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