blkif.h revision 231743
1/****************************************************************************** 2 * blkif.h 3 * 4 * Unified block-device I/O interface for Xen guest OSes. 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to 8 * deal in the Software without restriction, including without limitation the 9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or 10 * sell copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 22 * DEALINGS IN THE SOFTWARE. 23 * 24 * Copyright (c) 2003-2004, Keir Fraser 25 * Copyright (c) 2012, Spectra Logic Corporation 26 */ 27 28#ifndef __XEN_PUBLIC_IO_BLKIF_H__ 29#define __XEN_PUBLIC_IO_BLKIF_H__ 30 31#include <xen/interface/io/ring.h> 32#include <xen/interface/grant_table.h> 33 34/* 35 * Front->back notifications: When enqueuing a new request, sending a 36 * notification can be made conditional on req_event (i.e., the generic 37 * hold-off mechanism provided by the ring macros). Backends must set 38 * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()). 39 * 40 * Back->front notifications: When enqueuing a new response, sending a 41 * notification can be made conditional on rsp_event (i.e., the generic 42 * hold-off mechanism provided by the ring macros). Frontends must set 43 * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()). 44 */ 45 46#ifndef blkif_vdev_t 47#define blkif_vdev_t uint16_t 48#endif 49#define blkif_sector_t uint64_t 50 51/* 52 * Feature and Parameter Negotiation 53 * ================================= 54 * The two halves of a Xen block driver utilize nodes within the XenStore to 55 * communicate capabilities and to negotiate operating parameters. This 56 * section enumerates these nodes which reside in the respective front and 57 * backend portions of the XenStore, following the XenBus convention. 58 * 59 * All data in the XenStore is stored as strings. Nodes specifying numeric 60 * values are encoded in decimal. Integer value ranges listed below are 61 * expressed as fixed sized integer types capable of storing the conversion 62 * of a properly formated node string, without loss of information. 63 * 64 * Any specified default value is in effect if the corresponding XenBus node 65 * is not present in the XenStore. 66 * 67 * XenStore nodes in sections marked "PRIVATE" are solely for use by the 68 * driver side whose XenBus tree contains them. 69 * 70 * See the XenBus state transition diagram below for details on when XenBus 71 * nodes must be published and when they can be queried. 72 * 73 ***************************************************************************** 74 * Backend XenBus Nodes 75 ***************************************************************************** 76 * 77 *------------------ Backend Device Identification (PRIVATE) ------------------ 78 * 79 * mode 80 * Values: "r" (read only), "w" (writable) 81 * 82 * The read or write access permissions to the backing store to be 83 * granted to the frontend. 84 * 85 * params 86 * Values: string 87 * 88 * A free formatted string providing sufficient information for the 89 * backend driver to open the backing device. (e.g. the path to the 90 * file or block device representing the backing store.) 91 * 92 * type 93 * Values: "file", "phy", "tap" 94 * 95 * The type of the backing device/object. 96 * 97 *--------------------------------- Features --------------------------------- 98 * 99 * feature-barrier 100 * Values: 0/1 (boolean) 101 * Default Value: 0 102 * 103 * A value of "1" indicates that the backend can process requests 104 * containing the BLKIF_OP_WRITE_BARRIER request opcode. Requests 105 * of this type may still be returned at any time with the 106 * BLKIF_RSP_EOPNOTSUPP result code. 107 * 108 * feature-flush-cache 109 * Values: 0/1 (boolean) 110 * Default Value: 0 111 * 112 * A value of "1" indicates that the backend can process requests 113 * containing the BLKIF_OP_FLUSH_DISKCACHE request opcode. Requests 114 * of this type may still be returned at any time with the 115 * BLKIF_RSP_EOPNOTSUPP result code. 116 * 117 * feature-discard 118 * Values: 0/1 (boolean) 119 * Default Value: 0 120 * 121 * A value of "1" indicates that the backend can process requests 122 * containing the BLKIF_OP_DISCARD request opcode. Requests 123 * of this type may still be returned at any time with the 124 * BLKIF_RSP_EOPNOTSUPP result code. 125 * 126 *----------------------- Request Transport Parameters ------------------------ 127 * 128 * max-ring-page-order 129 * Values: <uint32_t> 130 * Default Value: 0 131 * Notes: 1, 3 132 * 133 * The maximum supported size of the request ring buffer in units of 134 * lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 135 * etc.). 