idnits 2.17.1 draft-ietf-nfsv4-minorversion2-09.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 : ---------------------------------------------------------------------------- == There are 5 instances of lines with non-RFC6890-compliant IPv4 addresses in the document. If these are example addresses, they should be changed. == There are 5 instances of lines with private range IPv4 addresses in the document. If these are generic example addresses, they should be changed to use any of the ranges defined in RFC 6890 (or successor): 192.0.2.x, 198.51.100.x or 203.0.113.x. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Using lowercase 'not' together with uppercase 'MUST', 'SHALL', 'SHOULD', or 'RECOMMENDED' is not an accepted usage according to RFC 2119. Please use uppercase 'NOT' together with RFC 2119 keywords (if that is what you mean). Found 'MUST not' in this paragraph: Furthermore, each DS MUST not report to a client either a sparse ADB or data which belongs to another DS. One implication of this requirement is that the app_data_block4's adb_block_size MUST be either be the stripe width or the stripe width must be an even multiple of it. == Using lowercase 'not' together with uppercase 'MUST', 'SHALL', 'SHOULD', or 'RECOMMENDED' is not an accepted usage according to RFC 2119. Please use uppercase 'NOT' together with RFC 2119 keywords (if that is what you mean). Found 'MUST not' in this paragraph: The second change is to provide a method for the server to notify the client that the attribute changed on an open file on the server. If the file is closed, then during the open attempt, the client will gather the new attribute value. The server MUST not communicate the new value of the attribute, the client MUST query it. This requirement stems from the need for the client to provide sufficient access rights to the attribute. == Using lowercase 'not' together with uppercase 'MUST', 'SHALL', 'SHOULD', or 'RECOMMENDED' is not an accepted usage according to RFC 2119. Please use uppercase 'NOT' together with RFC 2119 keywords (if that is what you mean). Found 'MUST not' in this paragraph: When a data server chooses to return a hole result, it has the option of returning hole information for the data stored on that data server (as defined by the data layout), but it MUST not return results for a byte range that includes data managed by another data server. Data servers that can obtain hole information for the parts of the file stored on that data server, the data server SHOULD return HOLE_INFO and the byte range of the hole stored on that data server. == 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 02, 2012) is 4370 days in the past. Is this intentional? -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: '0' is mentioned on line 3756, but not defined -- Looks like a reference, but probably isn't: '32K' on line 3756 -- Possible downref: Non-RFC (?) normative reference: ref. '1' ** Obsolete normative reference: RFC 5661 (ref. '2') (Obsoleted by RFC 8881) -- Possible downref: Non-RFC (?) normative reference: ref. '3' -- Possible downref: Non-RFC (?) normative reference: ref. '6' == Outdated reference: A later version (-05) exists of draft-ietf-nfsv4-labreqs-00 ** Downref: Normative reference to an Informational draft: draft-ietf-nfsv4-labreqs (ref. '7') == Outdated reference: A later version (-35) exists of draft-ietf-nfsv4-rfc3530bis-09 -- Obsolete informational reference (is this intentional?): RFC 2616 (ref. '13') (Obsoleted by RFC 7230, RFC 7231, RFC 7232, RFC 7233, RFC 7234, RFC 7235) -- Obsolete informational reference (is this intentional?): RFC 5226 (ref. '24') (Obsoleted by RFC 8126) Summary: 2 errors (**), 0 flaws (~~), 10 warnings (==), 8 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 NFSv4 T. Haynes 3 Internet-Draft Editor 4 Intended status: Standards Track May 02, 2012 5 Expires: November 3, 2012 7 NFS Version 4 Minor Version 2 8 draft-ietf-nfsv4-minorversion2-09.txt 10 Abstract 12 This Internet-Draft describes NFS version 4 minor version two, 13 focusing mainly on the protocol extensions made from NFS version 4 14 minor version 0 and NFS version 4 minor version 1. Major extensions 15 introduced in NFS version 4 minor version two include: Server-side 16 Copy, Space Reservations, and Support for Sparse Files. 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 RFC 2119 [1]. 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 3, 2012. 41 Copyright Notice 43 Copyright (c) 2012 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 . . . . . . . . . . . . . . . . . . . . . . . . . 6 71 1.1. The NFS Version 4 Minor Version 2 Protocol . . . . . . . 6 72 1.2. Scope of This Document . . . . . . . . . . . . . . . . . 6 73 1.3. NFSv4.2 Goals . . . . . . . . . . . . . . . . . . . . . . 6 74 1.4. Overview of NFSv4.2 Features . . . . . . . . . . . . . . 6 75 1.4.1. Sparse Files . . . . . . . . . . . . . . . . . . . . . 6 76 1.4.2. Application I/O Advise . . . . . . . . . . . . . . . . 7 77 1.5. Differences from NFSv4.1 . . . . . . . . . . . . . . . . 7 78 2. NFS Server-side Copy . . . . . . . . . . . . . . . . . . . . . 7 79 2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . 7 80 2.2. Protocol Overview . . . . . . . . . . . . . . . . . . . . 8 81 2.2.1. Intra-Server Copy . . . . . . . . . . . . . . . . . . 9 82 2.2.2. Inter-Server Copy . . . . . . . . . . . . . . . . . . 11 83 2.2.3. Server-to-Server Copy Protocol . . . . . . . . . . . . 13 84 2.3. Operations . . . . . . . . . . . . . . . . . . . . . . . 15 85 2.3.1. netloc4 - Network Locations . . . . . . . . . . . . . 15 86 2.3.2. Copy Offload Stateids . . . . . . . . . . . . . . . . 16 87 2.4. Security Considerations . . . . . . . . . . . . . . . . . 16 88 2.4.1. Inter-Server Copy Security . . . . . . . . . . . . . . 16 89 3. Sparse Files . . . . . . . . . . . . . . . . . . . . . . . . . 25 90 3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . 25 91 3.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 25 92 4. Space Reservation . . . . . . . . . . . . . . . . . . . . . . 26 93 4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . 26 94 5. Support for Application IO Hints . . . . . . . . . . . . . . . 28 95 5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . 28 96 5.2. POSIX Requirements . . . . . . . . . . . . . . . . . . . 29 97 5.3. Additional Requirements . . . . . . . . . . . . . . . . . 30 98 5.4. Security Considerations . . . . . . . . . . . . . . . . . 31 99 5.5. IANA Considerations . . . . . . . . . . . . . . . . . . . 31 100 6. Application Data Block Support . . . . . . . . . . . . . . . . 31 101 6.1. Generic Framework . . . . . . . . . . . . . . . . . . . . 32 102 6.1.1. Data Block Representation . . . . . . . . . . . . . . 32 103 6.1.2. Data Content . . . . . . . . . . . . . . . . . . . . . 33 104 6.2. pNFS Considerations . . . . . . . . . . . . . . . . . . . 33 105 6.3. An Example of Detecting Corruption . . . . . . . . . . . 34 106 6.4. Example of READ_PLUS . . . . . . . . . . . . . . . . . . 35 107 6.5. Zero Filled Holes . . . . . . . . . . . . . . . . . . . . 36 108 7. Labeled NFS . . . . . . . . . . . . . . . . . . . . . . . . . 36 109 7.1. Introduction . . . . . . . . . . . . . . . . . . . . . . 36 110 7.2. Definitions . . . . . . . . . . . . . . . . . . . . . . . 37 111 7.3. MAC Security Attribute . . . . . . . . . . . . . . . . . 37 112 7.3.1. Interpreting FATTR4_SEC_LABEL . . . . . . . . . . . . 38 113 7.3.2. Delegations . . . . . . . . . . . . . . . . . . . . . 39 114 7.3.3. Permission Checking . . . . . . . . . . . . . . . . . 39 115 7.3.4. Object Creation . . . . . . . . . . . . . . . . . . . 39 116 7.3.5. Existing Objects . . . . . . . . . . . . . . . . . . . 40 117 7.3.6. Label Changes . . . . . . . . . . . . . . . . . . . . 40 118 7.4. pNFS Considerations . . . . . . . . . . . . . . . . . . . 41 119 7.5. Discovery of Server LNFS Support . . . . . . . . . . . . 41 120 7.6. MAC Security NFS Modes of Operation . . . . . . . . . . . 41 121 7.6.1. Full Mode . . . . . . . . . . . . . . . . . . . . . . 42 122 7.6.2. Guest Mode . . . . . . . . . . . . . . . . . . . . . . 43 123 7.7. Security Considerations . . . . . . . . . . . . . . . . . 43 124 8. Sharing change attribute implementation details with NFSv4 125 clients . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 126 8.1. Introduction . . . . . . . . . . . . . . . . . . . . . . 44 127 8.2. Definition of the 'change_attr_type' per-file system 128 attribute . . . . . . . . . . . . . . . . . . . . . . . . 44 129 9. Security Considerations . . . . . . . . . . . . . . . . . . . 46 130 10. Error Values . . . . . . . . . . . . . . . . . . . . . . . . . 46 131 10.1. Error Definitions . . . . . . . . . . . . . . . . . . . . 46 132 10.1.1. General Errors . . . . . . . . . . . . . . . . . . . . 46 133 10.1.2. Server to Server Copy Errors . . . . . . . . . . . . . 46 134 10.1.3. Labeled NFS Errors . . . . . . . . . . . . . . . . . . 47 135 11. File Attributes . . . . . . . . . . . . . . . . . . . . . . . 47 136 11.1. Attribute Definitions . . . . . . . . . . . . . . . . . . 47 137 12. Operations: REQUIRED, RECOMMENDED, or OPTIONAL . . . . . . . . 48 138 13. NFSv4.2 Operations . . . . . . . . . . . . . . . . . . . . . . 52 139 13.1. Operation 59: COPY - Initiate a server-side copy . . . . 52 140 13.2. Operation 60: COPY_ABORT - Cancel a server-side copy . . 59 141 13.3. Operation 61: COPY_NOTIFY - Notify a source server of 142 a future copy . . . . . . . . . . . . . . . . . . . . . . 60 143 13.4. Operation 62: COPY_REVOKE - Revoke a destination 144 server's copy privileges . . . . . . . . . . . . . . . . 62 145 13.5. Operation 63: COPY_STATUS - Poll for status of a 146 server-side copy . . . . . . . . . . . . . . . . . . . . 63 147 13.6. Modification to Operation 42: EXCHANGE_ID - 148 Instantiate Client ID . . . . . . . . . . . . . . . . . . 64 149 13.7. Operation 64: INITIALIZE . . . . . . . . . . . . . . . . 65 150 13.8. Operation 67: IO_ADVISE - Application I/O access 151 pattern hints . . . . . . . . . . . . . . . . . . . . . . 69 152 13.9. Changes to Operation 51: LAYOUTRETURN . . . . . . . . . . 75 153 13.10. Operation 65: READ_PLUS . . . . . . . . . . . . . . . . . 78 154 13.11. Operation 66: SEEK . . . . . . . . . . . . . . . . . . . 83 155 14. NFSv4.2 Callback Operations . . . . . . . . . . . . . . . . . 84 156 14.1. Procedure 16: CB_ATTR_CHANGED - Notify Client that 157 the File's Attributes Changed . . . . . . . . . . . . . . 84 158 14.2. Operation 15: CB_COPY - Report results of a 159 server-side copy . . . . . . . . . . . . . . . . . . . . 85 160 15. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 87 161 16. References . . . . . . . . . . . . . . . . . . . . . . . . . . 87 162 16.1. Normative References . . . . . . . . . . . . . . . . . . 87 163 16.2. Informative References . . . . . . . . . . . . . . . . . 88 165 Appendix A. Acknowledgments . . . . . . . . . . . . . . . . . . . 89 166 Appendix B. RFC Editor Notes . . . . . . . . . . . . . . . . . . 90 167 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 90 169 1. Introduction 171 1.1. The NFS Version 4 Minor Version 2 Protocol 173 The NFS version 4 minor version 2 (NFSv4.2) protocol is the third 174 minor version of the NFS version 4 (NFSv4) protocol. The first minor 175 version, NFSv4.0, is described in [10] and the second minor version, 176 NFSv4.1, is described in [2]. It follows the guidelines for minor 177 versioning that are listed in Section 11 of [10]. 179 As a minor version, NFSv4.2 is consistent with the overall goals for 180 NFSv4, but extends the protocol so as to better meet those goals, 181 based on experiences with NFSv4.1. In addition, NFSv4.2 has adopted 182 some additional goals, which motivate some of the major extensions in 183 NFSv4.2. 185 1.2. Scope of This Document 187 This document describes the NFSv4.2 protocol. With respect to 188 NFSv4.0 and NFSv4.1, this document does not: 190 o describe the NFSv4.0 or NFSv4.1 protocols, except where needed to 191 contrast with NFSv4.2. 193 o modify the specification of the NFSv4.0 or NFSv4.1 protocols. 195 o clarify the NFSv4.0 or NFSv4.1 protocols. I.e., any 196 clarifications made here apply to NFSv4.2 and neither of the prior 197 protocols. 199 The full XDR for NFSv4.2 is presented in [3]. 201 1.3. NFSv4.2 Goals 203 [[Comment.1: This needs fleshing out! --TH]] 205 1.4. Overview of NFSv4.2 Features 207 [[Comment.2: This needs fleshing out! --TH]] 209 1.4.1. Sparse Files 211 Two new operations are defined to support the reading of sparse files 212 (READ_PLUS) and the punching of holes to remove backing storage 213 (INITIALIZE). 215 1.4.2. Application I/O Advise 217 We propose a new IO_ADVISE operation for NFSv4.2 that clients can use 218 to communicate expected I/O behavior to the server. By communicating 219 future I/O behavior such as whether a file will be accessed 220 sequentially or randomly, and whether a file will or will not be 221 accessed in the near future, servers can optimize future I/O requests 222 for a file by, for example, prefetching or evicting data. This 223 operation can be used to support the posix_fadvise function as well 224 as other applications such as databases and video editors. 226 1.5. Differences from NFSv4.1 228 In NFSv4.1, the only way to introduce new variants of an operation 229 was to introduce a new operation. I.e., READ becomes either READ2 or 230 READ_PLUS. With the use of discriminated unions as parameters to 231 such functions in NFSv4.2, it is possible to add a new arm in a 232 subsequent minor version. And it is also possible to move such an 233 operation from OPTIONAL/RECOMMENDED to REQUIRED. Forcing an 234 implementation to adopt each arm of a discriminated union at such a 235 time does not meet the spirit of the minor versioning rules. As 236 such, new arms of a discriminated union MUST follow the same 237 guidelines for minor versioning as operations in NFSv4.1 - i.e., they 238 may not be made REQUIRED. To support this, a new error code, 239 NFS4ERR_UNION_NOTSUPP, is introduced which allows the server to 240 communicate to the client that the operation is supported, but the 241 specific arm of the discriminated union is not. 243 2. NFS Server-side Copy 245 2.1. Introduction 247 This section describes a server-side copy feature for the NFS 248 protocol. 250 The server-side copy feature provides a mechanism for the NFS client 251 to perform a file copy on the server without the data being 252 transmitted back and forth over the network. 254 Without this feature, an NFS client copies data from one location to 255 another by reading the data from the server over the network, and 256 then writing the data back over the network to the server. Using 257 this server-side copy operation, the client is able to instruct the 258 server to copy the data locally without the data being sent back and 259 forth over the network unnecessarily. 261 In general, this feature is useful whenever data is copied from one 262 location to another on the server. It is particularly useful when 263 copying the contents of a file from a backup. Backup-versions of a 264 file are copied for a number of reasons, including restoring and 265 cloning data. 267 If the source object and destination object are on different file 268 servers, the file servers will communicate with one another to 269 perform the copy operation. The server-to-server protocol by which 270 this is accomplished is not defined in this document. 272 2.2. Protocol Overview 274 The server-side copy offload operations support both intra-server and 275 inter-server file copies. An intra-server copy is a copy in which 276 the source file and destination file reside on the same server. In 277 an inter-server copy, the source file and destination file are on 278 different servers. In both cases, the copy may be performed 279 synchronously or asynchronously. 281 Throughout the rest of this document, we refer to the NFS server 282 containing the source file as the "source server" and the NFS server 283 to which the file is transferred as the "destination server". In the 284 case of an intra-server copy, the source server and destination 285 server are the same server. Therefore in the context of an intra- 286 server copy, the terms source server and destination server refer to 287 the single server performing the copy. 289 The operations described below are designed to copy files. Other 290 file system objects can be copied by building on these operations or 291 using other techniques. For example if the user wishes to copy a 292 directory, the client can synthesize a directory copy by first 293 creating the destination directory and then copying the source 294 directory's files to the new destination directory. If the user 295 wishes to copy a namespace junction [11] [12], the client can use the 296 ONC RPC Federated Filesystem protocol [12] to perform the copy. 297 Specifically the client can determine the source junction's 298 attributes using the FEDFS_LOOKUP_FSN procedure and create a 299 duplicate junction using the FEDFS_CREATE_JUNCTION procedure. 301 For the inter-server copy protocol, the operations are defined to be 302 compatible with a server-to-server copy protocol in which the 303 destination server reads the file data from the source server. This 304 model in which the file data is pulled from the source by the 305 destination has a number of advantages over a model in which the 306 source pushes the file data to the destination. The advantages of 307 the pull model include: 309 o The pull model only requires a remote server (i.e., the 310 destination server) to be granted read access. A push model 311 requires a remote server (i.e., the source server) to be granted 312 write access, which is more privileged. 314 o The pull model allows the destination server to stop reading if it 315 has run out of space. In a push model, the destination server 316 must flow control the source server in this situation. 318 o The pull model allows the destination server to easily flow 319 control the data stream by adjusting the size of its read 320 operations. In a push model, the destination server does not have 321 this ability. The source server in a push model is capable of 322 writing chunks larger than the destination server has requested in 323 attributes and session parameters. In theory, the destination 324 server could perform a "short" write in this situation, but this 325 approach is known to behave poorly in practice. 327 The following operations are provided to support server-side copy: 329 COPY_NOTIFY: For inter-server copies, the client sends this 330 operation to the source server to notify it of a future file copy 331 from a given destination server for the given user. 333 COPY_REVOKE: Also for inter-server copies, the client sends this 334 operation to the source server to revoke permission to copy a file 335 for the given user. 337 COPY: Used by the client to request a file copy. 339 COPY_ABORT: Used by the client to abort an asynchronous file copy. 341 COPY_STATUS: Used by the client to poll the status of an 342 asynchronous file copy. 344 CB_COPY: Used by the destination server to report the results of an 345 asynchronous file copy to the client. 347 These operations are described in detail in Section 2.3. This 348 section provides an overview of how these operations are used to 349 perform server-side copies. 351 2.2.1. Intra-Server Copy 353 To copy a file on a single server, the client uses a COPY operation. 354 The server may respond to the copy operation with the final results 355 of the copy or it may perform the copy asynchronously and deliver the 356 results using a CB_COPY operation callback. If the copy is performed 357 asynchronously, the client may poll the status of the copy using 358 COPY_STATUS or cancel the copy using COPY_ABORT. 360 A synchronous intra-server copy is shown in Figure 1. In this 361 example, the NFS server chooses to perform the copy synchronously. 362 The copy operation is completed, either successfully or 363 unsuccessfully, before the server replies to the client's request. 364 The server's reply contains the final result of the operation. 366 Client Server 367 + + 368 | | 369 |--- COPY ---------------------------->| Client requests 370 |<------------------------------------/| a file copy 371 | | 372 | | 374 Figure 1: A synchronous intra-server copy. 376 An asynchronous intra-server copy is shown in Figure 2. In this 377 example, the NFS server performs the copy asynchronously. The 378 server's reply to the copy request indicates that the copy operation 379 was initiated and the final result will be delivered at a later time. 380 The server's reply also contains a copy stateid. The client may use 381 this copy stateid to poll for status information (as shown) or to 382 cancel the copy using a COPY_ABORT. When the server completes the 383 copy, the server performs a callback to the client and reports the 384 results. 386 Client Server 387 + + 388 | | 389 |--- COPY ---------------------------->| Client requests 390 |<------------------------------------/| a file copy 391 | | 392 | | 393 |--- COPY_STATUS --------------------->| Client may poll 394 |<------------------------------------/| for status 395 | | 396 | . | Multiple COPY_STATUS 397 | . | operations may be sent. 398 | . | 399 | | 400 |<-- CB_COPY --------------------------| Server reports results 401 |\------------------------------------>| 402 | | 404 Figure 2: An asynchronous intra-server copy. 406 2.2.2. Inter-Server Copy 408 A copy may also be performed between two servers. The copy protocol 409 is designed to accommodate a variety of network topologies. As shown 410 in Figure 3, the client and servers may be connected by multiple 411 networks. In particular, the servers may be connected by a 412 specialized, high speed network (network 192.168.33.0/24 in the 413 diagram) that does not include the client. The protocol allows the 414 client to setup the copy between the servers (over network 415 10.11.78.0/24 in the diagram) and for the servers to communicate on 416 the high speed network if they choose to do so. 418 192.168.33.0/24 419 +-------------------------------------+ 420 | | 421 | | 422 | 192.168.33.18 | 192.168.33.56 423 +-------+------+ +------+------+ 424 | Source | | Destination | 425 +-------+------+ +------+------+ 426 | 10.11.78.18 | 10.11.78.56 427 | | 428 | | 429 | 10.11.78.0/24 | 430 +------------------+------------------+ 431 | 432 | 433 | 10.11.78.243 434 +-----+-----+ 435 | Client | 436 +-----------+ 438 Figure 3: An example inter-server network topology. 440 For an inter-server copy, the client notifies the source server that 441 a file will be copied by the destination server using a COPY_NOTIFY 442 operation. The client then initiates the copy by sending the COPY 443 operation to the destination server. The destination server may 444 perform the copy synchronously or asynchronously. 