idnits 2.17.1 draft-ietf-nfsv4-rfc5667bis-11.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 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 (May 8, 2017) is 2543 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references to lower-maturity documents in RFCs) ** Obsolete normative reference: RFC 5661 (Obsoleted by RFC 8881) == Outdated reference: A later version (-11) exists of draft-ietf-nfsv4-versioning-09 -- Obsolete informational reference (is this intentional?): RFC 5666 (Obsoleted by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, Ed. 3 Internet-Draft Oracle 4 Obsoletes: 5667 (if approved) May 8, 2017 5 Intended status: Standards Track 6 Expires: November 9, 2017 8 Network File System (NFS) Upper Layer Binding To RPC-Over-RDMA Version 9 One 10 draft-ietf-nfsv4-rfc5667bis-11 12 Abstract 14 This document specifies Upper Layer Bindings of Network File System 15 (NFS) protocol versions to RPC-over-RDMA Version One, enabling the 16 use of Direct Data Placement. This document obsoletes RFC 5667. 18 Requirements Language 20 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 21 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 22 document are to be interpreted as described in [RFC2119]. 24 Status of This Memo 26 This Internet-Draft is submitted in full conformance with the 27 provisions of BCP 78 and BCP 79. 29 Internet-Drafts are working documents of the Internet Engineering 30 Task Force (IETF). Note that other groups may also distribute 31 working documents as Internet-Drafts. The list of current Internet- 32 Drafts is at http://datatracker.ietf.org/drafts/current/. 34 Internet-Drafts are draft documents valid for a maximum of six months 35 and may be updated, replaced, or obsoleted by other documents at any 36 time. It is inappropriate to use Internet-Drafts as reference 37 material or to cite them other than as "work in progress." 39 This Internet-Draft will expire on November 9, 2017. 41 Copyright Notice 43 Copyright (c) 2017 IETF Trust and the persons identified as the 44 document authors. All rights reserved. 46 This document is subject to BCP 78 and the IETF Trust's Legal 47 Provisions Relating to IETF Documents 48 (http://trustee.ietf.org/license-info) in effect on the date of 49 publication of this document. Please review these documents 50 carefully, as they describe your rights and restrictions with respect 51 to this document. Code Components extracted from this document must 52 include Simplified BSD License text as described in Section 4.e of 53 the Trust Legal Provisions and are provided without warranty as 54 described in the Simplified BSD License. 56 This document may contain material from IETF Documents or IETF 57 Contributions published or made publicly available before November 58 10, 2008. The person(s) controlling the copyright in some of this 59 material may not have granted the IETF Trust the right to allow 60 modifications of such material outside the IETF Standards Process. 61 Without obtaining an adequate license from the person(s) controlling 62 the copyright in such materials, this document may not be modified 63 outside the IETF Standards Process, and derivative works of it may 64 not be created outside the IETF Standards Process, except to format 65 it for publication as an RFC or to translate it into languages other 66 than English. 68 Table of Contents 70 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 71 2. Reply Size Estimation . . . . . . . . . . . . . . . . . . . . 3 72 2.1. Short Reply Chunk Retry . . . . . . . . . . . . . . . . . 4 73 3. Upper Layer Binding for NFS Versions 2 and 3 . . . . . . . . 5 74 3.1. Reply Size Estimation . . . . . . . . . . . . . . . . . . 5 75 3.2. RPC Binding Considerations . . . . . . . . . . . . . . . 5 76 4. Upper Layer Bindings for NFS Version 2 and 3 Auxiliary 77 Protocols . . . . . . . . . . . . . . . . . . . . . . . . . . 6 78 4.1. MOUNT, NLM, and NSM Protocols . . . . . . . . . . . . . . 6 79 4.2. NFSACL Protocol . . . . . . . . . . . . . . . . . . . . . 6 80 5. Upper Layer Binding For NFS Version 4 . . . . . . . . . . . . 7 81 5.1. DDP-Eligibility . . . . . . . . . . . . . . . . . . . . . 7 82 5.2. Reply Size Estimation . . . . . . . . . . . . . . . . . . 7 83 5.3. RPC Binding Considerations . . . . . . . . . . . . . . . 8 84 5.4. NFS COMPOUND Requests . . . . . . . . . . . . . . . . . . 8 85 5.5. NFS Callback Requests . . . . . . . . . . . . . . . . . . 11 86 5.6. Session-Related Considerations . . . . . . . . . . . . . 12 87 5.7. Transport Considerations . . . . . . . . . . . . . . . . 