idnits 2.17.1 draft-templin-intarea-seal-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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (July 9, 2010) is 5038 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references to lower-maturity documents in RFCs) ** Obsolete normative reference: RFC 2460 (Obsoleted by RFC 8200) == Outdated reference: A later version (-07) exists of draft-ietf-intarea-ipv4-id-update-00 == Outdated reference: A later version (-04) exists of draft-ietf-v6ops-tunnel-security-concerns-02 == Outdated reference: A later version (-40) exists of draft-templin-intarea-vet-15 == Outdated reference: A later version (-17) exists of draft-templin-iron-08 -- Obsolete informational reference (is this intentional?): RFC 1063 (Obsoleted by RFC 1191) -- Obsolete informational reference (is this intentional?): RFC 1981 (Obsoleted by RFC 8201) Summary: 1 error (**), 0 flaws (~~), 5 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group F. Templin, Ed. 3 Internet-Draft Boeing Research & Technology 4 Intended status: Standards Track July 9, 2010 5 Expires: January 10, 2011 7 The Subnetwork Encapsulation and Adaptation Layer (SEAL) 8 draft-templin-intarea-seal-16.txt 10 Abstract 12 For the purpose of this document, a subnetwork is defined as a 13 virtual topology configured over a connected IP network routing 14 region and bounded by encapsulating border nodes. These virtual 15 topologies may span multiple IP and/or sub-IP layer forwarding hops, 16 and can introduce failure modes due to packet duplication and/or 17 links with diverse Maximum Transmission Units (MTUs). This document 18 specifies a Subnetwork Encapsulation and Adaptation Layer (SEAL) that 19 accommodates such virtual topologies over diverse underlying link 20 technologies. 22 Status of this Memo 24 This Internet-Draft is submitted in full conformance with the 25 provisions of BCP 78 and BCP 79. 27 Internet-Drafts are working documents of the Internet Engineering 28 Task Force (IETF). Note that other groups may also distribute 29 working documents as Internet-Drafts. The list of current Internet- 30 Drafts is at http://datatracker.ietf.org/drafts/current/. 32 Internet-Drafts are draft documents valid for a maximum of six months 33 and may be updated, replaced, or obsoleted by other documents at any 34 time. It is inappropriate to use Internet-Drafts as reference 35 material or to cite them other than as "work in progress." 37 This Internet-Draft will expire on January 10, 2011. 39 Copyright Notice 41 Copyright (c) 2010 IETF Trust and the persons identified as the 42 document authors. All rights reserved. 44 This document is subject to BCP 78 and the IETF Trust's Legal 45 Provisions Relating to IETF Documents 46 (http://trustee.ietf.org/license-info) in effect on the date of 47 publication of this document. Please review these documents 48 carefully, as they describe your rights and restrictions with respect 49 to this document. Code Components extracted from this document must 50 include Simplified BSD License text as described in Section 4.e of 51 the Trust Legal Provisions and are provided without warranty as 52 described in the Simplified BSD License. 54 Table of Contents 56 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 57 1.1. Motivation . . . . . . . . . . . . . . . . . . . . . . . . 4 58 1.2. Approach . . . . . . . . . . . . . . . . . . . . . . . . . 6 59 2. Terminology and Requirements . . . . . . . . . . . . . . . . . 7 60 3. Applicability Statement . . . . . . . . . . . . . . . . . . . 9 61 4. SEAL Protocol Specification . . . . . . . . . . . . . . . . . 10 62 4.1. Model of Operation . . . . . . . . . . . . . . . . . . . . 10 63 4.2. SEAL Header Format . . . . . . . . . . . . . . . . . . . . 13 64 4.3. ITE Specification . . . . . . . . . . . . . . . . . . . . 14 65 4.3.1. Tunnel Interface MTU . . . . . . . . . . . . . . . . . 14 66 4.3.2. Tunnel Interface Soft State . . . . . . . . . . . . . 15 67 4.3.3. Admitting Packets into the Tunnel . . . . . . . . . . 16 68 4.3.4. Mid-Layer Encapsulation . . . . . . . . . . . . . . . 17 69 4.3.5. SEAL Segmentation . . . . . . . . . . . . . . . . . . 17 70 4.3.6. Outer Encapsulation . . . . . . . . . . . . . . . . . 17 71 4.3.7. Probing Strategy . . . . . . . . . . . . . . . . . . . 18 72 4.3.8. Identification . . . . . . . . . . . . . . . . . . . . 18 73 4.3.9. Sending SEAL Protocol Packets . . . . . . . . . . . . 19 74 4.3.10. Processing Raw ICMP Messages . . . . . . . . . . . . . 19 75 4.4. ETE Specification . . . . . . . . . . . . . . . . . . . . 19 76 4.4.1. Reassembly Buffer Requirements . . . . . . . . . . . . 19 77 4.4.2. IP-Layer Reassembly . . . . . . . . . . . . . . . . . 20 78 4.4.3. SEAL-Layer Reassembly . . . . . . . . . . . . . . . . 21 79 4.4.4. Decapsulation and Delivery to Upper Layers . . . . . . 22 80 4.5. The SEAL Control Message Protocol (SCMP) . . . . . . . . . 22 81 4.5.1. Generating SCMP Messages . . . . . . . . . . . . . . . 22 82 4.5.2. Processing SCMP Messages . . . . . . . . . . . . . . . 25 83 4.6. Tunnel Endpoint Synchronization . . . . . . . . . . . . . 27 84 5. Link Requirements . . . . . . . . . . . . . . . . . . . . . . 29 85 6. End System Requirements . . . . . . . . . . . . . . . . . . . 29 86 7. Router Requirements . . . . . . . . . . . . . . . . . . . . . 30 87 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 30 88 9. Security Considerations . . . . . . . . . . . . . . . . . . . 30 89 10. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . 31 90 11. SEAL Advantages over Classical Methods . . . . . . . . . . . . 31 91 12. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 32 92 13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 33 93 13.1. Normative References . . . . . . . . . . . . . . . . . . . 33 94 13.2. Informative References . . . . . . . . . . . . . . . . . . 33 95 Appendix A. Reliability . . . . . . . . . . . . . . . . . . . . . 36 96 Appendix B. Integrity . . . . . . . . . . . . . . . . . . . . . . 36 97 Appendix C. Transport Mode . . . . . . . . . . . . . . . . . . . 37 98 Appendix D. Historic Evolution of PMTUD . . . . . . . . . . . . . 38 99 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 39 101 1. Introduction 103 As Internet technology and communication has grown and matured, many 104 techniques have developed that use virtual topologies (including 105 tunnels of one form or another) over an actual network that supports 106 the Internet Protocol (IP) [RFC0791][RFC2460]. Those virtual 107 topologies have elements that appear as one hop in the virtual 108 topology, but are actually multiple IP or sub-IP layer hops. These 109 multiple hops often have quite diverse properties that are often not 110 even visible to the endpoints of the virtual hop. This introduces 111 failure modes that are not dealt with well in current approaches. 113 The use of IP encapsulation has long been considered as the means for 114 creating such virtual topologies. However, the insertion of an outer 115 IP header reduces the effective path MTU visible to the inner network 116 layer. When IPv4 is used, this reduced MTU can be accommodated 117 through the use of IPv4 fragmentation, but unmitigated in-the-network 118 fragmentation has been found to be harmful through operational 119 experience and studies conducted over the course of many years 120 [FRAG][FOLK][RFC4963]. Additionally, classical path MTU discovery 121 [RFC1191] has known operational issues that are exacerbated by in- 122 the-network tunnels [RFC2923][RFC4459]. The following subsections 123 present further details on the motivation and approach for addressing 124 these issues. 126 1.1. Motivation 128 Before discussing the approach, it is necessary to first understand 129 the problems. In both the Internet and private-use networks today, 130 IPv4 is ubiquitously deployed as the Layer 3 protocol. The two 131 primary functions of IPv4 are to provide for 1) addressing, and 2) a 132 fragmentation and reassembly capability used to accommodate links 133 with diverse MTUs. While it is well known that the IPv4 address 134 space is rapidly becoming depleted, there is a lesser-known but 135 growing consensus that other IPv4 protocol limitations have already 136 or may soon become problematic. 138 First, the IPv4 header Identification field is only 16 bits in 139 length, meaning that at most 2^16 unique packets with the same 140 (source, destination, protocol)-tuple may be active in the Internet 141 at a given time [I-D.ietf-intarea-ipv4-id-update]. Due to the 142 escalating deployment of high-speed links (e.g., 1Gbps Ethernet), 143 however, this number may soon become too small by several orders of 144 magnitude for high data rate packet sources such as tunnel endpoints 145 [RFC4963]. Furthermore, there are many well-known limitations 146 pertaining to IPv4 fragmentation and reassembly - even to the point 147 that it has been deemed "harmful" in both classic and modern-day 148 studies (cited above). In particular, IPv4 fragmentation raises 149 issues ranging from minor annoyances (e.g., in-the-network router 150 fragmentation) to the potential for major integrity issues (e.g., 151 mis-association of the fragments of multiple IP packets during 152 reassembly [RFC4963]). 154 As a result of these perceived limitations, a fragmentation-avoiding 155 technique for discovering the MTU of the forward path from a source 156 to a destination node was devised through the deliberations of the 157 Path MTU Discovery Working Group (PMTUDWG) during the late 1980's 158 through early 1990's (see Appendix D). In this method, the source 159 node provides explicit instructions to routers in the path to discard 160 the packet and return an ICMP error message if an MTU restriction is 161 encountered. However, this approach has several serious shortcomings 162 that lead to an overall "brittleness" [RFC2923]. 164 In particular, site border routers in the Internet are being 165 configured more and more to discard ICMP error messages coming from 166 the outside world. This is due in large part to the fact that 167 malicious spoofing of error messages in the Internet is made simple 168 since there is no way to authenticate the source of the messages 169 [I-D.ietf-tcpm-icmp-attacks]. Furthermore, when a source node that 170 requires ICMP error message feedback when a packet is dropped due to 171 an MTU restriction does not receive the messages, a path MTU-related 172 black hole occurs. This means that the source will continue to send 173 packets that are too large and never receive an indication from the 174 network that they are being discarded. This behavior has been 175 confirmed through documented studies showing clear evidence of path 176 MTU discovery failures in the Internet today [TBIT][WAND]. 178 The issues with both IPv4 fragmentation and this "classical" method 179 of path MTU discovery are exacerbated further when IP tunneling is 180 used [RFC4459]. For example, ingress tunnel endpoints (ITEs) may be 181 required to forward encapsulated packets into the subnetwork on 182 behalf of hundreds, thousands, or even more original sources in the 183 end site. If the ITE allows IPv4 fragmentation on the encapsulated 184 packets, persistent fragmentation could lead to undetected data 185 corruption due to Identification field wrapping. If the ITE instead 186 uses classical IPv4 path MTU discovery, it may be inconvenienced by 187 excessive ICMP error messages coming from the subnetwork that may be 188 either suspect or contain insufficient information for translation 189 into error messages to be returned to the original sources. 