xref: /linux-6.15/include/linux/blkdev.h (revision 0899431f)
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/bsg.h>
22 #include <linux/smp.h>
23 #include <linux/rcupdate.h>
24 #include <linux/percpu-refcount.h>
25 #include <linux/scatterlist.h>
26 #include <linux/blkzoned.h>
27 #include <linux/pm.h>
28 
29 struct module;
30 struct scsi_ioctl_command;
31 
32 struct request_queue;
33 struct elevator_queue;
34 struct blk_trace;
35 struct request;
36 struct sg_io_hdr;
37 struct bsg_job;
38 struct blkcg_gq;
39 struct blk_flush_queue;
40 struct pr_ops;
41 struct rq_qos;
42 struct blk_queue_stats;
43 struct blk_stat_callback;
44 struct blk_keyslot_manager;
45 
46 #define BLKDEV_MIN_RQ	4
47 #define BLKDEV_MAX_RQ	128	/* Default maximum */
48 
49 /* Must be consistent with blk_mq_poll_stats_bkt() */
50 #define BLK_MQ_POLL_STATS_BKTS 16
51 
52 /* Doing classic polling */
53 #define BLK_MQ_POLL_CLASSIC -1
54 
55 /*
56  * Maximum number of blkcg policies allowed to be registered concurrently.
57  * Defined here to simplify include dependency.
58  */
59 #define BLKCG_MAX_POLS		5
60 
61 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
62 
63 /*
64  * request flags */
65 typedef __u32 __bitwise req_flags_t;
66 
67 /* drive already may have started this one */
68 #define RQF_STARTED		((__force req_flags_t)(1 << 1))
69 /* may not be passed by ioscheduler */
70 #define RQF_SOFTBARRIER		((__force req_flags_t)(1 << 3))
71 /* request for flush sequence */
72 #define RQF_FLUSH_SEQ		((__force req_flags_t)(1 << 4))
73 /* merge of different types, fail separately */
74 #define RQF_MIXED_MERGE		((__force req_flags_t)(1 << 5))
75 /* track inflight for MQ */
76 #define RQF_MQ_INFLIGHT		((__force req_flags_t)(1 << 6))
77 /* don't call prep for this one */
78 #define RQF_DONTPREP		((__force req_flags_t)(1 << 7))
79 /* vaguely specified driver internal error.  Ignored by the block layer */
80 #define RQF_FAILED		((__force req_flags_t)(1 << 10))
81 /* don't warn about errors */
82 #define RQF_QUIET		((__force req_flags_t)(1 << 11))
83 /* elevator private data attached */
84 #define RQF_ELVPRIV		((__force req_flags_t)(1 << 12))
85 /* account into disk and partition IO statistics */
86 #define RQF_IO_STAT		((__force req_flags_t)(1 << 13))
87 /* runtime pm request */
88 #define RQF_PM			((__force req_flags_t)(1 << 15))
89 /* on IO scheduler merge hash */
90 #define RQF_HASHED		((__force req_flags_t)(1 << 16))
91 /* track IO completion time */
92 #define RQF_STATS		((__force req_flags_t)(1 << 17))
93 /* Look at ->special_vec for the actual data payload instead of the
94    bio chain. */
95 #define RQF_SPECIAL_PAYLOAD	((__force req_flags_t)(1 << 18))
96 /* The per-zone write lock is held for this request */
97 #define RQF_ZONE_WRITE_LOCKED	((__force req_flags_t)(1 << 19))
98 /* already slept for hybrid poll */
99 #define RQF_MQ_POLL_SLEPT	((__force req_flags_t)(1 << 20))
100 /* ->timeout has been called, don't expire again */
101 #define RQF_TIMED_OUT		((__force req_flags_t)(1 << 21))
102 
103 /* flags that prevent us from merging requests: */
104 #define RQF_NOMERGE_FLAGS \
105 	(RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
106 
107 /*
108  * Request state for blk-mq.
109  */
110 enum mq_rq_state {
111 	MQ_RQ_IDLE		= 0,
112 	MQ_RQ_IN_FLIGHT		= 1,
113 	MQ_RQ_COMPLETE		= 2,
114 };
115 
116 /*
117  * Try to put the fields that are referenced together in the same cacheline.
118  *
119  * If you modify this structure, make sure to update blk_rq_init() and
120  * especially blk_mq_rq_ctx_init() to take care of the added fields.
121  */
122 struct request {
123 	struct request_queue *q;
124 	struct blk_mq_ctx *mq_ctx;
125 	struct blk_mq_hw_ctx *mq_hctx;
126 
127 	unsigned int cmd_flags;		/* op and common flags */
128 	req_flags_t rq_flags;
129 
130 	int tag;
131 	int internal_tag;
132 
133 	/* the following two fields are internal, NEVER access directly */
134 	unsigned int __data_len;	/* total data len */
135 	sector_t __sector;		/* sector cursor */
136 
137 	struct bio *bio;
138 	struct bio *biotail;
139 
140 	struct list_head queuelist;
141 
142 	/*
143 	 * The hash is used inside the scheduler, and killed once the
144 	 * request reaches the dispatch list. The ipi_list is only used
145 	 * to queue the request for softirq completion, which is long
146 	 * after the request has been unhashed (and even removed from
147 	 * the dispatch list).
148 	 */
149 	union {
150 		struct hlist_node hash;	/* merge hash */
151 		struct llist_node ipi_list;
152 	};
153 
154 	/*
155 	 * The rb_node is only used inside the io scheduler, requests
156 	 * are pruned when moved to the dispatch queue. So let the
157 	 * completion_data share space with the rb_node.
158 	 */
159 	union {
160 		struct rb_node rb_node;	/* sort/lookup */
161 		struct bio_vec special_vec;
162 		void *completion_data;
163 		int error_count; /* for legacy drivers, don't use */
164 	};
165 
166 	/*
167 	 * Three pointers are available for the IO schedulers, if they need
168 	 * more they have to dynamically allocate it.  Flush requests are
169 	 * never put on the IO scheduler. So let the flush fields share
170 	 * space with the elevator data.
171 	 */
172 	union {
173 		struct {
174 			struct io_cq		*icq;
175 			void			*priv[2];
176 		} elv;
177 
178 		struct {
179 			unsigned int		seq;
180 			struct list_head	list;
181 			rq_end_io_fn		*saved_end_io;
182 		} flush;
183 	};
184 
185 	struct gendisk *rq_disk;
186 	struct block_device *part;
187 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
188 	/* Time that the first bio started allocating this request. */
189 	u64 alloc_time_ns;
190 #endif
191 	/* Time that this request was allocated for this IO. */
192 	u64 start_time_ns;
193 	/* Time that I/O was submitted to the device. */
194 	u64 io_start_time_ns;
195 
196 #ifdef CONFIG_BLK_WBT
197 	unsigned short wbt_flags;
198 #endif
199 	/*
200 	 * rq sectors used for blk stats. It has the same value
201 	 * with blk_rq_sectors(rq), except that it never be zeroed
202 	 * by completion.
203 	 */
204 	unsigned short stats_sectors;
205 
206 	/*
207 	 * Number of scatter-gather DMA addr+len pairs after
208 	 * physical address coalescing is performed.
209 	 */
210 	unsigned short nr_phys_segments;
211 
212 #if defined(CONFIG_BLK_DEV_INTEGRITY)
213 	unsigned short nr_integrity_segments;
214 #endif
215 
216 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
217 	struct bio_crypt_ctx *crypt_ctx;
218 	struct blk_ksm_keyslot *crypt_keyslot;
219 #endif
220 
221 	unsigned short write_hint;
222 	unsigned short ioprio;
223 
224 	enum mq_rq_state state;
225 	refcount_t ref;
226 
227 	unsigned int timeout;
228 	unsigned long deadline;
229 
230 	union {
231 		struct __call_single_data csd;
232 		u64 fifo_time;
233 	};
234 
235 	/*
236 	 * completion callback.