136 * 137 * max-ring-pages 138 * Values: <uint32_t> 139 * Default Value: 1 140 * Notes: 2, 3 141 * 142 * The maximum supported size of the request ring buffer in units of 143 * machine pages. The value must be a power of 2. 144 * 145 * max-requests <uint32_t> 146 * Default Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE) 147 * Maximum Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE * max-ring-pages) 148 * 149 * The maximum number of concurrent, logical requests that will be 150 * issued by the backend. 151 * 152 * Note: A logical request may span multiple ring entries. 153 * 154 * max-request-segments 155 * Values: <uint8_t> 156 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK 157 * Maximum Value: BLKIF_MAX_SEGMENTS_PER_REQUEST 158 * 159 * The maximum value of blkif_request.nr_segments supported by 160 * the backend. 161 * 162 * max-request-size 163 * Values: <uint32_t> 164 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK * PAGE_SIZE 165 * Maximum Value: BLKIF_MAX_SEGMENTS_PER_REQUEST * PAGE_SIZE 166 * 167 * The maximum amount of data, in bytes, that can be referenced by a 168 * request type that accesses frontend memory (currently BLKIF_OP_READ, 169 * BLKIF_OP_WRITE, or BLKIF_OP_WRITE_BARRIER). 170 * 171 *------------------------- Backend Device Properties ------------------------- 172 * 173 * discard-aligment 174 * Values: <uint32_t> 175 * Default Value: 0 176 * Notes: 4, 5 177 * 178 * The offset, in bytes from the beginning of the virtual block device, 179 * to the first, addressable, discard extent on the underlying device. 180 * 181 * discard-granularity 182 * Values: <uint32_t> 183 * Default Value: <"sector-size"> 184 * Notes: 4 185 * 186 * The size, in bytes, of the individually addressable discard extents 187 * of the underlying device. 188 * 189 * discard-secure 190 * Values: 0/1 (boolean) 191 * Default Value: 0 192 * 193 * A value of "1" indicates that the backend can process BLKIF_OP_DISCARD 194 * requests with the BLKIF_DISCARD_SECURE flag set. 195 * 196 * info 197 * Values: <uint32_t> (bitmap) 198 * 199 * A collection of bit flags describing attributes of the backing 200 * device. The VDISK_* macros define the meaning of each bit 201 * location. 202 * 203 * sector-size 204 * Values: <uint32_t> 205 * 206 * The native sector size, in bytes, of the backend device. 207 * 208 * sectors 209 * Values: <uint64_t> 210 * 211 * The size of the backend device, expressed in units of its native 212 * sector size ("sector-size"). 213 * 214 ***************************************************************************** 215 * Frontend XenBus Nodes 216 ***************************************************************************** 217 * 218 *----------------------- Request Transport Parameters ----------------------- 219 * 220 * event-channel 221 * Values: <uint32_t> 222 * 223 * The identifier of the Xen event channel used to signal activity 224 * in the ring buffer. 225 * 226 * ring-ref 227 * Values: <uint32_t> 228 * Notes: 6 229 * 230 * The Xen grant reference granting permission for the backend to map 231 * the sole page in a single page sized ring buffer. 232 * 233 * ring-ref%u 234 * Values: <uint32_t> 235 * Notes: 6 236 * 237 * For a frontend providing a multi-page ring, a "num-ring-pages" sized 238 * list of nodes, each containing a Xen grant reference granting 239 * permission for the backend to map the page of the ring located 240 * at page index "%u". Page indexes are zero based. 241 * 242 * protocol 243 * Values: string (XEN_IO_PROTO_ABI_*) 244 * Default Value: XEN_IO_PROTO_ABI_NATIVE 245 * 246 * The machine ABI rules governing the format of all ring request and 247 * response structures. 248 * 249 * ring-page-order 250 * Values: <uint32_t> 251 * Default Value: 0 252 * Maximum Value: MAX(ffs(max-ring-pages) - 1, max-ring-page-order) 253 * Notes: 1, 3 254 * 255 * The size of the frontend allocated request ring buffer in units 256 * of lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 257 * etc.). 