446 A synchronous inter-server copy is shown in Figure 4. In this case, 447 the destination server chooses to perform the copy before responding 448 to the client's COPY request. 450 An asynchronous copy is shown in Figure 5. In this case, the 451 destination server chooses to respond to the client's COPY request 452 immediately and then perform the copy asynchronously. 454 Client Source Destination 455 + + + 456 | | | 457 |--- COPY_NOTIFY --->| | 458 |<------------------/| | 459 | | | 460 | | | 461 |--- COPY ---------------------------->| 462 | | | 463 | | | 464 | |<----- read -----| 465 | |\--------------->| 466 | | | 467 | | . | Multiple reads may 468 | | . | be necessary 469 | | . | 470 | | | 471 | | | 472 |<------------------------------------/| Destination replies 473 | | | to COPY 475 Figure 4: A synchronous inter-server copy. 477 Client Source Destination 478 + + + 479 | | | 480 |--- COPY_NOTIFY --->| | 481 |<------------------/| | 482 | | | 483 | | | 484 |--- COPY ---------------------------->| 485 |<------------------------------------/| 486 | | | 487 | | | 488 | |<----- read -----| 489 | |\--------------->| 490 | | | 491 | | . | Multiple reads may 492 | | . | be necessary 493 | | . | 494 | | | 495 | | | 496 |--- COPY_STATUS --------------------->| Client may poll 497 |<------------------------------------/| for status 498 | | | 499 | | . | Multiple COPY_STATUS 500 | | . | operations may be sent 501 | | . | 502 | | | 503 | | | 504 | | | 505 |<-- CB_COPY --------------------------| Destination reports 506 |\------------------------------------>| results 507 | | | 509 Figure 5: An asynchronous inter-server copy. 511 2.2.3. Server-to-Server Copy Protocol 513 During an inter-server copy, the destination server reads the file 514 data from the source server. The source server and destination 515 server are not required to use a specific protocol to transfer the 516 file data. The choice of what protocol to use is ultimately the 517 destination server's decision. 519 2.2.3.1. Using NFSv4.x as a Server-to-Server Copy Protocol 521 The destination server MAY use standard NFSv4.x (where x >= 1) to 522 read the data from the source server. If NFSv4.x is used for the 523 server-to-server copy protocol, the destination server can use the 524 filehandle contained in the COPY request with standard NFSv4.x 525 operations to read data from the source server. Specifically, the 526 destination server may use the NFSv4.x OPEN operation's CLAIM_FH 527 facility to open the file being copied and obtain an open stateid. 528 Using the stateid, the destination server may then use NFSv4.x READ 529 operations to read the file. 531 2.2.3.2. Using an alternative Server-to-Server Copy Protocol 533 In a homogeneous environment, the source and destination servers 534 might be able to perform the file copy extremely efficiently using 535 specialized protocols. For example the source and destination 536 servers might be two nodes sharing a common file system format for 537 the source and destination file systems. Thus the source and 538 destination are in an ideal position to efficiently render the image 539 of the source file to the destination file by replicating the file 540 system formats at the block level. Another possibility is that the 541 source and destination might be two nodes sharing a common storage 542 area network, and thus there is no need to copy any data at all, and 543 instead ownership of the file and its contents might simply be re- 544 assigned to the destination. To allow for these possibilities, the 545 destination server is allowed to use a server-to-server copy protocol 546 of its choice. 548 In a heterogeneous environment, using a protocol other than NFSv4.x 549 (e.g,. HTTP [13] or FTP [14]) presents some challenges. In 550 particular, the destination server is presented with the challenge of 551 accessing the source file given only an NFSv4.x filehandle. 553 One option for protocols that identify source files with path names 554 is to use an ASCII hexadecimal representation of the source 555 filehandle as the file name. 557 Another option for the source server is to use URLs to direct the 558 destination server to a specialized service. For example, the 559 response to COPY_NOTIFY could include the URL 560 ftp://s1.example.com:9999/_FH/0x12345, where 0x12345 is the ASCII 561 hexadecimal representation of the source filehandle. When the 562 destination server receives the source server's URL, it would use 563 "_FH/0x12345" as the file name to pass to the FTP server listening on 564 port 9999 of s1.example.com. On port 9999 there would be a special 565 instance of the FTP service that understands how to convert NFS 566 filehandles to an open file descriptor (in many operating systems, 567 this would require a new system call, one which is the inverse of the 568 makefh() function that the pre-NFSv4 MOUNT service needs). 570 Authenticating and identifying the destination server to the source 571 server is also a challenge. Recommendations for how to accomplish 572 this are given in Section 2.4.1.2.4 and Section 2.4.1.4. 574 2.3. Operations 576 In the sections that follow, several operations are defined that 577 together provide the server-side copy feature. These operations are 578 intended to be OPTIONAL operations as defined in section 17 of [2]. 579 The COPY_NOTIFY, COPY_REVOKE, COPY, COPY_ABORT, and COPY_STATUS 580 operations are designed to be sent within an NFSv4 COMPOUND 581 procedure. The CB_COPY operation is designed to be sent within an 582 NFSv4 CB_COMPOUND procedure. 584 Each operation is performed in the context of the user identified by 585 the ONC RPC credential of its containing COMPOUND or CB_COMPOUND 586 request. For example, a COPY_ABORT operation issued by a given user 587 indicates that a specified COPY operation initiated by the same user 588 be canceled. Therefore a COPY_ABORT MUST NOT interfere with a copy 589 of the same file initiated by another user. 591 An NFS server MAY allow an administrative user to monitor or cancel 592 copy operations using an implementation specific interface. 594 2.3.1. netloc4 - Network Locations 596 The server-side copy operations specify network locations using the 597 netloc4 data type shown below: 599 enum netloc_type4 { 600 NL4_NAME = 0, 601 NL4_URL = 1, 602 NL4_NETADDR = 2 603 }; 604 union netloc4 switch (netloc_type4 nl_type) { 605 case NL4_NAME: utf8str_cis nl_name; 606 case NL4_URL: utf8str_cis nl_url; 607 case NL4_NETADDR: netaddr4 nl_addr; 608 }; 610 If the netloc4 is of type NL4_NAME, the nl_name field MUST be 611 specified as a UTF-8 string. The nl_name is expected to be resolved 612 to a network address via DNS, LDAP, NIS, /etc/hosts, or some other 613 means. If the netloc4 is of type NL4_URL, a server URL [4] 614 appropriate for the server-to-server copy operation is specified as a 615 UTF-8 string. If the netloc4 is of type NL4_NETADDR, the nl_addr 616 field MUST contain a valid netaddr4 as defined in Section 3.3.9 of 617 [2]. 619 When netloc4 values are used for an inter-server copy as shown in 620 Figure 3, their values may be evaluated on the source server, 621 destination server, and client. The network environment in which 622 these systems operate should be configured so that the netloc4 values 623 are interpreted as intended on each system. 625 2.3.2. Copy Offload Stateids 627 A server may perform a copy offload operation asynchronously. An 628 asynchronous copy is tracked using a copy offload stateid. Copy 629 offload stateids are included in the COPY, COPY_ABORT, COPY_STATUS, 630 and CB_COPY operations. 632 Section 8.2.4 of [2] specifies that stateids are valid until either 633 (A) the client or server restart or (B) the client returns the 634 resource. 636 A copy offload stateid will be valid until either (A) the client or 637 server restarts or (B) the client returns the resource by issuing a 638 COPY_ABORT operation or the client replies to a CB_COPY operation. 640 A copy offload stateid's seqid MUST NOT be 0 (zero). In the context 641 of a copy offload operation, it is ambiguous to indicate the most 642 recent copy offload operation using a stateid with seqid of 0 (zero). 643 Therefore a copy offload stateid with seqid of 0 (zero) MUST be 644 considered invalid. 646 2.4. Security Considerations 648 The security considerations pertaining to NFSv4 [10] apply to this 649 document. 651 The standard security mechanisms provide by NFSv4 [10] may be used to 652 secure the protocol described in this document. 654 NFSv4 clients and servers supporting the the inter-server copy 655 operations described in this document are REQUIRED to implement [5], 656 including the RPCSEC_GSSv3 privileges copy_from_auth and 657 copy_to_auth. If the server-to-server copy protocol is ONC RPC 658 based, the servers are also REQUIRED to implement the RPCSEC_GSSv3 659 privilege copy_confirm_auth. These requirements to implement are not 660 requirements to use. NFSv4 clients and servers are RECOMMENDED to 661 use [5] to secure server-side copy operations. 663 2.4.1. Inter-Server Copy Security 665 2.4.1.1. Requirements for Secure Inter-Server Copy 667 Inter-server copy is driven by several requirements: 669 o The specification MUST NOT mandate an inter-server copy protocol. 670 There are many ways to copy data. Some will be more optimal than 671 others depending on the identities of the source server and 672 destination server. For example the source and destination 673 servers might be two nodes sharing a common file system format for 674 the source and destination file systems. Thus the source and 675 destination are in an ideal position to efficiently render the 676 image of the source file to the destination file by replicating 677 the file system formats at the block level. In other cases, the 678 source and destination might be two nodes sharing a common storage 679 area network, and thus there is no need to copy any data at all, 680 and instead ownership of the file and its contents simply gets re- 681 assigned to the destination. 683 o The specification MUST provide guidance for using NFSv4.x as a 684 copy protocol. For those source and destination servers willing 685 to use NFSv4.x there are specific security considerations that 686 this specification can and does address. 688 o The specification MUST NOT mandate pre-configuration between the 689 source and destination server. Requiring that the source and 690 destination first have a "copying relationship" increases the 691 administrative burden. However the specification MUST NOT 692 preclude implementations that require pre-configuration. 694 o The specification MUST NOT mandate a trust relationship between 695 the source and destination server. The NFSv4 security model 696 requires mutual authentication between a principal on an NFS 697 client and a principal on an NFS server. This model MUST continue 698 with the introduction of COPY. 700 2.4.1.2. Inter-Server Copy with RPCSEC_GSSv3 702 When the client sends a COPY_NOTIFY to the source server to expect 703 the destination to attempt to copy data from the source server, it is 704 expected that this copy is being done on behalf of the principal 705 (called the "user principal") that sent the RPC request that encloses 706 the COMPOUND procedure that contains the COPY_NOTIFY operation. The 707 user principal is identified by the RPC credentials. A mechanism 708 that allows the user principal to authorize the destination server to 709 perform the copy in a manner that lets the source server properly 710 authenticate the destination's copy, and without allowing the 711 destination to exceed its authorization is necessary. 713 An approach that sends delegated credentials of the client's user 714 principal to the destination server is not used for the following 715 reasons. If the client's user delegated its credentials, the 716 destination would authenticate as the user principal. If the 717 destination were using the NFSv4 protocol to perform the copy, then 718 the source server would authenticate the destination server as the 719 user principal, and the file copy would securely proceed. However, 720 this approach would allow the destination server to copy other files. 721 The user principal would have to trust the destination server to not 722 do so. This is counter to the requirements, and therefore is not 723 considered. Instead an approach using RPCSEC_GSSv3 [5] privileges is 724 proposed. 726 One of the stated applications of the proposed RPCSEC_GSSv3 protocol 727 is compound client host and user authentication [+ privilege 728 assertion]. For inter-server file copy, we require compound NFS 729 server host and user authentication [+ privilege assertion]. The 730 distinction between the two is one without meaning. 732 RPCSEC_GSSv3 introduces the notion of privileges. We define three 733 privileges: 735 copy_from_auth: A user principal is authorizing a source principal 736 ("nfs@") to allow a destination principal ("nfs@ 737 ") to copy a file from the source to the destination. 738 This privilege is established on the source server before the user 739 principal sends a COPY_NOTIFY operation to the source server. 741 struct copy_from_auth_priv { 742 secret4 cfap_shared_secret; 743 netloc4 cfap_destination; 744 /* the NFSv4 user name that the user principal maps to */ 745 utf8str_mixed cfap_username; 746 /* equal to seq_num of rpc_gss_cred_vers_3_t */ 747 unsigned int cfap_seq_num; 748 }; 750 cfp_shared_secret is a secret value the user principal generates. 752 copy_to_auth: A user principal is authorizing a destination 753 principal ("nfs@") to allow it to copy a file from 754 the source to the destination. This privilege is established on 755 the destination server before the user principal sends a COPY 756 operation to the destination server. 758 struct copy_to_auth_priv { 759 /* equal to cfap_shared_secret */ 760 secret4 ctap_shared_secret; 761 netloc4 ctap_source; 762 /* the NFSv4 user name that the user principal maps to */ 763 utf8str_mixed ctap_username; 764 /* equal to seq_num of rpc_gss_cred_vers_3_t */ 765 unsigned int ctap_seq_num; 766 }; 768 ctap_shared_secret is a secret value the user principal generated 769 and was used to establish the copy_from_auth privilege with the 770 source principal. 772 copy_confirm_auth: A destination principal is confirming with the 773 source principal that it is authorized to copy data from the 774 source on behalf of the user principal. When the inter-server 775 copy protocol is NFSv4, or for that matter, any protocol capable 776 of being secured via RPCSEC_GSSv3 (i.e., any ONC RPC protocol), 777 this privilege is established before the file is copied from the 778 source to the destination. 780 struct copy_confirm_auth_priv { 781 /* equal to GSS_GetMIC() of cfap_shared_secret */ 782 opaque ccap_shared_secret_mic<>; 783 /* the NFSv4 user name that the user principal maps to */ 784 utf8str_mixed ccap_username; 785 /* equal to seq_num of rpc_gss_cred_vers_3_t */ 786 unsigned int ccap_seq_num; 787 }; 789 2.4.1.2.1. Establishing a Security Context 791 When the user principal wants to COPY a file between two servers, if 792 it has not established copy_from_auth and copy_to_auth privileges on 793 the servers, it establishes them: 795 o The user principal generates a secret it will share with the two 796 servers. This shared secret will be placed in the 797 cfap_shared_secret and ctap_shared_secret fields of the 798 appropriate privilege data types, copy_from_auth_priv and 799 copy_to_auth_priv. 801 o An instance of copy_from_auth_priv is filled in with the shared 802 secret, the destination server, and the NFSv4 user id of the user 803 principal. It will be sent with an RPCSEC_GSS3_CREATE procedure, 804 and so cfap_seq_num is set to the seq_num of the credential of the 805 RPCSEC_GSS3_CREATE procedure. Because cfap_shared_secret is a 806 secret, after XDR encoding copy_from_auth_priv, GSS_Wrap() (with 807 privacy) is invoked on copy_from_auth_priv. The 808 RPCSEC_GSS3_CREATE procedure's arguments are: 810 struct { 811 rpc_gss3_gss_binding *compound_binding; 812 rpc_gss3_chan_binding *chan_binding_mic; 813 rpc_gss3_assertion assertions<>; 814 rpc_gss3_extension extensions<>; 815 } rpc_gss3_create_args; 817 The string "copy_from_auth" is placed in assertions[0].privs. The 818 output of GSS_Wrap() is placed in extensions[0].data. The field 819 extensions[0].critical is set to TRUE. The source server calls 820 GSS_Unwrap() on the privilege, and verifies that the seq_num 821 matches the credential. It then verifies that the NFSv4 user id 822 being asserted matches the source server's mapping of the user 823 principal. If it does, the privilege is established on the source 824 server as: <"copy_from_auth", user id, destination>. The 825 successful reply to RPCSEC_GSS3_CREATE has: 827 struct { 828 opaque handle<>; 829 rpc_gss3_chan_binding *chan_binding_mic; 830 rpc_gss3_assertion granted_assertions<>; 831 rpc_gss3_assertion server_assertions<>; 832 rpc_gss3_extension extensions<>; 833 } rpc_gss3_create_res; 835 The field "handle" is the RPCSEC_GSSv3 handle that the client will 836 use on COPY_NOTIFY requests involving the source and destination 837 server. granted_assertions[0].privs will be equal to 838 "copy_from_auth". The server will return a GSS_Wrap() of 839 copy_to_auth_priv. 841 o An instance of copy_to_auth_priv is filled in with the shared 842 secret, the source server, and the NFSv4 user id. It will be sent 843 with an RPCSEC_GSS3_CREATE procedure, and so ctap_seq_num is set 844 to the seq_num of the credential of the RPCSEC_GSS3_CREATE 845 procedure. Because ctap_shared_secret is a secret, after XDR 846 encoding copy_to_auth_priv, GSS_Wrap() is invoked on 847 copy_to_auth_priv. The RPCSEC_GSS3_CREATE procedure's arguments 848 are: 850 struct { 851 rpc_gss3_gss_binding *compound_binding; 852 rpc_gss3_chan_binding *chan_binding_mic; 853 rpc_gss3_assertion assertions<>; 854 rpc_gss3_extension extensions<>; 855 } rpc_gss3_create_args; 857 The string "copy_to_auth" is placed in assertions[0].privs. The 858 output of GSS_Wrap() is placed in extensions[0].data. The field 859 extensions[0].critical is set to TRUE. After unwrapping, 860 verifying the seq_num, and the user principal to NFSv4 user ID 861 mapping, the destination establishes a privilege of 862 <"copy_to_auth", user id, source>. The successful reply to 863 RPCSEC_GSS3_CREATE has: 865 struct { 866 opaque handle<>; 867 rpc_gss3_chan_binding *chan_binding_mic; 868 rpc_gss3_assertion granted_assertions<>; 869 rpc_gss3_assertion server_assertions<>; 870 rpc_gss3_extension extensions<>; 871 } rpc_gss3_create_res; 873 The field "handle" is the RPCSEC_GSSv3 handle that the client will 874 use on COPY requests involving the source and destination server. 875 The field granted_assertions[0].privs will be equal to 876 "copy_to_auth". The server will return a GSS_Wrap() of 877 copy_to_auth_priv. 879 2.4.1.2.2. Starting a Secure Inter-Server Copy 881 When the client sends a COPY_NOTIFY request to the source server, it 882 uses the privileged "copy_from_auth" RPCSEC_GSSv3 handle. 883 cna_destination_server in COPY_NOTIFY MUST be the same as the name of 884 the destination server specified in copy_from_auth_priv. Otherwise, 885 COPY_NOTIFY will fail with NFS4ERR_ACCESS. The source server 886 verifies that the privilege <"copy_from_auth", user id, destination> 887 exists, and annotates it with the source filehandle, if the user 888 principal has read access to the source file, and if administrative 889 policies give the user principal and the NFS client read access to 890 the source file (i.e., if the ACCESS operation would grant read 891 access). Otherwise, COPY_NOTIFY will fail with NFS4ERR_ACCESS. 893 When the client sends a COPY request to the destination server, it 894 uses the privileged "copy_to_auth" RPCSEC_GSSv3 handle. 895 ca_source_server in COPY MUST be the same as the name of the source 896 server specified in copy_to_auth_priv. Otherwise, COPY will fail 897 with NFS4ERR_ACCESS. The destination server verifies that the 898 privilege <"copy_to_auth", user id, source> exists, and annotates it 899 with the source and destination filehandles. If the client has 900 failed to establish the "copy_to_auth" policy it will reject the 901 request with NFS4ERR_PARTNER_NO_AUTH. 903 If the client sends a COPY_REVOKE to the source server to rescind the 904 destination server's copy privilege, it uses the privileged 905 "copy_from_auth" RPCSEC_GSSv3 handle and the cra_destination_server 906 in COPY_REVOKE MUST be the same as the name of the destination server 907 specified in copy_from_auth_priv. The source server will then delete 908 the <"copy_from_auth", user id, destination> privilege and fail any 909 subsequent copy requests sent under the auspices of this privilege 910 from the destination server. 912 2.4.1.2.3. Securing ONC RPC Server-to-Server Copy Protocols 914 After a destination server has a "copy_to_auth" privilege established 915 on it, and it receives a COPY request, if it knows it will use an ONC 916 RPC protocol to copy data, it will establish a "copy_confirm_auth" 917 privilege on the source server, using nfs@ as the 918 initiator principal, and nfs@ as the target principal. 920 The value of the field ccap_shared_secret_mic is a GSS_VerifyMIC() of 921 the shared secret passed in the copy_to_auth privilege. The field 922 ccap_username is the mapping of the user principal to an NFSv4 user 923 name ("user"@"domain" form), and MUST be the same as ctap_username 924 and cfap_username. The field ccap_seq_num is the seq_num of the 925 RPCSEC_GSSv3 credential used for the RPCSEC_GSS3_CREATE procedure the 926 destination will send to the source server to establish the 927 privilege. 929 The source server verifies the privilege, and establishes a 930 <"copy_confirm_auth", user id, destination> privilege. If the source 931 server fails to verify the privilege, the COPY operation will be 932 rejected with NFS4ERR_PARTNER_NO_AUTH. All subsequent ONC RPC 933 requests sent from the destination to copy data from the source to 934 the destination will use the RPCSEC_GSSv3 handle returned by the 935 source's RPCSEC_GSS3_CREATE response. 937 Note that the use of the "copy_confirm_auth" privilege accomplishes 938 the following: 940 o if a protocol like NFS is being used, with export policies, export 941 policies can be overridden in case the destination server as-an- 942 NFS-client is not authorized 944 o manual configuration to allow a copy relationship between the 945 source and destination is not needed. 