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 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 	bool			bd_ro_warned;
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  * Alpha cannot write a byte atomically, so we need to use 32-bit value.
92  */
93 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
94 typedef u32 __bitwise blk_status_t;
95 typedef u32 blk_short_t;
96 #else
97 typedef u8 __bitwise blk_status_t;
98 typedef u16 blk_short_t;
99 #endif
100 #define	BLK_STS_OK 0
101 #define BLK_STS_NOTSUPP		((__force blk_status_t)1)
102 #define BLK_STS_TIMEOUT		((__force blk_status_t)2)
103 #define BLK_STS_NOSPC		((__force blk_status_t)3)
104 #define BLK_STS_TRANSPORT	((__force blk_status_t)4)
105 #define BLK_STS_TARGET		((__force blk_status_t)5)
106 #define BLK_STS_RESV_CONFLICT	((__force blk_status_t)6)
107 #define BLK_STS_MEDIUM		((__force blk_status_t)7)
108 #define BLK_STS_PROTECTION	((__force blk_status_t)8)
109 #define BLK_STS_RESOURCE	((__force blk_status_t)9)
110 #define BLK_STS_IOERR		((__force blk_status_t)10)
111 
112 /* hack for device mapper, don't use elsewhere: */
113 #define BLK_STS_DM_REQUEUE    ((__force blk_status_t)11)
114 
115 /*
116  * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set
117  * and the bio would block (cf bio_wouldblock_error())
118  */
119 #define BLK_STS_AGAIN		((__force blk_status_t)12)
120 
121 /*
122  * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
123  * device related resources are unavailable, but the driver can guarantee
124  * that the queue will be rerun in the future once resources become
125  * available again. This is typically the case for device specific
126  * resources that are consumed for IO. If the driver fails allocating these
127  * resources, we know that inflight (or pending) IO will free these
128  * resource upon completion.
129  *
130  * This is different from BLK_STS_RESOURCE in that it explicitly references
131  * a device specific resource. For resources of wider scope, allocation
132  * failure can happen without having pending IO. This means that we can't
133  * rely on request completions freeing these resources, as IO may not be in
134  * flight. Examples of that are kernel memory allocations, DMA mappings, or
135  * any other system wide resources.
136  */
137 #define BLK_STS_DEV_RESOURCE	((__force blk_status_t)13)
138 
139 /*
140  * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
141  * related resources are unavailable, but the driver can guarantee the queue
142  * will be rerun in the future once the resources become available again.
143  *
144  * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
145  * a zone specific resource and IO to a different zone on the same device could
146  * still be served. Examples of that are zones that are write-locked, but a read
147  * to the same zone could be served.
148  */
149 #define BLK_STS_ZONE_RESOURCE	((__force blk_status_t)14)
150 
151 /*
152  * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
153  * path if the device returns a status indicating that too many zone resources
154  * are currently open. The same command should be successful if resubmitted
155  * after the number of open zones decreases below the device's limits, which is
156  * reported in the request_queue's max_open_zones.
157  */
158 #define BLK_STS_ZONE_OPEN_RESOURCE	((__force blk_status_t)15)
159 
160 /*
161  * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
162  * path if the device returns a status indicating that too many zone resources
163  * are currently active. The same command should be successful if resubmitted
164  * after the number of active zones decreases below the device's limits, which
165  * is reported in the request_queue's max_active_zones.
166  */
167 #define BLK_STS_ZONE_ACTIVE_RESOURCE	((__force blk_status_t)16)
168 
169 /*
170  * BLK_STS_OFFLINE is returned from the driver when the target device is offline
171  * or is being taken offline. This could help differentiate the case where a
172  * device is intentionally being shut down from a real I/O error.
173  */
174 #define BLK_STS_OFFLINE		((__force blk_status_t)17)
175 
176 /*
177  * BLK_STS_DURATION_LIMIT is returned from the driver when the target device
178  * aborted the command because it exceeded one of its Command Duration Limits.
