1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Block data types and constants. Directly include this file only to 4 * break include dependency loop. 5 */ 6 #ifndef __LINUX_BLK_TYPES_H 7 #define __LINUX_BLK_TYPES_H 8 9 #include <linux/types.h> 10 #include <linux/bvec.h> 11 #include <linux/device.h> 12 #include <linux/ktime.h> 13 #include <linux/rw_hint.h> 14 15 struct bio_set; 16 struct bio; 17 struct bio_integrity_payload; 18 struct page; 19 struct io_context; 20 struct cgroup_subsys_state; 21 typedef void (bio_end_io_t) (struct bio *); 22 struct bio_crypt_ctx; 23 24 /* 25 * The basic unit of block I/O is a sector. It is used in a number of contexts 26 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 27 * bytes. Variables of type sector_t represent an offset or size that is a 28 * multiple of 512 bytes. Hence these two constants. 29 */ 30 #ifndef SECTOR_SHIFT 31 #define SECTOR_SHIFT 9 32 #endif 33 #ifndef SECTOR_SIZE 34 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 35 #endif 36 37 #define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT) 38 #define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT) 39 #define SECTOR_MASK (PAGE_SECTORS - 1) 40 41 struct block_device { 42 sector_t bd_start_sect; 43 sector_t bd_nr_sectors; 44 struct gendisk * bd_disk; 45 struct request_queue * bd_queue; 46 struct disk_stats __percpu *bd_stats; 47 unsigned long bd_stamp; 48 atomic_t __bd_flags; // partition number + flags 49 #define BD_PARTNO 255 // lower 8 bits; assign-once 50 #define BD_READ_ONLY (1u<<8) // read-only policy 51 #define BD_WRITE_HOLDER (1u<<9) 52 #define BD_HAS_SUBMIT_BIO (1u<<10) 53 #define BD_RO_WARNED (1u<<11) 54 #ifdef CONFIG_FAIL_MAKE_REQUEST 55 #define BD_MAKE_IT_FAIL (1u<<12) 56 #endif 57 dev_t bd_dev; 58 struct address_space *bd_mapping; /* page cache */ 59 60 atomic_t bd_openers; 61 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */ 62 void * bd_claiming; 63 void * bd_holder; 64 const struct blk_holder_ops *bd_holder_ops; 65 struct mutex bd_holder_lock; 66 int bd_holders; 67 struct kobject *bd_holder_dir; 68 69 atomic_t bd_fsfreeze_count; /* number of freeze requests */ 70 struct mutex bd_fsfreeze_mutex; /* serialize freeze/thaw */ 71 72 struct partition_meta_info *bd_meta_info; 73 int bd_writers; 74 /* 75 * keep this out-of-line as it's both big and not needed in the fast 76 * path 77 */ 78 struct device bd_device; 79 } __randomize_layout; 80 81 #define bdev_whole(_bdev) \ 82 ((_bdev)->bd_disk->part0) 83 84 #define dev_to_bdev(device) \ 85 container_of((device), struct block_device, bd_device) 86 87 #define bdev_kobj(_bdev) \ 88 (&((_bdev)->bd_device.kobj)) 89 90 /* 91 * Block error status values. See block/blk-core:blk_errors for the details. 92 */ 93 typedef u8 __bitwise blk_status_t; 94 typedef u16 blk_short_t; 95 #define BLK_STS_OK 0 96 #define BLK_STS_NOTSUPP ((__force blk_status_t)1) 97 #define BLK_STS_TIMEOUT ((__force blk_status_t)2) 98 #define BLK_STS_NOSPC ((__force blk_status_t)3) 99 #define BLK_STS_TRANSPORT ((__force blk_status_t)4) 100 #define BLK_STS_TARGET ((__force blk_status_t)5) 101 #define BLK_STS_RESV_CONFLICT ((__force blk_status_t)6) 102 #define BLK_STS_MEDIUM ((__force blk_status_t)7) 103 #define BLK_STS_PROTECTION ((__force blk_status_t)8) 104 #define BLK_STS_RESOURCE ((__force blk_status_t)9) 105 #define BLK_STS_IOERR ((__force blk_status_t)10) 106 107 /* hack for device mapper, don't use elsewhere: */ 108 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11) 109 110 /* 111 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set 112 * and the bio would block (cf bio_wouldblock_error()) 113 */ 114 #define BLK_STS_AGAIN ((__force blk_status_t)12) 115 116 /* 117 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if 118 * device related resources are unavailable, but the driver can guarantee 119 * that the queue will be rerun in the future once resources become 120 * available again. This is typically the case for device specific 121 * resources that are consumed for IO. If the driver fails allocating these 122 * resources, we know that inflight (or pending) IO will free these 123 * resource upon completion. 