idnits 2.17.1 draft-ietf-nfsv4-rfc5667bis-12.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 seems to lack the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords -- however, there's a paragraph with a matching beginning. Boilerplate error? (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). == The document seems to contain a disclaimer for pre-RFC5378 work, but was first submitted on or after 10 November 2008. The disclaimer is usually necessary only for documents that revise or obsolete older RFCs, and that take significant amounts of text from those RFCs. If you can contact all authors of the source material and they are willing to grant the BCP78 rights to the IETF Trust, you can and should remove the disclaimer. Otherwise, the disclaimer is needed and you can ignore this comment. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (August 7, 2017) is 2453 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 5661 (Obsoleted by RFC 8881) -- Obsolete informational reference (is this intentional?): RFC 5666 (Obsoleted by RFC 8166) -- Obsolete informational reference (is this intentional?): RFC 5667 (Obsoleted by RFC 8267) Summary: 1 error (**), 0 flaws (~~), 3 warnings (==), 3 comments (--). 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 Obsoletes: 5667 (if approved) August 7, 2017 5 Intended status: Standards Track 6 Expires: February 8, 2018 8 Network File System (NFS) Upper Layer Binding To RPC-Over-RDMA Version 1 9 draft-ietf-nfsv4-rfc5667bis-12 11 Abstract 13 This document specifies Upper Layer Bindings of Network File System 14 (NFS) protocol versions to RPC-over-RDMA version 1, enabling the use 15 of Direct Data Placement. This document obsoletes RFC 5667. 17 Status of This Memo 19 This Internet-Draft is submitted in full conformance with the 20 provisions of BCP 78 and BCP 79. 22 Internet-Drafts are working documents of the Internet Engineering 23 Task Force (IETF). Note that other groups may also distribute 24 working documents as Internet-Drafts. The list of current Internet- 25 Drafts is at http://datatracker.ietf.org/drafts/current/. 27 Internet-Drafts are draft documents valid for a maximum of six months 28 and may be updated, replaced, or obsoleted by other documents at any 29 time. It is inappropriate to use Internet-Drafts as reference 30 material or to cite them other than as "work in progress." 32 This Internet-Draft will expire on February 8, 2018. 34 Copyright Notice 36 Copyright (c) 2017 IETF Trust and the persons identified as the 37 document authors. All rights reserved. 39 This document is subject to BCP 78 and the IETF Trust's Legal 40 Provisions Relating to IETF Documents 41 (http://trustee.ietf.org/license-info) in effect on the date of 42 publication of this document. Please review these documents 43 carefully, as they describe your rights and restrictions with respect 44 to this document. Code Components extracted from this document must 45 include Simplified BSD License text as described in Section 4.e of 46 the Trust Legal Provisions and are provided without warranty as 47 described in the Simplified BSD License. 49 This document may contain material from IETF Documents or IETF 50 Contributions published or made publicly available before November 51 10, 2008. The person(s) controlling the copyright in some of this 52 material may not have granted the IETF Trust the right to allow 53 modifications of such material outside the IETF Standards Process. 54 Without obtaining an adequate license from the person(s) controlling 55 the copyright in such materials, this document may not be modified 56 outside the IETF Standards Process, and derivative works of it may 57 not be created outside the IETF Standards Process, except to format 58 it for publication as an RFC or to translate it into languages other 59 than English. 61 Table of Contents 63 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 64 2. Requirements Language . . . . . . . . . . . . . . . . . . . . 3 65 3. Reply Size Estimation . . . . . . . . . . . . . . . . . . . . 3 66 3.1. Short Reply Chunk Retry . . . . . . . . . . . . . . . . . 4 67 4. Upper Layer Binding for NFS Versions 2 and 3 . . . . . . . . 5 68 4.1. Reply Size Estimation . . . . . . . . . . . . . . . . . . 5 69 4.2. RPC Binding Considerations . . . . . . . . . . . . . . . 5 70 5. Upper Layer Bindings for NFS Version 2 and 3 Auxiliary 71 Protocols . . . . . . . . . . . . . . . . . . . . . . . . . . 6 72 5.1. MOUNT, NLM, and NSM Protocols . . . . . . . . . . . . . . 6 73 5.2. NFSACL Protocol . . . . . . . . . . . . . . . . . . . . . 6 74 6. Upper Layer Binding For NFS Version 4 . . . . . . . . . . . . 7 75 6.1. DDP-Eligibility . . . . . . . . . . . . . . . . . . . . . 7 76 6.2. Reply Size Estimation . . . . . . . . . . . . . . . . . . 