179  */
180 #define BLK_STS_DURATION_LIMIT	((__force blk_status_t)18)
181 
182 /**
183  * blk_path_error - returns true if error may be path related
184  * @error: status the request was completed with
185  *
186  * Description:
187  *     This classifies block error status into non-retryable errors and ones
188  *     that may be successful if retried on a failover path.
189  *
190  * Return:
191  *     %false - retrying failover path will not help
192  *     %true  - may succeed if retried
193  */
blk_path_error(blk_status_t error)194 static inline bool blk_path_error(blk_status_t error)
195 {
196 	switch (error) {
197 	case BLK_STS_NOTSUPP:
198 	case BLK_STS_NOSPC:
199 	case BLK_STS_TARGET:
200 	case BLK_STS_RESV_CONFLICT:
201 	case BLK_STS_MEDIUM:
202 	case BLK_STS_PROTECTION:
203 		return false;
204 	}
205 
206 	/* Anything else could be a path failure, so should be retried */
207 	return true;
208 }
209 
210 /*
211  * From most significant bit:
212  * 1 bit: reserved for other usage, see below
213  * 12 bits: original size of bio
214  * 51 bits: issue time of bio
215  */
216 #define BIO_ISSUE_RES_BITS      1
217 #define BIO_ISSUE_SIZE_BITS     12
218 #define BIO_ISSUE_RES_SHIFT     (64 - BIO_ISSUE_RES_BITS)
219 #define BIO_ISSUE_SIZE_SHIFT    (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
220 #define BIO_ISSUE_TIME_MASK     ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
221 #define BIO_ISSUE_SIZE_MASK     \
222 	(((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
223 #define BIO_ISSUE_RES_MASK      (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
224 
225 /* Reserved bit for blk-throtl */
226 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
227 
228 struct bio_issue {
229 	u64 value;
230 };
231 
__bio_issue_time(u64 time)232 static inline u64 __bio_issue_time(u64 time)
233 {
234 	return time & BIO_ISSUE_TIME_MASK;
235 }
236 
bio_issue_time(struct bio_issue * issue)237 static inline u64 bio_issue_time(struct bio_issue *issue)
238 {
239 	return __bio_issue_time(issue->value);
240 }
241 
bio_issue_size(struct bio_issue * issue)242 static inline sector_t bio_issue_size(struct bio_issue *issue)
243 {
244 	return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
245 }
246 
bio_issue_init(struct bio_issue * issue,sector_t size)247 static inline void bio_issue_init(struct bio_issue *issue,
248 				       sector_t size)
249 {
250 	size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
251 	issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
252 			(ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
253 			((u64)size << BIO_ISSUE_SIZE_SHIFT));
254 }
255 
256 typedef __u32 __bitwise blk_opf_t;
257 
258 typedef unsigned int blk_qc_t;
259 #define BLK_QC_T_NONE		-1U
260 
261 /*
262  * main unit of I/O for the block layer and lower layers (ie drivers and
263  * stacking drivers)
264  */
265 struct bio {
266 	struct bio		*bi_next;	/* request queue link */
267 	struct block_device	*bi_bdev;
268 	blk_opf_t		bi_opf;		/* bottom bits REQ_OP, top bits
269 						 * req_flags.
270 						 */
271 	unsigned short		bi_flags;	/* BIO_* below */
272 	unsigned short		bi_ioprio;
273 	blk_status_t		bi_status;
274 	atomic_t		__bi_remaining;
275 
276 	struct bvec_iter	bi_iter;
277 
278 	blk_qc_t		bi_cookie;
279 	bio_end_io_t		*bi_end_io;
280 	void			*bi_private;
281 #ifdef CONFIG_BLK_CGROUP
282 	/*
283 	 * Represents the association of the css and request_queue for the bio.
284 	 * If a bio goes direct to device, it will not have a blkg as it will
285 	 * not have a request_queue associated with it.  The reference is put
286 	 * on release of the bio.
