in 1523 Appendix C of [RFC4960]. This requires that the first 8 bytes of the 1524 SCTP common header are contained in the PTB message, which can be the 1525 case for ICMPv4 (but note the UDP header also consumes a part of the 1526 quoted packet header) and is normally the case for ICMPv6. When the 1527 validation is completed, the PTB_SIZE indicated in the PTB message 1528 SHOULD be used with the DPLPMTUD providing that the reported PTB_SIZE 1529 is less than the current probe size. 1531 6.2.3. DPLPMTUD for SCTP/DTLS 1533 The Datagram Transport Layer Security (DTLS) encapsulation of SCTP is 1534 specified in [RFC8261] . This is used for data channels in WebRTC 1535 implementations. This specification updates the reference to RFC 1536 4821 in section 5 of RFC 8261 to refer to XXXTHISRFCXXX. 1538 XXX RFC EDITOR - please replace XXXTHISRFCXXX when published XXX 1540 6.2.3.1. Initial Connectivity 1542 A sender can enter the BASE state as soon as SCTP connectivity has 1543 been confirmed. 1545 6.2.3.2. Sending SCTP/DTLS Probe Packets 1547 Packet probing can be done, as specified in Section 6.2.1.2. 1549 6.2.3.3. Validating the Path with SCTP/DTLS 1551 Since SCTP provides an acknowledged PL, a sender MUST NOT implement 1552 the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state. 1554 6.2.3.4. Handling of PTB Messages by SCTP/DTLS 1556 [RFC4960] does not specify a way to validate SCTP/DTLS ICMP message 1557 payload. This can prevent processing of PTB messages at the PL. 1559 6.3. DPLPMTUD for QUIC 1561 QUIC [I-D.ietf-quic-transport] is a UDP-based transport that provides 1562 reception feedback. The UDP payload includes the QUIC packet header, 1563 protected payload, and any authentication fields. QUIC depends on a 1564 PMTU of at least 1280 bytes. 1566 Section 14 of [I-D.ietf-quic-transport] describes the path 1567 considerations when sending QUIC packets. It recommends the use of 1568 PADDING frames to build the probe packet. Pure probe-only packets 1569 are constructed with PADDING frames and PING frames to create a 1570 padding only packet that will elicit an acknowledgement. Such 1571 padding only packets enable probing without affecting the transfer of 1572 other QUIC frames. 1574 The recommendation for QUIC endpoints implementing DPLPMTUD is that a 1575 MPS is maintained for each combination of local and remote IP 1576 addresses [I-D.ietf-quic-transport]. If a QUIC endpoint determines 1577 that the PMTU between any pair of local and remote IP addresses has 1578 fallen below an acceptable MPS, it immediately ceases to send QUIC 1579 packets on the affected path. This could result in termination of 1580 the connection if an alternative path cannot be found 1581 [I-D.ietf-quic-transport]. 1583 6.3.1. Initial Connectivity 1585 The base protocol is specified in [I-D.ietf-quic-transport]. This 1586 provides an acknowledged PL. A sender can therefore enter the BASE 1587 state as soon as connectivity has been confirmed. 1589 6.3.2. Sending QUIC Probe Packets 1591 A probe packet consists of a QUIC Header and a payload containing 1592 PADDING Frames and a PING Frame. PADDING Frames are a single octet 1593 (0x00) and several of these can be used to create a probe packet of 1594 size PROBED_SIZE. QUIC provides an acknowledged PL, a sender can 1595 therefore enter the BASE state as soon as connectivity has been 1596 confirmed. 1598 The current specification of QUIC sets the following: 1600 * BASE_PMTU: 1280. A QUIC sender pads initial packets to confirm 1601 the path can support packets of the required size. 1603 * MIN_PMTU: 1280 bytes. A QUIC sender that determines the PLPMTU 1604 has fallen below 1280 bytes MUST immediately stop sending on the 1605 affected path. 1607 6.3.3. Validating the Path with QUIC 1609 QUIC provides an acknowledged PL. A sender therefore MUST NOT 1610 implement the CONFIRMATION_TIMER while in the SEARCH_COMPLETE state. 