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 dev_t bd_dev; /* not a kdev_t - it's a search key */
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 struct gendisk * bd_disk;
42 struct request_queue * bd_queue;
43 struct backing_dev_info *bd_bdi;
44 struct list_head bd_list;
46 * Private data. You must have bd_claim'ed the block_device
47 * to use this. NOTE: bd_claim allows an owner to claim
48 * the same device multiple times, the owner must take special
49 * care to not mess up bd_private for that case.
51 unsigned long bd_private;
53 /* The counter of freeze processes */
54 int bd_fsfreeze_count;
55 /* Mutex for freeze */
56 struct mutex bd_fsfreeze_mutex;
60 * Block error status values. See block/blk-core:blk_errors for the details.
61 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
63 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
64 typedef u32 __bitwise blk_status_t;
66 typedef u8 __bitwise blk_status_t;
69 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
70 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
71 #define BLK_STS_NOSPC ((__force blk_status_t)3)
72 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
73 #define BLK_STS_TARGET ((__force blk_status_t)5)
74 #define BLK_STS_NEXUS ((__force blk_status_t)6)
75 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
76 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
77 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
78 #define BLK_STS_IOERR ((__force blk_status_t)10)
80 /* hack for device mapper, don't use elsewhere: */
81 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
83 #define BLK_STS_AGAIN ((__force blk_status_t)12)
86 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
87 * device related resources are unavailable, but the driver can guarantee
88 * that the queue will be rerun in the future once resources become
89 * available again. This is typically the case for device specific
90 * resources that are consumed for IO. If the driver fails allocating these
91 * resources, we know that inflight (or pending) IO will free these
92 * resource upon completion.
94 * This is different from BLK_STS_RESOURCE in that it explicitly references
95 * a device specific resource. For resources of wider scope, allocation
96 * failure can happen without having pending IO. This means that we can't
97 * rely on request completions freeing these resources, as IO may not be in
98 * flight. Examples of that are kernel memory allocations, DMA mappings, or
99 * any other system wide resources.
101 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
104 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
105 * related resources are unavailable, but the driver can guarantee the queue
106 * will be rerun in the future once the resources become available again.
108 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
109 * a zone specific resource and IO to a different zone on the same device could
110 * still be served. Examples of that are zones that are write-locked, but a read
111 * to the same zone could be served.
113 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
116 * blk_path_error - returns true if error may be path related
117 * @error: status the request was completed with
120 * This classifies block error status into non-retryable errors and ones
121 * that may be successful if retried on a failover path.
124 * %false - retrying failover path will not help
125 * %true - may succeed if retried
127 static inline bool blk_path_error(blk_status_t error)
130 case BLK_STS_NOTSUPP:
135 case BLK_STS_PROTECTION:
139 /* Anything else could be a path failure, so should be retried */
144 * From most significant bit:
145 * 1 bit: reserved for other usage, see below
146 * 12 bits: original size of bio
147 * 51 bits: issue time of bio
149 #define BIO_ISSUE_RES_BITS 1
150 #define BIO_ISSUE_SIZE_BITS 12
151 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
152 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
153 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
154 #define BIO_ISSUE_SIZE_MASK \
155 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
156 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
158 /* Reserved bit for blk-throtl */
159 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
165 static inline u64 __bio_issue_time(u64 time)
167 return time & BIO_ISSUE_TIME_MASK;
170 static inline u64 bio_issue_time(struct bio_issue *issue)
172 return __bio_issue_time(issue->value);
175 static inline sector_t bio_issue_size(struct bio_issue *issue)
177 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
180 static inline void bio_issue_init(struct bio_issue *issue,
183 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
184 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
185 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
186 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
190 * main unit of I/O for the block layer and lower layers (ie drivers and
194 struct bio *bi_next; /* request queue link */
195 struct gendisk *bi_disk;
196 unsigned int bi_opf; /* bottom bits req flags,
197 * top bits REQ_OP. Use
200 unsigned short bi_flags; /* status, etc and bvec pool number */
201 unsigned short bi_ioprio;
202 unsigned short bi_write_hint;
203 blk_status_t bi_status;
205 atomic_t __bi_remaining;
207 struct bvec_iter bi_iter;
209 bio_end_io_t *bi_end_io;
212 #ifdef CONFIG_BLK_CGROUP
214 * Represents the association of the css and request_queue for the bio.
