xref: /linux-6.15/include/linux/blkdev.h (revision c5b483d5)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
4 
5 #include <linux/sched.h>
6 #include <linux/genhd.h>
7 #include <linux/list.h>
8 #include <linux/llist.h>
9 #include <linux/minmax.h>
10 #include <linux/timer.h>
11 #include <linux/workqueue.h>
12 #include <linux/wait.h>
13 #include <linux/bio.h>
14 #include <linux/gfp.h>
15 #include <linux/rcupdate.h>
16 #include <linux/percpu-refcount.h>
17 #include <linux/blkzoned.h>
18 #include <linux/sbitmap.h>
19 #include <linux/srcu.h>
20 
21 struct module;
22 struct request_queue;
23 struct elevator_queue;
24 struct blk_trace;
25 struct request;
26 struct sg_io_hdr;
27 struct blkcg_gq;
28 struct blk_flush_queue;
29 struct kiocb;
30 struct pr_ops;
31 struct rq_qos;
32 struct blk_queue_stats;
33 struct blk_stat_callback;
34 struct blk_crypto_profile;
35 
36 /* Must be consistent with blk_mq_poll_stats_bkt() */
37 #define BLK_MQ_POLL_STATS_BKTS 16
38 
39 /* Doing classic polling */
40 #define BLK_MQ_POLL_CLASSIC -1
41 
42 /*
43  * Maximum number of blkcg policies allowed to be registered concurrently.
44  * Defined here to simplify include dependency.
45  */
46 #define BLKCG_MAX_POLS		6
47 
48 static inline int blk_validate_block_size(unsigned long bsize)
49 {
50 	if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize))
51 		return -EINVAL;
52 
53 	return 0;
54 }
55 
56 static inline bool blk_op_is_passthrough(unsigned int op)
57 {
58 	op &= REQ_OP_MASK;
59 	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
60 }
61 
62 /*
63  * Zoned block device models (zoned limit).
64  *
65  * Note: This needs to be ordered from the least to the most severe
66  * restrictions for the inheritance in blk_stack_limits() to work.
67  */
68 enum blk_zoned_model {
69 	BLK_ZONED_NONE = 0,	/* Regular block device */
70 	BLK_ZONED_HA,		/* Host-aware zoned block device */
71 	BLK_ZONED_HM,		/* Host-managed zoned block device */
72 };
73 
74 /*
75  * BLK_BOUNCE_NONE:	never bounce (default)
76  * BLK_BOUNCE_HIGH:	bounce all highmem pages
77  */
78 enum blk_bounce {
79 	BLK_BOUNCE_NONE,
80 	BLK_BOUNCE_HIGH,
81 };
82 
83 struct queue_limits {
84 	enum blk_bounce		bounce;
85 	unsigned long		seg_boundary_mask;
86 	unsigned long		virt_boundary_mask;
87 
88 	unsigned int		max_hw_sectors;
89 	unsigned int		max_dev_sectors;
90 	unsigned int		chunk_sectors;
91 	unsigned int		max_sectors;
92 	unsigned int		max_segment_size;
93 	unsigned int		physical_block_size;
94 	unsigned int		logical_block_size;
95 	unsigned int		alignment_offset;
96 	unsigned int		io_min;
97 	unsigned int		io_opt;
98 	unsigned int		max_discard_sectors;
99 	unsigned int		max_hw_discard_sectors;
100 	unsigned int		max_write_zeroes_sectors;
101 	unsigned int		max_zone_append_sectors;
102 	unsigned int		discard_granularity;
103 	unsigned int		discard_alignment;
104 	unsigned int		zone_write_granularity;
105 
106 	unsigned short		max_segments;
107 	unsigned short		max_integrity_segments;
108 	unsigned short		max_discard_segments;
109 
110 	unsigned char		misaligned;
111 	unsigned char		discard_misaligned;
112 	unsigned char		raid_partial_stripes_expensive;
113 	enum blk_zoned_model	zoned;
114 };
115 
116 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
117 			       void *data);
118 
119 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
120 
121 #ifdef CONFIG_BLK_DEV_ZONED
122 
123 #define BLK_ALL_ZONES  ((unsigned int)-1)
124 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
125 			unsigned int nr_zones, report_zones_cb cb, void *data);
126 unsigned int blkdev_nr_zones(struct gendisk *disk);
127 extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
128 			    sector_t sectors, sector_t nr_sectors,
129 			    gfp_t gfp_mask);
130 int blk_revalidate_disk_zones(struct gendisk *disk,
131 			      void (*update_driver_data)(struct gendisk *disk));
132 
133 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
134 				     unsigned int cmd, unsigned long arg);
135 extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
136 				  unsigned int cmd, unsigned long arg);
137 
138 #else /* CONFIG_BLK_DEV_ZONED */
139 
140 static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
141 {
142 	return 0;
143 }
144 
145 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
146 					    fmode_t mode, unsigned int cmd,
147 					    unsigned long arg)
148 {
149 	return -ENOTTY;
150 }
151 
152 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
153 					 fmode_t mode, unsigned int cmd,
154 					 unsigned long arg)
155 {
156 	return -ENOTTY;
157 }
158 
159 #endif /* CONFIG_BLK_DEV_ZONED */
160 
161 /*
162  * Independent access ranges: struct blk_independent_access_range describes
163  * a range of contiguous sectors that can be accessed using device command
164  * execution resources that are independent from the resources used for
165  * other access ranges. This is typically found with single-LUN multi-actuator
166  * HDDs where each access range is served by a different set of heads.