191 Although recent works have led to the development of a robust end-to- 192 end MTU determination scheme [RFC4821], this approach requires 193 tunnels to present a consistent MTU the same as for ordinary links on 194 the end-to-end path. Moreover, in current practice existing 195 tunneling protocols mask the MTU issues by selecting a "lowest common 196 denominator" MTU that may be much smaller than necessary for most 197 paths and difficult to change at a later date. Due to these many 198 consideration, a new approach to accommodate tunnels over links with 199 diverse MTUs is necessary. 201 1.2. Approach 203 For the purpose of this document, a subnetwork is defined as a 204 virtual topology configured over a connected network routing region 205 and bounded by encapsulating border nodes. Examples include the 206 global Internet interdomain routing core, Mobile Ad hoc Networks 207 (MANETs) and enterprise networks. Subnetwork border nodes forward 208 unicast and multicast packets over the virtual topology across 209 multiple IP and/or sub-IP layer forwarding hops that may introduce 210 packet duplication and/or traverse links with diverse Maximum 211 Transmission Units (MTUs). 213 This document introduces a Subnetwork Encapsulation and Adaptation 214 Layer (SEAL) for tunneling network layer protocols (e.g., IP, OSI, 215 etc.) over IP subnetworks that connect Ingress and Egress Tunnel 216 Endpoints (ITEs/ETEs) of border nodes. It provides a modular 217 specification designed to be tailored to specific associated 218 tunneling protocols. A transport-mode of operation is also possible, 219 and described in Appendix C. SEAL accommodates links with diverse 220 MTUs, protects against off-path denial-of-service attacks, and 221 supports efficient duplicate packet detection through the use of a 222 minimal mid-layer encapsulation. 224 SEAL specifically treats tunnels that traverse the subnetwork as 225 ordinary links that must support network layer services. As for any 226 link, tunnels that use SEAL must provide suitable networking services 227 including best-effort datagram delivery, integrity and consistent 228 handling of packets of various sizes. As for any link whose media 229 cannot provide suitable services natively, tunnels that use SEAL 230 employ link-level adaptation functions to meet the legitimate 231 expectations of the network layer service. As this is essentially a 232 link level adaptation, SEAL is therefore permitted to alter packets 233 within the subnetwork as long as it restores them to their original 234 form when they exit the subnetwork. The mechanisms described within 235 this document are designed precisely for this purpose. 237 SEAL encapsulation introduces an extended Identification field for 238 packet identification and a mid-layer segmentation and reassembly 239 capability that allows simplified cutting and pasting of packets. 240 Moreover, SEAL senses in-the-network fragmentation as a "noise" 241 indication that packet sizing parameters are "out of tune" with 242 respect to the network path. As a result, SEAL can naturally tune 243 its packet sizing parameters to eliminate the in-the-network 244 fragmentation. This approach is in contrast to existing tunneling 245 protocol practices which seek to avoid MTU issues by selecting a 246 "lowest common denominator" MTU that may be overly conservative for 247 many tunnels and difficult to change even when larger MTUs become 248 available. 250 The following sections provide the SEAL normative specifications, 251 while the appendices present non-normative additional considerations. 253 2. Terminology and Requirements 255 The following terms are defined within the scope of this document: 257 subnetwork 258 a virtual topology configured over a connected network routing 259 region and bounded by encapsulating border nodes. 261 Ingress Tunnel Endpoint 262 a virtual interface over which an encapsulating border node (host 263 or router) sends encapsulated packets into the subnetwork. 265 Egress Tunnel Endpoint 266 a virtual interface over which an encapsulating border node (host 267 or router) receives encapsulated packets from the subnetwork. 269 inner packet 270 an unencapsulated network layer protocol packet (e.g., IPv6 271 [RFC2460], IPv4 [RFC0791], OSI/CLNP [RFC1070], etc.) before any 272 mid-layer or outer encapsulations are added. Internet protocol 273 numbers that identify inner packets are found in the IANA Internet 274 Protocol registry [RFC3232]. 276 mid-layer packet 277 a packet resulting from adding mid-layer encapsulating headers to 278 an inner packet. 280 outer IP packet 281 a packet resulting from adding an outer IP header to a mid-layer 282 packet. 284 packet-in-error 285 the leading portion of an invoking data packet encapsulated in the 286 body of an error control message (e.g., an ICMPv4 [RFC0792] error 287 message, an ICMPv6 [RFC4443] error message, etc.). 289 IP, IPvX, IPvY 290 used to generically refer to either IP protocol version, i.e., 291 IPv4 or IPv6. 293 The following abbreviations correspond to terms used within this 294 document and elsewhere in common Internetworking nomenclature: 296 DF - the IPv4 header "Don't Fragment" flag [RFC0791] 298 ETE - Egress Tunnel Endpoint 300 HLEN - the sum of MHLEN and OHLEN 302 ITE - Ingress Tunnel Endpoint 304 MHLEN - the length of any mid-layer headers and trailers 306 OHLEN - the length of the outer encapsulating headers and 307 trailers, including the outer IP header, the SEAL header and any 308 other outer headers and trailers. 310 PTB - a Packet Too Big message recognized by the inner network 311 layer, e.g., an ICMPv6 "Packet Too Big" message [RFC4443], an 312 ICMPv4 "Fragmentation Needed" message [RFC0792], etc. 314 S_MRU - the SEAL Maximum Reassembly Unit 316 S_MSS - the SEAL Maximum Segment Size 318 SCMP - the SEAL Control Message Protocol 320 SEAL_ID - an Identification value, randomly initialized and 321 monotonically incremented for each SEAL protocol packet 323 SEAL_PORT - a TCP/UDP service port number used for SEAL 325 SEAL_PROTO - an IPv4 protocol number used for SEAL 327 TE - Tunnel Endpoint (i.e., either ingress or egress) 329 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 330 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 331 document are to be interpreted as described in [RFC2119]. When used 332 in lower case (e.g., must, must not, etc.), these words MUST NOT be 333 interpreted as described in [RFC2119], but are rather interpreted as 334 they would be in common English. 336 3. Applicability Statement 338 SEAL was originally motivated by the specific case of subnetwork 339 abstraction for Mobile Ad hoc Networks (MANETs), however it soon 340 became apparent that the domain of applicability also extends to 341 subnetwork abstractions of enterprise networks, ISP networks, SOHO 342 networks, the interdomain routing core, and any other networking 343 scenario involving IP encapsulation. SEAL and its associated 344 technologies (including Virtual Enterprise Traversal (VET) 345 [I-D.templin-intarea-vet]) are functional building blocks for a new 346 Internetworking architecture based on Routing and Addressing in 347 Networks with Global Enterprise Recursion (RANGER) 348 [RFC5720][I-D.russert-rangers] and the Internet Routing Overlay 349 Network (IRON) [I-D.templin-iron]. 351 SEAL provides a network sublayer for encapsulation of an inner 352 network layer packet within outer encapsulating headers. For 353 example, for IPvX in IPvY encapsulation (e.g., as IPv4/SEAL/IPv6), 354 the SEAL header appears as a subnetwork encapsulation as seen by the 355 inner IP layer. SEAL can also be used as a sublayer within a UDP 356 data payload (e.g., as IPv4/UDP/SEAL/IPv6 similar to Teredo 357 [RFC4380]), where UDP encapsulation is typically used for NAT 358 traversal as well as operation over subnetworks that give 359 preferential treatment to the "core" Internet protocols (i.e., TCP 360 and UDP). The SEAL header is processed the same as for IPv6 361 extension headers, i.e., it is not part of the outer IP header but 362 rather allows for the creation of an arbitrarily extensible chain of 363 headers in the same way that IPv6 does. 365 SEAL supports a segmentation and reassembly capability for adapting 366 the network layer to the underlying subnetwork characteristics, where 367 the Egress Tunnel Endpoint (ETE) determines how much or how little 368 reassembly it is willing to support. In the limiting case, the ETE 369 acts as a passive observer that simply informs the Ingress Tunnel 370 Endpoint (ITE) of any MTU limitations and otherwise discards all 371 packets that arrive as multiple fragments. This mode is useful for 372 determining an appropriate MTU for tunnels between performance- 373 critical routers connected to high data rate subnetworks such as the 374 Internet DFZ, as well as for other uses in which reassembly would 375 present too great of a burden for the routers or end systems. 377 When the ETE supports reassembly, the tunnel can be used to transport 378 packets that are too large to traverse the path without 379 fragmentation. In this mode, the ITE determines the tunnel MTU based 380 on the largest packet the ETE is capable of reassembling rather than 381 on the MTU of the smallest link in the path. Therefore, tunnel 382 endpoints that use SEAL can transport packets that are much larger 383 than the underlying subnetwork links themselves can carry in a single 384 piece. 386 SEAL tunnels may be configured over paths that include not only 387 ordinary physical links, but also virtual links that may include 388 other tunnels. An example application would be linking two 389 geographically remote supercomputer centers with large MTU links by 390 configuring a SEAL tunnel across the Internet. A second example 391 would be support for sub-IP segmentation over low-end links, i.e., 392 especially over wireless transmission media such as IEEE 802.15.4, 393 broadcast radio links in Mobile Ad-hoc Networks (MANETs), Very High 394 Frequency (VHF) civil aviation data links, etc. 396 Many other use case examples are anticipated, and will be identified 397 as further experience is gained. 399 4. SEAL Protocol Specification 401 The following sections specify the operation of the SEAL protocol. 403 4.1. Model of Operation 405 SEAL is an encapsulation sublayer that supports a multi-level 406 segmentation and reassembly capability for the transmission of 407 unicast and multicast packets across an underlying IP subnetwork with 408 heterogeneous links. First, the ITE can use IPv4 fragmentation to 409 fragment inner IPv4 packets before SEAL encapsulation if necessary. 410 Secondly, the SEAL layer itself provides a simple cutting-and-pasting 411 capability for mid-layer packets that can be used to avoid IP 412 fragmentation on the outer packet. Finally, ordinary IP 413 fragmentation is permitted on the outer packet after SEAL 414 encapsulation and is used to detect and tune out any in-the-network 415 fragmentation. 