947 If the attempt to establish a "copy_confirm_auth" privilege fails, 948 then when the user principal sends a COPY request to destination, the 949 destination server will reject it with NFS4ERR_PARTNER_NO_AUTH. 951 2.4.1.2.4. Securing Non ONC RPC Server-to-Server Copy Protocols 953 If the destination won't be using ONC RPC to copy the data, then the 954 source and destination are using an unspecified copy protocol. The 955 destination could use the shared secret and the NFSv4 user id to 956 prove to the source server that the user principal has authorized the 957 copy. 959 For protocols that authenticate user names with passwords (e.g., HTTP 960 [13] and FTP [14]), the nfsv4 user id could be used as the user name, 961 and an ASCII hexadecimal representation of the RPCSEC_GSSv3 shared 962 secret could be used as the user password or as input into non- 963 password authentication methods like CHAP [15]. 965 2.4.1.3. Inter-Server Copy via ONC RPC but without RPCSEC_GSSv3 967 ONC RPC security flavors other than RPCSEC_GSSv3 MAY be used with the 968 server-side copy offload operations described in this document. In 969 particular, host-based ONC RPC security flavors such as AUTH_NONE and 970 AUTH_SYS MAY be used. If a host-based security flavor is used, a 971 minimal level of protection for the server-to-server copy protocol is 972 possible. 974 In the absence of strong security mechanisms such as RPCSEC_GSSv3, 975 the challenge is how the source server and destination server 976 identify themselves to each other, especially in the presence of 977 multi-homed source and destination servers. In a multi-homed 978 environment, the destination server might not contact the source 979 server from the same network address specified by the client in the 980 COPY_NOTIFY. This can be overcome using the procedure described 981 below. 983 When the client sends the source server the COPY_NOTIFY operation, 984 the source server may reply to the client with a list of target 985 addresses, names, and/or URLs and assign them to the unique 986 quadruple: . If the destination uses one of these target netlocs to contact 988 the source server, the source server will be able to uniquely 989 identify the destination server, even if the destination server does 990 not connect from the address specified by the client in COPY_NOTIFY. 991 The level of assurance in this identification depends on the 992 unpredictability, strength and secrecy of the random number. 994 For example, suppose the network topology is as shown in Figure 3. 995 If the source filehandle is 0x12345, the source server may respond to 996 a COPY_NOTIFY for destination 10.11.78.56 with the URLs: 998 nfs://10.11.78.18//_COPY/FvhH1OKbu8VrxvV1erdjvR7N/10.11.78.56/_FH/ 999 0x12345 1001 nfs://192.168.33.18//_COPY/FvhH1OKbu8VrxvV1erdjvR7N/10.11.78.56/ 1002 _FH/0x12345 1004 The name component after _COPY is 24 characters of base 64, more than 1005 enough to encode a 128 bit random number. 1007 The client will then send these URLs to the destination server in the 1008 COPY operation. Suppose that the 192.168.33.0/24 network is a high 1009 speed network and the destination server decides to transfer the file 1010 over this network. If the destination contacts the source server 1011 from 192.168.33.56 over this network using NFSv4.1, it does the 1012 following: 1014 COMPOUND { PUTROOTFH, LOOKUP "_COPY" ; LOOKUP 1015 "FvhH1OKbu8VrxvV1erdjvR7N" ; LOOKUP "10.11.78.56"; LOOKUP "_FH" ; 1016 OPEN "0x12345" ; GETFH } 1018 Provided that the random number is unpredictable and has been kept 1019 secret by the parties involved, the source server will therefore know 1020 that these NFSv4.x operations are being issued by the destination 1021 server identified in the COPY_NOTIFY. This random number technique 1022 only provides initial authentication of the destination server, and 1023 cannot defend against man-in-the-middle attacks after authentication 1024 or an eavesdropper that observes the random number on the wire. 1025 Other secure communication techniques (e.g., IPsec) are necessary to 1026 block these attacks. 1028 2.4.1.4. Inter-Server Copy without ONC RPC and RPCSEC_GSSv3 1030 The same techniques as Section 2.4.1.3, using unique URLs for each 1031 destination server, can be used for other protocols (e.g., HTTP [13] 1032 and FTP [14]) as well. 1034 3. Sparse Files 1036 3.1. Introduction 1038 A sparse file is a common way of representing a large file without 1039 having to utilize all of the disk space for it. Consequently, a 1040 sparse file uses less physical space than its size indicates. This 1041 means the file contains 'holes', byte ranges within the file that 1042 contain no data. Most modern file systems support sparse files, 1043 including most UNIX file systems and NTFS, but notably not Apple's 1044 HFS+. Common examples of sparse files include Virtual Machine (VM) 1045 OS/disk images, database files, log files, and even checkpoint 1046 recovery files most commonly used by the HPC community. 1048 If an application reads a hole in a sparse file, the file system must 1049 return all zeros to the application. For local data access there is 1050 little penalty, but with NFS these zeroes must be transferred back to 1051 the client. If an application uses the NFS client to read data into 1052 memory, this wastes time and bandwidth as the application waits for 1053 the zeroes to be transferred. 1055 A sparse file is typically created by initializing the file to be all 1056 zeros - nothing is written to the data in the file, instead the hole 1057 is recorded in the metadata for the file. So a 8G disk image might 1058 be represented initially by a couple hundred bits in the inode and 1059 nothing on the disk. If the VM then writes 100M to a file in the 1060 middle of the image, there would now be two holes represented in the 1061 metadata and 100M in the data. 1063 This section introduces a new operation READ_PLUS (Section 13.10) 1064 which supports all the features of READ but includes an extension to 1065 support sparse pattern files. READ_PLUS is guaranteed to perform no 1066 worse than READ, and can dramatically improve performance with sparse 1067 files. READ_PLUS does not depend on pNFS protocol features, but can 1068 be used by pNFS to support sparse files. 1070 3.2. Terminology 1072 Regular file: An object of file type NF4REG or NF4NAMEDATTR. 1074 Sparse file: A Regular file that contains one or more Holes. 1076 Hole: A byte range within a Sparse file that contains regions of all 1077 zeroes. For block-based file systems, this could also be an 1078 unallocated region of the file. 1080 Hole Threshold: The minimum length of a Hole as determined by the 1081 server. If a server chooses to define a Hole Threshold, then it 1082 would not return hole information about holes with a length 1083 shorter than the Hole Threshold. 1085 4. Space Reservation 1087 4.1. Introduction 1089 This section describes a set of operations that allow applications 1090 such as hypervisors to reserve space for a file, report the amount of 1091 actual disk space a file occupies and freeup the backing space of a 1092 file when it is not required. In virtualized environments, virtual 1093 disk files are often stored on NFS mounted volumes. Since virtual 1094 disk files represent the hard disks of virtual machines, hypervisors 1095 often have to guarantee certain properties for the file. 1097 One such example is space reservation. When a hypervisor creates a 1098 virtual disk file, it often tries to preallocate the space for the 1099 file so that there are no future allocation related errors during the 1100 operation of the virtual machine. Such errors prevent a virtual 1101 machine from continuing execution and result in downtime. 1103 Currently, in order to achieve such a guarantee, applications zero 1104 the entire file. The initial zeroing allocates the backing blocks 1105 and all subsequent writes are overwrites of already allocated blocks. 1106 This approach is not only inefficient in terms of the amount of I/O 1107 done, it is also not guaranteed to work on filesystems that are log 1108 structured or deduplicated. An efficient way of guaranteeing space 1109 reservation would be beneficial to such applications. 1111 If the space_reserved attribute is set on a file, it is guaranteed 1112 that writes that do not grow the file will not fail with 1113 NFSERR_NOSPC. 1115 Another useful feature would be the ability to report the number of 1116 blocks that would be freed when a file is deleted. Currently, NFS 1117 reports two size attributes: 1119 size The logical file size of the file. 1121 space_used The size in bytes that the file occupies on disk 1123 While these attributes are sufficient for space accounting in 1124 traditional filesystems, they prove to be inadequate in modern 1125 filesystems that support block sharing. In such filesystems, 1126 multiple inodes can point to a single block with a block reference 1127 count to guard against premature freeing. Having a way to tell the 1128 number of blocks that would be freed if the file was deleted would be 1129 useful to applications that wish to migrate files when a volume is 1130 low on space. 1132 Since virtual disks represent a hard drive in a virtual machine, a 1133 virtual disk can be viewed as a filesystem within a file. Since not 1134 all blocks within a filesystem are in use, there is an opportunity to 1135 reclaim blocks that are no longer in use. A call to deallocate 1136 blocks could result in better space efficiency. Lesser space MAY be 1137 consumed for backups after block deallocation. 1139 The following operations and attributes can be used to resolve this 1140 issues: 1142 space_reserved This attribute specifies whether the blocks backing 1143 the file have been preallocated. 1145 space_freed This attribute specifies the space freed when a file is 1146 deleted, taking block sharing into consideration. 1148 INITIALIZED This operation zeroes and/or deallocates the blocks 1149 backing a region of the file. 1151 If space_used of a file is interpreted to mean the size in bytes of 1152 all disk blocks pointed to by the inode of the file, then shared 1153 blocks get double counted, over-reporting the space utilization. 1154 This also has the adverse effect that the deletion of a file with 1155 shared blocks frees up less than space_used bytes. 1157 On the other hand, if space_used is interpreted to mean the size in 1158 bytes of those disk blocks unique to the inode of the file, then 1159 shared blocks are not counted in any file, resulting in under- 1160 reporting of the space utilization. 1162 For example, two files A and B have 10 blocks each. Let 6 of these 1163 blocks be shared between them. Thus, the combined space utilized by 1164 the two files is 14 * BLOCK_SIZE bytes. In the former case, the 1165 combined space utilization of the two files would be reported as 20 * 1166 BLOCK_SIZE. However, deleting either would only result in 4 * 1167 BLOCK_SIZE being freed. Conversely, the latter interpretation would 1168 report that the space utilization is only 8 * BLOCK_SIZE. 1170 Adding another size attribute, space_freed, is helpful in solving 1171 this problem. space_freed is the number of blocks that are allocated 1172 to the given file that would be freed on its deletion. In the 1173 example, both A and B would report space_freed as 4 * BLOCK_SIZE and 1174 space_used as 10 * BLOCK_SIZE. If A is deleted, B will report 1175 space_freed as 10 * BLOCK_SIZE as the deletion of B would result in 1176 the deallocation of all 10 blocks. 1178 The addition of this problem doesn't solve the problem of space being 1179 over-reported. However, over-reporting is better than under- 1180 reporting. 1182 5. Support for Application IO Hints 1184 5.1. Introduction 1186 Applications currently have several options for communicating I/O 1187 access patterns to the NFS client. While this can help the NFS 1188 client optimize I/O and caching for a file, it does not allow the NFS 1189 server and its exported file system to do likewise. Therefore, here 1190 we put forth a proposal for the NFSv4.2 protocol to allow 1191 applications to communicate their expected behavior to the server. 1193 By communicating expected access pattern, e.g., sequential or random, 1194 and data re-use behavior, e.g., data range will be read multiple 1195 times and should be cached, the server will be able to better 1196 understand what optimizations it should implement for access to a 1197 file. For example, if a application indicates it will never read the 1198 data more than once, then the file system can avoid polluting the 1199 data cache and not cache the data. 1201 The first application that can issue client I/O hints is the 1202 posix_fadvise operation. For example, on Linux, when an application 1203 uses posix_fadvise to specify a file will be read sequentially, Linux 1204 doubles the readahead buffer size. 1206 Another instance where applications provide an indication of their 1207 desired I/O behavior is the use of direct I/O. By specifying direct 1208 I/O, clients will no longer cache data, but this information is not 1209 passed to the server, which will continue caching data. 1211 Application specific NFS clients such as those used by hypervisors 1212 and databases can also leverage application hints to communicate 1213 their specialized requirements. 1215 This section adds a new IO_ADVISE operation to communicate the client 1216 file access patterns to the NFS server. The NFS server upon 1217 receiving a IO_ADVISE operation MAY choose to alter its I/O and 1218 caching behavior, but is under no obligation to do so. 1220 5.2. POSIX Requirements 1222 The first key requirement of the IO_ADVISE operation is to support 1223 the posix_fadvise function [6], which is supported in Linux and many 1224 other operating systems. Examples and guidance on how to use 1225 posix_fadvise to improve performance can be found here [16]. 1226 posix_fadvise is defined as follows, 1228 int posix_fadvise(int fd, off_t offset, off_t len, int advice); 1230 The posix_fadvise() function shall advise the implementation on the 1231 expected behavior of the application with respect to the data in the 1232 file associated with the open file descriptor, fd, starting at offset 1233 and continuing for len bytes. The specified range need not currently 1234 exist in the file. If len is zero, all data following offset is 1235 specified. The implementation may use this information to optimize 1236 handling of the specified data. The posix_fadvise() function shall 1237 have no effect on the semantics of other operations on the specified 1238 data, although it may affect the performance of other operations. 1240 The advice to be applied to the data is specified by the advice 1241 parameter and may be one of the following values: 1243 POSIX_FADV_NORMAL - Specifies that the application has no advice to 1244 give on its behavior with respect to the specified data. It is 1245 the default characteristic if no advice is given for an open file. 1247 POSIX_FADV_SEQUENTIAL - Specifies that the application expects to 1248 access the specified data sequentially from lower offsets to 1249 higher offsets. 1251 POSIX_FADV_RANDOM - Specifies that the application expects to access 1252 the specified data in a random order. 1254 POSIX_FADV_WILLNEED - Specifies that the application expects to 1255 access the specified data in the near future. 1257 POSIX_FADV_DONTNEED - Specifies that the application expects that it 1258 will not access the specified data in the near future. 1260 POSIX_FADV_NOREUSE - Specifies that the application expects to 1261 access the specified data once and then not reuse it thereafter. 1263 Upon successful completion, posix_fadvise() shall return zero; 1264 otherwise, an error number shall be returned to indicate the error. 1266 5.3. Additional Requirements 1268 Many use cases exist for sending application I/O hints to the server 1269 that cannot utilize the POSIX supported interface. This is because 1270 some applications may benefit from additional hints not specified by 1271 posix_fadvise, and some applications may not use POSIX altogether. 1273 One use case is "Opportunistic Prefetch", which allows a stateid 1274 holder to tell the server that it is possible that it will access the 1275 specified data in the near future. This is similar to 1276 POSIX_FADV_WILLNEED, but the client is unsure it will in fact read 1277 the specified data, so the server should only prefetch the data if it 1278 can be done at a marginal cost. For example, when a server receives 1279 this hint, it could prefetch only the indirect blocks for a file 1280 instead of all the data. This would still improve performance if the 1281 client does read the data, but with less pressure on server memory. 1283 An example use case for this hint is a database that reads in a 1284 single record that points to additional records in either other areas 1285 of the same file or different files located on the same or different 1286 server. While it is likely that the application may access the 1287 additional records, it is far from guaranteed. Therefore, the 1288 database may issue an opportunistic prefetch (instead of 1289 POSIX_FADV_WILLNEED) for the data in the other files pointed to by 1290 the record. 1292 Another use case is "Direct I/O", which allows a stated holder to 1293 inform the server that it does not wish to cache data. Today, for 1294 applications that only intend to read data once, the use of direct 1295 I/O disables client caching, but does not affect server caching. By 1296 caching data that will not be re-read, the server is polluting its 1297 cache and possibly causing useful cached data to be evicted. By 1298 informing the server of its expected I/O access, this situation can 1299 be avoid. Direct I/O can be used in Linux and AIX via the open() 1300 O_DIRECT parameter, in Solaris via the directio() function, and in 1301 Windows via the CreateFile() FILE_FLAG_NO_BUFFERING flag. 1303 Another use case is "Backward Sequential Read", which allows a stated 1304 holder to inform the server that it intends to read the specified 1305 data backwards, i.e., back the end to the beginning. This is 1306 different than POSIX_FADV_SEQUENTIAL, whose implied intention was 1307 that data will be read from beginning to end. This hint allows 1308 servers to prefetch data at the end of the range first, and then 1309 prefetch data sequentially in a backwards manner to the start of the 1310 data range. One example of an application that can make use of this 1311 hint is video editing. 1313 5.4. Security Considerations 1315 None. 1317 5.5. IANA Considerations 1319 The IO_ADVISE_type4 will be extended through an IANA registry. 1321 6. Application Data Block Support 1323 At the OS level, files are contained on disk blocks. Applications 1324 are also free to impose structure on the data contained in a file and 1325 we can define an Application Data Block (ADB) to be such a structure. 1326 From the application's viewpoint, it only wants to handle ADBs and 1327 not raw bytes (see [17]). An ADB is typically comprised of two 1328 sections: a header and data. The header describes the 1329 characteristics of the block and can provide a means to detect 1330 corruption in the data payload. The data section is typically 1331 initialized to all zeros. 1333 The format of the header is application specific, but there are two 1334 main components typically encountered: 1336 1. An ADB Number (ADBN), which allows the application to determine 1337 which data block is being referenced. The ADBN is a logical 1338 block number and is useful when the client is not storing the 1339 blocks in contiguous memory. 1341 2. Fields to describe the state of the ADB and a means to detect 1342 block corruption. For both pieces of data, a useful property is 1343 that allowed values be unique in that if passed across the 1344 network, corruption due to translation between big and little 1345 endian architectures are detectable. For example, 0xF0DEDEF0 has 1346 the same bit pattern in both architectures. 1348 Applications already impose structures on files [17] and detect 1349 corruption in data blocks [18]. What they are not able to do is 1350 efficiently transfer and store ADBs. To initialize a file with ADBs, 1351 the client must send the full ADB to the server and that must be 1352 stored on the server. When the application is initializing a file to 1353 have the ADB structure, it could compress the ADBs to just the 1354 information to necessary to later reconstruct the header portion of 1355 the ADB when the contents are read back. Using sparse file 1356 techniques, the disk blocks described by would not be allocated. 1357 Unlike sparse file techniques, there would be a small cost to store 1358 the compressed header data. 1360 In this section, we are going to define a generic framework for an 1361 ADB, present one approach to detecting corruption in a given ADB 1362 implementation, and describe the model for how the client and server 1363 can support efficient initialization of ADBs, reading of ADB holes, 1364 punching holes in ADBs, and space reservation. Further, we need to 1365 be able to extend this model to applications which do not support 1366 ADBs, but wish to be able to handle sparse files, hole punching, and 1367 space reservation. 1369 6.1. Generic Framework 1371 We want the representation of the ADB to be flexible enough to 1372 support many different applications. The most basic approach is no 1373 imposition of a block at all, which means we are working with the raw 1374 bytes. Such an approach would be useful for storing holes, punching 1375 holes, etc. In more complex deployments, a server might be 1376 supporting multiple applications, each with their own definition of 1377 the ADB. One might store the ADBN at the start of the block and then 1378 have a guard pattern to detect corruption [19]. The next might store 1379 the ADBN at an offset of 100 bytes within the block and have no guard 1380 pattern at all. The point is that existing applications might 1381 already have well defined formats for their data blocks. 1383 The guard pattern can be used to represent the state of the block, to 1384 protect against corruption, or both. Again, it needs to be able to 1385 be placed anywhere within the ADB. 1387 We need to be able to represent the starting offset of the block and 1388 the size of the block. Note that nothing prevents the application 1389 from defining different sized blocks in a file. 1391 6.1.1. Data Block Representation 1393 struct app_data_block4 { 1394 offset4 adb_offset; 1395 length4 adb_block_size; 1396 length4 adb_block_count; 1397 length4 adb_reloff_blocknum; 1398 count4 adb_block_num; 1399 length4 adb_reloff_pattern; 1400 opaque adb_pattern<>; 1401 }; 1403 The app_data_block4 structure captures the abstraction presented for 1404 the ADB. The additional fields present are to allow the transmission 1405 of adb_block_count ADBs at one time. We also use adb_block_num to 1406 convey the ADBN of the first block in the sequence. Each ADB will 1407 contain the same adb_pattern string. 