237 	 */
238 	rq_end_io_fn *end_io;
239 	void *end_io_data;
240 };
241 
242 static inline bool blk_op_is_scsi(unsigned int op)
243 {
244 	return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
245 }
246 
247 static inline bool blk_op_is_private(unsigned int op)
248 {
249 	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
250 }
251 
252 static inline bool blk_rq_is_scsi(struct request *rq)
253 {
254 	return blk_op_is_scsi(req_op(rq));
255 }
256 
257 static inline bool blk_rq_is_private(struct request *rq)
258 {
259 	return blk_op_is_private(req_op(rq));
260 }
261 
262 static inline bool blk_rq_is_passthrough(struct request *rq)
263 {
264 	return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
265 }
266 
267 static inline bool bio_is_passthrough(struct bio *bio)
268 {
269 	unsigned op = bio_op(bio);
270 
271 	return blk_op_is_scsi(op) || blk_op_is_private(op);
272 }
273 
274 static inline bool blk_op_is_passthrough(unsigned int op)
275 {
276 	return (blk_op_is_scsi(op & REQ_OP_MASK) ||
277 			blk_op_is_private(op & REQ_OP_MASK));
278 }
279 
280 static inline unsigned short req_get_ioprio(struct request *req)
281 {
282 	return req->ioprio;
283 }
284 
285 #include <linux/elevator.h>
286 
287 struct blk_queue_ctx;
288 
289 struct bio_vec;
290 
291 enum blk_eh_timer_return {
292 	BLK_EH_DONE,		/* drivers has completed the command */
293 	BLK_EH_RESET_TIMER,	/* reset timer and try again */
294 };
295 
296 enum blk_queue_state {
297 	Queue_down,
298 	Queue_up,
299 };
300 
301 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
302 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
303 
304 #define BLK_SCSI_MAX_CMDS	(256)
305 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
306 
307 /*
308  * Zoned block device models (zoned limit).
309  *
310  * Note: This needs to be ordered from the least to the most severe
311  * restrictions for the inheritance in blk_stack_limits() to work.
312  */
313 enum blk_zoned_model {
314 	BLK_ZONED_NONE = 0,	/* Regular block device */
315 	BLK_ZONED_HA,		/* Host-aware zoned block device */
316 	BLK_ZONED_HM,		/* Host-managed zoned block device */
317 };
318 
319 /*
320  * BLK_BOUNCE_NONE:	never bounce (default)
321  * BLK_BOUNCE_HIGH:	bounce all highmem pages
322  */
323 enum blk_bounce {
324 	BLK_BOUNCE_NONE,
325 	BLK_BOUNCE_HIGH,
326 };
327 
328 struct queue_limits {
329 	enum blk_bounce		bounce;
330 	unsigned long		seg_boundary_mask;
331 	unsigned long		virt_boundary_mask;
332 
333 	unsigned int		max_hw_sectors;
334 	unsigned int		max_dev_sectors;
335 	unsigned int		chunk_sectors;
336 	unsigned int		max_sectors;
337 	unsigned int		max_segment_size;
338 	unsigned int		physical_block_size;
339 	unsigned int		logical_block_size;
340 	unsigned int		alignment_offset;
341 	unsigned int		io_min;
342 	unsigned int		io_opt;
343 	unsigned int		max_discard_sectors;
344 	unsigned int		max_hw_discard_sectors;
345 	unsigned int		max_write_same_sectors;
346 	unsigned int		max_write_zeroes_sectors;
347 	unsigned int		max_zone_append_sectors;
348 	unsigned int		discard_granularity;
349 	unsigned int		discard_alignment;
350 	unsigned int		zone_write_granularity;
351 
352 	unsigned short		max_segments;
353 	unsigned short		max_integrity_segments;
354 	unsigned short		max_discard_segments;
355 
356 	unsigned char		misaligned;
357 	unsigned char		discard_misaligned;
358 	unsigned char		raid_partial_stripes_expensive;
359 	enum blk_zoned_model	zoned;
360 };
361 
362 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
363 			       void *data);
364 
365 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
366 
367 #ifdef CONFIG_BLK_DEV_ZONED
368 
369 #define BLK_ALL_ZONES  ((unsigned int)-1)
370 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
371 			unsigned int nr_zones, report_zones_cb cb, void *data);
372 unsigned int blkdev_nr_zones(struct gendisk *disk);
373 extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
374 			    sector_t sectors, sector_t nr_sectors,
375 			    gfp_t gfp_mask);
376 int blk_revalidate_disk_zones(struct gendisk *disk,
377 			      void (*update_driver_data)(struct gendisk *disk));
378 
379 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
380 				     unsigned int cmd, unsigned long arg);
381 extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
382 				  unsigned int cmd, unsigned long arg);
383 
384 #else /* CONFIG_BLK_DEV_ZONED */
385 
386 static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
387 {
388 	return 0;
389 }
390 
391 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
392 					    fmode_t mode, unsigned int cmd,
393 					    unsigned long arg)
394 {
395 	return -ENOTTY;
396 }
397 
398 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
399 					 fmode_t mode, unsigned int cmd,
400 					 unsigned long arg)
401 {
402 	return -ENOTTY;
403 }
404 
405 #endif /* CONFIG_BLK_DEV_ZONED */
406 
407 struct request_queue {
408 	struct request		*last_merge;
409 	struct elevator_queue	*elevator;
410 
411 	struct percpu_ref	q_usage_counter;
412 
413 	struct blk_queue_stats	*stats;
414 	struct rq_qos		*rq_qos;
415 
416 	const struct blk_mq_ops	*mq_ops;
417 
418 	/* sw queues */
419 	struct blk_mq_ctx __percpu	*queue_ctx;
420 
421 	unsigned int		queue_depth;
422 
423 	/* hw dispatch queues */
424 	struct blk_mq_hw_ctx	**queue_hw_ctx;
425 	unsigned int		nr_hw_queues;
426 
427 	struct backing_dev_info	*backing_dev_info;
428 
429 	/*
430 	 * The queue owner gets to use this for whatever they like.
431 	 * ll_rw_blk doesn't touch it.
432 	 */
433 	void			*queuedata;
434 
435 	/*
436 	 * various queue flags, see QUEUE_* below
437 	 */
438 	unsigned long		queue_flags;
439 	/*
440 	 * Number of contexts that have called blk_set_pm_only(). If this
441 	 * counter is above zero then only RQF_PM requests are processed.
442 	 */
443 	atomic_t		pm_only;
444 
445 	/*
446 	 * ida allocated id for this queue.  Used to index queues from
447 	 * ioctx.
448 	 */
449 	int			id;
450 
451 	spinlock_t		queue_lock;
452 
453 	/*
454 	 * queue kobject
455 	 */
456 	struct kobject kobj;
457 
458 	/*
459 	 * mq queue kobject
460 	 */
461 	struct kobject *mq_kobj;
462 
463 #ifdef  CONFIG_BLK_DEV_INTEGRITY
464 	struct blk_integrity integrity;
465 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
466 
467 #ifdef CONFIG_PM
468 	struct device		*dev;
469 	enum rpm_status		rpm_status;
470 #endif
471 
472 	/*
473 	 * queue settings
474 	 */
475 	unsigned long		nr_requests;	/* Max # of requests */
476 
477 	unsigned int		dma_pad_mask;
478 	unsigned int		dma_alignment;
479 
480 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
481 	/* Inline crypto capabilities */
482 	struct blk_keyslot_manager *ksm;
483 #endif
484 
485 	unsigned int		rq_timeout;
486 	int			poll_nsec;
487 
488 	struct blk_stat_callback	*poll_cb;
489 	struct blk_rq_stat	poll_stat[BLK_MQ_POLL_STATS_BKTS];
490 
491 	struct timer_list	timeout;
492 	struct work_struct	timeout_work;
493 
494 	atomic_t		nr_active_requests_shared_sbitmap;
495 
496 	struct list_head	icq_list;
497 #ifdef CONFIG_BLK_CGROUP
498 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
499 	struct blkcg_gq		*root_blkg;
500 	struct list_head	blkg_list;
501 #endif
502 
503 	struct queue_limits	limits;
504 
505 	unsigned int		required_elevator_features;
506 
507 #ifdef CONFIG_BLK_DEV_ZONED
508 	/*
509 	 * Zoned block device information for request dispatch control.