13 88 6. Extending NFS Upper Layer Bindings . . . . . . . . . . . . . 14 89 7. Security Considerations . . . . . . . . . . . . . . . . . . . 14 90 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 14 91 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 15 92 9.1. Normative References . . . . . . . . . . . . . . . . . . 15 93 9.2. Informative References . . . . . . . . . . . . . . . . . 16 94 Appendix A. Changes Since RFC 5667 . . . . . . . . . . . . . . . 17 95 Appendix B. Acknowledgments . . . . . . . . . . . . . . . . . . 18 96 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 18 98 1. Introduction 100 The RPC-over-RDMA Version One transport may employ direct data 101 placement to convey data payloads associated with RPC transactions 102 [I-D.ietf-nfsv4-rfc5666bis]. To enable successful interoperation, 103 RPC client and server implementations using RPC-over-RDMA Version One 104 must agree which XDR data items and RPC procedures are eligible to 105 use direct data placement (DDP). 107 An Upper Layer Binding specifies this agreement for one RPC Program. 108 Other operational details, such as RPC binding assignments, pairing 109 Write chunks with result data items, and reply size estimation, are 110 also specified by this Binding. 112 This document contains material required of Upper Layer Bindings, as 113 specified in [I-D.ietf-nfsv4-rfc5666bis], for the following NFS 114 protocol versions: 116 o NFS Version 2 [RFC1094] 118 o NFS Version 3 [RFC1813] 120 o NFS Version 4.0 [RFC7530] 122 o NFS Version 4.1 [RFC5661] 124 o NFS Version 4.2 [RFC7862] 126 Upper Layer Bindings are also provided for auxiliary protocols used 127 with NFS versions 2 and 3. 129 This document assumes the reader is already familiar with concepts 130 and terminology defined in [I-D.ietf-nfsv4-rfc5666bis] and the 131 documents it references. 133 2. 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. 147 Therefore reliable reply size estimation is necessary to ensure 148 successful interoperation. 150 In most cases, the NFS protocol's XDR definition provides enough 151 information to enable an NFS client to predict the maximum size of 152 the expected Reply message. If there are variable-size data items in 153 the result, the maximum size of the RPC Reply message can be 154 estimated as follows: 156 o The client requests only a specific portion of an object (for 157 example, using the "count" and "offset" fields in an NFS READ). 159 o The client limits the number of results (e.g. using the "count" 160 field of an NFS READDIR request). 162 o The client has already cached the size of the whole object it is 163 about to request (say, via a previous NFS GETATTR request). 165 o The client and server have negotiated a maximum size for all calls 166 and responses (using a CREATE_SESSION operation, for instance). 168 2.1. Short Reply Chunk Retry 170 In a few cases, either the size of one or more returned data items or 171 the number of returned data items cannot be known in advance of 172 forming an RPC Call. 174 If an NFS server finds that the NFS client provided inadequate 175 receive resources to return the whole reply, it returns an RPC level 176 error or a transport error, such as ERR_CHUNK. 178 In response to these errors, an NFS client can choose to: 180 o Terminate the RPC transaction immediately with an error, or 182 o Allocate a larger Reply chunk and send the same request as a new 183 RPC transaction (to avoid hitting in a Duplicate Reply Cache). 184 The NFS client should avoid retrying the request indefinitely 185 because a responder may return ERR_CHUNK for a variety of reasons. 187 Subsequent sections of this document discuss exactly which operations 188 might have ultimate difficulty with Reply size estimation. These 189 operations are eligible for "short Reply chunk retry." Unless 190 explicitly mentioned as applicable, short Reply chunk retry should 191 not be used. 193 NFS server implementations can avoid connection loss by first 194 confirming that target RDMA segments are large enough to receive 195 results before initiating explicit RDMA operations. 197 3. Upper Layer Binding for NFS Versions 2 and 3 199 The Upper Layer Binding specification in this section applies to NFS 200 Version 2 [RFC1094] and NFS Version 3 [RFC1813]. For brevity, in 201 this document a "Legacy NFS client" refers to an NFS client using the 202 NFS version 2 or NFS version 3 RPC Programs (100003) to communicate 203 with an NFS server. Likewise, a "Legacy NFS server" is an NFS server 204 communicating with clients using NFS version 2 or NFS version 3. 