167  * The set of independent ranges supported by the device is defined using
168  * struct blk_independent_access_ranges. The independent ranges must not overlap
169  * and must include all sectors within the disk capacity (no sector holes
170  * allowed).
171  * For a device with multiple ranges, requests targeting sectors in different
172  * ranges can be executed in parallel. A request can straddle an access range
173  * boundary.
174  */
175 struct blk_independent_access_range {
176 	struct kobject		kobj;
177 	struct request_queue	*queue;
178 	sector_t		sector;
179 	sector_t		nr_sectors;
180 };
181 
182 struct blk_independent_access_ranges {
183 	struct kobject				kobj;
184 	bool					sysfs_registered;
185 	unsigned int				nr_ia_ranges;
186 	struct blk_independent_access_range	ia_range[];
187 };
188 
189 struct request_queue {
190 	struct request		*last_merge;
191 	struct elevator_queue	*elevator;
192 
193 	struct percpu_ref	q_usage_counter;
194 
195 	struct blk_queue_stats	*stats;
196 	struct rq_qos		*rq_qos;
197 
198 	const struct blk_mq_ops	*mq_ops;
199 
200 	/* sw queues */
201 	struct blk_mq_ctx __percpu	*queue_ctx;
202 
203 	unsigned int		queue_depth;
204 
205 	/* hw dispatch queues */
206 	struct blk_mq_hw_ctx	**queue_hw_ctx;
207 	unsigned int		nr_hw_queues;
208 
209 	/*
210 	 * The queue owner gets to use this for whatever they like.
211 	 * ll_rw_blk doesn't touch it.
212 	 */
213 	void			*queuedata;
214 
215 	/*
216 	 * various queue flags, see QUEUE_* below
217 	 */
218 	unsigned long		queue_flags;
219 	/*
220 	 * Number of contexts that have called blk_set_pm_only(). If this
221 	 * counter is above zero then only RQF_PM requests are processed.
222 	 */
223 	atomic_t		pm_only;
224 
225 	/*
226 	 * ida allocated id for this queue.  Used to index queues from
227 	 * ioctx.
228 	 */
229 	int			id;
230 
231 	spinlock_t		queue_lock;
232 
233 	struct gendisk		*disk;
234 
235 	/*
236 	 * queue kobject
237 	 */
238 	struct kobject kobj;
239 
240 	/*
241 	 * mq queue kobject
242 	 */
243 	struct kobject *mq_kobj;
244 
245 #ifdef  CONFIG_BLK_DEV_INTEGRITY
246 	struct blk_integrity integrity;
247 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
248 
249 #ifdef CONFIG_PM
250 	struct device		*dev;
251 	enum rpm_status		rpm_status;
252 #endif
253 
254 	/*
255 	 * queue settings
256 	 */
257 	unsigned long		nr_requests;	/* Max # of requests */
258 
259 	unsigned int		dma_pad_mask;
260 	unsigned int		dma_alignment;
261 
262 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
263 	struct blk_crypto_profile *crypto_profile;
264 #endif
265 
266 	unsigned int		rq_timeout;
267 	int			poll_nsec;
268 
269 	struct blk_stat_callback	*poll_cb;
270 	struct blk_rq_stat	*poll_stat;
271 
272 	struct timer_list	timeout;
273 	struct work_struct	timeout_work;
274 
275 	atomic_t		nr_active_requests_shared_tags;
276 
277 	struct blk_mq_tags	*sched_shared_tags;
278 
279 	struct list_head	icq_list;
280 #ifdef CONFIG_BLK_CGROUP
281 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
282 	struct blkcg_gq		*root_blkg;
283 	struct list_head	blkg_list;
284 #endif
285 
286 	struct queue_limits	limits;
287 
288 	unsigned int		required_elevator_features;
289 
290 #ifdef CONFIG_BLK_DEV_ZONED
291 	/*
292 	 * Zoned block device information for request dispatch control.
293 	 * nr_zones is the total number of zones of the device. This is always
294 	 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
295 	 * bits which indicates if a zone is conventional (bit set) or
296 	 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
297 	 * bits which indicates if a zone is write locked, that is, if a write
298 	 * request targeting the zone was dispatched. All three fields are
299 	 * initialized by the low level device driver (e.g. scsi/sd.c).
300 	 * Stacking drivers (device mappers) may or may not initialize
301 	 * these fields.
302 	 *
303 	 * Reads of this information must be protected with blk_queue_enter() /
304 	 * blk_queue_exit(). Modifying this information is only allowed while
305 	 * no requests are being processed. See also blk_mq_freeze_queue() and
306 	 * blk_mq_unfreeze_queue().
