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/ktime.h>
15 struct bio_integrity_payload;
18 struct cgroup_subsys_state;
19 typedef void (bio_end_io_t) (struct bio *);
23 sector_t bd_start_sect;
24 struct disk_stats __percpu *bd_stats;
25 unsigned long bd_stamp;
26 bool bd_read_only; /* read-only policy */
29 struct inode * bd_inode; /* will die */
30 struct super_block * bd_super;
31 struct mutex bd_mutex; /* open/close mutex */
37 struct list_head bd_holder_disks;
39 struct kobject *bd_holder_dir;
41 struct hd_struct * bd_part;
42 /* number of times partitions within this device have been opened. */
43 unsigned bd_part_count;
45 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
46 struct gendisk * bd_disk;
47 struct backing_dev_info *bd_bdi;
49 /* The counter of freeze processes */
50 int bd_fsfreeze_count;
51 /* Mutex for freeze */
52 struct mutex bd_fsfreeze_mutex;
53 struct super_block *bd_fsfreeze_sb;
55 struct partition_meta_info *bd_meta_info;
56 #ifdef CONFIG_FAIL_MAKE_REQUEST
61 #define bdev_whole(_bdev) \
62 ((_bdev)->bd_disk->part0)
64 #define bdev_kobj(_bdev) \
65 (&part_to_dev((_bdev)->bd_part)->kobj)
68 * Block error status values. See block/blk-core:blk_errors for the details.
69 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
71 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
72 typedef u32 __bitwise blk_status_t;
74 typedef u8 __bitwise blk_status_t;
77 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
78 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
79 #define BLK_STS_NOSPC ((__force blk_status_t)3)
80 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
81 #define BLK_STS_TARGET ((__force blk_status_t)5)
82 #define BLK_STS_NEXUS ((__force blk_status_t)6)
83 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
84 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
85 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
86 #define BLK_STS_IOERR ((__force blk_status_t)10)
88 /* hack for device mapper, don't use elsewhere: */
89 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
91 #define BLK_STS_AGAIN ((__force blk_status_t)12)
94 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
95 * device related resources are unavailable, but the driver can guarantee
96 * that the queue will be rerun in the future once resources become
97 * available again. This is typically the case for device specific
98 * resources that are consumed for IO. If the driver fails allocating these
99 * resources, we know that inflight (or pending) IO will free these
100 * resource upon completion.
102 * This is different from BLK_STS_RESOURCE in that it explicitly references
103 * a device specific resource. For resources of wider scope, allocation
104 * failure can happen without having pending IO. This means that we can't
105 * rely on request completions freeing these resources, as IO may not be in
106 * flight. Examples of that are kernel memory allocations, DMA mappings, or
107 * any other system wide resources.
109 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
112 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
113 * related resources are unavailable, but the driver can guarantee the queue
114 * will be rerun in the future once the resources become available again.
116 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
117 * a zone specific resource and IO to a different zone on the same device could
118 * still be served. Examples of that are zones that are write-locked, but a read
119 * to the same zone could be served.
121 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
124 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
125 * path if the device returns a status indicating that too many zone resources
126 * are currently open. The same command should be successful if resubmitted
127 * after the number of open zones decreases below the device's limits, which is
128 * reported in the request_queue's max_open_zones.
130 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
133 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
134 * path if the device returns a status indicating that too many zone resources
135 * are currently active. The same command should be successful if resubmitted
136 * after the number of active zones decreases below the device's limits, which
137 * is reported in the request_queue's max_active_zones.
139 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
142 * blk_path_error - returns true if error may be path related
143 * @error: status the request was completed with
146 * This classifies block error status into non-retryable errors and ones
147 * that may be successful if retried on a failover path.
