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 *);
25 struct inode * bd_inode; /* will die */
26 struct super_block * bd_super;
27 struct mutex bd_mutex; /* open/close mutex */
33 struct list_head bd_holder_disks;
36 struct hd_struct * bd_part;
37 /* number of times partitions within this device have been opened. */
38 unsigned bd_part_count;
40 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
41 struct gendisk * bd_disk;
42 struct backing_dev_info *bd_bdi;
44 /* The counter of freeze processes */
45 int bd_fsfreeze_count;
46 /* Mutex for freeze */
47 struct mutex bd_fsfreeze_mutex;
48 struct super_block *bd_fsfreeze_sb;
51 #define bdev_whole(_bdev) \
52 ((_bdev)->bd_disk->part0.bdev)
54 #define bdev_kobj(_bdev) \
55 (&part_to_dev((_bdev)->bd_part)->kobj)
58 * Block error status values. See block/blk-core:blk_errors for the details.
59 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
61 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
62 typedef u32 __bitwise blk_status_t;
64 typedef u8 __bitwise blk_status_t;
67 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
68 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
69 #define BLK_STS_NOSPC ((__force blk_status_t)3)
70 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
71 #define BLK_STS_TARGET ((__force blk_status_t)5)
72 #define BLK_STS_NEXUS ((__force blk_status_t)6)
73 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
74 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
75 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
76 #define BLK_STS_IOERR ((__force blk_status_t)10)
78 /* hack for device mapper, don't use elsewhere: */
79 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
81 #define BLK_STS_AGAIN ((__force blk_status_t)12)
84 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
85 * device related resources are unavailable, but the driver can guarantee
86 * that the queue will be rerun in the future once resources become
87 * available again. This is typically the case for device specific
88 * resources that are consumed for IO. If the driver fails allocating these
89 * resources, we know that inflight (or pending) IO will free these
90 * resource upon completion.
92 * This is different from BLK_STS_RESOURCE in that it explicitly references
93 * a device specific resource. For resources of wider scope, allocation
94 * failure can happen without having pending IO. This means that we can't
95 * rely on request completions freeing these resources, as IO may not be in
96 * flight. Examples of that are kernel memory allocations, DMA mappings, or
97 * any other system wide resources.
99 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
102 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
103 * related resources are unavailable, but the driver can guarantee the queue
104 * will be rerun in the future once the resources become available again.
106 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
107 * a zone specific resource and IO to a different zone on the same device could
108 * still be served. Examples of that are zones that are write-locked, but a read
109 * to the same zone could be served.
111 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
114 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
115 * path if the device returns a status indicating that too many zone resources
116 * are currently open. The same command should be successful if resubmitted
117 * after the number of open zones decreases below the device's limits, which is
118 * reported in the request_queue's max_open_zones.
120 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
123 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
124 * path if the device returns a status indicating that too many zone resources
125 * are currently active. The same command should be successful if resubmitted
126 * after the number of active zones decreases below the device's limits, which
127 * is reported in the request_queue's max_active_zones.
129 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
132 * blk_path_error - returns true if error may be path related
133 * @error: status the request was completed with
136 * This classifies block error status into non-retryable errors and ones
137 * that may be successful if retried on a failover path.
