1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Block data types and constants. Directly include this file only to
4 * break include dependency loop.
6 #ifndef __LINUX_BLK_TYPES_H
7 #define __LINUX_BLK_TYPES_H
9 #include <linux/types.h>
10 #include <linux/bvec.h>
11 #include <linux/device.h>
12 #include <linux/ktime.h>
16 struct bio_integrity_payload;
19 struct cgroup_subsys_state;
20 typedef void (bio_end_io_t) (struct bio *);
24 * The basic unit of block I/O is a sector. It is used in a number of contexts
25 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
26 * bytes. Variables of type sector_t represent an offset or size that is a
27 * multiple of 512 bytes. Hence these two constants.
30 #define SECTOR_SHIFT 9
33 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
36 #define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
37 #define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
38 #define SECTOR_MASK (PAGE_SECTORS - 1)
41 sector_t bd_start_sect;
42 sector_t bd_nr_sectors;
43 struct disk_stats __percpu *bd_stats;
44 unsigned long bd_stamp;
45 bool bd_read_only; /* read-only policy */
48 struct inode * bd_inode; /* will die */
49 struct super_block * bd_super;
51 struct device bd_device;
55 struct kobject *bd_holder_dir;
57 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
58 struct gendisk * bd_disk;
59 struct request_queue * bd_queue;
61 /* The counter of freeze processes */
62 int bd_fsfreeze_count;
63 /* Mutex for freeze */
64 struct mutex bd_fsfreeze_mutex;
65 struct super_block *bd_fsfreeze_sb;
67 struct partition_meta_info *bd_meta_info;
68 #ifdef CONFIG_FAIL_MAKE_REQUEST
73 #define bdev_whole(_bdev) \
74 ((_bdev)->bd_disk->part0)
76 #define dev_to_bdev(device) \
77 container_of((device), struct block_device, bd_device)
79 #define bdev_kobj(_bdev) \
80 (&((_bdev)->bd_device.kobj))
83 * Block error status values. See block/blk-core:blk_errors for the details.
84 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
86 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
87 typedef u32 __bitwise blk_status_t;
88 typedef u32 blk_short_t;
90 typedef u8 __bitwise blk_status_t;
91 typedef u16 blk_short_t;
94 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
95 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
96 #define BLK_STS_NOSPC ((__force blk_status_t)3)
97 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
98 #define BLK_STS_TARGET ((__force blk_status_t)5)
99 #define BLK_STS_NEXUS ((__force blk_status_t)6)
100 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
101 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
102 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
103 #define BLK_STS_IOERR ((__force blk_status_t)10)
105 /* hack for device mapper, don't use elsewhere: */
106 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
109 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set
110 * and the bio would block (cf bio_wouldblock_error())
112 #define BLK_STS_AGAIN ((__force blk_status_t)12)
115 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
116 * device related resources are unavailable, but the driver can guarantee
117 * that the queue will be rerun in the future once resources become
118 * available again. This is typically the case for device specific
119 * resources that are consumed for IO. If the driver fails allocating these
120 * resources, we know that inflight (or pending) IO will free these
121 * resource upon completion.
123 * This is different from BLK_STS_RESOURCE in that it explicitly references
124 * a device specific resource. For resources of wider scope, allocation
125 * failure can happen without having pending IO. This means that we can't
126 * rely on request completions freeing these resources, as IO may not be in
127 * flight. Examples of that are kernel memory allocations, DMA mappings, or
128 * any other system wide resources.
130 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
133 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
134 * related resources are unavailable, but the driver can guarantee the queue
135 * will be rerun in the future once the resources become available again.
137 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
138 * a zone specific resource and IO to a different zone on the same device could
139 * still be served. Examples of that are zones that are write-locked, but a read
140 * to the same zone could be served.
142 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
145 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
146 * path if the device returns a status indicating that too many zone resources
147 * are currently open. The same command should be successful if resubmitted
148 * after the number of open zones decreases below the device's limits, which is
149 * reported in the request_queue's max_open_zones.
151 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
154 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
155 * path if the device returns a status indicating that too many zone resources
156 * are currently active. The same command should be successful if resubmitted
157 * after the number of active zones decreases below the device's limits, which
158 * is reported in the request_queue's max_active_zones.
160 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
163 * BLK_STS_OFFLINE is returned from the driver when the target device is offline
164 * or is being taken offline. This could help differentiate the case where a
165 * device is intentionally being shut down from a real I/O error.