417 SEAL-enabled ITEs encapsulate each inner packet in any mid-layer 418 headers and trailers, segment the resulting mid-layer packet into 419 multiple segments if necessary, then append a SEAL header and any 420 outer encapsulations to each segment. As an example, for IPv6-in- 421 IPv4 encapsulation a single-segment inner IPv6 packet encapsulated in 422 any mid-layer headers and trailers, followed by the SEAL header, 423 followed by any outer headers and trailers, followed by an outer IPv4 424 header would appear as shown in Figure 1: 426 +--------------------+ 427 ~ outer IPv4 header ~ 428 +--------------------+ 429 I ~ other outer hdrs ~ 430 n +--------------------+ 431 n ~ SEAL Header ~ 432 e +--------------------+ +--------------------+ 433 r ~ mid-layer headers ~ ~ mid-layer headers ~ 434 +--------------------+ +--------------------+ 435 I --> | | --> | | 436 P --> ~ inner IPv6 ~ --> ~ inner IPv6 ~ 437 v --> ~ Packet ~ --> ~ Packet ~ 438 6 --> | | --> | | 439 +--------------------+ +--------------------+ 440 P ~ mid-layer trailers ~ ~ mid-layer trailers ~ 441 a +--------------------+ +--------------------+ 442 c ~ outer trailers ~ 443 k Mid-layer packet +--------------------+ 444 e after mid-layer encaps. 445 t Outer IPv4 packet 446 after SEAL and outer encaps. 448 Figure 1: SEAL Encapsulation - Single Segment 450 As a second example, for IPv4-in-IPv6 encapsulation an inner IPv4 451 packet requiring three SEAL segments would appear as three separate 452 outer IPv6 packets, where the mid-layer headers are carried only in 453 segment 0 and the mid-layer trailers are carried in segment 2 as 454 shown in Figure 2: 456 +------------------+ +------------------+ 457 ~ outer IPv6 hdr ~ ~ outer IPv6 hdr ~ 458 +------------------+ +------------------+ +------------------+ 459 ~ other outer hdrs ~ ~ outer IPv6 hdr ~ ~ other outer hdrs ~ 460 +------------------+ +------------------+ +------------------+ 461 ~ SEAL hdr (SEG=0) ~ ~ other outer hdrs ~ ~ SEAL hdr (SEG=2) ~ 462 +------------------+ +------------------+ +------------------+ 463 ~ mid-layer hdrs ~ ~ SEAL hdr (SEG=1) ~ | inner IPv4 | 464 +------------------+ +------------------+ ~ Packet ~ 465 | inner IPv4 | | inner IPv4 | | (Segment 2) | 466 ~ Packet ~ ~ Packet ~ +------------------+ 467 | (Segment 0) | | (Segment 1) | ~ mid-layer trails ~ 468 +------------------+ +------------------+ +------------------+ 469 ~ outer trailers ~ ~ outer trailers ~ ~ outer trailers ~ 470 +------------------+ +------------------+ +------------------+ 472 Segment 0 (includes Segment 1 (no mid- Segment 2 (includes 473 mid-layer hdrs) layer encaps) mid-layer trails) 475 Figure 2: SEAL Encapsulation - Multiple Segments 477 The SEAL header itself is inserted according to the specific 478 tunneling protocol. Examples include the following: 480 o For simple encapsulation of an inner network layer packet within 481 an outer IPvX header (e.g., [RFC1070][RFC2003][RFC2473][RFC4213], 482 etc.), the SEAL header is inserted between the inner packet and 483 outer IPvX headers as: IPvX/SEAL/{inner packet}. 485 o For encapsulations over transports such as UDP (e.g., [RFC4380]), 486 the SEAL header is inserted between the outer transport layer 487 header and the mid-layer packet, e.g., as IPvX/UDP/SEAL/{mid-layer 488 packet}. Here, the UDP header is seen as an "other outer header". 490 SEAL-encapsulated packets include a SEAL_ID that the TEs maintain as 491 either a monotonically-incrementing packet identification number or 492 as a static nonce to identify the tunnel. When the SEAL_ID is 493 maintained as a packet identifier, routers within the subnetwork can 494 use it for duplicate packet detection and the TEs can use it for SEAL 495 segmentation/reassembly. TEs can also use the SEAL_ID to detect off- 496 path attacks whether it is maintained as a packet identifier or a 497 nonce. 499 The following sections specify the SEAL header format and SEAL- 500 related operations of the ITE and ETE, respectively. 502 4.2. SEAL Header Format 504 The SEAL header is formatted as follows: 506 0 1 2 3 507 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 508 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 509 |VER|A|I|F|M|RSV| NEXTHDR/SEG | SEAL_ID (bits 48 - 32) | 510 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 511 | SEAL_ID (bits 31 - 0) | 512 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 514 Figure 3: SEAL Header Format 516 where the header fields are defined as: 518 VER (2) 519 a 2-bit version field. This document specifies Version 0 of the 520 SEAL protocol, i.e., the VER field encodes the value 0. 522 A (1) 523 the "Acknowledgement Requested" bit. Set to 1 by the ITE in data 524 packets if it wishes to receive an explicit acknowledgement from 525 the ETE. 527 I (1) 528 the "Identifier" bit. Set to 1 if the SEAL_ID contains a 529 monotonically-incrementing packet identifier; set to 0 if the 530 SEAL_ID contains a static nonce. 532 F (1) 533 the "First Segment" bit. Set to 1 if this SEAL protocol packet 534 contains the first segment (i.e., Segment #0) of a mid-layer 535 packet. 537 M (1) 538 the "More Segments" bit. Set to 1 if this SEAL protocol packet 539 contains a non-final segment of a multi-segment mid-layer packet. 541 RSV (1) 542 a 2-bit "Reserved" field. Set to zero by the ITE and ignored by 543 the ETE. 545 NEXTHDR/SEG (8) an 8-bit field. When 'F'=1, encodes the next header 546 Internet Protocol number the same as for the IPv4 protocol and 547 IPv6 next header fields. When 'F'=0, encodes a segment number of 548 a multi-segment mid-layer packet. (The segment number 0 is 549 reserved.) 551 SEAL_ID (48) 552 a 48-bit Identification or nonce field. 554 Setting of the various bits and fields of the SEAL header is 555 specified in the following sections. 557 4.3. ITE Specification 559 4.3.1. Tunnel Interface MTU 561 The ITE configures a tunnel virtual interface over one or more 562 underlying links that connect the border node to the subnetwork. The 563 tunnel interface must present a fixed MTU to Layer 3 as the size for 564 admission of inner packets into the tunnel. Since the tunnel 565 interface may support a large set of ETEs that accept widely varying 566 maximum packet sizes, however, a number of factors should be taken 567 into consideration when selecting a tunnel interface MTU. 569 Due to the ubiquitous deployment of standard Ethernet and similar 570 networking gear, the nominal Internet cell size has become 1500 571 bytes; this is the de facto size that end systems have come to expect 572 will either be delivered by the network without loss due to an MTU 573 restriction on the path or a suitable ICMP Packet Too Big (PTB) 574 message returned. When the 1500 byte packets sent by end systems 575 incur additional encapsulation at an ITE, however, they may be 576 dropped silently since the network may not always deliver the 577 necessary PTBs [RFC2923]. 579 The ITE should therefore set a tunnel virtual interface MTU of at 580 least 1500 bytes plus extra room to accommodate any additional 581 encapsulations that may occur on the path from the original source. 582 The ITE can set larger MTU values still, but should select a value 583 that is not so large as to cause excessive PTBs coming from within 584 the tunnel interface. The ITE can also set smaller MTU values; 585 however, care must be taken not to set so small a value that original 586 sources would experience an MTU underflow. In particular, IPv6 587 sources must see a minimum path MTU of 1280 bytes, and IPv4 sources 588 should see a minimum path MTU of 576 bytes. 590 The ITE can alternatively set an indefinite MTU on the tunnel virtual 591 interface such that all inner packets are admitted into the interface 592 without regard to size. For ITEs that host applications, this option 593 must be carefully coordinated with protocol stack upper layers, since 594 some upper layer protocols (e.g., TCP) derive their packet sizing 595 parameters from the MTU of the outgoing interface and as such may 596 select too large an initial size. This is not a problem for upper 597 layers that use conservative initial maximum segment size estimates 598 and/or when the tunnel interface can reduce the upper layer's maximum 599 segment size (e.g., the size advertised in the TCP MSS option) based 600 on the per-neighbor MTU. 602 The inner network layer protocol consults the tunnel interface MTU 603 when admitting a packet into the interface. For inner IPv4 packets 604 with the IPv4 Don't Fragment (DF) bit set to 0, if the packet is 605 larger than the tunnel interface MTU the inner IPv4 layer uses IPv4 606 fragmentation to break the packet into fragments no larger than the 607 tunnel interface MTU. The ITE then admits each fragment into the 608 tunnel as an independent packet. 610 For all other inner packets, the ITE admits the packet if it is no 611 larger than the tunnel interface MTU; otherwise, it drops the packet 612 and sends a PTB error message to the source with the MTU value set to 613 the tunnel interface MTU. The message must contain as much of the 614 invoking packet as possible without the entire message exceeding the 615 network layer minimum MTU (e.g., 576 bytes for IPv4, 1280 bytes for 616 IPv6, etc.). 618 Note that when the tunnel interface sets an indefinite MTU the ITE 619 unconditionally admits all packets into the interface without 620 fragmentation. In light of the above considerations, it is 621 RECOMMENDED that the ITE configure an indefinite MTU on the tunnel 622 virtual interface and adapt to any per-neighbor MTU limitations 623 within the tunnel virtual interface as described in the following 624 sections. 626 4.3.2. Tunnel Interface Soft State 628 For each ETE, the ITE maintains soft state within the tunnel 629 interface (e.g., in a neighbor cache) used to support inner 630 fragmentation and SEAL segmentation for packets admitted into the 631 tunnel interface. The soft state includes the following: 633 o a Mid-layer Header Length (MHLEN); set to the length of any mid- 634 layer encapsulation headers and trailers that must be added before 635 SEAL segmentation. 637 o an Outer Header Length (OHLEN); set to the length of the outer IP, 638 SEAL and other outer encapsulation headers and trailers. 640 o a total Header Length (HLEN); set to MHLEN plus OHLEN. 642 o a SEAL Maximum Segment Size (S_MSS). The ITE initializes S_MSS to 643 the underlying interface MTU if the underlying interface MTU can 644 be determined (otherwise, the ITE initializes S_MSS to 645 "infinity"). The ITE decreases or increased S_MSS based on any 646 SCMP "MTU Report" messages received (see Section 4.5). 648 o a SEAL Maximum Reassembly Unit (S_MRU). If the ITE is not 649 configured to use SEAL segmentation, it initializes S_MRU to the 650 static value 0. Otherwise, it initializes S_MRU to "infinity" and 651 decreases or increases S_MRU based on any SCMP MTU Report messages 652 received (see Section 4.5). When (S_MRU>(S_MSS*256)), the ITE 653 uses (S_MSS*256) as the effective S_MRU value. 