287 	 */
288 	struct blkcg_gq		*bi_blkg;
289 	struct bio_issue	bi_issue;
290 #ifdef CONFIG_BLK_CGROUP_IOCOST
291 	u64			bi_iocost_cost;
292 #endif
293 #endif
294 
295 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
296 	struct bio_crypt_ctx	*bi_crypt_context;
297 #endif
298 
299 	union {
300 #if defined(CONFIG_BLK_DEV_INTEGRITY)
301 		struct bio_integrity_payload *bi_integrity; /* data integrity */
302 #endif
303 	};
304 
305 	unsigned short		bi_vcnt;	/* how many bio_vec's */
306 
307 	/*
308 	 * Everything starting with bi_max_vecs will be preserved by bio_reset()
309 	 */
310 
311 	unsigned short		bi_max_vecs;	/* max bvl_vecs we can hold */
312 
313 	atomic_t		__bi_cnt;	/* pin count */
314 
315 	struct bio_vec		*bi_io_vec;	/* the actual vec list */
316 
317 	struct bio_set		*bi_pool;
318 
319 	/*
320 	 * We can inline a number of vecs at the end of the bio, to avoid
321 	 * double allocations for a small number of bio_vecs. This member
322 	 * MUST obviously be kept at the very end of the bio.
323 	 */
324 	struct bio_vec		bi_inline_vecs[];
325 };
326 
327 #define BIO_RESET_BYTES		offsetof(struct bio, bi_max_vecs)
328 #define BIO_MAX_SECTORS		(UINT_MAX >> SECTOR_SHIFT)
329 
330 /*
331  * bio flags
332  */
333 enum {
334 	BIO_PAGE_PINNED,	/* Unpin pages in bio_release_pages() */
335 	BIO_CLONED,		/* doesn't own data */
336 	BIO_BOUNCED,		/* bio is a bounce bio */
337 	BIO_QUIET,		/* Make BIO Quiet */
338 	BIO_CHAIN,		/* chained bio, ->bi_remaining in effect */
339 	BIO_REFFED,		/* bio has elevated ->bi_cnt */
340 	BIO_BPS_THROTTLED,	/* This bio has already been subjected to
341 				 * throttling rules. Don't do it again. */
342 	BIO_TRACE_COMPLETION,	/* bio_endio() should trace the final completion
343 				 * of this bio. */
344 	BIO_CGROUP_ACCT,	/* has been accounted to a cgroup */
345 	BIO_QOS_THROTTLED,	/* bio went through rq_qos throttle path */
346 	BIO_QOS_MERGED,		/* but went through rq_qos merge path */
347 	BIO_REMAPPED,
348 	BIO_ZONE_WRITE_LOCKED,	/* Owns a zoned device zone write lock */
349 	BIO_FLAG_LAST
350 };
351 
352 typedef __u32 __bitwise blk_mq_req_flags_t;
353 
354 #define REQ_OP_BITS	8
355 #define REQ_OP_MASK	(__force blk_opf_t)((1 << REQ_OP_BITS) - 1)
356 #define REQ_FLAG_BITS	24
357 
358 /**
359  * enum req_op - Operations common to the bio and request structures.
360  * We use 8 bits for encoding the operation, and the remaining 24 for flags.
361  *
362  * The least significant bit of the operation number indicates the data
363  * transfer direction:
364  *
365  *   - if the least significant bit is set transfers are TO the device
366  *   - if the least significant bit is not set transfers are FROM the device
367  *
368  * If a operation does not transfer data the least significant bit has no
369  * meaning.