1409 As both adb_block_num and adb_pattern are optional, if either 1410 adb_reloff_pattern or adb_reloff_blocknum is set to NFS4_UINT64_MAX, 1411 then the corresponding field is not set in any of the ADB. 1413 6.1.2. Data Content 1415 /* 1416 * Use an enum such that we can extend new types. 1417 */ 1418 enum data_content4 { 1419 NFS4_CONTENT_DATA = 0, 1420 NFS4_CONTENT_APP_BLOCK = 1, 1421 NFS4_CONTENT_HOLE = 2 1422 }; 1424 New operations might need to differentiate between wanting to access 1425 data versus an ADB. Also, future minor versions might want to 1426 introduce new data formats. This enumeration allows that to occur. 1428 6.2. pNFS Considerations 1430 While this document does not mandate how sparse ADBs are recorded on 1431 the server, it does make the assumption that such information is not 1432 in the file. I.e., the information is metadata. As such, the 1433 INITIALIZE operation is defined to be not supported by the DS - it 1434 must be issued to the MDS. But since the client must not assume a 1435 priori whether a read is sparse or not, the READ_PLUS operation MUST 1436 be supported by both the DS and the MDS. I.e., the client might 1437 impose on the MDS to asynchronously read the data from the DS. 1439 Furthermore, each DS MUST not report to a client either a sparse ADB 1440 or data which belongs to another DS. One implication of this 1441 requirement is that the app_data_block4's adb_block_size MUST be 1442 either be the stripe width or the stripe width must be an even 1443 multiple of it. 1445 The second implication here is that the DS must be able to use the 1446 Control Protocol to determine from the MDS where the sparse ADBs 1447 occur. [[Comment.3: Need to discuss what happens if after the file 1448 is being written to and an INITIALIZE occurs? --TH]] Perhaps instead 1449 of the DS pulling from the MDS, the MDS pushes to the DS? Thus an 1450 INITIALIZE causes a new push? [[Comment.4: Still need to consider 1451 race cases of the DS getting a WRITE and the MDS getting an 1452 INITIALIZE. --TH]] 1454 6.3. An Example of Detecting Corruption 1456 In this section, we define an ADB format in which corruption can be 1457 detected. Note that this is just one possible format and means to 1458 detect corruption. 1460 Consider a very basic implementation of an operating system's disk 1461 blocks. A block is either data or it is an indirect block which 1462 allows for files to be larger than one block. It is desired to be 1463 able to initialize a block. Lastly, to quickly unlink a file, a 1464 block can be marked invalid. The contents remain intact - which 1465 would enable this OS application to undelete a file. 1467 The application defines 4k sized data blocks, with an 8 byte block 1468 counter occurring at offset 0 in the block, and with the guard 1469 pattern occurring at offset 8 inside the block. Furthermore, the 1470 guard pattern can take one of four states: 1472 0xfeedface - This is the FREE state and indicates that the ADB 1473 format has been applied. 1475 0xcafedead - This is the DATA state and indicates that real data 1476 has been written to this block. 1478 0xe4e5c001 - This is the INDIRECT state and indicates that the 1479 block contains block counter numbers that are chained off of this 1480 block. 1482 0xba1ed4a3 - This is the INVALID state and indicates that the block 1483 contains data whose contents are garbage. 1485 Finally, it also defines an 8 byte checksum [20] starting at byte 16 1486 which applies to the remaining contents of the block. If the state 1487 is FREE, then that checksum is trivially zero. As such, the 1488 application has no need to transfer the checksum implicitly inside 1489 the ADB - it need not make the transfer layer aware of the fact that 1490 there is a checksum (see [18] for an example of checksums used to 1491 detect corruption in application data blocks). 1493 Corruption in each ADB can be detected thusly: 1495 o If the guard pattern is anything other than one of the allowed 1496 values, including all zeros. 1498 o If the guard pattern is FREE and any other byte in the remainder 1499 of the ADB is anything other than zero. 1501 o If the guard pattern is anything other than FREE, then if the 1502 stored checksum does not match the computed checksum. 1504 o If the guard pattern is INDIRECT and one of the stored indirect 1505 block numbers has a value greater than the number of ADBs in the 1506 file. 1508 o If the guard pattern is INDIRECT and one of the stored indirect 1509 block numbers is a duplicate of another stored indirect block 1510 number. 1512 As can be seen, the application can detect errors based on the 1513 combination of the guard pattern state and the checksum. But also, 1514 the application can detect corruption based on the state and the 1515 contents of the ADB. This last point is important in validating the 1516 minimum amount of data we incorporated into our generic framework. 1517 I.e., the guard pattern is sufficient in allowing applications to 1518 design their own corruption detection. 1520 Finally, it is important to note that none of these corruption checks 1521 occur in the transport layer. The server and client components are 1522 totally unaware of the file format and might report everything as 1523 being transferred correctly even in the case the application detects 1524 corruption. 1526 6.4. Example of READ_PLUS 1528 The hypothetical application presented in Section 6.3 can be used to 1529 illustrate how READ_PLUS would return an array of results. A file is 1530 created and initialized with 100 4k ADBs in the FREE state: 1532 INITIALIZE {0, 4k, 100, 0, 0, 8, 0xfeedface} 1534 Further, assume the application writes a single ADB at 16k, changing 1535 the guard pattern to 0xcafedead, we would then have in memory: 1537 0 -> (16k - 1) : 4k, 4, 0, 0, 8, 0xfeedface 1538 16k -> (20k - 1) : 00 00 00 05 ca fe de ad XX XX ... XX XX 1539 20k -> 400k : 4k, 95, 0, 6, 0xfeedface 1541 And when the client did a READ_PLUS of 64k at the start of the file, 1542 it would get back a result of an ADB, some data, and a final ADB: 1544 ADB {0, 4, 0, 0, 8, 0xfeedface} 1545 data 4k 1546 ADB {20k, 4k, 59, 0, 6, 0xfeedface} 1548 6.5. Zero Filled Holes 1550 As applications are free to define the structure of an ADB, it is 1551 trivial to define an ADB which supports zero filled holes. Such a 1552 case would encompass the traditional definitions of a sparse file and 1553 hole punching. For example, to punch a 64k hole, starting at 100M, 1554 into an existing file which has no ADB structure: 1556 INITIALIZE {100M, 64k, 1, NFS4_UINT64_MAX, 1557 0, NFS4_UINT64_MAX, 0x0} 1559 7. Labeled NFS 1561 7.1. Introduction 1563 Access control models such as Unix permissions or Access Control 1564 Lists are commonly referred to as Discretionary Access Control (DAC) 1565 models. These systems base their access decisions on user identity 1566 and resource ownership. In contrast Mandatory Access Control (MAC) 1567 models base their access control decisions on the label on the 1568 subject (usually a process) and the object it wishes to access [7]. 1569 These labels may contain user identity information but usually 1570 contain additional information. In DAC systems users are free to 1571 specify the access rules for resources that they own. MAC models 1572 base their security decisions on a system wide policy established by 1573 an administrator or organization which the users do not have the 1574 ability to override. In this section, we add a MAC model to NFSv4. 1576 The first change necessary is to devise a method for transporting and 1577 storing security label data on NFSv4 file objects. Security labels 1578 have several semantics that are met by NFSv4 recommended attributes 1579 such as the ability to set the label value upon object creation. 1580 Access control on these attributes are done through a combination of 1581 two mechanisms. As with other recommended attributes on file objects 1582 the usual DAC checks (ACLs and permission bits) will be performed to 1583 ensure that proper file ownership is enforced. In addition a MAC 1584 system MAY be employed on the client, server, or both to enforce 1585 additional policy on what subjects may modify security label 1586 information. 1588 The second change is to provide a method for the server to notify the 1589 client that the attribute changed on an open file on the server. If 1590 the file is closed, then during the open attempt, the client will 1591 gather the new attribute value. The server MUST not communicate the 1592 new value of the attribute, the client MUST query it. This 1593 requirement stems from the need for the client to provide sufficient 1594 access rights to the attribute. 1596 The final change necessary is a modification to the RPC layer used in 1597 NFSv4 in the form of a new version of the RPCSEC_GSS [8] framework. 1598 In order for an NFSv4 server to apply MAC checks it must obtain 1599 additional information from the client. Several methods were 1600 explored for performing this and it was decided that the best 1601 approach was to incorporate the ability to make security attribute 1602 assertions through the RPC mechanism. RPCSECGSSv3 [5] outlines a 1603 method to assert additional security information such as security 1604 labels on gss context creation and have that data bound to all RPC 1605 requests that make use of that context. 1607 7.2. Definitions 1609 Label Format Specifier (LFS): is an identifier used by the client to 1610 establish the syntactic format of the security label and the 1611 semantic meaning of its components. These specifiers exist in a 1612 registry associated with documents describing the format and 1613 semantics of the label. 1615 Label Format Registry: is the IANA registry containing all 1616 registered LFS along with references to the documents that 1617 describe the syntactic format and semantics of the security label. 1619 Policy Identifier (PI): is an optional part of the definition of a 1620 Label Format Specifier which allows for clients and server to 1621 identify specific security policies. 1623 Object: is a passive resource within the system that we wish to be 1624 protected. Objects can be entities such as files, directories, 1625 pipes, sockets, and many other system resources relevant to the 1626 protection of the system state. 1628 Subject: A subject is an active entity usually a process which is 1629 requesting access to an object. 1631 Multi-Level Security (MLS): is a traditional model where objects are 1632 given a sensitivity level (Unclassified, Secret, Top Secret, etc) 1633 and a category set [21]. 1635 7.3. MAC Security Attribute 1637 MAC models base access decisions on security attributes bound to 1638 subjects and objects. This information can range from a user 1639 identity for an identity based MAC model, sensitivity levels for 1640 Multi-level security, or a type for Type Enforcement. These models 1641 base their decisions on different criteria but the semantics of the 1642 security attribute remain the same. The semantics required by the 1643 security attributes are listed below: 1645 o Must provide flexibility with respect to MAC model. 1647 o Must provide the ability to atomically set security information 1648 upon object creation. 1650 o Must provide the ability to enforce access control decisions both 1651 on the client and the server. 1653 o Must not expose an object to either the client or server name 1654 space before its security information has been bound to it. 1656 NFSv4 implements the security attribute as a recommended attribute. 1657 These attributes have a fixed format and semantics, which conflicts 1658 with the flexible nature of the security attribute. To resolve this 1659 the security attribute consists of two components. The first 1660 component is a LFS as defined in [22] to allow for interoperability 1661 between MAC mechanisms. The second component is an opaque field 1662 which is the actual security attribute data. To allow for various 1663 MAC models NFSv4 should be used solely as a transport mechanism for 1664 the security attribute. It is the responsibility of the endpoints to 1665 consume the security attribute and make access decisions based on 1666 their respective models. In addition, creation of objects through 1667 OPEN and CREATE allows for the security attribute to be specified 1668 upon creation. By providing an atomic create and set operation for 1669 the security attribute it is possible to enforce the second and 1670 fourth requirements. The recommended attribute FATTR4_SEC_LABEL will 1671 be used to satisfy this requirement. 1673 7.3.1. Interpreting FATTR4_SEC_LABEL 1675 The XDR [23] necessary to implement Labeled NFSv4 is presented below: 1677 const FATTR4_SEC_LABEL = 81; 1679 typedef uint32_t policy4; 1681 Figure 6 1683 struct labelformat_spec4 { 1684 policy4 lfs_lfs; 1685 policy4 lfs_pi; 1686 }; 1688 struct sec_label_attr_info { 1689 labelformat_spec4 slai_lfs; 1690 opaque slai_data<>; 1691 }; 1692 The FATTR4_SEC_LABEL contains an array of two components with the 1693 first component being an LFS. It serves to provide the receiving end 1694 with the information necessary to translate the security attribute 1695 into a form that is usable by the endpoint. Label Formats assigned 1696 an LFS may optionally choose to include a Policy Identifier field to 1697 allow for complex policy deployments. The LFS and Label Format 1698 Registry are described in detail in [22]. The translation used to 1699 interpret the security attribute is not specified as part of the 1700 protocol as it may depend on various factors. The second component 1701 is an opaque section which contains the data of the attribute. This 1702 component is dependent on the MAC model to interpret and enforce. 1704 In particular, it is the responsibility of the LFS specification to 1705 define a maximum size for the opaque section, slai_data<>. When 1706 creating or modifying a label for an object, the client needs to be 1707 guaranteed that the server will accept a label that is sized 1708 correctly. By both client and server being part of a specific MAC 1709 model, the client will be aware of the size. 1711 7.3.2. Delegations 1713 In the event that a security attribute is changed on the server while 1714 a client holds a delegation on the file, the client should follow the 1715 existing protocol with respect to attribute changes. It should flush 1716 all changes back to the server and relinquish the delegation. 1718 7.3.3. Permission Checking 1720 It is not feasible to enumerate all possible MAC models and even 1721 levels of protection within a subset of these models. This means 1722 that the NFSv4 client and servers cannot be expected to directly make 1723 access control decisions based on the security attribute. Instead 1724 NFSv4 should defer permission checking on this attribute to the host 1725 system. These checks are performed in addition to existing DAC and 1726 ACL checks outlined in the NFSv4 protocol. Section 7.6 gives a 1727 specific example of how the security attribute is handled under a 1728 particular MAC model. 1730 7.3.4. Object Creation 1732 When creating files in NFSv4 the OPEN and CREATE operations are used. 1733 One of the parameters to these operations is an fattr4 structure 1734 containing the attributes the file is to be created with. This 1735 allows NFSv4 to atomically set the security attribute of files upon 1736 creation. When a client is MAC aware it must always provide the 1737 initial security attribute upon file creation. In the event that the 1738 server is the only MAC aware entity in the system it should ignore 1739 the security attribute specified by the client and instead make the 1740 determination itself. A more in depth explanation can be found in 1741 Section 7.6. 1743 7.3.5. Existing Objects 1745 Note that under the MAC model, all objects must have labels. 1746 Therefore, if an existing server is upgraded to include LNFS support, 1747 then it is the responsibility of the security system to define the 1748 behavior for existing objects. For example, if the security system 1749 is LFS 0, which means the server just stores and returns labels, then 1750 existing files should return labels which are set to an empty value. 1752 7.3.6. Label Changes 1754 As per the requirements, when a file's security label is modified, 1755 the server must notify all clients which have the file opened of the 1756 change in label. It does so with CB_ATTR_CHANGED. There are 1757 preconditions to making an attribute change imposed by NFSv4 and the 1758 security system might want to impose others. In the process of 1759 meeting these preconditions, the server may chose to either serve the 1760 request in whole or return NFS4ERR_DELAY to the SETATTR operation. 1762 If there are open delegations on the file belonging to client other 1763 than the one making the label change, then the process described in 1764 Section 7.3.2 must be followed. 1766 As the server is always presented with the subject label from the 1767 client, it does not necessarily need to communicate the fact that the 1768 label has changed to the client. In the cases where the change 1769 outright denies the client access, the client will be able to quickly 1770 determine that there is a new label in effect. It is in cases where 1771 the client may share the same object between multiple subjects or a 1772 security system which is not strictly hierarchical that the 1773 CB_ATTR_CHANGED callback is very useful. It allows the server to 1774 inform the clients that the cached security attribute is now stale. 1776 Consider a system in which the clients enforce MAC checks and and the 1777 server has a very simple security system which just stores the 1778 labels. In this system, the MAC label check always allows access, 1779 regardless of the subject label. 1781 The way in which MAC labels are enforced is by the client. So if 1782 client A changes a security label on a file, then the server MUST 1783 inform all clients that have the file opened that the label has 1784 changed via CB_ATTR_CHANGED. Then the clients MUST retrieve the new 1785 label and MUST enforce access via the new attribute values. 1787 7.4. pNFS Considerations 1789 This section examines the issues in deploying LNFS in a pNFS 1790 community of servers. 1792 7.4.1. MAC Label Checks 1794 The new FATTR4_SEC_LABEL attribute is metadata information and as 1795 such the DS is not aware of the value contained on the MDS. 1796 Fortunately, the NFSv4.1 protocol [2] already has provisions for 1797 doing access level checks from the DS to the MDS. In order for the 1798 DS to validate the subject label presented by the client, it SHOULD 1799 utilize this mechanism. 1801 If a file's FATTR4_SEC_LABEL is changed, then the MDS should utilize 1802 CB_ATTR_CHANGED to inform the client of that fact. If the MDS is 1803 maintaining 1805 7.5. Discovery of Server LNFS Support 1807 The server can easily determine that a client supports LNFS when it 1808 queries for the FATTR4_SEC_LABEL label for an object. Note that it 1809 cannot assume that the presence of RPCSEC_GSSv3 indicates LNFS 1810 support. The client might need to discover which LFS the server 1811 supports. 1813 A server which supports LNFS MUST allow a client with any subject 1814 label to retrieve the FATTR4_SEC_LABEL attribute for the root 1815 filehandle, ROOTFH. The following compound must always succeed as 1816 far as a MAC label check is concerned: 1818 PUTROOTFH, GETATTR {FATTR4_SEC_LABEL} 1820 Note that the server might have imposed a security flavor on the root 1821 that precludes such access. I.e., if the server requires kerberized 1822 access and the client presents a compound with AUTH_SYS, then the 1823 server is allowed to return NFS4ERR_WRONGSEC in this case. But if 1824 the client presents a correct security flavor, then the server MUST 1825 return the FATTR4_SEC_LABEL attribute with the supported LFS filled 1826 in. 1828 7.6. MAC Security NFS Modes of Operation 1830 A system using Labeled NFS may operate in two modes. The first mode 1831 provides the most protection and is called "full mode". In this mode 1832 both the client and server implement a MAC model allowing each end to 1833 make an access control decision. The remaining mode is called the 1834 "guest mode" and in this mode one end of the connection is not 1835 implementing a MAC model and thus offers less protection than full 1836 mode. 1838 7.6.1. Full Mode 1840 Full mode environments consist of MAC aware NFSv4 servers and clients 1841 and may be composed of mixed MAC models and policies. The system 1842 requires that both the client and server have an opportunity to 1843 perform an access control check based on all relevant information 1844 within the network. The file object security attribute is provided 1845 using the mechanism described in Section 7.3. The security attribute 1846 of the subject making the request is transported at the RPC layer 1847 using the mechanism described in RPCSECGSSv3 [5]. 1849 7.6.1.1. Initial Labeling and Translation 1851 The ability to create a file is an action that a MAC model may wish 1852 to mediate. The client is given the responsibility to determine the 1853 initial security attribute to be placed on a file. This allows the 1854 client to make a decision as to the acceptable security attributes to 1855 create a file with before sending the request to the server. Once 1856 the server receives the creation request from the client it may 1857 choose to evaluate if the security attribute is acceptable. 1859 Security attributes on the client and server may vary based on MAC 1860 model and policy. To handle this the security attribute field has an 1861 LFS component. This component is a mechanism for the host to 1862 identify the format and meaning of the opaque portion of the security 1863 attribute. A full mode environment may contain hosts operating in 1864 several different LFSs. In this case a mechanism for translating the 1865 opaque portion of the security attribute is needed. The actual 1866 translation function will vary based on MAC model and policy and is 1867 out of the scope of this document. If a translation is unavailable 1868 for a given LFS then the request SHOULD be denied. Another recourse 1869 is to allow the host to provide a fallback mapping for unknown 1870 security attributes. 