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 15struct bio_set; 16struct bio; 17struct bio_integrity_payload; 18struct page; 19struct io_context; 20struct cgroup_subsys_state; 21typedef void (bio_end_io_t) (struct bio *); 22struct 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 41struct 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 bool bd_read_only; /* read-only policy */ 49 u8 bd_partno; 50 bool bd_write_holder; 51 bool bd_has_submit_bio; 52 dev_t bd_dev; 53 struct inode *bd_inode; /* will die */ 54 55 atomic_t bd_openers; 56 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */ 57 void * bd_claiming; 58 void * bd_holder; 59 const struct blk_holder_ops *bd_holder_ops; 60 struct mutex bd_holder_lock; 61 int bd_holders; 62 struct kobject *bd_holder_dir; 63 64 atomic_t bd_fsfreeze_count; /* number of freeze requests */ 65 struct mutex bd_fsfreeze_mutex; /* serialize freeze/thaw */ 66 67 struct partition_meta_info *bd_meta_info; 68#ifdef CONFIG_FAIL_MAKE_REQUEST 69 bool bd_make_it_fail; 70#endif 71 bool bd_ro_warned; 72 int bd_writers; 73 /* 74 * keep this out-of-line as it's both big and not needed in the fast 75 * path 76 */ 77 struct device bd_device; 78} __randomize_layout; 79 80#define bdev_whole(_bdev) \ 81 ((_bdev)->bd_disk->part0) 82 83#define dev_to_bdev(device) \ 84 container_of((device), struct block_device, bd_device) 85 86#define bdev_kobj(_bdev) \ 87 (&((_bdev)->bd_device.kobj)) 88 89/* 90 * Block error status values. See block/blk-core:blk_errors for the details. 91 * Alpha cannot write a byte atomically, so we need to use 32-bit value. 92 */ 93#if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__) 94typedef u32 __bitwise blk_status_t; 95typedef u32 blk_short_t; 96#else 97typedef u8 __bitwise blk_status_t; 98typedef u16 blk_short_t; 99#endif 100#define BLK_STS_OK 0 101#define BLK_STS_NOTSUPP ((__force blk_status_t)1) 102#define BLK_STS_TIMEOUT ((__force blk_status_t)2) 103#define BLK_STS_NOSPC ((__force blk_status_t)3) 104#define BLK_STS_TRANSPORT ((__force blk_status_t)4) 105#define BLK_STS_TARGET ((__force blk_status_t)5) 106#define BLK_STS_RESV_CONFLICT ((__force blk_status_t)6) 107#define BLK_STS_MEDIUM ((__force blk_status_t)7) 108#define BLK_STS_PROTECTION ((__force blk_status_t)8) 109#define BLK_STS_RESOURCE ((__force blk_status_t)9) 110#define BLK_STS_IOERR ((__force blk_status_t)10) 111 112/* hack for device mapper, don't use elsewhere: */ 113#define BLK_STS_DM_REQUEUE ((__force blk_status_t)11) 114 115/* 116 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set 117 * and the bio would block (cf bio_wouldblock_error()) 118 */ 119#define BLK_STS_AGAIN ((__force blk_status_t)12) 120 121/* 122 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if 123 * device related resources are unavailable, but the driver can guarantee 124 * that the queue will be rerun in the future once resources become 125 * available again. This is typically the case for device specific 126 * resources that are consumed for IO. If the driver fails allocating these 127 * resources, we know that inflight (or pending) IO will free these 128 * resource upon completion. 129 * 130 * This is different from BLK_STS_RESOURCE in that it explicitly references 131 * a device specific resource. For resources of wider scope, allocation 132 * failure can happen without having pending IO. This means that we can't 133 * rely on request completions freeing these resources, as IO may not be in 134 * flight. Examples of that are kernel memory allocations, DMA mappings, or 135 * any other system wide resources. 136 */ 137#define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13) 138 139/* 140 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone 141 * related resources are unavailable, but the driver can guarantee the queue 142 * will be rerun in the future once the resources become available again. 