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_MAX_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 /* The disk state is being synchronized with the current 106 * blob state. 107 */ 108 SPDK_BLOB_STATE_SYNCING, 109 }; 110 111 struct spdk_blob_data { 112 struct spdk_blob_store *bs; 113 114 uint32_t open_ref; 115 116 spdk_blob_id id; 117 118 enum spdk_blob_state state; 119 120 /* Two copies of the mutable data. One is a version 121 * that matches the last known data on disk (clean). 122 * The other (active) is the current data. Syncing 123 * a blob makes the clean match the active. 124 */ 125 struct spdk_blob_mut_data clean; 126 struct spdk_blob_mut_data active; 127 128 bool invalid; 129 bool data_ro; 130 bool md_ro; 131 132 uint64_t invalid_flags; 133 uint64_t data_ro_flags; 134 uint64_t md_ro_flags; 135 136 /* TODO: The xattrs are mutable, but we don't want to be 137 * copying them unecessarily. Figure this out. 138 */ 139 TAILQ_HEAD(, spdk_xattr) xattrs; 140 141 TAILQ_ENTRY(spdk_blob_data) link; 142 }; 143 144 #define __blob_to_data(x) ((struct spdk_blob_data *)(x)) 145 #define __data_to_blob(x) ((struct spdk_blob *)(x)) 146 147 struct spdk_blob_store { 148 uint64_t md_start; /* Offset from beginning of disk, in pages */ 149 uint32_t md_len; /* Count, in pages */ 150 151 struct spdk_io_channel *md_channel; 152 uint32_t max_channel_ops; 153 154 struct spdk_bs_dev *dev; 155 156 struct spdk_bit_array *used_md_pages; 157 struct spdk_bit_array *used_clusters; 158 159 uint32_t cluster_sz; 160 uint64_t total_clusters; 161 uint64_t total_data_clusters; 162 uint64_t num_free_clusters; 163 uint32_t pages_per_cluster; 164 165 spdk_blob_id super_blob; 166 struct spdk_bs_type bstype; 167 168 struct spdk_bs_cpl unload_cpl; 169 int unload_err; 170 171 TAILQ_HEAD(, spdk_blob_data) blobs; 172 }; 173 174 struct spdk_bs_channel { 175 struct spdk_bs_request_set *req_mem; 176 TAILQ_HEAD(, spdk_bs_request_set) reqs; 177 178 struct spdk_blob_store *bs; 179 180 struct spdk_bs_dev *dev; 181 struct spdk_io_channel *dev_channel; 182 }; 183 184 /** operation type */ 185 enum spdk_blob_op_type { 186 SPDK_BLOB_WRITE, 187 SPDK_BLOB_READ, 188 SPDK_BLOB_UNMAP, 189 SPDK_BLOB_WRITE_ZEROES, 190 }; 191 192 /* On-Disk Data Structures 193 * 194 * The following data structures exist on disk. 195 */ 196 #define SPDK_BS_INITIAL_VERSION 1 197 #define SPDK_BS_VERSION 3 /* current version */ 198 199 #pragma pack(push, 1) 200 201 #define SPDK_MD_MASK_TYPE_USED_PAGES 0 202 #define SPDK_MD_MASK_TYPE_USED_CLUSTERS 1 203 204 struct spdk_bs_md_mask { 205 uint8_t type; 206 uint32_t length; /* In bits */ 207 uint8_t mask[0]; 208 }; 209 210 #define SPDK_MD_DESCRIPTOR_TYPE_PADDING 0 211 #define SPDK_MD_DESCRIPTOR_TYPE_EXTENT 1 212 #define SPDK_MD_DESCRIPTOR_TYPE_XATTR 2 213 #define SPDK_MD_DESCRIPTOR_TYPE_FLAGS 3 214 215 struct spdk_blob_md_descriptor_xattr { 216 uint8_t type; 217 uint32_t length; 218 219 uint16_t name_length; 220 uint16_t value_length; 221 222 char name[0]; 223 /* String name immediately followed by string value. */ 224 }; 225 226 struct spdk_blob_md_descriptor_extent { 227 uint8_t type; 228 uint32_t length; 229 230 struct { 231 uint32_t cluster_idx; 232 uint32_t length; /* In units of clusters */ 233 } extents[0]; 234 }; 235 236 /* 237 * As new flags are defined, these values will be updated to reflect the 238 * mask of all flag values understood by this application. 