1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * Block data types and constants. Directly include this file only to
4 * break include dependency loop.
5 */
6#ifndef __LINUX_BLK_TYPES_H
7#define __LINUX_BLK_TYPES_H
8
9#include <linux/types.h>
10#include <linux/bvec.h>
11#include <linux/device.h>
12#include <linux/ktime.h>
13
14struct bio_set;
15struct bio;
16struct bio_integrity_payload;
17struct page;
18struct io_context;
19struct cgroup_subsys_state;
20typedef void (bio_end_io_t) (struct bio *);
21struct bio_crypt_ctx;
22
23/*
24 * The basic unit of block I/O is a sector. It is used in a number of contexts
25 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
26 * bytes. Variables of type sector_t represent an offset or size that is a
27 * multiple of 512 bytes. Hence these two constants.
28 */
29#ifndef SECTOR_SHIFT
30#define SECTOR_SHIFT 9
31#endif
32#ifndef SECTOR_SIZE
33#define SECTOR_SIZE (1 << SECTOR_SHIFT)
34#endif
35
36#define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
37#define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
38#define SECTOR_MASK (PAGE_SECTORS - 1)
39
40struct block_device {
41 sector_t bd_start_sect;
42 sector_t bd_nr_sectors;
43 struct gendisk * bd_disk;
44 struct request_queue * bd_queue;
45 struct disk_stats __percpu *bd_stats;
46 unsigned long bd_stamp;
47 bool bd_read_only; /* read-only policy */
48 u8 bd_partno;
49 bool bd_write_holder;
50 bool bd_has_submit_bio;
51 dev_t bd_dev;
52 atomic_t bd_openers;
53 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
54 struct inode * bd_inode; /* will die */
55 void * bd_claiming;
56 void * bd_holder;
57 const struct blk_holder_ops *bd_holder_ops;
58 struct mutex bd_holder_lock;
59 /* The counter of freeze processes */
60 int bd_fsfreeze_count;
61 int bd_holders;
62 struct kobject *bd_holder_dir;
63
64 /* Mutex for freeze */
65 struct mutex bd_fsfreeze_mutex;
66 struct super_block *bd_fsfreeze_sb;
67
68 struct partition_meta_info *bd_meta_info;
69#ifdef CONFIG_FAIL_MAKE_REQUEST
70 bool bd_make_it_fail;
71#endif
72 /*
73 * keep this out-of-line as it's both big and not needed in the fast
74 * path
75 */
76 struct device bd_device;
77} __randomize_layout;
78
79#define bdev_whole(_bdev) \
80 ((_bdev)->bd_disk->part0)
81
82#define dev_to_bdev(device) \
83 container_of((device), struct block_device, bd_device)
84
85#define bdev_kobj(_bdev) \
86 (&((_bdev)->bd_device.kobj))
87
88/*
89 * Block error status values. See block/blk-core:blk_errors for the details.
90 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
91 */
92#if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
93typedef u32 __bitwise blk_status_t;
94typedef u32 blk_short_t;
95#else
96typedef u8 __bitwise blk_status_t;
97typedef u16 blk_short_t;
98#endif
99#define BLK_STS_OK 0
100#define BLK_STS_NOTSUPP ((__force blk_status_t)1)
101#define BLK_STS_TIMEOUT ((__force blk_status_t)2)
102#define BLK_STS_NOSPC ((__force blk_status_t)3)
103#define BLK_STS_TRANSPORT ((__force blk_status_t)4)
104#define BLK_STS_TARGET ((__force blk_status_t)5)
105#define BLK_STS_RESV_CONFLICT ((__force blk_status_t)6)
106#define BLK_STS_MEDIUM ((__force blk_status_t)7)
107#define BLK_STS_PROTECTION ((__force blk_status_t)8)
108#define BLK_STS_RESOURCE ((__force blk_status_t)9)
109#define BLK_STS_IOERR ((__force blk_status_t)10)
110
111/* hack for device mapper, don't use elsewhere: */
112#define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
113
114/*
115 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set
116 * and the bio would block (cf bio_wouldblock_error())
117 */
118#define BLK_STS_AGAIN ((__force blk_status_t)12)
119
120/*
121 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
122 * device related resources are unavailable, but the driver can guarantee
123 * that the queue will be rerun in the future once resources become
124 * available again. This is typically the case for device specific
125 * resources that are consumed for IO. If the driver fails allocating these
126 * resources, we know that inflight (or pending) IO will free these
127 * resource upon completion.
128 *
129 * This is different from BLK_STS_RESOURCE in that it explicitly references
130 * a device specific resource. For resources of wider scope, allocation
131 * failure can happen without having pending IO. This means that we can't
132 * rely on request completions freeing these resources, as IO may not be in
133 * flight. Examples of that are kernel memory allocations, DMA mappings, or
134 * any other system wide resources.