150 * %false - retrying failover path will not help
151 * %true - may succeed if retried
153 static inline bool blk_path_error(blk_status_t error)
156 case BLK_STS_NOTSUPP:
161 case BLK_STS_PROTECTION:
165 /* Anything else could be a path failure, so should be retried */
170 * From most significant bit:
171 * 1 bit: reserved for other usage, see below
172 * 12 bits: original size of bio
173 * 51 bits: issue time of bio
175 #define BIO_ISSUE_RES_BITS 1
176 #define BIO_ISSUE_SIZE_BITS 12
177 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
178 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
179 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
180 #define BIO_ISSUE_SIZE_MASK \
181 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
182 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
184 /* Reserved bit for blk-throtl */
185 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
191 static inline u64 __bio_issue_time(u64 time)
193 return time & BIO_ISSUE_TIME_MASK;
196 static inline u64 bio_issue_time(struct bio_issue *issue)
198 return __bio_issue_time(issue->value);
201 static inline sector_t bio_issue_size(struct bio_issue *issue)
203 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
206 static inline void bio_issue_init(struct bio_issue *issue,
209 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
210 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
211 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
212 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
216 * main unit of I/O for the block layer and lower layers (ie drivers and
220 struct bio *bi_next; /* request queue link */
221 struct gendisk *bi_disk;
222 unsigned int bi_opf; /* bottom bits req flags,
223 * top bits REQ_OP. Use
226 unsigned short bi_flags; /* status, etc and bvec pool number */
227 unsigned short bi_ioprio;
228 unsigned short bi_write_hint;
229 blk_status_t bi_status;
231 atomic_t __bi_remaining;
233 struct bvec_iter bi_iter;
235 bio_end_io_t *bi_end_io;
238 #ifdef CONFIG_BLK_CGROUP
240 * Represents the association of the css and request_queue for the bio.
241 * If a bio goes direct to device, it will not have a blkg as it will
242 * not have a request_queue associated with it. The reference is put
243 * on release of the bio.
245 struct blkcg_gq *bi_blkg;
246 struct bio_issue bi_issue;
247 #ifdef CONFIG_BLK_CGROUP_IOCOST
252 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
253 struct bio_crypt_ctx *bi_crypt_context;
257 #if defined(CONFIG_BLK_DEV_INTEGRITY)
258 struct bio_integrity_payload *bi_integrity; /* data integrity */
262 unsigned short bi_vcnt; /* how many bio_vec's */
265 * Everything starting with bi_max_vecs will be preserved by bio_reset()
268 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
270 atomic_t __bi_cnt; /* pin count */
272 struct bio_vec *bi_io_vec; /* the actual vec list */
274 struct bio_set *bi_pool;
277 * We can inline a number of vecs at the end of the bio, to avoid
278 * double allocations for a small number of bio_vecs. This member
279 * MUST obviously be kept at the very end of the bio.
281 struct bio_vec bi_inline_vecs[];
284 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
290 BIO_NO_PAGE_REF, /* don't put release vec pages */
291 BIO_CLONED, /* doesn't own data */
292 BIO_BOUNCED, /* bio is a bounce bio */
293 BIO_WORKINGSET, /* contains userspace workingset pages */
294 BIO_QUIET, /* Make BIO Quiet */
295 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
296 BIO_REFFED, /* bio has elevated ->bi_cnt */
297 BIO_THROTTLED, /* This bio has already been subjected to
298 * throttling rules. Don't do it again. */
299 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
301 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
302 BIO_TRACKED, /* set if bio goes through the rq_qos path */
306 /* See BVEC_POOL_OFFSET below before adding new flags */
309 * We support 6 different bvec pools, the last one is magic in that it
310 * is backed by a mempool.
312 #define BVEC_POOL_NR 6
313 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
316 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
317 * 1 to the actual index so that 0 indicates that there are no bvecs to be
320 #define BVEC_POOL_BITS (3)
321 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
322 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
323 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
324 # error "BVEC_POOL_BITS is too small"
328 * Flags starting here get preserved by bio_reset() - this includes
329 * only BVEC_POOL_IDX()
331 #define BIO_RESET_BITS BVEC_POOL_OFFSET
333 typedef __u32 __bitwise blk_mq_req_flags_t;
336 * Operations and flags common to the bio and request structures.