239 */ 240 #define SPDK_BLOB_INVALID_FLAGS_MASK 0 241 #define SPDK_BLOB_DATA_RO_FLAGS_MASK 0 242 #define SPDK_BLOB_MD_RO_FLAGS_MASK 0 243 244 struct spdk_blob_md_descriptor_flags { 245 uint8_t type; 246 uint32_t length; 247 248 /* 249 * If a flag in invalid_flags is set that the application is not aware of, 250 * it will not allow the blob to be opened. 251 */ 252 uint64_t invalid_flags; 253 254 /* 255 * If a flag in data_ro_flags is set that the application is not aware of, 256 * allow the blob to be opened in data_read_only and md_read_only mode. 257 */ 258 uint64_t data_ro_flags; 259 260 /* 261 * If a flag in md_ro_flags is set the the application is not aware of, 262 * allow the blob to be opened in md_read_only mode. 263 */ 264 uint64_t md_ro_flags; 265 }; 266 267 struct spdk_blob_md_descriptor { 268 uint8_t type; 269 uint32_t length; 270 }; 271 272 #define SPDK_INVALID_MD_PAGE UINT32_MAX 273 274 struct spdk_blob_md_page { 275 spdk_blob_id id; 276 277 uint32_t sequence_num; 278 uint32_t reserved0; 279 280 /* Descriptors here */ 281 uint8_t descriptors[4072]; 282 283 uint32_t next; 284 uint32_t crc; 285 }; 286 #define SPDK_BS_PAGE_SIZE 0x1000 287 SPDK_STATIC_ASSERT(SPDK_BS_PAGE_SIZE == sizeof(struct spdk_blob_md_page), "Invalid md page size"); 288 289 #define SPDK_BS_SUPER_BLOCK_SIG "SPDKBLOB" 290 291 struct spdk_bs_super_block { 292 uint8_t signature[8]; 293 uint32_t version; 294 uint32_t length; 295 uint32_t clean; /* If there was a clean shutdown, this is 1. */ 296 spdk_blob_id super_blob; 297 298 uint32_t cluster_size; /* In bytes */ 299 300 uint32_t used_page_mask_start; /* Offset from beginning of disk, in pages */ 301 uint32_t used_page_mask_len; /* Count, in pages */ 302 303 uint32_t used_cluster_mask_start; /* Offset from beginning of disk, in pages */ 304 uint32_t used_cluster_mask_len; /* Count, in pages */ 305 306 uint32_t md_start; /* Offset from beginning of disk, in pages */ 307 uint32_t md_len; /* Count, in pages */ 308 309 struct spdk_bs_type bstype; /* blobstore type */ 310 311 uint8_t reserved[4020]; 312 uint32_t crc; 313 }; 314 SPDK_STATIC_ASSERT(sizeof(struct spdk_bs_super_block) == 0x1000, "Invalid super block size"); 315 316 #pragma pack(pop) 317 318 /* Unit Conversions 319 * 320 * The blobstore works with several different units: 321 * - Byte: Self explanatory 322 * - LBA: The logical blocks on the backing storage device. 323 * - Page: The read/write units of blobs and metadata. This is 324 * an offset into a blob in units of 4KiB. 325 * - Cluster Index: The disk is broken into a sequential list of 326 * clusters. This is the offset from the beginning. 