xref: /linux-6.15/include/linux/blkdev.h (revision d6a6a555)
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 	unsigned int		bio_max_bytes;
330 
331 	enum blk_bounce		bounce;
332 	unsigned long		seg_boundary_mask;
333 	unsigned long		virt_boundary_mask;
334 
335 	unsigned int		max_hw_sectors;
336 	unsigned int		max_dev_sectors;
337 	unsigned int		chunk_sectors;
338 	unsigned int		max_sectors;
339 	unsigned int		max_segment_size;
340 	unsigned int		physical_block_size;
341 	unsigned int		logical_block_size;
342 	unsigned int		alignment_offset;
343 	unsigned int		io_min;
344 	unsigned int		io_opt;
345 	unsigned int		max_discard_sectors;
346 	unsigned int		max_hw_discard_sectors;
347 	unsigned int		max_write_same_sectors;
348 	unsigned int		max_write_zeroes_sectors;
349 	unsigned int		max_zone_append_sectors;
350 	unsigned int		discard_granularity;
351 	unsigned int		discard_alignment;
352 	unsigned int		zone_write_granularity;
353 
354 	unsigned short		max_segments;
355 	unsigned short		max_integrity_segments;
356 	unsigned short		max_discard_segments;
357 
358 	unsigned char		misaligned;
359 	unsigned char		discard_misaligned;
360 	unsigned char		raid_partial_stripes_expensive;
361 	enum blk_zoned_model	zoned;
362 };
363 
364 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
365 			       void *data);
366 
367 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
368 
369 #ifdef CONFIG_BLK_DEV_ZONED
370 
371 #define BLK_ALL_ZONES  ((unsigned int)-1)
372 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
373 			unsigned int nr_zones, report_zones_cb cb, void *data);
374 unsigned int blkdev_nr_zones(struct gendisk *disk);
375 extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
376 			    sector_t sectors, sector_t nr_sectors,
377 			    gfp_t gfp_mask);
378 int blk_revalidate_disk_zones(struct gendisk *disk,
379 			      void (*update_driver_data)(struct gendisk *disk));
380 
381 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
382 				     unsigned int cmd, unsigned long arg);
383 extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
384 				  unsigned int cmd, unsigned long arg);
385 
386 #else /* CONFIG_BLK_DEV_ZONED */
387 
388 static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
389 {
390 	return 0;
391 }
392 
393 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
394 					    fmode_t mode, unsigned int cmd,
395 					    unsigned long arg)
396 {
397 	return -ENOTTY;
398 }
399 
400 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
401 					 fmode_t mode, unsigned int cmd,
402 					 unsigned long arg)
403 {
404 	return -ENOTTY;
405 }
406 
407 #endif /* CONFIG_BLK_DEV_ZONED */
408 
409 struct request_queue {
410 	struct request		*last_merge;
411 	struct elevator_queue	*elevator;
412 
413 	struct percpu_ref	q_usage_counter;
414 
415 	struct blk_queue_stats	*stats;
416 	struct rq_qos		*rq_qos;
417 
418 	const struct blk_mq_ops	*mq_ops;
419 
420 	/* sw queues */
421 	struct blk_mq_ctx __percpu	*queue_ctx;
422 
423 	unsigned int		queue_depth;
424 
425 	/* hw dispatch queues */
426 	struct blk_mq_hw_ctx	**queue_hw_ctx;
427 	unsigned int		nr_hw_queues;
428 
429 	struct backing_dev_info	*backing_dev_info;
430 
431 	/*
432 	 * The queue owner gets to use this for whatever they like.
433 	 * ll_rw_blk doesn't touch it.
434 	 */
435 	void			*queuedata;
436 
437 	/*
438 	 * various queue flags, see QUEUE_* below
439 	 */
440 	unsigned long		queue_flags;
441 	/*
442 	 * Number of contexts that have called blk_set_pm_only(). If this
443 	 * counter is above zero then only RQF_PM requests are processed.