7 77 6.3. RPC Binding Considerations . . . . . . . . . . . . . . . 8 78 6.4. NFS COMPOUND Requests . . . . . . . . . . . . . . . . . . 9 79 6.5. NFS Callback Requests . . . . . . . . . . . . . . . . . . 11 80 6.6. Session-Related Considerations . . . . . . . . . . . . . 12 81 6.7. Transport Considerations . . . . . . . . . . . . . . . . 13 82 7. Extending NFS Upper Layer Bindings . . . . . . . . . . . . . 14 83 8. Security Considerations . . . . . . . . . . . . . . . . . . . 14 84 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15 85 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 15 86 10.1. Normative References . . . . . . . . . . . . . . . . . . 15 87 10.2. Informative References . . . . . . . . . . . . . . . . . 16 88 Appendix A. Changes Since RFC 5667 . . . . . . . . . . . . . . . 17 89 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 18 90 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 19 92 1. Introduction 94 The RPC-over-RDMA version 1 transport may employ direct data 95 placement to convey data payloads associated with RPC transactions 96 [RFC8166]. To enable successful interoperation, RPC client and 97 server implementations using RPC-over-RDMA version 1 must agree which 98 XDR data items and RPC procedures are eligible to use direct data 99 placement (DDP). 101 An Upper Layer Binding specifies this agreement for one RPC Program. 102 Other operational details, such as RPC binding assignments, pairing 103 Write chunks with result data items, and reply size estimation, are 104 also specified by this Binding. 106 This document contains material required of Upper Layer Bindings, as 107 specified in [RFC8166]. for the following NFS protocol versions: 109 o NFS version 2 [RFC1094] 111 o NFS version 3 [RFC1813] 113 o NFS version 4.0 [RFC7530] 115 o NFS version 4.1 [RFC5661] 117 o NFS version 4.2 [RFC7862] 119 Upper Layer Bindings are also provided for auxiliary protocols used 120 with NFS versions 2 and 3 (see Section 5). 122 This document assumes the reader is already familiar with concepts 123 and terminology defined in [RFC8166] and the documents it references. 125 2. Requirements Language 127 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 128 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 129 document are to be interpreted as described in BCP 14 [RFC2119] 130 [RFC8174] when, and only when, they appear in all capitals, as shown 131 here. 133 3. Reply Size Estimation 135 During the construction of each RPC Call message, a requester is 136 responsible for allocating appropriate resources for receiving the 137 corresponding Reply message. If the requester expects the RPC Reply 138 message will be larger than its inline threshold, it provides Write 139 and/or Reply chunks wherein the responder can place results and the 140 reply's Payload stream. 142 A reply resource overrun occurs if the RPC Reply Payload stream does 143 not fit into the provided Reply chunk, or no Reply chunk was provided 144 and the Payload stream does not fit inline. This prevents the 145 responder from returning the Upper Layer reply to the requester. 146 Therefore reliable reply size estimation is necessary to ensure 147 successful interoperation. 149 In most cases, the NFS protocol's XDR definition provides enough 150 information to enable an NFS client to predict the maximum size of 151 the expected Reply message. If there are variable-size data items in 152 the result, the maximum size of the RPC Reply message can be 153 estimated as follows: 155 o The client requests only a specific portion of an object (for 156 example, using the "count" and "offset" fields in an NFS READ). 158 o The client limits the number of results (e.g. using the "count" 159 field of an NFS READDIR request). 161 o The client has already cached the size of the whole object it is 162 about to request (say, via a previous NFS GETATTR request). 164 o The client and server have negotiated a maximum size for all calls 165 and responses (using a CREATE_SESSION operation, for instance). 167 3.1. Short Reply Chunk Retry 169 In a few cases, either the size of one or more returned data items or 170 the number of returned data items cannot be known in advance of 171 forming an RPC Call. 173 If an NFS server finds that the NFS client provided inadequate 174 receive resources to return the whole reply, it returns an RPC level 175 error or a transport error, such as ERR_CHUNK. 177 In response to these errors, an NFS client can choose to: 179 o Terminate the RPC transaction immediately with an error, or 181 o Allocate a larger Reply chunk and send the same request as a new 182 RPC transaction (a new XID should be assigned to the retransmitted 183 request to avoid matching a cached RPC Reply that caches the 184 original error). The NFS client should avoid retrying the request 185 indefinitely because a responder may return ERR_CHUNK for a 186 variety of reasons. 