206 The following XDR data items in NFS versions 2 and 3 are DDP- 207 eligible: 209 o The opaque file data argument in the NFS WRITE procedure 211 o The pathname argument in the NFS SYMLINK procedure 213 o The opaque file data result in the NFS READ procedure 215 o The pathname result in the NFS READLINK procedure 217 All other argument or result data items in NFS versions 2 and 3 are 218 not DDP-eligible. 220 A transport error does not give an indication of whether the server 221 has processed the arguments of the RPC Call, or whether the server 222 has accessed or modified client memory associated with that RPC. 224 3.1. Reply Size Estimation 226 A Legacy NFS client determines the maximum reply size for each 227 operation using the criteria outlined in Section 2. There are no 228 operations in NFS version 2 or 3 that benefit from short Reply chunk 229 retry. 231 3.2. RPC Binding Considerations 233 Legacy NFS servers traditionally listen for clients on UDP and TCP 234 port 2049. Additionally, they register these ports with a local 235 portmapper [RFC1833] service. 237 A Legacy NFS server supporting RPC-over-RDMA Version One on such a 238 network and registering itself with the RPC portmapper MAY choose an 239 arbitrary port, or MAY use the alternative well-known port number for 240 its RPC-over-RDMA service (see Section 8). The chosen port MAY be 241 registered with the RPC portmapper under the netids assigned in 242 [I-D.ietf-nfsv4-rfc5666bis]. 244 4. Upper Layer Bindings for NFS Version 2 and 3 Auxiliary Protocols 246 NFS versions 2 and 3 are typically deployed with several other 247 protocols, sometimes referred to as "NFS auxiliary protocols." These 248 are distinct RPC Programs that define procedures which are not part 249 of the NFS version 2 or version 3 RPC Programs. The Upper Layer 250 Bindings in this section apply to: 252 o Versions 2 and 3 of the MOUNT protocol [RFC1813] 254 o Versions 1, 3, and 4 of the NLM protocol [RFC1813] 256 o Version 1 of the NSM protocol, described in Chapter 11 of [XNFS] 258 o Version 1 of the NFSACL protocol, which does not have a public 259 definition. NFSACL is treated in this document as a de facto 260 standard, as there are several interoperating implementations. 262 4.1. MOUNT, NLM, and NSM Protocols 264 Historically, NFS/RDMA implementations have chosen to convey the 265 MOUNT, NLM, and NSM protocols via TCP. To enable interoperation of 266 these protocols when NFS/RDMA is in use, a legacy NFS server MUST 267 provide TCP-based MOUNT, NLM, and NSM services. 269 4.2. NFSACL Protocol 271 Legacy clients and servers that support the NFSACL RPC Program 272 typically convey NFSACL procedures on the same connection as NFS RPC 273 Programs. This obviates the need for separate rpcbind queries to 274 discover server support for this RPC Program. 276 ACLs are typically small, but even large ACLs must be encoded and 277 decoded to some degree. Thus no data item in this Upper Layer 278 Protocol is DDP-eligible. 280 For procedures whose replies do not include an ACL object, the size 281 of a reply is determined directly from the NFSACL RPC Program's XDR 282 definition. 284 There is no protocol-specified size limit for NFS version 3 ACLs, and 285 there is no mechanism in either the NFSACL or NFS RPC Programs for a 286 Legacy client to ascertain the largest ACL a Legacy server can 287 return. Legacy client implementations should choose a maximum size 288 for ACLs based on their own internal limits. 290 Because an NFSACL client cannot know in advance how large a returned 291 ACL will be, it can use short Reply chunk retry when an NFSACL GETACL 292 operation encounters a transport error. 294 5. Upper Layer Binding For NFS Version 4 296 The Upper Layer Binding specification in this section applies to RPC 297 Programs defined in NFS Version 4.0 [RFC7530], NFS Version 4.1 298 [RFC5661], and NFS Version 4.2 [RFC7862]. 300 5.1. DDP-Eligibility 302 Only the following XDR data items in the COMPOUND procedure of all 303 NFS version 4 minor versions are DDP-eligible: 305 o The opaque data field in the WRITE4args structure 307 o The linkdata field of the NF4LNK arm in the createtype4 union 309 o The opaque data field in the READ4resok structure 311 o The linkdata field in the READLINK4resok structure 313 5.2. Reply Size Estimation 315 Within NFS version 4, there are certain variable-length result data 316 items whose maximum size cannot be estimated by clients reliably 317 because there is no protocol-specified size limit on these arrays. 