xref: /linux-6.15/include/linux/blkdev.h (revision 54404d35)
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/sched/clock.h>
7 #include <linux/major.h>
8 #include <linux/genhd.h>
9 #include <linux/list.h>
10 #include <linux/llist.h>
11 #include <linux/minmax.h>
12 #include <linux/timer.h>
13 #include <linux/workqueue.h>
14 #include <linux/backing-dev-defs.h>
15 #include <linux/wait.h>
16 #include <linux/mempool.h>
17 #include <linux/pfn.h>
18 #include <linux/bio.h>
19 #include <linux/stringify.h>
20 #include <linux/gfp.h>
21 #include <linux/smp.h>
22 #include <linux/rcupdate.h>
23 #include <linux/percpu-refcount.h>
24 #include <linux/scatterlist.h>
25 #include <linux/blkzoned.h>
26 #include <linux/pm.h>
27 #include <linux/sbitmap.h>
28 
29 struct module;
30 struct request_queue;
31 struct elevator_queue;
32 struct blk_trace;
33 struct request;
34 struct sg_io_hdr;
35 struct blkcg_gq;
36 struct blk_flush_queue;
37 struct pr_ops;
38 struct rq_qos;
39 struct blk_queue_stats;
40 struct blk_stat_callback;
41 struct blk_keyslot_manager;
42 
43 #define BLKDEV_MIN_RQ	4
44 #define BLKDEV_MAX_RQ	128	/* Default maximum */
45 
46 /* Must be consistent with blk_mq_poll_stats_bkt() */
47 #define BLK_MQ_POLL_STATS_BKTS 16
48 
49 /* Doing classic polling */
50 #define BLK_MQ_POLL_CLASSIC -1
51 
52 /*
53  * Maximum number of blkcg policies allowed to be registered concurrently.
54  * Defined here to simplify include dependency.
55  */
56 #define BLKCG_MAX_POLS		5
57 
58 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
59 
60 /*
61  * request flags */
62 typedef __u32 __bitwise req_flags_t;
63 
64 /* drive already may have started this one */
65 #define RQF_STARTED		((__force req_flags_t)(1 << 1))
66 /* may not be passed by ioscheduler */
67 #define RQF_SOFTBARRIER		((__force req_flags_t)(1 << 3))
68 /* request for flush sequence */
69 #define RQF_FLUSH_SEQ		((__force req_flags_t)(1 << 4))
70 /* merge of different types, fail separately */
71 #define RQF_MIXED_MERGE		((__force req_flags_t)(1 << 5))
72 /* track inflight for MQ */
73 #define RQF_MQ_INFLIGHT		((__force req_flags_t)(1 << 6))
74 /* don't call prep for this one */
75 #define RQF_DONTPREP		((__force req_flags_t)(1 << 7))
76 /* vaguely specified driver internal error.  Ignored by the block layer */
77 #define RQF_FAILED		((__force req_flags_t)(1 << 10))
78 /* don't warn about errors */
79 #define RQF_QUIET		((__force req_flags_t)(1 << 11))
80 /* elevator private data attached */
81 #define RQF_ELVPRIV		((__force req_flags_t)(1 << 12))
82 /* account into disk and partition IO statistics */
83 #define RQF_IO_STAT		((__force req_flags_t)(1 << 13))
84 /* runtime pm request */
85 #define RQF_PM			((__force req_flags_t)(1 << 15))
86 /* on IO scheduler merge hash */
87 #define RQF_HASHED		((__force req_flags_t)(1 << 16))
88 /* track IO completion time */
89 #define RQF_STATS		((__force req_flags_t)(1 << 17))
90 /* Look at ->special_vec for the actual data payload instead of the
91    bio chain. */
92 #define RQF_SPECIAL_PAYLOAD	((__force req_flags_t)(1 << 18))
93 /* The per-zone write lock is held for this request */
94 #define RQF_ZONE_WRITE_LOCKED	((__force req_flags_t)(1 << 19))
95 /* already slept for hybrid poll */
96 #define RQF_MQ_POLL_SLEPT	((__force req_flags_t)(1 << 20))
97 /* ->timeout has been called, don't expire again */
98 #define RQF_TIMED_OUT		((__force req_flags_t)(1 << 21))
99 
100 /* flags that prevent us from merging requests: */
101 #define RQF_NOMERGE_FLAGS \
102 	(RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
103 
104 /*
105  * Request state for blk-mq.
106  */
107 enum mq_rq_state {
108 	MQ_RQ_IDLE		= 0,
109 	MQ_RQ_IN_FLIGHT		= 1,
110 	MQ_RQ_COMPLETE		= 2,
111 };
112 
113 /*
114  * Try to put the fields that are referenced together in the same cacheline.
115  *
116  * If you modify this structure, make sure to update blk_rq_init() and
117  * especially blk_mq_rq_ctx_init() to take care of the added fields.
118  */
119 struct request {
120 	struct request_queue *q;
121 	struct blk_mq_ctx *mq_ctx;
122 	struct blk_mq_hw_ctx *mq_hctx;
123 
124 	unsigned int cmd_flags;		/* op and common flags */
125 	req_flags_t rq_flags;
126 
127 	int tag;
128 	int internal_tag;
129 
130 	/* the following two fields are internal, NEVER access directly */
131 	unsigned int __data_len;	/* total data len */
132 	sector_t __sector;		/* sector cursor */
133 
134 	struct bio *bio;
135 	struct bio *biotail;
136 
137 	struct list_head queuelist;
138 
139 	/*
140 	 * The hash is used inside the scheduler, and killed once the
141 	 * request reaches the dispatch list. The ipi_list is only used
142 	 * to queue the request for softirq completion, which is long
143 	 * after the request has been unhashed (and even removed from
144 	 * the dispatch list).
145 	 */
146 	union {
147 		struct hlist_node hash;	/* merge hash */
148 		struct llist_node ipi_list;
149 	};
150 
151 	/*
152 	 * The rb_node is only used inside the io scheduler, requests
153 	 * are pruned when moved to the dispatch queue. So let the
154 	 * completion_data share space with the rb_node.
155 	 */
156 	union {
157 		struct rb_node rb_node;	/* sort/lookup */
158 		struct bio_vec special_vec;
159 		void *completion_data;
160 		int error_count; /* for legacy drivers, don't use */
161 	};
162 
163 	/*
164 	 * Three pointers are available for the IO schedulers, if they need
165 	 * more they have to dynamically allocate it.  Flush requests are
166 	 * never put on the IO scheduler. So let the flush fields share
167 	 * space with the elevator data.
168 	 */
169 	union {
170 		struct {
171 			struct io_cq		*icq;
172 			void			*priv[2];
173 		} elv;
174 
175 		struct {
176 			unsigned int		seq;
177 			struct list_head	list;
178 			rq_end_io_fn		*saved_end_io;
179 		} flush;
180 	};
181 
182 	struct gendisk *rq_disk;
183 	struct block_device *part;
184 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
185 	/* Time that the first bio started allocating this request. */
186 	u64 alloc_time_ns;
187 #endif
188 	/* Time that this request was allocated for this IO. */
189 	u64 start_time_ns;
190 	/* Time that I/O was submitted to the device. */
191 	u64 io_start_time_ns;
192 
193 #ifdef CONFIG_BLK_WBT
194 	unsigned short wbt_flags;
195 #endif
196 	/*
197 	 * rq sectors used for blk stats. It has the same value
198 	 * with blk_rq_sectors(rq), except that it never be zeroed
199 	 * by completion.
200 	 */
201 	unsigned short stats_sectors;
202 
203 	/*
204 	 * Number of scatter-gather DMA addr+len pairs after
205 	 * physical address coalescing is performed.