655 Note that S_MSS and S_MRU include the length of the outer and mid- 656 layer encapsulating headers and trailers (i.e., HLEN), since the ETE 657 must retain the headers and trailers during reassembly. Note also 658 that the ITE maintains S_MSS and S_MRU as 32-bit values such that 659 inner packets larger than 64KB (e.g., IPv6 jumbograms [RFC2675]) can 660 be accommodated when appropriate for a given subnetwork. 662 4.3.3. Admitting Packets into the Tunnel 664 After the ITE admits an inner packet/fragment into the tunnel 665 interface, it uses the following algorithm to determine whether the 666 packet can be accommodated and (if so) whether (further) inner IP 667 fragmentation is needed: 669 o if the inner packet is unfragmentable (e.g., an IPv6 packet, an 670 IPv4 packet with DF=1, etc.), and the packet is larger than 671 (MAX(S_MRU, S_MSS) - HLEN), the ITE drops the packet and sends a 672 PTB message to the original source with an MTU value of 673 (MAX(S_MRU, S_MSS) - HLEN); else, 675 o if the inner packet is fragmentable (e.g., an IPv4 packet with 676 DF=0), and the packet is larger than (foo) bytes, the ITE uses 677 inner fragmentation to break the packet into fragments no larger 678 than (foo) bytes; else, 680 o the ITE processes the packet without inner fragmentation. 682 In the above, the ITE must track whether the tunnel interface is 683 using header compression. If so, the ITE must include the length of 684 the uncompressed headers and trailers when calculating HLEN. Note 685 also in the above that the ITE is permitted to admit inner packets 686 into the tunnel that can be accommodated in a single SEAL segment 687 (i.e., no larger than S_MSS) even if they are larger than the ETE 688 would be willing to reassemble if fragmented (i.e., larger than 689 S_MRU) - see: Section 4.4.1. 691 When the ITE uses inner fragmentation, it should use a "safe" 692 fragment size of (foo) bytes that would be highly unlikely to incur 693 an outer IP MTU restriction within the tunnel. If the ITE can 694 determine a larger fragment size (e.g., via probing), it should use 695 the larger size for inner fragmentation. In the absence of 696 deterministic information, it is RECOMMENDED that the ITE set (foo) 697 to 1280. 699 4.3.4. Mid-Layer Encapsulation 701 After inner IP fragmentation (if necessary), the ITE next 702 encapsulates each inner packet/fragment in the MHLEN bytes of mid- 703 layer headers and trailers. The ITE then presents the mid-layer 704 packet for SEAL segmentation and outer encapsulation. 706 4.3.5. SEAL Segmentation 708 If the ITE is configured to use SEAL segmentation, it checks the 709 length of the resulting packet after mid-layer encapsulation to 710 determine whether SEAL segmentation is needed. If the length of the 711 resulting mid-layer packet plus OHLEN is larger than S_MSS but no 712 larger than S_MRU the ITE performs SEAL segmentation by breaking the 713 mid-layer packet into N segments (N <= 256) that are no larger than 714 (S_MSS - OHLEN) bytes each. Each segment, except the final one, MUST 715 be of equal length. The first byte of each segment MUST begin 716 immediately after the final byte of the previous segment, i.e., the 717 segments MUST NOT overlap. The ITE SHOULD generate the smallest 718 number of segments possible, e.g., it SHOULD NOT generate 6 smaller 719 segments when the packet could be accommodated with 4 larger 720 segments. 722 Note that this SEAL segmentation ignores the fact that the mid-layer 723 packet may be unfragmentable outside of the subnetwork. This 724 segmentation process is a mid-layer (not an IP layer) operation 725 employed by the ITE to adapt the mid-layer packet to the subnetwork 726 path characteristics, and the ETE will restore the packet to its 727 original form during reassembly. Therefore, the fact that the packet 728 may have been segmented within the subnetwork is not observable 729 outside of the subnetwork. 731 4.3.6. Outer Encapsulation 733 Following SEAL segmentation, the ITE next encapsulates each segment 734 in a SEAL header formatted as specified in Section 4.2. For the 735 first segment, the ITE sets F=1, then sets NEXTHDR to the Internet 736 Protocol number of the encapsulated inner packet, and finally sets 737 M=1 if there are more segments or sets M=0 otherwise. For each non- 738 initial segment of an N-segment mid-layer packet (N <= 256), the ITE 739 sets (F=0; M=1; SEG=1) in the SEAL header of the first non-initial 740 segment, sets (F=0; M=1; SEG=2) in the next non-initial segment, 741 etc., and sets (F=0; M=0; SEG=N-1) in the final segment. (Note that 742 the value SEG=0 is not used, since the initial segment encodes a 743 NEXTHDR value and not a SEG value.) 744 The ITE next encapsulates each segment in the requisite outer headers 745 and trailers according to the specific encapsulation format (e.g., 746 [RFC1070], [RFC2003], [RFC2473], [RFC4213], etc.), except that it 747 writes 'SEAL_PROTO' in the protocol field of the outer IP header 748 (when simple IP encapsulation is used) or writes 'SEAL_PORT' in the 749 outer destination service port field (e.g., when IP/UDP encapsulation 750 is used). The ITE finally sets A=1 if probing is necessary as 751 specified in Section 4.3.7, sets the packet identification values as 752 specified in Section 4.3.8 and sends the packets as specified in 753 Section 4.3.9. 755 4.3.7. Probing Strategy 757 All SEAL encapsulated packets sent by the ITE are considered implicit 758 probes. SEAL encapsulated packets that use IPv4 as the outer layer 759 of encapsulation will elicit SCMP PTB messages from the ETE (see: 760 Section 4.5) if any IPv4 fragmentation occurs in the path. SEAL 761 encapsulated packets that use IPv6 as the outer layer of 762 encapsulation may be dropped by an IPv6 router on the path to the ETE 763 which will also return an ICMPv6 PTB message to the ITE. The ITE can 764 then use the SEAL_ID within the packet-in-error to determine whether 765 the PTB message corresponds to one of its recent packet 766 transmissions. 768 The ITE should also send explicit probes, periodically, to verify 769 that the ETE is still reachable. The ITE sets A=1 in the SEAL header 770 of a segment to be used as an explicit probe, where the probe can be 771 either an ordinary data packet or a NULL packet created by setting 772 the NEXTHDR field to a value of "No Next Header" (see Section 4.7 of 773 [RFC2460]). The probe will elicit a solicited SCMP Neighbor 774 Advertisement (NA) message from the ETE as an acknowledgement (see 775 Section 4.5.1). 777 Finally, the ITE MAY send "expendable" outer IP probe packets (see 778 Section 4.3.9) as explicit probes in order to detect increases in the 779 path MTU to the ETE. One possible strategy is to send expendable 780 packets with A=1 in the SEAL header and DF=1 in the IP header. In 781 all cases, the ITE MUST be conservative in its use of the A bit in 782 order to limit the resultant control message overhead. 784 4.3.8. Identification 786 The ITE maintains a randomly-initialized SEAL_ID value as per-ETE 787 soft state (e.g., in the neighbor cache). If the SEAL_ID is to be 788 used as a packet identifier, the ITE monotonically increments the 789 value for each successive SEAL protocol packet it sends to the ETE. 790 If the SEAL_ID is to be used as a tunnel identifier, the ITE instead 791 maintains SEAL_ID as a static value. 793 For each successive SEAL protocol packet, the ITE writes the current 794 SEAL_ID value into the header field of the same name in the SEAL 795 header. It then sets I=1 if the SEAL_ID represents a packet 796 identifier and I=0 if the SEAL_ID represents a tunnel identifier. 798 Note that the ITE must be consistent in its setting of the I bit. 799 For example, it must not set I=1 in some packets and I=0 in others 800 since this may result in unpredictable behavior. 802 4.3.9. Sending SEAL Protocol Packets 804 Following SEAL segmentation and encapsulation, the ITE sets DF=0 in 805 the header of each outer IPv4 packet to ensure that they will be 806 delivered to the ETE even if they are fragmented within the 807 subnetwork. (The ITE can instead set DF=1 for "expendable" outer 808 IPv4 packets (e.g., for NULL packets used as probes -- see Section 809 4.3.7), but these may be lost due to an MTU restriction). For outer 810 IPv6 packets, the "DF" bit is always implicitly set to 1; hence, they 811 will not be fragmented within the subnetwork. 813 The ITE sends each outer packet that encapsulates a segment of the 814 same mid-layer packet into the tunnel in canonical order, i.e., 815 segment 0 first, followed by segment 1, etc., and finally segment 816 N-1. 818 4.3.10. Processing Raw ICMP Messages 820 The ITE may receive "raw" ICMP error messages [RFC0792][RFC4443] from 821 either the ETE or routers within the subnetwork that comprise an 822 outer IP header, followed by an ICMP header, followed by a portion of 823 the SEAL packet that generated the error (also known as the "packet- 824 in-error"). The ITE can use the SEAL_ID encoded in the packet-in- 825 error as a nonce to confirm that the ICMP message came from either 826 the ETE or an on-path router, and can use any additional information 827 to determine whether to accept or discard the message. 829 The ITE should specifically process raw ICMPv4 Protocol Unreachable 830 messages and ICMPv6 Parameter Problem messages with Code 831 "Unrecognized Next Header type encountered" as a hint that the ETE 832 does not implement the SEAL protocol; specific actions that the ITE 833 may take in this case are out of scope. 835 4.4. ETE Specification 837 4.4.1. Reassembly Buffer Requirements 839 The ETE SHOULD support IP-layer and SEAL-layer reassembly for inner 840 packets of at least 1280 bytes in length and MAY support reassembly 841 for larger inner packets. (The ETE may instead support only a 842 minimum-length reassembly buffer or even a zero-length buffer, but 843 this may cause MTU underruns in some environments.) The ETE must 844 retain the outer IP, SEAL and other outer headers and trailers during 845 both IP-layer and SEAL-layer reassembly for the purpose of 846 associating the fragments/segments of the same packet, and must also 847 configure a SEAL-layer reassembly buffer that is no smaller than the 848 IP-layer reassembly buffer. Hence, the ETE: 850 o SHOULD configure an outer IP-layer reassembly buffer size of at 851 least (1280 + HELN) bytes, and 853 o MUST be capable of discarding inner packets that require IP-layer 854 or SEAL-layer reassembly and that are larger than (S_MRU - HLEN). 856 The ETE can maintain S_MRU either as a single value to be applied for 857 all ITEs, or as a per-ITE value. In that case, the ETE can manage 858 each per-ITE S_MRU value separately (e.g., to reduce congestion 859 caused by excessive segmentation from specific ITEs) but should seek 860 to maintain as stable a value as possible for each ITE. 862 Note that the ETE is permitted to accept inner packets that did not 863 undergo IP-layer and/or SEAL-layer reassembly even if they are larger 864 than (S_MRU - HELN) bytes. Hence, S_MRU is a maximum *reassembly* 865 size, and may be less than the ETE is able to receive without 866 reassembly. 868 4.4.2. IP-Layer Reassembly 870 The ETE submits unfragmented SEAL protocol IP packets for SEAL-layer 871 reassembly as specified in Section 4.4.3. The ETE instead performs 872 standard IP-layer reassembly for multi-fragment SEAL protocol IP 873 packets as follows. 