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;
35 struct block_device * bd_contains;
37 struct hd_struct * bd_part;
38 /* number of times partitions within this device have been opened. */
39 unsigned bd_part_count;
41 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
42 struct gendisk * bd_disk;
43 struct backing_dev_info *bd_bdi;
45 /* The counter of freeze processes */
46 int bd_fsfreeze_count;
47 /* Mutex for freeze */
48 struct mutex bd_fsfreeze_mutex;
49 struct super_block *bd_fsfreeze_sb;
53 * Block error status values. See block/blk-core:blk_errors for the details.
54 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
56 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
57 typedef u32 __bitwise blk_status_t;
59 typedef u8 __bitwise blk_status_t;
62 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
63 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
64 #define BLK_STS_NOSPC ((__force blk_status_t)3)
65 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
66 #define BLK_STS_TARGET ((__force blk_status_t)5)
67 #define BLK_STS_NEXUS ((__force blk_status_t)6)
68 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
69 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
70 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
71 #define BLK_STS_IOERR ((__force blk_status_t)10)
73 /* hack for device mapper, don't use elsewhere: */
74 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
76 #define BLK_STS_AGAIN ((__force blk_status_t)12)
79 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
80 * device related resources are unavailable, but the driver can guarantee
81 * that the queue will be rerun in the future once resources become
82 * available again. This is typically the case for device specific
83 * resources that are consumed for IO. If the driver fails allocating these
84 * resources, we know that inflight (or pending) IO will free these
85 * resource upon completion.
87 * This is different from BLK_STS_RESOURCE in that it explicitly references
88 * a device specific resource. For resources of wider scope, allocation
89 * failure can happen without having pending IO. This means that we can't
90 * rely on request completions freeing these resources, as IO may not be in
91 * flight. Examples of that are kernel memory allocations, DMA mappings, or
92 * any other system wide resources.
94 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
97 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
98 * related resources are unavailable, but the driver can guarantee the queue
99 * will be rerun in the future once the resources become available again.
101 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
102 * a zone specific resource and IO to a different zone on the same device could
103 * still be served. Examples of that are zones that are write-locked, but a read
104 * to the same zone could be served.
106 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
109 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
110 * path if the device returns a status indicating that too many zone resources
111 * are currently open. The same command should be successful if resubmitted
112 * after the number of open zones decreases below the device's limits, which is
113 * reported in the request_queue's max_open_zones.
115 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
118 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
119 * path if the device returns a status indicating that too many zone resources
120 * are currently active. The same command should be successful if resubmitted
121 * after the number of active zones decreases below the device's limits, which
122 * is reported in the request_queue's max_active_zones.
124 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
127 * blk_path_error - returns true if error may be path related
128 * @error: status the request was completed with
131 * This classifies block error status into non-retryable errors and ones
132 * that may be successful if retried on a failover path.
135 * %false - retrying failover path will not help
136 * %true - may succeed if retried
138 static inline bool blk_path_error(blk_status_t error)
141 case BLK_STS_NOTSUPP:
146 case BLK_STS_PROTECTION:
150 /* Anything else could be a path failure, so should be retried */
155 * From most significant bit:
156 * 1 bit: reserved for other usage, see below
157 * 12 bits: original size of bio
158 * 51 bits: issue time of bio
160 #define BIO_ISSUE_RES_BITS 1
161 #define BIO_ISSUE_SIZE_BITS 12
162 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
163 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
164 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
165 #define BIO_ISSUE_SIZE_MASK \
166 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
167 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
169 /* Reserved bit for blk-throtl */
170 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
176 static inline u64 __bio_issue_time(u64 time)
178 return time & BIO_ISSUE_TIME_MASK;
181 static inline u64 bio_issue_time(struct bio_issue *issue)
183 return __bio_issue_time(issue->value);
186 static inline sector_t bio_issue_size(struct bio_issue *issue)
188 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
191 static inline void bio_issue_init(struct bio_issue *issue,
194 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
195 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
196 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
197 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
201 * main unit of I/O for the block layer and lower layers (ie drivers and
205 struct bio *bi_next; /* request queue link */
206 struct gendisk *bi_disk;
207 unsigned int bi_opf; /* bottom bits req flags,
208 * top bits REQ_OP. Use
211 unsigned short bi_flags; /* status, etc and bvec pool number */
212 unsigned short bi_ioprio;
213 unsigned short bi_write_hint;
214 blk_status_t bi_status;
216 atomic_t __bi_remaining;
218 struct bvec_iter bi_iter;
220 bio_end_io_t *bi_end_io;
223 #ifdef CONFIG_BLK_CGROUP
225 * Represents the association of the css and request_queue for the bio.