370  */
371 enum req_op {
372 	/* read sectors from the device */
373 	REQ_OP_READ		= (__force blk_opf_t)0,
374 	/* write sectors to the device */
375 	REQ_OP_WRITE		= (__force blk_opf_t)1,
376 	/* flush the volatile write cache */
377 	REQ_OP_FLUSH		= (__force blk_opf_t)2,
378 	/* discard sectors */
379 	REQ_OP_DISCARD		= (__force blk_opf_t)3,
380 	/* securely erase sectors */
381 	REQ_OP_SECURE_ERASE	= (__force blk_opf_t)5,
382 	/* write the zero filled sector many times */
383 	REQ_OP_WRITE_ZEROES	= (__force blk_opf_t)9,
384 	/* Open a zone */
385 	REQ_OP_ZONE_OPEN	= (__force blk_opf_t)10,
386 	/* Close a zone */
387 	REQ_OP_ZONE_CLOSE	= (__force blk_opf_t)11,
388 	/* Transition a zone to full */
389 	REQ_OP_ZONE_FINISH	= (__force blk_opf_t)12,
390 	/* write data at the current zone write pointer */
391 	REQ_OP_ZONE_APPEND	= (__force blk_opf_t)13,
392 	/* reset a zone write pointer */
393 	REQ_OP_ZONE_RESET	= (__force blk_opf_t)15,
394 	/* reset all the zone present on the device */
395 	REQ_OP_ZONE_RESET_ALL	= (__force blk_opf_t)17,
396 
397 	/* Driver private requests */
398 	REQ_OP_DRV_IN		= (__force blk_opf_t)34,
399 	REQ_OP_DRV_OUT		= (__force blk_opf_t)35,
400 
401 	REQ_OP_LAST		= (__force blk_opf_t)36,
402 };
403 
404 enum req_flag_bits {
405 	__REQ_FAILFAST_DEV =	/* no driver retries of device errors */
406 		REQ_OP_BITS,
407 	__REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
408 	__REQ_FAILFAST_DRIVER,	/* no driver retries of driver errors */
409 	__REQ_SYNC,		/* request is sync (sync write or read) */
410 	__REQ_META,		/* metadata io request */
411 	__REQ_PRIO,		/* boost priority in cfq */
412 	__REQ_NOMERGE,		/* don't touch this for merging */
413 	__REQ_IDLE,		/* anticipate more IO after this one */
414 	__REQ_INTEGRITY,	/* I/O includes block integrity payload */
415 	__REQ_FUA,		/* forced unit access */
416 	__REQ_PREFLUSH,		/* request for cache flush */
417 	__REQ_RAHEAD,		/* read ahead, can fail anytime */
418 	__REQ_BACKGROUND,	/* background IO */
419 	__REQ_NOWAIT,           /* Don't wait if request will block */
420 	__REQ_POLLED,		/* caller polls for completion using bio_poll */
421 	__REQ_ALLOC_CACHE,	/* allocate IO from cache if available */
422 	__REQ_SWAP,		/* swap I/O */
423 	__REQ_DRV,		/* for driver use */
424 	__REQ_FS_PRIVATE,	/* for file system (submitter) use */
425 
426 	/*
427 	 * Command specific flags, keep last:
428 	 */
429 	/* for REQ_OP_WRITE_ZEROES: */
430 	__REQ_NOUNMAP,		/* do not free blocks when zeroing */
431 
432 	__REQ_NR_BITS,		/* stops here */
433 };
434 
435 #define REQ_FAILFAST_DEV	\
436 			(__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV)
437 #define REQ_FAILFAST_TRANSPORT	\
438 			(__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT)
439 #define REQ_FAILFAST_DRIVER	\
440 			(__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER)
441 #define REQ_SYNC	(__force blk_opf_t)(1ULL << __REQ_SYNC)
442 #define REQ_META	(__force blk_opf_t)(1ULL << __REQ_META)
443 #define REQ_PRIO	(__force blk_opf_t)(1ULL << __REQ_PRIO)
444 #define REQ_NOMERGE	(__force blk_opf_t)(1ULL << __REQ_NOMERGE)
445 #define REQ_IDLE	(__force blk_opf_t)(1ULL << __REQ_IDLE)