258 * 259 * num-ring-pages 260 * Values: <uint32_t> 261 * Default Value: 1 262 * Maximum Value: MAX(max-ring-pages,(0x1 << max-ring-page-order)) 263 * Notes: 2, 3 264 * 265 * The size of the frontend allocated request ring buffer in units of 266 * machine pages. The value must be a power of 2. 267 * 268 * max-requests 269 * Values: <uint32_t> 270 * Default Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE) 271 * Maximum Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE * max-ring_pages) 272 * 273 * The maximum number of concurrent, logical requests that will be 274 * issued by the frontend. 275 * 276 * Note: A logical request may span multiple ring entries. 277 * 278 * max-request-segments 279 * Values: <uint8_t> 280 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK 281 * Maximum Value: MIN(255, backend/max-request-segments) 282 * 283 * The maximum value the frontend will set in the 284 * blkif_request.nr_segments field. 285 * 286 * max-request-size 287 * Values: <uint32_t> 288 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK * PAGE_SIZE 289 * Maximum Value: max-request-segments * PAGE_SIZE 290 * 291 * The maximum amount of data, in bytes, that can be referenced by 292 * a request type that accesses frontend memory (currently BLKIF_OP_READ, 293 * BLKIF_OP_WRITE, or BLKIF_OP_WRITE_BARRIER). 294 * 295 *------------------------- Virtual Device Properties ------------------------- 296 * 297 * device-type 298 * Values: "disk", "cdrom", "floppy", etc. 299 * 300 * virtual-device 301 * Values: <uint32_t> 302 * 303 * A value indicating the physical device to virtualize within the 304 * frontend's domain. (e.g. "The first ATA disk", "The third SCSI 305 * disk", etc.) 306 * 307 * See docs/misc/vbd-interface.txt for details on the format of this 308 * value. 309 * 310 * Notes 311 * ----- 312 * (1) Multi-page ring buffer scheme first developed in the Citrix XenServer 313 * PV drivers. 314 * (2) Multi-page ring buffer scheme first used in some RedHat distributions 315 * including a distribution deployed on certain nodes of the Amazon 316 * EC2 cluster. 317 * (3) Support for multi-page ring buffers was implemented independently, 318 * in slightly different forms, by both Citrix and RedHat/Amazon. 319 * For full interoperability, block front and backends should publish 320 * identical ring parameters, adjusted for unit differences, to the 321 * XenStore nodes used in both schemes. 322 * (4) Devices that support discard functionality may internally allocate 323 * space (discardable extents) in units that are larger than the 324 * exported logical block size. 325 * (5) The discard-alignment parameter allows a physical device to be 326 * partitioned into virtual devices that do not necessarily begin or 327 * end on a discardable extent boundary. 328 * (6) When there is only a single page allocated to the request ring, 329 * 'ring-ref' is used to communicate the grant reference for this 330 * page to the backend. When using a multi-page ring, the 'ring-ref' 331 * node is not created. Instead 'ring-ref0' - 'ring-refN' are used. 332 */ 333 334/* 335 * STATE DIAGRAMS 336 * 337 ***************************************************************************** 338 * Startup * 339 ***************************************************************************** 340 * 341 * Tool stack creates front and back nodes with state XenbusStateInitialising. 342 * 343 * Front Back 344 * ================================= ===================================== 345 * XenbusStateInitialising XenbusStateInitialising 346 * o Query virtual device o Query backend device identification 347 * properties. data. 348 * o Setup OS device instance. o Open and validate backend device. 349 * o Publish backend features and 350 * transport parameters. 351 * | 352 * | 353 * V 354 * XenbusStateInitWait 355 * 356 * o Query backend features and 357 * transport parameters. 358 * o Allocate and initialize the 359 * request ring. 360 * o Publish transport parameters 361 * that will be in effect during 362 * this connection. 363 * | 364 * | 365 * V 366 * XenbusStateInitialised 367 * 368 * o Query frontend transport parameters. 