188 Subsequent sections of this document discuss exactly which operations 189 might have ultimate difficulty with Reply size estimation. These 190 operations are eligible for "short Reply chunk retry." Unless 191 explicitly mentioned as applicable, short Reply chunk retry should 192 not be used since accurate reply size estimation is problematic in 193 only a few cases. In all other cases reply size underestimation is 194 considered a correctable implementation bug. 196 NFS server implementations can avoid connection loss by first 197 confirming that target RDMA segments are large enough to receive 198 results before initiating explicit RDMA operations. 200 4. Upper Layer Binding for NFS Versions 2 and 3 202 The Upper Layer Binding specification in this section applies to NFS 203 version 2 [RFC1094] and NFS version 3 [RFC1813]. For brevity, in 204 this document a "Legacy NFS client" refers to an NFS client using the 205 NFS version 2 or NFS version 3 RPC Programs (100003) to communicate 206 with an NFS server. Likewise, a "Legacy NFS server" is an NFS server 207 communicating with clients using NFS version 2 or NFS version 3. 209 The following XDR data items in NFS versions 2 and 3 are DDP- 210 eligible: 212 o The opaque file data argument in the NFS WRITE procedure 214 o The pathname argument in the NFS SYMLINK procedure 216 o The opaque file data result in the NFS READ procedure 218 o The pathname result in the NFS READLINK procedure 220 All other argument or result data items in NFS versions 2 and 3 are 221 not DDP-eligible. 223 A transport error does not give an indication of whether the server 224 has processed the arguments of the RPC Call, or whether the server 225 has accessed or modified client memory associated with that RPC. 227 4.1. Reply Size Estimation 229 A Legacy NFS client determines the maximum reply size for each 230 operation using the criteria outlined in Section 3. There are no 231 operations in NFS version 2 or 3 that benefit from short Reply chunk 232 retry. 234 4.2. RPC Binding Considerations 236 Legacy NFS servers traditionally listen for clients on UDP and TCP 237 port 2049. Additionally, they register these ports with a local 238 portmapper [RFC1833] service. 240 A Legacy NFS server supporting RPC-over-RDMA version 1 on such a 241 network and registering itself with the RPC portmapper MAY choose an 242 arbitrary port, or MAY use the alternative well-known port number for 243 its RPC-over-RDMA service (see Section 9). The chosen port MAY be 244 registered with the RPC portmapper under the netids assigned in 245 [RFC8166]. 247 5. Upper Layer Bindings for NFS Version 2 and 3 Auxiliary Protocols 249 NFS versions 2 and 3 are typically deployed with several other 250 protocols, sometimes referred to as "NFS auxiliary protocols." These 251 are distinct RPC Programs that define procedures which are not part 252 of the NFS version 2 or version 3 RPC Programs. The Upper Layer 253 Bindings in this section apply to: 255 o Versions 2 and 3 of the MOUNT protocol [RFC1813] 257 o Versions 1, 3, and 4 of the NLM protocol [RFC1813] 259 o Version 1 of the NSM protocol, described in Chapter 11 of [XNFS] 261 o Version 1 of the NFSACL protocol, which does not have a public 262 definition. NFSACL is treated in this document as a de facto 263 standard, as there are several interoperating implementations. 265 5.1. MOUNT, NLM, and NSM Protocols 267 Historically, NFS/RDMA implementations have chosen to convey the 268 MOUNT, NLM, and NSM protocols via TCP. To enable interoperation of 269 these protocols when NFS/RDMA is in use, a legacy NFS server MUST 270 provide support for these protocols via TCP. 272 5.2. NFSACL Protocol 274 Legacy clients and servers that support the NFSACL RPC Program 275 typically convey NFSACL procedures on the same connection as the NFS 276 RPC program (100003). This obviates the need for separate rpcbind 277 queries to discover server support for this RPC Program. 279 ACLs are typically small, but even large ACLs must be encoded and 280 decoded to some degree. Thus no data item in this Upper Layer 281 Protocol is DDP-eligible. 283 For procedures whose replies do not include an ACL object, the size 284 of a reply is determined directly from the NFSACL RPC Program's XDR 285 definition. 287 There is no protocol-specified size limit for NFS version 3 ACLs, and 288 there is no mechanism in either the NFSACL or NFS RPC Programs for a 289 Legacy client to ascertain the largest ACL a Legacy server can 290 return. Legacy client implementations should choose a maximum size 291 for ACLs based on their own internal limits. 293 Because an NFSACL client cannot know in advance how large a returned 294 ACL will be, it can use short Reply chunk retry when an NFSACL GETACL 295 operation encounters a transport error. 