1612 6.3.4. Handling of PTB Messages by QUIC 1614 QUIC validates ICMP PTB messages. In addition to UDP Port 1615 validation, QUIC can validate an ICMP message by using other PL 1616 information (e.g., validation of connection IDs in the quoted packet 1617 of any received ICMP message). 1619 7. Acknowledgements 1621 This work was partially funded by the European Union's Horizon 2020 1622 research and innovation programme under grant agreement No. 644334 1623 (NEAT). The views expressed are solely those of the author(s). 1625 Thanks to all that have commented or contributed, the TSVWG and QUIC 1626 working groups, and Mathew Calder and Julius Flohr for providing 1627 early implementations. 1629 8. IANA Considerations 1631 This memo includes no request to IANA. 1633 If there are no requirements for IANA, the section will be removed 1634 during conversion into an RFC by the RFC Editor. 1636 9. Security Considerations 1638 The security considerations for the use of UDP and SCTP are provided 1639 in the referenced RFCs. 1641 To avoid excessive load, the interval between individual probe 1642 packets MUST be at least one RTT, and the interval between rounds of 1643 probing is determined by the PMTU_RAISE_TIMER. 1645 A PL sender needs to ensure that the method used to confirm reception 1646 of probe packets protects from off-path attackers injecting packets 1647 into the path. This protection if provided in IETF-defined protocols 1648 (e.g., TCP, SCTP) using a randomly-initialized sequence number. A 1649 description of one way to do this when using UDP is provided in 1650 section 5.1 of [RFC8085]). 1652 There are cases where ICMP Packet Too Big (PTB) messages are not 1653 delivered due to policy, configuration or equipment design (see 1654 Section 1.1), this method therefore does not rely upon PTB messages 1655 being received, but is able to utilize these when they are received 1656 by the sender. PTB messages could potentially be used to cause a 1657 node to inappropriately reduce the PLPMTU. A node supporting 1658 DPLPMTUD MUST therefore appropriately validate the payload of PTB 1659 messages to ensure these are received in response to transmitted 1660 traffic (i.e., a reported error condition that corresponds to a 1661 datagram actually sent by the path layer, see Section 4.6.1). 1663 An on-path attacker, able to create a PTB message could forge PTB 1664 messages that include a valid quoted IP packet. Such an attack could 1665 be used to drive down the PLPMTU. There are two ways this method can 1666 be mitigated against such attacks: First, by ensuring that a PL 1667 sender never reduces the PLPMTU below the base size, solely in 1668 response to receiving a PTB message. This is achieved by first 1669 entering the BASE state when such a message is received. Second, the 1670 design does not require processing of PTB messages, a PL sender could 1671 therefore suspend processing of PTB messages (e.g., in a robustness 1672 mode after detecting that subsequent probes actually confirm that a 1673 size larger than the PTB_SIZE is supported by a path). 1675 The successful processing of an ICMP message can trigger a probe when 1676 the reported PTB size is valid, but this does not directly update the 1677 PLPMTU for the path. This prevents a message attempting to black 1678 hole data by indicating a size larger than supported by the path. 1680 Parallel forwarding paths SHOULD be considered. Section 5.4 1681 identifies the need for robustness in the method because the path 1682 information might be inconsistent. 