875 The ETE should maintain conservative IP-layer reassembly cache high- 876 and low-water marks. When the size of the reassembly cache exceeds 877 this high-water mark, the ETE should actively discard incomplete 878 reassemblies (e.g., using an Active Queue Management (AQM) strategy) 879 until the size falls below the low-water mark. The ETE should also 880 actively discard any pending reassemblies that clearly have no 881 opportunity for completion, e.g., when a considerable number of new 882 fragments have been received before a fragment that completes a 883 pending reassembly has arrived. Following successful IP-layer 884 reassembly, the ETE submits the reassembled packet for SEAL-layer 885 reassembly as specified in Section 4.4.3. 887 When the ETE processes the IP first fragment (i.e., one with MF=1 and 888 Offset=0 in the IP header) of a fragmented SEAL packet, it sends an 889 SCMP PTB message back to the ITE (see Section 4.5.1). When the ETE 890 processes an IP fragment that would cause the reassembled outer 891 packet to be larger than the IP-layer reassembly buffer following 892 reassembly, it discontinues the reassembly and discards any further 893 fragments of the same packet. 895 4.4.3. SEAL-Layer Reassembly 897 Following IP reassembly (if necessary), if the mid-layer packet has 898 an incorrect value in the SEAL header the ETE discards the packet and 899 returns an SCMP "Parameter Problem" message (see Section 4.5.1). 900 Next, if the SEAL header has A=1, the ETE sends a solicited SCMP 901 Neighbor Advertisement (NA) message back to the ITE (see Section 902 4.5.1). The ETE next submits single-segment mid-layer packets for 903 decapsulation and delivery to upper layers as specified in Section 904 4.4.4. The ETE instead performs SEAL-layer reassembly for multi- 905 segment mid-layer packets with I=1 in the SEAL header as follows. 907 The ETE adds each segment of a multi-segment mid-layer packet with 908 I=1 in the SEAL header to a SEAL-layer pending-reassembly queue 909 according to the (Source, Destination, SEAL_ID)-tuple found in the 910 outer IP and SEAL headers. The ETE performs SEAL-layer reassembly 911 through simple in-order concatenation of the encapsulated segments of 912 the same mid-layer packet from N consecutive SEAL segments. SEAL- 913 layer reassembly requires the ETE to maintain a cache of recently 914 received segments for a hold time that would allow for nominal inter- 915 segment delays. When a SEAL reassembly times out, the ETE discards 916 the incomplete reassembly and returns an SCMP "Time Exceeded" message 917 to the ITE (see Section 4.5.1). As for IP-layer reassembly, the ETE 918 should also maintain a conservative reassembly cache high- and low- 919 water mark and should actively discard any pending reassemblies that 920 clearly have no opportunity for completion, e.g., when a considerable 921 number of new SEAL packets have been received before a packet that 922 completes a pending reassembly has arrived. 924 If the ETE receives a SEAL packet for which a segment with the same 925 (Source, Destination, SEAL_ID)-tuple is already in the queue, it must 926 determine whether to accept the new segment and release the old, or 927 drop the new segment. If accepting the new segment would cause an 928 inconsistency with other segments already in the queue (e.g., 929 differing segment lengths), the ETE drops the segment that is least 930 likely to complete the reassembly. If the ETE accepts a new SEAL 931 segment that would cause the reassembled outer packet to be larger 932 than S_MRU following reassembly, it schedules the reassembly 933 resources for garbage collection and sends an SCMP PTB message back 934 to the ITE (see Section 4.5.1). 936 After all segments are gathered, the ETE reassembles the packet by 937 concatenating the segments encapsulated in the N consecutive SEAL 938 packets beginning with the initial segment (i.e., SEG=0) and followed 939 by any non-initial segments 1 through N-1. That is, for an N-segment 940 mid-layer packet, reassembly entails the concatenation of the SEAL- 941 encapsulated packet segments with (F=1, M=1, SEAL_ID=j) in the first 942 SEAL header, followed by (F=0, M=1, SEG=1, SEAL_ID=(j+1)) in the next 943 SEAL header, followed by (F=0, M=1, SEG=2, SEAL_ID=(j+2)), etc., up 944 to (F=0, M=0, SEG=(N-1), SEAL_ID=(j + N-1)) in the final SEAL header. 945 (Note that modulo arithmetic based on the length of the SEAL_ID field 946 is used). Following successful SEAL-layer reassembly, the ETE 947 submits the reassembled mid-layer packet for decapsulation and 948 delivery to upper layers as specified in Section 4.4.4. 950 4.4.4. Decapsulation and Delivery to Upper Layers 952 Following any necessary IP- and SEAL-layer reassembly, the ETE 953 discards the outer headers and trailers and performs any mid-layer 954 transformations on the mid-layer packet. The ETE next discards the 955 mid-layer headers and trailers, and delivers the inner packet to the 956 upper-layer protocol indicated either in the SEAL NEXTHDR field or 957 the next header field of the mid-layer packet (i.e., if the packet 958 included mid-layer encapsulations). The ETE instead silently 959 discards the inner packet if it was a NULL packet (see Section 960 4.3.9). 962 4.5. The SEAL Control Message Protocol (SCMP) 964 SEAL uses a companion SEAL Control Message Protocol (SCMP) based on 965 the same message format as the Internet Control Message Protocol for 966 IPv6 (ICMPv6) [RFC4443]. SCMP messages are further identified by the 967 NEXTHDR value '58' the same as for ICMPv6 messages, however the SCMP 968 message is *not* immediately preceded by an inner IPv6 header. 969 Instead, SCMP messages appear immediately following the SEAL header 970 which allows TEs to differentiate them from ordinary ICMPv6 messages. 971 Unlike ICMPv6 messages, SCMP messages are used only for the purpose 972 of conveying information between TEs, i.e., they are used only for 973 sharing control information within the tunnel and not beyond the 974 tunnel. 976 The following sections specify the generation and processing of SCMP 977 messages: 979 4.5.1. Generating SCMP Messages 981 SCMP messages may be generated by either ITEs or ETEs (i.e., by any 982 TE) using use the same message Type and Code values specified for 983 ordinary ICMPv6 messages in [RFC4443]. SCMP can also be used to 984 carry other message types and their associated options as specified 985 in other documents (e.g., [RFC4191][RFC4861]). The general format 986 for SCMP messages is shown in Figure 4: 988 0 1 2 3 989 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 990 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 991 | Type | Code | Checksum | 992 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 993 | | 994 ~ Message Body ~ 995 | | 996 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 997 | As much of invoking SEAL data | 998 ~ packet as possible without the SCMP ~ 999 | packet exceeding 576 bytes (*) | 1000 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 1002 (*) also known as the "packet-in-error" 1004 Figure 4: SCMP Message Format 1006 As for ordinary ICMPv6 messages, the SCMP message begins with a 4 1007 byte header that includes 8-bit Type and Code fields followed by a 1008 16-bit Checksum field followed by a variable-length Message Body. 1009 The Message Body is followed by the leading portion of the invoking 1010 SEAL data packet (i.e., the "packet-in-error") IFF the packet-in- 1011 error would also be included in the corresponding ICMPv6 message. 1012 The TE sets the Type and Code fields to the same values that would 1013 appear in the corresponding ICMPv6 message and also formats the 1014 message body the same as for the corresponding ICMPv6 message except 1015 as otherwise specified. 1017 If the SCMP message will include a packet-in-error, the TE then 1018 includes as much of the leading portion of the invoking SEAL data 1019 packet as possible beginning with the outer IP header and extending 1020 to a length that would not cause the entire SCMP message following 1021 encapsulation to exceed 576 bytes. The ITE finally calculates the 1022 Checksum the same as specified for ICMPv4 messages [RFC0792] and does 1023 not include a pseudo-header of the outer IP header since the SEAL_ID 1024 gives sufficient assurance against mis-delivery. The TE then 1025 encapsulates the SCMP message in the outer headers as shown in 1026 Figure 5: 1028 +--------------------+ 1029 ~ outer IPv4 header ~ 1030 +--------------------+ 1031 ~ other outer hdrs ~ 1032 +--------------------+ 1033 ~ SEAL Header ~ 1034 +--------------------+ +--------------------+ 1035 ~ SCMP message header~ --> ~ SCMP message header~ 1036 +--------------------+ --> +--------------------+ 1037 ~ SCMP message body ~ --> ~ SCMP message body ~ 1038 +--------------------+ --> +--------------------+ 1039 ~ packet-in-error ~ --> ~ packet-in-error ~ 1040 +--------------------+ +--------------------+ 1041 ~ outer trailers ~ 1042 SCMP Message +--------------------+ 1043 before encapsulation 1044 SCMP Message 1045 after encapsulation 1047 Figure 5: SCMP Message Encapsulation 1049 When a TE generates an SCMP message in response to a packet-in-error, 1050 it sets the outer IP destination and source addresses of the SCMP 1051 packet to the packet-in-error's source and destination addresses 1052 (respectively). (If the destination address in the packet-in-error 1053 was multicast, the TE instead sets the outer IP source address of the 1054 SCMP packet to an address assigned to the underlying IP interface.) 1055 When a TE generates an SCMP message that is not due to a packet-in- 1056 error, it sets the outer IP destination and source addresses of the 1057 SCMP packet the same as for ordinary data packets. The TE finally 1058 sets the NEXTHDR field in the SEAL header to the value '58' (i.e., 1059 the official IANA protocol number for the ICMPv6 protocol) and sends 1060 the SCMP message to the tunnel far end. 1062 4.5.1.1. Generating SCMP Packet Too Big (PTB) Messages 1064 An ETE generates an SCMP Packet Too Big (PTB) message when it 1065 receives the IP first fragment (i.e., one with MF=1 and Offset=0 in 1066 the outer IP header) of a SEAL protocol packet that arrived as 1067 multiple IP fragments, or when it discontinues reassembly of a SEAL 1068 protocol packet that arrived as multiple IP fragments and/or multiple 1069 SEAL segments and would exceed S_MRU following reassembly. 1071 The ETE prepares an SCMP PTB message the same as for the 1072 corresponding ICMPv6 PTB message, except that it writes the value 0 1073 in the MTU field of the message if the PTB is generated as a result 1074 of receiving an IP first fragment and writes the S_MRU value for this 1075 ITE in the MTU field otherwise. 1077 4.5.1.2. Generating SCMP Neighbor Discovery Messages 1079 An ITE generates an SCMP "Neighbor Solicitation" (NS) or "Router 1080 Solicitation" (RS) message when it needs to solicit a response from 1081 an ETE. An ETE generates a solicited SCMP "Neighbor Advertisement" 1082 (NA) or "Router Advertisement" (RA) message when it receives an NS/RS 1083 message, and also generates a solicited NA message when it receives a 1084 SEAL protocol packet with A=1 in the SEAL header. Any TE may also 1085 generate unsolicited NA/RA messages that are not triggered by a 1086 specific solicitation event. 