167 #define BLK_STS_OFFLINE ((__force blk_status_t)17)
170 * blk_path_error - returns true if error may be path related
171 * @error: status the request was completed with
174 * This classifies block error status into non-retryable errors and ones
175 * that may be successful if retried on a failover path.
178 * %false - retrying failover path will not help
179 * %true - may succeed if retried
181 static inline bool blk_path_error(blk_status_t error)
184 case BLK_STS_NOTSUPP:
189 case BLK_STS_PROTECTION:
193 /* Anything else could be a path failure, so should be retried */
198 * From most significant bit:
199 * 1 bit: reserved for other usage, see below
200 * 12 bits: original size of bio
201 * 51 bits: issue time of bio
203 #define BIO_ISSUE_RES_BITS 1
204 #define BIO_ISSUE_SIZE_BITS 12
205 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
206 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
207 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
208 #define BIO_ISSUE_SIZE_MASK \
209 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
210 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
212 /* Reserved bit for blk-throtl */
213 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
219 static inline u64 __bio_issue_time(u64 time)
221 return time & BIO_ISSUE_TIME_MASK;
224 static inline u64 bio_issue_time(struct bio_issue *issue)
226 return __bio_issue_time(issue->value);
229 static inline sector_t bio_issue_size(struct bio_issue *issue)
231 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
234 static inline void bio_issue_init(struct bio_issue *issue,
237 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
238 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
239 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
240 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
243 typedef unsigned int blk_qc_t;
244 #define BLK_QC_T_NONE -1U
247 * main unit of I/O for the block layer and lower layers (ie drivers and
251 struct bio *bi_next; /* request queue link */
252 struct block_device *bi_bdev;
253 unsigned int bi_opf; /* bottom bits REQ_OP, top bits
256 unsigned short bi_flags; /* BIO_* below */
257 unsigned short bi_ioprio;
258 blk_status_t bi_status;
259 atomic_t __bi_remaining;
261 struct bvec_iter bi_iter;
264 bio_end_io_t *bi_end_io;
266 #ifdef CONFIG_BLK_CGROUP
268 * Represents the association of the css and request_queue for the bio.
269 * If a bio goes direct to device, it will not have a blkg as it will
270 * not have a request_queue associated with it. The reference is put
271 * on release of the bio.
273 struct blkcg_gq *bi_blkg;
274 struct bio_issue bi_issue;
275 #ifdef CONFIG_BLK_CGROUP_IOCOST
280 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
281 struct bio_crypt_ctx *bi_crypt_context;
285 #if defined(CONFIG_BLK_DEV_INTEGRITY)
286 struct bio_integrity_payload *bi_integrity; /* data integrity */
290 unsigned short bi_vcnt; /* how many bio_vec's */
293 * Everything starting with bi_max_vecs will be preserved by bio_reset()
296 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
298 atomic_t __bi_cnt; /* pin count */
300 struct bio_vec *bi_io_vec; /* the actual vec list */
302 struct bio_set *bi_pool;
305 * We can inline a number of vecs at the end of the bio, to avoid
306 * double allocations for a small number of bio_vecs. This member
307 * MUST obviously be kept at the very end of the bio.
309 struct bio_vec bi_inline_vecs[];
312 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
313 #define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT)
319 BIO_NO_PAGE_REF, /* don't put release vec pages */
320 BIO_CLONED, /* doesn't own data */
321 BIO_BOUNCED, /* bio is a bounce bio */
322 BIO_WORKINGSET, /* contains userspace workingset pages */
323 BIO_QUIET, /* Make BIO Quiet */
324 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
325 BIO_REFFED, /* bio has elevated ->bi_cnt */
326 BIO_THROTTLED, /* This bio has already been subjected to
327 * throttling rules. Don't do it again. */
328 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
330 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
331 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */
332 BIO_QOS_MERGED, /* but went through rq_qos merge path */
334 BIO_ZONE_WRITE_LOCKED, /* Owns a zoned device zone write lock */
338 typedef __u32 __bitwise blk_mq_req_flags_t;
341 * Operations and flags common to the bio and request structures.