124 * 125 * This is different from BLK_STS_RESOURCE in that it explicitly references 126 * a device specific resource. For resources of wider scope, allocation 127 * failure can happen without having pending IO. This means that we can't 128 * rely on request completions freeing these resources, as IO may not be in 129 * flight. Examples of that are kernel memory allocations, DMA mappings, or 130 * any other system wide resources. 131 */ 132 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13) 133 134 /* 135 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion 136 * path if the device returns a status indicating that too many zone resources 137 * are currently open. The same command should be successful if resubmitted 138 * after the number of open zones decreases below the device's limits, which is 139 * reported in the request_queue's max_open_zones. 140 */ 141 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)14) 142 143 /* 144 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion 145 * path if the device returns a status indicating that too many zone resources 146 * are currently active. The same command should be successful if resubmitted 147 * after the number of active zones decreases below the device's limits, which 148 * is reported in the request_queue's max_active_zones. 149 */ 150 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)15) 151 152 /* 153 * BLK_STS_OFFLINE is returned from the driver when the target device is offline 154 * or is being taken offline. This could help differentiate the case where a 155 * device is intentionally being shut down from a real I/O error. 156 */ 157 #define BLK_STS_OFFLINE ((__force blk_status_t)16) 158 159 /* 160 * BLK_STS_DURATION_LIMIT is returned from the driver when the target device 161 * aborted the command because it exceeded one of its Command Duration Limits. 162 */ 163 #define BLK_STS_DURATION_LIMIT ((__force blk_status_t)17) 164 165 /** 166 * blk_path_error - returns true if error may be path related 167 * @error: status the request was completed with 168 * 169 * Description: 170 * This classifies block error status into non-retryable errors and ones 171 * that may be successful if retried on a failover path. 172 * 173 * Return: 174 * %false - retrying failover path will not help 175 * %true - may succeed if retried 176 */ 177 static inline bool blk_path_error(blk_status_t error) 178 { 179 switch (error) { 180 case BLK_STS_NOTSUPP: 181 case BLK_STS_NOSPC: 182 case BLK_STS_TARGET: 183 case BLK_STS_RESV_CONFLICT: 184 case BLK_STS_MEDIUM: 185 case BLK_STS_PROTECTION: 186 return false; 187 } 188 189 /* Anything else could be a path failure, so should be retried */ 190 return true; 191 } 192 193 struct bio_issue { 194 u64 value; 195 }; 196 197 typedef __u32 __bitwise blk_opf_t; 198 199 typedef unsigned int blk_qc_t; 200 #define BLK_QC_T_NONE -1U 201 202 /* 203 * main unit of I/O for the block layer and lower layers (ie drivers and 204 * stacking drivers) 205 */ 206 struct bio { 207 struct bio *bi_next; /* request queue link */ 208 struct block_device *bi_bdev; 209 blk_opf_t bi_opf; /* bottom bits REQ_OP, top bits 210 * req_flags. 211 */ 212 unsigned short bi_flags; /* BIO_* below */ 213 unsigned short bi_ioprio; 214 enum rw_hint bi_write_hint; 215 blk_status_t bi_status; 216 atomic_t __bi_remaining; 217 218 struct bvec_iter bi_iter; 219 220 union { 221 /* for polled bios: */ 222 blk_qc_t bi_cookie; 223 /* for plugged zoned writes only: */ 224 unsigned int __bi_nr_segments; 225 }; 226 bio_end_io_t *bi_end_io; 227 void *bi_private; 228 #ifdef CONFIG_BLK_CGROUP 229 /* 230 * Represents the association of the css and request_queue for the bio. 231 * If a bio goes direct to device, it will not have a blkg as it will 232 * not have a request_queue associated with it. The reference is put 233 * on release of the bio. 234 */ 235 struct blkcg_gq *bi_blkg; 236 struct bio_issue bi_issue; 237 #ifdef CONFIG_BLK_CGROUP_IOCOST 238 u64 bi_iocost_cost; 239 #endif 240 #endif 241 242 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 243 struct bio_crypt_ctx *bi_crypt_context; 244 #endif 245 246 union { 247 #if defined(CONFIG_BLK_DEV_INTEGRITY) 248 struct bio_integrity_payload *bi_integrity; /* data integrity */ 249 #endif 250 }; 251 252 unsigned short bi_vcnt; /* how many bio_vec's */ 253 254 /* 255 * Everything starting with bi_max_vecs will be preserved by bio_reset() 256 */ 257 258 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */ 259 260 atomic_t __bi_cnt; /* pin count */ 261 262 struct bio_vec *bi_io_vec; /* the actual vec list */ 263 264 struct bio_set *bi_pool; 265 266 /* 267 * We can inline a number of vecs at the end of the bio, to avoid 268 * double allocations for a small number of bio_vecs. This member 269 * MUST obviously be kept at the very end of the bio. 270 */ 271 struct bio_vec bi_inline_vecs[]; 272 }; 273 274 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs) 275 #define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT) 276 277 /* 278 * bio flags 279 */ 280 enum { 281 BIO_PAGE_PINNED, /* Unpin pages in bio_release_pages() */ 282 BIO_CLONED, /* doesn't own data */ 283 BIO_BOUNCED, /* bio is a bounce bio */ 284 BIO_QUIET, /* Make BIO Quiet */ 285 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */ 286 BIO_REFFED, /* bio has elevated ->bi_cnt */ 287 BIO_BPS_THROTTLED, /* This bio has already been subjected to 288 * throttling rules. Don't do it again. */ 289 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion 290 * of this bio. */ 291 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ 292 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */ 293 BIO_QOS_MERGED, /* but went through rq_qos merge path */ 294 BIO_REMAPPED, 295 BIO_ZONE_WRITE_PLUGGING, /* bio handled through zone write plugging */ 296 BIO_EMULATES_ZONE_APPEND, /* bio emulates a zone append operation */ 297 BIO_FLAG_LAST 298 }; 299 300 typedef __u32 __bitwise blk_mq_req_flags_t; 301 302 #define REQ_OP_BITS 8 303 #define REQ_OP_MASK (__force blk_opf_t)((1 << REQ_OP_BITS) - 1) 304 #define REQ_FLAG_BITS 24 305 306 /** 307 * enum req_op - Operations common to the bio and request structures. 308 * We use 8 bits for encoding the operation, and the remaining 24 for flags. 309 * 310 * The least significant bit of the operation number indicates the data 311 * transfer direction: 312 * 313 * - if the least significant bit is set transfers are TO the device 314 * - if the least significant bit is not set transfers are FROM the device 315 * 316 * If a operation does not transfer data the least significant bit has no 317 * meaning. 318 */ 319 enum req_op { 320 /* read sectors from the device */ 321 REQ_OP_READ = (__force blk_opf_t)0, 322 /* write sectors to the device */ 323 REQ_OP_WRITE = (__force blk_opf_t)1, 324 /* flush the volatile write cache */ 325 REQ_OP_FLUSH = (__force blk_opf_t)2, 326 /* discard sectors */ 327 REQ_OP_DISCARD = (__force blk_opf_t)3, 328 /* securely erase sectors */ 329 REQ_OP_SECURE_ERASE = (__force blk_opf_t)5, 330 /* write data at the current zone write pointer */ 331 REQ_OP_ZONE_APPEND = (__force blk_opf_t)7, 332 /* write the zero filled sector many times */ 333 REQ_OP_WRITE_ZEROES = (__force blk_opf_t)9, 334 /* Open a zone */ 335 REQ_OP_ZONE_OPEN = (__force blk_opf_t)10, 336 /* Close a zone */ 337 REQ_OP_ZONE_CLOSE = (__force blk_opf_t)11, 338 /* Transition a zone to full */ 339 REQ_OP_ZONE_FINISH = (__force blk_opf_t)12, 340 /* reset a zone write pointer */ 341 REQ_OP_ZONE_RESET = (__force blk_opf_t)13, 342 /* reset all the zone present on the device */ 343 REQ_OP_ZONE_RESET_ALL = (__force blk_opf_t)15, 344 345 /* Driver private requests */ 346 REQ_OP_DRV_IN = (__force blk_opf_t)34, 347 REQ_OP_DRV_OUT = (__force blk_opf_t)35, 348 349 REQ_OP_LAST = (__force blk_opf_t)36, 350 }; 351 352 enum req_flag_bits { 353 __REQ_FAILFAST_DEV = /* no driver retries of device errors */ 354 REQ_OP_BITS, 355 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */ 356 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */ 357 __REQ_SYNC, /* request is sync (sync write or read) */ 358 __REQ_META, /* metadata io request */ 359 __REQ_PRIO, /* boost priority in cfq */ 360 __REQ_NOMERGE, /* don't touch this for merging */ 361 __REQ_IDLE, /* anticipate more IO after this one */ 362 __REQ_INTEGRITY, /* I/O includes block integrity payload */ 363 __REQ_FUA, /* forced