blkif.h revision 232308
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 formatted 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 * XenStore nodes marked "DEPRECATED" in their notes section should only be 71 * used to provide interoperability with legacy implementations. 72 * 73 * See the XenBus state transition diagram below for details on when XenBus 74 * nodes must be published and when they can be queried. 75 * 76 ***************************************************************************** 77 * Backend XenBus Nodes 78 ***************************************************************************** 79 * 80 *------------------ Backend Device Identification (PRIVATE) ------------------ 81 * 82 * mode 83 * Values: "r" (read only), "w" (writable) 84 * 85 * The read or write access permissions to the backing store to be 86 * granted to the frontend. 87 * 88 * params 89 * Values: string 90 * 91 * Data used by the backend driver to locate and configure the backing 92 * device. The format and semantics of this data vary according to the 93 * backing device in use and are outside the scope of this specification. 94 * 95 * type 96 * Values: "file", "phy", "tap" 97 * 98 * The type of the backing device/object. 99 * 100 *--------------------------------- Features --------------------------------- 101 * 102 * feature-barrier 103 * Values: 0/1 (boolean) 104 * Default Value: 0 105 * 106 * A value of "1" indicates that the backend can process requests 107 * containing the BLKIF_OP_WRITE_BARRIER request opcode. Requests 108 * of this type may still be returned at any time with the 109 * BLKIF_RSP_EOPNOTSUPP result code. 110 * 111 * feature-flush-cache 112 * Values: 0/1 (boolean) 113 * Default Value: 0 114 * 115 * A value of "1" indicates that the backend can process requests 116 * containing the BLKIF_OP_FLUSH_DISKCACHE request opcode. Requests 117 * of this type may still be returned at any time with the 118 * BLKIF_RSP_EOPNOTSUPP result code. 119 * 120 * feature-discard 121 * Values: 0/1 (boolean) 122 * Default Value: 0 123 * 124 * A value of "1" indicates that the backend can process requests 125 * containing the BLKIF_OP_DISCARD request opcode. Requests 126 * of this type may still be returned at any time with the 127 * BLKIF_RSP_EOPNOTSUPP result code. 128 * 129 *----------------------- Request Transport Parameters ------------------------ 130 * 131 * max-ring-page-order 132 * Values: <uint32_t> 133 * Default Value: 0 134 * Notes: 1, 3 135 * 136 * The maximum supported size of the request ring buffer in units of 137 * lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 138 * etc.). 139 * 140 * max-ring-pages 141 * Values: <uint32_t> 142 * Default Value: 1 143 * Notes: DEPRECATED, 2, 3 144 * 145 * The maximum supported size of the request ring buffer in units of 146 * machine pages. The value must be a power of 2. 147 * 148 * max-requests <uint32_t> 149 * Default Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE) 150 * Maximum Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE * max-ring-pages) 151 * 152 * The maximum number of concurrent, logical requests supported by 153 * the backend. 154 * 155 * Note: A logical request may span multiple ring entries. 156 * 157 * max-request-segments 158 * Values: <uint8_t> 159 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK 160 * Maximum Value: BLKIF_MAX_SEGMENTS_PER_REQUEST 161 * 162 * The maximum value of blkif_request.nr_segments supported by 163 * the backend. 164 * 165 * max-request-size 166 * Values: <uint32_t> 167 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK * PAGE_SIZE 168 * Maximum Value: BLKIF_MAX_SEGMENTS_PER_REQUEST * PAGE_SIZE 169 * 170 * The maximum amount of data, in bytes, that can be referenced by a 171 * request type that accesses frontend memory (currently BLKIF_OP_READ, 172 * BLKIF_OP_WRITE, or BLKIF_OP_WRITE_BARRIER). 173 * 174 *------------------------- Backend Device Properties ------------------------- 175 * 176 * discard-alignment 177 * Values: <uint32_t> 178 * Default Value: 0 179 * Notes: 4, 5 180 * 181 * The offset, in bytes from the beginning of the virtual block device, 182 * to the first, addressable, discard extent on the underlying device. 183 * 184 * discard-granularity 185 * Values: <uint32_t> 186 * Default Value: <"sector-size"> 187 * Notes: 4 188 * 189 * The size, in bytes, of the individually addressable discard extents 190 * of the underlying device. 191 * 192 * discard-secure 193 * Values: 0/1 (boolean) 194 * Default Value: 0 195 * 196 * A value of "1" indicates that the backend can process BLKIF_OP_DISCARD 197 * requests with the BLKIF_DISCARD_SECURE flag set. 198 * 199 * info 200 * Values: <uint32_t> (bitmap) 201 * 202 * A collection of bit flags describing attributes of the backing 203 * device. The VDISK_* macros define the meaning of each bit 204 * location. 