510 	 * nr_zones is the total number of zones of the device. This is always
511 	 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
512 	 * bits which indicates if a zone is conventional (bit set) or
513 	 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
514 	 * bits which indicates if a zone is write locked, that is, if a write
515 	 * request targeting the zone was dispatched. All three fields are
516 	 * initialized by the low level device driver (e.g. scsi/sd.c).
517 	 * Stacking drivers (device mappers) may or may not initialize
518 	 * these fields.
519 	 *
520 	 * Reads of this information must be protected with blk_queue_enter() /
521 	 * blk_queue_exit(). Modifying this information is only allowed while
522 	 * no requests are being processed. See also blk_mq_freeze_queue() and
523 	 * blk_mq_unfreeze_queue().
524 	 */
525 	unsigned int		nr_zones;
526 	unsigned long		*conv_zones_bitmap;
527 	unsigned long		*seq_zones_wlock;
528 	unsigned int		max_open_zones;
529 	unsigned int		max_active_zones;
530 #endif /* CONFIG_BLK_DEV_ZONED */
531 
532 	/*
533 	 * sg stuff
534 	 */
535 	unsigned int		sg_timeout;
536 	unsigned int		sg_reserved_size;
537 	int			node;
538 	struct mutex		debugfs_mutex;
539 #ifdef CONFIG_BLK_DEV_IO_TRACE
540 	struct blk_trace __rcu	*blk_trace;
541 #endif
542 	/*
543 	 * for flush operations
544 	 */
545 	struct blk_flush_queue	*fq;
546 
547 	struct list_head	requeue_list;
548 	spinlock_t		requeue_lock;
549 	struct delayed_work	requeue_work;
550 
551 	struct mutex		sysfs_lock;
552 	struct mutex		sysfs_dir_lock;
553 
554 	/*
555 	 * for reusing dead hctx instance in case of updating
556 	 * nr_hw_queues
557 	 */
558 	struct list_head	unused_hctx_list;
559 	spinlock_t		unused_hctx_lock;
560 
561 	int			mq_freeze_depth;
562 
563 #if defined(CONFIG_BLK_DEV_BSG)
564 	struct bsg_class_device bsg_dev;
565 #endif
566 
567 #ifdef CONFIG_BLK_DEV_THROTTLING
568 	/* Throttle data */
569 	struct throtl_data *td;
570 #endif
571 	struct rcu_head		rcu_head;
572 	wait_queue_head_t	mq_freeze_wq;
573 	/*
574 	 * Protect concurrent access to q_usage_counter by
575 	 * percpu_ref_kill() and percpu_ref_reinit().
576 	 */
577 	struct mutex		mq_freeze_lock;
578 
579 	struct blk_mq_tag_set	*tag_set;
580 	struct list_head	tag_set_list;
581 	struct bio_set		bio_split;
582 
583 	struct dentry		*debugfs_dir;
584 
585 #ifdef CONFIG_BLK_DEBUG_FS
586 	struct dentry		*sched_debugfs_dir;
587 	struct dentry		*rqos_debugfs_dir;
588 #endif
589 
590 	bool			mq_sysfs_init_done;
591 
592 	size_t			cmd_size;
593 
594 #define BLK_MAX_WRITE_HINTS	5
595 	u64			write_hints[BLK_MAX_WRITE_HINTS];
596 };
597 
598 /* Keep blk_queue_flag_name[] in sync with the definitions below */
599 #define QUEUE_FLAG_STOPPED	0	/* queue is stopped */
600 #define QUEUE_FLAG_DYING	1	/* queue being torn down */
601 #define QUEUE_FLAG_NOMERGES     3	/* disable merge attempts */
602 #define QUEUE_FLAG_SAME_COMP	4	/* complete on same CPU-group */
603 #define QUEUE_FLAG_FAIL_IO	5	/* fake timeout */
604 #define QUEUE_FLAG_NONROT	6	/* non-rotational device (SSD) */
605 #define QUEUE_FLAG_VIRT		QUEUE_FLAG_NONROT /* paravirt device */
606 #define QUEUE_FLAG_IO_STAT	7	/* do disk/partitions IO accounting */
607 #define QUEUE_FLAG_DISCARD	8	/* supports DISCARD */
608 #define QUEUE_FLAG_NOXMERGES	9	/* No extended merges */
609 #define QUEUE_FLAG_ADD_RANDOM	10	/* Contributes to random pool */
610 #define QUEUE_FLAG_SECERASE	11	/* supports secure erase */
611 #define QUEUE_FLAG_SAME_FORCE	12	/* force complete on same CPU */
612 #define QUEUE_FLAG_DEAD		13	/* queue tear-down finished */
613 #define QUEUE_FLAG_INIT_DONE	14	/* queue is initialized */
614 #define QUEUE_FLAG_STABLE_WRITES 15	/* don't modify blks until WB is done */
615 #define QUEUE_FLAG_POLL		16	/* IO polling enabled if set */
616 #define QUEUE_FLAG_WC		17	/* Write back caching */
617 #define QUEUE_FLAG_FUA		18	/* device supports FUA writes */
618 #define QUEUE_FLAG_DAX		19	/* device supports DAX */
619 #define QUEUE_FLAG_STATS	20	/* track IO start and completion times */
620 #define QUEUE_FLAG_POLL_STATS	21	/* collecting stats for hybrid polling */
621 #define QUEUE_FLAG_REGISTERED	22	/* queue has been registered to a disk */
622 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23	/* queue supports SCSI commands */
623 #define QUEUE_FLAG_QUIESCED	24	/* queue has been quiesced */
624 #define QUEUE_FLAG_PCI_P2PDMA	25	/* device supports PCI p2p requests */
625 #define QUEUE_FLAG_ZONE_RESETALL 26	/* supports Zone Reset All */
626 #define QUEUE_FLAG_RQ_ALLOC_TIME 27	/* record rq->alloc_time_ns */
627 #define QUEUE_FLAG_HCTX_ACTIVE	28	/* at least one blk-mq hctx is active */
628 #define QUEUE_FLAG_NOWAIT       29	/* device supports NOWAIT */
629 
630 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
631 				 (1 << QUEUE_FLAG_SAME_COMP) |		\
632 				 (1 << QUEUE_FLAG_NOWAIT))
633 
634 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
635 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
636 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
637 
638 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
639 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
640 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
641 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
642 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
643 #define blk_queue_noxmerges(q)	\
644 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
645 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
646 #define blk_queue_stable_writes(q) \
647 	test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
648 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
649 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
650 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
651 #define blk_queue_zone_resetall(q)	\
652 	test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
653 #define blk_queue_secure_erase(q) \
654 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
655 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
656 #define blk_queue_scsi_passthrough(q)	\
657 	test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
658 #define blk_queue_pci_p2pdma(q)	\
659 	test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
660 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
661 #define blk_queue_rq_alloc_time(q)	\
662 	test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
663 #else
664 #define blk_queue_rq_alloc_time(q)	false
665 #endif
666 
667 #define blk_noretry_request(rq) \
668 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
669 			     REQ_FAILFAST_DRIVER))
670 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
671 #define blk_queue_pm_only(q)	atomic_read(&(q)->pm_only)
672 #define blk_queue_fua(q)	test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
673 #define blk_queue_registered(q)	test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
674 #define blk_queue_nowait(q)	test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
675 
676 extern void blk_set_pm_only(struct request_queue *q);
677 extern void blk_clear_pm_only(struct request_queue *q);
678 
679 static inline bool blk_account_rq(struct request *rq)
680 {
681 	return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
682 }
683 
684 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
685 
686 #define rq_data_dir(rq)		(op_is_write(req_op(rq)) ? WRITE : READ)
687 
688 #define rq_dma_dir(rq) \
689 	(op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
690 
691 #define dma_map_bvec(dev, bv, dir, attrs) \
692 	dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
693 	(dir), (attrs))
694 
695 #define queue_to_disk(q)	(dev_to_disk(kobj_to_dev((q)->kobj.parent)))
696 
697 static inline bool queue_is_mq(struct request_queue *q)
698 {
699 	return q->mq_ops;
700 }
701 
702 #ifdef CONFIG_PM
703 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