1872 7.6.1.2. Policy Enforcement 1874 In full mode access control decisions are made by both the clients 1875 and servers. When a client makes a request it takes the security 1876 attribute from the requesting process and makes an access control 1877 decision based on that attribute and the security attribute of the 1878 object it is trying to access. If the client denies that access an 1879 RPC call to the server is never made. If however the access is 1880 allowed the client will make a call to the NFS server. 1882 When the server receives the request from the client it extracts the 1883 security attribute conveyed in the RPC request. The server then uses 1884 this security attribute and the attribute of the object the client is 1885 trying to access to make an access control decision. If the server's 1886 policy allows this access it will fulfill the client's request, 1887 otherwise it will return NFS4ERR_ACCESS. 1889 Implementations MAY validate security attributes supplied over the 1890 network to ensure that they are within a set of attributes permitted 1891 from a specific peer, and if not, reject them. Note that a system 1892 may permit a different set of attributes to be accepted from each 1893 peer. 1895 7.6.1.3. Label Aware Only Server 1897 If the LFS is 0, then it indicates a server which is label aware, but 1898 does not enforce policies. Such a server will store and retrieve all 1899 object labels presented by clients, notify the clients of any label 1900 changes via CB_ATTR_CHANGED, but will not restrict access via the 1901 subject label. Instead, it will expect the clients to enforce all 1902 such access locally. 1904 7.6.2. Guest Mode 1906 Guest mode implies that either the client or the server does not 1907 handle labels. If the client is not LNFS aware, then it will not 1908 offer subject labels to the server. The server is the only entity 1909 enforcing policy, and may selectively provide standard NFS services 1910 to clients based on their authentication credentials and/or 1911 associated network attributes (e.g., IP address, network interface). 1912 The level of trust and access extended to a client in this mode is 1913 configuration-specific. If the server is not LNFS aware, then it 1914 will not return object labels to the client. Clients in this 1915 environment are may consist of groups implementing different MAC 1916 model policies. The system requires that all clients in the 1917 environment be responsible for access control checks. 1919 7.7. Security Considerations 1921 This entire document deals with security issues. 1923 Depending on the level of protection the MAC system offers there may 1924 be a requirement to tightly bind the security attribute to the data. 1926 When only one of the client or server enforces labels, it is 1927 important to realize that the other side is not enforcing MAC 1928 protections. Alternate methods might be in use to handle the lack of 1929 MAC support and care should be taken to identify and mitigate threats 1930 from possible tampering outside of these methods. 1932 An example of this is that a server that modifies READDIR or LOOKUP 1933 results based on the client's subject label might want to always 1934 construct the same subject label for a client which does not present 1935 one. This will prevent a non-LNFS client from mixing entries in the 1936 directory cache. 1938 8. Sharing change attribute implementation details with NFSv4 clients 1940 8.1. Introduction 1942 Although both the NFSv4 [10] and NFSv4.1 protocol [2], define the 1943 change attribute as being mandatory to implement, there is little in 1944 the way of guidance. The only feature that is mandated by them is 1945 that the value must change whenever the file data or metadata change. 1947 While this allows for a wide range of implementations, it also leaves 1948 the client with a conundrum: how does it determine which is the most 1949 recent value for the change attribute in a case where several RPC 1950 calls have been issued in parallel? In other words if two COMPOUNDs, 1951 both containing WRITE and GETATTR requests for the same file, have 1952 been issued in parallel, how does the client determine which of the 1953 two change attribute values returned in the replies to the GETATTR 1954 requests corresponds to the most recent state of the file? In some 1955 cases, the only recourse may be to send another COMPOUND containing a 1956 third GETATTR that is fully serialised with the first two. 1958 NFSv4.2 avoids this kind of inefficiency by allowing the server to 1959 share details about how the change attribute is expected to evolve, 1960 so that the client may immediately determine which, out of the 1961 several change attribute values returned by the server, is the most 1962 recent. 1964 8.2. Definition of the 'change_attr_type' per-file system attribute 1966 enum change_attr_typeinfo { 1967 NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR = 0, 1968 NFS4_CHANGE_TYPE_IS_VERSION_COUNTER = 1, 1969 NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS = 2, 1970 NFS4_CHANGE_TYPE_IS_TIME_METADATA = 3, 1971 NFS4_CHANGE_TYPE_IS_UNDEFINED = 4 1972 }; 1974 +------------------+----+---------------------------+-----+ 1975 | Name | Id | Data Type | Acc | 1976 +------------------+----+---------------------------+-----+ 1977 | change_attr_type | XX | enum change_attr_typeinfo | R | 1978 +------------------+----+---------------------------+-----+ 1980 The solution enables the NFS server to provide additional information 1981 about how it expects the change attribute value to evolve after the 1982 file data or metadata has changed. 'change_attr_type' is defined as a 1983 new recommended attribute, and takes values from enum 1984 change_attr_typeinfo as follows: 1986 NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR: The change attribute value MUST 1987 monotonically increase for every atomic change to the file 1988 attributes, data or directory contents. 1990 NFS4_CHANGE_TYPE_IS_VERSION_COUNTER: The change attribute value MUST 1991 be incremented by one unit for every atomic change to the file 1992 attributes, data or directory contents. This property is 1993 preserved when writing to pNFS data servers. 1995 NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS: The change attribute 1996 value MUST be incremented by one unit for every atomic change to 1997 the file attributes, data or directory contents. In the case 1998 where the client is writing to pNFS data servers, the number of 1999 increments is not guaranteed to exactly match the number of 2000 writes. 2002 NFS4_CHANGE_TYPE_IS_TIME_METADATA: The change attribute is 2003 implemented as suggested in the NFSv4 spec [10] in terms of the 2004 time_metadata attribute. 2006 NFS4_CHANGE_TYPE_IS_UNDEFINED: The change attribute does not take 2007 values that fit into any of these categories. 2009 If either NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR, 2010 NFS4_CHANGE_TYPE_IS_VERSION_COUNTER, or 2011 NFS4_CHANGE_TYPE_IS_TIME_METADATA are set, then the client knows at 2012 the very least that the change attribute is monotonically increasing, 2013 which is sufficient to resolve the question of which value is the 2014 most recent. 2016 If the client sees the value NFS4_CHANGE_TYPE_IS_TIME_METADATA, then 2017 by inspecting the value of the 'time_delta' attribute it additionally 2018 has the option of detecting rogue server implementations that use 2019 time_metadata in violation of the spec. 2021 Finally, if the client sees NFS4_CHANGE_TYPE_IS_VERSION_COUNTER, it 2022 has the ability to predict what the resulting change attribute value 2023 should be after a COMPOUND containing a SETATTR, WRITE, or CREATE. 2024 This again allows it to detect changes made in parallel by another 2025 client. The value NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS permits 2026 the same, but only if the client is not doing pNFS WRITEs. 2028 9. Security Considerations 2030 10. Error Values 2032 NFS error numbers are assigned to failed operations within a Compound 2033 (COMPOUND or CB_COMPOUND) request. A Compound request contains a 2034 number of NFS operations that have their results encoded in sequence 2035 in a Compound reply. The results of successful operations will 2036 consist of an NFS4_OK status followed by the encoded results of the 2037 operation. If an NFS operation fails, an error status will be 2038 entered in the reply and the Compound request will be terminated. 2040 10.1. Error Definitions 2042 Protocol Error Definitions 2044 +--------------------------+--------+------------------+ 2045 | Error | Number | Description | 2046 +--------------------------+--------+------------------+ 2047 | NFS4ERR_BADLABEL | 10093 | Section 10.1.3.1 | 2048 | NFS4ERR_METADATA_NOTSUPP | 10090 | Section 10.1.2.1 | 2049 | NFS4ERR_OFFLOAD_DENIED | 10091 | Section 10.1.2.2 | 2050 | NFS4ERR_PARTNER_NO_AUTH | 10089 | Section 10.1.2.3 | 2051 | NFS4ERR_PARTNER_NOTSUPP | 10088 | Section 10.1.2.4 | 2052 | NFS4ERR_UNION_NOTSUPP | 10094 | Section 10.1.1.1 | 2053 | NFS4ERR_WRONG_LFS | 10092 | Section 10.1.3.2 | 2054 +--------------------------+--------+------------------+ 2056 Table 1 2058 10.1.1. General Errors 2060 This section deals with errors that are applicable to a broad set of 2061 different purposes. 2063 10.1.1.1. NFS4ERR_UNION_NOTSUPP (Error Code 10094) 2065 One of the arguments to the operation is a discriminated union and 2066 while the server supports the given operation, it does not support 2067 the selected arm of the discriminated union. For an example, see 2068 READ_PLUS (Section 13.10). 2070 10.1.2. Server to Server Copy Errors 2072 These errors deal with the interaction between server to server 2073 copies. 2075 10.1.2.1. NFS4ERR_METADATA_NOTSUPP (Error Code 10090) 2077 The destination file cannot support the same metadata as the source 2078 file. 2080 10.1.2.2. NFS4ERR_OFFLOAD_DENIED (Error Code 10091) 2082 The copy offload operation is supported by both the source and the 2083 destination, but the destination is not allowing it for this file. 2084 If the client sees this error, it should fall back to the normal copy 2085 semantics. 2087 10.1.2.3. NFS4ERR_PARTNER_NO_AUTH (Error Code 10089) 2089 The remote server does not authorize a server-to-server copy offload 2090 operation. This may be due to the client's failure to send the 2091 COPY_NOTIFY operation to the remote server, the remote server 2092 receiving a server-to-server copy offload request after the copy 2093 lease time expired, or for some other permission problem. 2095 10.1.2.4. NFS4ERR_PARTNER_NOTSUPP (Error Code 10088) 2097 The remote server does not support the server-to-server copy offload 2098 protocol. 2100 10.1.3. Labeled NFS Errors 2102 These errors are used in LNFS. 2104 10.1.3.1. NFS4ERR_BADLABEL (Error Code 10093) 2106 The label specified is invalid in some manner. 2108 10.1.3.2. NFS4ERR_WRONG_LFS (Error Code 10092) 2110 The LFS specified in the subject label is not compatible with the LFS 2111 in object label. 2113 11. File Attributes 2115 11.1. Attribute Definitions 2117 11.1.1. Attribute 77: space_reserved 2119 The space_reserve attribute is a read/write attribute of type 2120 boolean. It is a per file attribute. When the space_reserved 2121 attribute is set via SETATTR, the server must ensure that there is 2122 disk space to accommodate every byte in the file before it can return 2123 success. If the server cannot guarantee this, it must return 2124 NFS4ERR_NOSPC. 2126 If the client tries to grow a file which has the space_reserved 2127 attribute set, the server must guarantee that there is disk space to 2128 accommodate every byte in the file with the new size before it can 2129 return success. If the server cannot guarantee this, it must return 2130 NFS4ERR_NOSPC. 2132 It is not required that the server allocate the space to the file 2133 before returning success. The allocation can be deferred, however, 2134 it must be guaranteed that it will not fail for lack of space. 2136 The value of space_reserved can be obtained at any time through 2137 GETATTR. 2139 In order to avoid ambiguity, the space_reserve bit cannot be set 2140 along with the size bit in SETATTR. Increasing the size of a file 2141 with space_reserve set will fail if space reservation cannot be 2142 guaranteed for the new size. If the file size is decreased, space 2143 reservation is only guaranteed for the new size and the extra blocks 2144 backing the file can be released. 2146 11.1.2. Attribute 78: space_freed 2148 space_freed gives the number of bytes freed if the file is deleted. 2149 This attribute is read only and is of type length4. It is a per file 2150 attribute. 2152 12. Operations: REQUIRED, RECOMMENDED, or OPTIONAL 2154 The following tables summarize the operations of the NFSv4.2 protocol 2155 and the corresponding designation of REQUIRED, RECOMMENDED, and 2156 OPTIONAL to implement or MUST NOT implement. The designation of MUST 2157 NOT implement is reserved for those operations that were defined in 2158 either NFSv4.0 or NFSV4.1 and MUST NOT be implemented in NFSv4.2. 2160 For the most part, the REQUIRED, RECOMMENDED, or OPTIONAL designation 2161 for operations sent by the client is for the server implementation. 2162 The client is generally required to implement the operations needed 2163 for the operating environment for which it serves. For example, a 2164 read-only NFSv4.2 client would have no need to implement the WRITE 2165 operation and is not required to do so. 2167 The REQUIRED or OPTIONAL designation for callback operations sent by 2168 the server is for both the client and server. Generally, the client 2169 has the option of creating the backchannel and sending the operations 2170 on the fore channel that will be a catalyst for the server sending 2171 callback operations. A partial exception is CB_RECALL_SLOT; the only 2172 way the client can avoid supporting this operation is by not creating 2173 a backchannel. 2175 Since this is a summary of the operations and their designation, 2176 there are subtleties that are not presented here. Therefore, if 2177 there is a question of the requirements of implementation, the 2178 operation descriptions themselves must be consulted along with other 2179 relevant explanatory text within this either specification or that of 2180 NFSv4.1 [2]. 2182 The abbreviations used in the second and third columns of the table 2183 are defined as follows. 2185 REQ REQUIRED to implement 2187 REC RECOMMEND to implement 2189 OPT OPTIONAL to implement 2191 MNI MUST NOT implement 2193 For the NFSv4.2 features that are OPTIONAL, the operations that 2194 support those features are OPTIONAL, and the server would return 2195 NFS4ERR_NOTSUPP in response to the client's use of those operations. 2196 If an OPTIONAL feature is supported, it is possible that a set of 2197 operations related to the feature become REQUIRED to implement. The 2198 third column of the table designates the feature(s) and if the 2199 operation is REQUIRED or OPTIONAL in the presence of support for the 2200 feature. 2202 The OPTIONAL features identified and their abbreviations are as 2203 follows: 2205 pNFS Parallel NFS 2207 FDELG File Delegations 2209 DDELG Directory Delegations 2211 COPY Server Side Copy 2213 ADB Application Data Blocks 2215 Operations 2217 +----------------------+--------------------+-----------------------+ 2218 | Operation | REQ, REC, OPT, or | Feature (REQ, REC, or | 2219 | | MNI | OPT) | 2220 +----------------------+--------------------+-----------------------+ 2221 | ACCESS | REQ | | 2222 | BACKCHANNEL_CTL | REQ | | 2223 | BIND_CONN_TO_SESSION | REQ | | 2224 | CLOSE | REQ | | 2225 | COMMIT | REQ | | 2226 | COPY | OPT | COPY (REQ) | 2227 | COPY_ABORT | OPT | COPY (REQ) | 2228 | COPY_NOTIFY | OPT | COPY (REQ) | 2229 | COPY_REVOKE | OPT | COPY (REQ) | 2230 | COPY_STATUS | OPT | COPY (REQ) | 2231 | CREATE | REQ | | 2232 | CREATE_SESSION | REQ | | 2233 | DELEGPURGE | OPT | FDELG (REQ) | 2234 | DELEGRETURN | OPT | FDELG, DDELG, pNFS | 2235 | | | (REQ) | 2236 | DESTROY_CLIENTID | REQ | | 2237 | DESTROY_SESSION | REQ | | 2238 | EXCHANGE_ID | REQ | | 2239 | FREE_STATEID | REQ | | 2240 | GETATTR | REQ | | 2241 | GETDEVICEINFO | OPT | pNFS (REQ) | 2242 | GETDEVICELIST | OPT | pNFS (OPT) | 2243 | GETFH | REQ | | 2244 | INITIALIZE | OPT | ADB (REQ) | 2245 | GET_DIR_DELEGATION | OPT | DDELG (REQ) | 2246 | LAYOUTCOMMIT | OPT | pNFS (REQ) | 2247 | LAYOUTGET | OPT | pNFS (REQ) | 2248 | LAYOUTRETURN | OPT | pNFS (REQ) | 2249 | LINK | OPT | | 2250 | LOCK | REQ | | 2251 | LOCKT | REQ | | 2252 | LOCKU | REQ | | 2253 | LOOKUP | REQ | | 2254 | LOOKUPP | REQ | | 2255 | NVERIFY | REQ | | 2256 | OPEN | REQ | | 2257 | OPENATTR | OPT | | 2258 | OPEN_CONFIRM | MNI | | 2259 | OPEN_DOWNGRADE | REQ | | 2260 | PUTFH | REQ | | 2261 | PUTPUBFH | REQ | | 2262 | PUTROOTFH | REQ | | 2263 | READ | OPT | | 2264 | READDIR | REQ | | 2265 | READLINK | OPT | | 2266 | READ_PLUS | OPT | ADB (REQ) | 2267 | RECLAIM_COMPLETE | REQ | | 2268 | RELEASE_LOCKOWNER | MNI | | 2269 | REMOVE | REQ | | 2270 | RENAME | REQ | | 2271 | RENEW | MNI | | 2272 | RESTOREFH | REQ | | 2273 | SAVEFH | REQ | | 2274 | SECINFO | REQ | | 2275 | SECINFO_NO_NAME | REC | pNFS file layout | 2276 | | | (REQ) | 2277 | SEQUENCE | REQ | | 2278 | SETATTR | REQ | | 2279 | SETCLIENTID | MNI | | 2280 | SETCLIENTID_CONFIRM | MNI | | 2281 | SET_SSV | REQ | | 2282 | TEST_STATEID | REQ | | 2283 | VERIFY | REQ | | 2284 | WANT_DELEGATION | OPT | FDELG (OPT) | 2285 | WRITE | REQ | | 2286 +----------------------+--------------------+-----------------------+ 2288 Callback Operations 2290 +-------------------------+-------------------+---------------------+ 2291 | Operation | REQ, REC, OPT, or | Feature (REQ, REC, | 2292 | | MNI | or OPT) | 2293 +-------------------------+-------------------+---------------------+ 2294 | CB_COPY | OPT | COPY (REQ) | 2295 | CB_GETATTR | OPT | FDELG (REQ) | 2296 | CB_LAYOUTRECALL | OPT | pNFS (REQ) | 2297 | CB_NOTIFY | OPT | DDELG (REQ) | 2298 | CB_NOTIFY_DEVICEID | OPT | pNFS (OPT) | 2299 | CB_NOTIFY_LOCK | OPT | | 2300 | CB_PUSH_DELEG | OPT | FDELG (OPT) | 2301 | CB_RECALL | OPT | FDELG, DDELG, pNFS | 2302 | | | (REQ) | 2303 | CB_RECALL_ANY | OPT | FDELG, DDELG, pNFS | 2304 | | | (REQ) | 2305 | CB_RECALL_SLOT | REQ | | 2306 | CB_RECALLABLE_OBJ_AVAIL | OPT | DDELG, pNFS (REQ) | 2307 | CB_SEQUENCE | OPT | FDELG, DDELG, pNFS | 2308 | | | (REQ) | 2309 | CB_WANTS_CANCELLED | OPT | FDELG, DDELG, pNFS | 2310 | | | (REQ) | 2311 +-------------------------+-------------------+---------------------+ 2313 13. NFSv4.2 Operations 2315 13.1. Operation 59: COPY - Initiate a server-side copy 2317 13.1.1. ARGUMENT 2319 const COPY4_GUARDED = 0x00000001; 2320 const COPY4_METADATA = 0x00000002; 2322 struct COPY4args { 2323 /* SAVED_FH: source file */ 2324 /* CURRENT_FH: destination file or */ 2325 /* directory */ 2326 offset4 ca_src_offset; 2327 offset4 ca_dst_offset; 2328 length4 ca_count; 2329 uint32_t ca_flags; 2330 component4 ca_destination; 2331 netloc4 ca_source_server<>; 2332 }; 2334 13.1.2. RESULT 2336 union COPY4res switch (nfsstat4 cr_status) { 2337 case NFS4_OK: 2338 stateid4 cr_callback_id<1>; 2339 default: 2340 length4 cr_bytes_copied; 2341 }; 2343 13.1.3. DESCRIPTION 2345 The COPY operation is used for both intra-server and inter-server 2346 copies. In both cases, the COPY is always sent from the client to 2347 the destination server of the file copy. The COPY operation requests 2348 that a file be copied from the location specified by the SAVED_FH 2349 value to the location specified by the combination of CURRENT_FH and 2350 ca_destination. 2352 The SAVED_FH must be a regular file. If SAVED_FH is not a regular 2353 file, the operation MUST fail and return NFS4ERR_WRONG_TYPE. 2355 In order to set SAVED_FH to the source file handle, the compound 2356 procedure requesting the COPY will include a sub-sequence of 2357 operations such as 2358 PUTFH source-fh 2359 SAVEFH 2361 If the request is for a server-to-server copy, the source-fh is a 2362 filehandle from the source server and the compound procedure is being 2363 executed on the destination server. In this case, the source-fh is a 2364 foreign filehandle on the server receiving the COPY request. If 2365 either PUTFH or SAVEFH checked the validity of the filehandle, the 2366 operation would likely fail and return NFS4ERR_STALE. 2368 In order to avoid this problem, the minor version incorporating the 2369 COPY operations will need to make a few small changes in the handling 2370 of existing operations. If a server supports the server-to-server 2371 COPY feature, a PUTFH followed by a SAVEFH MUST NOT return 2372 NFS4ERR_STALE for either operation. These restrictions do not pose 2373 substantial difficulties for servers. The CURRENT_FH and SAVED_FH 2374 may be validated in the context of the operation referencing them and 2375 an NFS4ERR_STALE error returned for an invalid file handle at that 2376 point. 2378 The CURRENT_FH and ca_destination together specify the destination of 2379 the copy operation. If ca_destination is of 0 (zero) length, then 2380 CURRENT_FH specifies the target file. In this case, CURRENT_FH MUST 2381 be a regular file and not a directory. If ca_destination is not of 0 2382 (zero) length, the ca_destination argument specifies the file name to 2383 which the data will be copied within the directory identified by 2384 CURRENT_FH. In this case, CURRENT_FH MUST be a directory and not a 2385 regular file. 2387 If the file named by ca_destination does not exist and the operation 2388 completes successfully, the file will be visible in the file system 2389 namespace. If the file does not exist and the operation fails, the 2390 file MAY be visible in the file system namespace depending on when 2391 the failure occurs and on the implementation of the NFS server 2392 receiving the COPY operation. If the ca_destination name cannot be 2393 created in the destination file system (due to file name 2394 restrictions, such as case or length), the operation MUST fail. 2396 The ca_src_offset is the offset within the source file from which the 2397 data will be read, the ca_dst_offset is the offset within the 2398 destination file to which the data will be written, and the ca_count 2399 is the number of bytes that will be copied. An offset of 0 (zero) 2400 specifies the start of the file. A count of 0 (zero) requests that 2401 all bytes from ca_src_offset through EOF be copied to the 2402 destination. If concurrent modifications to the source file overlap 2403 with the source file region being copied, the data copied may include 2404 all, some, or none of the modifications. The client can use standard 2405 NFS operations (e.g., OPEN with OPEN4_SHARE_DENY_WRITE or mandatory 2406 byte range locks) to protect against concurrent modifications if the 2407 client is concerned about this. If the source file's end of file is 2408 being modified in parallel with a copy that specifies a count of 0 2409 (zero) bytes, the amount of data copied is implementation dependent 2410 (clients may guard against this case by specifying a non-zero count 2411 value or preventing modification of the source file as mentioned 2412 above). 