143 * 144 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references 145 * a zone specific resource and IO to a different zone on the same device could 146 * still be served. Examples of that are zones that are write-locked, but a read 147 * to the same zone could be served. 148 */ 149#define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14) 150 151/* 152 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion 153 * path if the device returns a status indicating that too many zone resources 154 * are currently open. The same command should be successful if resubmitted 155 * after the number of open zones decreases below the device's limits, which is 156 * reported in the request_queue's max_open_zones. 157 */ 158#define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15) 159 160/* 161 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion 162 * path if the device returns a status indicating that too many zone resources 163 * are currently active. The same command should be successful if resubmitted 164 * after the number of active zones decreases below the device's limits, which 165 * is reported in the request_queue's max_active_zones. 166 */ 167#define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16) 168 169/* 170 * BLK_STS_OFFLINE is returned from the driver when the target device is offline 171 * or is being taken offline. This could help differentiate the case where a 172 * device is intentionally being shut down from a real I/O error. 173 */ 174#define BLK_STS_OFFLINE ((__force blk_status_t)17) 175 176/* 177 * BLK_STS_DURATION_LIMIT is returned from the driver when the target device 178 * aborted the command because it exceeded one of its Command Duration Limits. 179 */ 180#define BLK_STS_DURATION_LIMIT ((__force blk_status_t)18) 181 182/** 183 * blk_path_error - returns true if error may be path related 184 * @error: status the request was completed with 185 * 186 * Description: 187 * This classifies block error status into non-retryable errors and ones 188 * that may be successful if retried on a failover path. 189 * 190 * Return: 191 * %false - retrying failover path will not help 192 * %true - may succeed if retried 193 */ 194static inline bool blk_path_error(blk_status_t error) 195{ 196 switch (error) { 197 case BLK_STS_NOTSUPP: 198 case BLK_STS_NOSPC: 199 case BLK_STS_TARGET: 200 case BLK_STS_RESV_CONFLICT: 201 case BLK_STS_MEDIUM: 202 case BLK_STS_PROTECTION: 203 return false; 204 } 205 206 /* Anything else could be a path failure, so should be retried */ 207 return true; 208} 209 210struct bio_issue { 211 u64 value; 212}; 213 214typedef __u32 __bitwise blk_opf_t; 215 216typedef unsigned int blk_qc_t; 217#define BLK_QC_T_NONE -1U 218 219/* 220 * main unit of I/O for the block layer and lower layers (ie drivers and 221 * stacking drivers) 222 */ 223struct bio { 224 struct bio *bi_next; /* request queue link */ 225 struct block_device *bi_bdev; 226 blk_opf_t bi_opf; /* bottom bits REQ_OP, top bits 227 * req_flags. 228 */ 229 unsigned short bi_flags; /* BIO_* below */ 230 unsigned short bi_ioprio; 231 enum rw_hint bi_write_hint; 232 blk_status_t bi_status; 233 atomic_t __bi_remaining; 234 235 struct bvec_iter bi_iter; 236 237 blk_qc_t bi_cookie; 238 bio_end_io_t *bi_end_io; 239 void *bi_private; 240#ifdef CONFIG_BLK_CGROUP 241 /* 242 * Represents the association of the css and request_queue for the bio. 243 * If a bio goes direct to device, it will not have a blkg as it will 244 * not have a request_queue associated with it. The reference is put 245 * on release of the bio. 246 */ 247 struct blkcg_gq *bi_blkg; 248 struct bio_issue bi_issue; 249#ifdef CONFIG_BLK_CGROUP_IOCOST 250 u64 bi_iocost_cost; 251#endif 252#endif 253 254#ifdef CONFIG_BLK_INLINE_ENCRYPTION 255 struct bio_crypt_ctx *bi_crypt_context; 256#endif 257 258 union { 259#if defined(CONFIG_BLK_DEV_INTEGRITY) 260 struct bio_integrity_payload *bi_integrity; /* data integrity */ 261#endif 262 }; 263 264 unsigned short bi_vcnt; /* how many bio_vec's */ 265 266 /* 267 * Everything starting with bi_max_vecs will be preserved by bio_reset() 268 */ 