idnits 2.17.1 draft-ietf-nfsv4-nfs-ulb-v2-06.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 (16 November 2021) is 864 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-07) exists of draft-ietf-nfsv4-rpcrdma-version-two-05 Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network File System Version 4 C. Lever 3 Internet-Draft Oracle 4 Intended status: Standards Track 16 November 2021 5 Expires: 20 May 2022 7 Network File System (NFS) Upper-Layer Binding To RPC-Over-RDMA Version 2 8 draft-ietf-nfsv4-nfs-ulb-v2-06 10 Abstract 12 This document specifies Upper-Layer Bindings of Network File System 13 (NFS) protocol versions to RPC-over-RDMA version 2. 15 Note 17 Discussion of this draft takes place on the NFSv4 working group 18 mailing list (nfsv4@ietf.org), archived at 19 https://mailarchive.ietf.org/arch/browse/nfsv4/. Working Group 20 information is available at https://datatracker.ietf.org/wg/nfsv4/ 21 about/. 23 Submit suggestions and changes as pull requests at 24 https://github.com/chucklever/i-d-nfs-ulb-v2. Instructions are on 25 that page. 27 Status of This Memo 29 This Internet-Draft is submitted in full conformance with the 30 provisions of BCP 78 and BCP 79. 32 Internet-Drafts are working documents of the Internet Engineering 33 Task Force (IETF). Note that other groups may also distribute 34 working documents as Internet-Drafts. The list of current Internet- 35 Drafts is at https://datatracker.ietf.org/drafts/current/. 37 Internet-Drafts are draft documents valid for a maximum of six months 38 and may be updated, replaced, or obsoleted by other documents at any 39 time. It is inappropriate to use Internet-Drafts as reference 40 material or to cite them other than as "work in progress." 42 This Internet-Draft will expire on 20 May 2022. 44 Copyright Notice 46 Copyright (c) 2021 IETF Trust and the persons identified as the 47 document authors. All rights reserved. 49 This document is subject to BCP 78 and the IETF Trust's Legal 50 Provisions Relating to IETF Documents (https://trustee.ietf.org/ 51 license-info) in effect on the date of publication of this document. 52 Please review these documents carefully, as they describe your rights 53 and restrictions with respect to this document. Code Components 54 extracted from this document must include Simplified BSD License text 55 as described in Section 4.e of the Trust Legal Provisions and are 56 provided without warranty as described in the Simplified BSD License. 58 Table of Contents 60 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 61 2. Requirements Language . . . . . . . . . . . . . . . . . . . . 3 62 3. Upper-Layer Binding for NFS Versions 2 and 3 . . . . . . . . 4 63 3.1. Reply Size Estimation . . . . . . . . . . . . . . . . . . 4 64 3.2. RPC Binding Considerations . . . . . . . . . . . . . . . 5 65 3.3. Transport Considerations . . . . . . . . . . . . . . . . 5 66 3.3.1. Keep-Alive . . . . . . . . . . . . . . . . . . . . . 5 67 3.3.2. Replay Detection . . . . . . . . . . . . . . . . . . 6 68 4. Upper-Layer Bindings for NFS Version 2 and 3 Auxiliary 69 Protocols . . . . . . . . . . . . . . . . . . . . . . . . 6 70 4.1. MOUNT, NLM, and NSM Protocols . . . . . . . . . . . . . . 7 71 4.2. NFSACL Protocol . . . . . . . . . . . . . . . . . . . . . 7 72 5. Upper-Layer Binding For NFS Version 4 . . . . . . . . . . . . 7 73 5.1. DDP-Eligibility . . . . . . . . . . . . . . . . . . . . . 7 74 5.1.1. The NFSv4.2 READ_PLUS operation . . . . . . . . . . . 8 75 5.2. Reply Size Estimation . . . . . . . . . . . . . . . . . . 8 76 5.2.1. Reply Size Estimation for Minor Version 0 . . . . . . 9 77 5.2.2. Reply Size Estimation for Minor Version 1 and 78 Newer . . . . . . . . . . . . . . . . . . . . . . . . 9 79 5.3. RPC Binding Considerations . . . . . . . . . . . . . . . 9 80 5.4. NFS COMPOUND Requests . . . . . . . . . . . . . . . . . . 9 81 5.4.1. Multiple DDP-eligible Data Items . . . . . . . . . . 10 82 5.4.2. Chunk List Complexity . . . . . . . . . . . . . . . . 10 83 5.4.3. NFS Version 4 COMPOUND Example . . . . . . . . . . . 11 84 5.5. NFS Callback Requests . . . . . . . . . . . . . . . . . . 11 85 5.5.1. NFS Version 4.0 Callback . . . . . . . . . . . . . . 12 86 5.5.2. NFS Version 4.1 Callback . . . . . . . . . . . . . . 12 87 5.6. Session-Related Considerations . . . . . . . . . . . . . 