297 6. Upper Layer Binding For NFS Version 4 299 The Upper Layer Binding specification in this section applies to RPC 300 Programs defined in NFS version 4.0 [RFC7530], NFS version 4.1 301 [RFC5661], and NFS version 4.2 [RFC7862]. 303 6.1. DDP-Eligibility 305 Only the following XDR data items in the COMPOUND procedure of all 306 NFS version 4 minor versions are DDP-eligible: 308 o The opaque data field in the WRITE4args structure 310 o The linkdata field of the NF4LNK arm in the createtype4 union 312 o The opaque data field in the READ4resok structure 314 o The linkdata field in the READLINK4resok structure 316 6.2. Reply Size Estimation 318 Within NFS version 4, there are certain variable-length result data 319 items whose maximum size cannot be estimated by clients reliably 320 because there is no protocol-specified size limit on these arrays. 321 These include: 323 o The attrlist4 field 325 o Fields containing ACLs such as fattr4_acl, fattr4_dacl, 326 fattr4_sacl 328 o Fields in the fs_locations4 and fs_locations_info4 data structures 330 o Fields opaque to the NFS version 4 protocol which pertain to pNFS 331 layout metadata, such as loc_body, loh_body, da_addr_body, 332 lou_body, lrf_body, fattr_layout_types and fs_layout_types, 334 6.2.1. Reply Size Estimation for Minor Version 0 336 The NFS version 4.0 protocol itself does not impose any bound on the 337 size of NFS calls or responses. 339 Some of the data items enumerated in Section 6.2 (in particular, the 340 items related to ACLs and fs_locations) make it difficult to predict 341 the maximum size of NFS version 4.0 replies that interrogate 342 variable-length fattr4 attributes. Client implementations might rely 343 on their own internal architectural limits to constrain the reply 344 size, but such limits are not always guaranteed to be reliable. 346 When an especially large fattr4 result is expected, a Reply chunk 347 might be required. An NFS version 4.0 client can use short Reply 348 chunk retry when an NFS COMPOUND containing a GETATTR operation 349 encounters a transport error. 351 The use of NFS COMPOUND operations raises the possibility of requests 352 that combine a non-idempotent operation (e.g. RENAME) with a GETATTR 353 operation that requests one or more variable-length results. This 354 combination should be avoided by ensuring that any GETATTR operation 355 that requests a result of unpredictable length is sent in an NFS 356 COMPOUND by itself. 358 6.2.2. Reply Size Estimation for Minor Version 1 and Newer 360 In NFS version 4.1 and newer minor versions, the csa_fore_chan_attrs 361 argument of the CREATE_SESSION operation contains a 362 ca_maxresponsesize field. The value in this field can be taken as 363 the absolute maximum size of replies generated by an NFS version 4.1 364 server. 366 This value can be used in cases where it is not possible to estimate 367 a reply size upper bound precisely. In practice, objects such as 368 ACLs, named attributes, layout bodies, and security labels are much 369 smaller than this maximum. 371 6.3. RPC Binding Considerations 373 NFS version 4 servers are required to listen on TCP port 2049, and 374 they are not required to register with an rpcbind service [RFC7530]. 376 Therefore, an NFS version 4 server supporting RPC-over-RDMA version 1 377 MUST use the alternative well-known port number for its RPC-over-RDMA 378 service (see Section 9). Clients SHOULD connect to this well-known 379 port without consulting the RPC portmapper (as for NFS version 4 on 380 TCP transports). 382 6.4. NFS COMPOUND Requests 384 6.4.1. Multiple DDP-eligible Data Items 386 An NFS version 4 COMPOUND procedure can contain more than one 387 operation that carries a DDP-eligible data item. An NFS version 4 388 client provides XDR Position values in each Read chunk to 389 disambiguate which chunk is associated with which argument data item. 390 However NFS version 4 server and client implementations must agree in 391 advance on how to pair Write chunks with returned result data items. 393 In the following list, a "READ operation" refers to any NFS version 4 394 operation which has a DDP-eligible result data item. The mechanism 395 specified in Section 4.3.2 of [RFC8166] is applied to this class of 396 operations: 398 o If an NFS version 4 client wishes all DDP-eligible items in an NFS 399 reply to be conveyed inline, it leaves the Write list empty. 401 o The first chunk in the Write list MUST be used by the first READ 402 operation in an NFS version 4 COMPOUND procedure. The next Write 403 chunk is used by the next READ operation, and so on. 405 o If an NFS version 4 client has provided a matching non-empty Write 406 chunk, then the corresponding READ operation MUST return its DDP- 407 eligible data item using that chunk. 