307 	 */
308 	unsigned int		nr_zones;
309 	unsigned long		*conv_zones_bitmap;
310 	unsigned long		*seq_zones_wlock;
311 	unsigned int		max_open_zones;
312 	unsigned int		max_active_zones;
313 #endif /* CONFIG_BLK_DEV_ZONED */
314 
315 	int			node;
316 	struct mutex		debugfs_mutex;
317 #ifdef CONFIG_BLK_DEV_IO_TRACE
318 	struct blk_trace __rcu	*blk_trace;
319 #endif
320 	/*
321 	 * for flush operations
322 	 */
323 	struct blk_flush_queue	*fq;
324 
325 	struct list_head	requeue_list;
326 	spinlock_t		requeue_lock;
327 	struct delayed_work	requeue_work;
328 
329 	struct mutex		sysfs_lock;
330 	struct mutex		sysfs_dir_lock;
331 
332 	/*
333 	 * for reusing dead hctx instance in case of updating
334 	 * nr_hw_queues
335 	 */
336 	struct list_head	unused_hctx_list;
337 	spinlock_t		unused_hctx_lock;
338 
339 	int			mq_freeze_depth;
340 
341 #ifdef CONFIG_BLK_DEV_THROTTLING
342 	/* Throttle data */
343 	struct throtl_data *td;
344 #endif
345 	struct rcu_head		rcu_head;
346 	wait_queue_head_t	mq_freeze_wq;
347 	/*
348 	 * Protect concurrent access to q_usage_counter by
349 	 * percpu_ref_kill() and percpu_ref_reinit().
350 	 */
351 	struct mutex		mq_freeze_lock;
352 
353 	int			quiesce_depth;
354 
355 	struct blk_mq_tag_set	*tag_set;
356 	struct list_head	tag_set_list;
357 	struct bio_set		bio_split;
358 
359 	struct dentry		*debugfs_dir;
360 
361 #ifdef CONFIG_BLK_DEBUG_FS
362 	struct dentry		*sched_debugfs_dir;
363 	struct dentry		*rqos_debugfs_dir;
364 #endif
365 
366 	bool			mq_sysfs_init_done;
367 
368 #define BLK_MAX_WRITE_HINTS	5
369 	u64			write_hints[BLK_MAX_WRITE_HINTS];
370 
371 	/*
372 	 * Independent sector access ranges. This is always NULL for
373 	 * devices that do not have multiple independent access ranges.
374 	 */
375 	struct blk_independent_access_ranges *ia_ranges;
376 
377 	/**
378 	 * @srcu: Sleepable RCU. Use as lock when type of the request queue
379 	 * is blocking (BLK_MQ_F_BLOCKING). Must be the last member
380 	 */
381 	struct srcu_struct	srcu[];
382 };
383 
384 /* Keep blk_queue_flag_name[] in sync with the definitions below */
385 #define QUEUE_FLAG_STOPPED	0	/* queue is stopped */
386 #define QUEUE_FLAG_DYING	1	/* queue being torn down */
387 #define QUEUE_FLAG_HAS_SRCU	2	/* SRCU is allocated */
388 #define QUEUE_FLAG_NOMERGES     3	/* disable merge attempts */
389 #define QUEUE_FLAG_SAME_COMP	4	/* complete on same CPU-group */
390 #define QUEUE_FLAG_FAIL_IO	5	/* fake timeout */
391 #define QUEUE_FLAG_NONROT	6	/* non-rotational device (SSD) */
392 #define QUEUE_FLAG_VIRT		QUEUE_FLAG_NONROT /* paravirt device */
393 #define QUEUE_FLAG_IO_STAT	7	/* do disk/partitions IO accounting */
394 #define QUEUE_FLAG_DISCARD	8	/* supports DISCARD */
395 #define QUEUE_FLAG_NOXMERGES	9	/* No extended merges */
396 #define QUEUE_FLAG_ADD_RANDOM	10	/* Contributes to random pool */
397 #define QUEUE_FLAG_SECERASE	11	/* supports secure erase */
398 #define QUEUE_FLAG_SAME_FORCE	12	/* force complete on same CPU */
399 #define QUEUE_FLAG_DEAD		13	/* queue tear-down finished */
400 #define QUEUE_FLAG_INIT_DONE	14	/* queue is initialized */
401 #define QUEUE_FLAG_STABLE_WRITES 15	/* don't modify blks until WB is done */
402 #define QUEUE_FLAG_POLL		16	/* IO polling enabled if set */
403 #define QUEUE_FLAG_WC		17	/* Write back caching */
404 #define QUEUE_FLAG_FUA		18	/* device supports FUA writes */
405 #define QUEUE_FLAG_DAX		19	/* device supports DAX */
406 #define QUEUE_FLAG_STATS	20	/* track IO start and completion times */
407 #define QUEUE_FLAG_REGISTERED	22	/* queue has been registered to a disk */
408 #define QUEUE_FLAG_QUIESCED	24	/* queue has been quiesced */
409 #define QUEUE_FLAG_PCI_P2PDMA	25	/* device supports PCI p2p requests */
410 #define QUEUE_FLAG_ZONE_RESETALL 26	/* supports Zone Reset All */
411 #define QUEUE_FLAG_RQ_ALLOC_TIME 27	/* record rq->alloc_time_ns */
412 #define QUEUE_FLAG_HCTX_ACTIVE	28	/* at least one blk-mq hctx is active */
413 #define QUEUE_FLAG_NOWAIT       29	/* device supports NOWAIT */
414 
415 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
416 				 (1 << QUEUE_FLAG_SAME_COMP) |		\
417 				 (1 << QUEUE_FLAG_NOWAIT))
418 
419 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
420 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
421 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
422 
423 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
424 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
425 #define blk_queue_has_srcu(q)	test_bit(QUEUE_FLAG_HAS_SRCU, &(q)->queue_flags)
426 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
427 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
428 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
429 #define blk_queue_noxmerges(q)	\
430 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
431 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
432 #define blk_queue_stable_writes(q) \
433 	test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
434 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
435 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
436 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
437 #define blk_queue_zone_resetall(q)	\
438 	test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
439 #define blk_queue_secure_erase(q) \
440 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
441 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
442 #define blk_queue_pci_p2pdma(q)	\
443 	test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
444 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
445 #define blk_queue_rq_alloc_time(q)	\
446 	test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
447 #else
448 #define blk_queue_rq_alloc_time(q)	false
449 #endif
450 
451 #define blk_noretry_request(rq) \
452 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
453 			     REQ_FAILFAST_DRIVER))
454 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
455 #define blk_queue_pm_only(q)	atomic_read(&(q)->pm_only)
