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 *);
22 #define BDEV_NEED_PART_SCAN 0
27 struct inode * bd_inode; /* will die */
28 struct super_block * bd_super;
29 struct mutex bd_mutex; /* open/close mutex */
35 struct list_head bd_holder_disks;
37 struct block_device * bd_contains;
39 struct hd_struct * bd_part;
40 /* number of times partitions within this device have been opened. */
41 unsigned bd_part_count;
42 unsigned long bd_flags;
43 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
44 struct gendisk * bd_disk;
45 struct backing_dev_info *bd_bdi;
47 /* The counter of freeze processes */
48 int bd_fsfreeze_count;
49 /* Mutex for freeze */
50 struct mutex bd_fsfreeze_mutex;
54 * Block error status values. See block/blk-core:blk_errors for the details.
55 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
57 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
58 typedef u32 __bitwise blk_status_t;
60 typedef u8 __bitwise blk_status_t;
63 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
64 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
65 #define BLK_STS_NOSPC ((__force blk_status_t)3)
66 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
67 #define BLK_STS_TARGET ((__force blk_status_t)5)
68 #define BLK_STS_NEXUS ((__force blk_status_t)6)
69 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
70 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
71 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
72 #define BLK_STS_IOERR ((__force blk_status_t)10)
74 /* hack for device mapper, don't use elsewhere: */
75 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
77 #define BLK_STS_AGAIN ((__force blk_status_t)12)
80 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
81 * device related resources are unavailable, but the driver can guarantee
82 * that the queue will be rerun in the future once resources become
83 * available again. This is typically the case for device specific
84 * resources that are consumed for IO. If the driver fails allocating these
85 * resources, we know that inflight (or pending) IO will free these
86 * resource upon completion.
88 * This is different from BLK_STS_RESOURCE in that it explicitly references
89 * a device specific resource. For resources of wider scope, allocation
90 * failure can happen without having pending IO. This means that we can't
91 * rely on request completions freeing these resources, as IO may not be in
92 * flight. Examples of that are kernel memory allocations, DMA mappings, or
93 * any other system wide resources.
95 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
98 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
99 * related resources are unavailable, but the driver can guarantee the queue
100 * will be rerun in the future once the resources become available again.
102 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
103 * a zone specific resource and IO to a different zone on the same device could
104 * still be served. Examples of that are zones that are write-locked, but a read
105 * to the same zone could be served.
107 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
110 * blk_path_error - returns true if error may be path related
111 * @error: status the request was completed with
114 * This classifies block error status into non-retryable errors and ones
115 * that may be successful if retried on a failover path.
118 * %false - retrying failover path will not help
119 * %true - may succeed if retried
121 static inline bool blk_path_error(blk_status_t error)
124 case BLK_STS_NOTSUPP:
129 case BLK_STS_PROTECTION:
133 /* Anything else could be a path failure, so should be retried */
138 * From most significant bit:
139 * 1 bit: reserved for other usage, see below
140 * 12 bits: original size of bio
141 * 51 bits: issue time of bio
143 #define BIO_ISSUE_RES_BITS 1
144 #define BIO_ISSUE_SIZE_BITS 12
145 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
146 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
147 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
148 #define BIO_ISSUE_SIZE_MASK \
149 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
150 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
152 /* Reserved bit for blk-throtl */
153 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
159 static inline u64 __bio_issue_time(u64 time)
161 return time & BIO_ISSUE_TIME_MASK;
164 static inline u64 bio_issue_time(struct bio_issue *issue)
166 return __bio_issue_time(issue->value);
169 static inline sector_t bio_issue_size(struct bio_issue *issue)
171 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
174 static inline void bio_issue_init(struct bio_issue *issue,
177 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
178 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
179 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
180 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
184 * main unit of I/O for the block layer and lower layers (ie drivers and
188 struct bio *bi_next; /* request queue link */
189 struct gendisk *bi_disk;
190 unsigned int bi_opf; /* bottom bits req flags,
191 * top bits REQ_OP. Use
194 unsigned short bi_flags; /* status, etc and bvec pool number */
195 unsigned short bi_ioprio;
196 unsigned short bi_write_hint;
197 blk_status_t bi_status;
199 atomic_t __bi_remaining;
201 struct bvec_iter bi_iter;
203 bio_end_io_t *bi_end_io;
206 #ifdef CONFIG_BLK_CGROUP
208 * Represents the association of the css and request_queue for the bio.