318 These include: 320 o The attrlist4 field 322 o Fields containing ACLs such as fattr4_acl, fattr4_dacl, 323 fattr4_sacl 325 o Fields in the fs_locations4 and fs_locations_info4 data structures 327 o Fields opaque to the NFS version 4 protocol which pertain to pNFS 328 layout metadata, such as loc_body, loh_body, da_addr_body, 329 lou_body, lrf_body, fattr_layout_types and fs_layout_types, 331 5.2.1. Reply Size Estimation for Minor Version 0 333 The NFS version 4.0 protocol itself does not impose any bound on the 334 size of NFS calls or responses. 336 Some of the data items enumerated in Section 5.2 (in particular, the 337 items related to ACLs and fs_locations) make it difficult to predict 338 the maximum size of NFS version 4.0 replies that interrogate 339 variable-length fattr4 attributes. Client implementations might rely 340 on their own internal architectural limits to constrain the reply 341 size, but such limits are not always guaranteed to be reliable. 343 When an especially large fattr4 result is expected, a Reply chunk 344 might be required. An NFS version 4.0 client can use short Reply 345 chunk retry when an NFS COMPOUND containing a GETATTR operation 346 encounters a transport error. 348 The use of NFS COMPOUND operations raises the possibility of requests 349 that combine a non-idempotent operation (e.g. RENAME) with a GETATTR 350 operation that requests one or more variable-length results. This 351 combination should be avoided by ensuring that any GETATTR operation 352 that requests a result of unpredictable length is sent in an NFS 353 COMPOUND by itself. 355 5.2.2. Reply Size Estimation for Minor Version 1 and Newer 357 In NFS version 4.1 and newer minor versions, the csa_fore_chan_attrs 358 argument of the CREATE_SESSION operation contains a 359 ca_maxresponsesize field. The value in this field can be taken as 360 the absolute maximum size of replies generated by an NFS version 4.1 361 server. 363 This value can be used in cases where it is not possible to estimate 364 a reply size upper bound precisely. In practice, objects such as 365 ACLs, named attributes, layout bodies, and security labels are much 366 smaller than this maximum. 368 5.3. RPC Binding Considerations 370 NFS version 4 servers are required to listen on TCP port 2049, and 371 they are not required to register with an rpcbind service [RFC7530]. 373 Therefore, an NFS version 4 server supporting RPC-over-RDMA Version 374 One MUST use the alternative well-known port number for its RPC-over- 375 RDMA service (see Section 8). Clients SHOULD connect to this well- 376 known port without consulting the RPC portmapper (as for NFS version 377 4 on TCP transports). 379 5.4. NFS COMPOUND Requests 381 5.4.1. Multiple DDP-eligible Data Items 383 An NFS version 4 COMPOUND procedure can contain more than one 384 operation that carries a DDP-eligible data item. An NFS version 4 385 client provides XDR Position values in each Read chunk to 386 disambiguate which chunk is associated with which argument data item. 387 However NFS version 4 server and client implementations must agree in 388 advance on how to pair Write chunks with returned result data items. 390 In the following list, a "READ operation" refers to any NFS Version 4 391 operation which has a DDP-eligible result data item. The mechanism 392 specified in Section 4.3.2 of [I-D.ietf-nfsv4-rfc5666bis]) is applied 393 to this class of operations: 395 o If an NFS version 4 client wishes all DDP-eligible items in an NFS 396 reply to be conveyed inline, it leaves the Write list empty. 398 o The first chunk in the Write list MUST be used by the first READ 399 operation in an NFS version 4 COMPOUND procedure. The next Write 400 chunk is used by the next READ operation, and so on. 402 o If an NFS version 4 client has provided a matching non-empty Write 403 chunk, then the corresponding READ operation MUST return its DDP- 404 eligible data item using that chunk. 406 o If an NFS version 4 client has provided an empty matching Write 407 chunk, then the corresponding READ operation MUST return all of 408 its result data items inline. 410 o If a READ operation returns a union arm which does not contain a 411 DDP-eligible result, and the NFS version 4 client has provided a 412 matching non-empty Write chunk, an NFS version 4 server MUST 413 return an empty Write chunk in that Write list position. 415 o If there are more READ operations than Write chunks, then 416 remaining NFS Read operations in an NFS version 4 COMPOUND that 417 have no matching Write chunk MUST return their results inline. 