2414 If the source offset or the source offset plus count is greater than 2415 or equal to the size of the source file, the operation will fail with 2416 NFS4ERR_INVAL. The destination offset or destination offset plus 2417 count may be greater than the size of the destination file. This 2418 allows for the client to issue parallel copies to implement 2419 operations such as "cat file1 file2 file3 file4 > dest". 2421 If the destination file is created as a result of this command, the 2422 destination file's size will be equal to the number of bytes 2423 successfully copied. If the destination file already existed, the 2424 destination file's size may increase as a result of this operation 2425 (e.g. if ca_dst_offset plus ca_count is greater than the 2426 destination's initial size). 2428 If the ca_source_server list is specified, then this is an inter- 2429 server copy operation and the source file is on a remote server. The 2430 client is expected to have previously issued a successful COPY_NOTIFY 2431 request to the remote source server. The ca_source_server list 2432 SHOULD be the same as the COPY_NOTIFY response's cnr_source_server 2433 list. If the client includes the entries from the COPY_NOTIFY 2434 response's cnr_source_server list in the ca_source_server list, the 2435 source server can indicate a specific copy protocol for the 2436 destination server to use by returning a URL, which specifies both a 2437 protocol service and server name. Server-to-server copy protocol 2438 considerations are described in Section 2.2.3 and Section 2.4.1. 2440 The ca_flags argument allows the copy operation to be customized in 2441 the following ways using the guarded flag (COPY4_GUARDED) and the 2442 metadata flag (COPY4_METADATA). 2444 If the guarded flag is set and the destination exists on the server, 2445 this operation will fail with NFS4ERR_EXIST. 2447 If the guarded flag is not set and the destination exists on the 2448 server, the behavior is implementation dependent. 2450 If the metadata flag is set and the client is requesting a whole file 2451 copy (i.e., ca_count is 0 (zero)), a subset of the destination file's 2452 attributes MUST be the same as the source file's corresponding 2453 attributes and a subset of the destination file's attributes SHOULD 2454 be the same as the source file's corresponding attributes. The 2455 attributes in the MUST and SHOULD copy subsets will be defined for 2456 each NFS version. 2458 For NFSv4.1, Table 2 and Table 3 list the REQUIRED and RECOMMENDED 2459 attributes respectively. A "MUST" in the "Copy to destination file?" 2460 column indicates that the attribute is part of the MUST copy set. A 2461 "SHOULD" in the "Copy to destination file?" column indicates that the 2462 attribute is part of the SHOULD copy set. 2464 +--------------------+----+---------------------------+ 2465 | Name | Id | Copy to destination file? | 2466 +--------------------+----+---------------------------+ 2467 | supported_attrs | 0 | no | 2468 | type | 1 | MUST | 2469 | fh_expire_type | 2 | no | 2470 | change | 3 | SHOULD | 2471 | size | 4 | MUST | 2472 | link_support | 5 | no | 2473 | symlink_support | 6 | no | 2474 | named_attr | 7 | no | 2475 | fsid | 8 | no | 2476 | unique_handles | 9 | no | 2477 | lease_time | 10 | no | 2478 | rdattr_error | 11 | no | 2479 | filehandle | 19 | no | 2480 | suppattr_exclcreat | 75 | no | 2481 +--------------------+----+---------------------------+ 2483 Table 2 2485 +--------------------+----+---------------------------+ 2486 | Name | Id | Copy to destination file? | 2487 +--------------------+----+---------------------------+ 2488 | acl | 12 | MUST | 2489 | aclsupport | 13 | no | 2490 | archive | 14 | no | 2491 | cansettime | 15 | no | 2492 | case_insensitive | 16 | no | 2493 | case_preserving | 17 | no | 2494 | change_policy | 60 | no | 2495 | chown_restricted | 18 | MUST | 2496 | dacl | 58 | MUST | 2497 | dir_notif_delay | 56 | no | 2498 | dirent_notif_delay | 57 | no | 2499 | fileid | 20 | no | 2500 | files_avail | 21 | no | 2501 | files_free | 22 | no | 2502 | files_total | 23 | no | 2503 | fs_charset_cap | 76 | no | 2504 | fs_layout_type | 62 | no | 2505 | fs_locations | 24 | no | 2506 | fs_locations_info | 67 | no | 2507 | fs_status | 61 | no | 2508 | hidden | 25 | MUST | 2509 | homogeneous | 26 | no | 2510 | layout_alignment | 66 | no | 2511 | layout_blksize | 65 | no | 2512 | layout_hint | 63 | no | 2513 | layout_type | 64 | no | 2514 | maxfilesize | 27 | no | 2515 | maxlink | 28 | no | 2516 | maxname | 29 | no | 2517 | maxread | 30 | no | 2518 | maxwrite | 31 | no | 2519 | mdsthreshold | 68 | no | 2520 | mimetype | 32 | MUST | 2521 | mode | 33 | MUST | 2522 | mode_set_masked | 74 | no | 2523 | mounted_on_fileid | 55 | no | 2524 | no_trunc | 34 | no | 2525 | numlinks | 35 | no | 2526 | owner | 36 | MUST | 2527 | owner_group | 37 | MUST | 2528 | quota_avail_hard | 38 | no | 2529 | quota_avail_soft | 39 | no | 2530 | quota_used | 40 | no | 2531 | rawdev | 41 | no | 2532 | retentevt_get | 71 | MUST | 2533 | retentevt_set | 72 | no | 2534 | retention_get | 69 | MUST | 2535 | retention_hold | 73 | MUST | 2536 | retention_set | 70 | no | 2537 | sacl | 59 | MUST | 2538 | space_avail | 42 | no | 2539 | space_free | 43 | no | 2540 | space_freed | 78 | no | 2541 | space_reserved | 77 | MUST | 2542 | space_total | 44 | no | 2543 | space_used | 45 | no | 2544 | system | 46 | MUST | 2545 | time_access | 47 | MUST | 2546 | time_access_set | 48 | no | 2547 | time_backup | 49 | no | 2548 | time_create | 50 | MUST | 2549 | time_delta | 51 | no | 2550 | time_metadata | 52 | SHOULD | 2551 | time_modify | 53 | MUST | 2552 | time_modify_set | 54 | no | 2553 +--------------------+----+---------------------------+ 2555 Table 3 2557 [NOTE: The source file's attribute values will take precedence over 2558 any attribute values inherited by the destination file.] 2560 In the case of an inter-server copy or an intra-server copy between 2561 file systems, the attributes supported for the source file and 2562 destination file could be different. By definition,the REQUIRED 2563 attributes will be supported in all cases. If the metadata flag is 2564 set and the source file has a RECOMMENDED attribute that is not 2565 supported for the destination file, the copy MUST fail with 2566 NFS4ERR_ATTRNOTSUPP. 2568 Any attribute supported by the destination server that is not set on 2569 the source file SHOULD be left unset. 2571 Metadata attributes not exposed via the NFS protocol SHOULD be copied 2572 to the destination file where appropriate. 2574 The destination file's named attributes are not duplicated from the 2575 source file. After the copy process completes, the client MAY 2576 attempt to duplicate named attributes using standard NFSv4 2577 operations. However, the destination file's named attribute 2578 capabilities MAY be different from the source file's named attribute 2579 capabilities. 2581 If the metadata flag is not set and the client is requesting a whole 2582 file copy (i.e., ca_count is 0 (zero)), the destination file's 2583 metadata is implementation dependent. 2585 If the client is requesting a partial file copy (i.e., ca_count is 2586 not 0 (zero)), the client SHOULD NOT set the metadata flag and the 2587 server MUST ignore the metadata flag. 2589 If the operation does not result in an immediate failure, the server 2590 will return NFS4_OK, and the CURRENT_FH will remain the destination's 2591 filehandle. 2593 If an immediate failure does occur, cr_bytes_copied will be set to 2594 the number of bytes copied to the destination file before the error 2595 occurred. The cr_bytes_copied value indicates the number of bytes 2596 copied but not which specific bytes have been copied. 2598 A return of NFS4_OK indicates that either the operation is complete 2599 or the operation was initiated and a callback will be used to deliver 2600 the final status of the operation. 2602 If the cr_callback_id is returned, this indicates that the operation 2603 was initiated and a CB_COPY callback will deliver the final results 2604 of the operation. The cr_callback_id stateid is termed a copy 2605 stateid in this context. The server is given the option of returning 2606 the results in a callback because the data may require a relatively 2607 long period of time to copy. 2609 If no cr_callback_id is returned, the operation completed 2610 synchronously and no callback will be issued by the server. The 2611 completion status of the operation is indicated by cr_status. 2613 If the copy completes successfully, either synchronously or 2614 asynchronously, the data copied from the source file to the 2615 destination file MUST appear identical to the NFS client. However, 2616 the NFS server's on disk representation of the data in the source 2617 file and destination file MAY differ. For example, the NFS server 2618 might encrypt, compress, deduplicate, or otherwise represent the on 2619 disk data in the source and destination file differently. 2621 In the event of a failure the state of the destination file is 2622 implementation dependent. The COPY operation may fail for the 2623 following reasons (this is a partial list). 2625 o NFS4ERR_MOVED 2627 o NFS4ERR_NOTSUPP 2629 o NFS4ERR_PARTNER_NOTSUPP 2631 o NFS4ERR_OFFLOAD_DENIED 2633 o NFS4ERR_PARTNER_NO_AUTH 2635 o NFS4ERR_FBIG 2637 o NFS4ERR_NOTDIR 2639 o NFS4ERR_WRONG_TYPE 2641 o NFS4ERR_ISDIR 2643 o NFS4ERR_INVAL 2644 o NFS4ERR_DELAY 2646 o NFS4ERR_METADATA_NOTSUPP 2648 o NFS4ERR_WRONGSEC 2650 13.2. Operation 60: COPY_ABORT - Cancel a server-side copy 2652 13.2.1. ARGUMENT 2654 struct COPY_ABORT4args { 2655 /* CURRENT_FH: desination file */ 2656 stateid4 caa_stateid; 2657 }; 2659 13.2.2. RESULT 2661 struct COPY_ABORT4res { 2662 nfsstat4 car_status; 2663 }; 2665 13.2.3. DESCRIPTION 2667 COPY_ABORT is used for both intra- and inter-server asynchronous 2668 copies. The COPY_ABORT operation allows the client to cancel a 2669 server-side copy operation that it initiated. This operation is sent 2670 in a COMPOUND request from the client to the destination server. 2671 This operation may be used to cancel a copy when the application that 2672 requested the copy exits before the operation is completed or for 2673 some other reason. 2675 The request contains the filehandle and copy stateid cookies that act 2676 as the context for the previously initiated copy operation. 2678 The result's car_status field indicates whether the cancel was 2679 successful or not. A value of NFS4_OK indicates that the copy 2680 operation was canceled and no callback will be issued by the server. 2681 A copy operation that is successfully canceled may result in none, 2682 some, or all of the data copied. 2684 If the server supports asynchronous copies, the server is REQUIRED to 2685 support the COPY_ABORT operation. 2687 The COPY_ABORT operation may fail for the following reasons (this is 2688 a partial list): 2690 o NFS4ERR_NOTSUPP 2692 o NFS4ERR_RETRY 2694 o NFS4ERR_COMPLETE_ALREADY 2696 o NFS4ERR_SERVERFAULT 2698 13.3. Operation 61: COPY_NOTIFY - Notify a source server of a future 2699 copy 2701 13.3.1. ARGUMENT 2703 struct COPY_NOTIFY4args { 2704 /* CURRENT_FH: source file */ 2705 netloc4 cna_destination_server; 2706 }; 2708 13.3.2. RESULT 2710 struct COPY_NOTIFY4resok { 2711 nfstime4 cnr_lease_time; 2712 netloc4 cnr_source_server<>; 2713 }; 2715 union COPY_NOTIFY4res switch (nfsstat4 cnr_status) { 2716 case NFS4_OK: 2717 COPY_NOTIFY4resok resok4; 2718 default: 2719 void; 2720 }; 2722 13.3.3. DESCRIPTION 2724 This operation is used for an inter-server copy. A client sends this 2725 operation in a COMPOUND request to the source server to authorize a 2726 destination server identified by cna_destination_server to read the 2727 file specified by CURRENT_FH on behalf of the given user. 2729 The cna_destination_server MUST be specified using the netloc4 2730 network location format. The server is not required to resolve the 2731 cna_destination_server address before completing this operation. 2733 If this operation succeeds, the source server will allow the 2734 cna_destination_server to copy the specified file on behalf of the 2735 given user. If COPY_NOTIFY succeeds, the destination server is 2736 granted permission to read the file as long as both of the following 2737 conditions are met: 2739 o The destination server begins reading the source file before the 2740 cnr_lease_time expires. If the cnr_lease_time expires while the 2741 destination server is still reading the source file, the 2742 destination server is allowed to finish reading the file. 2744 o The client has not issued a COPY_REVOKE for the same combination 2745 of user, filehandle, and destination server. 2747 The cnr_lease_time is chosen by the source server. A cnr_lease_time 2748 of 0 (zero) indicates an infinite lease. To renew the copy lease 2749 time the client should resend the same copy notification request to 2750 the source server. 2752 To avoid the need for synchronized clocks, copy lease times are 2753 granted by the server as a time delta. However, there is a 2754 requirement that the client and server clocks do not drift 2755 excessively over the duration of the lease. There is also the issue 2756 of propagation delay across the network which could easily be several 2757 hundred milliseconds as well as the possibility that requests will be 2758 lost and need to be retransmitted. 2760 To take propagation delay into account, the client should subtract it 2761 from copy lease times (e.g., if the client estimates the one-way 2762 propagation delay as 200 milliseconds, then it can assume that the 2763 lease is already 200 milliseconds old when it gets it). In addition, 2764 it will take another 200 milliseconds to get a response back to the 2765 server. So the client must send a lease renewal or send the copy 2766 offload request to the cna_destination_server at least 400 2767 milliseconds before the copy lease would expire. If the propagation 2768 delay varies over the life of the lease (e.g., the client is on a 2769 mobile host), the client will need to continuously subtract the 2770 increase in propagation delay from the copy lease times. 2772 The server's copy lease period configuration should take into account 2773 the network distance of the clients that will be accessing the 2774 server's resources. It is expected that the lease period will take 2775 into account the network propagation delays and other network delay 2776 factors for the client population. Since the protocol does not allow 2777 for an automatic method to determine an appropriate copy lease 2778 period, the server's administrator may have to tune the copy lease 2779 period. 2781 A successful response will also contain a list of names, addresses, 2782 and URLs called cnr_source_server, on which the source is willing to 2783 accept connections from the destination. These might not be 2784 reachable from the client and might be located on networks to which 2785 the client has no connection. 2787 If the client wishes to perform an inter-server copy, the client MUST 2788 send a COPY_NOTIFY to the source server. Therefore, the source 2789 server MUST support COPY_NOTIFY. 2791 For a copy only involving one server (the source and destination are 2792 on the same server), this operation is unnecessary. 2794 The COPY_NOTIFY operation may fail for the following reasons (this is 2795 a partial list): 2797 o NFS4ERR_MOVED 2799 o NFS4ERR_NOTSUPP 2801 o NFS4ERR_WRONGSEC 2803 13.4. Operation 62: COPY_REVOKE - Revoke a destination server's copy 2804 privileges 2806 13.4.1. ARGUMENT 2808 struct COPY_REVOKE4args { 2809 /* CURRENT_FH: source file */ 2810 netloc4 cra_destination_server; 2811 }; 2813 13.4.2. RESULT 2815 struct COPY_REVOKE4res { 2816 nfsstat4 crr_status; 2817 }; 2819 13.4.3. DESCRIPTION 2821 This operation is used for an inter-server copy. A client sends this 2822 operation in a COMPOUND request to the source server to revoke the 2823 authorization of a destination server identified by 2824 cra_destination_server from reading the file specified by CURRENT_FH 2825 on behalf of given user. If the cra_destination_server has already 2826 begun copying the file, a successful return from this operation 2827 indicates that further access will be prevented. 2829 The cra_destination_server MUST be specified using the netloc4 2830 network location format. The server is not required to resolve the 2831 cra_destination_server address before completing this operation. 2833 The COPY_REVOKE operation is useful in situations in which the source 2834 server granted a very long or infinite lease on the destination 2835 server's ability to read the source file and all copy operations on 2836 the source file have been completed. 2838 For a copy only involving one server (the source and destination are 2839 on the same server), this operation is unnecessary. 2841 If the server supports COPY_NOTIFY, the server is REQUIRED to support 2842 the COPY_REVOKE operation. 2844 The COPY_REVOKE operation may fail for the following reasons (this is 2845 a partial list): 2847 o NFS4ERR_MOVED 2849 o NFS4ERR_NOTSUPP 2851 13.5. Operation 63: COPY_STATUS - Poll for status of a server-side copy 2853 13.5.1. ARGUMENT 2855 struct COPY_STATUS4args { 2856 /* CURRENT_FH: destination file */ 2857 stateid4 csa_stateid; 2858 }; 2860 13.5.2. RESULT 2862 struct COPY_STATUS4resok { 2863 length4 csr_bytes_copied; 2864 nfsstat4 csr_complete<1>; 2865 }; 2867 union COPY_STATUS4res switch (nfsstat4 csr_status) { 2868 case NFS4_OK: 2869 COPY_STATUS4resok resok4; 2870 default: 2871 void; 2872 }; 2874 13.5.3. DESCRIPTION 2876 COPY_STATUS is used for both intra- and inter-server asynchronous 2877 copies. The COPY_STATUS operation allows the client to poll the 2878 server to determine the status of an asynchronous copy operation. 2879 This operation is sent by the client to the destination server. 2881 If this operation is successful, the number of bytes copied are 2882 returned to the client in the csr_bytes_copied field. The 2883 csr_bytes_copied value indicates the number of bytes copied but not 2884 which specific bytes have been copied. 2886 If the optional csr_complete field is present, the copy has 2887 completed. In this case the status value indicates the result of the 2888 asynchronous copy operation. In all cases, the server will also 2889 deliver the final results of the asynchronous copy in a CB_COPY 2890 operation. 2892 The failure of this operation does not indicate the result of the 2893 asynchronous copy in any way. 2895 If the server supports asynchronous copies, the server is REQUIRED to 2896 support the COPY_STATUS operation. 2898 The COPY_STATUS operation may fail for the following reasons (this is 2899 a partial list): 2901 o NFS4ERR_NOTSUPP 2903 o NFS4ERR_BAD_STATEID 2905 o NFS4ERR_EXPIRED 2907 13.6. Modification to Operation 42: EXCHANGE_ID - Instantiate Client ID 2909 13.6.1. ARGUMENT 2911 /* new */ 2912 const EXCHGID4_FLAG_SUPP_FENCE_OPS = 0x00000004; 2914 13.6.2. RESULT 2916 Unchanged 2918 13.6.3. MOTIVATION 2920 Enterprise applications require guarantees that an operation has 2921 either aborted or completed. NFSv4.1 provides this guarantee as long 2922 as the session is alive: simply send a SEQUENCE operation on the same 2923 slot with a new sequence number, and the successful return of 2924 SEQUENCE indicates the previous operation has completed. However, if 2925 the session is lost, there is no way to know when any in progress 2926 operations have aborted or completed. In hindsight, the NFSv4.1 2927 specification should have mandated that DESTROY_SESSION abort/ 2928 complete all outstanding operations. 2930 13.6.4. DESCRIPTION 2932 A client SHOULD request the EXCHGID4_FLAG_SUPP_FENCE_OPS capability 2933 when it sends an EXCHANGE_ID operation. The server SHOULD set this 2934 capability in the EXCHANGE_ID reply whether the client requests it or 2935 not. If the client ID is created with this capability then the 2936 following will occur: 2938 o The server will not reply to DESTROY_SESSION until all operations 2939 in progress are completed or aborted. 2941 o The server will not reply to subsequent EXCHANGE_ID invoked on the 2942 same Client Owner with a new verifier until all operations in 2943 progress on the Client ID's session are completed or aborted. 2945 o When DESTROY_CLIENTID is invoked, if there are sessions (both idle 2946 and non-idle), opens, locks, delegations, layouts, and/or wants 2947 (Section 18.49) associated with the client ID are removed. 2948 Pending operations will be completed or aborted before the 2949 sessions, opens, locks, delegations, layouts, and/or wants are 2950 deleted. 2952 o The NFS server SHOULD support client ID trunking, and if it does 2953 and the EXCHGID4_FLAG_SUPP_FENCE_OPS capability is enabled, then a 2954 session ID created on one node of the storage cluster MUST be 2955 destroyable via DESTROY_SESSION. In addition, DESTROY_CLIENTID 2956 and an EXCHANGE_ID with a new verifier affects all sessions 2957 regardless what node the sessions were created on. 2959 13.7. Operation 64: INITIALIZE 2961 This operation can be used to initialize the structure imposed by an 2962 application onto a file, i.e., ADBs, and to punch a hole into a file. 2964 13.7.1. ARGUMENT 2966 /* 2967 * We use data_content4 in case we wish to 2968 * extend new types later. Note that we 2969 * are explicitly disallowing data. 2970 */ 2971 union initialize_arg4 switch (data_content4 content) { 2972 case NFS4_CONTENT_APP_BLOCK: 2973 app_data_block4 ia_adb; 2974 case NFS4_CONTENT_HOLE: 2975 data_info4 ia_hole; 2976 default: 2977 void; 2978 }; 2980 struct INITIALIZE4args { 2981 /* CURRENT_FH: file */ 2982 stateid4 ia_stateid; 2983 stable_how4 ia_stable; 2984 initialize_arg4 ia_data<>; 2985 }; 2987 13.7.2. RESULT 2989 struct INITIALIZE4resok { 2990 count4 ir_count; 2991 stable_how4 ir_committed; 2992 verifier4 ir_writeverf; 2993 data_content4 ir_sparse; 2994 }; 2996 union INITIALIZE4res switch (nfsstat4 status) { 2997 case NFS4_OK: 2998 INITIALIZE4resok resok4; 2999 default: 3000 void; 3001 }; 3003 13.7.3. DESCRIPTION 3004 13.7.3.1. Hole punching 3006 Whenever a client wishes to zero the blocks backing a particular 3007 region in the file, it calls the INITIALIZE operation with the 3008 current filehandle set to the filehandle of the file in question, and 3009 the equivalent of start offset and length in bytes of the region set 3010 in ia_hole.di_offset and ia_hole.di_length respectively. If the 3011 ia_hole.di_allocated is set to TRUE, then the blocks will be zeroed 3012 and if it is set to FALSE, then they will be deallocated. All 3013 further reads to this region MUST return zeros until overwritten. 