140 * %false - retrying failover path will not help
141 * %true - may succeed if retried
143 static inline bool blk_path_error(blk_status_t error)
146 case BLK_STS_NOTSUPP:
151 case BLK_STS_PROTECTION:
155 /* Anything else could be a path failure, so should be retried */
160 * From most significant bit:
161 * 1 bit: reserved for other usage, see below
162 * 12 bits: original size of bio
163 * 51 bits: issue time of bio
165 #define BIO_ISSUE_RES_BITS 1
166 #define BIO_ISSUE_SIZE_BITS 12
167 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
168 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
169 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
170 #define BIO_ISSUE_SIZE_MASK \
171 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
172 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
174 /* Reserved bit for blk-throtl */
175 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
181 static inline u64 __bio_issue_time(u64 time)
183 return time & BIO_ISSUE_TIME_MASK;
186 static inline u64 bio_issue_time(struct bio_issue *issue)
188 return __bio_issue_time(issue->value);
191 static inline sector_t bio_issue_size(struct bio_issue *issue)
193 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
196 static inline void bio_issue_init(struct bio_issue *issue,
199 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
200 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
201 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
202 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
206 * main unit of I/O for the block layer and lower layers (ie drivers and
210 struct bio *bi_next; /* request queue link */
211 struct gendisk *bi_disk;
212 unsigned int bi_opf; /* bottom bits req flags,
213 * top bits REQ_OP. Use
216 unsigned short bi_flags; /* status, etc and bvec pool number */
217 unsigned short bi_ioprio;
218 unsigned short bi_write_hint;
219 blk_status_t bi_status;
221 atomic_t __bi_remaining;
223 struct bvec_iter bi_iter;
225 bio_end_io_t *bi_end_io;
228 #ifdef CONFIG_BLK_CGROUP
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.
235 struct blkcg_gq *bi_blkg;
236 struct bio_issue bi_issue;
237 #ifdef CONFIG_BLK_CGROUP_IOCOST
242 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
243 struct bio_crypt_ctx *bi_crypt_context;
247 #if defined(CONFIG_BLK_DEV_INTEGRITY)
248 struct bio_integrity_payload *bi_integrity; /* data integrity */
252 unsigned short bi_vcnt; /* how many bio_vec's */
255 * Everything starting with bi_max_vecs will be preserved by bio_reset()
258 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
260 atomic_t __bi_cnt; /* pin count */
262 struct bio_vec *bi_io_vec; /* the actual vec list */
264 struct bio_set *bi_pool;
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.
271 struct bio_vec bi_inline_vecs[];
274 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
280 BIO_NO_PAGE_REF, /* don't put release vec pages */
281 BIO_CLONED, /* doesn't own data */
282 BIO_BOUNCED, /* bio is a bounce bio */
283 BIO_WORKINGSET, /* contains userspace workingset pages */
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_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
291 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
292 BIO_TRACKED, /* set if bio goes through the rq_qos path */
296 /* See BVEC_POOL_OFFSET below before adding new flags */
299 * We support 6 different bvec pools, the last one is magic in that it
300 * is backed by a mempool.
302 #define BVEC_POOL_NR 6
303 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
306 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
307 * 1 to the actual index so that 0 indicates that there are no bvecs to be
310 #define BVEC_POOL_BITS (3)
311 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
312 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
313 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
314 # error "BVEC_POOL_BITS is too small"
318 * Flags starting here get preserved by bio_reset() - this includes
319 * only BVEC_POOL_IDX()
321 #define BIO_RESET_BITS BVEC_POOL_OFFSET
323 typedef __u32 __bitwise blk_mq_req_flags_t;
326 * Operations and flags common to the bio and request structures.
327 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
329 * The least significant bit of the operation number indicates the data
330 * transfer direction:
332 * - if the least significant bit is set transfers are TO the device
333 * - if the least significant bit is not set transfers are FROM the device
335 * If a operation does not transfer data the least significant bit has no
338 #define REQ_OP_BITS 8
339 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
340 #define REQ_FLAG_BITS 24
343 /* read sectors from the device */
345 /* write sectors to the device */
347 /* flush the volatile write cache */
349 /* discard sectors */
351 /* securely erase sectors */
352 REQ_OP_SECURE_ERASE = 5,
353 /* write the same sector many times */
354 REQ_OP_WRITE_SAME = 7,
355 /* write the zero filled sector many times */
356 REQ_OP_WRITE_ZEROES = 9,
358 REQ_OP_ZONE_OPEN = 10,
360 REQ_OP_ZONE_CLOSE = 11,
361 /* Transition a zone to full */
362 REQ_OP_ZONE_FINISH = 12,
363 /* write data at the current zone write pointer */
364 REQ_OP_ZONE_APPEND = 13,
365 /* reset a zone write pointer */
366 REQ_OP_ZONE_RESET = 15,
367 /* reset all the zone present on the device */
368 REQ_OP_ZONE_RESET_ALL = 17,
370 /* SCSI passthrough using struct scsi_request */
372 REQ_OP_SCSI_OUT = 33,
373 /* Driver private requests */
381 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
383 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
384 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
385 __REQ_SYNC, /* request is sync (sync write or read) */
386 __REQ_META, /* metadata io request */
387 __REQ_PRIO, /* boost priority in cfq */
388 __REQ_NOMERGE, /* don't touch this for merging */
389 __REQ_IDLE, /* anticipate more IO after this one */
390 __REQ_INTEGRITY, /* I/O includes block integrity payload */
391 __REQ_FUA, /* forced unit access */
392 __REQ_PREFLUSH, /* request for cache flush */
393 __REQ_RAHEAD, /* read ahead, can fail anytime */
394 __REQ_BACKGROUND, /* background IO */
395 __REQ_NOWAIT, /* Don't wait if request will block */
397 * When a shared kthread needs to issue a bio for a cgroup, doing
398 * so synchronously can lead to priority inversions as the kthread
399 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
400 * submit_bio() punt the actual issuing to a dedicated per-blkcg
401 * work item to avoid such priority inversions.