205 * 206 * sector-size 207 * Values: <uint32_t> 208 * 209 * The size, in bytes, of the individually addressible data blocks 210 * on the backend device. 211 * 212 * sectors 213 * Values: <uint64_t> 214 * 215 * The size of the backend device, expressed in units of its native 216 * sector size ("sector-size"). 217 * 218 ***************************************************************************** 219 * Frontend XenBus Nodes 220 ***************************************************************************** 221 * 222 *----------------------- Request Transport Parameters ----------------------- 223 * 224 * event-channel 225 * Values: <uint32_t> 226 * 227 * The identifier of the Xen event channel used to signal activity 228 * in the ring buffer. 229 * 230 * ring-ref 231 * Values: <uint32_t> 232 * Notes: 6 233 * 234 * The Xen grant reference granting permission for the backend to map 235 * the sole page in a single page sized ring buffer. 236 * 237 * ring-ref%u 238 * Values: <uint32_t> 239 * Notes: 6 240 * 241 * For a frontend providing a multi-page ring, a "number of ring pages" 242 * sized list of nodes, each containing a Xen grant reference granting 243 * permission for the backend to map the page of the ring located 244 * at page index "%u". Page indexes are zero based. 245 * 246 * protocol 247 * Values: string (XEN_IO_PROTO_ABI_*) 248 * Default Value: XEN_IO_PROTO_ABI_NATIVE 249 * 250 * The machine ABI rules governing the format of all ring request and 251 * response structures. 252 * 253 * ring-page-order 254 * Values: <uint32_t> 255 * Default Value: 0 256 * Maximum Value: MAX(ffs(max-ring-pages) - 1, max-ring-page-order) 257 * Notes: 1, 3 258 * 259 * The size of the frontend allocated request ring buffer in units 260 * of lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 261 * etc.). 262 * 263 * num-ring-pages 264 * Values: <uint32_t> 265 * Default Value: 1 266 * Maximum Value: MAX(max-ring-pages,(0x1 << max-ring-page-order)) 267 * Notes: DEPRECATED, 2, 3 268 * 269 * The size of the frontend allocated request ring buffer in units of 270 * machine pages. The value must be a power of 2. 271 * 272 * max-requests 273 * Values: <uint32_t> 274 * Default Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE) 275 * Maximum Value: BLKIF_MAX_RING_REQUESTS(PAGE_SIZE * max-ring-pages) 276 * 277 * The maximum number of concurrent, logical requests that will be 278 * issued by the frontend. 279 * 280 * Note: A logical request may span multiple ring entries. 281 * 282 * max-request-segments 283 * Values: <uint8_t> 284 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK 285 * Maximum Value: MIN(255, backend/max-request-segments) 286 * 287 * The maximum value the frontend will set in the 288 * blkif_request.nr_segments field. 289 * 290 * max-request-size 291 * Values: <uint32_t> 292 * Default Value: BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK * PAGE_SIZE 293 * Maximum Value: max-request-segments * PAGE_SIZE 294 * 295 * The maximum amount of data, in bytes, that can be referenced by 296 * a request type that accesses frontend memory (currently BLKIF_OP_READ, 297 * BLKIF_OP_WRITE, or BLKIF_OP_WRITE_BARRIER). 298 * 299 *------------------------- Virtual Device Properties ------------------------- 300 * 301 * device-type 302 * Values: "disk", "cdrom", "floppy", etc. 303 * 304 * virtual-device 305 * Values: <uint32_t> 306 * 307 * A value indicating the physical device to virtualize within the 308 * frontend's domain. (e.g. "The first ATA disk", "The third SCSI 309 * disk", etc.) 310 * 311 * See docs/misc/vbd-interface.txt for details on the format of this 312 * value. 313 * 314 * Notes 315 * ----- 316 * (1) Multi-page ring buffer scheme first developed in the Citrix XenServer 317 * PV drivers. 318 * (2) Multi-page ring buffer scheme first used in some Red Hat distributions 319 * including a distribution deployed on certain nodes of the Amazon 320 * EC2 cluster. 321 * (3) Support for multi-page ring buffers was implemented independently, 322 * in slightly different forms, by both Citrix and Red Hat/Amazon. 323 * For full interoperability, block front and backends should publish 324 * identical ring parameters, adjusted for unit differences, to the 325 * XenStore nodes used in both schemes. 326 * (4) Devices that support discard functionality may internally allocate 327 * space (discardable extents) in units that are larger than the 328 * exported logical block size. 329 * (5) The discard-alignment parameter allows a physical device to be 330 * partitioned into virtual devices that do not necessarily begin or 331 * end on a discardable extent boundary. 