3014 The filehandle specified must be that of a regular file. 3016 Situations may arise where di_offset and/or di_offset + di_length 3017 will not be aligned to a boundary that the server does allocations/ 3018 deallocations in. For most filesystems, this is the block size of 3019 the file system. In such a case, the server can deallocate as many 3020 bytes as it can in the region. The blocks that cannot be deallocated 3021 MUST be zeroed. Except for the block deallocation and maximum hole 3022 punching capability, a INITIALIZE operation is to be treated similar 3023 to a write of zeroes. 3025 The server is not required to complete deallocating the blocks 3026 specified in the operation before returning. It is acceptable to 3027 have the deallocation be deferred. In fact, INITIALIZE is merely a 3028 hint; it is valid for a server to return success without ever doing 3029 anything towards deallocating the blocks backing the region 3030 specified. However, any future reads to the region MUST return 3031 zeroes. 3033 If used to hole punch, INITIALIZE will result in the space_used 3034 attribute being decreased by the number of bytes that were 3035 deallocated. The space_freed attribute may or may not decrease, 3036 depending on the support and whether the blocks backing the specified 3037 range were shared or not. The size attribute will remain unchanged. 3039 The INITIALIZE operation MUST NOT change the space reservation 3040 guarantee of the file. While the server can deallocate the blocks 3041 specified by di_offset and di_length, future writes to this region 3042 MUST NOT fail with NFSERR_NOSPC. 3044 The INITIALIZE operation may fail for the following reasons (this is 3045 a partial list): 3047 NFS4ERR_NOTSUPP The Hole punch operations are not supported by the 3048 NFS server receiving this request. 3050 NFS4ERR_DIR The current filehandle is of type NF4DIR. 3052 NFS4ERR_SYMLINK The current filehandle is of type NF4LNK. 3054 NFS4ERR_WRONG_TYPE The current filehandle does not designate an 3055 ordinary file. 3057 13.7.3.2. ADBs 3059 If the server supports ADBs, then it MUST support the 3060 NFS4_CONTENT_APP_BLOCK arm of the INITIALIZE operation. The server 3061 has no concept of the structure imposed by the application. It is 3062 only when the application writes to a section of the file does order 3063 get imposed. In order to detect corruption even before the 3064 application utilizes the file, the application will want to 3065 initialize a range of ADBs using INITIALIZE. 3067 For ADBs, when the client invokes the INITIALIZE operation, it has 3068 two desired results: 3070 1. The structure described by the app_data_block4 be imposed on the 3071 file. 3073 2. The contents described by the app_data_block4 be sparse. 3075 If the server supports the INITIALIZE operation, it still might not 3076 support sparse files. So if it receives the INITIALIZE operation, 3077 then it MUST populate the contents of the file with the initialized 3078 ADBs. 3080 If the data was already initialized, there are two interesting 3081 scenarios: 3083 1. The data blocks are allocated. 3085 2. Initializing in the middle of an existing ADB. 3087 If the data blocks were already allocated, then the INITIALIZE is a 3088 hole punch operation. If INITIALIZE supports sparse files, then the 3089 data blocks are to be deallocated. If not, then the data blocks are 3090 to be rewritten in the indicated ADB format. 3092 Since the server has no knowledge of ADBs, it should not report 3093 misaligned creation of ADBs. Even while it can detect them, it 3094 cannot disallow them, as the application might be in the process of 3095 changing the size of the ADBs. Thus the server must be prepared to 3096 handle an INITIALIZE into an existing ADB. 3098 This document does not mandate the manner in which the server stores 3099 ADBs sparsely for a file. It does assume that if ADBs are stored 3100 sparsely, then the server can detect when an INITIALIZE arrives that 3101 will force a new ADB to start inside an existing ADB. For example, 3102 assume that ADBi has a adb_block_size of 4k and that an INITIALIZE 3103 starts 1k inside ADBi. The server should [[Comment.5: Need to flesh 3104 this out. --TH]] 3106 13.8. Operation 67: IO_ADVISE - Application I/O access pattern hints 3108 This section introduces a new operation, named IO_ADVISE, which 3109 allows NFS clients to communicate application I/O access pattern 3110 hints to the NFS server. This new operation will allow hints to be 3111 sent to the server when applications use posix_fadvise, direct I/O, 3112 or at any other point at which the client finds useful. 3114 13.8.1. ARGUMENT 3116 enum IO_ADVISE_type4 { 3117 IO_ADVISE4_NORMAL = 0, 3118 IO_ADVISE4_SEQUENTIAL = 1, 3119 IO_ADVISE4_SEQUENTIAL_BACKWARDS = 2, 3120 IO_ADVISE4_RANDOM = 3, 3121 IO_ADVISE4_WILLNEED = 4, 3122 IO_ADVISE4_WILLNEED_OPPORTUNISTIC = 5, 3123 IO_ADVISE4_DONTNEED = 6, 3124 IO_ADVISE4_NOREUSE = 7, 3125 IO_ADVISE4_READ = 8, 3126 IO_ADVISE4_WRITE = 9, 3127 IO_ADVISE4_INIT_PROXIMITY = 10 3128 }; 3130 struct IO_ADVISE4args { 3131 /* CURRENT_FH: file */ 3132 stateid4 iar_stateid; 3133 offset4 iar_offset; 3134 length4 iar_count; 3135 bitmap4 iar_hints; 3136 }; 3138 13.8.2. RESULT 3140 struct IO_ADVISE4resok { 3141 bitmap4 ior_hints; 3142 }; 3144 union IO_ADVISE4res switch (nfsstat4 _status) { 3145 case NFS4_OK: 3146 IO_ADVISE4resok resok4; 3147 default: 3148 void; 3149 }; 3151 13.8.3. DESCRIPTION 3153 The IO_ADVISE operation sends an I/O access pattern hint to the 3154 server for the owner of stated for a given byte range specified by 3155 iar_offset and iar_count. The byte range specified by iar_offset and 3156 iar_count need not currently exist in the file, but the iar_hints 3157 will apply to the byte range when it does exist. If iar_count is 0, 3158 all data following iar_offset is specified. The server MAY ignore 3159 the advice. 3161 The following are the possible hints: 3163 IO_ADVISE4_NORMAL Specifies that the application has no advice to 3164 give on its behavior with respect to the specified data. It is 3165 the default characteristic if no advice is given. 3167 IO_ADVISE4_SEQUENTIAL Specifies that the stated holder expects to 3168 access the specified data sequentially from lower offsets to 3169 higher offsets. 3171 IO_ADVISE4_SEQUENTIAL BACKWARDS Specifies that the stated holder 3172 expects to access the specified data sequentially from higher 3173 offsets to lower offsets. 3175 IO_ADVISE4_RANDOM Specifies that the stated holder expects to access 3176 the specified data in a random order. 3178 IO_ADVISE4_WILLNEED Specifies that the stated holder expects to 3179 access the specified data in the near future. 3181 IO_ADVISE4_WILLNEED_OPPORTUNISTIC Specifies that the stated holder 3182 expects to possibly access the data in the near future. This is a 3183 speculative hint, and therefore the server should prefetch data or 3184 indirect blocks only if it can be done at a marginal cost. 3186 IO_ADVISE_DONTNEED Specifies that the stated holder expects that it 3187 will not access the specified data in the near future. 3189 IO_ADVISE_NOREUSE Specifies that the stated holder expects to access 3190 the specified data once and then not reuse it thereafter. 3192 IO_ADVISE4_READ Specifies that the stated holder expects to read the 3193 specified data in the near future. 3195 IO_ADVISE4_WRITE Specifies that the stated holder expects to write 3196 the specified data in the near future. 3198 IO_ADVISE4_INIT_PROXIMITY The client has recently accessed the byte 3199 range in its own cache. This informs the server that the data in 3200 the byte range remains important to the client. When the server 3201 reaches resource exhaustion, knowing which data is more important 3202 allows the server to make better choices about which data to, for 3203 example purge from a cache, or move to secondary storage. It also 3204 informs the server which delegations are more important, since if 3205 delegations are working correctly, once delegated to a client, a 3206 server might never receive another I/O request for the file. 3208 The server will return success if the operation is properly formed, 3209 otherwise the server will return an error. The server MUST NOT 3210 return an error if it does not recognize or does not support the 3211 requested advice. This is also true even if the client sends 3212 contradictory hints to the server, e.g., IO_ADVISE4_SEQUENTIAL and 3213 IO_ADVISE4_RANDOM in a single IO_ADVISE operation. In this case, the 3214 server MUST return success and a ior_hints value that indicates the 3215 hint it intends to optimize. For contradictory hints, this may mean 3216 simply returning IO_ADVISE4_NORMAL for example. 3218 The ior_hints returned by the server is primarily for debugging 3219 purposes since the server is under no obligation to carry out the 3220 hints that it describes in the ior_hints result. In addition, while 3221 the server may have intended to implement the hints returned in 3222 ior_hints, as time progresses, the server may need to change its 3223 handling of a given file due to several reasons including, but not 3224 limited to, memory pressure, additional IO_ADVISE hints sent by other 3225 clients, and heuristically detected file access patterns. 3227 The server MAY return different advice than what the client 3228 requested. If it does, then this might be due to one of several 3229 conditions, including, but not limited to another client advising of 3230 a different I/O access pattern; a different I/O access pattern from 3231 another client that that the server has heuristically detected; or 3232 the server is not able to support the requested I/O access pattern, 3233 perhaps due to a temporary resource limitation. 3235 Each issuance of the IO_ADVISE operation overrides all previous 3236 issuances of IO_ADVISE for a given byte range. This effectively 3237 follows a strategy of last hint wins for a given stated and byte 3238 range. 3240 Clients should assume that hints included in an IO_ADVISE operation 3241 will be forgotten once the file is closed. 3243 13.8.4. IMPLEMENTATION 3245 The NFS client may choose to issue an IO_ADVISE operation to the 3246 server in several different instances. 3248 The most obvious is in direct response to an application's execution 3249 of posix_fadvise. In this case, IO_ADVISE4_WRITE and IO_ADVISE4_READ 3250 may be set based upon the type of file access specified when the file 3251 was opened. 3253 Another useful point would be when an application indicates it is 3254 using direct I/O. Direct I/O may be specified at file open, in which 3255 case a IO_ADVISE may be included in the same compound as the OPEN 3256 operation with the IO_ADVISE4_NOREUSE flag set. Direct I/O may also 3257 be specified separately, in which case a IO_ADVISE operation can be 3258 sent to the server separately. As above, IO_ADVISE4_WRITE and 3259 IO_ADVISE4_READ may be set based upon the type of file access 3260 specified when the file was opened. 3262 13.8.5. pNFS File Layout Data Type Considerations 3264 The IO_ADVISE considerations for pNFS are very similar to the COMMIT 3265 considerations for pNFS. That is, as with COMMIT, some NFS server 3266 implementations prefer IO_ADVISE be done on the DS, and some prefer 3267 it be done on the MDS. 3269 So for the file's layout type, it is proposed that NFSv4.2 include an 3270 additional hint NFL42_CARE_IO_ADVISE_THRU_MDS which is valid only on 3271 NFSv4.2 or higher. Any file's layout obtained with NFSv4.1 MUST NOT 3272 have NFL42_UFLG_IO_ADVISE_THRU_MDS set. Any file's layout obtained 3273 with NFSv4.2 MAY have NFL42_UFLG_IO_ADVISE_THRU_MDS set. If the 3274 client does not implement IO_ADVISE, then it MUST ignore 3275 NFL42_UFLG_IO_ADVISE_THRU_MDS. 3277 If NFL42_UFLG_IO_ADVISE_THRU_MDS is set, then if the client 3278 implements IO_ADVISE, then if it wants the DS to honor IO_ADVISE, the 3279 client MUST send the operation to the MDS, and the server will 3280 communicate the advice back each DS. If the client sends IO_ADVISE 3281 to the DS, then the server MAY return NFS4ERR_NOTSUPP. 3283 If NFL42_UFLG_IO_ADVISE_THRU_MDS is not set, then this indicates to 3284 client that if wants to inform the server via IO_ADVISE of the 3285 client's intended use of the file, then the client SHOULD send an 3286 IO_ADVISE to each DS. While the client MAY always send IO_ADVISE to 3287 the MDS, if the server has not set NFL42_UFLG_IO_ADVISE_THRU_MDS, the 3288 client should expect that such an IO_ADVISE is futile. Note that a 3289 client SHOULD use the same set of arguments on each IO_ADVISE sent to 3290 a DS for the same open file reference. 3292 The server is not required to support different advice for different 3293 DS's with the same open file reference. 3295 13.8.5.1. Dense and Sparse Packing Considerations 3297 The IO_ADVISE operation MUST use the iar_offset and byte range as 3298 dictated by the presence or absence of NFL4_UFLG_DENSE. 3300 E.g., if NFL4_UFLG_DENSE is present, and a READ or WRITE to the DS 3301 for iar_offset 0 really means iar_offset 10000 in the logical file, 3302 then an IO_ADVISE for iar_offset 0 means iar_offset 10000. 3304 E.g., if NFL4_UFLG_DENSE is absent, then a READ or WRITE to the DS 3305 for iar_offset 0 really means iar_offset 0 in the logical file, then 3306 an IO_ADVISE for iar_offset 0 means iar_offset 0 in the logical file. 3308 E.g., if NFL4_UFLG_DENSE is present, the stripe unit is 1000 bytes 3309 and the stripe count is 10, and the dense DS file is serving 3310 iar_offset 0. A READ or WRITE to the DS for iar_offsets 0, 1000, 3311 2000, and 3000, really mean iar_offsets 10000, 20000, 30000, and 3312 40000 (implying a stripe count of 10 and a stripe unit of 1000), then 3313 an IO_ADVISE sent to the same DS with an iar_offset of 500, and a 3314 iar_count of 3000 means that the IO_ADVISE applies to these byte 3315 ranges of the dense DS file: 3317 - 500 to 999 3318 - 1000 to 1999 3319 - 2000 to 2999 3320 - 3000 to 3499 3322 I.e., the contiguous range 500 to 3499 as specified in IO_ADVISE. 3324 It also applies to these byte ranges of the logical file: 3326 - 10500 to 10999 (500 bytes) 3327 - 20000 to 20999 (1000 bytes) 3328 - 30000 to 30999 (1000 bytes) 3329 - 40000 to 40499 (500 bytes) 3330 (total 3000 bytes) 3332 E.g., if NFL4_UFLG_DENSE is absent, the stripe unit is 250 bytes, the 3333 stripe count is 4, and the sparse DS file is serving iar_offset 0. 3334 Then a READ or WRITE to the DS for iar_offsets 0, 1000, 2000, and 3335 3000, really mean iar_offsets 0, 1000, 2000, and 3000 in the logical 3336 file, keeping in mind that on the DS file,. byte ranges 250 to 999, 3337 1250 to 1999, 2250 to 2999, and 3250 to 3999 are not accessible. 3338 Then an IO_ADVISE sent to the same DS with an iar_offset of 500, and 3339 a iar_count of 3000 means that the IO_ADVISE applies to these byte 3340 ranges of the logical file and the sparse DS file: 3342 - 500 to 999 (500 bytes) - no effect 3343 - 1000 to 1249 (250 bytes) - effective 3344 - 1250 to 1999 (750 bytes) - no effect 3345 - 2000 to 2249 (250 bytes) - effective 3346 - 2250 to 2999 (750 bytes) - no effect 3347 - 3000 to 3249 (250 bytes) - effective 3348 - 3250 to 3499 (250 bytes) - no effect 3349 (subtotal 2250 bytes) - no effect 3350 (subtotal 750 bytes) - effective 3351 (grand total 3000 bytes) - no effect + effective 3353 If neither of the flags NFL42_UFLG_IO_ADVISE_THRU_MDS and 3354 NFL4_UFLG_DENSE are set in the layout, then any IO_ADVISE request 3355 sent to the data server with a byte range that overlaps stripe unit 3356 that the data server does not serve MUST NOT result in the status 3357 NFS4ERR_PNFS_IO_HOLE. Instead, the response SHOULD be successful and 3358 if the server applies IO_ADVISE hints on any stripe units that 3359 overlap with the specified range, those hints SHOULD be indicated in 3360 the response. 3362 13.8.6. Number of Supported File Segments 3364 In theory IO_ADVISE allows a client and server to support multiple 3365 file segments, meaning that different, possibly overlapping, byte 3366 ranges of the same open file reference will support different hints. 3367 This is not practical, and in general the server will support just 3368 one set of hints, and these will apply to the entire file. However, 3369 there are some hints that very ephemeral, and are essentially amount 3370 to one time instructions to the NFS server, which will be forgotten 3371 momentarily after IO_ADVISE is executed. 3373 The following hints will always apply to the entire file, regardless 3374 of the specified byte range: 3376 o IO_ADVISE4_NORMAL 3378 o IO_ADVISE4_SEQUENTIAL 3380 o IO_ADVISE4_SEQUENTIAL_BACKWARDS 3382 o IO_ADVISE4_RANDOM 3384 The following hints will always apply to specified byte range, and 3385 will treated as one time instructions: 3387 o IO_ADVISE4_WILLNEED 3389 o IO_ADVISE4_WILLNEED_OPPORTUNISTIC 3391 o IO_ADVISE4_DONTNEED 3393 o IO_ADVISE4_NOREUSE 3395 The following hints are modifiers to all other hints, and will apply 3396 to the entire file and/or to a one time instruction on the specified 3397 byte range: 3399 o IO_ADVISE4_READ 3401 o IO_ADVISE4_WRITE 3403 13.9. Changes to Operation 51: LAYOUTRETURN 3405 13.9.1. Introduction 3407 In the pNFS description provided in [2], the client is not enabled to 3408 relay an error code from the DS to the MDS. In the specification of 3409 the Objects-Based Layout protocol [9], use is made of the opaque 3410 lrf_body field of the LAYOUTRETURN argument to do such a relaying of 3411 error codes. In this section, we define a new data structure to 3412 enable the passing of error codes back to the MDS and provide some 3413 guidelines on what both the client and MDS should expect in such 3414 circumstances. 3416 There are two broad classes of errors, transient and persistent. The 3417 client SHOULD strive to only use this new mechanism to report 3418 persistent errors. It MUST be able to deal with transient issues by 3419 itself. Also, while the client might consider an issue to be 3420 persistent, it MUST be prepared for the MDS to consider such issues 3421 to be persistent. A prime example of this is if the MDS fences off a 3422 client from either a stateid or a filehandle. The client will get an 3423 error from the DS and might relay either NFS4ERR_ACCESS or 3424 NFS4ERR_STALE_STATEID back to the MDS, with the belief that this is a 3425 hard error. The MDS on the other hand, is waiting for the client to 3426 report such an error. For it, the mission is accomplished in that 3427 the client has returned a layout that the MDS had most likley 3428 recalled. 3430 The existing LAYOUTRETURN operation is extended by introducing a new 3431 data structure to report errors, layoutreturn_device_error4. Also, 3432 layoutreturn_device_error4 is introduced to enable an array of errors 3433 to be reported. 3435 13.9.2. ARGUMENT 3437 The ARGUMENT specification of the LAYOUTRETURN operation in section 3438 18.44.1 of [2] is augmented by the following XDR code [23]: 3440 struct layoutreturn_device_error4 { 3441 deviceid4 lrde_deviceid; 3442 nfsstat4 lrde_status; 3443 nfs_opnum4 lrde_opnum; 3444 }; 3446 struct layoutreturn_error_report4 { 3447 layoutreturn_device_error4 lrer_errors<>; 3448 }; 3450 13.9.3. RESULT 3452 The RESULT of the LAYOUTRETURN operation is unchanged; see section 3453 18.44.2 of [2]. 3455 13.9.4. DESCRIPTION 3457 The following text is added to the end of the LAYOUTRETURN operation 3458 DESCRIPTION in section 18.44.3 of [2]. 3460 When a client used LAYOUTRETURN with a type of LAYOUTRETURN4_FILE, 3461 then if the lrf_body field is NULL, it indicates to the MDS that the 3462 client experienced no errors. If lrf_body is non-NULL, then the 3463 field references error information which is layout type specific. 3464 I.e., the Objects-Based Layout protocol can continue to utilize 3465 lrf_body as specified in [9]. For both Files-Based Layouts, the 3466 field references a layoutreturn_device_error4, which contains an 3467 array of layoutreturn_device_error4. 3469 Each individual layoutreturn_device_error4 descibes a single error 3470 associated with a DS, which is identfied via lrde_deviceid. The 3471 operation which returned the error is identified via lrde_opnum. 3472 Finally the NFS error value (nfsstat4) encountered is provided via 3473 lrde_status and may consist of the following error codes: 3475 NFS4_OKAY: No issues were found for this device. 3477 NFS4ERR_NXIO: The client was unable to establish any communication 3478 with the DS. 3480 NFS4ERR_*: The client was able to establish communication with the 3481 DS and is returning one of the allowed error codes for the 3482 operation denoted by lrde_opnum. 3484 13.9.5. IMPLEMENTATION 3486 The following text is added to the end of the LAYOUTRETURN operation 3487 IMPLEMENTATION in section 18.4.4 of [2]. 3489 A client that expects to use pNFS for a mounted filesystem SHOULD 3490 check for pNFS support at mount time. This check SHOULD be performed 3491 by sending a GETDEVICELIST operation, followed by layout-type- 3492 specific checks for accessibility of each storage device returned by 3493 GETDEVICELIST. If the NFS server does not support pNFS, the 3494 GETDEVICELIST operation will be rejected with an NFS4ERR_NOTSUPP 3495 error; in this situation it is up to the client to determine whether 3496 it is acceptable to proceed with NFS-only access. 3498 Clients are expected to tolerate transient storage device errors, and 3499 hence clients SHOULD NOT use the LAYOUTRETURN error handling for 3500 device access problems that may be transient. The methods by which a 3501 client decides whether an access problem is transient vs. persistent 3502 are implementation-specific, but may include retrying I/Os to a data 3503 server under appropriate conditions. 3505 When an I/O fails to a storage device, the client SHOULD retry the 3506 failed I/O via the MDS. In this situation, before retrying the I/O, 3507 the client SHOULD return the layout, or the affected portion thereof, 3508 and SHOULD indicate which storage device or devices was problematic. 3509 If the client does not do this, the MDS may issue a layout recall 3510 callback in order to perform the retried I/O. 3512 The client needs to be cognizant that since this error handling is 3513 optional in the MDS, the MDS may silently ignore this functionality. 3514 Also, as the MDS may consider some issues the client reports to be 3515 expected (see Section 13.9.1), the client might find it difficult to 3516 detect a MDS which has not implemented error handling via 3517 LAYOUTRETURN. 