444 	 */
445 	atomic_t		pm_only;
446 
447 	/*
448 	 * ida allocated id for this queue.  Used to index queues from
449 	 * ioctx.
450 	 */
451 	int			id;
452 
453 	spinlock_t		queue_lock;
454 
455 	/*
456 	 * queue kobject
457 	 */
458 	struct kobject kobj;
459 
460 	/*
461 	 * mq queue kobject
462 	 */
463 	struct kobject *mq_kobj;
464 
465 #ifdef  CONFIG_BLK_DEV_INTEGRITY
466 	struct blk_integrity integrity;
467 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
468 
469 #ifdef CONFIG_PM
470 	struct device		*dev;
471 	enum rpm_status		rpm_status;
472 #endif
473 
474 	/*
475 	 * queue settings
476 	 */
477 	unsigned long		nr_requests;	/* Max # of requests */
478 
479 	unsigned int		dma_pad_mask;
480 	unsigned int		dma_alignment;
481 
482 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
483 	/* Inline crypto capabilities */
484 	struct blk_keyslot_manager *ksm;
485 #endif
486 
487 	unsigned int		rq_timeout;
488 	int			poll_nsec;
489 
490 	struct blk_stat_callback	*poll_cb;
491 	struct blk_rq_stat	poll_stat[BLK_MQ_POLL_STATS_BKTS];
492 
493 	struct timer_list	timeout;
494 	struct work_struct	timeout_work;
495 
496 	atomic_t		nr_active_requests_shared_sbitmap;
497 
498 	struct list_head	icq_list;
499 #ifdef CONFIG_BLK_CGROUP
500 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
501 	struct blkcg_gq		*root_blkg;
502 	struct list_head	blkg_list;
503 #endif
504 
505 	struct queue_limits	limits;
506 
507 	unsigned int		required_elevator_features;
508 
509 #ifdef CONFIG_BLK_DEV_ZONED
510 	/*
511 	 * Zoned block device information for request dispatch control.
512 	 * nr_zones is the total number of zones of the device. This is always
513 	 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
514 	 * bits which indicates if a zone is conventional (bit set) or
515 	 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
516 	 * bits which indicates if a zone is write locked, that is, if a write
517 	 * request targeting the zone was dispatched. All three fields are
518 	 * initialized by the low level device driver (e.g. scsi/sd.c).
519 	 * Stacking drivers (device mappers) may or may not initialize
520 	 * these fields.
521 	 *
522 	 * Reads of this information must be protected with blk_queue_enter() /
523 	 * blk_queue_exit(). Modifying this information is only allowed while
524 	 * no requests are being processed. See also blk_mq_freeze_queue() and
525 	 * blk_mq_unfreeze_queue().
526 	 */
527 	unsigned int		nr_zones;
528 	unsigned long		*conv_zones_bitmap;
529 	unsigned long		*seq_zones_wlock;
530 	unsigned int		max_open_zones;
531 	unsigned int		max_active_zones;
532 #endif /* CONFIG_BLK_DEV_ZONED */
533 
534 	/*
535 	 * sg stuff
536 	 */
537 	unsigned int		sg_timeout;
538 	unsigned int		sg_reserved_size;
539 	int			node;
540 	struct mutex		debugfs_mutex;
541 #ifdef CONFIG_BLK_DEV_IO_TRACE
542 	struct blk_trace __rcu	*blk_trace;
543 #endif
544 	/*
545 	 * for flush operations
546 	 */
547 	struct blk_flush_queue	*fq;
548 
549 	struct list_head	requeue_list;
550 	spinlock_t		requeue_lock;
551 	struct delayed_work	requeue_work;
552 
553 	struct mutex		sysfs_lock;
554 	struct mutex		sysfs_dir_lock;
555 
556 	/*
557 	 * for reusing dead hctx instance in case of updating
558 	 * nr_hw_queues
559 	 */
560 	struct list_head	unused_hctx_list;
561 	spinlock_t		unused_hctx_lock;
562 
563 	int			mq_freeze_depth;
564 
565 #if defined(CONFIG_BLK_DEV_BSG)
566 	struct bsg_class_device bsg_dev;
567 #endif
568 
569 #ifdef CONFIG_BLK_DEV_THROTTLING
570 	/* Throttle data */
571 	struct throtl_data *td;
572 #endif
573 	struct rcu_head		rcu_head;
574 	wait_queue_head_t	mq_freeze_wq;
575 	/*
576 	 * Protect concurrent access to q_usage_counter by
577 	 * percpu_ref_kill() and percpu_ref_reinit().