409 o If an NFS version 4 client has provided an empty matching Write 410 chunk, then the corresponding READ operation MUST return all of 411 its result data items inline. 413 o If a READ operation returns a union arm which does not contain a 414 DDP-eligible result, and the NFS version 4 client has provided a 415 matching non-empty Write chunk, an NFS version 4 server MUST 416 return an empty Write chunk in that Write list position. 418 o If there are more READ operations than Write chunks, then 419 remaining NFS Read operations in an NFS version 4 COMPOUND that 420 have no matching Write chunk MUST return their results inline. 422 6.4.2. Chunk List Complexity 424 The RPC-over-RDMA version 1 protocol does not place any limit on the 425 number of chunks or segments that may appear in Read or Write lists. 426 However, for various reasons NFS version 4 server implementations 427 often have practical limits on the number of chunks or segments they 428 are prepared to process in a single RPC transaction conveyed via RPC- 429 over-RDMA version 1. 431 These implementation limits are especially important when Kerberos 432 integrity or privacy is in use [RFC7861]. GSS services increase the 433 size of credential material in RPC headers, potentially requiring 434 more frequent use of Long messages. This can increase the complexity 435 of chunk lists independent of the NFS version 4 COMPOUND being 436 conveyed. 438 In the absence of explicit knowledge of the server's limits, NFS 439 version 4 clients SHOULD follow the prescriptions listed below when 440 constructing RPC-over-RDMA version 1 messages. NFS version 4 servers 441 MUST accept and process such requests. 443 o The Read list can contain either a Position-Zero Read chunk, one 444 Read chunk with a non-zero Position, or both. 446 o The Write list can contain no more than one Write chunk. 448 o Any chunk can contain up to sixteen RDMA segments. 450 NFS version 4 clients wishing to send more complex chunk lists can 451 provide configuration interfaces to bound the complexity of NFS 452 version 4 COMPOUNDs, limit the number of elements in scatter-gather 453 operations, and avoid other sources of chunk overruns at the 454 receiving peer. 456 An NFS version 4 server SHOULD return one of the following responses 457 to a client that has sent an RPC transaction via RPC-over-RDMA 458 version 1 which cannot be processed due to chunk list complexity 459 limits on the server: 461 o A problem is detected by the transport layer while parsing the 462 transport header in an RPC Call message. The server responds with 463 an RDMA_ERROR message with the err field set to ERR_CHUNK. 465 o A problem is detected during XDR decoding of the RPC Call message 466 while the RPC layer reassembles the call's XDR stream. The server 467 responds with an RPC reply with its "reply_stat" field set to 468 MSG_ACCEPTED and its "accept_stat" field set to GARBAGE_ARGS. 470 After receiving one of these errors, an NFS version 4 client SHOULD 471 NOT retransmit the failing request, as the result would be the same 472 error. It SHOULD immediately terminate the RPC transaction 473 associated with the XID in the reply. 475 6.4.3. NFS Version 4 COMPOUND Example 477 The following example shows a Write list with three Write chunks, A, 478 B, and C. The NFS version 4 server consumes the provided Write 479 chunks by writing the results of the designated operations in the 480 compound request (READ and READLINK) back to each chunk. 482 Write list: 484 A --> B --> C 486 NFS version 4 COMPOUND request: 488 PUTFH LOOKUP READ PUTFH LOOKUP READLINK PUTFH LOOKUP READ 489 | | | 490 v v v 491 A B C 493 If the NFS version 4 client does not want to have the READLINK result 494 returned via RDMA, it provides an empty Write chunk for buffer B to 495 indicate that the READLINK result must be returned inline. 497 6.5. NFS Callback Requests 499 The NFS version 4 family of protocols support server-initiated 500 callbacks to notify NFS version 4 clients of events such as recalled 501 delegations. 503 6.5.1. NFS Version 4.0 Callback 505 NFS version 4.0 implementations typically employ a separate TCP 506 connection to handle callback operations, even when the forward 507 channel uses an RPC-over-RDMA version 1 transport. 509 No operation in the NFS version 4.0 callback RPC Program conveys a 510 significant data payload. Therefore, no XDR data items in this RPC 511 Program is DDP-eligible. 513 A CB_RECALL reply is small and fixed in size. The CB_GETATTR reply 514 contains a variable-length fattr4 data item. See Section 6.2.1 for a 515 discussion of reply size prediction for this data item. 517 An NFS version 4.0 client advertises netids and ad hoc port addresses 518 for contacting its NFS version 4.0 callback service using the 519 SETCLIENTID operation. 