419 5.4.2. Chunk List Complexity 421 The RPC-over-RDMA Version One protocol does not place any limit on 422 the number of chunks or segments that may appear in Read or Write 423 lists. However, for various reasons NFS version 4 server 424 implementations often have practical limits on the number of chunks 425 or segments they are prepared to process in a single RPC transaction 426 conveyed via RPC-over-RDMA Version One. 428 These implementation limits are especially important when Kerberos 429 integrity or privacy is in use [RFC7861]. GSS services increase the 430 size of credential material in RPC headers, potentially requiring 431 more frequent use of Long messages. This can increase the complexity 432 of chunk lists independent of the NFS version 4 COMPOUND being 433 conveyed. 435 In the absence of explicit knowledge of the server's limits, NFS 436 Version 4 clients SHOULD follow the prescriptions listed below when 437 constructing RPC-over-RDMA Version One messages. NFS Version 4 438 servers MUST accept and process such requests. 440 o The Read list can contain either a Position-Zero Read chunk, one 441 Read chunk with a non-zero Position, or both. 443 o The Write list can contain no more than one Write chunk. 445 o Any chunk can contain up to sixteen RDMA segments. 447 NFS version 4 clients wishing to send more complex chunk lists can 448 provide configuration interfaces to bound the complexity of NFS 449 version 4 COMPOUNDs, limit the number of elements in scatter-gather 450 operations, and avoid other sources of chunk overruns at the 451 receiving peer. 453 An NFS Version 4 server SHOULD return one of the following responses 454 to a client that has sent an RPC transaction via RPC-over-RDMA 455 Version One which cannot be processed due to chunk list complexity 456 limits on the server: 458 o A problem is detected by the transport layer while parsing the 459 transport header in an RPC Call message. The server responds with 460 an RDMA_ERROR message with the err field set to ERR_CHUNK. 462 o A problem is detected during XDR decoding of the RPC Call message 463 while the RPC layer reassembles the call's XDR stream. The server 464 responds with an RPC reply with its "reply_stat" field set to 465 MSG_ACCEPTED and its "accept_stat" field set to GARBAGE_ARGS. 467 After receiving one of these errors, an NFS version 4 client SHOULD 468 NOT retransmit the failing request, as the result would be the same 469 error. It SHOULD immediately terminate the RPC transaction 470 associated with the XID in the reply. 472 5.4.3. NFS Version 4 COMPOUND Example 474 The following example shows a Write list with three Write chunks, A, 475 B, and C. The NFS version 4 server consumes the provided Write 476 chunks by writing the results of the designated operations in the 477 compound request (READ and READLINK) back to each chunk. 479 Write list: 481 A --> B --> C 483 NFS version 4 COMPOUND request: 485 PUTFH LOOKUP READ PUTFH LOOKUP READLINK PUTFH LOOKUP READ 486 | | | 487 v v v 488 A B C 490 If the NFS version 4 client does not want to have the READLINK result 491 returned via RDMA, it provides an empty Write chunk for buffer B to 492 indicate that the READLINK result must be returned inline. 494 5.5. NFS Callback Requests 496 The NFS version 4 family of protocols support server-initiated 497 callbacks to notify NFS version 4 clients of events such as recalled 498 delegations. 500 5.5.1. NFS Version 4.0 Callback 502 NFS version 4.0 implementations typically employ a separate TCP 503 connection to handle callback operations, even when the forward 504 channel uses an RPC-over-RDMA Version One transport. 506 No operation in the NFS version 4.0 callback RPC Program conveys a 507 significant data payload. Therefore, no XDR data items in this RPC 508 Program is DDP-eligible. 510 A CB_RECALL reply is small and fixed in size. The CB_GETATTR reply 511 contains a variable-length fattr4 data item. See Section 5.2.1 for a 512 discussion of reply size prediction for this data item. 514 An NFS version 4.0 client advertises netids and ad hoc port addresses 515 for contacting its NFS version 4.0 callback service using the 516 SETCLIENTID operation. 518 5.5.2. NFS Version 4.1 Callback 520 In NFS version 4.1 and newer minor versions, callback operations may 521 appear on the same connection as is used for NFS version 4 forward 522 channel client requests. NFS version 4 clients and servers MUST use 523 the approach described in [I-D.ietf-nfsv4-rpcrdma-bidirection] when 524 backchannel operations are conveyed on RPC-over-RDMA Version One 525 transports. 