unit access */ 364 __REQ_PREFLUSH, /* request for cache flush */ 365 __REQ_RAHEAD, /* read ahead, can fail anytime */ 366 __REQ_BACKGROUND, /* background IO */ 367 __REQ_NOWAIT, /* Don't wait if request will block */ 368 __REQ_POLLED, /* caller polls for completion using bio_poll */ 369 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */ 370 __REQ_SWAP, /* swap I/O */ 371 __REQ_DRV, /* for driver use */ 372 __REQ_FS_PRIVATE, /* for file system (submitter) use */ 373 374 /* 375 * Command specific flags, keep last: 376 */ 377 /* for REQ_OP_WRITE_ZEROES: */ 378 __REQ_NOUNMAP, /* do not free blocks when zeroing */ 379 380 __REQ_NR_BITS, /* stops here */ 381 }; 382 383 #define REQ_FAILFAST_DEV \ 384 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV) 385 #define REQ_FAILFAST_TRANSPORT \ 386 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT) 387 #define REQ_FAILFAST_DRIVER \ 388 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER) 389 #define REQ_SYNC (__force blk_opf_t)(1ULL << __REQ_SYNC) 390 #define REQ_META (__force blk_opf_t)(1ULL << __REQ_META) 391 #define REQ_PRIO (__force blk_opf_t)(1ULL << __REQ_PRIO) 392 #define REQ_NOMERGE (__force blk_opf_t)(1ULL << __REQ_NOMERGE) 393 #define REQ_IDLE (__force blk_opf_t)(1ULL << __REQ_IDLE) 394 #define REQ_INTEGRITY (__force blk_opf_t)(1ULL << __REQ_INTEGRITY) 395 #define REQ_FUA (__force blk_opf_t)(1ULL << __REQ_FUA) 396 #define REQ_PREFLUSH (__force blk_opf_t)(1ULL << __REQ_PREFLUSH) 397 #define REQ_RAHEAD (__force blk_opf_t)(1ULL << __REQ_RAHEAD) 398 #define REQ_BACKGROUND (__force blk_opf_t)(1ULL << __REQ_BACKGROUND) 399 #define REQ_NOWAIT (__force blk_opf_t)(1ULL << __REQ_NOWAIT) 400 #define REQ_POLLED (__force blk_opf_t)(1ULL << __REQ_POLLED) 401 #define REQ_ALLOC_CACHE (__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE) 402 #define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP) 403 #define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV) 404 #define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE) 405 406 #define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP) 407 408 #define REQ_FAILFAST_MASK \ 409 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER) 410 411 #define REQ_NOMERGE_FLAGS \ 412 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA) 413 414 enum stat_group { 415 STAT_READ, 416 STAT_WRITE, 417 STAT_DISCARD, 418 STAT_FLUSH, 419 420 NR_STAT_GROUPS 421 }; 422 423 static inline enum req_op bio_op(const struct bio *bio) 424 { 425 return bio->bi_opf & REQ_OP_MASK; 426 } 427 428 static inline bool op_is_write(blk_opf_t op) 429 { 430 return !!(op & (__force blk_opf_t)1); 431 } 432 433 /* 434 * Check if the bio or request is one that needs special treatment in the 435 * flush state machine. 436 */ 437 static inline bool op_is_flush(blk_opf_t op) 438 { 439 return op & (REQ_FUA | REQ_PREFLUSH); 440 } 441 442 /* 443 * Reads are always treated as synchronous, as are requests with the FUA or 444 * PREFLUSH flag. Other operations may be marked as synchronous using the 445 * REQ_SYNC flag. 446 */ 447 static inline bool op_is_sync(blk_opf_t op) 448 { 449 return (op & REQ_OP_MASK) == REQ_OP_READ || 450 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH)); 451 } 452 453 static inline bool op_is_discard(blk_opf_t op) 454 { 455 return (op & REQ_OP_MASK) == REQ_OP_DISCARD; 456 } 457 458 /* 459 * Check if a bio or request operation is a zone management operation, with 460 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case 461 * due to its different handling in the block layer and device response in 462 * case of command failure. 463 */ 464 static inline bool op_is_zone_mgmt(enum req_op op) 465 { 466 switch (op & REQ_OP_MASK) { 467 case REQ_OP_ZONE_RESET: 468 case REQ_OP_ZONE_OPEN: 469 case REQ_OP_ZONE_CLOSE: 470 case REQ_OP_ZONE_FINISH: 471 return true; 472 default: 473 return false; 474 } 475 } 476 477 static inline int op_stat_group(enum req_op op) 478 { 479 if (op_is_discard(op)) 480 return STAT_DISCARD; 481 return op_is_write(op); 482 } 483 484 struct blk_rq_stat { 485 u64 mean; 486 u64 min; 487 u64 max; 488 u32 nr_samples; 489 u64 batch; 490 }; 491 492 #endif /* __LINUX_BLK_TYPES_H */ 493