704 {
705 	return q->rpm_status;
706 }
707 #else
708 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
709 {
710 	return RPM_ACTIVE;
711 }
712 #endif
713 
714 static inline enum blk_zoned_model
715 blk_queue_zoned_model(struct request_queue *q)
716 {
717 	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
718 		return q->limits.zoned;
719 	return BLK_ZONED_NONE;
720 }
721 
722 static inline bool blk_queue_is_zoned(struct request_queue *q)
723 {
724 	switch (blk_queue_zoned_model(q)) {
725 	case BLK_ZONED_HA:
726 	case BLK_ZONED_HM:
727 		return true;
728 	default:
729 		return false;
730 	}
731 }
732 
733 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
734 {
735 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
736 }
737 
738 #ifdef CONFIG_BLK_DEV_ZONED
739 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
740 {
741 	return blk_queue_is_zoned(q) ? q->nr_zones : 0;
742 }
743 
744 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
745 					     sector_t sector)
746 {
747 	if (!blk_queue_is_zoned(q))
748 		return 0;
749 	return sector >> ilog2(q->limits.chunk_sectors);
750 }
751 
752 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
753 					 sector_t sector)
754 {
755 	if (!blk_queue_is_zoned(q))
756 		return false;
757 	if (!q->conv_zones_bitmap)
758 		return true;
759 	return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
760 }
761 
762 static inline void blk_queue_max_open_zones(struct request_queue *q,
763 		unsigned int max_open_zones)
764 {
765 	q->max_open_zones = max_open_zones;
766 }
767 
768 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
769 {
770 	return q->max_open_zones;
771 }
772 
773 static inline void blk_queue_max_active_zones(struct request_queue *q,
774 		unsigned int max_active_zones)
775 {
776 	q->max_active_zones = max_active_zones;
777 }
778 
779 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
780 {
781 	return q->max_active_zones;
782 }
783 #else /* CONFIG_BLK_DEV_ZONED */
784 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
785 {
786 	return 0;
787 }
788 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
789 					 sector_t sector)
790 {
791 	return false;
792 }
793 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
794 					     sector_t sector)
795 {
796 	return 0;
797 }
798 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
799 {
800 	return 0;
801 }
802 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
803 {
804 	return 0;
805 }
806 #endif /* CONFIG_BLK_DEV_ZONED */
807 
808 static inline bool rq_is_sync(struct request *rq)
809 {
810 	return op_is_sync(rq->cmd_flags);
811 }
812 
813 static inline bool rq_mergeable(struct request *rq)
814 {
815 	if (blk_rq_is_passthrough(rq))
816 		return false;
817 
818 	if (req_op(rq) == REQ_OP_FLUSH)
819 		return false;
820 
821 	if (req_op(rq) == REQ_OP_WRITE_ZEROES)
822 		return false;
823 
824 	if (req_op(rq) == REQ_OP_ZONE_APPEND)
825 		return false;
826 
827 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
828 		return false;
829 	if (rq->rq_flags & RQF_NOMERGE_FLAGS)
830 		return false;
831 
832 	return true;
833 }
834 
835 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
836 {
837 	if (bio_page(a) == bio_page(b) &&
838 	    bio_offset(a) == bio_offset(b))
839 		return true;
840 
841 	return false;
842 }
843 
844 static inline unsigned int blk_queue_depth(struct request_queue *q)
845 {
846 	if (q->queue_depth)
847 		return q->queue_depth;
848 
849 	return q->nr_requests;
850 }
851 
852 /*
853  * default timeout for SG_IO if none specified
854  */
855 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
856 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
857 
858 struct rq_map_data {
859 	struct page **pages;
860 	int page_order;
861 	int nr_entries;
862 	unsigned long offset;
863 	int null_mapped;
864 	int from_user;
865 };
866 
867 struct req_iterator {
868 	struct bvec_iter iter;
869 	struct bio *bio;
870 };
871 
872 /* This should not be used directly - use rq_for_each_segment */
873 #define for_each_bio(_bio)		\
874 	for (; _bio; _bio = _bio->bi_next)
875 #define __rq_for_each_bio(_bio, rq)	\
876 	if ((rq->bio))			\
877 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
878 
879 #define rq_for_each_segment(bvl, _rq, _iter)			\
880 	__rq_for_each_bio(_iter.bio, _rq)			\
881 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
882 
883 #define rq_for_each_bvec(bvl, _rq, _iter)			\
884 	__rq_for_each_bio(_iter.bio, _rq)			\
885 		bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
886 
887 #define rq_iter_last(bvec, _iter)				\
888 		(_iter.bio->bi_next == NULL &&			\
889 		 bio_iter_last(bvec, _iter.iter))
890 
891 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
892 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
893 #endif
894 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
895 extern void rq_flush_dcache_pages(struct request *rq);
896 #else
897 static inline void rq_flush_dcache_pages(struct request *rq)
898 {
899 }
900 #endif
901 
902 extern int blk_register_queue(struct gendisk *disk);
903 extern void blk_unregister_queue(struct gendisk *disk);
904 blk_qc_t submit_bio_noacct(struct bio *bio);
905 extern void blk_rq_init(struct request_queue *q, struct request *rq);
906 extern void blk_put_request(struct request *);
907 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
908 				       blk_mq_req_flags_t flags);
909 extern int blk_lld_busy(struct request_queue *q);
910 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
911 			     struct bio_set *bs, gfp_t gfp_mask,
912 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
913 			     void *data);
914 extern void blk_rq_unprep_clone(struct request *rq);
915 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
916 				     struct request *rq);
917 int blk_rq_append_bio(struct request *rq, struct bio *bio);
918 extern void blk_queue_split(struct bio **);
919 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
920 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
921 			      unsigned int, void __user *);
922 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
923 			  unsigned int, void __user *);
924 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
925 			 struct scsi_ioctl_command __user *);
926 extern int get_sg_io_hdr(struct sg_io_hdr *hdr, const void __user *argp);
927 extern int put_sg_io_hdr(const struct sg_io_hdr *hdr, void __user *argp);
928 
929 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
930 extern void blk_queue_exit(struct request_queue *q);
931 extern void blk_sync_queue(struct request_queue *q);
932 extern int blk_rq_map_user(struct request_queue *, struct request *,
933 			   struct rq_map_data *, void __user *, unsigned long,
934 			   gfp_t);
935 extern int blk_rq_unmap_user(struct bio *);
936 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
937 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
938 			       struct rq_map_data *, const struct iov_iter *,
939 			       gfp_t);
940 extern void blk_execute_rq(struct gendisk *, struct request *, int);
941 extern void blk_execute_rq_nowait(struct gendisk *,
942 				  struct request *, int, rq_end_io_fn *);
943 
944 /* Helper to convert REQ_OP_XXX to its string format XXX */
945 extern const char *blk_op_str(unsigned int op);
946 
947 int blk_status_to_errno(blk_status_t status);
948 blk_status_t errno_to_blk_status(int errno);
949 
950 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
951 
952 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
953 {
954 	return bdev->bd_disk->queue;	/* this is never NULL */
955 }
956 
957 /*
958  * The basic unit of block I/O is a sector. It is used in a number of contexts
959  * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
960  * bytes. Variables of type sector_t represent an offset or size that is a
961  * multiple of 512 bytes. Hence these two constants.
