501 lines · c
1/* SPDX-License-Identifier: GPL-2.0 */2/*3 * Block data types and constants. Directly include this file only to4 * break include dependency loop.5 */6#ifndef __LINUX_BLK_TYPES_H7#define __LINUX_BLK_TYPES_H8 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 contexts26 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**927 * bytes. Variables of type sector_t represent an offset or size that is a28 * multiple of 512 bytes. Hence these two constants.29 */30#ifndef SECTOR_SHIFT31#define SECTOR_SHIFT 932#endif33#ifndef SECTOR_SIZE34#define SECTOR_SIZE (1 << SECTOR_SHIFT)35#endif36 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 atomic_t __bd_flags; // partition number + flags49#define BD_PARTNO 255 // lower 8 bits; assign-once50#define BD_READ_ONLY (1u<<8) // read-only policy51#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_REQUEST55#define BD_MAKE_IT_FAIL (1u<<12)56#endif57 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#ifdef CONFIG_SECURITY75 void *bd_security;76#endif77 /*78 * keep this out-of-line as it's both big and not needed in the fast79 * path80 */81 struct device bd_device;82} __randomize_layout;83 84#define bdev_whole(_bdev) \85 ((_bdev)->bd_disk->part0)86 87#define dev_to_bdev(device) \88 container_of((device), struct block_device, bd_device)89 90#define bdev_kobj(_bdev) \91 (&((_bdev)->bd_device.kobj))92 93/*94 * Block error status values. See block/blk-core:blk_errors for the details.95 */96typedef u8 __bitwise blk_status_t;97typedef u16 blk_short_t;98#define BLK_STS_OK 099#define BLK_STS_NOTSUPP ((__force blk_status_t)1)100#define BLK_STS_TIMEOUT ((__force blk_status_t)2)101#define BLK_STS_NOSPC ((__force blk_status_t)3)102#define BLK_STS_TRANSPORT ((__force blk_status_t)4)103#define BLK_STS_TARGET ((__force blk_status_t)5)104#define BLK_STS_RESV_CONFLICT ((__force blk_status_t)6)105#define BLK_STS_MEDIUM ((__force blk_status_t)7)106#define BLK_STS_PROTECTION ((__force blk_status_t)8)107#define BLK_STS_RESOURCE ((__force blk_status_t)9)108#define BLK_STS_IOERR ((__force blk_status_t)10)109 110/* hack for device mapper, don't use elsewhere: */111#define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)112 113/*114 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set115 * and the bio would block (cf bio_wouldblock_error())116 */117#define BLK_STS_AGAIN ((__force blk_status_t)12)118 119/*120 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if121 * device related resources are unavailable, but the driver can guarantee122 * that the queue will be rerun in the future once resources become123 * available again. This is typically the case for device specific124 * resources that are consumed for IO. If the driver fails allocating these125 * resources, we know that inflight (or pending) IO will free these126 * resource upon completion.127 *128 * This is different from BLK_STS_RESOURCE in that it explicitly references129 * a device specific resource. For resources of wider scope, allocation130 * failure can happen without having pending IO. This means that we can't131 * rely on request completions freeing these resources, as IO may not be in132 * flight. Examples of that are kernel memory allocations, DMA mappings, or133 * any other system wide resources.134 */135#define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)136 137/*138 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion139 * path if the device returns a status indicating that too many zone resources140 * are currently open. The same command should be successful if resubmitted141 * after the number of open zones decreases below the device's limits, which is142 * reported in the request_queue's max_open_zones.143 */144#define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)14)145 146/*147 