1 /*- 2 * BSD LICENSE 3 * 4 * Copyright (c) Intel Corporation. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * * Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * * Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * * Neither the name of Intel Corporation nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #ifndef SPDK_BLOBSTORE_H 35 #define SPDK_BLOBSTORE_H 36 37 #include "spdk/assert.h" 38 #include "spdk/blob.h" 39 #include "spdk/queue.h" 40 #include "spdk/util.h" 41 42 #include "request.h" 43 44 /* In Memory Data Structures 45 * 46 * The following data structures exist only in memory. 47 */ 48 49 #define SPDK_BLOB_OPTS_CLUSTER_SZ (1024 * 1024) 50 #define SPDK_BLOB_OPTS_NUM_MD_PAGES UINT32_MAX 51 #define SPDK_BLOB_OPTS_MAX_MD_OPS 32 52 #define SPDK_BLOB_OPTS_DEFAULT_CHANNEL_OPS 512 53 #define SPDK_BLOB_BLOBID_HIGH_BIT (1ULL << 32) 54 55 struct spdk_xattr { 56 /* TODO: reorder for best packing */ 57 uint32_t index; 58 char *name; 59 void *value; 60 uint16_t value_len; 61 TAILQ_ENTRY(spdk_xattr) link; 62 }; 63 64 /* The mutable part of the blob data that is sync'd to 65 * disk. The data in here is both mutable and persistent. 66 */ 67 struct spdk_blob_mut_data { 68 /* Number of data clusters in the blob */ 69 uint64_t num_clusters; 70 71 /* Array LBAs that are the beginning of a cluster, in 72 * the order they appear in the blob. 73 */ 74 uint64_t *clusters; 75 76 /* The size of the clusters array. This is greater than or 77 * equal to 'num_clusters'. 78 */ 79 size_t cluster_array_size; 80 81 /* Number of metadata pages */ 82 uint32_t num_pages; 83 84 /* Array of page offsets into the metadata region, in 85 * the order of the metadata page sequence. 86 */ 87 uint32_t *pages; 88 }; 89 90 enum spdk_blob_state { 91 /* The blob in-memory version does not match the on-disk 92 * version. 93 */ 94 SPDK_BLOB_STATE_DIRTY, 95 96 /* The blob in memory version of the blob matches the on disk 97 * version. 98 */ 99 SPDK_BLOB_STATE_CLEAN, 100 101 /* The in-memory state being synchronized with the on-disk 102 * blob state. */ 103 SPDK_BLOB_STATE_LOADING, 104 }; 105 106 TAILQ_HEAD(spdk_xattr_tailq, spdk_xattr); 107 108 struct spdk_blob { 109 struct spdk_blob_store *bs; 110 111 uint32_t open_ref; 112 113 spdk_blob_id id; 114 115 enum spdk_blob_state state; 116 117 /* Two copies of the mutable data. One is a version 118 * that matches the last known data on disk (clean). 119 * The other (active) is the current data. Syncing 120 * a blob makes the clean match the active. 121 */ 122 struct spdk_blob_mut_data clean; 123 struct spdk_blob_mut_data active; 124 125 bool invalid; 126 bool data_ro; 127 bool md_ro; 128 129 uint64_t invalid_flags; 130 uint64_t data_ro_flags; 131 uint64_t md_ro_flags; 132 133 struct spdk_bs_dev *back_bs_dev; 134 135 /* TODO: The xattrs are mutable, but we don't want to be 136 * copying them unecessarily. Figure this out. 137 */ 138 struct spdk_xattr_tailq xattrs; 139 struct spdk_xattr_tailq xattrs_internal; 140 141 TAILQ_ENTRY(spdk_blob) link; 142 }; 143 144 struct spdk_blob_store { 145 uint64_t md_start; /* Offset from beginning of disk, in pages */ 146 uint32_t md_len; /* Count, in pages */ 147 148 struct spdk_io_channel *md_channel; 149 uint32_t max_channel_ops; 150 151 struct spdk_thread *md_thread; 152 153 struct spdk_bs_dev *dev; 154 155 struct spdk_bit_array *used_md_pages; 156 struct spdk_bit_array *used_clusters; 157 struct spdk_bit_array *used_blobids; 158 159 pthread_mutex_t used_clusters_mutex; 160 161 uint32_t cluster_sz; 162 uint64_t total_clusters; 163 uint64_t