962  */
963 #ifndef SECTOR_SHIFT
964 #define SECTOR_SHIFT 9
965 #endif
966 #ifndef SECTOR_SIZE
967 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
968 #endif
969 
970 /*
971  * blk_rq_pos()			: the current sector
972  * blk_rq_bytes()		: bytes left in the entire request
973  * blk_rq_cur_bytes()		: bytes left in the current segment
974  * blk_rq_err_bytes()		: bytes left till the next error boundary
975  * blk_rq_sectors()		: sectors left in the entire request
976  * blk_rq_cur_sectors()		: sectors left in the current segment
977  * blk_rq_stats_sectors()	: sectors of the entire request used for stats
978  */
979 static inline sector_t blk_rq_pos(const struct request *rq)
980 {
981 	return rq->__sector;
982 }
983 
984 static inline unsigned int blk_rq_bytes(const struct request *rq)
985 {
986 	return rq->__data_len;
987 }
988 
989 static inline int blk_rq_cur_bytes(const struct request *rq)
990 {
991 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
992 }
993 
994 extern unsigned int blk_rq_err_bytes(const struct request *rq);
995 
996 static inline unsigned int blk_rq_sectors(const struct request *rq)
997 {
998 	return blk_rq_bytes(rq) >> SECTOR_SHIFT;
999 }
1000 
1001 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1002 {
1003 	return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1004 }
1005 
1006 static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
1007 {
1008 	return rq->stats_sectors;
1009 }
1010 
1011 #ifdef CONFIG_BLK_DEV_ZONED
1012 
1013 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
1014 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
1015 
1016 static inline unsigned int blk_rq_zone_no(struct request *rq)
1017 {
1018 	return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1019 }
1020 
1021 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1022 {
1023 	return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1024 }
1025 #endif /* CONFIG_BLK_DEV_ZONED */
1026 
1027 /*
1028  * Some commands like WRITE SAME have a payload or data transfer size which
1029  * is different from the size of the request.  Any driver that supports such
1030  * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1031  * calculate the data transfer size.
1032  */
1033 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1034 {
1035 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1036 		return rq->special_vec.bv_len;
1037 	return blk_rq_bytes(rq);
1038 }
1039 
1040 /*
1041  * Return the first full biovec in the request.  The caller needs to check that
1042  * there are any bvecs before calling this helper.
1043  */
1044 static inline struct bio_vec req_bvec(struct request *rq)
1045 {
1046 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1047 		return rq->special_vec;
1048 	return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
1049 }
1050 
1051 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1052 						     int op)
1053 {
1054 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1055 		return min(q->limits.max_discard_sectors,
1056 			   UINT_MAX >> SECTOR_SHIFT);
1057 
1058 	if (unlikely(op == REQ_OP_WRITE_SAME))
1059 		return q->limits.max_write_same_sectors;
1060 
1061 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
1062 		return q->limits.max_write_zeroes_sectors;
1063 
1064 	return q->limits.max_sectors;
1065 }
1066 
1067 /*
1068  * Return maximum size of a request at given offset. Only valid for
1069  * file system requests.
1070  */
1071 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1072 					       sector_t offset,
1073 					       unsigned int chunk_sectors)
1074 {
1075 	if (!chunk_sectors) {
1076 		if (q->limits.chunk_sectors)
1077 			chunk_sectors = q->limits.chunk_sectors;
1078 		else
1079 			return q->limits.max_sectors;
1080 	}
1081 
1082 	if (likely(is_power_of_2(chunk_sectors)))
1083 		chunk_sectors -= offset & (chunk_sectors - 1);
1084 	else
1085 		chunk_sectors -= sector_div(offset, chunk_sectors);
1086 
1087 	return min(q->limits.max_sectors, chunk_sectors);
1088 }
1089 
1090 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1091 						  sector_t offset)
1092 {
1093 	struct request_queue *q = rq->q;
1094 
1095 	if (blk_rq_is_passthrough(rq))
1096 		return q->limits.max_hw_sectors;
1097 
1098 	if (!q->limits.chunk_sectors ||
1099 	    req_op(rq) == REQ_OP_DISCARD ||
1100 	    req_op(rq) == REQ_OP_SECURE_ERASE)
1101 		return blk_queue_get_max_sectors(q, req_op(rq));
1102 
1103 	return min(blk_max_size_offset(q, offset, 0),
1104 			blk_queue_get_max_sectors(q, req_op(rq)));
1105 }
1106 
1107 static inline unsigned int blk_rq_count_bios(struct request *rq)
1108 {
1109 	unsigned int nr_bios = 0;
1110 	struct bio *bio;
1111 
1112 	__rq_for_each_bio(bio, rq)
1113 		nr_bios++;
1114 
1115 	return nr_bios;
1116 }
1117 
1118 void blk_steal_bios(struct bio_list *list, struct request *rq);
1119 
1120 /*
1121  * Request completion related functions.
1122  *
1123  * blk_update_request() completes given number of bytes and updates
1124  * the request without completing it.
1125  */
1126 extern bool blk_update_request(struct request *rq, blk_status_t error,
1127 			       unsigned int nr_bytes);
1128 
1129 extern void blk_abort_request(struct request *);
1130 
1131 /*
1132  * Access functions for manipulating queue properties
1133  */
1134 extern void blk_cleanup_queue(struct request_queue *);
1135 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
1136 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1137 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1138 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1139 extern void blk_queue_max_discard_segments(struct request_queue *,
1140 		unsigned short);
1141 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1142 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1143 		unsigned int max_discard_sectors);
1144 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1145 		unsigned int max_write_same_sectors);
1146 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1147 		unsigned int max_write_same_sectors);
1148 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1149 extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
1150 		unsigned int max_zone_append_sectors);
1151 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1152 void blk_queue_zone_write_granularity(struct request_queue *q,
1153 				      unsigned int size);
1154 extern void blk_queue_alignment_offset(struct request_queue *q,
1155 				       unsigned int alignment);
1156 void blk_queue_update_readahead(struct request_queue *q);
1157 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1158 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1159 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1160 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1161 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1162 extern void blk_set_default_limits(struct queue_limits *lim);
1163 extern void blk_set_stacking_limits(struct queue_limits *lim);
1164 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1165 			    sector_t offset);
1166 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1167 			      sector_t offset);
1168 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1169 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1170 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1171 extern void blk_queue_dma_alignment(struct request_queue *, int);
1172 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1173 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1174 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1175 extern void blk_queue_required_elevator_features(struct request_queue *q,
1176 						 unsigned int features);
1177 extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
1178 					      struct device *dev);
1179 
1180 /*
1181  * Number of physical segments as sent to the device.
1182  *
1183  * Normally this is the number of discontiguous data segments sent by the
1184  * submitter.  But for data-less command like discard we might have no
1185  * actual data segments submitted, but the driver might have to add it's
1186  * own special payload.  In that case we still return 1 here so that this
1187  * special payload will be mapped.
1188  */
1189 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1190 {
1191 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1192 		return 1;
1193 	return rq->nr_phys_segments;
1194 }
1195 
1196 /*
1197  * Number of discard segments (or ranges) the driver needs to fill in.
1198  * Each discard bio merged into a request is counted as one segment.