521 6.5.2. NFS Version 4.1 Callback 523 In NFS version 4.1 and newer minor versions, callback operations may 524 appear on the same connection as is used for NFS version 4 forward 525 channel client requests. NFS version 4 clients and servers MUST use 526 the approach described in [RFC8167] when backchannel operations are 527 conveyed on RPC-over-RDMA version 1 transports. 529 The csa_back_chan_attrs argument of the CREATE_SESSION operation 530 contains a ca_maxresponsesize field. The value in this field can be 531 taken as the absolute maximum size of backchannel replies generated 532 by a replying NFS version 4 client. 534 There are no DDP-eligible data items in callback procedures defined 535 in NFS version 4.1 or NFS version 4.2. However, some callback 536 operations, such as messages that convey device ID information, can 537 be large, in which case a Long Call or Reply might be required. 539 When an NFS version 4.1 client can support Long Calls in its 540 backchannel, it reports a backchannel ca_maxrequestsize that is 541 larger than the connection's inline thresholds. Otherwise an NFS 542 version 4 server MUST use only Short messages to convey backchannel 543 operations. 545 6.6. Session-Related Considerations 547 The presence of an NFS session (defined in [RFC5661]) has no effect 548 on the operation of RPC-over-RDMA version 1. None of the operations 549 introduced to support NFS sessions (e.g. the SEQUENCE operation) 550 contain DDP-eligible data items. There is no need to match the 551 number of session slots with the number of available RPC-over-RDMA 552 credits. 554 However, there are a few new cases where an RPC transaction can fail. 555 For example, a requester might receive, in response to an RPC 556 request, an RDMA_ERROR message with an rdma_err value of ERR_CHUNK. 557 These situations are not different from existing RPC errors which an 558 NFS session implementation is already prepared to handle for other 559 transports. And as with other transports during such a failure, 560 there might be no SEQUENCE result available to the requester to 561 distinguish whether failure occurred before or after the requested 562 operations were executed on the responder. 564 When a transport error occurs (e.g. RDMA_ERROR), the requester 565 proceeds as usual to match the incoming XID value to a waiting RPC 566 Call. The RPC transaction is terminated, and the result status is 567 reported to the Upper Layer Protocol. The requester's session 568 implementation then determines the session ID and slot for the failed 569 request, and performs slot recovery to make that slot usable again. 570 If this were not done, that slot could be rendered permanently 571 unavailable. 573 When an NFS session is not present (for example, when NFS version 4.0 574 is in use), a transport error does not provide an indication of 575 whether the server has processed the arguments of the RPC Call, or 576 whether the server has accessed or modified client memory associated 577 with that RPC. 579 6.7. Transport Considerations 581 6.7.1. Congestion Avoidance 583 Section 3.1 of [RFC7530] states: 585 Where an NFS version 4 implementation supports operation over the 586 IP network protocol, the supported transport layer between NFS and 587 IP MUST be an IETF standardized transport protocol that is 588 specified to avoid network congestion; such transports include TCP 589 and the Stream Control Transmission Protocol (SCTP). 591 Section 2.9.1 of [RFC5661] also states: 593 Even if NFS version 4.1 is used over a non-IP network protocol, it 594 is RECOMMENDED that the transport support congestion control. 596 It is permissible for a connectionless transport to be used under 597 NFS version 4.1; however, reliable and in-order delivery of data 598 combined with congestion control by the connectionless transport 599 is REQUIRED. As a consequence, UDP by itself MUST NOT be used as 600 an NFS version 4.1 transport. 602 RPC-over-RDMA version 1 is constructed on a platform of RDMA Reliable 603 Connections [RFC8166] [RFC5041]. RDMA Reliable Connections are 604 reliable, connection-oriented transports that guarantee in-order 605 delivery, meeting all above requirements for NFS version 4 606 transports. 608 6.7.2. Retransmission and Keep-alive 610 NFS version 4 client implementations often rely on a transport-layer 611 keep-alive mechanism to detect when an NFS version 4 server has 612 become unresponsive. When an NFS server is no longer responsive, 613 client-side keep-alive terminates the connection, which in turn 614 triggers reconnection and RPC retransmission. 