blkif.h revision 287801
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 "ring.h" 32#include "../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 *------------------------- Backend Device Properties ------------------------- 149 * 150 * discard-alignment 151 * Values: <uint32_t> 152 * Default Value: 0 153 * Notes: 4, 5 154 * 155 * The offset, in bytes from the beginning of the virtual block device, 156 * to the first, addressable, discard extent on the underlying device. 157 * 158 * discard-granularity 159 * Values: <uint32_t> 160 * Default Value: <"sector-size"> 161 * Notes: 4 162 * 163 * The size, in bytes, of the individually addressable discard extents 164 * of the underlying device. 165 * 166 * discard-secure 167 * Values: 0/1 (boolean) 168 * Default Value: 0 169 * 170 * A value of "1" indicates that the backend can process BLKIF_OP_DISCARD 171 * requests with the BLKIF_DISCARD_SECURE flag set. 172 * 173 * info 174 * Values: <uint32_t> (bitmap) 175 * 176 * A collection of bit flags describing attributes of the backing 177 * device. The VDISK_* macros define the meaning of each bit 178 * location. 179 * 180 * sector-size 181 * Values: <uint32_t> 182 * 183 * The size, in bytes, of the individually addressible data blocks 184 * on the backend device. 185 * 186 * sectors 187 * Values: <uint64_t> 188 * 189 * The size of the backend device, expressed in units of its native 190 * sector size ("sector-size"). 191 * 192 ***************************************************************************** 193 * Frontend XenBus Nodes 194 ***************************************************************************** 195 * 196 *----------------------- Request Transport Parameters ----------------------- 197 * 198 * event-channel 199 * Values: <uint32_t> 200 * 201 * The identifier of the Xen event channel used to signal activity 202 * in the ring buffer. 203 * 204 * ring-ref 205 * Values: <uint32_t> 206 * Notes: 6 207 * 208 * The Xen grant reference granting permission for the backend to map 209 * the sole page in a single page sized ring buffer. 210 * 211 * ring-ref%u 212 * Values: <uint32_t> 213 * Notes: 6 214 * 215 * For a frontend providing a multi-page ring, a "number of ring pages" 216 * sized list of nodes, each containing a Xen grant reference granting 217 * permission for the backend to map the page of the ring located 218 * at page index "%u". Page indexes are zero based. 219 * 220 * protocol 221 * Values: string (XEN_IO_PROTO_ABI_*) 222 * Default Value: XEN_IO_PROTO_ABI_NATIVE 223 * 224 * The machine ABI rules governing the format of all ring request and 225 * response structures. 226 * 227 * ring-page-order 228 * Values: <uint32_t> 229 * Default Value: 0 230 * Maximum Value: MAX(ffs(max-ring-pages) - 1, max-ring-page-order) 231 * Notes: 1, 3 232 * 233 * The size of the frontend allocated request ring buffer in units 234 * of lb(machine pages). (e.g. 0 == 1 page, 1 = 2 pages, 2 == 4 pages, 235 * etc.). 236 * 237 * num-ring-pages 238 * Values: <uint32_t> 239 * Default Value: 1 240 * Maximum Value: MAX(max-ring-pages,(0x1 << max-ring-page-order)) 241 * Notes: DEPRECATED, 2, 3 242 * 243 * The size of the frontend allocated request ring buffer in units of 244 * machine pages. The value must be a power of 2. 245 * 246 *------------------------- Virtual Device Properties ------------------------- 247 * 248 * device-type 249 * Values: "disk", "cdrom", "floppy", etc. 250 * 251 * virtual-device 252 * Values: <uint32_t> 253 * 254 * A value indicating the physical device to virtualize within the 255 * frontend's domain. (e.g. "The first ATA disk", "The third SCSI 256 * disk", etc.) 257 * 258 * See docs/misc/vbd-interface.txt for details on the format of this 259 * value. 260 * 261 * Notes 262 * ----- 263 * (1) Multi-page ring buffer scheme first developed in the Citrix XenServer 264 * PV drivers. 265 * (2) Multi-page ring buffer scheme first used in some Red Hat distributions 266 * including a distribution deployed on certain nodes of the Amazon 267 * EC2 cluster. 268 * (3) Support for multi-page ring buffers was implemented independently, 269 * in slightly different forms, by both Citrix and Red Hat/Amazon. 