1199  */
1200 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1201 {
1202 	return max_t(unsigned short, rq->nr_phys_segments, 1);
1203 }
1204 
1205 int __blk_rq_map_sg(struct request_queue *q, struct request *rq,
1206 		struct scatterlist *sglist, struct scatterlist **last_sg);
1207 static inline int blk_rq_map_sg(struct request_queue *q, struct request *rq,
1208 		struct scatterlist *sglist)
1209 {
1210 	struct scatterlist *last_sg = NULL;
1211 
1212 	return __blk_rq_map_sg(q, rq, sglist, &last_sg);
1213 }
1214 extern void blk_dump_rq_flags(struct request *, char *);
1215 
1216 bool __must_check blk_get_queue(struct request_queue *);
1217 struct request_queue *blk_alloc_queue(int node_id);
1218 extern void blk_put_queue(struct request_queue *);
1219 extern void blk_set_queue_dying(struct request_queue *);
1220 
1221 #ifdef CONFIG_BLOCK
1222 /*
1223  * blk_plug permits building a queue of related requests by holding the I/O
1224  * fragments for a short period. This allows merging of sequential requests
1225  * into single larger request. As the requests are moved from a per-task list to
1226  * the device's request_queue in a batch, this results in improved scalability
1227  * as the lock contention for request_queue lock is reduced.
1228  *
1229  * It is ok not to disable preemption when adding the request to the plug list
1230  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1231  * the plug list when the task sleeps by itself. For details, please see
1232  * schedule() where blk_schedule_flush_plug() is called.
1233  */
1234 struct blk_plug {
1235 	struct list_head mq_list; /* blk-mq requests */
1236 	struct list_head cb_list; /* md requires an unplug callback */
1237 	unsigned short rq_count;
1238 	bool multiple_queues;
1239 	bool nowait;
1240 };
1241 #define BLK_MAX_REQUEST_COUNT 16
1242 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1243 
1244 struct blk_plug_cb;
1245 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1246 struct blk_plug_cb {
1247 	struct list_head list;
1248 	blk_plug_cb_fn callback;
1249 	void *data;
1250 };
1251 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1252 					     void *data, int size);
1253 extern void blk_start_plug(struct blk_plug *);
1254 extern void blk_finish_plug(struct blk_plug *);
1255 extern void blk_flush_plug_list(struct blk_plug *, bool);
1256 
1257 static inline void blk_flush_plug(struct task_struct *tsk)
1258 {
1259 	struct blk_plug *plug = tsk->plug;
1260 
1261 	if (plug)
1262 		blk_flush_plug_list(plug, false);
1263 }
1264 
1265 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1266 {
1267 	struct blk_plug *plug = tsk->plug;
1268 
1269 	if (plug)
1270 		blk_flush_plug_list(plug, true);
1271 }
1272 
1273 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1274 {
1275 	struct blk_plug *plug = tsk->plug;
1276 
1277 	return plug &&
1278 		 (!list_empty(&plug->mq_list) ||
1279 		 !list_empty(&plug->cb_list));
1280 }
1281 
1282 int blkdev_issue_flush(struct block_device *bdev);
1283 long nr_blockdev_pages(void);
1284 #else /* CONFIG_BLOCK */
1285 struct blk_plug {
1286 };
1287 
1288 static inline void blk_start_plug(struct blk_plug *plug)
1289 {
1290 }
1291 
1292 static inline void blk_finish_plug(struct blk_plug *plug)
1293 {
1294 }
1295 
1296 static inline void blk_flush_plug(struct task_struct *task)
1297 {
1298 }
1299 
1300 static inline void blk_schedule_flush_plug(struct task_struct *task)
1301 {
1302 }
1303 
1304 
1305 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1306 {
1307 	return false;
1308 }
1309 
1310 static inline int blkdev_issue_flush(struct block_device *bdev)
1311 {
1312 	return 0;
1313 }
1314 
1315 static inline long nr_blockdev_pages(void)
1316 {
1317 	return 0;
1318 }
1319 #endif /* CONFIG_BLOCK */
1320 
1321 extern void blk_io_schedule(void);
1322 
1323 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1324 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1325 
1326 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
1327 
1328 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1329 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1330 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1331 		sector_t nr_sects, gfp_t gfp_mask, int flags,
1332 		struct bio **biop);
1333 
1334 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1335 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1336 
1337 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1338 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1339 		unsigned flags);
1340 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1341 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1342 
1343 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1344 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1345 {
1346 	return blkdev_issue_discard(sb->s_bdev,
1347 				    block << (sb->s_blocksize_bits -
1348 					      SECTOR_SHIFT),
1349 				    nr_blocks << (sb->s_blocksize_bits -
1350 						  SECTOR_SHIFT),
1351 				    gfp_mask, flags);
1352 }
1353 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1354 		sector_t nr_blocks, gfp_t gfp_mask)
1355 {
1356 	return blkdev_issue_zeroout(sb->s_bdev,
1357 				    block << (sb->s_blocksize_bits -
1358 					      SECTOR_SHIFT),
1359 				    nr_blocks << (sb->s_blocksize_bits -
1360 						  SECTOR_SHIFT),
1361 				    gfp_mask, 0);
1362 }
1363 
1364 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1365 
1366 static inline bool bdev_is_partition(struct block_device *bdev)
1367 {
1368 	return bdev->bd_partno;
1369 }
1370 
1371 enum blk_default_limits {
1372 	BLK_MAX_SEGMENTS	= 128,
1373 	BLK_SAFE_MAX_SECTORS	= 255,
1374 	BLK_DEF_MAX_SECTORS	= 2560,
1375 	BLK_MAX_SEGMENT_SIZE	= 65536,
1376 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1377 };
1378 
1379 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1380 {
1381 	return q->limits.seg_boundary_mask;
1382 }
1383 
1384 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1385 {
1386 	return q->limits.virt_boundary_mask;
1387 }
1388 
1389 static inline unsigned int queue_max_sectors(const struct request_queue *q)
1390 {
1391 	return q->limits.max_sectors;
1392 }
1393 
1394 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1395 {
1396 	return q->limits.max_hw_sectors;
1397 }
1398 
1399 static inline unsigned short queue_max_segments(const struct request_queue *q)
1400 {
1401 	return q->limits.max_segments;
1402 }
1403 
1404 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1405 {
1406 	return q->limits.max_discard_segments;
1407 }
1408 
1409 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1410 {
1411 	return q->limits.max_segment_size;
1412 }
1413 
1414 static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
1415 {
1416 
1417 	const struct queue_limits *l = &q->limits;
1418 
1419 	return min(l->max_zone_append_sectors, l->max_sectors);
1420 }
1421 
1422 static inline unsigned queue_logical_block_size(const struct request_queue *q)
1423 {
1424 	int retval = 512;
1425 
1426 	if (q && q->limits.logical_block_size)
1427 		retval = q->limits.logical_block_size;
1428 
1429 	return retval;
1430 }
1431 
1432 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1433 {
1434 	return queue_logical_block_size(bdev_get_queue(bdev));
1435 }
1436 
1437 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1438 {
1439 	return q->limits.physical_block_size;
1440 }
1441 
1442 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1443 {
1444 	return queue_physical_block_size(bdev_get_queue(bdev));
1445 }
1446 
1447 static inline unsigned int queue_io_min(const struct request_queue *q)
1448 {
1449 	return q->limits.io_min;
1450 }
1451 
1452 static inline int bdev_io_min(struct block_device *bdev)
1453 {
1454 	return queue_io_min(bdev_get_queue(bdev));
1455 }
1456 
1457 static inline unsigned int queue_io_opt(const struct request_queue *q)
1458 {
1459 	return q->limits.io_opt;
1460 }
1461 
1462 static inline int bdev_io_opt(struct block_device *bdev)
1463 {
1464 	return queue_io_opt(bdev_get_queue(bdev));
1465 }
1466 
1467 static inline unsigned int
1468 queue_zone_write_granularity(const struct request_queue *q)
1469 {
1470 	return q->limits.zone_write_granularity;
1471 }
1472 
