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