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion148 * path if the device returns a status indicating that too many zone resources149 * are currently active. The same command should be successful if resubmitted150 * after the number of active zones decreases below the device's limits, which151 * is reported in the request_queue's max_active_zones.152 */153#define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)15)154 155/*156 * BLK_STS_OFFLINE is returned from the driver when the target device is offline157 * or is being taken offline. This could help differentiate the case where a158 * device is intentionally being shut down from a real I/O error.159 */160#define BLK_STS_OFFLINE ((__force blk_status_t)16)161 162/*163 * BLK_STS_DURATION_LIMIT is returned from the driver when the target device164 * aborted the command because it exceeded one of its Command Duration Limits.165 */166#define BLK_STS_DURATION_LIMIT ((__force blk_status_t)17)167 168/*169 * Invalid size or alignment.170 */171#define BLK_STS_INVAL ((__force blk_status_t)19)172 173/**174 * blk_path_error - returns true if error may be path related175 * @error: status the request was completed with176 *177 * Description:178 * This classifies block error status into non-retryable errors and ones179 * that may be successful if retried on a failover path.180 *181 * Return:182 * %false - retrying failover path will not help183 * %true - may succeed if retried184 */185static inline bool blk_path_error(blk_status_t error)186{187 switch (error) {188 case BLK_STS_NOTSUPP:189 case BLK_STS_NOSPC:190 case BLK_STS_TARGET:191 case BLK_STS_RESV_CONFLICT:192 case BLK_STS_MEDIUM:193 case BLK_STS_PROTECTION:194 return false;195 }196 197 /* Anything else could be a path failure, so should be retried */198 return true;199}200 201struct bio_issue {202 u64 value;203};204 205typedef __u32 __bitwise blk_opf_t;206 207typedef unsigned int blk_qc_t;208#define BLK_QC_T_NONE -1U209 210/*211 * main unit of I/O for the block layer and lower layers (ie drivers and212 * stacking drivers)213 */214struct bio {215 struct bio *bi_next; /* request queue link */216 struct block_device *bi_bdev;217 blk_opf_t bi_opf; /* bottom bits REQ_OP, top bits218 * req_flags.219 */220 unsigned short bi_flags; /* BIO_* below */221 unsigned short bi_ioprio;222 enum rw_hint bi_write_hint;223 blk_status_t bi_status;224 atomic_t __bi_remaining;225 226 struct bvec_iter bi_iter;227 228 union {229 /* for polled bios: */230 blk_qc_t bi_cookie;231 /* for plugged zoned writes only: */232 unsigned int __bi_nr_segments;233 };234 bio_end_io_t *bi_end_io;235 void *bi_private;236#ifdef CONFIG_BLK_CGROUP237 /*238 * Represents the association of the css and request_queue for the bio.239 * If a bio goes direct to device, it will not have a blkg as it will240 * not have a request_queue associated with it. The reference is put241 * on release of the bio.242 */243 struct blkcg_gq *bi_blkg;244 struct bio_issue bi_issue;245#ifdef CONFIG_BLK_CGROUP_IOCOST246 u64 bi_iocost_cost;247#endif248#endif249 250#ifdef CONFIG_BLK_INLINE_ENCRYPTION251 struct bio_crypt_ctx *bi_crypt_context;252#endif253 254#if defined(CONFIG_BLK_DEV_INTEGRITY)255 struct bio_integrity_payload *bi_integrity; /* data integrity */256#endif257 258 unsigned short bi_vcnt; /* how many bio_vec's */259 260 /*261 * Everything starting with bi_max_vecs will be preserved by bio_reset()262 */263 264 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */265 266 atomic_t __bi_cnt; /* pin count */267 268 struct bio_vec *bi_io_vec; /* the actual vec list */269 270 struct bio_set *bi_pool;271 272 /*273 * We can inline a number of vecs at the end of the bio, to avoid274 * double allocations for a small number of bio_vecs. This member275 * MUST obviously be kept at the very end of the bio.276 */277 struct bio_vec