1473 static inline unsigned int
1474 bdev_zone_write_granularity(struct block_device *bdev)
1475 {
1476 	return queue_zone_write_granularity(bdev_get_queue(bdev));
1477 }
1478 
1479 static inline int queue_alignment_offset(const struct request_queue *q)
1480 {
1481 	if (q->limits.misaligned)
1482 		return -1;
1483 
1484 	return q->limits.alignment_offset;
1485 }
1486 
1487 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1488 {
1489 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1490 	unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1491 		<< SECTOR_SHIFT;
1492 
1493 	return (granularity + lim->alignment_offset - alignment) % granularity;
1494 }
1495 
1496 static inline int bdev_alignment_offset(struct block_device *bdev)
1497 {
1498 	struct request_queue *q = bdev_get_queue(bdev);
1499 
1500 	if (q->limits.misaligned)
1501 		return -1;
1502 	if (bdev_is_partition(bdev))
1503 		return queue_limit_alignment_offset(&q->limits,
1504 				bdev->bd_start_sect);
1505 	return q->limits.alignment_offset;
1506 }
1507 
1508 static inline int queue_discard_alignment(const struct request_queue *q)
1509 {
1510 	if (q->limits.discard_misaligned)
1511 		return -1;
1512 
1513 	return q->limits.discard_alignment;
1514 }
1515 
1516 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1517 {
1518 	unsigned int alignment, granularity, offset;
1519 
1520 	if (!lim->max_discard_sectors)
1521 		return 0;
1522 
1523 	/* Why are these in bytes, not sectors? */
1524 	alignment = lim->discard_alignment >> SECTOR_SHIFT;
1525 	granularity = lim->discard_granularity >> SECTOR_SHIFT;
1526 	if (!granularity)
1527 		return 0;
1528 
1529 	/* Offset of the partition start in 'granularity' sectors */
1530 	offset = sector_div(sector, granularity);
1531 
1532 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1533 	offset = (granularity + alignment - offset) % granularity;
1534 
1535 	/* Turn it back into bytes, gaah */
1536 	return offset << SECTOR_SHIFT;
1537 }
1538 
1539 static inline int bdev_discard_alignment(struct block_device *bdev)
1540 {
1541 	struct request_queue *q = bdev_get_queue(bdev);
1542 
1543 	if (bdev_is_partition(bdev))
1544 		return queue_limit_discard_alignment(&q->limits,
1545 				bdev->bd_start_sect);
1546 	return q->limits.discard_alignment;
1547 }
1548 
1549 static inline unsigned int bdev_write_same(struct block_device *bdev)
1550 {
1551 	struct request_queue *q = bdev_get_queue(bdev);
1552 
1553 	if (q)
1554 		return q->limits.max_write_same_sectors;
1555 
1556 	return 0;
1557 }
1558 
1559 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1560 {
1561 	struct request_queue *q = bdev_get_queue(bdev);
1562 
1563 	if (q)
1564 		return q->limits.max_write_zeroes_sectors;
1565 
1566 	return 0;
1567 }
1568 
1569 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1570 {
1571 	struct request_queue *q = bdev_get_queue(bdev);
1572 
1573 	if (q)
1574 		return blk_queue_zoned_model(q);
1575 
1576 	return BLK_ZONED_NONE;
1577 }
1578 
1579 static inline bool bdev_is_zoned(struct block_device *bdev)
1580 {
1581 	struct request_queue *q = bdev_get_queue(bdev);
1582 
1583 	if (q)
1584 		return blk_queue_is_zoned(q);
1585 
1586 	return false;
1587 }
1588 
1589 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1590 {
1591 	struct request_queue *q = bdev_get_queue(bdev);
1592 
1593 	if (q)
1594 		return blk_queue_zone_sectors(q);
1595 	return 0;
1596 }
1597 
1598 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1599 {
1600 	struct request_queue *q = bdev_get_queue(bdev);
1601 
1602 	if (q)
1603 		return queue_max_open_zones(q);
1604 	return 0;
1605 }
1606 
1607 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1608 {
1609 	struct request_queue *q = bdev_get_queue(bdev);
1610 
1611 	if (q)
1612 		return queue_max_active_zones(q);
1613 	return 0;
1614 }
1615 
1616 static inline int queue_dma_alignment(const struct request_queue *q)
1617 {
1618 	return q ? q->dma_alignment : 511;
1619 }
1620 
1621 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1622 				 unsigned int len)
1623 {
1624 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1625 	return !(addr & alignment) && !(len & alignment);
1626 }
1627 
1628 /* assumes size > 256 */
1629 static inline unsigned int blksize_bits(unsigned int size)
1630 {
1631 	unsigned int bits = 8;
1632 	do {
1633 		bits++;
1634 		size >>= 1;
1635 	} while (size > 256);
1636 	return bits;
1637 }
1638 
1639 static inline unsigned int block_size(struct block_device *bdev)
1640 {
1641 	return 1 << bdev->bd_inode->i_blkbits;
1642 }
1643 
1644 int kblockd_schedule_work(struct work_struct *work);
1645 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1646 
1647 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1648 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1649 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1650 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1651 
1652 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1653 
1654 enum blk_integrity_flags {
1655 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1656 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1657 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1658 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1659 };
1660 
1661 struct blk_integrity_iter {
1662 	void			*prot_buf;
1663 	void			*data_buf;
1664 	sector_t		seed;
1665 	unsigned int		data_size;
1666 	unsigned short		interval;
1667 	const char		*disk_name;
1668 };
1669 
1670 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1671 typedef void (integrity_prepare_fn) (struct request *);
1672 typedef void (integrity_complete_fn) (struct request *, unsigned int);
1673 
1674 struct blk_integrity_profile {
1675 	integrity_processing_fn		*generate_fn;
1676 	integrity_processing_fn		*verify_fn;
1677 	integrity_prepare_fn		*prepare_fn;
1678 	integrity_complete_fn		*complete_fn;
1679 	const char			*name;
1680 };
1681 
1682 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1683 extern void blk_integrity_unregister(struct gendisk *);
1684 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1685 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1686 				   struct scatterlist *);
1687 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1688 
1689 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1690 {
1691 	struct blk_integrity *bi = &disk->queue->integrity;
1692 
1693 	if (!bi->profile)
1694 		return NULL;
1695 
1696 	return bi;
1697 }
1698 
1699 static inline
1700 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1701 {
1702 	return blk_get_integrity(bdev->bd_disk);
1703 }
1704 
1705 static inline bool
1706 blk_integrity_queue_supports_integrity(struct request_queue *q)
1707 {
1708 	return q->integrity.profile;
1709 }
1710 
1711 static inline bool blk_integrity_rq(struct request *rq)
1712 {
1713 	return rq->cmd_flags & REQ_INTEGRITY;
1714 }
1715 
1716 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1717 						    unsigned int segs)
1718 {
1719 	q->limits.max_integrity_segments = segs;
1720 }
1721 
1722 static inline unsigned short
1723 queue_max_integrity_segments(const struct request_queue *q)
1724 {
1725 	return q->limits.max_integrity_segments;
1726 }
1727 
1728 /**
1729  * bio_integrity_intervals - Return number of integrity intervals for a bio
1730  * @bi:		blk_integrity profile for device
1731  * @sectors:	Size of the bio in 512-byte sectors
1732  *
1733  * Description: The block layer calculates everything in 512 byte
1734  * sectors but integrity metadata is done in terms of the data integrity
1735  * interval size of the storage device.  Convert the block layer sectors
1736  * to the appropriate number of integrity intervals.
1737  */
1738 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1739 						   unsigned int sectors)
1740 {
1741 	return sectors >> (bi->interval_exp - 9);
1742 }
1743 
1744 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1745 					       unsigned int sectors)
1746 {
1747 	return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1748 }
1749 
1750 /*
1751  * Return the first bvec that contains integrity data.  Only drivers that are
1752  * limited to a single integrity segment should use this helper.