327 * 328 * NOTE: These conversions all act on simple magnitudes, not with any sort 329 * of knowledge about the blobs themselves. For instance, converting 330 * a page to an lba with the conversion function below simply converts 331 * a number of pages to an equivalent number of lbas, but that 332 * lba certainly isn't the right lba that corresponds to a page offset 333 * for a particular blob. 334 */ 335 static inline uint64_t 336 _spdk_bs_byte_to_lba(struct spdk_blob_store *bs, uint64_t length) 337 { 338 assert(length % bs->dev->blocklen == 0); 339 340 return length / bs->dev->blocklen; 341 } 342 343 static inline uint64_t 344 _spdk_bs_lba_to_byte(struct spdk_blob_store *bs, uint64_t lba) 345 { 346 return lba * bs->dev->blocklen; 347 } 348 349 static inline uint64_t 350 _spdk_bs_page_to_lba(struct spdk_blob_store *bs, uint64_t page) 351 { 352 return page * SPDK_BS_PAGE_SIZE / bs->dev->blocklen; 353 } 354 355 static inline uint32_t 356 _spdk_bs_lba_to_page(struct spdk_blob_store *bs, uint64_t lba) 357 { 358 uint64_t lbas_per_page; 359 360 lbas_per_page = SPDK_BS_PAGE_SIZE / bs->dev->blocklen; 361 362 assert(lba % lbas_per_page == 0); 363 364 return lba / lbas_per_page; 365 } 366 367 static inline uint64_t 368 _spdk_bs_cluster_to_page(struct spdk_blob_store *bs, uint32_t cluster) 369 { 370 return cluster * bs->pages_per_cluster; 371 } 372 373 static inline uint32_t 374 _spdk_bs_page_to_cluster(struct spdk_blob_store *bs, uint64_t page) 375 { 376 assert(page % bs->pages_per_cluster == 0); 377 378 return page / bs->pages_per_cluster; 379 } 380 381 static inline uint64_t 382 _spdk_bs_cluster_to_lba(struct spdk_blob_store *bs, uint32_t cluster) 383 { 384 return cluster * (bs->cluster_sz / bs->dev->blocklen); 385 } 386 387 static inline uint32_t 388 _spdk_bs_lba_to_cluster(struct spdk_blob_store *bs, uint64_t lba) 389 { 390 assert(lba % (bs->cluster_sz / bs->dev->blocklen) == 0); 391 392 return lba / (bs->cluster_sz / bs->dev->blocklen); 393 } 394 395 /* End basic conversions */ 396 397 static inline uint32_t 398 _spdk_bs_blobid_to_page(spdk_blob_id id) 399 { 400 return id & 0xFFFFFFFF; 401 } 402 403 /* The blob id is a 64 bit number. The lower 32 bits are the page_idx. The upper 404 * 32 bits are not currently used. Stick a 1 there just to catch bugs where the 405 * code assumes blob id == page_idx. 406 */ 407 static inline spdk_blob_id 408 _spdk_bs_page_to_blobid(uint32_t page_idx) 409 { 410 return SPDK_BLOB_BLOBID_HIGH_BIT | page_idx; 411 } 412 413 /* Given a page offset into a blob, look up the LBA for the 414 * start of that page. 415 */ 416 static inline uint64_t 417 _spdk_bs_blob_page_to_lba(struct spdk_blob_data *blob, uint32_t page) 418 { 419 uint64_t lba; 420 uint32_t pages_per_cluster; 421 422 pages_per_cluster = blob->bs->pages_per_cluster; 423 424 assert(page < blob->active.num_clusters * pages_per_cluster); 425 426 lba = blob->active.clusters[page / pages_per_cluster]; 427 lba += _spdk_bs_page_to_lba(blob->bs, page % pages_per_cluster); 428 429 return lba; 430 } 431 432 /* Given a page offset into a blob, look up the number of pages until the 433 * next cluster boundary. 434 */ 435 static inline uint32_t 436 _spdk_bs_num_pages_to_cluster_boundary(struct spdk_blob_data *blob, uint32_t page) 437 { 438 uint32_t pages_per_cluster; 439 440 pages_per_cluster = blob->bs->pages_per_cluster; 441 442 return pages_per_cluster - (page % pages_per_cluster); 443 } 444 445 #endif 446