332 * (6) When there is only a single page allocated to the request ring, 333 * 'ring-ref' is used to communicate the grant reference for this 334 * page to the backend. When using a multi-page ring, the 'ring-ref' 335 * node is not created. Instead 'ring-ref0' - 'ring-refN' are used. 336 */ 337 338/* 339 * STATE DIAGRAMS 340 * 341 ***************************************************************************** 342 * Startup * 343 ***************************************************************************** 344 * 345 * Tool stack creates front and back nodes with state XenbusStateInitialising. 346 * 347 * Front Back 348 * ================================= ===================================== 349 * XenbusStateInitialising XenbusStateInitialising 350 * o Query virtual device o Query backend device identification 351 * properties. data. 352 * o Setup OS device instance. o Open and validate backend device. 353 * o Publish backend features and 354 * transport parameters. 355 * | 356 * | 357 * V 358 * XenbusStateInitWait 359 * 360 * o Query backend features and 361 * transport parameters. 362 * o Allocate and initialize the 363 * request ring. 364 * o Publish transport parameters 365 * that will be in effect during 366 * this connection. 367 * | 368 * | 369 * V 370 * XenbusStateInitialised 371 * 372 * o Query frontend transport parameters. 373 * o Connect to the request ring and 374 * event channel. 375 * o Publish backend device properties. 376 * | 377 * | 378 * V 379 * XenbusStateConnected 380 * 381 * o Query backend device properties. 382 * o Finalize OS virtual device 383 * instance. 384 * | 385 * | 386 * V 387 * XenbusStateConnected 388 * 389 * Note: Drivers that do not support any optional features, or the negotiation 390 * of transport parameters, can skip certain states in the state machine: 391 * 392 * o A frontend may transition to XenbusStateInitialised without 393 * waiting for the backend to enter XenbusStateInitWait. In this 394 * case, default transport parameters are in effect and any 395 * transport parameters published by the frontend must contain 396 * their default values. 397 * 398 * o A backend may transition to XenbusStateInitialised, bypassing 399 * XenbusStateInitWait, without waiting for the frontend to first 400 * enter the XenbusStateInitialised state. In this case, default 401 * transport parameters are in effect and any transport parameters 402 * published by the backend must contain their default values. 403 * 404 * Drivers that support optional features and/or transport parameter 405 * negotiation must tolerate these additional state transition paths. 406 * In general this means performing the work of any skipped state 407 * transition, if it has not already been performed, in addition to the 408 * work associated with entry into the current state. 409 */ 410 411/* 412 * REQUEST CODES. 413 */ 414#define BLKIF_OP_READ 0 415#define BLKIF_OP_WRITE 1 416/* 417 * All writes issued prior to a request with the BLKIF_OP_WRITE_BARRIER 418 * operation code ("barrier request") must be completed prior to the 419 * execution of the barrier request. All writes issued after the barrier 420 * request must not execute until after the completion of the barrier request. 421 * 422 * Optional. See "feature-barrier" XenBus node documentation above. 423 */ 424#define BLKIF_OP_WRITE_BARRIER 2 425/* 426 * Commit any uncommitted contents of the backing device's volatile cache 427 * to stable storage. 428 * 429 * Optional. See "feature-flush-cache" XenBus node documentation above. 430 */ 431#define BLKIF_OP_FLUSH_DISKCACHE 3 432/* 433 * Used in SLES sources for device specific command packet 434 * contained within the request. Reserved for that purpose. 435 */ 436#define BLKIF_OP_RESERVED_1 4 437/* 438 * Indicate to the backend device that a region of storage is no longer in 439 * use, and may be discarded at any time without impact to the client. If 440 * the BLKIF_DISCARD_SECURE flag is set on the request, all copies of the 441 * discarded region on the device must be rendered unrecoverable before the 442 * command returns. 443 * 444 * This operation is analogous to performing a trim (ATA) or unmap (SCSI), 445 * command on a native device. 446 * 447 * More information about trim/unmap operations can be found at: 448 * http://t13.org/Documents/UploadedDocuments/docs2008/ 449 * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc 450 * http://www.seagate.com/staticfiles/support/disc/manuals/ 451 * Interface%20manuals/100293068c.pdf 452 * 453 * Optional. See "feature-discard", "discard-alignment", 454 * "discard-granularity", and "discard-secure" in the XenBus node 455 * documentation above. 456 */ 457#define BLKIF_OP_DISCARD 5 458 459/* 460 * Maximum scatter/gather segments associated with a request header block. 461 * This is carefully chosen so that sizeof(blkif_ring_t) <= PAGE_SIZE. 462 * NB. This could be 12 if the ring indexes weren't stored in the same page. 463 */ 464#define BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK 11 465 466/* 467 * Maximum scatter/gather segments associated with a segment block. 