1753  */
1754 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1755 {
1756 	if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1757 		return NULL;
1758 	return rq->bio->bi_integrity->bip_vec;
1759 }
1760 
1761 #else /* CONFIG_BLK_DEV_INTEGRITY */
1762 
1763 struct bio;
1764 struct block_device;
1765 struct gendisk;
1766 struct blk_integrity;
1767 
1768 static inline int blk_integrity_rq(struct request *rq)
1769 {
1770 	return 0;
1771 }
1772 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1773 					    struct bio *b)
1774 {
1775 	return 0;
1776 }
1777 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1778 					  struct bio *b,
1779 					  struct scatterlist *s)
1780 {
1781 	return 0;
1782 }
1783 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1784 {
1785 	return NULL;
1786 }
1787 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1788 {
1789 	return NULL;
1790 }
1791 static inline bool
1792 blk_integrity_queue_supports_integrity(struct request_queue *q)
1793 {
1794 	return false;
1795 }
1796 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1797 {
1798 	return 0;
1799 }
1800 static inline void blk_integrity_register(struct gendisk *d,
1801 					 struct blk_integrity *b)
1802 {
1803 }
1804 static inline void blk_integrity_unregister(struct gendisk *d)
1805 {
1806 }
1807 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1808 						    unsigned int segs)
1809 {
1810 }
1811 static inline unsigned short queue_max_integrity_segments(const struct request_queue *q)
1812 {
1813 	return 0;
1814 }
1815 
1816 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1817 						   unsigned int sectors)
1818 {
1819 	return 0;
1820 }
1821 
1822 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1823 					       unsigned int sectors)
1824 {
1825 	return 0;
1826 }
1827 
1828 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1829 {
1830 	return NULL;
1831 }
1832 
1833 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1834 
1835 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1836 
1837 bool blk_ksm_register(struct blk_keyslot_manager *ksm, struct request_queue *q);
1838 
1839 void blk_ksm_unregister(struct request_queue *q);
1840 
1841 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1842 
1843 static inline bool blk_ksm_register(struct blk_keyslot_manager *ksm,
1844 				    struct request_queue *q)
1845 {
1846 	return true;
1847 }
1848 
1849 static inline void blk_ksm_unregister(struct request_queue *q) { }
1850 
1851 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1852 
1853 
1854 struct block_device_operations {
1855 	blk_qc_t (*submit_bio) (struct bio *bio);
1856 	int (*open) (struct block_device *, fmode_t);
1857 	void (*release) (struct gendisk *, fmode_t);
1858 	int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1859 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1860 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1861 	unsigned int (*check_events) (struct gendisk *disk,
1862 				      unsigned int clearing);
1863 	void (*unlock_native_capacity) (struct gendisk *);
1864 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1865 	int (*set_read_only)(struct block_device *bdev, bool ro);
1866 	/* this callback is with swap_lock and sometimes page table lock held */
1867 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1868 	int (*report_zones)(struct gendisk *, sector_t sector,
1869 			unsigned int nr_zones, report_zones_cb cb, void *data);
1870 	char *(*devnode)(struct gendisk *disk, umode_t *mode);
1871 	struct module *owner;
1872 	const struct pr_ops *pr_ops;
1873 };
1874 
1875 #ifdef CONFIG_COMPAT
1876 extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1877 				      unsigned int, unsigned long);
1878 #else
1879 #define blkdev_compat_ptr_ioctl NULL
1880 #endif
1881 
1882 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1883 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1884 						struct writeback_control *);
1885 
1886 #ifdef CONFIG_BLK_DEV_ZONED
1887 bool blk_req_needs_zone_write_lock(struct request *rq);
1888 bool blk_req_zone_write_trylock(struct request *rq);
1889 void __blk_req_zone_write_lock(struct request *rq);
1890 void __blk_req_zone_write_unlock(struct request *rq);
1891 
1892 static inline void blk_req_zone_write_lock(struct request *rq)
1893 {
1894 	if (blk_req_needs_zone_write_lock(rq))
1895 		__blk_req_zone_write_lock(rq);
1896 }
1897 
1898 static inline void blk_req_zone_write_unlock(struct request *rq)
1899 {
1900 	if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1901 		__blk_req_zone_write_unlock(rq);
1902 }
1903 
1904 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1905 {
1906 	return rq->q->seq_zones_wlock &&
1907 		test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1908 }
1909 
1910 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1911 {
1912 	if (!blk_req_needs_zone_write_lock(rq))
1913 		return true;
1914 	return !blk_req_zone_is_write_locked(rq);
1915 }
1916 #else
1917 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1918 {
1919 	return false;
1920 }
1921 
1922 static inline void blk_req_zone_write_lock(struct request *rq)
1923 {
1924 }
1925 
1926 static inline void blk_req_zone_write_unlock(struct request *rq)
1927 {
1928 }
1929 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1930 {
1931 	return false;
1932 }
1933 
1934 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1935 {
1936 	return true;
1937 }
1938 #endif /* CONFIG_BLK_DEV_ZONED */
1939 
1940 static inline void blk_wake_io_task(struct task_struct *waiter)
1941 {
1942 	/*
1943 	 * If we're polling, the task itself is doing the completions. For
1944 	 * that case, we don't need to signal a wakeup, it's enough to just
1945 	 * mark us as RUNNING.
1946 	 */
1947 	if (waiter == current)
1948 		__set_current_state(TASK_RUNNING);
1949 	else
1950 		wake_up_process(waiter);
1951 }
1952 
1953 unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1954 		unsigned int op);
1955 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1956 		unsigned long start_time);
1957 
1958 unsigned long bio_start_io_acct(struct bio *bio);
1959 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1960 		struct block_device *orig_bdev);
1961 
1962 /**
1963  * bio_end_io_acct - end I/O accounting for bio based drivers
1964  * @bio:	bio to end account for
1965  * @start:	start time returned by bio_start_io_acct()
1966  */
1967 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1968 {
1969 	return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1970 }
1971 
1972 int bdev_read_only(struct block_device *bdev);
1973 int set_blocksize(struct block_device *bdev, int size);
1974 
1975 const char *bdevname(struct block_device *bdev, char *buffer);
1976 int lookup_bdev(const char *pathname, dev_t *dev);
1977 
1978 void blkdev_show(struct seq_file *seqf, off_t offset);
1979 
1980 #define BDEVNAME_SIZE	32	/* Largest string for a blockdev identifier */
1981 #define BDEVT_SIZE	10	/* Largest string for MAJ:MIN for blkdev */
1982 #ifdef CONFIG_BLOCK
1983 #define BLKDEV_MAJOR_MAX	512
1984 #else
1985 #define BLKDEV_MAJOR_MAX	0
1986 #endif
1987 
1988 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1989 		void *holder);
1990 struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
1991 int bd_prepare_to_claim(struct block_device *bdev, void *holder);
1992 void bd_abort_claiming(struct block_device *bdev, void *holder);
1993 void blkdev_put(struct block_device *bdev, fmode_t mode);
1994 
1995 /* just for blk-cgroup, don't use elsewhere */
1996 struct block_device *blkdev_get_no_open(dev_t dev);
1997 void blkdev_put_no_open(struct block_device *bdev);
1998 
1999 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
2000 void bdev_add(struct block_device *bdev, dev_t dev);
2001 struct block_device *I_BDEV(struct inode *inode);
2002 struct block_device *bdgrab(struct block_device *bdev);
2003 void bdput(struct block_device *);
2004 int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
2005 		loff_t lend);
2006 
2007 #ifdef CONFIG_BLOCK
2008 void invalidate_bdev(struct block_device *bdev);
2009 int sync_blockdev(struct block_device *bdev);
2010 #else
2011 static inline void invalidate_bdev(struct block_device *bdev)
2012 {
2013 }
2014 static inline int sync_blockdev(struct block_device *bdev)
2015 {
2016 	return 0;
2017 }
2018 #endif
2019 int fsync_bdev(struct block_device *bdev);
2020 
2021 int freeze_bdev(struct block_device *bdev);
2022 int thaw_bdev(struct block_device *bdev);
2023 
2024 #endif /* _LINUX_BLKDEV_H */
2025