342 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
344 * The least significant bit of the operation number indicates the data
345 * transfer direction:
347 * - if the least significant bit is set transfers are TO the device
348 * - if the least significant bit is not set transfers are FROM the device
350 * If a operation does not transfer data the least significant bit has no
353 #define REQ_OP_BITS 8
354 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
355 #define REQ_FLAG_BITS 24
358 /* read sectors from the device */
360 /* write sectors to the device */
362 /* flush the volatile write cache */
364 /* discard sectors */
366 /* securely erase sectors */
367 REQ_OP_SECURE_ERASE = 5,
368 /* write the zero filled sector many times */
369 REQ_OP_WRITE_ZEROES = 9,
371 REQ_OP_ZONE_OPEN = 10,
373 REQ_OP_ZONE_CLOSE = 11,
374 /* Transition a zone to full */
375 REQ_OP_ZONE_FINISH = 12,
376 /* write data at the current zone write pointer */
377 REQ_OP_ZONE_APPEND = 13,
378 /* reset a zone write pointer */
379 REQ_OP_ZONE_RESET = 15,
380 /* reset all the zone present on the device */
381 REQ_OP_ZONE_RESET_ALL = 17,
383 /* Driver private requests */
391 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
393 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
394 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
395 __REQ_SYNC, /* request is sync (sync write or read) */
396 __REQ_META, /* metadata io request */
397 __REQ_PRIO, /* boost priority in cfq */
398 __REQ_NOMERGE, /* don't touch this for merging */
399 __REQ_IDLE, /* anticipate more IO after this one */
400 __REQ_INTEGRITY, /* I/O includes block integrity payload */
401 __REQ_FUA, /* forced unit access */
402 __REQ_PREFLUSH, /* request for cache flush */
403 __REQ_RAHEAD, /* read ahead, can fail anytime */
404 __REQ_BACKGROUND, /* background IO */
405 __REQ_NOWAIT, /* Don't wait if request will block */
407 * When a shared kthread needs to issue a bio for a cgroup, doing
408 * so synchronously can lead to priority inversions as the kthread
409 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
410 * submit_bio() punt the actual issuing to a dedicated per-blkcg
411 * work item to avoid such priority inversions.
414 __REQ_POLLED, /* caller polls for completion using bio_poll */
415 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */
416 __REQ_SWAP, /* swap I/O */
417 __REQ_DRV, /* for driver use */
420 * Command specific flags, keep last:
422 /* for REQ_OP_WRITE_ZEROES: */
423 __REQ_NOUNMAP, /* do not free blocks when zeroing */
425 __REQ_NR_BITS, /* stops here */
428 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
429 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
430 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
431 #define REQ_SYNC (1ULL << __REQ_SYNC)
432 #define REQ_META (1ULL << __REQ_META)
433 #define REQ_PRIO (1ULL << __REQ_PRIO)
434 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
435 #define REQ_IDLE (1ULL << __REQ_IDLE)
436 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
437 #define REQ_FUA (1ULL << __REQ_FUA)
438 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
439 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
440 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
441 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
442 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
444 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
445 #define REQ_POLLED (1ULL << __REQ_POLLED)
446 #define REQ_ALLOC_CACHE (1ULL << __REQ_ALLOC_CACHE)
448 #define REQ_DRV (1ULL << __REQ_DRV)
449 #define REQ_SWAP (1ULL << __REQ_SWAP)
451 #define REQ_FAILFAST_MASK \
452 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
454 #define REQ_NOMERGE_FLAGS \
455 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
466 #define bio_op(bio) \
467 ((bio)->bi_opf & REQ_OP_MASK)
469 /* obsolete, don't use in new code */
470 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
473 bio->bi_opf = op | op_flags;
476 static inline bool op_is_write(unsigned int op)
482 * Check if the bio or request is one that needs special treatment in the
483 * flush state machine.
485 static inline bool op_is_flush(unsigned int op)
487 return op & (REQ_FUA | REQ_PREFLUSH);
491 * Reads are always treated as synchronous, as are requests with the FUA or
492 * PREFLUSH flag. Other operations may be marked as synchronous using the
495 static inline bool op_is_sync(unsigned int op)
497 return (op & REQ_OP_MASK) == REQ_OP_READ ||
498 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
501 static inline bool op_is_discard(unsigned int op)
503 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
507 * Check if a bio or request operation is a zone management operation, with
508 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
509 * due to its different handling in the block layer and device response in
510 * case of command failure.
512 static inline bool op_is_zone_mgmt(enum req_opf op)
514 switch (op & REQ_OP_MASK) {
515 case REQ_OP_ZONE_RESET:
516 case REQ_OP_ZONE_OPEN:
517 case REQ_OP_ZONE_CLOSE:
518 case REQ_OP_ZONE_FINISH:
525 static inline int op_stat_group(unsigned int op)
527 if (op_is_discard(op))
529 return op_is_write(op);
540 #endif /* __LINUX_BLK_TYPES_H */