226 * If a bio goes direct to device, it will not have a blkg as it will
227 * not have a request_queue associated with it. The reference is put
228 * on release of the bio.
230 struct blkcg_gq *bi_blkg;
231 struct bio_issue bi_issue;
232 #ifdef CONFIG_BLK_CGROUP_IOCOST
237 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
238 struct bio_crypt_ctx *bi_crypt_context;
242 #if defined(CONFIG_BLK_DEV_INTEGRITY)
243 struct bio_integrity_payload *bi_integrity; /* data integrity */
247 unsigned short bi_vcnt; /* how many bio_vec's */
250 * Everything starting with bi_max_vecs will be preserved by bio_reset()
253 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
255 atomic_t __bi_cnt; /* pin count */
257 struct bio_vec *bi_io_vec; /* the actual vec list */
259 struct bio_set *bi_pool;
262 * We can inline a number of vecs at the end of the bio, to avoid
263 * double allocations for a small number of bio_vecs. This member
264 * MUST obviously be kept at the very end of the bio.
266 struct bio_vec bi_inline_vecs[];
269 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
275 BIO_NO_PAGE_REF, /* don't put release vec pages */
276 BIO_CLONED, /* doesn't own data */
277 BIO_BOUNCED, /* bio is a bounce bio */
278 BIO_WORKINGSET, /* contains userspace workingset pages */
279 BIO_QUIET, /* Make BIO Quiet */
280 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
281 BIO_REFFED, /* bio has elevated ->bi_cnt */
282 BIO_THROTTLED, /* This bio has already been subjected to
283 * throttling rules. Don't do it again. */
284 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
286 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
287 BIO_TRACKED, /* set if bio goes through the rq_qos path */
291 /* See BVEC_POOL_OFFSET below before adding new flags */
294 * We support 6 different bvec pools, the last one is magic in that it
295 * is backed by a mempool.
297 #define BVEC_POOL_NR 6
298 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
301 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
302 * 1 to the actual index so that 0 indicates that there are no bvecs to be
305 #define BVEC_POOL_BITS (3)
306 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
307 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
308 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
309 # error "BVEC_POOL_BITS is too small"
313 * Flags starting here get preserved by bio_reset() - this includes
314 * only BVEC_POOL_IDX()
316 #define BIO_RESET_BITS BVEC_POOL_OFFSET
318 typedef __u32 __bitwise blk_mq_req_flags_t;
321 * Operations and flags common to the bio and request structures.
322 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
324 * The least significant bit of the operation number indicates the data
325 * transfer direction:
327 * - if the least significant bit is set transfers are TO the device
328 * - if the least significant bit is not set transfers are FROM the device
330 * If a operation does not transfer data the least significant bit has no
333 #define REQ_OP_BITS 8
334 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
335 #define REQ_FLAG_BITS 24
338 /* read sectors from the device */
340 /* write sectors to the device */
342 /* flush the volatile write cache */
344 /* discard sectors */
346 /* securely erase sectors */
347 REQ_OP_SECURE_ERASE = 5,
348 /* write the same sector many times */
349 REQ_OP_WRITE_SAME = 7,
350 /* write the zero filled sector many times */
351 REQ_OP_WRITE_ZEROES = 9,
353 REQ_OP_ZONE_OPEN = 10,
355 REQ_OP_ZONE_CLOSE = 11,
356 /* Transition a zone to full */
357 REQ_OP_ZONE_FINISH = 12,
358 /* write data at the current zone write pointer */
359 REQ_OP_ZONE_APPEND = 13,
360 /* reset a zone write pointer */
361 REQ_OP_ZONE_RESET = 15,
362 /* reset all the zone present on the device */
363 REQ_OP_ZONE_RESET_ALL = 17,
365 /* SCSI passthrough using struct scsi_request */
367 REQ_OP_SCSI_OUT = 33,
368 /* Driver private requests */
376 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
378 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
379 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
380 __REQ_SYNC, /* request is sync (sync write or read) */
381 __REQ_META, /* metadata io request */
382 __REQ_PRIO, /* boost priority in cfq */
383 __REQ_NOMERGE, /* don't touch this for merging */
384 __REQ_IDLE, /* anticipate more IO after this one */
385 __REQ_INTEGRITY, /* I/O includes block integrity payload */
386 __REQ_FUA, /* forced unit access */
387 __REQ_PREFLUSH, /* request for cache flush */
388 __REQ_RAHEAD, /* read ahead, can fail anytime */
389 __REQ_BACKGROUND, /* background IO */
390 __REQ_NOWAIT, /* Don't wait if request will block */
392 * When a shared kthread needs to issue a bio for a cgroup, doing
393 * so synchronously can lead to priority inversions as the kthread
394 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
395 * submit_bio() punt the actual issuing to a dedicated per-blkcg
396 * work item to avoid such priority inversions.