206 	 */
207 	unsigned short nr_phys_segments;
208 
209 #if defined(CONFIG_BLK_DEV_INTEGRITY)
210 	unsigned short nr_integrity_segments;
211 #endif
212 
213 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
214 	struct bio_crypt_ctx *crypt_ctx;
215 	struct blk_ksm_keyslot *crypt_keyslot;
216 #endif
217 
218 	unsigned short write_hint;
219 	unsigned short ioprio;
220 
221 	enum mq_rq_state state;
222 	refcount_t ref;
223 
224 	unsigned int timeout;
225 	unsigned long deadline;
226 
227 	union {
228 		struct __call_single_data csd;
229 		u64 fifo_time;
230 	};
231 
232 	/*
233 	 * completion callback.
234 	 */
235 	rq_end_io_fn *end_io;
236 	void *end_io_data;
237 };
238 
239 static inline bool blk_op_is_passthrough(unsigned int op)
240 {
241 	op &= REQ_OP_MASK;
242 	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
243 }
244 
245 static inline bool blk_rq_is_passthrough(struct request *rq)
246 {
247 	return blk_op_is_passthrough(req_op(rq));
248 }
249 
250 static inline unsigned short req_get_ioprio(struct request *req)
251 {
252 	return req->ioprio;
253 }
254 
255 #include <linux/elevator.h>
256 
257 struct blk_queue_ctx;
258 
259 struct bio_vec;
260 
261 enum blk_eh_timer_return {
262 	BLK_EH_DONE,		/* drivers has completed the command */
263 	BLK_EH_RESET_TIMER,	/* reset timer and try again */
264 };
265 
266 enum blk_queue_state {
267 	Queue_down,
268 	Queue_up,
269 };
270 
271 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
272 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
273 
274 /*
275  * Zoned block device models (zoned limit).
276  *
277  * Note: This needs to be ordered from the least to the most severe
278  * restrictions for the inheritance in blk_stack_limits() to work.
279  */
280 enum blk_zoned_model {
281 	BLK_ZONED_NONE = 0,	/* Regular block device */
282 	BLK_ZONED_HA,		/* Host-aware zoned block device */
283 	BLK_ZONED_HM,		/* Host-managed zoned block device */
284 };
285 
286 /*
287  * BLK_BOUNCE_NONE:	never bounce (default)
288  * BLK_BOUNCE_HIGH:	bounce all highmem pages
289  */
290 enum blk_bounce {
291 	BLK_BOUNCE_NONE,
292 	BLK_BOUNCE_HIGH,
293 };
294 
295 struct queue_limits {
296 	enum blk_bounce		bounce;
297 	unsigned long		seg_boundary_mask;
298 	unsigned long		virt_boundary_mask;
299 
300 	unsigned int		max_hw_sectors;
301 	unsigned int		max_dev_sectors;
302 	unsigned int		chunk_sectors;
303 	unsigned int		max_sectors;
304 	unsigned int		max_segment_size;
305 	unsigned int		physical_block_size;
306 	unsigned int		logical_block_size;
307 	unsigned int		alignment_offset;
308 	unsigned int		io_min;
309 	unsigned int		io_opt;
310 	unsigned int		max_discard_sectors;
311 	unsigned int		max_hw_discard_sectors;
312 	unsigned int		max_write_same_sectors;
313 	unsigned int		max_write_zeroes_sectors;
314 	unsigned int		max_zone_append_sectors;
315 	unsigned int		discard_granularity;
316 	unsigned int		discard_alignment;
317 	unsigned int		zone_write_granularity;
318 
319 	unsigned short		max_segments;
320 	unsigned short		max_integrity_segments;
321 	unsigned short		max_discard_segments;
322 
323 	unsigned char		misaligned;
324 	unsigned char		discard_misaligned;
325 	unsigned char		raid_partial_stripes_expensive;
326 	enum blk_zoned_model	zoned;
327 };
328 
329 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
330 			       void *data);
331 
332 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
333 
334 #ifdef CONFIG_BLK_DEV_ZONED
335 
336 #define BLK_ALL_ZONES  ((unsigned int)-1)
337 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
338 			unsigned int nr_zones, report_zones_cb cb, void *data);
339 unsigned int blkdev_nr_zones(struct gendisk *disk);
340 extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
341 			    sector_t sectors, sector_t nr_sectors,
342 			    gfp_t gfp_mask);
343 int blk_revalidate_disk_zones(struct gendisk *disk,
344 			      void (*update_driver_data)(struct gendisk *disk));
345 
346 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
347 				     unsigned int cmd, unsigned long arg);
348 extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
349 				  unsigned int cmd, unsigned long arg);
350 
351 #else /* CONFIG_BLK_DEV_ZONED */
352 
353 static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
354 {
355 	return 0;
356 }
357 
358 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
359 					    fmode_t mode, unsigned int cmd,
360 					    unsigned long arg)
361 {
362 	return -ENOTTY;
363 }
364 
365 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
366 					 fmode_t mode, unsigned int cmd,
367 					 unsigned long arg)
368 {
369 	return -ENOTTY;
370 }
371 
372 #endif /* CONFIG_BLK_DEV_ZONED */
373 
374 struct request_queue {
375 	struct request		*last_merge;
376 	struct elevator_queue	*elevator;
377 
378 	struct percpu_ref	q_usage_counter;
379 
380 	struct blk_queue_stats	*stats;
381 	struct rq_qos		*rq_qos;
382 
383 	const struct blk_mq_ops	*mq_ops;
384 
385 	/* sw queues */
386 	struct blk_mq_ctx __percpu	*queue_ctx;
387 
388 	unsigned int		queue_depth;
389 
390 	/* hw dispatch queues */
391 	struct blk_mq_hw_ctx	**queue_hw_ctx;
392 	unsigned int		nr_hw_queues;
393 
394 	struct backing_dev_info	*backing_dev_info;
395 
396 	/*
397 	 * The queue owner gets to use this for whatever they like.
398 	 * ll_rw_blk doesn't touch it.
399 	 */
400 	void			*queuedata;
401 
402 	/*
403 	 * various queue flags, see QUEUE_* below
404 	 */
405 	unsigned long		queue_flags;
406 	/*
407 	 * Number of contexts that have called blk_set_pm_only(). If this
408 	 * counter is above zero then only RQF_PM requests are processed.
409 	 */
410 	atomic_t		pm_only;
411 
412 	/*
413 	 * ida allocated id for this queue.  Used to index queues from
414 	 * ioctx.
415 	 */
416 	int			id;
417 
418 	spinlock_t		queue_lock;
419 
420 	/*
421 	 * queue kobject
422 	 */
423 	struct kobject kobj;
424 
425 	/*
426 	 * mq queue kobject
427 	 */
428 	struct kobject *mq_kobj;
429 
430 #ifdef  CONFIG_BLK_DEV_INTEGRITY
431 	struct blk_integrity integrity;
432 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
433 
434 #ifdef CONFIG_PM
435 	struct device		*dev;
436 	enum rpm_status		rpm_status;
437 #endif
438 
439 	/*
440 	 * queue settings
441 	 */
442 	unsigned long		nr_requests;	/* Max # of requests */
443 
444 	unsigned int		dma_pad_mask;
445 	unsigned int		dma_alignment;
446 
447 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
448 	/* Inline crypto capabilities */
449 	struct blk_keyslot_manager *ksm;
450 #endif
451 
452 	unsigned int		rq_timeout;
453 	int			poll_nsec;
454 
455 	struct blk_stat_callback	*poll_cb;
456 	struct blk_rq_stat	poll_stat[BLK_MQ_POLL_STATS_BKTS];
457 
458 	struct timer_list	timeout;
459 	struct work_struct	timeout_work;
460 
461 	atomic_t		nr_active_requests_shared_sbitmap;
462 
463 	struct sbitmap_queue	sched_bitmap_tags;
464 	struct sbitmap_queue	sched_breserved_tags;
465 
466 	struct list_head	icq_list;
467 #ifdef CONFIG_BLK_CGROUP
468 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
469 	struct blkcg_gq		*root_blkg;
470 	struct list_head	blkg_list;
471 #endif
472 
473 	struct queue_limits	limits;
474 
475 	unsigned int		required_elevator_features;
476 
477 #ifdef CONFIG_BLK_DEV_ZONED
478 	/*
479 	 * Zoned block device information for request dispatch control.