1684 A node performing DPLPMTUD could experience conflicting information 1685 about the size of supported probe packets. This could occur when 1686 there are multiple paths are concurrently in use and these exhibit a 1687 different PMTU. If not considered, this could result in packets not 1688 being delivered (black holed) when the PLPMTU is larger than the 1689 smallest actual PMTU. 1691 DPLPMTUD methods can introduce padding data to inflate the length of 1692 the datagram to the total size required for a probe packet. The 1693 total size of a probe packet includes all headers and padding added 1694 to the payload data being sent (e.g., including security-related 1695 fields such as an AEAD tag and TLS record layer padding). The value 1696 of the padding data does not influence the DPLPMTUD search algorithm, 1697 and therefore needs to be set consistent with the policy of the PL. 1699 If a PL can make use of cryptographic confidentiality or data- 1700 integrity mechanisms, then adding anything (e.g., padding) for 1701 DPLPMTUD that is not protected by those cryptographic mechanisms is 1702 an anti-pattern to be avoided. 1704 10. References 1706 10.1. Normative References 1708 [I-D.ietf-quic-transport] 1709 Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed 1710 and Secure Transport", Work in Progress, Internet-Draft, 1711 draft-ietf-quic-transport-27, 21 February 2020, 1712 . 1715 [RFC0768] Postel, J., "User Datagram Protocol", STD 6, RFC 768, 1716 DOI 10.17487/RFC0768, August 1980, 1717 . 1719 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, 1720 DOI 10.17487/RFC0791, September 1981, 1721 . 1723 [RFC1191] Mogul, J.C. and S.E. Deering, "Path MTU discovery", 1724 RFC 1191, DOI 10.17487/RFC1191, November 1990, 1725 . 1727 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1728 Requirement Levels", BCP 14, RFC 2119, 1729 DOI 10.17487/RFC2119, March 1997, 1730 . 1732 [RFC3828] Larzon, L-A., Degermark, M., Pink, S., Jonsson, L-E., Ed., 1733 and G. Fairhurst, Ed., "The Lightweight User Datagram 1734 Protocol (UDP-Lite)", RFC 3828, DOI 10.17487/RFC3828, July 1735 2004, . 1737 [RFC4820] Tuexen, M., Stewart, R., and P. Lei, "Padding Chunk and 1738 Parameter for the Stream Control Transmission Protocol 1739 (SCTP)", RFC 4820, DOI 10.17487/RFC4820, March 2007, 1740 . 1742 [RFC4960] Stewart, R., Ed., "Stream Control Transmission Protocol", 1743 RFC 4960, DOI 10.17487/RFC4960, September 2007, 1744 . 1746 [RFC6951] Tuexen, M. and R. Stewart, "UDP Encapsulation of Stream 1747 Control Transmission Protocol (SCTP) Packets for End-Host 1748 to End-Host Communication", RFC 6951, 1749 DOI 10.17487/RFC6951, May 2013, 1750 . 1752 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 1753 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1754 March 2017, . 1756 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 1757 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 1758 May 2017, . 1760 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1761 (IPv6) Specification", STD 86, RFC 8200, 1762 DOI 10.17487/RFC8200, July 2017, 1763 . 1765 [RFC8201] McCann, J., Deering, S., Mogul, J., and R. Hinden, Ed., 1766 "Path MTU Discovery for IP version 6", STD 87, RFC 8201, 1767 DOI 10.17487/RFC8201, July 2017, 1768 . 1770 [RFC8261] Tuexen, M., Stewart, R., Jesup, R., and S. Loreto, 1771 "Datagram Transport Layer Security (DTLS) Encapsulation of 1772 SCTP Packets", RFC 8261, DOI 10.17487/RFC8261, November 1773 2017, . 