12 88 5.7. Transport Considerations . . . . . . . . . . . . . . . . 13 89 5.7.1. Congestion Avoidance . . . . . . . . . . . . . . . . 13 90 5.7.2. Retransmission and Keep-alive . . . . . . . . . . . . 14 91 6. Extending NFS Upper-Layer Bindings . . . . . . . . . . . . . 15 92 7. Security Considerations . . . . . . . . . . . . . . . . . . . 15 93 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15 94 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 15 95 9.1. Normative References . . . . . . . . . . . . . . . . . . 15 96 9.2. Informative References . . . . . . . . . . . . . . . . . 16 98 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 17 99 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 17 101 1. Introduction 103 The RPC-over-RDMA version 2 transport can employ direct data 104 placement to convey data payloads associated with RPC transactions, 105 as described in [I-D.ietf-nfsv4-rpcrdma-version-two]. As mandated by 106 that document, RPC client and server implementations using RPC-over- 107 RDMA version 2 MUST agree in advance which XDR data items and RPC 108 procedures are eligible for direct data placement (DDP). 110 An Upper-Layer Binding specifies this agreement for one or more 111 versions of one RPC program. Other operational details, such as RPC 112 binding assignments, pairing Write chunks with result data items, and 113 reply size estimation, are also specified by such a Binding. 115 This document contains material required of Upper-Layer Bindings, as 116 specified in Appendix A of [I-D.ietf-nfsv4-rpcrdma-version-two], for 117 the following NFS protocol versions: 119 * NFS version 2 [RFC1094] 121 * NFS version 3 [RFC1813] 123 * NFS version 4.0 [RFC7530] 125 * NFS version 4.1 [RFC8881] 127 * NFS version 4.2 [RFC7862] 129 The current document also provides Upper-Layer Bindings for auxiliary 130 protocols used with NFS versions 2 and 3 (see Section 4). 132 This document assumes the reader is already familiar with concepts 133 and terminology defined throughout 134 [I-D.ietf-nfsv4-rpcrdma-version-two] and the documents it references. 136 2. Requirements Language 138 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 139 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 140 "OPTIONAL" in this document are to be interpreted as described in 141 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all 142 capitals, as shown here. 144 3. Upper-Layer Binding for NFS Versions 2 and 3 146 The Upper-Layer Binding specification in this section applies to NFS 147 version 2 [RFC1094] and NFS version 3 [RFC1813]. For brevity, in 148 this document, a "Legacy NFS client" refers to an NFS client using 149 version 2 or version 3 of the NFS RPC program (100003) to communicate 150 with an NFS server. Likewise, a "Legacy NFS server" is an NFS server 151 communicating with clients using NFS version 2 or NFS version 3. 153 The following XDR data items in NFS versions 2 and 3 are DDP- 154 eligible: 156 * The opaque file data argument in the NFS WRITE procedure 158 * The pathname argument in the NFS SYMLINK procedure 160 * The opaque file data result in the NFS READ procedure 162 * The pathname result in the NFS READLINK procedure 164 All other argument or result data items in NFS versions 2 and 3 are 165 not DDP-eligible. 167 Regardless of whether an NFS operation is considered non-idempotent, 168 a transport error might not indicate whether the server has processed 169 the arguments of the RPC Call or whether the server has accessed or 170 modified client memory associated with that RPC. 172 3.1. Reply Size Estimation 174 During the construction of each RPC Call message, a Requester is 175 responsible for allocating appropriate RDMA resources to receive the 176 corresponding Reply message. These resources must be capable of 177 holding the entire Reply. Therefore the Requester needs to estimate 178 the maximum possible size of the expected Reply message. 180 * Often, the expected Reply can fit in a limited number of RDMA Send 181 messages. The Requester need not provision any RDMA resources for 182 the Reply, relying instead on message continuation to handle the 183 entire Reply message. 