369 * o Connect to the request ring and 370 * event channel. 371 * o Publish backend device properties. 372 * | 373 * | 374 * V 375 * XenbusStateConnected 376 * 377 * o Query backend device properties. 378 * o Finalize OS virtual device 379 * instance. 380 * | 381 * | 382 * V 383 * XenbusStateConnected 384 * 385 * Note: Drivers that do not support any optional features, or the negotiation 386 * of transport parameters, can skip certain states in the state machine: 387 * 388 * o A frontend may transition to XenbusStateInitialised without 389 * waiting for the backend to enter XenbusStateInitWait. In this 390 * case, default transport parameters are in effect and any 391 * transport parameters published by the frontend must contain 392 * their default values. 393 * 394 * o A backend may transition to XenbusStateInitialised, bypassing 395 * XenbusStateInitWait, without waiting for the frontend to first 396 * enter the XenbusStateInitialised state. In this case, default 397 * transport parameters are in effect and any transport parameters 398 * published by the backend must contain their default values. 399 * 400 * Drivers that support optional features and/or transport parameter 401 * negotiation must tolerate these additional state transition paths. 402 * In general this means performing the work of any skipped state 403 * transition, if it has not already been performed, in addition to the 404 * work associated with entry into the current state. 405 */ 406 407/* 408 * REQUEST CODES. 409 */ 410#define BLKIF_OP_READ 0 411#define BLKIF_OP_WRITE 1 412/* 413 * All writes issued prior to a request with the BLKIF_OP_WRITE_BARRIER 414 * operation code ("barrier request") must be completed prior to the 415 * execution of the barrier request. All writes issued after the barrier 416 * request must not execute until after the completion of the barrier request. 417 * 418 * Optional. See "feature-barrier" XenBus node documentation above. 419 */ 420#define BLKIF_OP_WRITE_BARRIER 2 421/* 422 * Commit any uncommitted contents of the backing device's volatile cache 423 * to stable storage. 424 * 425 * Optional. See "feature-flush-cache" XenBus node documentation above. 426 */ 427#define BLKIF_OP_FLUSH_DISKCACHE 3 428/* 429 * Used in SLES sources for device specific command packet 430 * contained within the request. Reserved for that purpose. 431 */ 432#define BLKIF_OP_RESERVED_1 4 433/* 434 * Indicate to the backend device that a region of storage is no longer in 435 * use, and may be discarded at any time without impact to the client. If 436 * the BLKIF_DISCARD_SECURE flag is set on the request, all copies of the 437 * discarded region on the device must be rendered unrecoverable before the 438 * command returns. 439 * 440 * This operation is analogous to performing a trim (ATA) or unamp (SCSI), 441 * command on a native device. 442 * 443 * More information about trim/unmap operations can be found at: 444 * http://t13.org/Documents/UploadedDocuments/docs2008/ 445 * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc 446 * http://www.seagate.com/staticfiles/support/disc/manuals/ 447 * Interface%20manuals/100293068c.pdf 448 * 449 * Optional. See "feature-discard", "discard-alignment", 450 * "discard-granularity", and "discard-secure" in the XenBus node 451 * documentation above. 452 */ 453#define BLKIF_OP_DISCARD 5 454 455/* 456 * Maximum scatter/gather segments associated with a request header block. 457 * This is carefully chosen so that sizeof(blkif_ring_t) <= PAGE_SIZE. 458 * NB. This could be 12 if the ring indexes weren't stored in the same page. 459 */ 460#define BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK 11 461 462/* 463 * Maximum scatter/gather segments associated with a segment block. 464 */ 465#define BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK 14 466 467/* 468 * Maximum scatter/gather segments per request (header + segment blocks). 469 */ 470#define BLKIF_MAX_SEGMENTS_PER_REQUEST 255 471 472/* 473 * NB. first_sect and last_sect in blkif_request_segment, as well as 474 * sector_number in blkif_request, are always expressed in 512-byte units. 475 * However they must be properly aligned to the real sector size of the 476 * physical disk, which is reported in the "sector-size" node in the backend 477 * xenbus info. Also the xenbus "sectors" node is expressed in 512-byte units. 