400 /* command specific flags for REQ_OP_WRITE_ZEROES: */
401 __REQ_NOUNMAP, /* do not free blocks when zeroing */
407 __REQ_SWAP, /* swapping request. */
408 __REQ_NR_BITS, /* stops here */
411 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
412 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
413 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
414 #define REQ_SYNC (1ULL << __REQ_SYNC)
415 #define REQ_META (1ULL << __REQ_META)
416 #define REQ_PRIO (1ULL << __REQ_PRIO)
417 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
418 #define REQ_IDLE (1ULL << __REQ_IDLE)
419 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
420 #define REQ_FUA (1ULL << __REQ_FUA)
421 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
422 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
423 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
424 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
425 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
427 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
428 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
430 #define REQ_DRV (1ULL << __REQ_DRV)
431 #define REQ_SWAP (1ULL << __REQ_SWAP)
433 #define REQ_FAILFAST_MASK \
434 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
436 #define REQ_NOMERGE_FLAGS \
437 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
448 #define bio_op(bio) \
449 ((bio)->bi_opf & REQ_OP_MASK)
450 #define req_op(req) \
451 ((req)->cmd_flags & REQ_OP_MASK)
453 /* obsolete, don't use in new code */
454 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
457 bio->bi_opf = op | op_flags;
460 static inline bool op_is_write(unsigned int op)
466 * Check if the bio or request is one that needs special treatment in the
467 * flush state machine.
469 static inline bool op_is_flush(unsigned int op)
471 return op & (REQ_FUA | REQ_PREFLUSH);
475 * Reads are always treated as synchronous, as are requests with the FUA or
476 * PREFLUSH flag. Other operations may be marked as synchronous using the
479 static inline bool op_is_sync(unsigned int op)
481 return (op & REQ_OP_MASK) == REQ_OP_READ ||
482 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
485 static inline bool op_is_discard(unsigned int op)
487 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
491 * Check if a bio or request operation is a zone management operation, with
492 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
493 * due to its different handling in the block layer and device response in
494 * case of command failure.
496 static inline bool op_is_zone_mgmt(enum req_opf op)
498 switch (op & REQ_OP_MASK) {
499 case REQ_OP_ZONE_RESET:
500 case REQ_OP_ZONE_OPEN:
501 case REQ_OP_ZONE_CLOSE:
502 case REQ_OP_ZONE_FINISH:
509 static inline int op_stat_group(unsigned int op)
511 if (op_is_discard(op))
513 return op_is_write(op);
516 typedef unsigned int blk_qc_t;
517 #define BLK_QC_T_NONE -1U
518 #define BLK_QC_T_SHIFT 16
519 #define BLK_QC_T_INTERNAL (1U << 31)
521 static inline bool blk_qc_t_valid(blk_qc_t cookie)
523 return cookie != BLK_QC_T_NONE;
526 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
528 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
531 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
533 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
536 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
538 return (cookie & BLK_QC_T_INTERNAL) != 0;
549 #endif /* __LINUX_BLK_TYPES_H */