616 Some RDMA transports (such as Reliable Connections on InfiniBand) 617 have no keep-alive mechanism. Without a disconnect or new RPC 618 traffic, such connections can remain alive long after an NFS server 619 has become unresponsive. Once an NFS client has consumed all 620 available RPC-over-RDMA credits on that transport connection, it will 621 forever await a reply before sending another RPC request. 623 NFS version 4 clients SHOULD reserve one RPC-over-RDMA credit to use 624 for periodic server or connection health assessment. This credit can 625 be used to drive an RPC request on an otherwise idle connection, 626 triggering either a quick affirmative server response or immediate 627 connection termination. 629 In addition to network partition and request loss scenarios, RPC- 630 over-RDMA transport connections can be terminated when a Transport 631 header is malformed, Reply messages are larger than receive 632 resources, or when too many RPC-over-RDMA messages are sent at once. 633 In such cases: 635 o If there is a transport error indicated (ie, RDMA_ERROR) before 636 the disconnect or instead of a disconnect, the requester MUST 637 respond to that error as prescribed by the specification of the 638 RPC transport. Then the NFS version 4 rules for handling 639 retransmission apply. 641 o If there is a transport disconnect and the responder has provided 642 no other response for a request, then only the NFS version 4 rules 643 for handling retransmission apply. 645 7. Extending NFS Upper Layer Bindings 647 RPC Programs such as NFS are required to have an Upper Layer Binding 648 specification to interoperate on RPC-over-RDMA version 1 transports 649 [RFC8166]. Via standards action, the Upper Layer Binding specified 650 in this document can be extended to cover versions of the NFS version 651 4 protocol specified after NFS version 4 minor version 2, or 652 separately published extensions to an existing NFS version 4 minor 653 version, as described in [RFC8178]. 655 8. Security Considerations 657 RPC-over-RDMA version 1 supports all RPC security models, including 658 RPCSEC_GSS security and transport-level security [RFC7861]. The 659 choice of what Direct Data Placement mechanism to convey RPC argument 660 and results does not affect this, since it changes only the method of 661 data transfer. Specifically, the requirements of [RFC8166] ensure 662 that this choice does not introduce new vulnerabilities. 664 Because this document defines only the binding of the NFS protocols 665 atop [RFC8166], all relevant security considerations are therefore to 666 be described at that layer. 668 9. IANA Considerations 670 The use of direct data placement in NFS introduces a need for an 671 additional port number assignment for networks that share traditional 672 UDP and TCP port spaces with RDMA services. The iWARP protocol is 673 such an example [RFC5041] [RFC5040]. 675 For this purpose, a set of transport protocol port number assignments 676 is specified by this document. IANA has assigned the following ports 677 for NFS/RDMA in the IANA port registry, according to the guidelines 678 described in [RFC6335]. 680 nfsrdma 20049/tcp Network File System (NFS) over RDMA 681 nfsrdma 20049/udp Network File System (NFS) over RDMA 682 nfsrdma 20049/sctp Network File System (NFS) over RDMA 684 This document should be listed as the reference for the nfsrdma port 685 assignments. This document does not alter these assignments. 687 10. References 689 10.1. Normative References 691 [RFC1833] Srinivasan, R., "Binding Protocols for ONC RPC Version 2", 692 RFC 1833, DOI 10.17487/RFC1833, August 1995, 693 . 695 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 696 Requirement Levels", BCP 14, RFC 2119, 697 DOI 10.17487/RFC2119, March 1997, 698 . 700 [RFC5661] Shepler, S., Ed., Eisler, M., Ed., and D. Noveck, Ed., 701 "Network File System (NFS) Version 4 Minor Version 1 702 Protocol", RFC 5661, DOI 10.17487/RFC5661, January 2010, 703 . 705 [RFC6335] Cotton, M., Eggert, L., Touch, J., Westerlund, M., and S. 706 Cheshire, "Internet Assigned Numbers Authority (IANA) 707 Procedures for the Management of the Service Name and 708 Transport Protocol Port Number Registry", BCP 165, 709 RFC 6335, DOI 10.17487/RFC6335, August 2011, 710 . 712 [RFC7530] Haynes, T., Ed. and D. Noveck, Ed., "Network File System 713 (NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530, 714 March 2015, . 716 [RFC7861] Adamson, A. and N. Williams, "Remote Procedure Call (RPC) 717 Security Version 3", RFC 7861, DOI 10.17487/RFC7861, 718 November 2016, . 720 [RFC7862] Haynes, T., "Network File System (NFS) Version 4 Minor 721 Version 2 Protocol", RFC 7862, DOI 10.17487/RFC7862, 722 November 2016, . 724 [RFC8166] Lever, C., Ed., Simpson, W., and T. Talpey, "Remote Direct 725 Memory Access Transport for Remote Procedure Call Version 726 1", RFC 8166, DOI 10.17487/RFC8166, June 2017, 727 . 729 [RFC8167] Lever, C., "Bidirectional Remote Procedure Call on RPC- 730 over-RDMA Transports", RFC 8167, DOI 10.17487/RFC8167, 731 June 2017, . 