527 The csa_back_chan_attrs argument of the CREATE_SESSION operation 528 contains a ca_maxresponsesize field. The value in this field can be 529 taken as the absolute maximum size of backchannel replies generated 530 by a replying NFS version 4 client. 532 There are no DDP-eligible data items in callback procedures defined 533 in NFS version 4.1 or NFS version 4.2. However, some callback 534 operations, such as messages that convey device ID information, can 535 be large, in which case a Long Call or Reply might be required. 537 When an NFS version 4.1 client can support Long Calls in its 538 backchannel, it reports a backchannel ca_maxrequestsize that is 539 larger than the connection's inline thresholds. Otherwise an NFS 540 version 4 server MUST use only Short messages to convey backchannel 541 operations. 543 5.6. Session-Related Considerations 545 The presence of an NFS session (defined in [RFC5661]) has no effect 546 on the operation of RPC-over-RDMA Version One. None of the 547 operations introduced to support NFS sessions (e.g. the SEQUENCE 548 operation) contain DDP-eligible data items. There is no need to 549 match the number of session slots with the number of available RPC- 550 over-RDMA credits. 552 However, there are a few new cases where an RPC transaction can fail. 553 For example, a requester might receive, in response to an RPC 554 request, an RDMA_ERROR message with an rdma_err value of ERR_CHUNK. 555 These situations are not different from existing RPC errors which an 556 NFS session implementation is already prepared to handle for other 557 transports. And as with other transports during such a failure, 558 there might be no SEQUENCE result available to the requester to 559 distinguish whether failure occurred before or after the requested 560 operations were executed on the responder. 562 When a transport error occurs (e.g. RDMA_ERROR), the requester 563 proceeds as usual to match the incoming XID value to a waiting RPC 564 Call. The RPC transaction is terminated, and the result status is 565 reported to the Upper Layer Protocol. The requester's session 566 implementation then determines the session ID and slot for the failed 567 request, and performs slot recovery to make that slot usable again. 568 If this were not done, that slot could be rendered permanently 569 unavailable. 571 5.7. Transport Considerations 573 5.7.1. Congestion Avoidance 575 Section 3.1 of [RFC7530] states: 577 Where an NFS version 4 implementation supports operation over the 578 IP network protocol, the supported transport layer between NFS and 579 IP MUST be an IETF standardized transport protocol that is 580 specified to avoid network congestion; such transports include TCP 581 and the Stream Control Transmission Protocol (SCTP). 583 Section 2.9.1 of [RFC5661] also states: 585 Even if NFS version 4.1 is used over a non-IP network protocol, it 586 is RECOMMENDED that the transport support congestion control. 588 It is permissible for a connectionless transport to be used under 589 NFS version 4.1; however, reliable and in-order delivery of data 590 combined with congestion control by the connectionless transport 591 is REQUIRED. As a consequence, UDP by itself MUST NOT be used as 592 an NFS version 4.1 transport. 594 RPC-over-RDMA Version One is constructed on a platform of RDMA 595 Reliable Connections [I-D.ietf-nfsv4-rfc5666bis] [RFC5041]. RDMA 596 Reliable Connections are reliable, connection-oriented transports 597 that guarantee in-order delivery, meeting all above requirements for 598 NFS version 4 transports. 600 5.7.2. Retransmission and Keep-alive 602 NFS version 4 client implementations often rely on a transport-layer 603 keep-alive mechanism to detect when an NFS version 4 server has 604 become unresponsive. When an NFS server is no longer responsive, 605 client-side keep-alive terminates the connection, which in turn 606 triggers reconnection and RPC retransmission. 608 Some RDMA transports (such as Reliable Connections on InfiniBand) 609 have no keep-alive mechanism. Without a disconnect or new RPC 610 traffic, such connections can remain alive long after an NFS server 611 has become unresponsive. Once an NFS client has consumed all 612 available RPC-over-RDMA credits on that transport connection, it will 613 forever await a reply before sending another RPC request. 615 NFS version 4 clients SHOULD reserve one RPC-over-RDMA credit to use 616 for periodic server or connection health assessment. This credit can 617 be used to drive an RPC request on an otherwise idle connection, 618 triggering either a quick affirmative server response or immediate 619 connection termination. 