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