total_data_clusters; 164 uint64_t num_free_clusters; 165 uint32_t pages_per_cluster; 166 167 spdk_blob_id super_blob; 168 struct spdk_bs_type bstype; 169 170 struct spdk_bs_cpl unload_cpl; 171 int unload_err; 172 173 TAILQ_HEAD(, spdk_blob) blobs; 174 }; 175 176 struct spdk_bs_channel { 177 struct spdk_bs_request_set *req_mem; 178 TAILQ_HEAD(, spdk_bs_request_set) reqs; 179 180 struct spdk_blob_store *bs; 181 182 struct spdk_bs_dev *dev; 183 struct spdk_io_channel *dev_channel; 184 185 TAILQ_HEAD(, spdk_bs_request_set) need_cluster_alloc; 186 }; 187 188 /** operation type */ 189 enum spdk_blob_op_type { 190 SPDK_BLOB_WRITE, 191 SPDK_BLOB_READ, 192 SPDK_BLOB_UNMAP, 193 SPDK_BLOB_WRITE_ZEROES, 194 SPDK_BLOB_WRITEV, 195 SPDK_BLOB_READV, 196 }; 197 198 /* back bs_dev */ 199 200 #define BLOB_SNAPSHOT "SNAP" 201 202 struct spdk_blob_bs_dev { 203 struct spdk_bs_dev bs_dev; 204 struct spdk_blob *blob; 205 }; 206 207 /* On-Disk Data Structures 208 * 209 * The following data structures exist on disk. 210 */ 211 #define SPDK_BS_INITIAL_VERSION 1 212 #define SPDK_BS_VERSION 3 /* current version */ 213 214 #pragma pack(push, 1) 215 216 #define SPDK_MD_MASK_TYPE_USED_PAGES 0 217 #define SPDK_MD_MASK_TYPE_USED_CLUSTERS 1 218 #define SPDK_MD_MASK_TYPE_USED_BLOBIDS 2 219 220 struct spdk_bs_md_mask { 221 uint8_t type; 222 uint32_t length; /* In bits */ 223 uint8_t mask[0]; 224 }; 225 226 #define SPDK_MD_DESCRIPTOR_TYPE_PADDING 0 227 #define SPDK_MD_DESCRIPTOR_TYPE_EXTENT 1 228 #define SPDK_MD_DESCRIPTOR_TYPE_XATTR 2 229 #define SPDK_MD_DESCRIPTOR_TYPE_FLAGS 3 230 #define SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL 4 231 232 struct spdk_blob_md_descriptor_xattr { 233 uint8_t type; 234 uint32_t length; 235 236 uint16_t name_length; 237 uint16_t value_length; 238 239 char name[0]; 240 /* String name immediately followed by string value. */ 241 }; 242 243 struct spdk_blob_md_descriptor_extent { 244 uint8_t type; 245 uint32_t length; 246 247 struct { 248 uint32_t cluster_idx; 249 uint32_t length; /* In units of clusters */ 250 } extents[0]; 251 }; 252 253 #define SPDK_BLOB_THIN_PROV (1ULL << 0) 254 #define SPDK_BLOB_INTERNAL_XATTR (1ULL << 1) 255 #define SPDK_BLOB_INVALID_FLAGS_MASK (SPDK_BLOB_THIN_PROV | SPDK_BLOB_INTERNAL_XATTR) 256 257 #define SPDK_BLOB_READ_ONLY (1ULL << 0) 258 #define SPDK_BLOB_DATA_RO_FLAGS_MASK SPDK_BLOB_READ_ONLY 259 #define SPDK_BLOB_MD_RO_FLAGS_MASK 0 260 261 #define spdk_blob_is_thin_provisioned(blob) (blob->invalid_flags & SPDK_BLOB_THIN_PROV) 262 263 struct spdk_blob_md_descriptor_flags { 264 uint8_t type; 265 uint32_t length; 266 267 /* 268 * If a flag in invalid_flags is set that the application is not aware of, 269 * it will not allow the blob to be opened. 270 */ 271 uint64_t invalid_flags; 272 273 /* 274 * If a flag in data_ro_flags is set that the application is not aware of, 275 * allow the blob to be opened in data_read_only and md_read_only mode. 276 */ 277 uint64_t data_ro_flags; 278 279 /* 280 * If a flag in md_ro_flags is set the the application is not aware of, 281 * allow the blob to be opened in md_read_only mode. 282 */ 283 uint64_t md_ro_flags; 284 }; 285 286 struct spdk_blob_md_descriptor { 287 uint8_t type; 288 uint32_t length; 289 }; 290 291 #define SPDK_INVALID_MD_PAGE UINT32_MAX 292 293 struct spdk_blob_md_page { 294 spdk_blob_id id; 295 296 uint32_t sequence_num; 297 uint32_t reserved0; 298 299 /* Descriptors here */ 300 uint8_t descriptors[4072]; 301 302 uint32_t next; 303 uint32_t crc; 304 }; 305 #define SPDK_BS_PAGE_SIZE 0x1000 306 SPDK_STATIC_ASSERT(SPDK_BS_PAGE_SIZE == sizeof(struct spdk_blob_md_page), "Invalid md page size"); 307 308 #define SPDK_BS_SUPER_BLOCK_SIG "SPDKBLOB" 309 310 struct spdk_bs_super_block { 311 uint8_t signature[8]; 312 uint32_t version; 313 uint32_t length; 314 uint32_t clean; /* If there was a clean shutdown, this is 1. */ 315 spdk_blob_id super_blob; 316 317 uint32_t cluster_size; /* In bytes */ 318 319 uint32_t used_page_mask_start; /* Offset from beginning of disk, in pages */ 320 uint32_t used_page_mask_len; /* Count, in pages */ 321 322 uint32_t used_cluster_mask_start; /* Offset from beginning of disk, in pages */ 323 uint32_t used_cluster_mask_len; /* Count, in pages */ 324 325 uint32_t md_start; /* Offset from beginning of disk, in pages */ 326 uint32_t md_len; /* Count, in pages */ 327 328 struct spdk_bs_type bstype; /* blobstore type */ 329 330 uint32_t used_blobid_mask_start; /* Offset from beginning of disk, in pages */ 331 uint32_t used_blobid_mask_len; /* Count, in pages */ 332 333 uint8_t reserved[4012]; 334 uint32_t crc; 335 }; 336 SPDK_STATIC_ASSERT(sizeof(struct spdk_bs_super_block) == 0x1000, "Invalid super block size"); 337 338 #pragma pack(pop) 339 340 struct spdk_bs_dev *spdk_bs_create_zeroes_dev(void); 341 struct spdk_bs_dev *spdk_bs_create_blob_bs_dev(struct spdk_blob *blob); 342 343 /* Unit Conversions 344 * 345 * The blobstore works with several different units: 346 * - Byte: Self explanatory 347 * - LBA: The logical blocks on the backing storage device. 348 * - Page: The read/write units of blobs and metadata. This is 349 * an offset into a blob in units of 4KiB. 350 * - Cluster Index: The disk is broken into a sequential list of 351 * clusters. This is the offset from the beginning. 352 * 353 * NOTE: These conversions all act on simple magnitudes, not with any sort 354 * of knowledge about the blobs themselves. For instance, converting 355 * a page to an lba with the conversion function below simply converts 356 * a number of pages to an equivalent number of lbas, but that 357 * lba certainly isn't the right lba that corresponds to a page offset 358 * for a particular blob. 359 */ 360 static inline uint64_t 361 _spdk_bs_byte_to_lba(struct spdk_blob_store *bs, uint64_t length) 362 { 363 assert(length % bs->dev->blocklen == 0); 364 365 return length / bs->dev->blocklen; 366 } 367 368 static inline uint64_t 369 _spdk_bs_dev_byte_to_lba(struct spdk_bs_dev *bs_dev, uint64_t length) 370 { 371 assert(length % bs_dev->blocklen == 0); 372 373 return length / bs_dev->blocklen; 374 } 375 376 static inline uint64_t 377 _spdk_bs_lba_to_byte(struct spdk_blob_store *bs, uint64_t lba) 378 { 379 return lba * bs->dev->blocklen; 380 } 381 382 static inline uint64_t 383 _spdk_bs_page_to_lba(struct spdk_blob_store *bs, uint64_t page) 384 { 385 return page * SPDK_BS_PAGE_SIZE / bs->dev->blocklen; 386 } 387 388 static inline uint64_t 389 _spdk_bs_dev_page_to_lba(struct spdk_bs_dev *bs_dev, uint64_t page) 390 { 391 return page * SPDK_BS_PAGE_SIZE / bs_dev->blocklen; 392 } 393 394 static inline uint32_t 395 _spdk_bs_lba_to_page(struct spdk_blob_store *bs, uint64_t lba) 396 { 397 uint64_t lbas_per_page; 398 399 lbas_per_page = SPDK_BS_PAGE_SIZE / bs->dev->blocklen; 400 401 assert(lba % lbas_per_page == 0); 402 403 return lba / lbas_per_page; 404 } 405 406 static inline uint64_t 407 _spdk_bs_dev_lba_to_page(struct spdk_bs_dev *bs_dev, uint64_t lba) 408 { 409 uint64_t lbas_per_page; 410 411 lbas_per_page = SPDK_BS_PAGE_SIZE / bs_dev->blocklen; 412 413 assert(lba % lbas_per_page == 0); 414 415 return lba / lbas_per_page; 