bi_inline_vecs[];278};279 280#define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)281#define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT)282 283/*284 * bio flags285 */286enum {287 BIO_PAGE_PINNED, /* Unpin pages in bio_release_pages() */288 BIO_CLONED, /* doesn't own data */289 BIO_BOUNCED, /* bio is a bounce bio */290 BIO_QUIET, /* Make BIO Quiet */291 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */292 BIO_REFFED, /* bio has elevated ->bi_cnt */293 BIO_BPS_THROTTLED, /* This bio has already been subjected to294 * throttling rules. Don't do it again. */295 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion296 * of this bio. */297 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */298 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */299 BIO_QOS_MERGED, /* but went through rq_qos merge path */300 BIO_REMAPPED,301 BIO_ZONE_WRITE_PLUGGING, /* bio handled through zone write plugging */302 BIO_EMULATES_ZONE_APPEND, /* bio emulates a zone append operation */303 BIO_FLAG_LAST304};305 306typedef __u32 __bitwise blk_mq_req_flags_t;307 308#define REQ_OP_BITS 8309#define REQ_OP_MASK (__force blk_opf_t)((1 << REQ_OP_BITS) - 1)310#define REQ_FLAG_BITS 24311 312/**313 * enum req_op - Operations common to the bio and request structures.314 * We use 8 bits for encoding the operation, and the remaining 24 for flags.315 *316 * The least significant bit of the operation number indicates the data317 * transfer direction:318 *319 * - if the least significant bit is set transfers are TO the device320 * - if the least significant bit is not set transfers are FROM the device321 *322 * If a operation does not transfer data the least significant bit has no323 * meaning.324 */325enum req_op {326 /* read sectors from the device */327 REQ_OP_READ = (__force blk_opf_t)0,328 /* write sectors to the device */329 REQ_OP_WRITE = (__force blk_opf_t)1,330 /* flush the volatile write cache */331 REQ_OP_FLUSH = (__force blk_opf_t)2,332 /* discard sectors */333 REQ_OP_DISCARD = (__force blk_opf_t)3,334 /* securely erase sectors */335 REQ_OP_SECURE_ERASE = (__force blk_opf_t)5,336 /* write data at the current zone write pointer */337 REQ_OP_ZONE_APPEND = (__force blk_opf_t)7,338 /* write the zero filled sector many times */339 REQ_OP_WRITE_ZEROES = (__force blk_opf_t)9,340 /* Open a zone */341 REQ_OP_ZONE_OPEN = (__force blk_opf_t)10,342 /* Close a zone */343 REQ_OP_ZONE_CLOSE = (__force blk_opf_t)11,344 /* Transition a zone to full */345 REQ_OP_ZONE_FINISH = (__force blk_opf_t)12,346 /* reset a zone write pointer */347 REQ_OP_ZONE_RESET = (__force blk_opf_t)13,348 /* reset all the zone present on the device */349 REQ_OP_ZONE_RESET_ALL = (__force blk_opf_t)15,350 351 /* Driver private requests */352 REQ_OP_DRV_IN = (__force blk_opf_t)34,353 REQ_OP_DRV_OUT = (__force blk_opf_t)35,354 355 REQ_OP_LAST = (__force blk_opf_t)36,356};357 358/* Keep cmd_flag_name[] in sync with the definitions below */359enum req_flag_bits {360 __REQ_FAILFAST_DEV = /* no driver retries of device errors */361 REQ_OP_BITS,362 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */363 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */364 __REQ_SYNC, /* request is sync (sync write or read) */365 __REQ_META, /* metadata io request */366 __REQ_PRIO, /* boost priority in cfq */367 __REQ_NOMERGE, /* don't touch this for merging */368 __REQ_IDLE, /* anticipate more IO after this one */369 __REQ_INTEGRITY, /* I/O includes block integrity payload */370 __REQ_FUA, /* forced unit access */371 __REQ_PREFLUSH, /* request for cache flush */372 __REQ_RAHEAD, /* read ahead, can fail anytime */373 __REQ_BACKGROUND, /* background IO */374 __REQ_NOWAIT, /* Don't wait if request will block */375 __REQ_POLLED, /* caller polls for completion using bio_poll */376 