215 * If a bio goes direct to device, it will not have a blkg as it will
216 * not have a request_queue associated with it. The reference is put
217 * on release of the bio.
219 struct blkcg_gq *bi_blkg;
220 struct bio_issue bi_issue;
221 #ifdef CONFIG_BLK_CGROUP_IOCOST
226 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
227 struct bio_crypt_ctx *bi_crypt_context;
231 #if defined(CONFIG_BLK_DEV_INTEGRITY)
232 struct bio_integrity_payload *bi_integrity; /* data integrity */
236 unsigned short bi_vcnt; /* how many bio_vec's */
239 * Everything starting with bi_max_vecs will be preserved by bio_reset()
242 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
244 atomic_t __bi_cnt; /* pin count */
246 struct bio_vec *bi_io_vec; /* the actual vec list */
248 struct bio_set *bi_pool;
251 * We can inline a number of vecs at the end of the bio, to avoid
252 * double allocations for a small number of bio_vecs. This member
253 * MUST obviously be kept at the very end of the bio.
255 struct bio_vec bi_inline_vecs[];
258 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
264 BIO_NO_PAGE_REF, /* don't put release vec pages */
265 BIO_CLONED, /* doesn't own data */
266 BIO_BOUNCED, /* bio is a bounce bio */
267 BIO_USER_MAPPED, /* contains user pages */
268 BIO_NULL_MAPPED, /* contains invalid user pages */
269 BIO_WORKINGSET, /* contains userspace workingset pages */
270 BIO_QUIET, /* Make BIO Quiet */
271 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
272 BIO_REFFED, /* bio has elevated ->bi_cnt */
273 BIO_THROTTLED, /* This bio has already been subjected to
274 * throttling rules. Don't do it again. */
275 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
277 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
278 BIO_TRACKED, /* set if bio goes through the rq_qos path */
282 /* See BVEC_POOL_OFFSET below before adding new flags */
285 * We support 6 different bvec pools, the last one is magic in that it
286 * is backed by a mempool.
288 #define BVEC_POOL_NR 6
289 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
292 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
293 * 1 to the actual index so that 0 indicates that there are no bvecs to be
296 #define BVEC_POOL_BITS (3)
297 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
298 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
299 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
300 # error "BVEC_POOL_BITS is too small"
304 * Flags starting here get preserved by bio_reset() - this includes
305 * only BVEC_POOL_IDX()
307 #define BIO_RESET_BITS BVEC_POOL_OFFSET
309 typedef __u32 __bitwise blk_mq_req_flags_t;
312 * Operations and flags common to the bio and request structures.
313 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
315 * The least significant bit of the operation number indicates the data
316 * transfer direction:
318 * - if the least significant bit is set transfers are TO the device
319 * - if the least significant bit is not set transfers are FROM the device
321 * If a operation does not transfer data the least significant bit has no
324 #define REQ_OP_BITS 8
325 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
326 #define REQ_FLAG_BITS 24
329 /* read sectors from the device */
331 /* write sectors to the device */
333 /* flush the volatile write cache */
335 /* discard sectors */
337 /* securely erase sectors */
338 REQ_OP_SECURE_ERASE = 5,
339 /* reset a zone write pointer */
340 REQ_OP_ZONE_RESET = 6,
341 /* write the same sector many times */
342 REQ_OP_WRITE_SAME = 7,
343 /* reset all the zone present on the device */
344 REQ_OP_ZONE_RESET_ALL = 8,
345 /* write the zero filled sector many times */
346 REQ_OP_WRITE_ZEROES = 9,
348 REQ_OP_ZONE_OPEN = 10,
350 REQ_OP_ZONE_CLOSE = 11,
351 /* Transition a zone to full */
352 REQ_OP_ZONE_FINISH = 12,
353 /* write data at the current zone write pointer */
354 REQ_OP_ZONE_APPEND = 13,
356 /* SCSI passthrough using struct scsi_request */
358 REQ_OP_SCSI_OUT = 33,
359 /* Driver private requests */
367 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
369 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
370 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
371 __REQ_SYNC, /* request is sync (sync write or read) */
372 __REQ_META, /* metadata io request */
373 __REQ_PRIO, /* boost priority in cfq */
374 __REQ_NOMERGE, /* don't touch this for merging */
375 __REQ_IDLE, /* anticipate more IO after this one */
376 __REQ_INTEGRITY, /* I/O includes block integrity payload */
377 __REQ_FUA, /* forced unit access */
378 __REQ_PREFLUSH, /* request for cache flush */
379 __REQ_RAHEAD, /* read ahead, can fail anytime */
380 __REQ_BACKGROUND, /* background IO */
381 __REQ_NOWAIT, /* Don't wait if request will block */
383 * When a shared kthread needs to issue a bio for a cgroup, doing
384 * so synchronously can lead to priority inversions as the kthread
385 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
386 * submit_bio() punt the actual issuing to a dedicated per-blkcg
387 * work item to avoid such priority inversions.