185 * In cases where the Upper Layer Binding permits direct data 186 placement of the results (DDP), a Requester can provision Write 187 chunks to receive those results. The Requester MUST reliably 188 estimate the maximum size of each result receive via a Write 189 chunk. 191 * A Requester that expects a large Reply message can provision a 192 Reply chunk. The Requester MUST reliably estimate the maximum 193 size of the payload received via the Reply chunk. 195 * If RDMA resources are not available to send a Reply, a Responder 196 falls back to message continuation. 198 A correctly implemented Legacy NFS client thus avoids retransmission 199 of non-idempotent NFS requests due to improperly estimated Reply 200 resources. 202 3.2. RPC Binding Considerations 204 Legacy NFS servers typically listen for clients on UDP and TCP port 205 2049. Additionally, they register these ports with a local 206 portmapper service [RFC1833]. 208 A Legacy NFS server supporting RPC-over-RDMA version 2 and 209 registering itself with the RPC portmapper MAY choose an arbitrary 210 port or MAY use the alternative well-known port number for its RPC- 211 over-RDMA service (see Section 8). The chosen port MAY be registered 212 with the RPC portmapper using the netids assigned in Section 12 of 213 [I-D.ietf-nfsv4-rpcrdma-version-two]. 215 3.3. Transport Considerations 217 3.3.1. Keep-Alive 219 Legacy NFS client implementations can rely on connection keep-alive 220 to detect when a Legacy NFS server has become unresponsive. When an 221 NFS server is no longer responsive, client-side keep-alive terminates 222 the connection, triggering reconnection and retransmission of 223 outstanding RPC transactions. 225 Some RDMA transports (such as the Reliable Connected QP type on 226 InfiniBand) have no keep-alive mechanism. Without a disconnect or 227 new RPC traffic, such connections can remain alive long after an NFS 228 server has become unresponsive or unreachable. Once an NFS client 229 has consumed all available RPC-over-RDMA version 2 credits on that 230 transport connection, it awaits a reply indefinitely before sending 231 another RPC request. 233 Legacy NFS clients SHOULD reserve one RPC-over-RDMA version 2 credit 234 to use for periodic server or connection health assessment. Either 235 peer can use this credit to drive an RPC request on an otherwise idle 236 connection, triggering either an affirmative server response or a 237 connection termination. 239 3.3.2. Replay Detection 241 Like NFSv4.0, Legacy NFS servers typically employ request replay 242 detection to reduce the risk of data and file namespace corruption 243 that could result when an NFS client retransmits a non-idempotent NFS 244 request. A Legacy NFS server can send a cached response when a 245 replay is detected, rather than executing the request again. Replay 246 detection is not perfect, but it is usually adequate. 248 For Legacy NFS servers, replay detection commonly utilizes heuristic 249 indicators such as the IP address of the NFS client, the source port 250 of the connection, the transaction ID of the request, and the 251 contents of the request's RPC and upper-layer protocol headers. A 252 Legacy NFS client is careful to re-use the same source port when 253 reconnecting so that Legacy NFS servers can better detect RPC 254 retransmission. 256 However, a Legacy NFS client operating over an RDMA transport has no 257 control over connection source ports. It is almost certain that an 258 RPC request retransmitted on a new connection can never be detected 259 as a replay if the receiving Legacy NFS server includes the 260 connection source port in its replay detection heuristics. 262 Therefore a Legacy NFS server using an RDMA transport should never 263 use a connection's source port as part of its NFS request replay 264 detection mechanism. 266 4. Upper-Layer Bindings for NFS Version 2 and 3 Auxiliary Protocols 268 Storage administrators typically deploy NFS versions 2 and 3 with 269 several other protocols, sometimes called the "NFS auxiliary 270 protocols." These are distinct RPC programs that define procedures 271 not part of the NFS RPC program (100003). The Upper-Layer Bindings 272 in this section apply to: 274 * Versions 2 and 3 of the MOUNT RPC program (100005) [RFC1813] 276 * Versions 1, 3, and 4 of the NLM RPC program (100021) [RFC1813] 278 * Version 1 of the NSM RPC program (100024), described in Chapter 11 279 of [XNFS] 281 * Versions 2 and 3 of the NFSACL RPC program (100227). The NFSACL 282 program does not have a public definition. This document treats 283 the NFSACL program as a de facto standard, as there are several 284 interoperating implementations. 