621 In addition to network partition and request loss scenarios, RPC- 622 over-RDMA transport connections can be terminated when a Transport 623 header is malformed, Reply messages are larger than receive 624 resources, or when too many RPC-over-RDMA messages are sent at once. 625 In such cases: 627 o If there is a transport error indicated (ie, RDMA_ERROR) before 628 the disconnect or instead of a disconnect, the requester MUST 629 respond to that error as prescribed by the specification of the 630 RPC transport. Then the NFS version 4 rules for handling 631 retransmission apply. 633 o If there is a transport disconnect and the responder has provided 634 no other response for a request, then only the NFS version 4 rules 635 for handling retransmission apply. 637 6. Extending NFS Upper Layer Bindings 639 RPC Programs such as NFS are required to have an Upper Layer Binding 640 specification to interoperate on RPC-over-RDMA Version One transports 641 [I-D.ietf-nfsv4-rfc5666bis]. Via standards action, the Upper Layer 642 Binding specified in this document can be extended to cover versions 643 of the NFS version 4 protocol specified after NFS version 4 minor 644 version 2, or separately published extensions to an existing NFS 645 version 4 minor version, as described in [I-D.ietf-nfsv4-versioning]. 647 7. Security Considerations 649 RPC-over-RDMA Version One supports all RPC security models, including 650 RPCSEC_GSS security and transport-level security [RFC7861]. The 651 choice of what Direct Data Placement mechanism to convey RPC argument 652 and results does not affect this, since it changes only the method of 653 data transfer. Specifically, the requirements of 654 [I-D.ietf-nfsv4-rfc5666bis] ensure that this choice does not 655 introduce new vulnerabilities. 657 Because this document defines only the binding of the NFS protocols 658 atop [I-D.ietf-nfsv4-rfc5666bis], all relevant security 659 considerations are therefore to be described at that layer. 661 8. IANA Considerations 663 The use of direct data placement in NFS introduces a need for an 664 additional port number assignment for networks that share traditional 665 UDP and TCP port spaces with RDMA services. The iWARP protocol is 666 such an example [RFC5041] [RFC5040]. 668 For this purpose, a set of transport protocol port number assignments 669 is specified by this document. IANA has assigned the following ports 670 for NFS/RDMA in the IANA port registry, according to the guidelines 671 described in [RFC6335]. 673 nfsrdma 20049/tcp Network File System (NFS) over RDMA 674 nfsrdma 20049/udp Network File System (NFS) over RDMA 675 nfsrdma 20049/sctp Network File System (NFS) over RDMA 677 This document should be listed as the reference for the nfsrdma port 678 assignments. This document does not alter these assignments. 680 9. References 682 9.1. Normative References 684 [I-D.ietf-nfsv4-rfc5666bis] 685 Lever, C., Simpson, W., and T. Talpey, "Remote Direct 686 Memory Access Transport for Remote Procedure Call, Version 687 One", draft-ietf-nfsv4-rfc5666bis-11 (work in progress), 688 March 2017. 690 [I-D.ietf-nfsv4-rpcrdma-bidirection] 691 Lever, C., "Bi-directional Remote Procedure Call On RPC- 692 over-RDMA Transports", draft-ietf-nfsv4-rpcrdma- 693 bidirection-08 (work in progress), March 2017. 695 [RFC1833] Srinivasan, R., "Binding Protocols for ONC RPC Version 2", 696 RFC 1833, DOI 10.17487/RFC1833, August 1995, 697 . 699 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 700 Requirement Levels", BCP 14, RFC 2119, 701 DOI 10.17487/RFC2119, March 1997, 702 . 704 [RFC5661] Shepler, S., Ed., Eisler, M., Ed., and D. Noveck, Ed., 705 "Network File System (NFS) Version 4 Minor Version 1 706 Protocol", RFC 5661, DOI 10.17487/RFC5661, January 2010, 707 . 709 [RFC6335] Cotton, M., Eggert, L., Touch, J., Westerlund, M., and S. 710 Cheshire, "Internet Assigned Numbers Authority (IANA) 711 Procedures for the Management of the Service Name and 712 Transport Protocol Port Number Registry", BCP 165, 713 RFC 6335, DOI 10.17487/RFC6335, August 2011, 714 . 716 [RFC7530] Haynes, T., Ed. and D. Noveck, Ed., "Network File System 717 (NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530, 718 March 2015, . 720 [RFC7861] Adamson, A. and N. Williams, "Remote Procedure Call (RPC) 721 Security Version 3", RFC 7861, DOI 10.17487/RFC7861, 722 November 2016, . 724 [RFC7862] Haynes, T., "Network File System (NFS) Version 4 Minor 725 Version 2 Protocol", RFC 7862, DOI 10.17487/RFC7862, 726 November 2016, . 