446 #define REQ_INTEGRITY	(__force blk_opf_t)(1ULL << __REQ_INTEGRITY)
447 #define REQ_FUA		(__force blk_opf_t)(1ULL << __REQ_FUA)
448 #define REQ_PREFLUSH	(__force blk_opf_t)(1ULL << __REQ_PREFLUSH)
449 #define REQ_RAHEAD	(__force blk_opf_t)(1ULL << __REQ_RAHEAD)
450 #define REQ_BACKGROUND	(__force blk_opf_t)(1ULL << __REQ_BACKGROUND)
451 #define REQ_NOWAIT	(__force blk_opf_t)(1ULL << __REQ_NOWAIT)
452 #define REQ_POLLED	(__force blk_opf_t)(1ULL << __REQ_POLLED)
453 #define REQ_ALLOC_CACHE	(__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE)
454 #define REQ_SWAP	(__force blk_opf_t)(1ULL << __REQ_SWAP)
455 #define REQ_DRV		(__force blk_opf_t)(1ULL << __REQ_DRV)
456 #define REQ_FS_PRIVATE	(__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE)
457 
458 #define REQ_NOUNMAP	(__force blk_opf_t)(1ULL << __REQ_NOUNMAP)
459 
460 #define REQ_FAILFAST_MASK \
461 	(REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
462 
463 #define REQ_NOMERGE_FLAGS \
464 	(REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
465 
466 enum stat_group {
467 	STAT_READ,
468 	STAT_WRITE,
469 	STAT_DISCARD,
470 	STAT_FLUSH,
471 
472 	NR_STAT_GROUPS
473 };
474 
bio_op(const struct bio * bio)475 static inline enum req_op bio_op(const struct bio *bio)
476 {
477 	return bio->bi_opf & REQ_OP_MASK;
478 }
479 
op_is_write(blk_opf_t op)480 static inline bool op_is_write(blk_opf_t op)
481 {
482 	return !!(op & (__force blk_opf_t)1);
483 }
484 
485 /*
486  * Check if the bio or request is one that needs special treatment in the
487  * flush state machine.
488  */
op_is_flush(blk_opf_t op)489 static inline bool op_is_flush(blk_opf_t op)
490 {
491 	return op & (REQ_FUA | REQ_PREFLUSH);
492 }
493 
494 /*
495  * Reads are always treated as synchronous, as are requests with the FUA or
496  * PREFLUSH flag.  Other operations may be marked as synchronous using the
497  * REQ_SYNC flag.
498  */
op_is_sync(blk_opf_t op)499 static inline bool op_is_sync(blk_opf_t op)
500 {
501 	return (op & REQ_OP_MASK) == REQ_OP_READ ||
502 		(op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
503 }
504 
op_is_discard(blk_opf_t op)505 static inline bool op_is_discard(blk_opf_t op)
506 {
507 	return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
508 }
509 
510 /*
511  * Check if a bio or request operation is a zone management operation, with
512  * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
513  * due to its different handling in the block layer and device response in
514  * case of command failure.
515  */
op_is_zone_mgmt(enum req_op op)516 static inline bool op_is_zone_mgmt(enum req_op op)
517 {
518 	switch (op & REQ_OP_MASK) {
519 	case REQ_OP_ZONE_RESET:
520 	case REQ_OP_ZONE_OPEN:
521 	case REQ_OP_ZONE_CLOSE:
522 	case REQ_OP_ZONE_FINISH:
523 		return true;
524 	default:
525 		return false;
526 	}
527 }
528 
op_stat_group(enum req_op op)529 static inline int op_stat_group(enum req_op op)
530 {
531 	if (op_is_discard(op))
532 		return STAT_DISCARD;
533 	return op_is_write(op);
534 }
535 
536 struct blk_rq_stat {
537 	u64 mean;
538 	u64 min;
539 	u64 max;
540 	u32 nr_samples;
541 	u64 batch;
542 };
543 
544 #endif /* __LINUX_BLK_TYPES_H */
545