405 /* command specific flags for REQ_OP_WRITE_ZEROES: */
406 __REQ_NOUNMAP, /* do not free blocks when zeroing */
412 __REQ_SWAP, /* swapping request. */
413 __REQ_NR_BITS, /* stops here */
416 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
417 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
418 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
419 #define REQ_SYNC (1ULL << __REQ_SYNC)
420 #define REQ_META (1ULL << __REQ_META)
421 #define REQ_PRIO (1ULL << __REQ_PRIO)
422 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
423 #define REQ_IDLE (1ULL << __REQ_IDLE)
424 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
425 #define REQ_FUA (1ULL << __REQ_FUA)
426 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
427 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
428 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
429 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
430 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
432 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
433 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
435 #define REQ_DRV (1ULL << __REQ_DRV)
436 #define REQ_SWAP (1ULL << __REQ_SWAP)
438 #define REQ_FAILFAST_MASK \
439 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
441 #define REQ_NOMERGE_FLAGS \
442 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
453 #define bio_op(bio) \
454 ((bio)->bi_opf & REQ_OP_MASK)
455 #define req_op(req) \
456 ((req)->cmd_flags & REQ_OP_MASK)
458 /* obsolete, don't use in new code */
459 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
462 bio->bi_opf = op | op_flags;
465 static inline bool op_is_write(unsigned int op)
471 * Check if the bio or request is one that needs special treatment in the
472 * flush state machine.
474 static inline bool op_is_flush(unsigned int op)
476 return op & (REQ_FUA | REQ_PREFLUSH);
480 * Reads are always treated as synchronous, as are requests with the FUA or
481 * PREFLUSH flag. Other operations may be marked as synchronous using the
484 static inline bool op_is_sync(unsigned int op)
486 return (op & REQ_OP_MASK) == REQ_OP_READ ||
487 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
490 static inline bool op_is_discard(unsigned int op)
492 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
496 * Check if a bio or request operation is a zone management operation, with
497 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
498 * due to its different handling in the block layer and device response in
499 * case of command failure.
501 static inline bool op_is_zone_mgmt(enum req_opf op)
503 switch (op & REQ_OP_MASK) {
504 case REQ_OP_ZONE_RESET:
505 case REQ_OP_ZONE_OPEN:
506 case REQ_OP_ZONE_CLOSE:
507 case REQ_OP_ZONE_FINISH:
514 static inline int op_stat_group(unsigned int op)
516 if (op_is_discard(op))
518 return op_is_write(op);
521 typedef unsigned int blk_qc_t;
522 #define BLK_QC_T_NONE -1U
523 #define BLK_QC_T_SHIFT 16
524 #define BLK_QC_T_INTERNAL (1U << 31)
526 static inline bool blk_qc_t_valid(blk_qc_t cookie)
528 return cookie != BLK_QC_T_NONE;
531 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
533 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
536 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
538 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
541 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
543 return (cookie & BLK_QC_T_INTERNAL) != 0;
554 #endif /* __LINUX_BLK_TYPES_H */