270 * For full interoperability, block front and backends should publish 271 * identical ring parameters, adjusted for unit differences, to the 272 * XenStore nodes used in both schemes. 273 * (4) Devices that support discard functionality may internally allocate 274 * space (discardable extents) in units that are larger than the 275 * exported logical block size. 276 * (5) The discard-alignment parameter allows a physical device to be 277 * partitioned into virtual devices that do not necessarily begin or 278 * end on a discardable extent boundary. 279 * (6) When there is only a single page allocated to the request ring, 280 * 'ring-ref' is used to communicate the grant reference for this 281 * page to the backend. When using a multi-page ring, the 'ring-ref' 282 * node is not created. Instead 'ring-ref0' - 'ring-refN' are used. 283 */ 284 285/* 286 * STATE DIAGRAMS 287 * 288 ***************************************************************************** 289 * Startup * 290 ***************************************************************************** 291 * 292 * Tool stack creates front and back nodes with state XenbusStateInitialising. 293 * 294 * Front Back 295 * ================================= ===================================== 296 * XenbusStateInitialising XenbusStateInitialising 297 * o Query virtual device o Query backend device identification 298 * properties. data. 299 * o Setup OS device instance. o Open and validate backend device. 300 * o Publish backend features and 301 * transport parameters. 302 * | 303 * | 304 * V 305 * XenbusStateInitWait 306 * 307 * o Query backend features and 308 * transport parameters. 309 * o Allocate and initialize the 310 * request ring. 311 * o Publish transport parameters 312 * that will be in effect during 313 * this connection. 314 * | 315 * | 316 * V 317 * XenbusStateInitialised 318 * 319 * o Query frontend transport parameters. 320 * o Connect to the request ring and 321 * event channel. 322 * o Publish backend device properties. 323 * | 324 * | 325 * V 326 * XenbusStateConnected 327 * 328 * o Query backend device properties. 329 * o Finalize OS virtual device 330 * instance. 331 * | 332 * | 333 * V 334 * XenbusStateConnected 335 * 336 * Note: Drivers that do not support any optional features, or the negotiation 337 * of transport parameters, can skip certain states in the state machine: 338 * 339 * o A frontend may transition to XenbusStateInitialised without 340 * waiting for the backend to enter XenbusStateInitWait. In this 341 * case, default transport parameters are in effect and any 342 * transport parameters published by the frontend must contain 343 * their default values. 344 * 345 * o A backend may transition to XenbusStateInitialised, bypassing 346 * XenbusStateInitWait, without waiting for the frontend to first 347 * enter the XenbusStateInitialised state. In this case, default 348 * transport parameters are in effect and any transport parameters 349 * published by the backend must contain their default values. 350 * 351 * Drivers that support optional features and/or transport parameter 352 * negotiation must tolerate these additional state transition paths. 353 * In general this means performing the work of any skipped state 354 * transition, if it has not already been performed, in addition to the 355 * work associated with entry into the current state. 356 */ 357 358/* 359 * REQUEST CODES. 360 */ 361#define BLKIF_OP_READ 0 362#define BLKIF_OP_WRITE 1 363/* 364 * All writes issued prior to a request with the BLKIF_OP_WRITE_BARRIER 365 * operation code ("barrier request") must be completed prior to the 366 * execution of the barrier request. All writes issued after the barrier 367 * request must not execute until after the completion of the barrier request. 368 * 369 * Optional. See "feature-barrier" XenBus node documentation above. 370 */ 371#define BLKIF_OP_WRITE_BARRIER 2 372/* 373 * Commit any uncommitted contents of the backing device's volatile cache 374 * to stable storage. 375 * 376 * Optional. See "feature-flush-cache" XenBus node documentation above. 377 */ 378#define BLKIF_OP_FLUSH_DISKCACHE 3 379/* 380 * Used in SLES sources for device specific command packet 381 * contained within the request. Reserved for that purpose. 