1088 The TE generates NS/RS and NA/RA messages the same as described for 1089 the corresponding IPv6 Neighbor Discovery (ND) messages (see: 1090 [RFC4861]), except that for solicited NA/RA messages it also includes 1091 a Redirected Header option formatted the same as for an IPv6 ND 1092 Redirect message. The messages may also be used in conjunction with 1093 the tunnel endpoint synchronization procedure specified in Section 1094 4.6. 1096 4.5.1.3. Generating Other SCMP Messages 1098 An ETE generates an SCMP "Destination Unreachable - Communication 1099 with Destination Administratively Prohibited" message when it is 1100 operating in synchronized mode and receives a SEAL packet with a 1101 SEAL_ID that is outside of the current window for this ITE (see: 1102 Section 4.6). An ETE also generates an SCMP "Destination 1103 Unreachable" message with an appropriate code under the same 1104 circumstances that an IPv6 system would generate an ICMPv6 1105 Destination Unreachable message using the same code. The SCMP 1106 Destination Unreachable message is formatted the same as for ICMPv6 1107 Destination Unreachable messages. 1109 An ETE generates an SCMP "Parameter Problem" message when it receives 1110 a SEAL packet with an incorrect value in the SEAL header, and 1111 generates an SCMP "Time Exceeded" message when it garbage collects an 1112 incomplete SEAL data packet reassembly. The message formats used are 1113 the same as for the corresponding ICMPv6 messages. 1115 Generation of all other SCMP message types is outside the scope of 1116 this document. 1118 4.5.2. Processing SCMP Messages 1120 An ITE processes any SCMP messages it receives as long as it can 1121 verify that the message was sent from an on-path ETE. The ITE can 1122 verify that the SCMP message came from an on-path ETE by checking 1123 that the SEAL_ID in the encapsulated packet-in-error corresponds to 1124 one of its recently-sent SEAL data packets. 1126 An ITE maintains a window of SEAL_IDs of packets that it has recently 1127 sent to each ETE. For each SCMP message it receives, the ITE first 1128 verifies that the SEAL_ID encoded in the packet-in-error is within 1129 the window of packets that it has recently sent to the ETE.. The ITE 1130 then verifies that the Checksum in the SCMP message header is 1131 correct. If the SEAL_ID is outside of the window and/or the checksum 1132 is incorrect, the ITE discards the message; otherwise, it processes 1133 the message the same as for ordinary ICMPv6 messages. 1135 Any TE may also receive unsolicited SCMP messages from the tunnel far 1136 end. When the TEs are synchronized, they can also check that the 1137 SEAL_ID in the SEAL header of an SCMP message is within the window of 1138 recently received packets from this tunnel far end (see Section 4.6). 1140 Finally, TEs process SCMP messages as an indication that the tunnel 1141 far end is responsive, i.e., in the same manner implied for IPv6 1142 Neighbor Unreachability Detection "hints of forward progress" (see: 1143 [RFC4861]). 1145 4.5.2.1. Processing SCMP PTB Messages 1147 An ITE may receive an SCMP PTB message after it sends a SEAL data 1148 packet (see: Section 4.5.1). When the ITE receives an SCMP PTB 1149 message, it examines the MTU field in the message. If the MTU field 1150 is non-zero, the PTB was the result of a reassembly buffer 1151 limitation; in that case, the ITE records the value in the MTU field 1152 as the new S_MRU value for this ETE then (optionally) sends a 1153 translated PTB message of the inner network layer protocol to the 1154 original source with MTU set to (MAX(S_MRU, S_MSS) - HLEN). If the 1155 MTU field is zero, however, the PTB was the result of an IP 1156 fragmentation event; in that case, the ITE does not send back a 1157 translated PTB message but determines a new S_MSS value according to 1158 the length recorded in the IP header of the packet-in-error as 1159 follows: 1161 o If the length is no less than 1280, the ITE records the length as 1162 the new S_MSS value. 1164 o If the length is less than the current S_MSS value and also less 1165 than 1280, the ITE can discern that IP fragmentation is occurring 1166 but it cannot determine the true MTU of the restricting link due 1167 to the possibility that a router on the path is generating runt 1168 first fragments. 1170 In this latter case, the ITE must search for a reduced S_MSS value 1171 through an iterative searching strategy that parallels (Section 5 of 1172 [RFC1191]). This searching strategy may require multiple iterations 1173 in which the ITE sends SEAL data packets using a reduced S_MSS and 1174 receives additional SCMP MTU Report messages. During this process, 1175 it is essential that the ITE reduce S_MSS based on the first SCMP MTU 1176 Report message received under the current S_MSS size, and refrain 1177 from further reducing S_MSS until SCMP MTU Report messages pertaining 1178 to packets sent under the new S_MSS are received. 1180 4.5.2.2. Processing SCMP Neighbor Discovery Messages 1182 An ETE may received NS/RS messages from an ITE as an the initial leg 1183 in a neighbor discovery exchange. An ITE may receive both solicited 1184 and unsolicited NA/RA messages from an ETE, where solicited NA/RA 1185 messages are distinguished by their inclusion of a Redirected header 1186 option (see: Section 4.5.1). 1188 The TE processes NS/RS and NA/RA messages the same as described for 1189 the corresponding IPv6 Neighbor Discovery (ND) messages (see: 1190 [RFC4861]). The messages may also be used in conjunction with the 1191 tunnel endpoint synchronization procedure specified in Section 4.6. 1193 4.5.2.3. Processing Other SCMP Messages 1195 An ITE may receive an SCMP "Destination Unreachable - Communication 1196 with Destination Administratively Prohibited" message after it sends 1197 a SEAL data packet. The ITE processes this message as an indication 1198 that it needs to (re)synchronize with the ETE (see: Section 4.6). An 1199 ITE may also receive an SCMP "Destination Unreachable" message with 1200 an appropriate code under the same circumstances that an IPv6 host 1201 would receive an ICMPv6 Destination Unreachable message. 1203 An ITE may receive an SCMP "Parameter Problem" message when the ETE 1204 receives a SEAL packet with an incorrect value in the SEAL header. 1205 The ITE should examine the incorrect SEAL header field setting to 1206 determine whether a different setting should be used in subsequent 1207 packets. 1209 .An ITE may receive an SCMP "Time Exceeded" message when the ETE 1210 garbage collects an incomplete SEAL data packet reassembly. The ITE 1211 should consider the message as an indication of congestion. 1213 Processing of all other SCMP message types is outside the scope of 1214 this document. 1216 4.6. Tunnel Endpoint Synchronization 1218 The SEAL ITE maintains a per-ETE window of SEAL_IDs of its recently- 1219 sent packets, but by default the SEAL ETE does not retain inter- 1220 packet state. When closer synchronization is required, SEAL Tunnel 1221 Endpoints (TEs) can exchange initial sequence numbers in a procedure 1222 that parallels IPv6 neighbor discovery and the TCP 3-way handshake. 1223 When the TEs are synchronized, the ETE can also maintain a per-ITE 1224 window of SEAL_IDs of its recently-received packets. 1226 When an initiating TE ("TE(A)") needs to synchronize with a new 1227 tunnel far end ("TE(B)"), it first chooses a randomly-initialized 48- 1228 bit SEAL_ID value that it would like TE(B) to use (i.e., 1229 "SEAL_ID(B)"). TE(A) then creates a neighbor cache entry for TE(B) 1230 and records SEAL_ID(B) in the neighbor cache entry. Next, TE(A) 1231 creates an SCMP NS or RS message that includes a Nonce option (see: 1232 [RFC3971], Section 5.3). TE(A) then writes the value SEAL_ID(B) in 1233 the Nonce option, writes the value 0 in the SEAL_ID field of the SEAL 1234 header and sends the NS/RS message to TE(B). 1236 When TE(B) receives an NS/RS message with a Nonce option and with the 1237 value 0 in the SEAL_ID of the SEAL header, it considers the message 1238 as a potential synchronization request. TE(B) first extracts the 1239 value SEAL_ID(B) from the Nonce option then chooses a randomly- 1240 initialized 48-bit SEAL_ID value that it would like TE(A) to use 1241 (i.e., "SEAL_ID(A)"). TE(B) then stores the tuple (ip_src, 1242 SEAL_ID(A), SEAL_ID(B)) in a minimal temporary fast path data 1243 structure, where "ip_src" is the outer IP source address of the SCMP 1244 message. (For efficiency and security purposes, the data structure 1245 should be indexed, e.g., by a secret hash of the -tuple). TE(B) then 1246 creates a solicited SCMP NA or RA message that includes a Nonce 1247 option. It then writes the value SEAL_ID(A) in the Nonce option, 1248 writes the value SEAL_ID(B) in the SEAL_ID field of the SEAL header 1249 and sends the NA/RA message back to TE(A). 1251 When TE(A) receives the NA/RA, it considers the message as a 1252 potential synchronization acknowledgement. TE(A) first verifies that 1253 the value encoded in the SEAL_ID of the SEAL header matches the 1254 SEAL_ID(B) in the neighbor cache entry. If the values match, TE(A) 1255 extracts SEAL_ID(A) from the nonce option and records it in the 1256 neighbor cache entry; otherwise, it drops the packet. If instead 1257 TE(A) does not receive a timely NA/RA response, it retransmits the 1258 initial NS/RS message for a total of 3 tries before giving up the 1259 same as for ordinary IPv6 neighbor discovery. 1261 After TE(A) receives the synchronization acknowledgement, it begins 1262 sending either unsolicited NA/RA messages or ordinary data packets 1263 back to TE(B) using SEAL_ID(A) as the initial sequence number. When 1264 TE(B) receives these packets, it first checks its neighbor cache to 1265 see if there is a matching neighbor cache entry. If there is a 1266 neighbor cache entry, and the SEAL_ID in the header of the packet is 1267 within the window of the SEAL_ID recorded in the neighbor cache 1268 entry, TE(B) accepts the packet. If the SEAL_ID in the packet is 1269 newer than the SEAL_ID in the neighbor cache entry, TE(B) also 1270 updates the neighbor cache value. If there is no neighbor cache 1271 entry, TE(B) instead checks the fast path cache to see if the packet 1272 is a match for an in-progress synchronization event. If there is a 1273 fast path cache entry with a SEAL_ID(A) that is within the window of 1274 the SEAL_ID in the packet header, TE(B) accepts the packet and also 1275 creates a new neighbor cache entry with the tuple (ip_src, 1276 SEAL_ID(A), SEAL_ID(B)). If there is no matching fast path cache 1277 entry, TE(B) instead simply discards the packet. 