337 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
339 * The least significant bit of the operation number indicates the data
340 * transfer direction:
342 * - if the least significant bit is set transfers are TO the device
343 * - if the least significant bit is not set transfers are FROM the device
345 * If a operation does not transfer data the least significant bit has no
348 #define REQ_OP_BITS 8
349 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
350 #define REQ_FLAG_BITS 24
353 /* read sectors from the device */
355 /* write sectors to the device */
357 /* flush the volatile write cache */
359 /* discard sectors */
361 /* securely erase sectors */
362 REQ_OP_SECURE_ERASE = 5,
363 /* write the same sector many times */
364 REQ_OP_WRITE_SAME = 7,
365 /* write the zero filled sector many times */
366 REQ_OP_WRITE_ZEROES = 9,
368 REQ_OP_ZONE_OPEN = 10,
370 REQ_OP_ZONE_CLOSE = 11,
371 /* Transition a zone to full */
372 REQ_OP_ZONE_FINISH = 12,
373 /* write data at the current zone write pointer */
374 REQ_OP_ZONE_APPEND = 13,
375 /* reset a zone write pointer */
376 REQ_OP_ZONE_RESET = 15,
377 /* reset all the zone present on the device */
378 REQ_OP_ZONE_RESET_ALL = 17,
380 /* SCSI passthrough using struct scsi_request */
382 REQ_OP_SCSI_OUT = 33,
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.
415 /* command specific flags for REQ_OP_WRITE_ZEROES: */
416 __REQ_NOUNMAP, /* do not free blocks when zeroing */
422 __REQ_SWAP, /* swapping request. */
423 __REQ_NR_BITS, /* stops here */
426 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
427 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
428 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
429 #define REQ_SYNC (1ULL << __REQ_SYNC)
430 #define REQ_META (1ULL << __REQ_META)
431 #define REQ_PRIO (1ULL << __REQ_PRIO)
432 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
433 #define REQ_IDLE (1ULL << __REQ_IDLE)
434 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
435 #define REQ_FUA (1ULL << __REQ_FUA)
436 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
437 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
438 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
439 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
440 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
442 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
443 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
445 #define REQ_DRV (1ULL << __REQ_DRV)
446 #define REQ_SWAP (1ULL << __REQ_SWAP)
448 #define REQ_FAILFAST_MASK \
449 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
451 #define REQ_NOMERGE_FLAGS \
452 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
463 #define bio_op(bio) \
464 ((bio)->bi_opf & REQ_OP_MASK)
465 #define req_op(req) \
466 ((req)->cmd_flags & REQ_OP_MASK)
468 /* obsolete, don't use in new code */
469 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
472 bio->bi_opf = op | op_flags;
475 static inline bool op_is_write(unsigned int op)
481 * Check if the bio or request is one that needs special treatment in the
482 * flush state machine.
484 static inline bool op_is_flush(unsigned int op)
486 return op & (REQ_FUA | REQ_PREFLUSH);
490 * Reads are always treated as synchronous, as are requests with the FUA or
491 * PREFLUSH flag. Other operations may be marked as synchronous using the
494 static inline bool op_is_sync(unsigned int op)
496 return (op & REQ_OP_MASK) == REQ_OP_READ ||
497 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
500 static inline bool op_is_discard(unsigned int op)
502 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
506 * Check if a bio or request operation is a zone management operation, with
507 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
508 * due to its different handling in the block layer and device response in
509 * case of command failure.
511 static inline bool op_is_zone_mgmt(enum req_opf op)
513 switch (op & REQ_OP_MASK) {
514 case REQ_OP_ZONE_RESET:
515 case REQ_OP_ZONE_OPEN:
516 case REQ_OP_ZONE_CLOSE:
517 case REQ_OP_ZONE_FINISH:
524 static inline int op_stat_group(unsigned int op)
526 if (op_is_discard(op))
528 return op_is_write(op);
531 typedef unsigned int blk_qc_t;
532 #define BLK_QC_T_NONE -1U
533 #define BLK_QC_T_SHIFT 16
534 #define BLK_QC_T_INTERNAL (1U << 31)
536 static inline bool blk_qc_t_valid(blk_qc_t cookie)
538 return cookie != BLK_QC_T_NONE;
541 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
543 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
546 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
548 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
551 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
553 return (cookie & BLK_QC_T_INTERNAL) != 0;
564 #endif /* __LINUX_BLK_TYPES_H */