391 /* command specific flags for REQ_OP_WRITE_ZEROES: */
392 __REQ_NOUNMAP, /* do not free blocks when zeroing */
398 __REQ_SWAP, /* swapping request. */
399 __REQ_NR_BITS, /* stops here */
402 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
403 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
404 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
405 #define REQ_SYNC (1ULL << __REQ_SYNC)
406 #define REQ_META (1ULL << __REQ_META)
407 #define REQ_PRIO (1ULL << __REQ_PRIO)
408 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
409 #define REQ_IDLE (1ULL << __REQ_IDLE)
410 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
411 #define REQ_FUA (1ULL << __REQ_FUA)
412 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
413 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
414 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
415 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
416 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
418 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
419 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
421 #define REQ_DRV (1ULL << __REQ_DRV)
422 #define REQ_SWAP (1ULL << __REQ_SWAP)
424 #define REQ_FAILFAST_MASK \
425 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
427 #define REQ_NOMERGE_FLAGS \
428 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
439 #define bio_op(bio) \
440 ((bio)->bi_opf & REQ_OP_MASK)
441 #define req_op(req) \
442 ((req)->cmd_flags & REQ_OP_MASK)
444 /* obsolete, don't use in new code */
445 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
448 bio->bi_opf = op | op_flags;
451 static inline bool op_is_write(unsigned int op)
457 * Check if the bio or request is one that needs special treatment in the
458 * flush state machine.
460 static inline bool op_is_flush(unsigned int op)
462 return op & (REQ_FUA | REQ_PREFLUSH);
466 * Reads are always treated as synchronous, as are requests with the FUA or
467 * PREFLUSH flag. Other operations may be marked as synchronous using the
470 static inline bool op_is_sync(unsigned int op)
472 return (op & REQ_OP_MASK) == REQ_OP_READ ||
473 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
476 static inline bool op_is_discard(unsigned int op)
478 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
482 * Check if a bio or request operation is a zone management operation, with
483 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
484 * due to its different handling in the block layer and device response in
485 * case of command failure.
487 static inline bool op_is_zone_mgmt(enum req_opf op)
489 switch (op & REQ_OP_MASK) {
490 case REQ_OP_ZONE_RESET:
491 case REQ_OP_ZONE_OPEN:
492 case REQ_OP_ZONE_CLOSE:
493 case REQ_OP_ZONE_FINISH:
500 static inline int op_stat_group(unsigned int op)
502 if (op_is_discard(op))
504 return op_is_write(op);
507 typedef unsigned int blk_qc_t;
508 #define BLK_QC_T_NONE -1U
509 #define BLK_QC_T_EAGAIN -2U
510 #define BLK_QC_T_SHIFT 16
511 #define BLK_QC_T_INTERNAL (1U << 31)
513 static inline bool blk_qc_t_valid(blk_qc_t cookie)
515 return cookie != BLK_QC_T_NONE && cookie != BLK_QC_T_EAGAIN;
518 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
520 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
523 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
525 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
528 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
530 return (cookie & BLK_QC_T_INTERNAL) != 0;
541 #endif /* __LINUX_BLK_TYPES_H */