1279 By maintaining the fast path cache, each TE is able to mitigate 1280 buffer exhaustion attacks that may be launched by off-path attackers 1281 [RFC4987]. The TE will receive positive confirmation that the 1282 synchronization request came from an on-path tunnel far end after it 1283 receives a stream of in-window packets as the "third leg" of this 1284 three-way handshake as described above. The TEs should maintain 1285 neighbor cache entries as long as they receive hints of forward 1286 progress from the tunnel far end, but should delete the neighbor 1287 cache entries after a nominal stale time (e.g., 30 seconds). The TEs 1288 should also purge fast-path cache entries for which no window 1289 synchronization messages are received within a nominal stale time 1290 (e.g., 5 seconds). 1292 After synchronization is complete, when a TE receives a SEAL packet 1293 it checks in its neighbor cache to determine whether the SEAL_ID is 1294 within the current window, and discards any packets that are outside 1295 the window. Since packets may be lost or reordered, and since SEAL 1296 presents only a best effort (i.e., and not reliable) link model, the 1297 TE should set a coarse-grained window size (e.g., 32768) and accept 1298 any packet with a SEAL_ID that is within the window. 1300 Note that when the ITE sends SEAL packets with I=0, the window is 1301 trivial and a constant SEAL_ID nonce value instead of an incrementing 1302 sequence number is used. 1304 5. Link Requirements 1306 Subnetwork designers are expected to follow the recommendations in 1307 Section 2 of [RFC3819] when configuring link MTUs. 1309 6. End System Requirements 1311 SEAL provides robust mechanisms for returning PTB messages; however, 1312 end systems that send unfragmentable IP packets larger than 1500 1313 bytes are strongly encouraged to implement their own end-to-end MTU 1314 assurance, e.g., using Packetization Layer Path MTU Discovery per 1315 [RFC4821]. 1317 7. Router Requirements 1319 IPv4 routers within the subnetwork are strongly encouraged to 1320 implement IPv4 fragmentation such that the first fragment is the 1321 largest and approximately the size of the underlying link MTU, i.e., 1322 they should avoid generating runt first fragments. 1324 IPv6 routers within the subnetwork are required to generate the 1325 necessary PTB messages when they drop outer IPv6 packets due to an 1326 MTU restriction. 1328 8. IANA Considerations 1330 The IANA is instructed to allocate an IP protocol number for 1331 'SEAL_PROTO' in the 'protocol-numbers' registry. 1333 The IANA is instructed to allocate a Well-Known Port number for 1334 'SEAL_PORT' in the 'port-numbers' registry. 1336 The IANA is instructed to establish a "SEAL Protocol" registry to 1337 record SEAL Version values. This registry should be initialized to 1338 include the initial SEAL Version number, i.e., Version 0. 1340 9. Security Considerations 1342 Unlike IPv4 fragmentation, overlapping fragment attacks are not 1343 possible due to the requirement that SEAL segments be non- 1344 overlapping. This condition is naturally enforced due to the fact 1345 that each consecutive SEAL segment begins at offset 0 with respect to 1346 the previous SEAL segment. 1348 An amplification/reflection attack is possible when an attacker sends 1349 IP first fragments with spoofed source addresses to an ETE, resulting 1350 in a stream of SCMP messages returned to a victim ITE. The SEAL_ID 1351 in the encapsulated segment of the spoofed IP first fragment provides 1352 mitigation for the ITE to detect and discard spurious SCMP messages. 1354 The SEAL header is sent in-the-clear (outside of any IPsec/ESP 1355 encapsulations) the same as for the outer IP and other outer headers. 1356 In this respect, the threat model is no different than for IPv6 1357 extension headers. As for IPv6 extension headers, the SEAL header is 1358 protected only by L2 integrity checks and is not covered under any L3 1359 integrity checks. 1361 SCMP messages carry the SEAL_ID of the packet-in-error. Therefore, 1362 when an ITE receives an SCMP message it can unambiguously associate 1363 it with the SEAL data packet that triggered the error. When the TEs 1364 are synchronized, the ETE can also detect off-path spoofing attacks. 1366 Security issues that apply to tunneling in general are discussed in 1367 [I-D.ietf-v6ops-tunnel-security-concerns]. 1369 10. Related Work 1371 Section 3.1.7 of [RFC2764] provides a high-level sketch for 1372 supporting large tunnel MTUs via a tunnel-level segmentation and 1373 reassembly capability to avoid IP level fragmentation, which is in 1374 part the same approach used by SEAL. SEAL could therefore be 1375 considered as a fully functioned manifestation of the method 1376 postulated by that informational reference. 1378 Section 3 of [RFC4459] describes inner and outer fragmentation at the 1379 tunnel endpoints as alternatives for accommodating the tunnel MTU; 1380 however, the SEAL protocol specifies a mid-layer segmentation and 1381 reassembly capability that is distinct from both inner and outer 1382 fragmentation. 1384 Section 4 of [RFC2460] specifies a method for inserting and 1385 processing extension headers between the base IPv6 header and 1386 transport layer protocol data. The SEAL header is inserted and 1387 processed in exactly the same manner. 1389 The concepts of path MTU determination through the report of 1390 fragmentation and extending the IP Identification field were first 1391 proposed in deliberations of the TCP-IP mailing list and the Path MTU 1392 Discovery Working Group (MTUDWG) during the late 1980's and early 1393 1990's. SEAL supports a report fragmentation capability using bits 1394 in an extension header (the original proposal used a spare bit in the 1395 IP header) and supports ID extension through a 16-bit field in an 1396 extension header (the original proposal used a new IP option). A 1397 historical analysis of the evolution of these concepts, as well as 1398 the development of the eventual path MTU discovery mechanism for IP, 1399 appears in Appendix D of this document. 1401 11. SEAL Advantages over Classical Methods 1403 The SEAL approach offers a number of distinct advantages over the 1404 classical path MTU discovery methods [RFC1191] [RFC1981]: 1406 1. Classical path MTU discovery always results in packet loss when 1407 an MTU restriction is encountered. Using SEAL, IP fragmentation 1408 provides a short-term interim mechanism for ensuring that packets 1409 are delivered while SEAL adjusts its packet sizing parameters. 1411 2. Classical path MTU may require several iterations of dropping 1412 packets and returning PTB messages until an acceptable path MTU 1413 value is determined. Under normal circumstances, SEAL determines 1414 the correct packet sizing parameters in a single iteration. 1416 3. Using SEAL, ordinary packets serve as implicit probes without 1417 exposing data to unnecessary loss. SEAL also provides an 1418 explicit probing mode not available in the classic methods. 1420 4. Using SEAL, ETEs encapsulate SCMP error messages in outer and 1421 mid-layer headers such that packet-filtering network middleboxes 1422 will not filter them the same as for "raw" ICMP messages that may 1423 be generated by an attacker. 1425 5. The SEAL approach ensures that the tunnel either delivers or 1426 deterministically drops packets according to their size, which is 1427 a required characteristic of any IP link. 1429 6. Most importantly, all SEAL packets have an Identification field 1430 that is sufficiently long to be used for duplicate packet 1431 detection purposes and to associate ICMP error messages with 1432 actual packets sent without requiring per-packet state; hence, 1433 SEAL avoids certain denial-of-service attack vectors open to the 1434 classical methods. 1436 12. Acknowledgments 1438 The following individuals are acknowledged for helpful comments and 1439 suggestions: Jari Arkko, Fred Baker, Iljitsch van Beijnum, Oliver 1440 Bonaventure, Teco Boot, Bob Braden, Brian Carpenter, Steve Casner, 1441 Ian Chakeres, Noel Chiappa, Remi Denis-Courmont, Remi Despres, Ralph 1442 Droms, Aurnaud Ebalard, Gorry Fairhurst, Dino Farinacci, Joel 1443 Halpern, Sam Hartman, John Heffner, Thomas Henderson, Bob Hinden, 1444 Christian Huitema, Eliot Lear, Darrel Lewis, Joe Macker, Matt Mathis, 1445 Erik Nordmark, Dan Romascanu, Dave Thaler, Joe Touch, Mark Townsley, 1446 Ole Troan, Margaret Wasserman, Magnus Westerlund, Robin Whittle, 1447 James Woodyatt, and members of the Boeing Research & Technology NST 1448 DC&NT group. 1450 Path MTU determination through the report of fragmentation was first 1451 proposed by Charles Lynn on the TCP-IP mailing list in 1987. 1452 Extending the IP identification field was first proposed by Steve 1453 Deering on the MTUDWG mailing list in 1989. 1455 13. References 1457 13.1. Normative References 1459 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, 1460 September 1981. 1462 [RFC0792] Postel, J., "Internet Control Message Protocol", STD 5, 1463 RFC 792, September 1981. 1465 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1466 Requirement Levels", BCP 14, RFC 2119, March 1997. 1468 [RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1469 (IPv6) Specification", RFC 2460, December 1998. 1471 [RFC3971] Arkko, J., Kempf, J., Zill, B., and P. Nikander, "SEcure 1472 Neighbor Discovery (SEND)", RFC 3971, March 2005. 1474 [RFC4443] Conta, A., Deering, S., and M. Gupta, "Internet Control 1475 Message Protocol (ICMPv6) for the Internet Protocol 1476 Version 6 (IPv6) Specification", RFC 4443, March 2006. 1478 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 1479 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 1480 September 2007. 1482 13.2. Informative References 1484 [FOLK] Shannon, C., Moore, D., and k. claffy, "Beyond Folklore: 1485 Observations on Fragmented Traffic", December 2002. 1487 [FRAG] Kent, C. and J. Mogul, "Fragmentation Considered Harmful", 1488 October 1987. 1490 [I-D.ietf-intarea-ipv4-id-update] 1491 Touch, J., "Updated Specification of the IPv4 ID Field", 1492 draft-ietf-intarea-ipv4-id-update-00 (work in progress), 1493 March 2010. 1495 [I-D.ietf-tcpm-icmp-attacks] 1496 Gont, F., "ICMP attacks against TCP", 1497 draft-ietf-tcpm-icmp-attacks-12 (work in progress), 1498 March 2010. 1500 [I-D.ietf-v6ops-tunnel-security-concerns] 1501 Hoagland, J., Krishnan, S., and D. Thaler, "Security 1502 Concerns With IP Tunneling", 1503 draft-ietf-v6ops-tunnel-security-concerns-02 (work in 1504 progress), March 2010. 1506 [I-D.russert-rangers] 1507 Russert, S., Fleischman, E., and F. Templin, "RANGER 1508 Scenarios", draft-russert-rangers-05 (work in progress), 1509 July 2010. 1511 [I-D.templin-intarea-vet] 1512 Templin, F., "Virtual Enterprise Traversal (VET)", 1513 draft-templin-intarea-vet-15 (work in progress), 1514 June 2010. 1516 [I-D.templin-iron] 1517 Templin, F., "The Internet Routing Overlay Network 1518 (IRON)", draft-templin-iron-08 (work in progress), 1519 July 2010. 1521 [MTUDWG] "IETF MTU Discovery Working Group mailing list, 1522 gatekeeper.dec.com/pub/DEC/WRL/mogul/mtudwg-log, November 1523 1989 - February 1995.". 1525 [RFC1063] Mogul, J., Kent, C., Partridge, C., and K. McCloghrie, "IP 1526 MTU discovery options", RFC 1063, July 1988. 1528 [RFC1070] Hagens, R., Hall, N., and M. Rose, "Use of the Internet as 1529 a subnetwork for experimentation with the OSI network 1530 layer", RFC 1070, February 1989. 1532 [RFC1191] Mogul, J. and S. Deering, "Path MTU discovery", RFC 1191, 1533 November 1990. 