456 #define blk_queue_fua(q)	test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
457 #define blk_queue_registered(q)	test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
458 #define blk_queue_nowait(q)	test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
459 
460 extern void blk_set_pm_only(struct request_queue *q);
461 extern void blk_clear_pm_only(struct request_queue *q);
462 
463 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
464 
465 #define dma_map_bvec(dev, bv, dir, attrs) \
466 	dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
467 	(dir), (attrs))
468 
469 static inline bool queue_is_mq(struct request_queue *q)
470 {
471 	return q->mq_ops;
472 }
473 
474 #ifdef CONFIG_PM
475 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
476 {
477 	return q->rpm_status;
478 }
479 #else
480 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
481 {
482 	return RPM_ACTIVE;
483 }
484 #endif
485 
486 static inline enum blk_zoned_model
487 blk_queue_zoned_model(struct request_queue *q)
488 {
489 	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
490 		return q->limits.zoned;
491 	return BLK_ZONED_NONE;
492 }
493 
494 static inline bool blk_queue_is_zoned(struct request_queue *q)
495 {
496 	switch (blk_queue_zoned_model(q)) {
497 	case BLK_ZONED_HA:
498 	case BLK_ZONED_HM:
499 		return true;
500 	default:
501 		return false;
502 	}
503 }
504 
505 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
506 {
507 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
508 }
509 
510 #ifdef CONFIG_BLK_DEV_ZONED
511 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
512 {
513 	return blk_queue_is_zoned(q) ? q->nr_zones : 0;
514 }
515 
516 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
517 					     sector_t sector)
518 {
519 	if (!blk_queue_is_zoned(q))
520 		return 0;
521 	return sector >> ilog2(q->limits.chunk_sectors);
522 }
523 
524 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
525 					 sector_t sector)
526 {
527 	if (!blk_queue_is_zoned(q))
528 		return false;
529 	if (!q->conv_zones_bitmap)
530 		return true;
531 	return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
532 }
533 
534 static inline void blk_queue_max_open_zones(struct request_queue *q,
535 		unsigned int max_open_zones)
536 {
537 	q->max_open_zones = max_open_zones;
538 }
539 
540 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
541 {
542 	return q->max_open_zones;
543 }
544 
545 static inline void blk_queue_max_active_zones(struct request_queue *q,
546 		unsigned int max_active_zones)
547 {
548 	q->max_active_zones = max_active_zones;
549 }
550 
551 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
552 {
553 	return q->max_active_zones;
554 }
555 #else /* CONFIG_BLK_DEV_ZONED */
556 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
557 {
558 	return 0;
559 }
560 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
561 					 sector_t sector)
562 {
563 	return false;
564 }
565 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
566 					     sector_t sector)
567 {
568 	return 0;
569 }
570 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
571 {
572 	return 0;
573 }
574 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
575 {
576 	return 0;
577 }
578 #endif /* CONFIG_BLK_DEV_ZONED */
579 
580 static inline unsigned int blk_queue_depth(struct request_queue *q)
581 {
582 	if (q->queue_depth)
583 		return q->queue_depth;
584 
585 	return q->nr_requests;
586 }
587 
588 /*
589  * default timeout for SG_IO if none specified
590  */
591 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
592 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
593 
594 /* This should not be used directly - use rq_for_each_segment */
595 #define for_each_bio(_bio)		\
596 	for (; _bio; _bio = _bio->bi_next)
597 
598 
599 extern int blk_register_queue(struct gendisk *disk);
600 extern void blk_unregister_queue(struct gendisk *disk);
601 void submit_bio_noacct(struct bio *bio);
602 
603 extern int blk_lld_busy(struct request_queue *q);
604 extern void blk_queue_split(struct bio **);
605 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
606 extern void blk_queue_exit(struct request_queue *q);
607 extern void blk_sync_queue(struct request_queue *q);
608 
609 /* Helper to convert REQ_OP_XXX to its string format XXX */
610 extern const char *blk_op_str(unsigned int op);
611 
612 int blk_status_to_errno(blk_status_t status);
613 blk_status_t errno_to_blk_status(int errno);
614 
615 /* only poll the hardware once, don't continue until a completion was found */
616 #define BLK_POLL_ONESHOT		(1 << 0)
617 /* do not sleep to wait for the expected completion time */
618 #define BLK_POLL_NOSLEEP		(1 << 1)
619 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags);
620 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob,
621 			unsigned int flags);
622 
623 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
624 {
625 	return bdev->bd_queue;	/* this is never NULL */
626 }
627 
628 #ifdef CONFIG_BLK_DEV_ZONED
629 
630 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
631 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
632 
633 static inline unsigned int bio_zone_no(struct bio *bio)
634 {
635 	return blk_queue_zone_no(bdev_get_queue(bio->bi_bdev),
636 				 bio->bi_iter.bi_sector);
637 }
638 
639 static inline unsigned int bio_zone_is_seq(struct bio *bio)
640 {
641 	return blk_queue_zone_is_seq(bdev_get_queue(bio->bi_bdev),
642 				     bio->bi_iter.bi_sector);
643 }
644 #endif /* CONFIG_BLK_DEV_ZONED */
645 
646 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
647 						     int op)
648 {
649 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
650 		return min(q->limits.max_discard_sectors,
651 			   UINT_MAX >> SECTOR_SHIFT);
652 
653 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
654 		return q->limits.max_write_zeroes_sectors;
655 
656 	return q->limits.max_sectors;
657 }
658 
659 /*
660  * Return maximum size of a request at given offset. Only valid for
661  * file system requests.