480 	 * nr_zones is the total number of zones of the device. This is always
481 	 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
482 	 * bits which indicates if a zone is conventional (bit set) or
483 	 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
484 	 * bits which indicates if a zone is write locked, that is, if a write
485 	 * request targeting the zone was dispatched. All three fields are
486 	 * initialized by the low level device driver (e.g. scsi/sd.c).
487 	 * Stacking drivers (device mappers) may or may not initialize
488 	 * these fields.
489 	 *
490 	 * Reads of this information must be protected with blk_queue_enter() /
491 	 * blk_queue_exit(). Modifying this information is only allowed while
492 	 * no requests are being processed. See also blk_mq_freeze_queue() and
493 	 * blk_mq_unfreeze_queue().
494 	 */
495 	unsigned int		nr_zones;
496 	unsigned long		*conv_zones_bitmap;
497 	unsigned long		*seq_zones_wlock;
498 	unsigned int		max_open_zones;
499 	unsigned int		max_active_zones;
500 #endif /* CONFIG_BLK_DEV_ZONED */
501 
502 	int			node;
503 	struct mutex		debugfs_mutex;
504 #ifdef CONFIG_BLK_DEV_IO_TRACE
505 	struct blk_trace __rcu	*blk_trace;
506 #endif
507 	/*
508 	 * for flush operations
509 	 */
510 	struct blk_flush_queue	*fq;
511 
512 	struct list_head	requeue_list;
513 	spinlock_t		requeue_lock;
514 	struct delayed_work	requeue_work;
515 
516 	struct mutex		sysfs_lock;
517 	struct mutex		sysfs_dir_lock;
518 
519 	/*
520 	 * for reusing dead hctx instance in case of updating
521 	 * nr_hw_queues
522 	 */
523 	struct list_head	unused_hctx_list;
524 	spinlock_t		unused_hctx_lock;
525 
526 	int			mq_freeze_depth;
527 
528 #ifdef CONFIG_BLK_DEV_THROTTLING
529 	/* Throttle data */
530 	struct throtl_data *td;
531 #endif
532 	struct rcu_head		rcu_head;
533 	wait_queue_head_t	mq_freeze_wq;
534 	/*
535 	 * Protect concurrent access to q_usage_counter by
536 	 * percpu_ref_kill() and percpu_ref_reinit().
537 	 */
538 	struct mutex		mq_freeze_lock;
539 
540 	struct blk_mq_tag_set	*tag_set;
541 	struct list_head	tag_set_list;
542 	struct bio_set		bio_split;
543 
544 	struct dentry		*debugfs_dir;
545 
546 #ifdef CONFIG_BLK_DEBUG_FS
547 	struct dentry		*sched_debugfs_dir;
548 	struct dentry		*rqos_debugfs_dir;
549 #endif
550 
551 	bool			mq_sysfs_init_done;
552 
553 	size_t			cmd_size;
554 
555 #define BLK_MAX_WRITE_HINTS	5
556 	u64			write_hints[BLK_MAX_WRITE_HINTS];
557 };
558 
559 /* Keep blk_queue_flag_name[] in sync with the definitions below */
560 #define QUEUE_FLAG_STOPPED	0	/* queue is stopped */
561 #define QUEUE_FLAG_DYING	1	/* queue being torn down */
562 #define QUEUE_FLAG_NOMERGES     3	/* disable merge attempts */
563 #define QUEUE_FLAG_SAME_COMP	4	/* complete on same CPU-group */
564 #define QUEUE_FLAG_FAIL_IO	5	/* fake timeout */
565 #define QUEUE_FLAG_NONROT	6	/* non-rotational device (SSD) */
566 #define QUEUE_FLAG_VIRT		QUEUE_FLAG_NONROT /* paravirt device */
567 #define QUEUE_FLAG_IO_STAT	7	/* do disk/partitions IO accounting */
568 #define QUEUE_FLAG_DISCARD	8	/* supports DISCARD */
569 #define QUEUE_FLAG_NOXMERGES	9	/* No extended merges */
570 #define QUEUE_FLAG_ADD_RANDOM	10	/* Contributes to random pool */
571 #define QUEUE_FLAG_SECERASE	11	/* supports secure erase */
572 #define QUEUE_FLAG_SAME_FORCE	12	/* force complete on same CPU */
573 #define QUEUE_FLAG_DEAD		13	/* queue tear-down finished */
574 #define QUEUE_FLAG_INIT_DONE	14	/* queue is initialized */
575 #define QUEUE_FLAG_STABLE_WRITES 15	/* don't modify blks until WB is done */
576 #define QUEUE_FLAG_POLL		16	/* IO polling enabled if set */
577 #define QUEUE_FLAG_WC		17	/* Write back caching */
578 #define QUEUE_FLAG_FUA		18	/* device supports FUA writes */
579 #define QUEUE_FLAG_DAX		19	/* device supports DAX */
580 #define QUEUE_FLAG_STATS	20	/* track IO start and completion times */
581 #define QUEUE_FLAG_POLL_STATS	21	/* collecting stats for hybrid polling */
582 #define QUEUE_FLAG_REGISTERED	22	/* queue has been registered to a disk */
583 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23	/* queue supports SCSI commands */
584 #define QUEUE_FLAG_QUIESCED	24	/* queue has been quiesced */
585 #define QUEUE_FLAG_PCI_P2PDMA	25	/* device supports PCI p2p requests */
586 #define QUEUE_FLAG_ZONE_RESETALL 26	/* supports Zone Reset All */
587 #define QUEUE_FLAG_RQ_ALLOC_TIME 27	/* record rq->alloc_time_ns */
588 #define QUEUE_FLAG_HCTX_ACTIVE	28	/* at least one blk-mq hctx is active */
589 #define QUEUE_FLAG_NOWAIT       29	/* device supports NOWAIT */
590 
591 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
592 				 (1 << QUEUE_FLAG_SAME_COMP) |		\
593 				 (1 << QUEUE_FLAG_NOWAIT))
594 
595 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
596 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
597 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
598 
599 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
600 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
601 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
602 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
603 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
604 #define blk_queue_noxmerges(q)	\
605 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
606 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
607 #define blk_queue_stable_writes(q) \
608 	test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
609 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
610 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
611 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
612 #define blk_queue_zone_resetall(q)	\
613 	test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
614 #define blk_queue_secure_erase(q) \
615 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
616 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
617 #define blk_queue_scsi_passthrough(q)	\
618 	test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
619 #define blk_queue_pci_p2pdma(q)	\
620 	test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
621 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
622 #define blk_queue_rq_alloc_time(q)	\
623 	test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
624 #else
625 #define blk_queue_rq_alloc_time(q)	false
626 #endif
627 
628 #define blk_noretry_request(rq) \
629 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
630 			     REQ_FAILFAST_DRIVER))
631 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
632 #define blk_queue_pm_only(q)	atomic_read(&(q)->pm_only)
633 #define blk_queue_fua(q)	test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
634 #define blk_queue_registered(q)	test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
635 #define blk_queue_nowait(q)	test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
636 
637 extern void blk_set_pm_only(struct request_queue *q);
638 extern void blk_clear_pm_only(struct request_queue *q);
639 
640 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
641 
642 #define rq_data_dir(rq)		(op_is_write(req_op(rq)) ? WRITE : READ)
643 
644 #define rq_dma_dir(rq) \
645 	(op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
646 
647 #define dma_map_bvec(dev, bv, dir, attrs) \
648 	dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
649 	(dir), (attrs))
650 
651 #define queue_to_disk(q)	(dev_to_disk(kobj_to_dev((q)->kobj.parent)))
652 
653 static inline bool queue_is_mq(struct request_queue *q)
654 {
655 	return q->mq_ops;