286 4.1. MOUNT, NLM, and NSM Protocols 288 Historically, NFS/RDMA implementations have conveyed the MOUNT, NLM, 289 and NSM protocols via TCP. A Legacy NFS server implementation MUST 290 provide support for these auxiliary protocols via TCP. 292 Moreover, there is little benefit from transporting these protocols 293 via RDMA. Thus this document does not provide an Upper-Layer binding 294 for them. 296 4.2. NFSACL Protocol 298 Legacy NFS clients and servers convey NFSACL procedures on the same 299 transport connection and port as the NFS RPC program (100003). 300 Utilizing the same port obviates the need for a separate rpcbind 301 query to discover server support for this RPC program. 303 ACLs are typically small, but even large ACLs must be encoded and 304 decoded to some degree before being being stored in local 305 filesystems. Thus no data item in this Upper-Layer Protocol is DDP- 306 eligible. 308 For procedures whose replies do not include an ACL object, the size 309 of each Reply is determined directly from the NFSACL RPC program's 310 XDR definition. 312 The NFSACL protocol does not provide a mechanism to determine the 313 size of a received ACL in advance. When preparing for responses that 314 include ACLs, Legacy NFS clients estimate a maximum reply size based 315 on limits within their local file systems. If that estimation is 316 inadequate, a Responder falls back to message continuation. 318 5. Upper-Layer Binding For NFS Version 4 320 The Upper-Layer Binding specification in this section applies to 321 versions of the NFS RPC program defined in NFS version 4.0 [RFC7530], 322 NFS version 4.1 [RFC8881], and NFS version 4.2 [RFC7862]. 324 5.1. DDP-Eligibility 326 Only the following XDR data items in the COMPOUND procedure of all 327 NFS version 4 minor versions are DDP-eligible: 329 * The opaque data field in the WRITE4args structure 331 * The linkdata field of the NF4LNK arm in the createtype4 union 333 * The opaque data field in the READ4resok structure 334 * The linkdata field in the READLINK4resok structure 336 5.1.1. The NFSv4.2 READ_PLUS operation 338 NFS version 4.2 introduces an enhanced READ operation called 339 READ_PLUS [RFC7862]. READ_PLUS enables an NFS server to compact 340 returned READ data payloads. No part of a READ_PLUS Reply is DDP- 341 eligible. 343 In a READ_PLUS result, returned file content appears as a list of one 344 or more of the following items: 346 * Regular data content, the same as the result of a traditional READ 347 operation 349 * Unallocated space in a file, where no data has been written, or 350 previously-written data has been removed via a hole-punch 351 operation 353 * A counted pattern 355 Upon receipt of a READ_PLUS result, an NFSv4.2 client expands the 356 returned list into its preferred representation of the original file 357 content. 359 Before receiving that result, an NFSv4.2 client is unaware of how the 360 NFS server has organized the file content. Thus it is not possible 361 to predict the size or structure of a READ_PLUS Reply in advance. 362 The use of direct data placement is therefore challenging. Moreover, 363 the usual benefits of hardware-assisted data placement are entirely 364 lost if the client must parse the result of each READ I/O. 366 Therefore this Upper Layer Binding does not make elements of an 367 NFSv4.2 READ_PLUS Reply DDP-eligible. Further, this Upper Layer 368 Binding recommends that NFS client implemenations avoid using the 369 READ_PLUS operation on NFS/RDMA mount points. 371 5.2. Reply Size Estimation 373 Within NFS version 4, there are certain variable-length result data 374 items whose maximum size cannot be estimated by clients reliably 375 because there is no protocol-specified size limit on these result 376 arrays. These include: 378 * The attrlist4 field 380 * Fields containing ACLs such as fattr4_acl, fattr4_dacl, and 381 fattr4_sacl 383 * Fields in the fs_locations4 and fs_locations_info4 data structures 385 * Fields which pertain to pNFS layout metadata, such as loc_body, 386 loh_body, da_addr_body, lou_body, lrf_body, fattr_layout_types, 387 and fs_layout_types 389 5.2.1. Reply Size Estimation for Minor Version 0 391 The NFS version 4.0 protocol itself does not impose any bound on the 392 size of NFS Calls or Replies. 