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