135 */
136#define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
137
138/*
139 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
140 * related resources are unavailable, but the driver can guarantee the queue
141 * will be rerun in the future once the resources become available again.
142 *
143 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
144 * a zone specific resource and IO to a different zone on the same device could
145 * still be served. Examples of that are zones that are write-locked, but a read
146 * to the same zone could be served.
147 */
148#define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
149
150/*
151 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
152 * path if the device returns a status indicating that too many zone resources
153 * are currently open. The same command should be successful if resubmitted
154 * after the number of open zones decreases below the device's limits, which is
155 * reported in the request_queue's max_open_zones.
156 */
157#define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
158
159/*
160 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
161 * path if the device returns a status indicating that too many zone resources
162 * are currently active. The same command should be successful if resubmitted
163 * after the number of active zones decreases below the device's limits, which
164 * is reported in the request_queue's max_active_zones.
165 */
166#define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
167
168/*
169 * BLK_STS_OFFLINE is returned from the driver when the target device is offline
170 * or is being taken offline. This could help differentiate the case where a
171 * device is intentionally being shut down from a real I/O error.
172 */
173#define BLK_STS_OFFLINE ((__force blk_status_t)17)
174
175/*
176 * BLK_STS_DURATION_LIMIT is returned from the driver when the target device
177 * aborted the command because it exceeded one of its Command Duration Limits.
178 */
179#define BLK_STS_DURATION_LIMIT ((__force blk_status_t)18)
180
181/**
182 * blk_path_error - returns true if error may be path related
183 * @error: status the request was completed with
184 *
185 * Description:
186 * This classifies block error status into non-retryable errors and ones
187 * that may be successful if retried on a failover path.
188 *
189 * Return:
190 * %false - retrying failover path will not help
191 * %true - may succeed if retried
192 */
193static inline bool blk_path_error(blk_status_t error)
194{
195 switch (error) {
196 case BLK_STS_NOTSUPP:
197 case BLK_STS_NOSPC:
198 case BLK_STS_TARGET:
199 case BLK_STS_RESV_CONFLICT:
200 case BLK_STS_MEDIUM:
201 case BLK_STS_PROTECTION:
202 return false;
203 }
204
205 /* Anything else could be a path failure, so should be retried */
206 return true;
207}
208
209/*
210 * From most significant bit:
211 * 1 bit: reserved for other usage, see below
212 * 12 bits: original size of bio
213 * 51 bits: issue time of bio
214 */
215#define BIO_ISSUE_RES_BITS 1
216#define BIO_ISSUE_SIZE_BITS 12
217#define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
218#define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
219#define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
220#define BIO_ISSUE_SIZE_MASK \
221 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
222#define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
223
224/* Reserved bit for blk-throtl */
225#define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
226
227struct bio_issue {
228 u64 value;
229};
230
231static inline u64 __bio_issue_time(u64 time)
232{
233 return time & BIO_ISSUE_TIME_MASK;
234}
235
236static inline u64 bio_issue_time(struct bio_issue *issue)
237{
238 return __bio_issue_time(time: issue->value);
239}
240
241static inline sector_t bio_issue_size(struct bio_issue *issue)
242{
243 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
244}
245
246static inline void bio_issue_init(struct bio_issue *issue,
247 sector_t size)
248{
249 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
250 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
251 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
252 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
253}
254
255typedef __u32 __bitwise blk_opf_t;
256
257typedef unsigned int blk_qc_t;
258#define BLK_QC_T_NONE -1U
259
260/*
261 * main unit of I/O for the block layer and lower layers (ie drivers and
262 * stacking drivers)
263 */
264struct bio {
265 struct bio *bi_next; /* request queue link */
266 struct block_device *bi_bdev;
267 blk_opf_t bi_opf; /* bottom bits REQ_OP, top bits
268 * req_flags.
269 */
270 unsigned short bi_flags; /* BIO_* below */
271 unsigned short bi_ioprio;
272 blk_status_t bi_status;
273 atomic_t __bi_remaining;
274
275 struct bvec_iter bi_iter;
276
277 blk_qc_t bi_cookie;
278 bio_end_io_t *bi_end_io;
279 void *bi_private;
280#ifdef CONFIG_BLK_CGROUP
281 /*
282 * Represents the association of the css and request_queue for the bio.
283 * If a bio goes direct to device, it will not have a blkg as it will
284 * not have a request_queue associated with it. The reference is put
285 * on release of the bio.