1775 10.2. Informative References 1777 [I-D.ietf-intarea-frag-fragile] 1778 Bonica, R., Baker, F., Huston, G., Hinden, R., Troan, O., 1779 and F. Gont, "IP Fragmentation Considered Fragile", Work 1780 in Progress, Internet-Draft, draft-ietf-intarea-frag- 1781 fragile-17, 30 September 2019, . 1784 [I-D.ietf-intarea-tunnels] 1785 Touch, J. and M. Townsley, "IP Tunnels in the Internet 1786 Architecture", Work in Progress, Internet-Draft, draft- 1787 ietf-intarea-tunnels-10, 12 September 2019, 1788 . 1791 [RFC0792] Postel, J., "Internet Control Message Protocol", STD 5, 1792 RFC 792, DOI 10.17487/RFC0792, September 1981, 1793 . 1795 [RFC1122] Braden, R., Ed., "Requirements for Internet Hosts - 1796 Communication Layers", STD 3, RFC 1122, 1797 DOI 10.17487/RFC1122, October 1989, 1798 . 1800 [RFC1812] Baker, F., Ed., "Requirements for IP Version 4 Routers", 1801 RFC 1812, DOI 10.17487/RFC1812, June 1995, 1802 . 1804 [RFC2923] Lahey, K., "TCP Problems with Path MTU Discovery", 1805 RFC 2923, DOI 10.17487/RFC2923, September 2000, 1806 . 1808 [RFC4340] Kohler, E., Handley, M., and S. Floyd, "Datagram 1809 Congestion Control Protocol (DCCP)", RFC 4340, 1810 DOI 10.17487/RFC4340, March 2006, 1811 . 1813 [RFC4443] Conta, A., Deering, S., and M. Gupta, Ed., "Internet 1814 Control Message Protocol (ICMPv6) for the Internet 1815 Protocol Version 6 (IPv6) Specification", STD 89, 1816 RFC 4443, DOI 10.17487/RFC4443, March 2006, 1817 . 1819 [RFC4821] Mathis, M. and J. Heffner, "Packetization Layer Path MTU 1820 Discovery", RFC 4821, DOI 10.17487/RFC4821, March 2007, 1821 . 1823 [RFC4890] Davies, E. and J. Mohacsi, "Recommendations for Filtering 1824 ICMPv6 Messages in Firewalls", RFC 4890, 1825 DOI 10.17487/RFC4890, May 2007, 1826 . 1828 [RFC5508] Srisuresh, P., Ford, B., Sivakumar, S., and S. Guha, "NAT 1829 Behavioral Requirements for ICMP", BCP 148, RFC 5508, 1830 DOI 10.17487/RFC5508, April 2009, 1831 . 1833 Appendix A. Revision Notes 1835 Note to RFC-Editor: please remove this entire section prior to 1836 publication. 1838 Individual draft -00: 1840 * Comments and corrections are welcome directly to the authors or 1841 via the IETF TSVWG working group mailing list. 1843 * This update is proposed for WG comments. 1845 Individual draft -01: 1847 * Contains the first representation of the algorithm, showing the 1848 states and timers 1850 * This update is proposed for WG comments. 1852 Individual draft -02: 1854 * Contains updated representation of the algorithm, and textual 1855 corrections. 1857 * The text describing when to set the effective PMTU has not yet 1858 been validated by the authors 1860 * To determine security to off-path-attacks: We need to decide 1861 whether a received PTB message SHOULD/MUST be validated? The text 1862 on how to handle a PTB message indicating a link MTU larger than 1863 the probe has yet not been validated by the authors 1865 * No text currently describes how to handle inconsistent results 1866 from arbitrary re-routing along different parallel paths 1868 * This update is proposed for WG comments. 1870 Working Group draft -00: 1872 * This draft follows a successful adoption call for TSVWG 1874 * There is still work to complete, please comment on this draft. 1876 Working Group draft -01: 1878 * This draft includes improved introduction. 1880 * The draft is updated to require ICMP validation prior to accepting 1881 PTB messages - this to be confirmed by WG 1883 * Section added to discuss Selection of Probe Size - methods to be 1884 evaluated and recommendations to be considered 1886 * Section added to align with work proposed in the QUIC WG. 1888 Working Group draft -02: 1890 * The draft was updated based on feedback from the WG, and a 1891 detailed review by Magnus Westerlund. 1893 * The document updates RFC 4821. 