382 */ 383#define BLKIF_OP_RESERVED_1 4 384/* 385 * Indicate to the backend device that a region of storage is no longer in 386 * use, and may be discarded at any time without impact to the client. If 387 * the BLKIF_DISCARD_SECURE flag is set on the request, all copies of the 388 * discarded region on the device must be rendered unrecoverable before the 389 * command returns. 390 * 391 * This operation is analogous to performing a trim (ATA) or unmap (SCSI), 392 * command on a native device. 393 * 394 * More information about trim/unmap operations can be found at: 395 * http://t13.org/Documents/UploadedDocuments/docs2008/ 396 * e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc 397 * http://www.seagate.com/staticfiles/support/disc/manuals/ 398 * Interface%20manuals/100293068c.pdf 399 * 400 * Optional. See "feature-discard", "discard-alignment", 401 * "discard-granularity", and "discard-secure" in the XenBus node 402 * documentation above. 403 */ 404#define BLKIF_OP_DISCARD 5 405 406/* 407 * Maximum scatter/gather segments per request. 408 * This is carefully chosen so that sizeof(blkif_ring_t) <= PAGE_SIZE. 409 * NB. This could be 12 if the ring indexes weren't stored in the same page. 410 */ 411#define BLKIF_MAX_SEGMENTS_PER_REQUEST 11 412 413/* 414 * NB. first_sect and last_sect in blkif_request_segment, as well as 415 * sector_number in blkif_request, are always expressed in 512-byte units. 416 * However they must be properly aligned to the real sector size of the 417 * physical disk, which is reported in the "sector-size" node in the backend 418 * xenbus info. Also the xenbus "sectors" node is expressed in 512-byte units. 419 */ 420struct blkif_request_segment { 421 grant_ref_t gref; /* reference to I/O buffer frame */ 422 /* @first_sect: first sector in frame to transfer (inclusive). */ 423 /* @last_sect: last sector in frame to transfer (inclusive). */ 424 uint8_t first_sect, last_sect; 425}; 426typedef struct blkif_request_segment blkif_request_segment_t; 427 428/* 429 * Starting ring element for any I/O request. 430 */ 431struct blkif_request { 432 uint8_t operation; /* BLKIF_OP_??? */ 433 uint8_t nr_segments; /* number of segments */ 434 blkif_vdev_t handle; /* only for read/write requests */ 435 uint64_t id; /* private guest value, echoed in resp */ 436 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 437 blkif_request_segment_t seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 438}; 439typedef struct blkif_request blkif_request_t; 440 441/* 442 * Cast to this structure when blkif_request.operation == BLKIF_OP_DISCARD 443 * sizeof(struct blkif_request_discard) <= sizeof(struct blkif_request) 444 */ 445struct blkif_request_discard { 446 uint8_t operation; /* BLKIF_OP_DISCARD */ 447 uint8_t flag; /* BLKIF_DISCARD_SECURE or zero */ 448#define BLKIF_DISCARD_SECURE (1<<0) /* ignored if discard-secure=0 */ 449 blkif_vdev_t handle; /* same as for read/write requests */ 450 uint64_t id; /* private guest value, echoed in resp */ 451 blkif_sector_t sector_number;/* start sector idx on disk */ 452 uint64_t nr_sectors; /* number of contiguous sectors to discard*/ 453}; 454typedef struct blkif_request_discard blkif_request_discard_t; 455 456struct blkif_response { 457 uint64_t id; /* copied from request */ 458 uint8_t operation; /* copied from request */ 459 int16_t status; /* BLKIF_RSP_??? */ 460}; 461typedef struct blkif_response blkif_response_t; 462 463/* 464 * STATUS RETURN CODES. 465 */ 466 /* Operation not supported (only happens on barrier writes). */ 467#define BLKIF_RSP_EOPNOTSUPP -2 468 /* Operation failed for some unspecified reason (-EIO). */ 469#define BLKIF_RSP_ERROR -1 470 /* Operation completed successfully. */ 471#define BLKIF_RSP_OKAY 0 472 473/* 474 * Generate blkif ring structures and types. 475 */ 476DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response); 477 478#define VDISK_CDROM 0x1 479#define VDISK_REMOVABLE 0x2 480#define VDISK_READONLY 0x4 481 482#endif /* __XEN_PUBLIC_IO_BLKIF_H__ */ 483 484/* 485 * Local variables: 486 * mode: C 487 * c-set-style: "BSD" 488 * c-basic-offset: 4 489 * tab-width: 4 490 * indent-tabs-mode: nil 491 * End: 492 */ 493