733 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 734 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 735 May 2017, . 737 10.2. Informative References 739 [RFC1094] Nowicki, B., "NFS: Network File System Protocol 740 specification", RFC 1094, DOI 10.17487/RFC1094, March 741 1989, . 743 [RFC1813] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS 744 Version 3 Protocol Specification", RFC 1813, 745 DOI 10.17487/RFC1813, June 1995, 746 . 748 [RFC5040] Recio, R., Metzler, B., Culley, P., Hilland, J., and D. 749 Garcia, "A Remote Direct Memory Access Protocol 750 Specification", RFC 5040, DOI 10.17487/RFC5040, October 751 2007, . 753 [RFC5041] Shah, H., Pinkerton, J., Recio, R., and P. Culley, "Direct 754 Data Placement over Reliable Transports", RFC 5041, 755 DOI 10.17487/RFC5041, October 2007, 756 . 758 [RFC5666] Talpey, T. and B. Callaghan, "Remote Direct Memory Access 759 Transport for Remote Procedure Call", RFC 5666, 760 DOI 10.17487/RFC5666, January 2010, 761 . 763 [RFC5667] Talpey, T. and B. Callaghan, "Network File System (NFS) 764 Direct Data Placement", RFC 5667, DOI 10.17487/RFC5667, 765 January 2010, . 767 [RFC8178] Noveck, D., "Rules for NFSv4 Extensions and Minor 768 Versions", RFC 8178, DOI 10.17487/RFC8178, July 2017, 769 . 771 [XNFS] The Open Group, "Protocols for Interworking: XNFS, Version 772 3W", February 1998. 774 Appendix A. Changes Since RFC 5667 776 Corrections and updates made necessary by new language in [RFC8166] 777 have been introduced. For example, references to deprecated features 778 of RPC-over-RDMA version 1, such as RDMA_MSGP, and the use of the 779 Read list for handling RPC replies, have been removed. The term 780 "mapping" has been replaced with the term "binding" or "Upper Layer 781 Binding" throughout the document. Material that duplicates what is 782 in [RFC8166] has been deleted. 784 Material required by [RFC8166] for Upper Layer Bindings that was not 785 present in [RFC5667] has been added. A complete discussion of reply 786 size estimation has been introduced for all protocols covered by the 787 Upper Layer Bindings in this document. 789 Technical corrections have been made. For example, the mention of 790 12KB and 36KB inline thresholds have been removed. The reference to 791 a non-existant NFS version 4 SYMLINK operation has been replaced. 793 The discussion of NFS version 4 COMPOUND handling has been completed. 794 Some changes were made to the algorithm for matching DDP-eligible 795 results to Write chunks. 797 Requirements to ignore extra Read or Write chunks have been removed 798 from the NFS version 2 and 3 Upper Layer Binding, as they conflict 799 with [RFC8166]. 801 A section discussing NFS version 4 retransmission and connection loss 802 has been added. 804 The following additional improvements have been made, relative to 805 [RFC5667]: 807 o An explicit discussion of NFS version 4.0 and NFS version 4.1 808 backchannel operation has replaced the previous treatment of 809 callback operations. 811 o A section describing considerations when an NFS session is in use 812 has been added. 814 o An Upper Layer Binding for NFS version 4.2 has been added. 816 o A section suggesting a mechanism for periodically assessing 817 connection health has been introduced. 819 o Ambiguous or erroneous uses of RFC2119 terms have been corrected. 821 o References to obsolete RFCs have been updated. 823 o An IANA Considerations Section has been added, which specifies the 824 port assignments for NFS/RDMA. This replaces the example 825 assignment that appeared in [RFC5666]. 827 o Code excerpts have been removed, and figures have been modernized. 829 Acknowledgments 831 The author gratefully acknowledges the work of Brent Callaghan and 832 Tom Talpey on the original NFS Direct Data Placement specification 833 [RFC5667]. Tom contributed the text of Section 6.4.2. 835 Dave Noveck provided excellent review, constructive suggestions, and 836 consistent navigational guidance throughout the process of drafting 837 this document. Dave contributed the text of Section 6.6 and 838 Section 7, and insisted on precise discussion of reply size 839 estimation. 841 Thanks to Karen Deitke for her sharp observations about idempotency, 842 NFS COMPOUNDs, and NFS sessions. 844 Special thanks go to Transport Area Director Spencer Dawkins, NFSV4 845 Working Group Chair and Document Shepherd Spencer Shepler, and NFSV4 846 Working Group Secretary Thomas Haynes for their support. The author 847 also wishes to thank Bill Baker and Greg Marsden for their support of 848 this work. 850 Author's Address 852 Charles Lever 853 Oracle Corporation 854 1015 Granger Avenue 855 Ann Arbor, MI 48104 856 United States of America 858 Phone: +1 248 816 6463 859 Email: chuck.lever@oracle.com