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;
28 struct inode * bd_inode; /* will die */
29 struct super_block * bd_super;
30 struct mutex bd_mutex; /* open/close mutex */
36 struct list_head bd_holder_disks;
38 struct kobject *bd_holder_dir;
40 struct hd_struct * bd_part;
41 /* number of times partitions within this device have been opened. */
42 unsigned bd_part_count;
44 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
45 struct gendisk * bd_disk;
46 struct backing_dev_info *bd_bdi;
48 /* The counter of freeze processes */
49 int bd_fsfreeze_count;
50 /* Mutex for freeze */
51 struct mutex bd_fsfreeze_mutex;
52 struct super_block *bd_fsfreeze_sb;
54 struct partition_meta_info *bd_meta_info;
57 #define bdev_whole(_bdev) \
58 ((_bdev)->bd_disk->part0.bdev)
60 #define bdev_kobj(_bdev) \
61 (&part_to_dev((_bdev)->bd_part)->kobj)
64 * Block error status values. See block/blk-core:blk_errors for the details.
65 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
67 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
68 typedef u32 __bitwise blk_status_t;
70 typedef u8 __bitwise blk_status_t;
73 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
74 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
75 #define BLK_STS_NOSPC ((__force blk_status_t)3)
76 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
77 #define BLK_STS_TARGET ((__force blk_status_t)5)
78 #define BLK_STS_NEXUS ((__force blk_status_t)6)
79 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
80 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
81 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
82 #define BLK_STS_IOERR ((__force blk_status_t)10)
84 /* hack for device mapper, don't use elsewhere: */
85 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
87 #define BLK_STS_AGAIN ((__force blk_status_t)12)
90 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
91 * device related resources are unavailable, but the driver can guarantee
92 * that the queue will be rerun in the future once resources become
93 * available again. This is typically the case for device specific
94 * resources that are consumed for IO. If the driver fails allocating these
95 * resources, we know that inflight (or pending) IO will free these
96 * resource upon completion.
98 * This is different from BLK_STS_RESOURCE in that it explicitly references
99 * a device specific resource. For resources of wider scope, allocation
100 * failure can happen without having pending IO. This means that we can't
101 * rely on request completions freeing these resources, as IO may not be in
102 * flight. Examples of that are kernel memory allocations, DMA mappings, or
103 * any other system wide resources.
105 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
108 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
109 * related resources are unavailable, but the driver can guarantee the queue
110 * will be rerun in the future once the resources become available again.
112 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
113 * a zone specific resource and IO to a different zone on the same device could
114 * still be served. Examples of that are zones that are write-locked, but a read
115 * to the same zone could be served.
117 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
120 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
121 * path if the device returns a status indicating that too many zone resources
122 * are currently open. The same command should be successful if resubmitted
123 * after the number of open zones decreases below the device's limits, which is
124 * reported in the request_queue's max_open_zones.
126 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
129 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
130 * path if the device returns a status indicating that too many zone resources
131 * are currently active. The same command should be successful if resubmitted
132 * after the number of active zones decreases below the device's limits, which
133 * is reported in the request_queue's max_active_zones.
135 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
138 * blk_path_error - returns true if error may be path related
139 * @error: status the request was completed with
142 * This classifies block error status into non-retryable errors and ones
143 * that may be successful if retried on a failover path.
146 * %false - retrying failover path will not help
147 * %true - may succeed if retried
149 static inline bool blk_path_error(blk_status_t error)
152 case BLK_STS_NOTSUPP:
157 case BLK_STS_PROTECTION:
161 /* Anything else could be a path failure, so should be retried */
166 * From most significant bit:
167 * 1 bit: reserved for other usage, see below
168 * 12 bits: original size of bio
169 * 51 bits: issue time of bio
171 #define BIO_ISSUE_RES_BITS 1
172 #define BIO_ISSUE_SIZE_BITS 12
173 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
174 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
175 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
176 #define BIO_ISSUE_SIZE_MASK \
177 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
178 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
180 /* Reserved bit for blk-throtl */
181 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
187 static inline u64 __bio_issue_time(u64 time)
189 return time & BIO_ISSUE_TIME_MASK;
192 static inline u64 bio_issue_time(struct bio_issue *issue)
194 return __bio_issue_time(issue->value);
197 static inline sector_t bio_issue_size(struct bio_issue *issue)
199 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
202 static inline void bio_issue_init(struct bio_issue *issue,
205 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
206 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
207 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
208 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
212 * main unit of I/O for the block layer and lower layers (ie drivers and
216 struct bio *bi_next; /* request queue link */
217 struct gendisk *bi_disk;
218 unsigned int bi_opf; /* bottom bits req flags,
219 * top bits REQ_OP. Use
222 unsigned short bi_flags; /* status, etc and bvec pool number */
223 unsigned short bi_ioprio;
224 unsigned short bi_write_hint;
225 blk_status_t bi_status;
227 atomic_t __bi_remaining;
229 struct bvec_iter bi_iter;
231 bio_end_io_t *bi_end_io;
234 #ifdef CONFIG_BLK_CGROUP
236 * Represents the association of the css and request_queue for the bio.