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */377 __REQ_SWAP, /* swap I/O */378 __REQ_DRV, /* for driver use */379 __REQ_FS_PRIVATE, /* for file system (submitter) use */380 __REQ_ATOMIC, /* for atomic write operations */381 /*382 * Command specific flags, keep last:383 */384 /* for REQ_OP_WRITE_ZEROES: */385 __REQ_NOUNMAP, /* do not free blocks when zeroing */386 387 __REQ_NR_BITS, /* stops here */388};389 390#define REQ_FAILFAST_DEV \391 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV)392#define REQ_FAILFAST_TRANSPORT \393 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT)394#define REQ_FAILFAST_DRIVER \395 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER)396#define REQ_SYNC (__force blk_opf_t)(1ULL << __REQ_SYNC)397#define REQ_META (__force blk_opf_t)(1ULL << __REQ_META)398#define REQ_PRIO (__force blk_opf_t)(1ULL << __REQ_PRIO)399#define REQ_NOMERGE (__force blk_opf_t)(1ULL << __REQ_NOMERGE)400#define REQ_IDLE (__force blk_opf_t)(1ULL << __REQ_IDLE)401#define REQ_INTEGRITY (__force blk_opf_t)(1ULL << __REQ_INTEGRITY)402#define REQ_FUA (__force blk_opf_t)(1ULL << __REQ_FUA)403#define REQ_PREFLUSH (__force blk_opf_t)(1ULL << __REQ_PREFLUSH)404#define REQ_RAHEAD (__force blk_opf_t)(1ULL << __REQ_RAHEAD)405#define REQ_BACKGROUND (__force blk_opf_t)(1ULL << __REQ_BACKGROUND)406#define REQ_NOWAIT (__force blk_opf_t)(1ULL << __REQ_NOWAIT)407#define REQ_POLLED (__force blk_opf_t)(1ULL << __REQ_POLLED)408#define REQ_ALLOC_CACHE (__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE)409#define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP)410#define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV)411#define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE)412#define REQ_ATOMIC (__force blk_opf_t)(1ULL << __REQ_ATOMIC)413 414#define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP)415 416#define REQ_FAILFAST_MASK \417 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)418 419#define REQ_NOMERGE_FLAGS \420 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)421 422enum stat_group {423 STAT_READ,424 STAT_WRITE,425 STAT_DISCARD,426 STAT_FLUSH,427 428 NR_STAT_GROUPS429};430 431static inline enum req_op bio_op(const struct bio *bio)432{433 return bio->bi_opf & REQ_OP_MASK;434}435 436static inline bool op_is_write(blk_opf_t op)437{438 return !!(op & (__force blk_opf_t)1);439}440 441/*442 * Check if the bio or request is one that needs special treatment in the443 * flush state machine.444 */445static inline bool op_is_flush(blk_opf_t op)446{447 return op & (REQ_FUA | REQ_PREFLUSH);448}449 450/*451 * Reads are always treated as synchronous, as are requests with the FUA or452 * PREFLUSH flag. Other operations may be marked as synchronous using the453 * REQ_SYNC flag.454 */455static inline bool op_is_sync(blk_opf_t op)456{457 return (op & REQ_OP_MASK) == REQ_OP_READ ||458 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));459}460 461static inline bool op_is_discard(blk_opf_t op)462{463 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;464}465 466/*467 * Check if a bio or request operation is a zone management operation, with468 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case469 * due to its different handling in the block layer and device response in470 * case of command failure.471 */472static inline bool op_is_zone_mgmt(enum req_op op)473{474 switch (op & REQ_OP_MASK) {475 case REQ_OP_ZONE_RESET:476 case REQ_OP_ZONE_OPEN:477 case REQ_OP_ZONE_CLOSE:478 case REQ_OP_ZONE_FINISH:479 return true;480 default:481 return false;482 }483}484 485static inline int op_stat_group(enum req_op op)486{487 if (op_is_discard(op))488 return STAT_DISCARD;489 return op_is_write(op);490}491 492struct blk_rq_stat {493 u64 mean;494 u64 min;495 u64 max;496 u32 nr_samples;497 u64 batch;498};499 500#endif /* __LINUX_BLK_TYPES_H */501