656 }
657 
658 #ifdef CONFIG_PM
659 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
660 {
661 	return q->rpm_status;
662 }
663 #else
664 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
665 {
666 	return RPM_ACTIVE;
667 }
668 #endif
669 
670 static inline enum blk_zoned_model
671 blk_queue_zoned_model(struct request_queue *q)
672 {
673 	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
674 		return q->limits.zoned;
675 	return BLK_ZONED_NONE;
676 }
677 
678 static inline bool blk_queue_is_zoned(struct request_queue *q)
679 {
680 	switch (blk_queue_zoned_model(q)) {
681 	case BLK_ZONED_HA:
682 	case BLK_ZONED_HM:
683 		return true;
684 	default:
685 		return false;
686 	}
687 }
688 
689 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
690 {
691 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
692 }
693 
694 #ifdef CONFIG_BLK_DEV_ZONED
695 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
696 {
697 	return blk_queue_is_zoned(q) ? q->nr_zones : 0;
698 }
699 
700 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
701 					     sector_t sector)
702 {
703 	if (!blk_queue_is_zoned(q))
704 		return 0;
705 	return sector >> ilog2(q->limits.chunk_sectors);
706 }
707 
708 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
709 					 sector_t sector)
710 {
711 	if (!blk_queue_is_zoned(q))
712 		return false;
713 	if (!q->conv_zones_bitmap)
714 		return true;
715 	return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
716 }
717 
718 static inline void blk_queue_max_open_zones(struct request_queue *q,
719 		unsigned int max_open_zones)
720 {
721 	q->max_open_zones = max_open_zones;
722 }
723 
724 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
725 {
726 	return q->max_open_zones;
727 }
728 
729 static inline void blk_queue_max_active_zones(struct request_queue *q,
730 		unsigned int max_active_zones)
731 {
732 	q->max_active_zones = max_active_zones;
733 }
734 
735 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
736 {
737 	return q->max_active_zones;
738 }
739 #else /* CONFIG_BLK_DEV_ZONED */
740 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
741 {
742 	return 0;
743 }
744 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
745 					 sector_t sector)
746 {
747 	return false;
748 }
749 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
750 					     sector_t sector)
751 {
752 	return 0;
753 }
754 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
755 {
756 	return 0;
757 }
758 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
759 {
760 	return 0;
761 }
762 #endif /* CONFIG_BLK_DEV_ZONED */
763 
764 static inline bool rq_is_sync(struct request *rq)
765 {
766 	return op_is_sync(rq->cmd_flags);
767 }
768 
769 static inline bool rq_mergeable(struct request *rq)
770 {
771 	if (blk_rq_is_passthrough(rq))
772 		return false;
773 
774 	if (req_op(rq) == REQ_OP_FLUSH)
775 		return false;
776 
777 	if (req_op(rq) == REQ_OP_WRITE_ZEROES)
778 		return false;
779 
780 	if (req_op(rq) == REQ_OP_ZONE_APPEND)
781 		return false;
782 
783 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
784 		return false;
785 	if (rq->rq_flags & RQF_NOMERGE_FLAGS)
786 		return false;
787 
788 	return true;
789 }
790 
791 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
792 {
793 	if (bio_page(a) == bio_page(b) &&
794 	    bio_offset(a) == bio_offset(b))
795 		return true;
796 
797 	return false;
798 }
799 
800 static inline unsigned int blk_queue_depth(struct request_queue *q)
801 {
802 	if (q->queue_depth)
803 		return q->queue_depth;
804 
805 	return q->nr_requests;
806 }
807 
808 /*
809  * default timeout for SG_IO if none specified
810  */
811 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
812 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
813 
814 struct rq_map_data {
815 	struct page **pages;
816 	int page_order;
817 	int nr_entries;
818 	unsigned long offset;
819 	int null_mapped;
820 	int from_user;
821 };
822 
823 struct req_iterator {
824 	struct bvec_iter iter;
825 	struct bio *bio;
826 };
827 
828 /* This should not be used directly - use rq_for_each_segment */
829 #define for_each_bio(_bio)		\
830 	for (; _bio; _bio = _bio->bi_next)
831 #define __rq_for_each_bio(_bio, rq)	\
832 	if ((rq->bio))			\
833 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
834 
835 #define rq_for_each_segment(bvl, _rq, _iter)			\
836 	__rq_for_each_bio(_iter.bio, _rq)			\
837 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
838 
839 #define rq_for_each_bvec(bvl, _rq, _iter)			\
840 	__rq_for_each_bio(_iter.bio, _rq)			\
841 		bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
842 
843 #define rq_iter_last(bvec, _iter)				\
844 		(_iter.bio->bi_next == NULL &&			\
845 		 bio_iter_last(bvec, _iter.iter))
846 
847 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
848 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
849 #endif
850 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
851 extern void rq_flush_dcache_pages(struct request *rq);
852 #else
853 static inline void rq_flush_dcache_pages(struct request *rq)
854 {
855 }
856 #endif
857 
858 extern int blk_register_queue(struct gendisk *disk);
859 extern void blk_unregister_queue(struct gendisk *disk);
860 blk_qc_t submit_bio_noacct(struct bio *bio);
861 extern void blk_rq_init(struct request_queue *q, struct request *rq);
862 extern void blk_put_request(struct request *);
863 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
864 				       blk_mq_req_flags_t flags);
865 extern int blk_lld_busy(struct request_queue *q);
866 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
867 			     struct bio_set *bs, gfp_t gfp_mask,
868 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
869 			     void *data);
870 extern void blk_rq_unprep_clone(struct request *rq);
871 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
872 				     struct request *rq);
873 int blk_rq_append_bio(struct request *rq, struct bio *bio);
874 extern void blk_queue_split(struct bio **);
875 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
876 extern void blk_queue_exit(struct request_queue *q);
877 extern void blk_sync_queue(struct request_queue *q);
878 extern int blk_rq_map_user(struct request_queue *, struct request *,
879 			   struct rq_map_data *, void __user *, unsigned long,
880 			   gfp_t);
881 extern int blk_rq_unmap_user(struct bio *);
882 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
883 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
884 			       struct rq_map_data *, const struct iov_iter *,
885 			       gfp_t);
886 extern void blk_execute_rq_nowait(struct gendisk *,
887 				  struct request *, int, rq_end_io_fn *);
888 
889 blk_status_t blk_execute_rq(struct gendisk *bd_disk, struct request *rq,
890 			    int at_head);
891 
892 /* Helper to convert REQ_OP_XXX to its string format XXX */
893 extern const char *blk_op_str(unsigned int op);
894 
895 int blk_status_to_errno(blk_status_t status);
896 blk_status_t errno_to_blk_status(int errno);
897 
898 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
899 
900 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
901 {
902 	return bdev->bd_disk->queue;	/* this is never NULL */
903 }
904 
905 /*
906  * The basic unit of block I/O is a sector. It is used in a number of contexts
907  * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
908  * bytes. Variables of type sector_t represent an offset or size that is a
909  * multiple of 512 bytes. Hence these two constants.