209 * If a bio goes direct to device, it will not have a blkg as it will
210 * not have a request_queue associated with it. The reference is put
211 * on release of the bio.
213 struct blkcg_gq *bi_blkg;
214 struct bio_issue bi_issue;
215 #ifdef CONFIG_BLK_CGROUP_IOCOST
220 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
221 struct bio_crypt_ctx *bi_crypt_context;
225 #if defined(CONFIG_BLK_DEV_INTEGRITY)
226 struct bio_integrity_payload *bi_integrity; /* data integrity */
230 unsigned short bi_vcnt; /* how many bio_vec's */
233 * Everything starting with bi_max_vecs will be preserved by bio_reset()
236 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
238 atomic_t __bi_cnt; /* pin count */
240 struct bio_vec *bi_io_vec; /* the actual vec list */
242 struct bio_set *bi_pool;
245 * We can inline a number of vecs at the end of the bio, to avoid
246 * double allocations for a small number of bio_vecs. This member
247 * MUST obviously be kept at the very end of the bio.
249 struct bio_vec bi_inline_vecs[];
252 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
258 BIO_NO_PAGE_REF, /* don't put release vec pages */
259 BIO_CLONED, /* doesn't own data */
260 BIO_BOUNCED, /* bio is a bounce bio */
261 BIO_WORKINGSET, /* contains userspace workingset pages */
262 BIO_QUIET, /* Make BIO Quiet */
263 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
264 BIO_REFFED, /* bio has elevated ->bi_cnt */
265 BIO_THROTTLED, /* This bio has already been subjected to
266 * throttling rules. Don't do it again. */
267 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
269 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
270 BIO_TRACKED, /* set if bio goes through the rq_qos path */
274 /* See BVEC_POOL_OFFSET below before adding new flags */
277 * We support 6 different bvec pools, the last one is magic in that it
278 * is backed by a mempool.
280 #define BVEC_POOL_NR 6
281 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
284 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
285 * 1 to the actual index so that 0 indicates that there are no bvecs to be
288 #define BVEC_POOL_BITS (3)
289 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
290 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
291 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
292 # error "BVEC_POOL_BITS is too small"
296 * Flags starting here get preserved by bio_reset() - this includes
297 * only BVEC_POOL_IDX()
299 #define BIO_RESET_BITS BVEC_POOL_OFFSET
301 typedef __u32 __bitwise blk_mq_req_flags_t;
304 * Operations and flags common to the bio and request structures.
305 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
307 * The least significant bit of the operation number indicates the data
308 * transfer direction:
310 * - if the least significant bit is set transfers are TO the device
311 * - if the least significant bit is not set transfers are FROM the device
313 * If a operation does not transfer data the least significant bit has no
316 #define REQ_OP_BITS 8
317 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
318 #define REQ_FLAG_BITS 24
321 /* read sectors from the device */
323 /* write sectors to the device */
325 /* flush the volatile write cache */
327 /* discard sectors */
329 /* securely erase sectors */
330 REQ_OP_SECURE_ERASE = 5,
331 /* write the same sector many times */
332 REQ_OP_WRITE_SAME = 7,
333 /* write the zero filled sector many times */
334 REQ_OP_WRITE_ZEROES = 9,
336 REQ_OP_ZONE_OPEN = 10,
338 REQ_OP_ZONE_CLOSE = 11,
339 /* Transition a zone to full */
340 REQ_OP_ZONE_FINISH = 12,
341 /* write data at the current zone write pointer */
342 REQ_OP_ZONE_APPEND = 13,
343 /* reset a zone write pointer */
344 REQ_OP_ZONE_RESET = 15,
345 /* reset all the zone present on the device */
346 REQ_OP_ZONE_RESET_ALL = 17,
348 /* SCSI passthrough using struct scsi_request */
350 REQ_OP_SCSI_OUT = 33,
351 /* Driver private requests */
359 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
361 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
362 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
363 __REQ_SYNC, /* request is sync (sync write or read) */
364 __REQ_META, /* metadata io request */
365 __REQ_PRIO, /* boost priority in cfq */
366 __REQ_NOMERGE, /* don't touch this for merging */
367 __REQ_IDLE, /* anticipate more IO after this one */
368 __REQ_INTEGRITY, /* I/O includes block integrity payload */
369 __REQ_FUA, /* forced unit access */
370 __REQ_PREFLUSH, /* request for cache flush */
371 __REQ_RAHEAD, /* read ahead, can fail anytime */
372 __REQ_BACKGROUND, /* background IO */
373 __REQ_NOWAIT, /* Don't wait if request will block */
375 * When a shared kthread needs to issue a bio for a cgroup, doing
376 * so synchronously can lead to priority inversions as the kthread
377 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
378 * submit_bio() punt the actual issuing to a dedicated per-blkcg
379 * work item to avoid such priority inversions.