1895 * Requirements list updated. 1897 * Added more explicit discussion of a simpler black-hole detection 1898 mode. 1900 * This draft includes reorganisation of the section on IETF 1901 protocols. 1903 * Added more discussion of implementation within an application. 1905 * Added text on flapping paths. 1907 * Replaced 'effective MTU' with new term PLPMTU. 1909 Working Group draft -03: 1911 * Updated figures 1913 * Added more discussion on blackhole detection 1915 * Added figure describing just blackhole detection 1917 * Added figure relating MPS sizes 1919 Working Group draft -04: 1921 * Described phases and named these consistently. 1923 * Corrected transition from confirmation directly to the search 1924 phase (Base has been checked). 1926 * Redrawn state diagrams. 1928 * Renamed BASE_MTU to BASE_PMTU (because it is a base for the PMTU). 1930 * Clarified Error state. 1932 * Clarified suspending DPLPMTUD. 1934 * Verified normative text in requirements section. 1936 * Removed duplicate text. 1938 * Changed all text to refer to /packet probe/probe packet/ 1939 /validation/verification/ added term /Probe Confirmation/ and 1940 clarified BlackHole detection. 1942 Working Group draft -05: 1944 * Updated security considerations. 1946 * Feedback after speaking with Joe Touch helped improve UDP-Options 1947 description. 1949 Working Group draft -06: 1951 * Updated description of ICMP issues in section 1.1 1953 * Update to description of QUIC. 1955 Working group draft -07: 1957 * Moved description of the PTB processing method from the PTB 1958 requirements section. 1960 * Clarified what is performed in the PTB validation check. 1962 * Updated security consideration to explain PTB security without 1963 needing to read the rest of the document. 1965 * Reformatted state machine diagram 1967 Working group draft -08: 1969 * Moved to rfcxml v3+ 1971 * Rendered diagrams to svg in html version. 1973 * Removed Appendix A. Event-driven state changes. 1975 * Removed section on DPLPMTUD with UDP Options. 1977 * Shortened the description of phases. 1979 Working group draft -09: 1981 * Remove final mention of UDP Options 1983 * Add Initial Connectivity sections to each PL 1985 * Add to disable outgoing pmtu enforcement of packets 1986 Working group draft -10: 1988 * Address comments from Lars Eggert 1990 * Reinforce that PROBE_COUNT is successive attempts to probe for any 1991 size 1993 * Redefine MAx_PROBES to 3 1995 * Address PTB_SIZE of 0 or less that MIN_PMTU 1997 Working group draft -11: 1999 * Restore a sentence removed in previous rev 2001 * De-acronymise QUIC 2003 * Address some nits 2005 Working group draft -12: 2007 * Add TSVWG, QUIC and implementers to acknowledgements 2009 * Shorten a diagram line. 2011 * Address nits from Julius and Wes. 2013 * Be clearer when talking about IP layer caches 2015 Authors' Addresses 2017 Godred Fairhurst 2018 University of Aberdeen 2019 School of Engineering 2020 Fraser Noble Building 2021 Aberdeen 2022 AB24 3UE 2023 United Kingdom 2025 Email: gorry@erg.abdn.ac.uk 2027 Tom Jones 2028 University of Aberdeen 2029 School of Engineering 2030 Fraser Noble Building 2031 Aberdeen 2032 AB24 3UE 2033 United Kingdom 2035 Email: tom@erg.abdn.ac.uk 2037 Michael Tuexen 2038 Muenster University of Applied Sciences 2039 Stegerwaldstrasse 39 2040 48565 Steinfurt 2041 Germany 2043 Email: tuexen@fh-muenster.de 2045 Irene Ruengeler 2046 Muenster University of Applied Sciences 2047 Stegerwaldstrasse 39 2048 48565 Steinfurt 2049 Germany 2051 Email: i.ruengeler@fh-muenster.de 2053 Timo Voelker 2054 Muenster University of Applied Sciences 2055 Stegerwaldstrasse 39 2056 48565 Steinfurt 2057 Germany 2059 Email: timo.voelker@fh-muenster.de