910  */
911 #ifndef SECTOR_SHIFT
912 #define SECTOR_SHIFT 9
913 #endif
914 #ifndef SECTOR_SIZE
915 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
916 #endif
917 
918 /*
919  * blk_rq_pos()			: the current sector
920  * blk_rq_bytes()		: bytes left in the entire request
921  * blk_rq_cur_bytes()		: bytes left in the current segment
922  * blk_rq_err_bytes()		: bytes left till the next error boundary
923  * blk_rq_sectors()		: sectors left in the entire request
924  * blk_rq_cur_sectors()		: sectors left in the current segment
925  * blk_rq_stats_sectors()	: sectors of the entire request used for stats
926  */
927 static inline sector_t blk_rq_pos(const struct request *rq)
928 {
929 	return rq->__sector;
930 }
931 
932 static inline unsigned int blk_rq_bytes(const struct request *rq)
933 {
934 	return rq->__data_len;
935 }
936 
937 static inline int blk_rq_cur_bytes(const struct request *rq)
938 {
939 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
940 }
941 
942 extern unsigned int blk_rq_err_bytes(const struct request *rq);
943 
944 static inline unsigned int blk_rq_sectors(const struct request *rq)
945 {
946 	return blk_rq_bytes(rq) >> SECTOR_SHIFT;
947 }
948 
949 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
950 {
951 	return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
952 }
953 
954 static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
955 {
956 	return rq->stats_sectors;
957 }
958 
959 #ifdef CONFIG_BLK_DEV_ZONED
960 
961 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
962 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
963 
964 static inline unsigned int bio_zone_no(struct bio *bio)
965 {
966 	return blk_queue_zone_no(bdev_get_queue(bio->bi_bdev),
967 				 bio->bi_iter.bi_sector);
968 }
969 
970 static inline unsigned int bio_zone_is_seq(struct bio *bio)
971 {
972 	return blk_queue_zone_is_seq(bdev_get_queue(bio->bi_bdev),
973 				     bio->bi_iter.bi_sector);
974 }
975 
976 static inline unsigned int blk_rq_zone_no(struct request *rq)
977 {
978 	return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
979 }
980 
981 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
982 {
983 	return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
984 }
985 #endif /* CONFIG_BLK_DEV_ZONED */
986 
987 /*
988  * Some commands like WRITE SAME have a payload or data transfer size which
989  * is different from the size of the request.  Any driver that supports such
990  * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
991  * calculate the data transfer size.
992  */
993 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
994 {
995 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
996 		return rq->special_vec.bv_len;
997 	return blk_rq_bytes(rq);
998 }
999 
1000 /*
1001  * Return the first full biovec in the request.  The caller needs to check that
1002  * there are any bvecs before calling this helper.
1003  */
1004 static inline struct bio_vec req_bvec(struct request *rq)
1005 {
1006 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1007 		return rq->special_vec;
1008 	return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
1009 }
1010 
1011 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1012 						     int op)
1013 {
1014 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1015 		return min(q->limits.max_discard_sectors,
1016 			   UINT_MAX >> SECTOR_SHIFT);
1017 
1018 	if (unlikely(op == REQ_OP_WRITE_SAME))
1019 		return q->limits.max_write_same_sectors;
1020 
1021 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
1022 		return q->limits.max_write_zeroes_sectors;
1023 
1024 	return q->limits.max_sectors;
1025 }
1026 
1027 /*
1028  * Return maximum size of a request at given offset. Only valid for
1029  * file system requests.
1030  */
1031 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1032 					       sector_t offset,
1033 					       unsigned int chunk_sectors)
1034 {
1035 	if (!chunk_sectors) {
1036 		if (q->limits.chunk_sectors)
1037 			chunk_sectors = q->limits.chunk_sectors;
1038 		else
1039 			return q->limits.max_sectors;
1040 	}
1041 
1042 	if (likely(is_power_of_2(chunk_sectors)))
1043 		chunk_sectors -= offset & (chunk_sectors - 1);
1044 	else
1045 		chunk_sectors -= sector_div(offset, chunk_sectors);
1046 
1047 	return min(q->limits.max_sectors, chunk_sectors);
1048 }
1049 
1050 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1051 						  sector_t offset)
1052 {
1053 	struct request_queue *q = rq->q;
1054 
1055 	if (blk_rq_is_passthrough(rq))
1056 		return q->limits.max_hw_sectors;
1057 
1058 	if (!q->limits.chunk_sectors ||
1059 	    req_op(rq) == REQ_OP_DISCARD ||
1060 	    req_op(rq) == REQ_OP_SECURE_ERASE)
1061 		return blk_queue_get_max_sectors(q, req_op(rq));
1062 
1063 	return min(blk_max_size_offset(q, offset, 0),
1064 			blk_queue_get_max_sectors(q, req_op(rq)));
1065 }
1066 
1067 static inline unsigned int blk_rq_count_bios(struct request *rq)
1068 {
1069 	unsigned int nr_bios = 0;
1070 	struct bio *bio;
1071 
1072 	__rq_for_each_bio(bio, rq)
1073 		nr_bios++;
1074 
1075 	return nr_bios;
1076 }
1077 
1078 void blk_steal_bios(struct bio_list *list, struct request *rq);
1079 
1080 /*
1081  * Request completion related functions.
1082  *
1083  * blk_update_request() completes given number of bytes and updates
1084  * the request without completing it.
1085  */
1086 extern bool blk_update_request(struct request *rq, blk_status_t error,
1087 			       unsigned int nr_bytes);
1088 
1089 extern void blk_abort_request(struct request *);
1090 
1091 /*
1092  * Access functions for manipulating queue properties
1093  */
1094 extern void blk_cleanup_queue(struct request_queue *);
1095 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
1096 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1097 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1098 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1099 extern void blk_queue_max_discard_segments(struct request_queue *,
1100 		unsigned short);
1101 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1102 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1103 		unsigned int max_discard_sectors);
1104 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1105 		unsigned int max_write_same_sectors);
1106 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1107 		unsigned int max_write_same_sectors);
1108 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1109 extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
1110 		unsigned int max_zone_append_sectors);
1111 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1112 void blk_queue_zone_write_granularity(struct request_queue *q,
1113 				      unsigned int size);
1114 extern void blk_queue_alignment_offset(struct request_queue *q,
1115 				       unsigned int alignment);
1116 void blk_queue_update_readahead(struct request_queue *q);
1117 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1118 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1119 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1120 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1121 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1122 extern void blk_set_default_limits(struct queue_limits *lim);
1123 extern void blk_set_stacking_limits(struct queue_limits *lim);
1124 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1125 			    sector_t offset);
1126 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1127 			      sector_t offset);
1128 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1129 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1130 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1131 extern void blk_queue_dma_alignment(struct request_queue *, int);
1132 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1133 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1134 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1135 extern void blk_queue_required_elevator_features(struct request_queue *q,
1136 						 unsigned int features);
1137 extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
1138 					      struct device *dev);
1139 
1140 /*
1141  * Number of physical segments as sent to the device.
1142  *
1143  * Normally this is the number of discontiguous data segments sent by the
1144  * submitter.  But for data-less command like discard we might have no
1145  * actual data segments submitted, but the driver might have to add it's
1146  * own special payload.  In that case we still return 1 here so that this
1147  * special payload will be mapped.
1148  */
1149 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1150 {
1151 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1152 		return 1;
1153 	return rq->nr_phys_segments;
1154 }
1155 
1156 /*
1157  * Number of discard segments (or ranges) the driver needs to fill in.
1158  * Each discard bio merged into a request is counted as one segment.
1159  */
1160 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1161 {
1162 	return max_t(unsigned short, rq->nr_phys_segments, 1);
1163 }
1164 
1165 int __blk_rq_map_sg(struct request_queue *q, struct request *rq,
1166 		struct scatterlist *sglist, struct scatterlist **last_sg);
1167 static inline int blk_rq_map_sg(struct request_queue *q, struct request *rq,
1168 		struct scatterlist *sglist)
1169 {
1170 	struct scatterlist *last_sg = NULL;
1171 
1172 	return __blk_rq_map_sg(q, rq, sglist, &last_sg);
1173 }
1174 extern void blk_dump_rq_flags(struct request *, char *);
1175 
1176 bool __must_check blk_get_queue(struct request_queue *);
1177 extern void blk_put_queue(struct request_queue *);
1178 extern void blk_set_queue_dying(struct request_queue *);
1179 
1180 #ifdef CONFIG_BLOCK
1181 /*
1182  * blk_plug permits building a queue of related requests by holding the I/O
1183  * fragments for a short period. This allows merging of sequential requests
1184  * into single larger request. As the requests are moved from a per-task list to
1185  * the device's request_queue in a batch, this results in improved scalability
1186  * as the lock contention for request_queue lock is reduced.