269 270 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */ 271 272 atomic_t __bi_cnt; /* pin count */ 273 274 struct bio_vec *bi_io_vec; /* the actual vec list */ 275 276 struct bio_set *bi_pool; 277 278 /* 279 * We can inline a number of vecs at the end of the bio, to avoid 280 * double allocations for a small number of bio_vecs. This member 281 * MUST obviously be kept at the very end of the bio. 282 */ 283 struct bio_vec bi_inline_vecs[]; 284}; 285 286#define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs) 287#define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT) 288 289/* 290 * bio flags 291 */ 292enum { 293 BIO_PAGE_PINNED, /* Unpin pages in bio_release_pages() */ 294 BIO_CLONED, /* doesn't own data */ 295 BIO_BOUNCED, /* bio is a bounce bio */ 296 BIO_QUIET, /* Make BIO Quiet */ 297 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */ 298 BIO_REFFED, /* bio has elevated ->bi_cnt */ 299 BIO_BPS_THROTTLED, /* This bio has already been subjected to 300 * throttling rules. Don't do it again. */ 301 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion 302 * of this bio. */ 303 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ 304 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */ 305 BIO_QOS_MERGED, /* but went through rq_qos merge path */ 306 BIO_REMAPPED, 307 BIO_ZONE_WRITE_LOCKED, /* Owns a zoned device zone write lock */ 308 BIO_FLAG_LAST 309}; 310 311typedef __u32 __bitwise blk_mq_req_flags_t; 312 313#define REQ_OP_BITS 8 314#define REQ_OP_MASK (__force blk_opf_t)((1 << REQ_OP_BITS) - 1) 315#define REQ_FLAG_BITS 24 316 317/** 318 * enum req_op - Operations common to the bio and request structures. 319 * We use 8 bits for encoding the operation, and the remaining 24 for flags. 320 * 321 * The least significant bit of the operation number indicates the data 322 * transfer direction: 323 * 324 * - if the least significant bit is set transfers are TO the device 325 * - if the least significant bit is not set transfers are FROM the device 326 * 327 * If a operation does not transfer data the least significant bit has no 328 * meaning. 329 */ 330enum req_op { 331 /* read sectors from the device */ 332 REQ_OP_READ = (__force blk_opf_t)0, 333 /* write sectors to the device */ 334 REQ_OP_WRITE = (__force blk_opf_t)1, 335 /* flush the volatile write cache */ 336 REQ_OP_FLUSH = (__force blk_opf_t)2, 337 /* discard sectors */ 338 REQ_OP_DISCARD = (__force blk_opf_t)3, 339 /* securely erase sectors */ 340 REQ_OP_SECURE_ERASE = (__force blk_opf_t)5, 341 /* write data at the current zone write pointer */ 342 REQ_OP_ZONE_APPEND = (__force blk_opf_t)7, 343 /* write the zero filled sector many times */ 344 REQ_OP_WRITE_ZEROES = (__force blk_opf_t)9, 345 /* Open a zone */ 346 REQ_OP_ZONE_OPEN = (__force blk_opf_t)10, 347 /* Close a zone */ 348 REQ_OP_ZONE_CLOSE = (__force blk_opf_t)11, 349 /* Transition a zone to full */ 350 REQ_OP_ZONE_FINISH = (__force blk_opf_t)12, 351 /* reset a zone write pointer */ 352 REQ_OP_ZONE_RESET = (__force blk_opf_t)13, 353 /* reset all the zone present on the device */ 354 REQ_OP_ZONE_RESET_ALL = (__force blk_opf_t)15, 355 356 /* Driver private requests */ 357 REQ_OP_DRV_IN = (__force blk_opf_t)34, 358 REQ_OP_DRV_OUT = (__force blk_opf_t)35, 359 360 REQ_OP_LAST = (__force blk_opf_t)36, 361}; 362 363enum req_flag_bits { 364 __REQ_FAILFAST_DEV = /* no driver retries of device errors */ 365 REQ_OP_BITS, 366 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */ 367 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */ 368 __REQ_SYNC, /* request is sync (sync write or read) */ 369 __REQ_META, /* metadata io request */ 370 __REQ_PRIO, /* boost priority in cfq */ 371 __REQ_NOMERGE, /* don't touch this for merging */ 372 __REQ_IDLE, /* anticipate more IO after this one */ 373 __REQ_INTEGRITY, /* I/O includes block integrity payload */ 374 __REQ_FUA, /* forced unit access */ 375 __REQ_PREFLUSH, /* request for cache flush */ 376 __REQ_RAHEAD, /* read ahead, can fail anytime */ 