394 Variable-length fattr4 attributes make it particularly difficult for 395 clients to predict the maximum size of some NFS version 4.0 Replies. 396 Client implementations might rely upon internal architectural limits 397 to constrain the reply size, but such limits are not always reliable. 398 When an NFS version 4.0 client cannot predict the size of a Reply, it 399 can rely on message continuation to enable a Reply under any 400 circumstances. 402 5.2.2. Reply Size Estimation for Minor Version 1 and Newer 404 In NFS version 4.1 and newer minor versions, the csa_fore_chan_attrs 405 argument of the CREATE_SESSION operation contains a 406 ca_maxresponsesize field. The value in this field is the absolute 407 maximum size of replies generated by an NFS version 4.1 server. 409 An NFS version 4 client can use this value when it is impossible to 410 estimate a reply size upper bound precisely. In practice, objects 411 such as ACLs, named attributes, layout bodies, and security labels 412 are much smaller than this maximum. 414 5.3. RPC Binding Considerations 416 NFS version 4 servers are required to listen on TCP port 2049 and are 417 not required to register with an rpcbind service [RFC7530]. 418 Therefore, an NFS version 4 server supporting RPC-over-RDMA version 2 419 MUST use the alternative well-known port number for its RPC-over-RDMA 420 service defined in Section 8. 422 5.4. NFS COMPOUND Requests 423 5.4.1. Multiple DDP-eligible Data Items 425 An NFS version 4 COMPOUND procedure can contain more than one 426 operation that carries a DDP-eligible data item. An NFS version 4 427 client provides XDR Position values in each Read chunk to determine 428 which chunk is associated with which argument data item. However, 429 NFS version 4 server and client implementations must agree on how to 430 pair Write chunks with returned result data items. 432 A "READ operation" refers to any NFS version 4 operation with a DDP- 433 eligible result data item in the following lists. An NFS version 4 434 client applies the mechanism specified in Section 4.3.2 of 435 [I-D.ietf-nfsv4-rpcrdma-version-two] to this class of operations as 436 follows: 438 * If an NFS version 4 client wishes all DDP-eligible items in an NFS 439 reply to be conveyed inline, it leaves the Write list empty. 441 An NFS version 4 server acts as follows: 443 * The first READ operation MUST use the first chunk in the Write 444 list in an NFS version 4 COMPOUND procedure. The next READ 445 operation uses the next Write chunk, and so on. 447 * If an NFS version 4 client has provided a matching non-empty Write 448 chunk, then the corresponding READ operation MUST return its DDP- 449 eligible data item using that chunk. 451 * If an NFS version 4 client has provided an empty matching Write 452 chunk, then the corresponding READ operation MUST return all of 453 its result data items inline. 455 * If a READ operation returns a union arm which does not contain a 456 DDP-eligible result, and the NFS version 4 client has provided a 457 matching non-empty Write chunk, an NFS version 4 server MUST 458 return an empty Write chunk in that Write list position. 460 * If there are more READ operations than Write chunks, then 461 remaining NFS Read operations in an NFS version 4 COMPOUND that 462 have no matching Write chunk MUST return their results inline. 464 5.4.2. Chunk List Complexity 466 By default, the RPC-over-RDMA version 2 protocol limits the number of 467 chunks or segments that may appear in Read or Write lists (see 468 Section 5.2 of [I-D.ietf-nfsv4-rpcrdma-version-two]). 470 These implementation limits are significant when Kerberos integrity 471 or privacy is in use [RFC7861]. GSS services increase the size of 472 credential material in RPC headers, potentially requiring the more 473 frequent use of less efficient Special Payload or Continued Payload 474 messages. 476 NFS version 4 clients follow the prescriptions listed below when 477 constructing RPC-over-RDMA version 2 messages in the absence of an 478 explicit transport property exchange that alters these limits. NFS 479 version 4 servers MUST accept and process all such requests. 