1187  *
1188  * It is ok not to disable preemption when adding the request to the plug list
1189  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1190  * the plug list when the task sleeps by itself. For details, please see
1191  * schedule() where blk_schedule_flush_plug() is called.
1192  */
1193 struct blk_plug {
1194 	struct list_head mq_list; /* blk-mq requests */
1195 	struct list_head cb_list; /* md requires an unplug callback */
1196 	unsigned short rq_count;
1197 	bool multiple_queues;
1198 	bool nowait;
1199 };
1200 #define BLK_MAX_REQUEST_COUNT 16
1201 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1202 
1203 struct blk_plug_cb;
1204 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1205 struct blk_plug_cb {
1206 	struct list_head list;
1207 	blk_plug_cb_fn callback;
1208 	void *data;
1209 };
1210 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1211 					     void *data, int size);
1212 extern void blk_start_plug(struct blk_plug *);
1213 extern void blk_finish_plug(struct blk_plug *);
1214 extern void blk_flush_plug_list(struct blk_plug *, bool);
1215 
1216 static inline void blk_flush_plug(struct task_struct *tsk)
1217 {
1218 	struct blk_plug *plug = tsk->plug;
1219 
1220 	if (plug)
1221 		blk_flush_plug_list(plug, false);
1222 }
1223 
1224 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1225 {
1226 	struct blk_plug *plug = tsk->plug;
1227 
1228 	if (plug)
1229 		blk_flush_plug_list(plug, true);
1230 }
1231 
1232 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1233 {
1234 	struct blk_plug *plug = tsk->plug;
1235 
1236 	return plug &&
1237 		 (!list_empty(&plug->mq_list) ||
1238 		 !list_empty(&plug->cb_list));
1239 }
1240 
1241 int blkdev_issue_flush(struct block_device *bdev);
1242 long nr_blockdev_pages(void);
1243 #else /* CONFIG_BLOCK */
1244 struct blk_plug {
1245 };
1246 
1247 static inline void blk_start_plug(struct blk_plug *plug)
1248 {
1249 }
1250 
1251 static inline void blk_finish_plug(struct blk_plug *plug)
1252 {
1253 }
1254 
1255 static inline void blk_flush_plug(struct task_struct *task)
1256 {
1257 }
1258 
1259 static inline void blk_schedule_flush_plug(struct task_struct *task)
1260 {
1261 }
1262 
1263 
1264 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1265 {
1266 	return false;
1267 }
1268 
1269 static inline int blkdev_issue_flush(struct block_device *bdev)
1270 {
1271 	return 0;
1272 }
1273 
1274 static inline long nr_blockdev_pages(void)
1275 {
1276 	return 0;
1277 }
1278 #endif /* CONFIG_BLOCK */
1279 
1280 extern void blk_io_schedule(void);
1281 
1282 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1283 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1284 
1285 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
1286 
1287 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1288 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1289 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1290 		sector_t nr_sects, gfp_t gfp_mask, int flags,
1291 		struct bio **biop);
1292 
1293 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1294 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1295 
1296 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1297 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1298 		unsigned flags);
1299 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1300 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1301 
1302 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1303 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1304 {
1305 	return blkdev_issue_discard(sb->s_bdev,
1306 				    block << (sb->s_blocksize_bits -
1307 					      SECTOR_SHIFT),
1308 				    nr_blocks << (sb->s_blocksize_bits -
1309 						  SECTOR_SHIFT),
1310 				    gfp_mask, flags);
1311 }
1312 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1313 		sector_t nr_blocks, gfp_t gfp_mask)
1314 {
1315 	return blkdev_issue_zeroout(sb->s_bdev,
1316 				    block << (sb->s_blocksize_bits -
1317 					      SECTOR_SHIFT),
1318 				    nr_blocks << (sb->s_blocksize_bits -
1319 						  SECTOR_SHIFT),
1320 				    gfp_mask, 0);
1321 }
1322 
1323 static inline bool bdev_is_partition(struct block_device *bdev)
1324 {
1325 	return bdev->bd_partno;
1326 }
1327 
1328 enum blk_default_limits {
1329 	BLK_MAX_SEGMENTS	= 128,
1330 	BLK_SAFE_MAX_SECTORS	= 255,
1331 	BLK_DEF_MAX_SECTORS	= 2560,
1332 	BLK_MAX_SEGMENT_SIZE	= 65536,
1333 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1334 };
1335 
1336 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1337 {
1338 	return q->limits.seg_boundary_mask;
1339 }
1340 
1341 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1342 {
1343 	return q->limits.virt_boundary_mask;
1344 }
1345 
1346 static inline unsigned int queue_max_sectors(const struct request_queue *q)
1347 {
1348 	return q->limits.max_sectors;
1349 }
1350 
1351 static inline unsigned int queue_max_bytes(struct request_queue *q)
1352 {
1353 	return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9;
1354 }
1355 
1356 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1357 {
1358 	return q->limits.max_hw_sectors;
1359 }
1360 
1361 static inline unsigned short queue_max_segments(const struct request_queue *q)
1362 {
1363 	return q->limits.max_segments;
1364 }
1365 
1366 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1367 {
1368 	return q->limits.max_discard_segments;
1369 }
1370 
1371 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1372 {
1373 	return q->limits.max_segment_size;
1374 }
1375 
1376 static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
1377 {
1378 
1379 	const struct queue_limits *l = &q->limits;
1380 
1381 	return min(l->max_zone_append_sectors, l->max_sectors);
1382 }
1383 
1384 static inline unsigned queue_logical_block_size(const struct request_queue *q)
1385 {
1386 	int retval = 512;
1387 
1388 	if (q && q->limits.logical_block_size)
1389 		retval = q->limits.logical_block_size;
1390 
1391 	return retval;
1392 }
1393 
1394 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1395 {
1396 	return queue_logical_block_size(bdev_get_queue(bdev));
1397 }
1398 
1399 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1400 {
1401 	return q->limits.physical_block_size;
1402 }
1403 
1404 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1405 {
1406 	return queue_physical_block_size(bdev_get_queue(bdev));
1407 }
1408 
1409 static inline unsigned int queue_io_min(const struct request_queue *q)
1410 {
1411 	return q->limits.io_min;
1412 }
1413 
1414 static inline int bdev_io_min(struct block_device *bdev)
1415 {
1416 	return queue_io_min(bdev_get_queue(bdev));
1417 }
1418 
1419 static inline unsigned int queue_io_opt(const struct request_queue *q)
1420 {
1421 	return q->limits.io_opt;
1422 }
1423 
1424 static inline int bdev_io_opt(struct block_device *bdev)
1425 {
1426 	return queue_io_opt(bdev_get_queue(bdev));
1427 }
1428 
1429 static inline unsigned int
1430 queue_zone_write_granularity(const struct request_queue *q)
1431 {
1432 	return q->limits.zone_write_granularity;
1433 }
1434 
1435 static inline unsigned int
1436 bdev_zone_write_granularity(struct block_device *bdev)
1437 {
1438 	return queue_zone_write_granularity(bdev_get_queue(bdev));
1439 }
1440 
1441 static inline int queue_alignment_offset(const struct request_queue *q)
1442 {
1443 	if (q->limits.misaligned)
1444 		return -1;
1445 
1446 	return q->limits.alignment_offset;
1447 }
1448 
1449 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1450 {
1451 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1452 	unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1453 		<< SECTOR_SHIFT;