728 9.2. Informative References 730 [I-D.ietf-nfsv4-versioning] 731 Noveck, D., "Rules for NFSv4 Extensions and Minor 732 Versions", draft-ietf-nfsv4-versioning-09 (work in 733 progress), December 2016. 735 [RFC1094] Nowicki, B., "NFS: Network File System Protocol 736 specification", RFC 1094, DOI 10.17487/RFC1094, March 737 1989, . 739 [RFC1813] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS 740 Version 3 Protocol Specification", RFC 1813, 741 DOI 10.17487/RFC1813, June 1995, 742 . 744 [RFC5040] Recio, R., Metzler, B., Culley, P., Hilland, J., and D. 745 Garcia, "A Remote Direct Memory Access Protocol 746 Specification", RFC 5040, DOI 10.17487/RFC5040, October 747 2007, . 749 [RFC5041] Shah, H., Pinkerton, J., Recio, R., and P. Culley, "Direct 750 Data Placement over Reliable Transports", RFC 5041, 751 DOI 10.17487/RFC5041, October 2007, 752 . 754 [RFC5666] Talpey, T. and B. Callaghan, "Remote Direct Memory Access 755 Transport for Remote Procedure Call", RFC 5666, 756 DOI 10.17487/RFC5666, January 2010, 757 . 759 [RFC5667] Talpey, T. and B. Callaghan, "Network File System (NFS) 760 Direct Data Placement", RFC 5667, DOI 10.17487/RFC5667, 761 January 2010, . 763 [XNFS] The Open Group, "Protocols for Interworking: XNFS, Version 764 3W", February 1998. 766 Appendix A. Changes Since RFC 5667 768 Corrections and updates made necessary by new language in 769 [I-D.ietf-nfsv4-rfc5666bis] have been introduced. For example, 770 references to deprecated features of RPC-over-RDMA Version One, such 771 as RDMA_MSGP, and the use of the Read list for handling RPC replies, 772 have been removed. The term "mapping" has been replaced with the 773 term "binding" or "Upper Layer Binding" throughout the document. 774 Material that duplicates what is in [I-D.ietf-nfsv4-rfc5666bis] has 775 been deleted. 777 Material required by [I-D.ietf-nfsv4-rfc5666bis] for Upper Layer 778 Bindings that was not present in [RFC5667] has been added. A 779 complete discussion of reply size estimation has been introduced for 780 all protocols covered by the Upper Layer Bindings in this document. 782 Technical corrections have been made. For example, the mention of 783 12KB and 36KB inline thresholds have been removed. The reference to 784 a non-existant NFS version 4 SYMLINK operation has been replaced. 786 The discussion of NFS version 4 COMPOUND handling has been completed. 787 Some changes were made to the algorithm for matching DDP-eligible 788 results to Write chunks. 790 Requirements to ignore extra Read or Write chunks have been removed 791 from the NFS version 2 and 3 Upper Layer Binding, as they conflict 792 with [I-D.ietf-nfsv4-rfc5666bis]. 794 A section discussing NFS version 4 retransmission and connection loss 795 has been added. 797 The following additional improvements have been made, relative to 798 [RFC5667]: 800 o An explicit discussion of NFS version 4.0 and NFS version 4.1 801 backchannel operation has replaced the previous treatment of 802 callback operations. 804 o A binding for NFS version 4.2 has been added. 806 o A section suggesting a mechanism for periodically assessing 807 connection health has been introduced. 809 o Ambiguous or erroneous uses of RFC2119 terms have been corrected. 811 o References to obsolete RFCs have been updated. 813 o An IANA Considerations Section has been added, which specifies the 814 port assignments for NFS/RDMA. This replaces the example 815 assignment that appeared in [RFC5666]. 817 o Code excerpts have been removed, and figures have been modernized. 819 Appendix B. Acknowledgments 821 The author gratefully acknowledges the work of Brent Callaghan and 822 Tom Talpey on the original NFS Direct Data Placement specification 823 [RFC5667]. Tom contributed the text of Section 5.4.2. 825 Dave Noveck provided excellent review, constructive suggestions, and 826 consistent navigational guidance throughout the process of drafting 827 this document. Dave contributed the text of Section 5.6 and 828 Section 6, and insisted on precise discussion of reply size 829 estimation. 831 Thanks to Karen Deitke for her sharp observations about idempotency, 832 NFS COMPOUNDs, and NFS sessions. 834 Special thanks go to Transport Area Director Spencer Dawkins, nfsv4 835 Working Group Chair Spencer Shepler, and nfsv4 Working Group 836 Secretary Thomas Haynes for their support. The author also wishes to 837 thank Bill Baker and Greg Marsden for their support of this work. 839 Author's Address 840 Charles Lever (editor) 841 Oracle Corporation 842 1015 Granger Avenue 843 Ann Arbor, MI 48104 844 USA 846 Phone: +1 248 816 6463 847 Email: chuck.lever@oracle.com