481 * The Read list can contain either a Call chunk, no more than one 482 Read chunk, or both a Call chunk and one Read chunk. 484 * The Write list can contain no more than one Write chunk. 486 NFS version 4 clients wishing to send more complex chunk lists can 487 use transport properties to bound the complexity of NFS version 4 488 COMPOUNDs, limit the number of elements in scatter-gather operations, 489 and avoid other sources of chunk overruns at the receiving peer. 491 5.4.3. NFS Version 4 COMPOUND Example 493 The following example shows a Write list with three Write chunks, A, 494 B, and C. The NFS version 4 server consumes the provided Write 495 chunks by writing the results of the designated operations in the 496 compound request (READ and READLINK) back to each chunk. 498 Write list: 500 A --> B --> C 502 NFS version 4 COMPOUND request: 504 PUTFH LOOKUP READ PUTFH LOOKUP READLINK PUTFH LOOKUP READ 505 | | | 506 v v v 507 A B C 509 If the NFS version 4 client does not want the READLINK result 510 returned via RDMA, it provides an empty Write chunk for buffer B to 511 indicate that the READLINK result must be returned inline. 513 5.5. NFS Callback Requests 515 The NFS version 4 family of protocols supports server-initiated 516 callbacks to notify NFS version 4 clients of events such as recalled 517 delegations. 519 5.5.1. NFS Version 4.0 Callback 521 An NFS version 4.0 client uses the SETCLIENTID operation for 522 advertising the IP address, port, and netid of its NFS version 4.0 523 callback service. When an NFS version 4.0 server provides a 524 backchannel service to an NFS version 4.0 client that uses RPC-over- 525 RDMA version 2 for its forward channel, the server MUST advertise the 526 backchannel service using either the "tcp" or "tcp6" netid. 528 Because the NFSv4.0 backchannel does not operate on RPC-over-RDMA, 529 this document does not specify an Upper-Layer binding for the NFSv4.0 530 backchannel RPC program. 532 5.5.2. NFS Version 4.1 Callback 534 In NFS version 4.1 and newer minor versions, callback operations may 535 appear on the same connection that is in use for NFS version 4 536 forward channel client requests. NFS version 4 clients and servers 537 MUST use the mechanisms described in Section 4.5 of 538 [I-D.ietf-nfsv4-rpcrdma-version-two] to convey backchannel operations 539 on an RPC-over-RDMA version 2 transport. 541 The csa_back_chan_attrs argument of the CREATE_SESSION operation 542 contains a ca_maxresponsesize field. The value in this field is the 543 absolute maximum size of backchannel replies generated by a replying 544 NFS version 4 client. 546 There are no DDP-eligible data items in callback procedures defined 547 in NFS version 4.1 or NFS version 4.2. However, some callback 548 operations, such as messages that convey device ID information, can 549 be sizeable. A sender can use Message Continuation or a Special 550 Payload message in this situation. 552 When an NFS version 4.1 client can support Special Payload Calls in 553 its backchannel, it reports a backchannel ca_maxrequestsize that is 554 larger than the connection's inline thresholds. Otherwise, an NFS 555 version 4 server MUST use only Simple Payload or Continued Payload 556 messages to convey backchannel operations. 558 5.6. Session-Related Considerations 560 The presence of an NFS version 4 session (as defined in [RFC8881]) 561 does not affect the operation of RPC-over-RDMA version 2. None of 562 the operations introduced to support NFS sessions (e.g., the SEQUENCE 563 operation) contain DDP-eligible data items. There is no need to 564 match the number of session slots with the available RPC-over-RDMA 565 version 2 credits. 567 However, there are a few new cases where an RPC transaction can fail. 568 For example, a Requester might receive, in response to an RPC 569 request, an RDMA2_ERROR message with a rdma_err value of 570 RDMA2_ERR_BADXDR. These situations are not different from existing 571 RPC errors, which an NFS session implementation can already handle 572 for other transport types. Moreover, there might be no SEQUENCE 573 result available to the Requester to distinguish whether failure 574 occurred before or after the Responder executed the requested 575 operations. 