1454 
1455 	return (granularity + lim->alignment_offset - alignment) % granularity;
1456 }
1457 
1458 static inline int bdev_alignment_offset(struct block_device *bdev)
1459 {
1460 	struct request_queue *q = bdev_get_queue(bdev);
1461 
1462 	if (q->limits.misaligned)
1463 		return -1;
1464 	if (bdev_is_partition(bdev))
1465 		return queue_limit_alignment_offset(&q->limits,
1466 				bdev->bd_start_sect);
1467 	return q->limits.alignment_offset;
1468 }
1469 
1470 static inline int queue_discard_alignment(const struct request_queue *q)
1471 {
1472 	if (q->limits.discard_misaligned)
1473 		return -1;
1474 
1475 	return q->limits.discard_alignment;
1476 }
1477 
1478 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1479 {
1480 	unsigned int alignment, granularity, offset;
1481 
1482 	if (!lim->max_discard_sectors)
1483 		return 0;
1484 
1485 	/* Why are these in bytes, not sectors? */
1486 	alignment = lim->discard_alignment >> SECTOR_SHIFT;
1487 	granularity = lim->discard_granularity >> SECTOR_SHIFT;
1488 	if (!granularity)
1489 		return 0;
1490 
1491 	/* Offset of the partition start in 'granularity' sectors */
1492 	offset = sector_div(sector, granularity);
1493 
1494 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1495 	offset = (granularity + alignment - offset) % granularity;
1496 
1497 	/* Turn it back into bytes, gaah */
1498 	return offset << SECTOR_SHIFT;
1499 }
1500 
1501 static inline int bdev_discard_alignment(struct block_device *bdev)
1502 {
1503 	struct request_queue *q = bdev_get_queue(bdev);
1504 
1505 	if (bdev_is_partition(bdev))
1506 		return queue_limit_discard_alignment(&q->limits,
1507 				bdev->bd_start_sect);
1508 	return q->limits.discard_alignment;
1509 }
1510 
1511 static inline unsigned int bdev_write_same(struct block_device *bdev)
1512 {
1513 	struct request_queue *q = bdev_get_queue(bdev);
1514 
1515 	if (q)
1516 		return q->limits.max_write_same_sectors;
1517 
1518 	return 0;
1519 }
1520 
1521 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1522 {
1523 	struct request_queue *q = bdev_get_queue(bdev);
1524 
1525 	if (q)
1526 		return q->limits.max_write_zeroes_sectors;
1527 
1528 	return 0;
1529 }
1530 
1531 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1532 {
1533 	struct request_queue *q = bdev_get_queue(bdev);
1534 
1535 	if (q)
1536 		return blk_queue_zoned_model(q);
1537 
1538 	return BLK_ZONED_NONE;
1539 }
1540 
1541 static inline bool bdev_is_zoned(struct block_device *bdev)
1542 {
1543 	struct request_queue *q = bdev_get_queue(bdev);
1544 
1545 	if (q)
1546 		return blk_queue_is_zoned(q);
1547 
1548 	return false;
1549 }
1550 
1551 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1552 {
1553 	struct request_queue *q = bdev_get_queue(bdev);
1554 
1555 	if (q)
1556 		return blk_queue_zone_sectors(q);
1557 	return 0;
1558 }
1559 
1560 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1561 {
1562 	struct request_queue *q = bdev_get_queue(bdev);
1563 
1564 	if (q)
1565 		return queue_max_open_zones(q);
1566 	return 0;
1567 }
1568 
1569 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1570 {
1571 	struct request_queue *q = bdev_get_queue(bdev);
1572 
1573 	if (q)
1574 		return queue_max_active_zones(q);
1575 	return 0;
1576 }
1577 
1578 static inline int queue_dma_alignment(const struct request_queue *q)
1579 {
1580 	return q ? q->dma_alignment : 511;
1581 }
1582 
1583 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1584 				 unsigned int len)
1585 {
1586 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1587 	return !(addr & alignment) && !(len & alignment);
1588 }
1589 
1590 /* assumes size > 256 */
1591 static inline unsigned int blksize_bits(unsigned int size)
1592 {
1593 	unsigned int bits = 8;
1594 	do {
1595 		bits++;
1596 		size >>= 1;
1597 	} while (size > 256);
1598 	return bits;
1599 }
1600 
1601 static inline unsigned int block_size(struct block_device *bdev)
1602 {
1603 	return 1 << bdev->bd_inode->i_blkbits;
1604 }
1605 
1606 int kblockd_schedule_work(struct work_struct *work);
1607 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1608 
1609 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1610 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1611 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1612 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1613 
1614 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1615 
1616 enum blk_integrity_flags {
1617 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1618 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1619 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1620 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1621 };
1622 
1623 struct blk_integrity_iter {
1624 	void			*prot_buf;
1625 	void			*data_buf;
1626 	sector_t		seed;
1627 	unsigned int		data_size;
1628 	unsigned short		interval;
1629 	const char		*disk_name;
1630 };
1631 
1632 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1633 typedef void (integrity_prepare_fn) (struct request *);
1634 typedef void (integrity_complete_fn) (struct request *, unsigned int);
1635 
1636 struct blk_integrity_profile {
1637 	integrity_processing_fn		*generate_fn;
1638 	integrity_processing_fn		*verify_fn;
1639 	integrity_prepare_fn		*prepare_fn;
1640 	integrity_complete_fn		*complete_fn;
1641 	const char			*name;
1642 };
1643 
1644 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1645 extern void blk_integrity_unregister(struct gendisk *);
1646 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1647 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1648 				   struct scatterlist *);
1649 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1650 
1651 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1652 {
1653 	struct blk_integrity *bi = &disk->queue->integrity;
1654 
1655 	if (!bi->profile)
1656 		return NULL;
1657 
1658 	return bi;
1659 }
1660 
1661 static inline
1662 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1663 {
1664 	return blk_get_integrity(bdev->bd_disk);
1665 }
1666 
1667 static inline bool
1668 blk_integrity_queue_supports_integrity(struct request_queue *q)
1669 {
1670 	return q->integrity.profile;
1671 }
1672 
1673 static inline bool blk_integrity_rq(struct request *rq)
1674 {
1675 	return rq->cmd_flags & REQ_INTEGRITY;
1676 }
1677 
1678 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1679 						    unsigned int segs)
1680 {
1681 	q->limits.max_integrity_segments = segs;
1682 }
1683 
1684 static inline unsigned short
1685 queue_max_integrity_segments(const struct request_queue *q)
1686 {
1687 	return q->limits.max_integrity_segments;
1688 }
1689 
1690 /**
1691  * bio_integrity_intervals - Return number of integrity intervals for a bio
1692  * @bi:		blk_integrity profile for device
1693  * @sectors:	Size of the bio in 512-byte sectors
1694  *
1695  * Description: The block layer calculates everything in 512 byte
1696  * sectors but integrity metadata is done in terms of the data integrity
1697  * interval size of the storage device.  Convert the block layer sectors
1698  * to the appropriate number of integrity intervals.
1699  */
1700 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1701 						   unsigned int sectors)
1702 {
1703 	return sectors >> (bi->interval_exp - 9);
1704 }
1705 
1706 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1707 					       unsigned int sectors)
1708 {
1709 	return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1710 }
1711 
1712 /*
1713  * Return the first bvec that contains integrity data.  Only drivers that are
1714  * limited to a single integrity segment should use this helper.