1535 [RFC1981] McCann, J., Deering, S., and J. Mogul, "Path MTU Discovery 1536 for IP version 6", RFC 1981, August 1996. 1538 [RFC2003] Perkins, C., "IP Encapsulation within IP", RFC 2003, 1539 October 1996. 1541 [RFC2473] Conta, A. and S. Deering, "Generic Packet Tunneling in 1542 IPv6 Specification", RFC 2473, December 1998. 1544 [RFC2675] Borman, D., Deering, S., and R. Hinden, "IPv6 Jumbograms", 1545 RFC 2675, August 1999. 1547 [RFC2764] Gleeson, B., Heinanen, J., Lin, A., Armitage, G., and A. 1548 Malis, "A Framework for IP Based Virtual Private 1549 Networks", RFC 2764, February 2000. 1551 [RFC2923] Lahey, K., "TCP Problems with Path MTU Discovery", 1552 RFC 2923, September 2000. 1554 [RFC3232] Reynolds, J., "Assigned Numbers: RFC 1700 is Replaced by 1555 an On-line Database", RFC 3232, January 2002. 1557 [RFC3366] Fairhurst, G. and L. Wood, "Advice to link designers on 1558 link Automatic Repeat reQuest (ARQ)", BCP 62, RFC 3366, 1559 August 2002. 1561 [RFC3819] Karn, P., Bormann, C., Fairhurst, G., Grossman, D., 1562 Ludwig, R., Mahdavi, J., Montenegro, G., Touch, J., and L. 1563 Wood, "Advice for Internet Subnetwork Designers", BCP 89, 1564 RFC 3819, July 2004. 1566 [RFC4191] Draves, R. and D. Thaler, "Default Router Preferences and 1567 More-Specific Routes", RFC 4191, November 2005. 1569 [RFC4213] Nordmark, E. and R. Gilligan, "Basic Transition Mechanisms 1570 for IPv6 Hosts and Routers", RFC 4213, October 2005. 1572 [RFC4380] Huitema, C., "Teredo: Tunneling IPv6 over UDP through 1573 Network Address Translations (NATs)", RFC 4380, 1574 February 2006. 1576 [RFC4459] Savola, P., "MTU and Fragmentation Issues with In-the- 1577 Network Tunneling", RFC 4459, April 2006. 1579 [RFC4821] Mathis, M. and J. Heffner, "Packetization Layer Path MTU 1580 Discovery", RFC 4821, March 2007. 1582 [RFC4963] Heffner, J., Mathis, M., and B. Chandler, "IPv4 Reassembly 1583 Errors at High Data Rates", RFC 4963, July 2007. 1585 [RFC4987] Eddy, W., "TCP SYN Flooding Attacks and Common 1586 Mitigations", RFC 4987, August 2007. 1588 [RFC5445] Watson, M., "Basic Forward Error Correction (FEC) 1589 Schemes", RFC 5445, March 2009. 1591 [RFC5720] Templin, F., "Routing and Addressing in Networks with 1592 Global Enterprise Recursion (RANGER)", RFC 5720, 1593 February 2010. 1595 [TBIT] Medina, A., Allman, M., and S. Floyd, "Measuring 1596 Interactions Between Transport Protocols and Middleboxes", 1597 October 2004. 1599 [TCP-IP] "Archive/Hypermail of Early TCP-IP Mail List, 1600 http://www-mice.cs.ucl.ac.uk/multimedia/misc/tcp_ip/, May 1601 1987 - May 1990.". 1603 [WAND] Luckie, M., Cho, K., and B. Owens, "Inferring and 1604 Debugging Path MTU Discovery Failures", October 2005. 1606 Appendix A. Reliability 1608 Although a SEAL tunnel may span an arbitrarily-large subnetwork 1609 expanse, the IP layer sees the tunnel as a simple link that supports 1610 the IP service model. Since SEAL supports segmentation at a layer 1611 below IP, SEAL therefore presents a case in which the link unit of 1612 loss (i.e., a SEAL segment) is smaller than the end-to-end 1613 retransmission unit (e.g., a TCP segment). 1615 Links with high bit error rates (BERs) (e.g., IEEE 802.11) use 1616 Automatic Repeat-ReQuest (ARQ) mechanisms [RFC3366] to increase 1617 packet delivery ratios, while links with much lower BERs typically 1618 omit such mechanisms. Since SEAL tunnels may traverse arbitrarily- 1619 long paths over links of various types that are already either 1620 performing or omitting ARQ as appropriate, it would therefore often 1621 be inefficient to also require the tunnel to perform ARQ. 1623 When the SEAL ITE has knowledge that the tunnel will traverse a 1624 subnetwork with non-negligible loss due to, e.g., interference, link 1625 errors, congestion, etc., it can solicit Segment Reports from the ETE 1626 periodically to discover missing segments for retransmission within a 1627 single round-trip time. However, retransmission of missing segments 1628 may require the ITE to maintain considerable state and may also 1629 result in considerable delay variance and packet reordering. 1631 SEAL may also use alternate reliability mechanisms such as Forward 1632 Error Correction (FEC). A simple FEC mechanism may merely entail 1633 gratuitous retransmissions of duplicate data, however more efficient 1634 alternatives are also possible. Basic FEC schemes are discussed in 1635 [RFC5445]. 1637 The use of ARQ and FEC mechanisms for improved reliability are for 1638 further study. 1640 Appendix B. Integrity 1642 Each link in the path over which a SEAL tunnel is configured is 1643 responsible for link layer integrity verification for packets that 1644 traverse the link. As such, when a multi-segment SEAL packet with N 1645 segments is reassembled, its segments will have been inspected by N 1646 independent link layer integrity check streams instead of a single 1647 stream that a single segment SEAL packet of the same size would have 1648 received. Intuitively, a reassembled packet subjected to N 1649 independent integrity check streams of shorter-length segments would 1650 seem to have integrity assurance that is no worse than a single- 1651 segment packet subjected to only a single integrity check steam, 1652 since the integrity check strength diminishes in inverse proportion 1653 with segment length. In any case, the link-layer integrity assurance 1654 for a multi-segment SEAL packet is no different than for a multi- 1655 fragment IPv6 packet. 1657 Fragmentation and reassembly schemes must also consider packet- 1658 splicing errors, e.g., when two segments from the same packet are 1659 concatenated incorrectly, when a segment from packet X is reassembled 1660 with segments from packet Y, etc. The primary sources of such errors 1661 include implementation bugs and wrapping IP ID fields. In terms of 1662 implementation bugs, the SEAL segmentation and reassembly algorithm 1663 is much simpler than IP fragmentation resulting in simplified 1664 implementations. In terms of wrapping ID fields, when IPv4 is used 1665 as the outer IP protocol, the 16-bit IP ID field can wrap with only 1666 64K packets with the same (src, dst, protocol)-tuple alive in the 1667 system at a given time [RFC4963] increasing the likelihood of 1668 reassembly mis-associations. However, SEAL ensures that any outer 1669 IPv4 fragmentation and reassembly will be short-lived and tuned out 1670 as soon as the ITE receives a Reassembly Repot, and SEAL segmentation 1671 and reassembly uses a much longer ID field. Therefore, reassembly 1672 mis-associations of IP fragments nor of SEAL segments should be 1673 prohibitively rare. 1675 Appendix C. Transport Mode 1677 SEAL can also be used in "transport-mode", e.g., when the inner layer 1678 comprises upper-layer protocol data rather than an encapsulated IP 1679 packet. For instance, TCP peers can negotiate the use of SEAL for 1680 the carriage of protocol data encapsulated as IPv4/SEAL/TCP. In this 1681 sense, the "subnetwork" becomes the entire end-to-end path between 1682 the TCP peers and may potentially span the entire Internet. 1684 Section specifies the operation of SEAL in "tunnel mode", i.e., when 1685 there are both an inner and outer IP layer with a SEAL encapsulation 1686 layer between. However, the SEAL protocol can also be used in a 1687 "transport mode" of operation within a subnetwork region in which the 1688 inner-layer corresponds to a transport layer protocol (e.g., UDP, 1689 TCP, etc.) instead of an inner IP layer. 1691 For example, two TCP endpoints connected to the same subnetwork 1692 region can negotiate the use of transport-mode SEAL for a connection 1693 by inserting a 'SEAL_OPTION' TCP option during the connection 1694 establishment phase. If both TCPs agree on the use of SEAL, their 1695 protocol messages will be carried as TCP/SEAL/IPv4 and the connection 1696 will be serviced by the SEAL protocol using TCP (instead of an 1697 encapsulating tunnel endpoint) as the transport layer protocol. The 1698 SEAL protocol for transport mode otherwise observes the same 1699 specifications as for Section 4. 1701 Appendix D. Historic Evolution of PMTUD 1703 The topic of Path MTU discovery (PMTUD) saw a flurry of discussion 1704 and numerous proposals in the late 1980's through early 1990. The 1705 initial problem was posed by Art Berggreen on May 22, 1987 in a 1706 message to the TCP-IP discussion group [TCP-IP]. The discussion that 1707 followed provided significant reference material for [FRAG]. An IETF 1708 Path MTU Discovery Working Group [MTUDWG] was formed in late 1989 1709 with charter to produce an RFC. Several variations on a very few 1710 basic proposals were entertained, including: 1712 1. Routers record the PMTUD estimate in ICMP-like path probe 1713 messages (proposed in [FRAG] and later [RFC1063]) 1715 2. The destination reports any fragmentation that occurs for packets 1716 received with the "RF" (Report Fragmentation) bit set (Steve 1717 Deering's 1989 adaptation of Charles Lynn's Nov. 1987 proposal) 1719 3. A hybrid combination of 1) and Charles Lynn's Nov. 1987 (straw 1720 RFC draft by McCloughrie, Fox and Mogul on Jan 12, 1990) 1722 4. Combination of the Lynn proposal with TCP (Fred Bohle, Jan 30, 1723 1990) 1725 5. Fragmentation avoidance by setting "IP_DF" flag on all packets 1726 and retransmitting if ICMPv4 "fragmentation needed" messages 1727 occur (Geof Cooper's 1987 proposal; later adapted into [RFC1191] 1728 by Mogul and Deering). 1730 Option 1) seemed attractive to the group at the time, since it was 1731 believed that routers would migrate more quickly than hosts. Option 1732 2) was a strong contender, but repeated attempts to secure an "RF" 1733 bit in the IPv4 header from the IESG failed and the proponents became 1734 discouraged. 3) was abandoned because it was perceived as too 1735 complicated, and 4) never received any apparent serious 1736 consideration. Proposal 5) was a late entry into the discussion from 1737 Steve Deering on Feb. 24th, 1990. The discussion group soon 1738 thereafter seemingly lost track of all other proposals and adopted 1739 5), which eventually evolved into [RFC1191] and later [RFC1981]. 1741 In retrospect, the "RF" bit postulated in 2) is not needed if a 1742 "contract" is first established between the peers, as in proposal 4) 1743 and a message to the MTUDWG mailing list from jrd@PTT.LCS.MIT.EDU on 1744 Feb 19. 1990. These proposals saw little discussion or rebuttal, and 1745 were dismissed based on the following the assertions: 1747 o routers upgrade their software faster than hosts 1749 o PCs could not reassemble fragmented packets 1751 o Proteon and Wellfleet routers did not reproduce the "RF" bit 1752 properly in fragmented packets 1754 o Ethernet-FDDI bridges would need to perform fragmentation (i.e., 1755 "translucent" not "transparent" bridging) 1757 o the 16-bit IP_ID field could wrap around and disrupt reassembly at 1758 high packet arrival rates 1760 The first four assertions, although perhaps valid at the time, have 1761 been overcome by historical events. The final assertion is addressed 1762 by the mechanisms specified in SEAL. 1764 Author's Address 1766 Fred L. Templin (editor) 1767 Boeing Research & Technology 1768 P.O. Box 3707 1769 Seattle, WA 98124 1770 USA 1772 Email: fltemplin@acm.org