286 */
287 struct blkcg_gq *bi_blkg;
288 struct bio_issue bi_issue;
289#ifdef CONFIG_BLK_CGROUP_IOCOST
290 u64 bi_iocost_cost;
291#endif
292#endif
293
294#ifdef CONFIG_BLK_INLINE_ENCRYPTION
295 struct bio_crypt_ctx *bi_crypt_context;
296#endif
297
298 union {
299#if defined(CONFIG_BLK_DEV_INTEGRITY)
300 struct bio_integrity_payload *bi_integrity; /* data integrity */
301#endif
302 };
303
304 unsigned short bi_vcnt; /* how many bio_vec's */
305
306 /*
307 * Everything starting with bi_max_vecs will be preserved by bio_reset()
308 */
309
310 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
311
312 atomic_t __bi_cnt; /* pin count */
313
314 struct bio_vec *bi_io_vec; /* the actual vec list */
315
316 struct bio_set *bi_pool;
317
318 /*
319 * We can inline a number of vecs at the end of the bio, to avoid
320 * double allocations for a small number of bio_vecs. This member
321 * MUST obviously be kept at the very end of the bio.
322 */
323 struct bio_vec bi_inline_vecs[];
324};
325
326#define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
327#define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT)
328
329/*
330 * bio flags
331 */
332enum {
333 BIO_PAGE_PINNED, /* Unpin pages in bio_release_pages() */
334 BIO_CLONED, /* doesn't own data */
335 BIO_BOUNCED, /* bio is a bounce bio */
336 BIO_QUIET, /* Make BIO Quiet */
337 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
338 BIO_REFFED, /* bio has elevated ->bi_cnt */
339 BIO_BPS_THROTTLED, /* This bio has already been subjected to
340 * throttling rules. Don't do it again. */
341 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
342 * of this bio. */
343 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
344 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */
345 BIO_QOS_MERGED, /* but went through rq_qos merge path */
346 BIO_REMAPPED,
347 BIO_ZONE_WRITE_LOCKED, /* Owns a zoned device zone write lock */
348 BIO_FLAG_LAST
349};
350
351typedef __u32 __bitwise blk_mq_req_flags_t;
352
353#define REQ_OP_BITS 8
354#define REQ_OP_MASK (__force blk_opf_t)((1 << REQ_OP_BITS) - 1)
355#define REQ_FLAG_BITS 24
356
357/**
358 * enum req_op - Operations common to the bio and request structures.
359 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
360 *
361 * The least significant bit of the operation number indicates the data
362 * transfer direction:
363 *
364 * - if the least significant bit is set transfers are TO the device
365 * - if the least significant bit is not set transfers are FROM the device
366 *
367 * If a operation does not transfer data the least significant bit has no
368 * meaning.
369 */
370enum req_op {
371 /* read sectors from the device */
372 REQ_OP_READ = (__force blk_opf_t)0,
373 /* write sectors to the device */
374 REQ_OP_WRITE = (__force blk_opf_t)1,
375 /* flush the volatile write cache */
376 REQ_OP_FLUSH = (__force blk_opf_t)2,
377 /* discard sectors */
378 REQ_OP_DISCARD = (__force blk_opf_t)3,
379 /* securely erase sectors */
380 REQ_OP_SECURE_ERASE = (__force blk_opf_t)5,
381 /* write the zero filled sector many times */
382 REQ_OP_WRITE_ZEROES = (__force blk_opf_t)9,
383 /* Open a zone */
384 REQ_OP_ZONE_OPEN = (__force blk_opf_t)10,
385 /* Close a zone */
386 REQ_OP_ZONE_CLOSE = (__force blk_opf_t)11,
387 /* Transition a zone to full */
388 REQ_OP_ZONE_FINISH = (__force blk_opf_t)12,
389 /* write data at the current zone write pointer */
390 REQ_OP_ZONE_APPEND = (__force blk_opf_t)13,
391 /* reset a zone write pointer */
392 REQ_OP_ZONE_RESET = (__force blk_opf_t)15,
393 /* reset all the zone present on the device */
394 REQ_OP_ZONE_RESET_ALL = (__force blk_opf_t)17,
395
396 /* Driver private requests */
397 REQ_OP_DRV_IN = (__force blk_opf_t)34,
398 REQ_OP_DRV_OUT = (__force blk_opf_t)35,
399
400 REQ_OP_LAST = (__force blk_opf_t)36,
401};
402
403enum req_flag_bits {
404 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
405 REQ_OP_BITS,
406 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
407 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
408 __REQ_SYNC, /* request is sync (sync write or read) */
409 __REQ_META, /* metadata io request */
410 __REQ_PRIO, /* boost priority in cfq */
411 __REQ_NOMERGE, /* don't touch this for merging */
412 __REQ_IDLE, /* anticipate more IO after this one */