662  */
663 static inline unsigned int blk_max_size_offset(struct request_queue *q,
664 					       sector_t offset,
665 					       unsigned int chunk_sectors)
666 {
667 	if (!chunk_sectors) {
668 		if (q->limits.chunk_sectors)
669 			chunk_sectors = q->limits.chunk_sectors;
670 		else
671 			return q->limits.max_sectors;
672 	}
673 
674 	if (likely(is_power_of_2(chunk_sectors)))
675 		chunk_sectors -= offset & (chunk_sectors - 1);
676 	else
677 		chunk_sectors -= sector_div(offset, chunk_sectors);
678 
679 	return min(q->limits.max_sectors, chunk_sectors);
680 }
681 
682 /*
683  * Access functions for manipulating queue properties
684  */
685 extern void blk_cleanup_queue(struct request_queue *);
686 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
687 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
688 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
689 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
690 extern void blk_queue_max_discard_segments(struct request_queue *,
691 		unsigned short);
692 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
693 extern void blk_queue_max_discard_sectors(struct request_queue *q,
694 		unsigned int max_discard_sectors);
695 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
696 		unsigned int max_write_same_sectors);
697 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
698 extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
699 		unsigned int max_zone_append_sectors);
700 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
701 void blk_queue_zone_write_granularity(struct request_queue *q,
702 				      unsigned int size);
703 extern void blk_queue_alignment_offset(struct request_queue *q,
704 				       unsigned int alignment);
705 void disk_update_readahead(struct gendisk *disk);
706 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
707 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
708 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
709 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
710 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
711 extern void blk_set_default_limits(struct queue_limits *lim);
712 extern void blk_set_stacking_limits(struct queue_limits *lim);
713 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
714 			    sector_t offset);
715 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
716 			      sector_t offset);
717 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
718 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
719 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
720 extern void blk_queue_dma_alignment(struct request_queue *, int);
721 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
722 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
723 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
724 
725 struct blk_independent_access_ranges *
726 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges);
727 void disk_set_independent_access_ranges(struct gendisk *disk,
728 				struct blk_independent_access_ranges *iars);
729 
730 /*
731  * Elevator features for blk_queue_required_elevator_features:
732  */
733 /* Supports zoned block devices sequential write constraint */
734 #define ELEVATOR_F_ZBD_SEQ_WRITE	(1U << 0)
735 /* Supports scheduling on multiple hardware queues */
736 #define ELEVATOR_F_MQ_AWARE		(1U << 1)
737 
738 extern void blk_queue_required_elevator_features(struct request_queue *q,
739 						 unsigned int features);
740 extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
741 					      struct device *dev);
742 
743 bool __must_check blk_get_queue(struct request_queue *);
744 extern void blk_put_queue(struct request_queue *);
745 extern void blk_set_queue_dying(struct request_queue *);
746 
747 #ifdef CONFIG_BLOCK
748 /*
749  * blk_plug permits building a queue of related requests by holding the I/O
750  * fragments for a short period. This allows merging of sequential requests
751  * into single larger request. As the requests are moved from a per-task list to
752  * the device's request_queue in a batch, this results in improved scalability
753  * as the lock contention for request_queue lock is reduced.
754  *
755  * It is ok not to disable preemption when adding the request to the plug list
756  * or when attempting a merge. For details, please see schedule() where
757  * blk_flush_plug() is called.