416 } 417 418 static inline uint64_t 419 _spdk_bs_cluster_to_page(struct spdk_blob_store *bs, uint32_t cluster) 420 { 421 return cluster * bs->pages_per_cluster; 422 } 423 424 static inline uint32_t 425 _spdk_bs_page_to_cluster(struct spdk_blob_store *bs, uint64_t page) 426 { 427 assert(page % bs->pages_per_cluster == 0); 428 429 return page / bs->pages_per_cluster; 430 } 431 432 static inline uint64_t 433 _spdk_bs_cluster_to_lba(struct spdk_blob_store *bs, uint32_t cluster) 434 { 435 return cluster * (bs->cluster_sz / bs->dev->blocklen); 436 } 437 438 static inline uint32_t 439 _spdk_bs_lba_to_cluster(struct spdk_blob_store *bs, uint64_t lba) 440 { 441 assert(lba % (bs->cluster_sz / bs->dev->blocklen) == 0); 442 443 return lba / (bs->cluster_sz / bs->dev->blocklen); 444 } 445 446 static inline uint64_t 447 _spdk_bs_blob_lba_to_back_dev_lba(struct spdk_blob *blob, uint64_t lba) 448 { 449 return lba * blob->bs->dev->blocklen / blob->back_bs_dev->blocklen; 450 } 451 452 static inline uint64_t 453 _spdk_bs_blob_lba_from_back_dev_lba(struct spdk_blob *blob, uint64_t lba) 454 { 455 return lba * blob->back_bs_dev->blocklen / blob->bs->dev->blocklen; 456 } 457 458 /* End basic conversions */ 459 460 static inline uint32_t 461 _spdk_bs_blobid_to_page(spdk_blob_id id) 462 { 463 return id & 0xFFFFFFFF; 464 } 465 466 /* The blob id is a 64 bit number. The lower 32 bits are the page_idx. The upper 467 * 32 bits are not currently used. Stick a 1 there just to catch bugs where the 468 * code assumes blob id == page_idx. 469 */ 470 static inline spdk_blob_id 471 _spdk_bs_page_to_blobid(uint32_t page_idx) 472 { 473 return SPDK_BLOB_BLOBID_HIGH_BIT | page_idx; 474 } 475 476 /* Given a page offset into a blob, look up the LBA for the 477 * start of that page. 478 */ 479 static inline uint64_t 480 _spdk_bs_blob_page_to_lba(struct spdk_blob *blob, uint32_t page) 481 { 482 uint64_t lba; 483 uint32_t pages_per_cluster; 484 485 pages_per_cluster = blob->bs->pages_per_cluster; 486 487 assert(page < blob->active.num_clusters * pages_per_cluster); 488 489 lba = blob->active.clusters[page / pages_per_cluster]; 490 lba += _spdk_bs_page_to_lba(blob->bs, page % pages_per_cluster); 491 492 return lba; 493 } 494 495 /* Given a page offset into a blob, look up the number of pages until the 496 * next cluster boundary. 497 */ 498 static inline uint32_t 499 _spdk_bs_num_pages_to_cluster_boundary(struct spdk_blob *blob, uint32_t page) 500 { 501 uint32_t pages_per_cluster; 502 503 pages_per_cluster = blob->bs->pages_per_cluster; 504 505 return pages_per_cluster - (page % pages_per_cluster); 506 } 507 508 /* Given a page offset into a blob, look up the number of pages into blob to beginning of current cluster */ 509 static inline uint32_t 510 _spdk_bs_page_to_cluster_start(struct spdk_blob *blob, uint32_t page) 511 { 512 uint32_t pages_per_cluster; 513 514 pages_per_cluster = blob->bs->pages_per_cluster; 515 516 return page - (page % pages_per_cluster); 517 } 518 519 /* Given a page offset into a blob, look up if it is from allocated cluster. */ 520 static inline bool 521 _spdk_bs_page_is_allocated(struct spdk_blob *blob, uint32_t page) 522 { 523 uint64_t lba; 524 uint32_t pages_per_cluster; 525 526 pages_per_cluster = blob->bs->pages_per_cluster; 527 528 assert(page < blob->active.num_clusters * pages_per_cluster); 529 530 lba = blob->active.clusters[page / pages_per_cluster]; 531 532 if (lba == 0) { 533 assert(spdk_blob_is_thin_provisioned(blob)); 534 return false; 535 } else { 536 return true; 537 } 538 } 539 540 #endif 541