577 When a transport error occurs (e.g., an RDMA2_ERROR type message is 578 received), the Requester proceeds, as usual, to match the incoming 579 XID value to a waiting RPC Call. The Requester terminates the RPC 580 transaction and reports the result status to the RPC consumer. The 581 Requester's session implementation then determines the session ID and 582 slot for the failed request and performs slot recovery to make that 583 slot usable again. Otherwise, that slot is rendered permanently 584 unavailable. 586 When an NFS session is not present (for example, when NFS version 4.0 587 is in use), a transport error does not indicate whether the server 588 has processed the arguments of the RPC Call, or whether the server 589 has accessed or modified client memory associated with that RPC. 591 5.7. Transport Considerations 593 5.7.1. Congestion Avoidance 595 Section 3.1 of [RFC7530] states: 597 Where an NFS version 4 implementation supports operation over the 598 IP network protocol, the supported transport layer between NFS and 599 IP MUST be an IETF standardized transport protocol that is 600 specified to avoid network congestion; such transports include TCP 601 and the Stream Control Transmission Protocol (SCTP). 603 Section 2.9.1 of [RFC8881] further states: 605 Even if NFS version 4.1 is used over a non-IP network protocol, it 606 is RECOMMENDED that the transport support congestion control. 608 It is permissible for a connectionless transport to be used under 609 NFS version 4.1; however, reliable and in-order delivery of data 610 combined with congestion control by the connectionless transport 611 is REQUIRED. As a consequence, UDP by itself MUST NOT be used as 612 an NFS version 4.1 transport. 614 RPC-over-RDMA version 2 utilizes only reliable, connection-oriented 615 transports that guarantee in-order delivery, meeting all the above 616 requirements for NFS version 4.0 and 4.1. See Section 4.2.1 of 617 [I-D.ietf-nfsv4-rpcrdma-version-two] for more details. 619 5.7.2. Retransmission and Keep-alive 621 NFS version 4 client implementations often rely on a transport-layer 622 connection keep-alive mechanism to detect when an NFS version 4 623 server has become unresponsive. When an NFS server is no longer 624 responsive, client-side keep-alive terminates the connection, 625 triggering reconnection and RPC retransmission. 627 Some RDMA transports (such as the Reliable Connected QP type on 628 InfiniBand) have no keep-alive mechanism. Without a disconnect or 629 new RPC traffic, such connections can remain alive long after an NFS 630 server has become unresponsive. Once an NFS client has consumed all 631 available RPC-over-RDMA version 2 credits on that transport 632 connection, it indefinitely awaits a reply before sending another RPC 633 request. 635 NFS version 4 peers SHOULD reserve one RPC-over-RDMA version 2 credit 636 for periodic server or connection health assessment. Either peer can 637 use this credit to drive an RPC request on an otherwise idle 638 connection, triggering either a quick affirmative server response or 639 immediate connection termination. 641 In addition to network partition and request loss scenarios, RPC- 642 over-RDMA version 2 peers can terminate a connection when a Transport 643 header is malformed or when too many RPC-over-RDMA messages are sent 644 without a credit update. In such cases: 646 * If a transport error occurs (e.g., an RDMA2_ERROR type message is 647 received) just before the disconnect or instead of a disconnect, 648 the Requester MUST respond to that error as prescribed by the 649 specification of the RPC transport. Then the NFS version 4 rules 650 for handling retransmission apply. 652 * If there is a transport disconnect and the Responder has provided 653 no other response for a request, then only the NFS version 4 rules 654 for handling retransmission apply. 656 6. Extending NFS Upper-Layer Bindings 658 RPC programs such as NFS must have an Upper-Layer Binding 659 specification to operate on an RPC-over-RDMA version 2 transport 660 [I-D.ietf-nfsv4-rpcrdma-version-two]. Via standards action, the 661 Upper-Layer Binding specified in this document can be extended to 662 cover versions of the NFS version 4 protocol specified after NFS 663 version 4 minor version 2, or to cover separately published 664 extensions to an existing NFS version 4 minor version, as described 665 in [RFC8178]. 