758  */
759 struct blk_plug {
760 	struct request *mq_list; /* blk-mq requests */
761 
762 	/* if ios_left is > 1, we can batch tag/rq allocations */
763 	struct request *cached_rq;
764 	unsigned short nr_ios;
765 
766 	unsigned short rq_count;
767 
768 	bool multiple_queues;
769 	bool has_elevator;
770 	bool nowait;
771 
772 	struct list_head cb_list; /* md requires an unplug callback */
773 };
774 
775 struct blk_plug_cb;
776 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
777 struct blk_plug_cb {
778 	struct list_head list;
779 	blk_plug_cb_fn callback;
780 	void *data;
781 };
782 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
783 					     void *data, int size);
784 extern void blk_start_plug(struct blk_plug *);
785 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short);
786 extern void blk_finish_plug(struct blk_plug *);
787 
788 void blk_flush_plug(struct blk_plug *plug, bool from_schedule);
789 
790 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
791 {
792 	struct blk_plug *plug = tsk->plug;
793 
794 	return plug &&
795 		 (plug->mq_list || !list_empty(&plug->cb_list));
796 }
797 
798 int blkdev_issue_flush(struct block_device *bdev);
799 long nr_blockdev_pages(void);
800 #else /* CONFIG_BLOCK */
801 struct blk_plug {
802 };
803 
804 static inline void blk_start_plug_nr_ios(struct blk_plug *plug,
805 					 unsigned short nr_ios)
806 {
807 }
808 
809 static inline void blk_start_plug(struct blk_plug *plug)
810 {
811 }
812 
813 static inline void blk_finish_plug(struct blk_plug *plug)
814 {
815 }
816 
817 static inline void blk_flush_plug(struct blk_plug *plug, bool async)
818 {
819 }
820 
821 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
822 {
823 	return false;
824 }
825 
826 static inline int blkdev_issue_flush(struct block_device *bdev)
827 {
828 	return 0;
829 }
830 
831 static inline long nr_blockdev_pages(void)
832 {
833 	return 0;
834 }
835 #endif /* CONFIG_BLOCK */
836 
837 extern void blk_io_schedule(void);
838 
839 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
840 
841 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
842 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
843 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
844 		sector_t nr_sects, gfp_t gfp_mask, int flags,
845 		struct bio **biop);
846 
847 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
848 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
849 
850 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
851 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
852 		unsigned flags);
853 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
854 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
855 
856 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
857 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
858 {
859 	return blkdev_issue_discard(sb->s_bdev,
860 				    block << (sb->s_blocksize_bits -
861 					      SECTOR_SHIFT),
862 				    nr_blocks << (sb->s_blocksize_bits -
863 						  SECTOR_SHIFT),
864 				    gfp_mask, flags);
865 }
866 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
867 		sector_t nr_blocks, gfp_t gfp_mask)
868 {
869 	return blkdev_issue_zeroout(sb->s_bdev,
870 				    block << (sb->s_blocksize_bits -
871 					      SECTOR_SHIFT),
872 				    nr_blocks << (sb->s_blocksize_bits -
873 						  SECTOR_SHIFT),
874 				    gfp_mask, 0);
875 }
876 
877 static inline bool bdev_is_partition(struct block_device *bdev)
878 {
879 	return bdev->bd_partno;
880 }
881 
882 enum blk_default_limits {
883 	BLK_MAX_SEGMENTS	= 128,
884 	BLK_SAFE_MAX_SECTORS	= 255,
885 	BLK_DEF_MAX_SECTORS	= 2560,
886 	BLK_MAX_SEGMENT_SIZE	= 65536,
887 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
888 };
889 
890 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
891 {
892 	return q->limits.seg_boundary_mask;
893 }
894 
895 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
896 {
897 	return q->limits.virt_boundary_mask;
898 }
899 
900 static inline unsigned int queue_max_sectors(const struct request_queue *q)
901 {
902 	return q->limits.max_sectors;
903 }
904 
905 static inline unsigned int queue_max_bytes(struct request_queue *q)
906 {
907 	return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9;
908 }
909 
910 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
911 {
912 	return q->limits.max_hw_sectors;
913 }
914 
915 static inline unsigned short queue_max_segments(const struct request_queue *q)
916 {
917 	return q->limits.max_segments;
918 }
919 
920 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
921 {
922 	return q->limits.max_discard_segments;
923 }
924 
925 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
926 {
927 	return q->limits.max_segment_size;
928 }
929 
930 static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
931 {
932 
933 	const struct queue_limits *l = &q->limits;
934 
935 	return min(l->max_zone_append_sectors, l->max_sectors);
936 }
937 
938 static inline unsigned queue_logical_block_size(const struct request_queue *q)
939 {
940 	int retval = 512;
941 
942 	if (q && q->limits.logical_block_size)
943 		retval = q->limits.logical_block_size;
944 
945 	return retval;
946 }
947 
948 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
949 {
950 	return queue_logical_block_size(bdev_get_queue(bdev));
951 }
952 
953 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
954 {
955 	return q->limits.physical_block_size;
956 }
957 
958 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
959 {
960 	return queue_physical_block_size(bdev_get_queue(bdev));
961 }
962 
963 static inline unsigned int queue_io_min(const struct request_queue *q)
964 {
965 	return q->limits.io_min;
966 }
967 
968 static inline int bdev_io_min(struct block_device *bdev)
969 {
970 	return queue_io_min(bdev_get_queue(bdev));
971 }
972 
973 static inline unsigned int queue_io_opt(const struct request_queue *q)
974 {
975 	return q->limits.io_opt;
976 }
977 
978 static inline int bdev_io_opt(struct block_device *bdev)
979 {
980 	return queue_io_opt(bdev_get_queue(bdev));
981 }
982 
983 static inline unsigned int
984 queue_zone_write_granularity(const struct request_queue *q)
985 {
986 	return q->limits.zone_write_granularity;
987 }
988 
989 static inline unsigned int
990 bdev_zone_write_granularity(struct block_device *bdev)
991 {
992 	return queue_zone_write_granularity(bdev_get_queue(bdev));
993 }
994 