3519 If an MDS is aware that a storage device is proving problematic to a 3520 client, the MDS SHOULD NOT include that storage device in any pNFS 3521 layouts sent to that client. If the MDS is aware that a storage 3522 device is affecting many clients, then the MDS SHOULD NOT include 3523 that storage device in any pNFS layouts sent out. Clients must still 3524 be aware that the MDS might not have any choice in using the storage 3525 device, i.e., there might only be one possible layout for the system. 3527 Another interesting complication is that for existing files, the MDS 3528 might have no choice in which storage devices to hand out to clients. 3529 The MDS might try to restripe a file across a different storage 3530 device, but clients need to be aware that not all implementations 3531 have restriping support. 3533 An MDS SHOULD react to a client return of layouts with errors by not 3534 using the problematic storage devices in layouts for that client, but 3535 the MDS is not required to indefinitely retain per-client storage 3536 device error information. An MDS is also not required to 3537 automatically reinstate use of a previously problematic storage 3538 device; administrative intervention may be required instead. 3540 A client MAY perform I/O via the MDS even when the client holds a 3541 layout that covers the I/O; servers MUST support this client 3542 behavior, and MAY recall layouts as needed to complete I/Os. 3544 13.10. Operation 65: READ_PLUS 3546 READ_PLUS is a new read operation which allows NFS clients to avoid 3547 reading holes in a sparse file and to efficiently transfer ADBs. 3548 READ_PLUS supports all the features of the existing NFSv4.1 READ 3549 operation [2] but also extends the response to avoid returning data 3550 for portions of the file which are either initialized and contain no 3551 backing store or if the result would appear to be so. I.e., if the 3552 result was a data block composed entirely of zeros, then it is easier 3553 to return a hole. Returning data blocks of unitialized data wastes 3554 computational and network resources, thus reducing performance. 3555 READ_PLUS uses a new result structure that tells the client that the 3556 result is all zeroes AND the byte-range of the hole in which the 3557 request was made. 3559 If the client sends a READ operation, it is explicitly stating that 3560 it is neither supporting sparse files nor ADBs. So if a READ occurs 3561 on a sparse ADB or file, then the server must expand such data to be 3562 raw bytes. If a READ occurs in the middle of a hole or ADB, the 3563 server can only send back bytes starting from that offset. 3565 Such an operation is inefficient for transfer of sparse sections of 3566 the file. As such, READ is marked as OBSOLETE in NFSv4.2. Instead, 3567 a client should issue READ_PLUS. Note that as the client has no a 3568 priori knowledge of whether either an ADB or a hole is present or 3569 not, it should always use READ_PLUS. 3571 13.10.1. ARGUMENT 3573 struct READ_PLUS4args { 3574 /* CURRENT_FH: file */ 3575 stateid4 rpa_stateid; 3576 offset4 rpa_offset; 3577 count4 rpa_count; 3578 }; 3580 13.10.2. RESULT 3582 union read_plus_content switch (data_content4 content) { 3583 case NFS4_CONTENT_DATA: 3584 opaque rpc_data<>; 3585 case NFS4_CONTENT_APP_BLOCK: 3586 app_data_block4 rpc_block; 3587 case NFS4_CONTENT_HOLE: 3588 data_info4 rpc_hole; 3589 default: 3590 void; 3591 }; 3593 /* 3594 * Allow a return of an array of contents. 3595 */ 3596 struct read_plus_res4 { 3597 bool rpr_eof; 3598 read_plus_content rpr_contents<>; 3599 }; 3601 union READ_PLUS4res switch (nfsstat4 status) { 3602 case NFS4_OK: 3603 read_plus_res4 resok4; 3604 default: 3605 void; 3606 }; 3608 13.10.3. DESCRIPTION 3610 The READ_PLUS operation is based upon the NFSv4.1 READ operation [2] 3611 and similarly reads data from the regular file identified by the 3612 current filehandle. 3614 The client provides a rpa_offset of where the READ_PLUS is to start 3615 and a rpa_count of how many bytes are to be read. A rpa_offset of 3616 zero means to read data starting at the beginning of the file. If 3617 rpa_offset is greater than or equal to the size of the file, the 3618 status NFS4_OK is returned with di_length (the data length) set to 3619 zero and eof set to TRUE. READ_PLUS is subject to access permissions 3620 checking. 3622 The READ_PLUS result is comprised of an array of rpr_contents, each 3623 of which describe a data_content4 type of data. For NFSv4.2, the 3624 allowed values are data, ADB, and hole. A server is required to 3625 support the data type, but neither ADB nor hole. Both an ADB and a 3626 hole must be returned in its entirety - clients must be prepared to 3627 get more information than they requested. 3629 READ_PLUS has to support all of the errors which are returned by READ 3630 plus NFS4ERR_UNION_NOTSUPP. If the client asks for a hole and the 3631 server does not support that arm of the discriminated union, but does 3632 support one or more additional arms, it can signal to the client that 3633 it supports the operation, but not the arm with 3634 NFS4ERR_UNION_NOTSUPP. 3636 If the data to be returned is comprised entirely of zeros, then the 3637 server may elect to return that data as a hole. The server 3638 differentiates this to the client by setting di_allocated to TRUE in 3639 this case. Note that in such a scenario, the server is not required 3640 to determine the full extent of the "hole" - it does not need to 3641 determine where the zeros start and end. 3643 The server may elect to return adjacent elements of the same type. 3644 For example, the guard pattern or block size of an ADB might change, 3645 which would require adjacent elements of type ADB. Likewise if the 3646 server has a range of data comprised entirely of zeros and then a 3647 hole, it might want to return two adjacent holes to the client. 3649 If the client specifies a rpa_count value of zero, the READ_PLUS 3650 succeeds and returns zero bytes of data, again subject to access 3651 permissions checking. In all situations, the server may choose to 3652 return fewer bytes than specified by the client. The client needs to 3653 check for this condition and handle the condition appropriately. 3655 If the client specifies an rpa_offset and rpa_count value that is 3656 entirely contained within a hole of the file, then the di_offset and 3657 di_length returned must be for the entire hole. This result is 3658 considered valid until the file is changed (detected via the change 3659 attribute). The server MUST provide the same semantics for the hole 3660 as if the client read the region and received zeroes; the implied 3661 holes contents lifetime MUST be exactly the same as any other read 3662 data. 3664 If the client specifies an rpa_offset and rpa_count value that begins 3665 in a non-hole of the file but extends into hole the server should 3666 return an array comprised of both data and a hole. The client MUST 3667 be prepared for the server to return a short read describing just the 3668 data. The client will then issue another READ_PLUS for the remaining 3669 bytes, which the server will respond with information about the hole 3670 in the file. 3672 Except when special stateids are used, the stateid value for a 3673 READ_PLUS request represents a value returned from a previous byte- 3674 range lock or share reservation request or the stateid associated 3675 with a delegation. The stateid identifies the associated owners if 3676 any and is used by the server to verify that the associated locks are 3677 still valid (e.g., have not been revoked). 3679 If the read ended at the end-of-file (formally, in a correctly formed 3680 READ_PLUS operation, if rpa_offset + rpa_count is equal to the size 3681 of the file), or the READ_PLUS operation extends beyond the size of 3682 the file (if rpa_offset + rpa_count is greater than the size of the 3683 file), eof is returned as TRUE; otherwise, it is FALSE. A successful 3684 READ_PLUS of an empty file will always return eof as TRUE. 3686 If the current filehandle is not an ordinary file, an error will be 3687 returned to the client. In the case that the current filehandle 3688 represents an object of type NF4DIR, NFS4ERR_ISDIR is returned. If 3689 the current filehandle designates a symbolic link, NFS4ERR_SYMLINK is 3690 returned. In all other cases, NFS4ERR_WRONG_TYPE is returned. 3692 For a READ_PLUS with a stateid value of all bits equal to zero, the 3693 server MAY allow the READ_PLUS to be serviced subject to mandatory 3694 byte-range locks or the current share deny modes for the file. For a 3695 READ_PLUS with a stateid value of all bits equal to one, the server 3696 MAY allow READ_PLUS operations to bypass locking checks at the 3697 server. 3699 On success, the current filehandle retains its value. 3701 13.10.4. IMPLEMENTATION 3703 In general, the IMPLEMENTATION notes for READ in Section 18.22.4 of 3704 [2] also apply to READ_PLUS. One delta is that when the owner has a 3705 locked byte range, the server MUST return an array of rpr_contents 3706 with values inside that range. 3708 13.10.4.1. Additional pNFS Implementation Information 3710 With pNFS, the semantics of using READ_PLUS remains the same. Any 3711 data server MAY return a hole or ADB result for a READ_PLUS request 3712 that it receives. 3714 When a data server chooses to return a hole result, it has the option 3715 of returning hole information for the data stored on that data server 3716 (as defined by the data layout), but it MUST not return results for a 3717 byte range that includes data managed by another data server. Data 3718 servers that can obtain hole information for the parts of the file 3719 stored on that data server, the data server SHOULD return HOLE_INFO 3720 and the byte range of the hole stored on that data server. 3722 A data server should do its best to return as much information about 3723 a hole as is feasible without having to contact the metadata server. 3724 If communication with the metadata server is required, then every 3725 attempt should be taken to minimize the number of requests. 3727 If mandatory locking is enforced, then the data server must also 3728 ensure that to return only information for a Hole that is within the 3729 owner's locked byte range. 3731 13.10.5. READ_PLUS with Sparse Files Example 3733 The following table describes a sparse file. For each byte range, 3734 the file contains either non-zero data or a hole. In addition, the 3735 server in this example uses a Hole Threshold of 32K. 3737 +-------------+----------+ 3738 | Byte-Range | Contents | 3739 +-------------+----------+ 3740 | 0-15999 | Hole | 3741 | 16K-31999 | Non-Zero | 3742 | 32K-255999 | Hole | 3743 | 256K-287999 | Non-Zero | 3744 | 288K-353999 | Hole | 3745 | 354K-417999 | Non-Zero | 3746 +-------------+----------+ 3748 Table 4 3750 Under the given circumstances, if a client was to read from the file 3751 with a max read size of 64K, the following will be the results for 3752 the given READ_PLUS calls. This assumes the client has already 3753 opened the file, acquired a valid stateid ('s' in the example), and 3754 just needs to issue READ_PLUS requests. 3756 1. READ_PLUS(s, 0, 64K) --> NFS_OK, eof = false, . Since the first hole is less than the server's 3758 Hole Threshhold, the first 32K of the file is returned as data 3759 and the remaining 32K is returned as a hole which actually 3760 extends to 256K. 3762 2. READ_PLUS(s, 32K, 64K) --> NFS_OK, eof = false, 3763 The requested range was all zeros, and the current hole begins at 3764 offset 32K and is 224K in length. Note that the client should 3765 not have followed up the previous READ_PLUS request with this one 3766 as the hole information from the previous call extended past what 3767 the client was requesting. 3769 3. READ_PLUS(s, 256K, 64K) --> NFS_OK, eof = false, . Returns an array of the 32K data and 3771 the hole which extends to 354K. 3773 4. READ_PLUS(s, 354K, 64K) --> NFS_OK, eof = true, . Returns the final 64K of data and informs the client 3775 there is no more data in the file. 3777 13.11. Operation 66: SEEK 3779 SEEK is an operation that allows a client to determine the location 3780 of the next data_content4 in a file. It allows an implementation of 3781 the emerging extension to lseek(2) to allow clients to determine 3782 SEEK_HOLE and SEEK_DATA. 3784 13.11.1. ARGUMENT 3786 struct SEEK4args { 3787 /* CURRENT_FH: file */ 3788 stateid4 sa_stateid; 3789 offset4 sa_offset; 3790 data_content4 sa_what; 3791 }; 3793 13.11.2. RESULT 3795 union seek_content switch (data_content4 content) { 3796 case NFS4_CONTENT_DATA: 3797 data_info4 sc_data; 3798 case NFS4_CONTENT_APP_BLOCK: 3799 app_data_block4 sc_block; 3800 case NFS4_CONTENT_HOLE: 3801 data_info4 sc_hole; 3802 default: 3803 void; 3804 }; 3806 struct seek_res4 { 3807 bool sr_eof; 3808 seek_content sr_contents; 3809 }; 3811 union SEEK4res switch (nfsstat4 status) { 3812 case NFS4_OK: 3813 seek_res4 resok4; 3814 default: 3815 void; 3816 }; 3818 13.11.3. DESCRIPTION 3820 From the given sa_offset, find the next data_content4 of type sa_what 3821 in the file. For either a hole or ADB, this must return the 3822 data_content4 in its entirety. For data, it must not return the 3823 actual data. 3825 SEEK must follow the same rules for stateids as READ_PLUS 3826 (Section 13.10.3). 3828 If the server could not find a corresponding sa_what, then the status 3829 would still be NFS4_OK, but sr_eof would be TRUE. The sr_contents 3830 would contain a zero-ed out content of the appropriate type. 3832 14. NFSv4.2 Callback Operations 3834 14.1. Procedure 16: CB_ATTR_CHANGED - Notify Client that the File's 3835 Attributes Changed 3837 14.1.1. ARGUMENTS 3839 struct CB_ATTR_CHANGED4args { 3840 nfs_fh4 acca_fh; 3841 bitmap4 acca_critical; 3842 bitmap4 acca_info; 3843 }; 3845 14.1.2. RESULTS 3847 struct CB_ATTR_CHANGED4res { 3848 nfsstat4 accr_status; 3849 }; 3851 14.1.3. DESCRIPTION 3853 The CB_ATTR_CHANGED callback operation is used by the server to 3854 indicate to the client that the file's attributes have been modified 3855 on the server. The server does not convey how the attributes have 3856 changed, just that they have been modified. The server can inform 3857 the client about both critical and informational attribute changes in 3858 the bitmask arguments. The client SHOULD query the server about all 3859 attributes set in acca_critical. For all changes reflected in 3860 acca_info, the client can decide whether or not it wants to poll the 3861 server. 3863 The CB_ATTR_CHANGED callback operation with the FATTR4_SEC_LABEL set 3864 in acca_critical is the method used by the server to indicate that 3865 the MAC label for the file referenced by acca_fh has changed. In 3866 many ways, the server does not care about the result returned by the 3867 client. 3869 14.2. Operation 15: CB_COPY - Report results of a server-side copy 3870 14.2.1. ARGUMENT 3872 union copy_info4 switch (nfsstat4 cca_status) { 3873 case NFS4_OK: 3874 void; 3875 default: 3876 length4 cca_bytes_copied; 3877 }; 3879 struct CB_COPY4args { 3880 nfs_fh4 cca_fh; 3881 stateid4 cca_stateid; 3882 copy_info4 cca_copy_info; 3883 }; 3885 14.2.2. RESULT 3887 struct CB_COPY4res { 3888 nfsstat4 ccr_status; 3889 }; 3891 14.2.3. DESCRIPTION 3893 CB_COPY is used for both intra- and inter-server asynchronous copies. 3894 The CB_COPY callback informs the client of the result of an 3895 asynchronous server-side copy. This operation is sent by the 3896 destination server to the client in a CB_COMPOUND request. The copy 3897 is identified by the filehandle and stateid arguments. The result is 3898 indicated by the status field. If the copy failed, cca_bytes_copied 3899 contains the number of bytes copied before the failure occurred. The 3900 cca_bytes_copied value indicates the number of bytes copied but not 3901 which specific bytes have been copied. 3903 In the absence of an established backchannel, the server cannot 3904 signal the completion of the COPY via a CB_COPY callback. The loss 3905 of a callback channel would be indicated by the server setting the 3906 SEQ4_STATUS_CB_PATH_DOWN flag in the sr_status_flags field of the 3907 SEQUENCE operation. The client must re-establish the callback 3908 channel to receive the status of the COPY operation. Prolonged loss 3909 of the callback channel could result in the server dropping the COPY 3910 operation state and invalidating the copy stateid. 3912 If the client supports the COPY operation, the client is REQUIRED to 3913 support the CB_COPY operation. 3915 The CB_COPY operation may fail for the following reasons (this is a 3916 partial list): 3918 NFS4ERR_NOTSUPP: The copy offload operation is not supported by the 3919 NFS client receiving this request. 3921 15. IANA Considerations 3923 This section uses terms that are defined in [24]. 3925 16. References 3927 16.1. Normative References 3929 [1] Bradner, S., "Key words for use in RFCs to Indicate Requirement 3930 Levels", March 1997. 3932 [2] Shepler, S., Eisler, M., and D. Noveck, "Network File System 3933 (NFS) Version 4 Minor Version 1 Protocol", RFC 5661, 3934 January 2010. 3936 [3] Haynes, T., "Network File System (NFS) Version 4 Minor Version 3937 2 External Data Representation Standard (XDR) Description", 3938 March 2011. 3940 [4] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform 3941 Resource Identifier (URI): Generic Syntax", STD 66, RFC 3986, 3942 January 2005. 3944 [5] Haynes, T. and N. Williams, "Remote Procedure Call (RPC) 3945 Security Version 3", draft-williams-rpcsecgssv3 (work in 3946 progress), 2011. 3948 [6] The Open Group, "Section 'posix_fadvise()' of System Interfaces 3949 of The Open Group Base Specifications Issue 6, IEEE Std 1003.1, 3950 2004 Edition", 2004. 3952 [7] Haynes, T., "Requirements for Labeled NFS", 3953 draft-ietf-nfsv4-labreqs-00 (work in progress). 3955 [8] Eisler, M., Chiu, A., and L. Ling, "RPCSEC_GSS Protocol 3956 Specification", RFC 2203, September 1997. 3958 [9] Halevy, B., Welch, B., and J. Zelenka, "Object-Based Parallel 3959 NFS (pNFS) Operations", RFC 5664, January 2010. 3961 16.2. Informative References 3963 [10] Haynes, T. and D. Noveck, "Network File System (NFS) version 4 3964 Protocol", draft-ietf-nfsv4-rfc3530bis-09 (Work In Progress), 3965 March 2011. 3967 [11] Lentini, J., Everhart, C., Ellard, D., Tewari, R., and M. Naik, 3968 "NSDB Protocol for Federated Filesystems", 3969 draft-ietf-nfsv4-federated-fs-protocol (Work In Progress), 3970 2010. 3972 [12] Lentini, J., Everhart, C., Ellard, D., Tewari, R., and M. Naik, 3973 "Administration Protocol for Federated Filesystems", 3974 draft-ietf-nfsv4-federated-fs-admin (Work In Progress), 2010. 3976 [13] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L., 3977 Leach, P., and T. Berners-Lee, "Hypertext Transfer Protocol -- 3978 HTTP/1.1", RFC 2616, June 1999. 3980 [14] Postel, J. and J. Reynolds, "File Transfer Protocol", STD 9, 3981 RFC 959, October 1985. 3983 [15] Simpson, W., "PPP Challenge Handshake Authentication Protocol 3984 (CHAP)", RFC 1994, August 1996. 3986 [16] VanDeBogart, S., Frost, C., and E. Kohler, "Reducing Seek 3987 Overhead with Application-Directed Prefetching", Proceedings of 3988 USENIX Annual Technical Conference , June 2009. 3990 [17] Strohm, R., "Chapter 2, Data Blocks, Extents, and Segments, of 3991 Oracle Database Concepts 11g Release 1 (11.1)", January 2011. 3993 [18] Ashdown, L., "Chapter 15, Validating Database Files and 3994 Backups, of Oracle Database Backup and Recovery User's Guide 3995 11g Release 1 (11.1)", August 2008. 3997 [19] McDougall, R. and J. Mauro, "Section 11.4.3, Detecting Memory 3998 Corruption of Solaris Internals", 2007. 4000 [20] Bairavasundaram, L., Goodson, G., Schroeder, B., Arpaci- 4001 Dusseau, A., and R. Arpaci-Dusseau, "An Analysis of Data 4002 Corruption in the Storage Stack", Proceedings of the 6th USENIX 4003 Symposium on File and Storage Technologies (FAST '08) , 2008. 4005 [21] "Section 46.6. Multi-Level Security (MLS) of Deployment Guide: 4006 Deployment, configuration and administration of Red Hat 4007 Enterprise Linux 5, Edition 6", 2011. 4009 [22] Quigley, D. and J. Lu, "Registry Specification for MAC Security 4010 Label Formats", draft-quigley-label-format-registry (work in 4011 progress), 2011. 4013 [23] Eisler, M., "XDR: External Data Representation Standard", 4014 RFC 4506, May 2006. 4016 [24] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA 4017 Considerations Section in RFCs", BCP 26, RFC 5226, May 2008. 4019 Appendix A. Acknowledgments 4021 For the pNFS Access Permissions Check, the original draft was by 4022 Sorin Faibish, David Black, Mike Eisler, and Jason Glasgow. The work 4023 was influenced by discussions with Benny Halevy and Bruce Fields. A 4024 review was done by Tom Haynes. 4026 For the Sharing change attribute implementation details with NFSv4 4027 clients, the original draft was by Trond Myklebust. 4029 For the NFS Server-side Copy, the original draft was by James 4030 Lentini, Mike Eisler, Deepak Kenchammana, Anshul Madan, and Rahul 4031 Iyer. Tom Talpey co-authored an unpublished version of that 4032 document. It was also was reviewed by a number of individuals: 4033 Pranoop Erasani, Tom Haynes, Arthur Lent, Trond Myklebust, Dave 4034 Noveck, Theresa Lingutla-Raj, Manjunath Shankararao, Satyam Vaghani, 4035 and Nico Williams. 4037 For the NFS space reservation operations, the original draft was by 4038 Mike Eisler, James Lentini, Manjunath Shankararao, and Rahul Iyer. 4040 For the sparse file support, the original draft was by Dean 4041 Hildebrand and Marc Eshel. Valuable input and advice was received 4042 from Sorin Faibish, Bruce Fields, Benny Halevy, Trond Myklebust, and 4043 Richard Scheffenegger. 4045 For the Application IO Hints, the original draft was by Dean 4046 Hildebrand, Mike Eisler, Trond Myklebust, and Sam Falkner. Some 4047 early reviwers included Benny Halevy and Pranoop Erasani. 4049 For Labeled NFS, the original draft was by David Quigley, James 4050 Morris, Jarret Lu, and Tom Haynes. Peter Staubach, Trond Myklebust, 4051 Stephen Smalley, Sorrin Faibish, Nico Williams, and David Black also 4052 contributed in the final push to get this accepted. 4054 Appendix B. RFC Editor Notes 4056 [RFC Editor: please remove this section prior to publishing this 4057 document as an RFC] 4059 [RFC Editor: prior to publishing this document as an RFC, please 4060 replace all occurrences of RFCTBD10 with RFCxxxx where xxxx is the 4061 RFC number of this document] 4063 Author's Address 4065 Thomas Haynes 4066 NetApp 4067 9110 E 66th St 4068 Tulsa, OK 74133 4069 USA 4071 Phone: +1 918 307 1415 4072 Email: thomas@netapp.com 4073 URI: http://www.tulsalabs.com