377 __REQ_BACKGROUND, /* background IO */ 378 __REQ_NOWAIT, /* Don't wait if request will block */ 379 __REQ_POLLED, /* caller polls for completion using bio_poll */ 380 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */ 381 __REQ_SWAP, /* swap I/O */ 382 __REQ_DRV, /* for driver use */ 383 __REQ_FS_PRIVATE, /* for file system (submitter) use */ 384 385 /* 386 * Command specific flags, keep last: 387 */ 388 /* for REQ_OP_WRITE_ZEROES: */ 389 __REQ_NOUNMAP, /* do not free blocks when zeroing */ 390 391 __REQ_NR_BITS, /* stops here */ 392}; 393 394#define REQ_FAILFAST_DEV \ 395 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV) 396#define REQ_FAILFAST_TRANSPORT \ 397 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT) 398#define REQ_FAILFAST_DRIVER \ 399 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER) 400#define REQ_SYNC (__force blk_opf_t)(1ULL << __REQ_SYNC) 401#define REQ_META (__force blk_opf_t)(1ULL << __REQ_META) 402#define REQ_PRIO (__force blk_opf_t)(1ULL << __REQ_PRIO) 403#define REQ_NOMERGE (__force blk_opf_t)(1ULL << __REQ_NOMERGE) 404#define REQ_IDLE (__force blk_opf_t)(1ULL << __REQ_IDLE) 405#define REQ_INTEGRITY (__force blk_opf_t)(1ULL << __REQ_INTEGRITY) 406#define REQ_FUA (__force blk_opf_t)(1ULL << __REQ_FUA) 407#define REQ_PREFLUSH (__force blk_opf_t)(1ULL << __REQ_PREFLUSH) 408#define REQ_RAHEAD (__force blk_opf_t)(1ULL << __REQ_RAHEAD) 409#define REQ_BACKGROUND (__force blk_opf_t)(1ULL << __REQ_BACKGROUND) 410#define REQ_NOWAIT (__force blk_opf_t)(1ULL << __REQ_NOWAIT) 411#define REQ_POLLED (__force blk_opf_t)(1ULL << __REQ_POLLED) 412#define REQ_ALLOC_CACHE (__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE) 413#define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP) 414#define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV) 415#define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE) 416 417#define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP) 418 419#define REQ_FAILFAST_MASK \ 420 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER) 421 422#define REQ_NOMERGE_FLAGS \ 423 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA) 424 425enum stat_group { 426 STAT_READ, 427 STAT_WRITE, 428 STAT_DISCARD, 429 STAT_FLUSH, 430 431 NR_STAT_GROUPS 432}; 433 434static inline enum req_op bio_op(const struct bio *bio) 435{ 436 return bio->bi_opf & REQ_OP_MASK; 437} 438 439static inline bool op_is_write(blk_opf_t op) 440{ 441 return !!(op & (__force blk_opf_t)1); 442} 443 444/* 445 * Check if the bio or request is one that needs special treatment in the 446 * flush state machine. 447 */ 448static inline bool op_is_flush(blk_opf_t op) 449{ 450 return op & (REQ_FUA | REQ_PREFLUSH); 451} 452 453/* 454 * Reads are always treated as synchronous, as are requests with the FUA or 455 * PREFLUSH flag. Other operations may be marked as synchronous using the 456 * REQ_SYNC flag. 457 */ 458static inline bool op_is_sync(blk_opf_t op) 459{ 460 return (op & REQ_OP_MASK) == REQ_OP_READ || 461 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH)); 462} 463 464static inline bool op_is_discard(blk_opf_t op) 465{ 466 return (op & REQ_OP_MASK) == REQ_OP_DISCARD; 467} 468 469/* 470 * Check if a bio or request operation is a zone management operation, with 471 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case 472 * due to its different handling in the block layer and device response in 473 * case of command failure. 474 */ 475static inline bool op_is_zone_mgmt(enum req_op op) 476{ 477 switch (op & REQ_OP_MASK) { 478 case REQ_OP_ZONE_RESET: 479 case REQ_OP_ZONE_OPEN: 480 case REQ_OP_ZONE_CLOSE: 481 case REQ_OP_ZONE_FINISH: 482 return true; 483 default: 484 return false; 485 } 486} 487 488static inline int op_stat_group(enum req_op op) 489{ 490 if (op_is_discard(op)) 491 return STAT_DISCARD; 492 return op_is_write(op); 493} 494 495struct blk_rq_stat { 496 u64 mean; 497 u64 min; 498 u64 max; 499 u32 nr_samples; 500 u64 batch; 501}; 502 503#endif /* __LINUX_BLK_TYPES_H */ 504