1715  */
1716 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1717 {
1718 	if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1719 		return NULL;
1720 	return rq->bio->bi_integrity->bip_vec;
1721 }
1722 
1723 #else /* CONFIG_BLK_DEV_INTEGRITY */
1724 
1725 struct bio;
1726 struct block_device;
1727 struct gendisk;
1728 struct blk_integrity;
1729 
1730 static inline int blk_integrity_rq(struct request *rq)
1731 {
1732 	return 0;
1733 }
1734 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1735 					    struct bio *b)
1736 {
1737 	return 0;
1738 }
1739 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1740 					  struct bio *b,
1741 					  struct scatterlist *s)
1742 {
1743 	return 0;
1744 }
1745 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1746 {
1747 	return NULL;
1748 }
1749 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1750 {
1751 	return NULL;
1752 }
1753 static inline bool
1754 blk_integrity_queue_supports_integrity(struct request_queue *q)
1755 {
1756 	return false;
1757 }
1758 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1759 {
1760 	return 0;
1761 }
1762 static inline void blk_integrity_register(struct gendisk *d,
1763 					 struct blk_integrity *b)
1764 {
1765 }
1766 static inline void blk_integrity_unregister(struct gendisk *d)
1767 {
1768 }
1769 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1770 						    unsigned int segs)
1771 {
1772 }
1773 static inline unsigned short queue_max_integrity_segments(const struct request_queue *q)
1774 {
1775 	return 0;
1776 }
1777 
1778 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1779 						   unsigned int sectors)
1780 {
1781 	return 0;
1782 }
1783 
1784 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1785 					       unsigned int sectors)
1786 {
1787 	return 0;
1788 }
1789 
1790 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1791 {
1792 	return NULL;
1793 }
1794 
1795 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1796 
1797 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1798 
1799 bool blk_ksm_register(struct blk_keyslot_manager *ksm, struct request_queue *q);
1800 
1801 void blk_ksm_unregister(struct request_queue *q);
1802 
1803 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1804 
1805 static inline bool blk_ksm_register(struct blk_keyslot_manager *ksm,
1806 				    struct request_queue *q)
1807 {
1808 	return true;
1809 }
1810 
1811 static inline void blk_ksm_unregister(struct request_queue *q) { }
1812 
1813 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1814 
1815 
1816 struct block_device_operations {
1817 	blk_qc_t (*submit_bio) (struct bio *bio);
1818 	int (*open) (struct block_device *, fmode_t);
1819 	void (*release) (struct gendisk *, fmode_t);
1820 	int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1821 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1822 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1823 	unsigned int (*check_events) (struct gendisk *disk,
1824 				      unsigned int clearing);
1825 	void (*unlock_native_capacity) (struct gendisk *);
1826 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1827 	int (*set_read_only)(struct block_device *bdev, bool ro);
1828 	/* this callback is with swap_lock and sometimes page table lock held */
1829 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1830 	int (*report_zones)(struct gendisk *, sector_t sector,
1831 			unsigned int nr_zones, report_zones_cb cb, void *data);
1832 	char *(*devnode)(struct gendisk *disk, umode_t *mode);
1833 	struct module *owner;
1834 	const struct pr_ops *pr_ops;
1835 };
1836 
1837 #ifdef CONFIG_COMPAT
1838 extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1839 				      unsigned int, unsigned long);
1840 #else
1841 #define blkdev_compat_ptr_ioctl NULL
1842 #endif
1843 
1844 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1845 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1846 						struct writeback_control *);
1847 
1848 #ifdef CONFIG_BLK_DEV_ZONED
1849 bool blk_req_needs_zone_write_lock(struct request *rq);
1850 bool blk_req_zone_write_trylock(struct request *rq);
1851 void __blk_req_zone_write_lock(struct request *rq);
1852 void __blk_req_zone_write_unlock(struct request *rq);
1853 
1854 static inline void blk_req_zone_write_lock(struct request *rq)
1855 {
1856 	if (blk_req_needs_zone_write_lock(rq))
1857 		__blk_req_zone_write_lock(rq);
1858 }
1859 
1860 static inline void blk_req_zone_write_unlock(struct request *rq)
1861 {
1862 	if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1863 		__blk_req_zone_write_unlock(rq);
1864 }
1865 
1866 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1867 {
1868 	return rq->q->seq_zones_wlock &&
1869 		test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1870 }
1871 
1872 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1873 {
1874 	if (!blk_req_needs_zone_write_lock(rq))
1875 		return true;
1876 	return !blk_req_zone_is_write_locked(rq);
1877 }
1878 #else
1879 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1880 {
1881 	return false;
1882 }
1883 
1884 static inline void blk_req_zone_write_lock(struct request *rq)
1885 {
1886 }
1887 
1888 static inline void blk_req_zone_write_unlock(struct request *rq)
1889 {
1890 }
1891 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1892 {
1893 	return false;
1894 }
1895 
1896 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1897 {
1898 	return true;
1899 }
1900 #endif /* CONFIG_BLK_DEV_ZONED */
1901 
1902 static inline void blk_wake_io_task(struct task_struct *waiter)
1903 {
1904 	/*
1905 	 * If we're polling, the task itself is doing the completions. For
1906 	 * that case, we don't need to signal a wakeup, it's enough to just
1907 	 * mark us as RUNNING.
1908 	 */
1909 	if (waiter == current)
1910 		__set_current_state(TASK_RUNNING);
1911 	else
1912 		wake_up_process(waiter);
1913 }
1914 
1915 unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1916 		unsigned int op);
1917 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1918 		unsigned long start_time);
1919 
1920 unsigned long bio_start_io_acct(struct bio *bio);
1921 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1922 		struct block_device *orig_bdev);
1923 
1924 /**
1925  * bio_end_io_acct - end I/O accounting for bio based drivers
1926  * @bio:	bio to end account for
1927  * @start:	start time returned by bio_start_io_acct()
1928  */
1929 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1930 {
1931 	return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1932 }
1933 
1934 int bdev_read_only(struct block_device *bdev);
1935 int set_blocksize(struct block_device *bdev, int size);
1936 
1937 const char *bdevname(struct block_device *bdev, char *buffer);
1938 int lookup_bdev(const char *pathname, dev_t *dev);
1939 
1940 void blkdev_show(struct seq_file *seqf, off_t offset);
1941 
1942 #define BDEVNAME_SIZE	32	/* Largest string for a blockdev identifier */
1943 #define BDEVT_SIZE	10	/* Largest string for MAJ:MIN for blkdev */
1944 #ifdef CONFIG_BLOCK
1945 #define BLKDEV_MAJOR_MAX	512
1946 #else
1947 #define BLKDEV_MAJOR_MAX	0
1948 #endif
1949 
1950 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1951 		void *holder);
1952 struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
1953 int bd_prepare_to_claim(struct block_device *bdev, void *holder);
1954 void bd_abort_claiming(struct block_device *bdev, void *holder);
1955 void blkdev_put(struct block_device *bdev, fmode_t mode);
1956 
1957 /* just for blk-cgroup, don't use elsewhere */
1958 struct block_device *blkdev_get_no_open(dev_t dev);
1959 void blkdev_put_no_open(struct block_device *bdev);
1960 
1961 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
1962 void bdev_add(struct block_device *bdev, dev_t dev);
1963 struct block_device *I_BDEV(struct inode *inode);
1964 struct block_device *bdgrab(struct block_device *bdev);
1965 void bdput(struct block_device *);
1966 int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
1967 		loff_t lend);
1968 
1969 #ifdef CONFIG_BLOCK
1970 void invalidate_bdev(struct block_device *bdev);
1971 int sync_blockdev(struct block_device *bdev);
1972 #else
1973 static inline void invalidate_bdev(struct block_device *bdev)
1974 {
1975 }
1976 static inline int sync_blockdev(struct block_device *bdev)
1977 {
1978 	return 0;
1979 }
1980 #endif
1981 int fsync_bdev(struct block_device *bdev);
1982 
1983 int freeze_bdev(struct block_device *bdev);
1984 int thaw_bdev(struct block_device *bdev);
1985 
1986 #endif /* _LINUX_BLKDEV_H */
1987