995 static inline int queue_alignment_offset(const struct request_queue *q)
996 {
997 	if (q->limits.misaligned)
998 		return -1;
999 
1000 	return q->limits.alignment_offset;
1001 }
1002 
1003 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1004 {
1005 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1006 	unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1007 		<< SECTOR_SHIFT;
1008 
1009 	return (granularity + lim->alignment_offset - alignment) % granularity;
1010 }
1011 
1012 static inline int bdev_alignment_offset(struct block_device *bdev)
1013 {
1014 	struct request_queue *q = bdev_get_queue(bdev);
1015 
1016 	if (q->limits.misaligned)
1017 		return -1;
1018 	if (bdev_is_partition(bdev))
1019 		return queue_limit_alignment_offset(&q->limits,
1020 				bdev->bd_start_sect);
1021 	return q->limits.alignment_offset;
1022 }
1023 
1024 static inline int queue_discard_alignment(const struct request_queue *q)
1025 {
1026 	if (q->limits.discard_misaligned)
1027 		return -1;
1028 
1029 	return q->limits.discard_alignment;
1030 }
1031 
1032 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1033 {
1034 	unsigned int alignment, granularity, offset;
1035 
1036 	if (!lim->max_discard_sectors)
1037 		return 0;
1038 
1039 	/* Why are these in bytes, not sectors? */
1040 	alignment = lim->discard_alignment >> SECTOR_SHIFT;
1041 	granularity = lim->discard_granularity >> SECTOR_SHIFT;
1042 	if (!granularity)
1043 		return 0;
1044 
1045 	/* Offset of the partition start in 'granularity' sectors */
1046 	offset = sector_div(sector, granularity);
1047 
1048 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1049 	offset = (granularity + alignment - offset) % granularity;
1050 
1051 	/* Turn it back into bytes, gaah */
1052 	return offset << SECTOR_SHIFT;
1053 }
1054 
1055 static inline int bdev_discard_alignment(struct block_device *bdev)
1056 {
1057 	struct request_queue *q = bdev_get_queue(bdev);
1058 
1059 	if (bdev_is_partition(bdev))
1060 		return queue_limit_discard_alignment(&q->limits,
1061 				bdev->bd_start_sect);
1062 	return q->limits.discard_alignment;
1063 }
1064 
1065 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1066 {
1067 	struct request_queue *q = bdev_get_queue(bdev);
1068 
1069 	if (q)
1070 		return q->limits.max_write_zeroes_sectors;
1071 
1072 	return 0;
1073 }
1074 
1075 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1076 {
1077 	struct request_queue *q = bdev_get_queue(bdev);
1078 
1079 	if (q)
1080 		return blk_queue_zoned_model(q);
1081 
1082 	return BLK_ZONED_NONE;
1083 }
1084 
1085 static inline bool bdev_is_zoned(struct block_device *bdev)
1086 {
1087 	struct request_queue *q = bdev_get_queue(bdev);
1088 
1089 	if (q)
1090 		return blk_queue_is_zoned(q);
1091 
1092 	return false;
1093 }
1094 
1095 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1096 {
1097 	struct request_queue *q = bdev_get_queue(bdev);
1098 
1099 	if (q)
1100 		return blk_queue_zone_sectors(q);
1101 	return 0;
1102 }
1103 
1104 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1105 {
1106 	struct request_queue *q = bdev_get_queue(bdev);
1107 
1108 	if (q)
1109 		return queue_max_open_zones(q);
1110 	return 0;
1111 }
1112 
1113 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1114 {
1115 	struct request_queue *q = bdev_get_queue(bdev);
1116 
1117 	if (q)
1118 		return queue_max_active_zones(q);
1119 	return 0;
1120 }
1121 
1122 static inline int queue_dma_alignment(const struct request_queue *q)
1123 {
1124 	return q ? q->dma_alignment : 511;
1125 }
1126 
1127 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1128 				 unsigned int len)
1129 {
1130 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1131 	return !(addr & alignment) && !(len & alignment);
1132 }
1133 
1134 /* assumes size > 256 */
1135 static inline unsigned int blksize_bits(unsigned int size)
1136 {
1137 	unsigned int bits = 8;
1138 	do {
1139 		bits++;
1140 		size >>= 1;
1141 	} while (size > 256);
1142 	return bits;
1143 }
1144 
1145 static inline unsigned int block_size(struct block_device *bdev)
1146 {
1147 	return 1 << bdev->bd_inode->i_blkbits;
1148 }
1149 
1150 int kblockd_schedule_work(struct work_struct *work);
1151 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1152 
1153 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1154 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1155 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1156 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1157 
1158 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1159 
1160 bool blk_crypto_register(struct blk_crypto_profile *profile,
1161 			 struct request_queue *q);
1162 
1163 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1164 
1165 static inline bool blk_crypto_register(struct blk_crypto_profile *profile,
1166 				       struct request_queue *q)
1167 {
1168 	return true;
1169 }
1170 
1171 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1172 
1173 enum blk_unique_id {
1174 	/* these match the Designator Types specified in SPC */
1175 	BLK_UID_T10	= 1,
1176 	BLK_UID_EUI64	= 2,
1177 	BLK_UID_NAA	= 3,
1178 };
1179 
1180 #define NFL4_UFLG_MASK			0x0000003F
1181 
1182 struct block_device_operations {
1183 	void (*submit_bio)(struct bio *bio);
1184 	int (*open) (struct block_device *, fmode_t);
1185 	void (*release) (struct gendisk *, fmode_t);
1186 	int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1187 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1188 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1189 	unsigned int (*check_events) (struct gendisk *disk,
1190 				      unsigned int clearing);
1191 	void (*unlock_native_capacity) (struct gendisk *);
1192 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1193 	int (*set_read_only)(struct block_device *bdev, bool ro);
1194 	/* this callback is with swap_lock and sometimes page table lock held */
1195 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1196 	int (*report_zones)(struct gendisk *, sector_t sector,
1197 			unsigned int nr_zones, report_zones_cb cb, void *data);
1198 	char *(*devnode)(struct gendisk *disk, umode_t *mode);
1199 	/* returns the length of the identifier or a negative errno: */
1200 	int (*get_unique_id)(struct gendisk *disk, u8 id[16],
1201 			enum blk_unique_id id_type);
1202 	struct module *owner;
1203 	const struct pr_ops *pr_ops;
1204 
1205 	/*
1206 	 * Special callback for probing GPT entry at a given sector.