667 7. Security Considerations 669 RPC-over-RDMA version 2 supports all RPC security models, including 670 RPCSEC_GSS security and transport-level security [RFC7861]. The 671 choice of what Direct Data Placement mechanism to convey RPC argument 672 and results does not affect this since it changes only the method of 673 data transfer. Because the current document defines only the binding 674 of the NFS protocols atop RPC-over-RDMA version 2 675 [I-D.ietf-nfsv4-rpcrdma-version-two], all relevant security 676 considerations are, therefore, described at that layer. 678 8. IANA Considerations 680 The use of direct data placement in NFS introduces a need for an 681 additional port number assignment for networks that share traditional 682 UDP and TCP port spaces with RDMA services. The DDP protocol is such 683 an example [RFC5041]. 685 For this purpose, the current document lists a set of port number 686 assignments that IANA has already assigned for NFS/RDMA in the IANA 687 port registry, according to the guidelines described in [RFC6335]. 689 nfsrdma 20049/tcp Network File System (NFS) over RDMA 690 nfsrdma 20049/udp Network File System (NFS) over RDMA 691 nfsrdma 20049/sctp Network File System (NFS) over RDMA 693 The author requests that IANA add the current document as a reference 694 for the existing nfsrdma port assignments. This document does not 695 alter these assignments. 697 9. References 699 9.1. Normative References 701 [I-D.ietf-nfsv4-rpcrdma-version-two] 702 Lever, C. and D. Noveck, "RPC-over-RDMA Version 2 703 Protocol", Work in Progress, Internet-Draft, draft-ietf- 704 nfsv4-rpcrdma-version-two-05, 6 July 2021, 705 . 708 [RFC1833] Srinivasan, R., "Binding Protocols for ONC RPC Version 2", 709 RFC 1833, DOI 10.17487/RFC1833, August 1995, 710 . 712 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 713 Requirement Levels", BCP 14, RFC 2119, 714 DOI 10.17487/RFC2119, March 1997, 715 . 717 [RFC6335] Cotton, M., Eggert, L., Touch, J., Westerlund, M., and S. 718 Cheshire, "Internet Assigned Numbers Authority (IANA) 719 Procedures for the Management of the Service Name and 720 Transport Protocol Port Number Registry", BCP 165, 721 RFC 6335, DOI 10.17487/RFC6335, August 2011, 722 . 724 [RFC7530] Haynes, T., Ed. and D. Noveck, Ed., "Network File System 725 (NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530, 726 March 2015, . 728 [RFC7861] Adamson, A. and N. Williams, "Remote Procedure Call (RPC) 729 Security Version 3", RFC 7861, DOI 10.17487/RFC7861, 730 November 2016, . 732 [RFC7862] Haynes, T., "Network File System (NFS) Version 4 Minor 733 Version 2 Protocol", RFC 7862, DOI 10.17487/RFC7862, 734 November 2016, . 736 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 737 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 738 May 2017, . 740 [RFC8881] Noveck, D., Ed. and C. Lever, "Network File System (NFS) 741 Version 4 Minor Version 1 Protocol", RFC 8881, 742 DOI 10.17487/RFC8881, August 2020, 743 . 745 9.2. Informative References 747 [RFC1094] Nowicki, B., "NFS: Network File System Protocol 748 specification", RFC 1094, DOI 10.17487/RFC1094, March 749 1989, . 751 [RFC1813] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS 752 Version 3 Protocol Specification", RFC 1813, 753 DOI 10.17487/RFC1813, June 1995, 754 . 756 [RFC5041] Shah, H., Pinkerton, J., Recio, R., and P. Culley, "Direct 757 Data Placement over Reliable Transports", RFC 5041, 758 DOI 10.17487/RFC5041, October 2007, 759 . 761 [RFC8178] Noveck, D., "Rules for NFSv4 Extensions and Minor 762 Versions", RFC 8178, DOI 10.17487/RFC8178, July 2017, 763 . 765 [XNFS] The Open Group, "Protocols for Interworking: XNFS, Version 766 3W", January 1998. 768 Acknowledgments 770 Thanks to Tom Talpey, who contributed the text of Section 5.4.2. 771 David Noveck contributed the text of Section 5.6 and Section 6. The 772 author also wishes to thank Bill Baker and Greg Marsden for their 773 support of this work. 775 Special thanks go to Transport Area Directors Zaheduzzaman Sarker, 776 NFSV4 Working Group Chairs Brian Pawlowski, and David Noveck, and 777 NFSV4 Working Group Secretary Thomas Haynes for their support. 779 Author's Address 781 Charles Lever 782 Oracle Corporation 783 United States of America 785 Email: chuck.lever@oracle.com