413 __REQ_INTEGRITY, /* I/O includes block integrity payload */
414 __REQ_FUA, /* forced unit access */
415 __REQ_PREFLUSH, /* request for cache flush */
416 __REQ_RAHEAD, /* read ahead, can fail anytime */
417 __REQ_BACKGROUND, /* background IO */
418 __REQ_NOWAIT, /* Don't wait if request will block */
419 __REQ_POLLED, /* caller polls for completion using bio_poll */
420 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */
421 __REQ_SWAP, /* swap I/O */
422 __REQ_DRV, /* for driver use */
423 __REQ_FS_PRIVATE, /* for file system (submitter) use */
424
425 /*
426 * Command specific flags, keep last:
427 */
428 /* for REQ_OP_WRITE_ZEROES: */
429 __REQ_NOUNMAP, /* do not free blocks when zeroing */
430
431 __REQ_NR_BITS, /* stops here */
432};
433
434#define REQ_FAILFAST_DEV \
435 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV)
436#define REQ_FAILFAST_TRANSPORT \
437 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT)
438#define REQ_FAILFAST_DRIVER \
439 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER)
440#define REQ_SYNC (__force blk_opf_t)(1ULL << __REQ_SYNC)
441#define REQ_META (__force blk_opf_t)(1ULL << __REQ_META)
442#define REQ_PRIO (__force blk_opf_t)(1ULL << __REQ_PRIO)
443#define REQ_NOMERGE (__force blk_opf_t)(1ULL << __REQ_NOMERGE)
444#define REQ_IDLE (__force blk_opf_t)(1ULL << __REQ_IDLE)
445#define REQ_INTEGRITY (__force blk_opf_t)(1ULL << __REQ_INTEGRITY)
446#define REQ_FUA (__force blk_opf_t)(1ULL << __REQ_FUA)
447#define REQ_PREFLUSH (__force blk_opf_t)(1ULL << __REQ_PREFLUSH)
448#define REQ_RAHEAD (__force blk_opf_t)(1ULL << __REQ_RAHEAD)
449#define REQ_BACKGROUND (__force blk_opf_t)(1ULL << __REQ_BACKGROUND)
450#define REQ_NOWAIT (__force blk_opf_t)(1ULL << __REQ_NOWAIT)
451#define REQ_POLLED (__force blk_opf_t)(1ULL << __REQ_POLLED)
452#define REQ_ALLOC_CACHE (__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE)
453#define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP)
454#define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV)
455#define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE)
456
457#define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP)
458
459#define REQ_FAILFAST_MASK \
460 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
461
462#define REQ_NOMERGE_FLAGS \
463 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
464
465enum stat_group {
466 STAT_READ,
467 STAT_WRITE,
468 STAT_DISCARD,
469 STAT_FLUSH,
470
471 NR_STAT_GROUPS
472};
473
474static inline enum req_op bio_op(const struct bio *bio)
475{
476 return bio->bi_opf & REQ_OP_MASK;
477}
478
479static inline bool op_is_write(blk_opf_t op)
480{
481 return !!(op & (__force blk_opf_t)1);
482}
483
484/*
485 * Check if the bio or request is one that needs special treatment in the
486 * flush state machine.
487 */
488static inline bool op_is_flush(blk_opf_t op)
489{
490 return op & (REQ_FUA | REQ_PREFLUSH);
491}
492
493/*
494 * Reads are always treated as synchronous, as are requests with the FUA or
495 * PREFLUSH flag. Other operations may be marked as synchronous using the
496 * REQ_SYNC flag.
497 */
498static inline bool op_is_sync(blk_opf_t op)
499{
500 return (op & REQ_OP_MASK) == REQ_OP_READ ||
501 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
502}
503
504static inline bool op_is_discard(blk_opf_t op)
505{
506 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
507}
508
509/*
510 * Check if a bio or request operation is a zone management operation, with
511 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
512 * due to its different handling in the block layer and device response in
513 * case of command failure.
514 */
515static inline bool op_is_zone_mgmt(enum req_op op)
516{
517 switch (op & REQ_OP_MASK) {
518 case REQ_OP_ZONE_RESET:
519 case REQ_OP_ZONE_OPEN:
520 case REQ_OP_ZONE_CLOSE:
521 case REQ_OP_ZONE_FINISH:
522 return true;
523 default:
524 return false;
525 }
526}
527
528static inline int op_stat_group(enum req_op op)
529{
530 if (op_is_discard(op))
531 return STAT_DISCARD;
532 return op_is_write(op);
533}
534
535struct blk_rq_stat {
536 u64 mean;
537 u64 min;
538 u64 max;
539 u32 nr_samples;
540 u64 batch;
541};
542
543#endif /* __LINUX_BLK_TYPES_H */
544

source code of linux/include/linux/blk_types.h