1207 	 * Needed by Android devices, used by GPT scanner and MMC blk
1208 	 * driver.
1209 	 */
1210 	int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector);
1211 };
1212 
1213 #ifdef CONFIG_COMPAT
1214 extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1215 				      unsigned int, unsigned long);
1216 #else
1217 #define blkdev_compat_ptr_ioctl NULL
1218 #endif
1219 
1220 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1221 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1222 						struct writeback_control *);
1223 
1224 static inline void blk_wake_io_task(struct task_struct *waiter)
1225 {
1226 	/*
1227 	 * If we're polling, the task itself is doing the completions. For
1228 	 * that case, we don't need to signal a wakeup, it's enough to just
1229 	 * mark us as RUNNING.
1230 	 */
1231 	if (waiter == current)
1232 		__set_current_state(TASK_RUNNING);
1233 	else
1234 		wake_up_process(waiter);
1235 }
1236 
1237 unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1238 		unsigned int op);
1239 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1240 		unsigned long start_time);
1241 
1242 unsigned long bio_start_io_acct(struct bio *bio);
1243 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1244 		struct block_device *orig_bdev);
1245 
1246 /**
1247  * bio_end_io_acct - end I/O accounting for bio based drivers
1248  * @bio:	bio to end account for
1249  * @start:	start time returned by bio_start_io_acct()
1250  */
1251 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1252 {
1253 	return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1254 }
1255 
1256 int bdev_read_only(struct block_device *bdev);
1257 int set_blocksize(struct block_device *bdev, int size);
1258 
1259 const char *bdevname(struct block_device *bdev, char *buffer);
1260 int lookup_bdev(const char *pathname, dev_t *dev);
1261 
1262 void blkdev_show(struct seq_file *seqf, off_t offset);
1263 
1264 #define BDEVNAME_SIZE	32	/* Largest string for a blockdev identifier */
1265 #define BDEVT_SIZE	10	/* Largest string for MAJ:MIN for blkdev */
1266 #ifdef CONFIG_BLOCK
1267 #define BLKDEV_MAJOR_MAX	512
1268 #else
1269 #define BLKDEV_MAJOR_MAX	0
1270 #endif
1271 
1272 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1273 		void *holder);
1274 struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
1275 int bd_prepare_to_claim(struct block_device *bdev, void *holder);
1276 void bd_abort_claiming(struct block_device *bdev, void *holder);
1277 void blkdev_put(struct block_device *bdev, fmode_t mode);
1278 
1279 /* just for blk-cgroup, don't use elsewhere */
1280 struct block_device *blkdev_get_no_open(dev_t dev);
1281 void blkdev_put_no_open(struct block_device *bdev);
1282 
1283 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
1284 void bdev_add(struct block_device *bdev, dev_t dev);
1285 struct block_device *I_BDEV(struct inode *inode);
1286 int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
1287 		loff_t lend);
1288 
1289 #ifdef CONFIG_BLOCK
1290 void invalidate_bdev(struct block_device *bdev);
1291 int sync_blockdev(struct block_device *bdev);
1292 int sync_blockdev_nowait(struct block_device *bdev);
1293 void sync_bdevs(bool wait);
1294 #else
1295 static inline void invalidate_bdev(struct block_device *bdev)
1296 {
1297 }
1298 static inline int sync_blockdev(struct block_device *bdev)
1299 {
1300 	return 0;
1301 }
1302 static inline int sync_blockdev_nowait(struct block_device *bdev)
1303 {
1304 	return 0;
1305 }
1306 static inline void sync_bdevs(bool wait)
1307 {
1308 }
1309 #endif
1310 int fsync_bdev(struct block_device *bdev);
1311 
1312 int freeze_bdev(struct block_device *bdev);
1313 int thaw_bdev(struct block_device *bdev);
1314 
1315 struct io_comp_batch {
1316 	struct request *req_list;
1317 	bool need_ts;
1318 	void (*complete)(struct io_comp_batch *);
1319 };
1320 
1321 #define DEFINE_IO_COMP_BATCH(name)	struct io_comp_batch name = { }
1322 
1323 #endif /* _LINUX_BLKDEV_H */
1324