237 * If a bio goes direct to device, it will not have a blkg as it will
238 * not have a request_queue associated with it. The reference is put
239 * on release of the bio.
241 struct blkcg_gq *bi_blkg;
242 struct bio_issue bi_issue;
243 #ifdef CONFIG_BLK_CGROUP_IOCOST
248 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
249 struct bio_crypt_ctx *bi_crypt_context;
253 #if defined(CONFIG_BLK_DEV_INTEGRITY)
254 struct bio_integrity_payload *bi_integrity; /* data integrity */
258 unsigned short bi_vcnt; /* how many bio_vec's */
261 * Everything starting with bi_max_vecs will be preserved by bio_reset()
264 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
266 atomic_t __bi_cnt; /* pin count */
268 struct bio_vec *bi_io_vec; /* the actual vec list */
270 struct bio_set *bi_pool;
273 * We can inline a number of vecs at the end of the bio, to avoid
274 * double allocations for a small number of bio_vecs. This member
275 * MUST obviously be kept at the very end of the bio.
277 struct bio_vec bi_inline_vecs[];
280 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
286 BIO_NO_PAGE_REF, /* don't put release vec pages */
287 BIO_CLONED, /* doesn't own data */
288 BIO_BOUNCED, /* bio is a bounce bio */
289 BIO_WORKINGSET, /* contains userspace workingset pages */
290 BIO_QUIET, /* Make BIO Quiet */
291 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
292 BIO_REFFED, /* bio has elevated ->bi_cnt */
293 BIO_THROTTLED, /* This bio has already been subjected to
294 * throttling rules. Don't do it again. */
295 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
297 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
298 BIO_TRACKED, /* set if bio goes through the rq_qos path */
302 /* See BVEC_POOL_OFFSET below before adding new flags */
305 * We support 6 different bvec pools, the last one is magic in that it
306 * is backed by a mempool.
308 #define BVEC_POOL_NR 6
309 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
312 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
313 * 1 to the actual index so that 0 indicates that there are no bvecs to be
316 #define BVEC_POOL_BITS (3)
317 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
318 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
319 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
320 # error "BVEC_POOL_BITS is too small"
324 * Flags starting here get preserved by bio_reset() - this includes
325 * only BVEC_POOL_IDX()
327 #define BIO_RESET_BITS BVEC_POOL_OFFSET
329 typedef __u32 __bitwise blk_mq_req_flags_t;
332 * Operations and flags common to the bio and request structures.