468 */ 469#define BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK 14 470 471/* 472 * Maximum scatter/gather segments per request (header + segment blocks). 473 */ 474#define BLKIF_MAX_SEGMENTS_PER_REQUEST 255 475 476/* 477 * NB. first_sect and last_sect in blkif_request_segment, as well as 478 * sector_number in blkif_request, are always expressed in 512-byte units. 479 * However they must be properly aligned to the real sector size of the 480 * physical disk, which is reported in the "sector-size" node in the backend 481 * xenbus info. Also the xenbus "sectors" node is expressed in 512-byte units. 482 */ 483struct blkif_request_segment { 484 grant_ref_t gref; /* reference to I/O buffer frame */ 485 /* @first_sect: first sector in frame to transfer (inclusive). */ 486 /* @last_sect: last sector in frame to transfer (inclusive). */ 487 uint8_t first_sect, last_sect; 488}; 489typedef struct blkif_request_segment blkif_request_segment_t; 490 491/* 492 * Starting ring element for any I/O request. 493 * 494 * One or more segment blocks can be inserted into the request ring 495 * just after a blkif_request_t, allowing requests to operate on 496 * up to BLKIF_MAX_SEGMENTS_PER_REQUEST. 497 * 498 * BLKIF_SEGS_TO_BLOCKS() can be used on blkif_requst.nr_segments 499 * to determine the number of contiguous ring entries associated 500 * with this request. 501 * 502 * Note: Due to the way Xen request rings operate, the producer and 503 * consumer indices of the ring must be incremented by the 504 * BLKIF_SEGS_TO_BLOCKS() value of the associated request. 505 * (e.g. a response to a 3 ring entry request must also consume 506 * 3 entries in the ring, even though only the first ring entry 507 * in the response has any data.) 508 */ 509struct blkif_request { 510 uint8_t operation; /* BLKIF_OP_??? */ 511 uint8_t nr_segments; /* number of segments */ 512 blkif_vdev_t handle; /* only for read/write requests */ 513 uint64_t id; /* private guest value, echoed in resp */ 514 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 515 blkif_request_segment_t seg[BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK]; 516}; 517typedef struct blkif_request blkif_request_t; 518 519/* 520 * A segment block is a ring request structure that contains only 521 * segment data. 522 * 523 * sizeof(struct blkif_segment_block) <= sizeof(struct blkif_request) 524 */ 525struct blkif_segment_block { 526 blkif_request_segment_t seg[BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK]; 527}; 528typedef struct blkif_segment_block blkif_segment_block_t; 529 530/* 531 * Cast to this structure when blkif_request.operation == BLKIF_OP_DISCARD 532 * sizeof(struct blkif_request_discard) <= sizeof(struct blkif_request) 533 */ 534struct blkif_request_discard { 535 uint8_t operation; /* BLKIF_OP_DISCARD */ 536 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */ 537#define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */ 538 blkif_vdev_t handle; /* same as for read/write requests */ 539 uint64_t id; /* private guest value, echoed in resp */ 540 blkif_sector_t sector_number;/* start sector idx on disk */ 541 uint64_t nr_sectors; /* number of contiguous sectors to discard*/ 542}; 543typedef struct blkif_request_discard blkif_request_discard_t; 544 545struct blkif_response { 546 uint64_t id; /* copied from request */ 547 uint8_t operation; /* copied from request */ 548 int16_t status; /* BLKIF_RSP_??? */ 549}; 550typedef struct blkif_response blkif_response_t; 551 552/* 553 * STATUS RETURN CODES. 554 */ 555 /* Operation not supported (only happens on barrier writes). */ 556#define BLKIF_RSP_EOPNOTSUPP -2 557 /* Operation failed for some unspecified reason (-EIO). */ 558#define BLKIF_RSP_ERROR -1 559 /* Operation completed successfully. */ 560#define BLKIF_RSP_OKAY 0 561 562/* 563 * Generate blkif ring structures and types. 564 */ 565DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response); 566 567/* 568 * Index to, and treat as a segment block, an entry in the ring. 569 */ 570#define BLKRING_GET_SEG_BLOCK(_r, _idx) \ 571 (((blkif_segment_block_t *)RING_GET_REQUEST(_r, _idx))->seg) 572 573/* 574 * The number of ring request blocks required to handle an I/O 575 * request containing _segs segments. 576 */ 577#define BLKIF_SEGS_TO_BLOCKS(_segs) \ 578 ((((_segs - BLKIF_MAX_SEGMENTS_PER_HEADER_BLOCK) \ 579 + (BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK - 1)) \ 580 / BLKIF_MAX_SEGMENTS_PER_SEGMENT_BLOCK) + /*header_block*/1) 581 582#define VDISK_CDROM 0x1 583#define VDISK_REMOVABLE 0x2 584#define VDISK_READONLY 0x4 585 586#endif /* __XEN_PUBLIC_IO_BLKIF_H__ */ 587 588/* 589 * Local variables: 590 * mode: C 591 * c-set-style: "BSD" 592 * c-basic-offset: 4 593 * tab-width: 4 594 * indent-tabs-mode: nil 595 * End: 596 */ 597