383 /* command specific flags for REQ_OP_WRITE_ZEROES: */
384 __REQ_NOUNMAP, /* do not free blocks when zeroing */
390 __REQ_SWAP, /* swapping request. */
391 __REQ_NR_BITS, /* stops here */
394 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
395 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
396 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
397 #define REQ_SYNC (1ULL << __REQ_SYNC)
398 #define REQ_META (1ULL << __REQ_META)
399 #define REQ_PRIO (1ULL << __REQ_PRIO)
400 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
401 #define REQ_IDLE (1ULL << __REQ_IDLE)
402 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
403 #define REQ_FUA (1ULL << __REQ_FUA)
404 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
405 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
406 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
407 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
408 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
410 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
411 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
413 #define REQ_DRV (1ULL << __REQ_DRV)
414 #define REQ_SWAP (1ULL << __REQ_SWAP)
416 #define REQ_FAILFAST_MASK \
417 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
419 #define REQ_NOMERGE_FLAGS \
420 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
431 #define bio_op(bio) \
432 ((bio)->bi_opf & REQ_OP_MASK)
433 #define req_op(req) \
434 ((req)->cmd_flags & REQ_OP_MASK)
436 /* obsolete, don't use in new code */
437 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
440 bio->bi_opf = op | op_flags;
443 static inline bool op_is_write(unsigned int op)
449 * Check if the bio or request is one that needs special treatment in the
450 * flush state machine.
452 static inline bool op_is_flush(unsigned int op)
454 return op & (REQ_FUA | REQ_PREFLUSH);
458 * Reads are always treated as synchronous, as are requests with the FUA or
459 * PREFLUSH flag. Other operations may be marked as synchronous using the
462 static inline bool op_is_sync(unsigned int op)
464 return (op & REQ_OP_MASK) == REQ_OP_READ ||
465 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
468 static inline bool op_is_discard(unsigned int op)
470 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
474 * Check if a bio or request operation is a zone management operation, with
475 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
476 * due to its different handling in the block layer and device response in
477 * case of command failure.
479 static inline bool op_is_zone_mgmt(enum req_opf op)
481 switch (op & REQ_OP_MASK) {
482 case REQ_OP_ZONE_RESET:
483 case REQ_OP_ZONE_OPEN:
484 case REQ_OP_ZONE_CLOSE:
485 case REQ_OP_ZONE_FINISH:
492 static inline int op_stat_group(unsigned int op)
494 if (op_is_discard(op))
496 return op_is_write(op);
499 typedef unsigned int blk_qc_t;
500 #define BLK_QC_T_NONE -1U
501 #define BLK_QC_T_EAGAIN -2U
502 #define BLK_QC_T_SHIFT 16
503 #define BLK_QC_T_INTERNAL (1U << 31)
505 static inline bool blk_qc_t_valid(blk_qc_t cookie)
507 return cookie != BLK_QC_T_NONE && cookie != BLK_QC_T_EAGAIN;
510 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
512 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
515 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
517 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
520 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
522 return (cookie & BLK_QC_T_INTERNAL) != 0;
533 #endif /* __LINUX_BLK_TYPES_H */