478 */ 479struct blkif_request_segment { 480 grant_ref_t gref; /* reference to I/O buffer frame */ 481 /* @first_sect: first sector in frame to transfer (inclusive). */ 482 /* @last_sect: last sector in frame to transfer (inclusive). */ 483 uint8_t first_sect, last_sect; 484}; 485typedef struct blkif_request_segment blkif_request_segment_t; 486 487/* 488 * Starting ring element for any I/O request. 489 * 490 * One or more segment blocks can be inserted into the request ring 491 * just after a blkif_request_t, allowing requests to operate on 492 * up to BLKIF_MAX_SEGMENTS_PER_REQUEST. 493 * 494 * BLKIF_SEGS_TO_BLOCKS() can be used on blkif_requst.nr_segments 495 * to determine the number of contiguous ring entries associated 496 * with this request. 497 * 498 * Note: Due to the way Xen request rings operate, the producer and 499 * consumer indices of the ring must be incremented by the 500 * BLKIF_SEGS_TO_BLOCKS() value of the associated request. 501 * (e.g. a response to a 3 ring entry request must also consume 502 * 3 entries in the ring, even though only the first ring entry 503 * in the response has any data.) 504 */ 505struct blkif_request { 506 uint8_t operation; /* BLKIF_OP_??? */ 507 uint8_t nr_segments; /* number of segments */ 508 blkif_vdev_t handle; /* only for read/write requests */ 509 uint64_t id; /* private guest value, echoed in resp */ 510 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 511 blkif_request_segment_t seg[BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK]; 512}; 513typedef struct blkif_request blkif_request_t; 514 515/* 516 * A segment block is a ring request structure that contains only 517 * segment data. 518 * 519 * sizeof(struct blkif_segment_block) <= sizeof(struct blkif_request) 520 */ 521struct blkif_segment_block { 522 blkif_request_segment_t seg[BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK]; 523}; 524typedef struct blkif_segment_block blkif_segment_block_t; 525 526/* 527 * Cast to this structure when blkif_request.operation == BLKIF_OP_DISCARD 528 * sizeof(struct blkif_request_discard) <= sizeof(struct blkif_request) 529 */ 530struct blkif_request_discard { 531 uint8_t operation; /* BLKIF_OP_DISCARD */ 532 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */ 533#define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */ 534 blkif_vdev_t handle; /* same as for read/write requests */ 535 uint64_t id; /* private guest value, echoed in resp */ 536 blkif_sector_t sector_number;/* start sector idx on disk */ 537 uint64_t nr_sectors; /* number of contiguous sectors to discard*/ 538}; 539typedef struct blkif_request_discard blkif_request_discard_t; 540 541struct blkif_response { 542 uint64_t id; /* copied from request */ 543 uint8_t operation; /* copied from request */ 544 int16_t status; /* BLKIF_RSP_??? */ 545}; 546typedef struct blkif_response blkif_response_t; 547 548/* 549 * STATUS RETURN CODES. 550 */ 551 /* Operation not supported (only happens on barrier writes). */ 552#define BLKIF_RSP_EOPNOTSUPP -2 553 /* Operation failed for some unspecified reason (-EIO). */ 554#define BLKIF_RSP_ERROR -1 555 /* Operation completed successfully. */ 556#define BLKIF_RSP_OKAY 0 557 558/* 559 * Generate blkif ring structures and types. 560 */ 561DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response); 562 563/* 564 * Index to, and treat as a segment block, an entry in the ring. 565 */ 566#define BLKRING_GET_SEG_BLOCK(_r, _idx) \ 567 (((blkif_segment_block_t *)RING_GET_REQUEST(_r, _idx))->seg) 568 569/* 570 * The number of ring request blocks required to handle an I/O 571 * request containing _segs segments. 572 */ 573#define BLKIF_SEGS_TO_BLOCKS(_segs) \ 574 ((((_segs - BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK) \ 575 + (BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK - 1)) \ 576 / BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK) + /*header_block*/1) 577 578#define VDISK_CDROM 0x1 579#define VDISK_REMOVABLE 0x2 580#define VDISK_READONLY 0x4 581 582#endif /* __XEN_PUBLIC_IO_BLKIF_H__ */ 583 584/* 585 * Local variables: 586 * mode: C 587 * c-set-style: "BSD" 588 * c-basic-offset: 4 589 * tab-width: 4 590 * indent-tabs-mode: nil 591 * End: 592 */ 593