333 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
335 * The least significant bit of the operation number indicates the data
336 * transfer direction:
338 * - if the least significant bit is set transfers are TO the device
339 * - if the least significant bit is not set transfers are FROM the device
341 * If a operation does not transfer data the least significant bit has no
344 #define REQ_OP_BITS 8
345 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
346 #define REQ_FLAG_BITS 24
349 /* read sectors from the device */
351 /* write sectors to the device */
353 /* flush the volatile write cache */
355 /* discard sectors */
357 /* securely erase sectors */
358 REQ_OP_SECURE_ERASE = 5,
359 /* write the same sector many times */
360 REQ_OP_WRITE_SAME = 7,
361 /* write the zero filled sector many times */
362 REQ_OP_WRITE_ZEROES = 9,
364 REQ_OP_ZONE_OPEN = 10,
366 REQ_OP_ZONE_CLOSE = 11,
367 /* Transition a zone to full */
368 REQ_OP_ZONE_FINISH = 12,
369 /* write data at the current zone write pointer */
370 REQ_OP_ZONE_APPEND = 13,
371 /* reset a zone write pointer */
372 REQ_OP_ZONE_RESET = 15,
373 /* reset all the zone present on the device */
374 REQ_OP_ZONE_RESET_ALL = 17,
376 /* SCSI passthrough using struct scsi_request */
378 REQ_OP_SCSI_OUT = 33,
379 /* Driver private requests */
387 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
389 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
390 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
391 __REQ_SYNC, /* request is sync (sync write or read) */
392 __REQ_META, /* metadata io request */
393 __REQ_PRIO, /* boost priority in cfq */
394 __REQ_NOMERGE, /* don't touch this for merging */
395 __REQ_IDLE, /* anticipate more IO after this one */
396 __REQ_INTEGRITY, /* I/O includes block integrity payload */
397 __REQ_FUA, /* forced unit access */
398 __REQ_PREFLUSH, /* request for cache flush */
399 __REQ_RAHEAD, /* read ahead, can fail anytime */
400 __REQ_BACKGROUND, /* background IO */
401 __REQ_NOWAIT, /* Don't wait if request will block */
403 * When a shared kthread needs to issue a bio for a cgroup, doing
404 * so synchronously can lead to priority inversions as the kthread
405 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
406 * submit_bio() punt the actual issuing to a dedicated per-blkcg
407 * work item to avoid such priority inversions.
411 /* command specific flags for REQ_OP_WRITE_ZEROES: */
412 __REQ_NOUNMAP, /* do not free blocks when zeroing */
418 __REQ_SWAP, /* swapping request. */
419 __REQ_NR_BITS, /* stops here */
422 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
423 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
424 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
425 #define REQ_SYNC (1ULL << __REQ_SYNC)
426 #define REQ_META (1ULL << __REQ_META)
427 #define REQ_PRIO (1ULL << __REQ_PRIO)
428 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
429 #define REQ_IDLE (1ULL << __REQ_IDLE)
430 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
431 #define REQ_FUA (1ULL << __REQ_FUA)
432 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
433 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
434 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
435 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
436 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
438 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
439 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
441 #define REQ_DRV (1ULL << __REQ_DRV)
442 #define REQ_SWAP (1ULL << __REQ_SWAP)
444 #define REQ_FAILFAST_MASK \
445 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
447 #define REQ_NOMERGE_FLAGS \
448 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
459 #define bio_op(bio) \
460 ((bio)->bi_opf & REQ_OP_MASK)
461 #define req_op(req) \
462 ((req)->cmd_flags & REQ_OP_MASK)
464 /* obsolete, don't use in new code */
465 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
468 bio->bi_opf = op | op_flags;
471 static inline bool op_is_write(unsigned int op)
477 * Check if the bio or request is one that needs special treatment in the
478 * flush state machine.
480 static inline bool op_is_flush(unsigned int op)
482 return op & (REQ_FUA | REQ_PREFLUSH);
486 * Reads are always treated as synchronous, as are requests with the FUA or
487 * PREFLUSH flag. Other operations may be marked as synchronous using the
490 static inline bool op_is_sync(unsigned int op)
492 return (op & REQ_OP_MASK) == REQ_OP_READ ||
493 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
496 static inline bool op_is_discard(unsigned int op)
498 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
502 * Check if a bio or request operation is a zone management operation, with
503 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
504 * due to its different handling in the block layer and device response in
505 * case of command failure.
507 static inline bool op_is_zone_mgmt(enum req_opf op)
509 switch (op & REQ_OP_MASK) {
510 case REQ_OP_ZONE_RESET:
511 case REQ_OP_ZONE_OPEN:
512 case REQ_OP_ZONE_CLOSE:
513 case REQ_OP_ZONE_FINISH:
520 static inline int op_stat_group(unsigned int op)
522 if (op_is_discard(op))
524 return op_is_write(op);
527 typedef unsigned int blk_qc_t;
528 #define BLK_QC_T_NONE -1U
529 #define BLK_QC_T_SHIFT 16
530 #define BLK_QC_T_INTERNAL (1U << 31)
532 static inline bool blk_qc_t_valid(blk_qc_t cookie)
534 return cookie != BLK_QC_T_NONE;
537 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
539 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
542 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
544 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
547 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
549 return (cookie & BLK_QC_T_INTERNAL) != 0;
560 #endif /* __LINUX_BLK_TYPES_H */