578 	 */
579 	struct mutex		mq_freeze_lock;
580 
581 	struct blk_mq_tag_set	*tag_set;
582 	struct list_head	tag_set_list;
583 	struct bio_set		bio_split;
584 
585 	struct dentry		*debugfs_dir;
586 
587 #ifdef CONFIG_BLK_DEBUG_FS
588 	struct dentry		*sched_debugfs_dir;
589 	struct dentry		*rqos_debugfs_dir;
590 #endif
591 
592 	bool			mq_sysfs_init_done;
593 
594 	size_t			cmd_size;
595 
596 #define BLK_MAX_WRITE_HINTS	5
597 	u64			write_hints[BLK_MAX_WRITE_HINTS];
598 };
599 
600 /* Keep blk_queue_flag_name[] in sync with the definitions below */
601 #define QUEUE_FLAG_STOPPED	0	/* queue is stopped */
602 #define QUEUE_FLAG_DYING	1	/* queue being torn down */
603 #define QUEUE_FLAG_NOMERGES     3	/* disable merge attempts */
604 #define QUEUE_FLAG_SAME_COMP	4	/* complete on same CPU-group */
605 #define QUEUE_FLAG_FAIL_IO	5	/* fake timeout */
606 #define QUEUE_FLAG_NONROT	6	/* non-rotational device (SSD) */
607 #define QUEUE_FLAG_VIRT		QUEUE_FLAG_NONROT /* paravirt device */
608 #define QUEUE_FLAG_IO_STAT	7	/* do disk/partitions IO accounting */
609 #define QUEUE_FLAG_DISCARD	8	/* supports DISCARD */
610 #define QUEUE_FLAG_NOXMERGES	9	/* No extended merges */
611 #define QUEUE_FLAG_ADD_RANDOM	10	/* Contributes to random pool */
612 #define QUEUE_FLAG_SECERASE	11	/* supports secure erase */
613 #define QUEUE_FLAG_SAME_FORCE	12	/* force complete on same CPU */
614 #define QUEUE_FLAG_DEAD		13	/* queue tear-down finished */
615 #define QUEUE_FLAG_INIT_DONE	14	/* queue is initialized */
616 #define QUEUE_FLAG_STABLE_WRITES 15	/* don't modify blks until WB is done */
617 #define QUEUE_FLAG_POLL		16	/* IO polling enabled if set */
618 #define QUEUE_FLAG_WC		17	/* Write back caching */
619 #define QUEUE_FLAG_FUA		18	/* device supports FUA writes */
620 #define QUEUE_FLAG_DAX		19	/* device supports DAX */
621 #define QUEUE_FLAG_STATS	20	/* track IO start and completion times */
622 #define QUEUE_FLAG_POLL_STATS	21	/* collecting stats for hybrid polling */
623 #define QUEUE_FLAG_REGISTERED	22	/* queue has been registered to a disk */
624 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23	/* queue supports SCSI commands */
625 #define QUEUE_FLAG_QUIESCED	24	/* queue has been quiesced */
626 #define QUEUE_FLAG_PCI_P2PDMA	25	/* device supports PCI p2p requests */
627 #define QUEUE_FLAG_ZONE_RESETALL 26	/* supports Zone Reset All */
628 #define QUEUE_FLAG_RQ_ALLOC_TIME 27	/* record rq->alloc_time_ns */
629 #define QUEUE_FLAG_HCTX_ACTIVE	28	/* at least one blk-mq hctx is active */
630 #define QUEUE_FLAG_NOWAIT       29	/* device supports NOWAIT */
631 
632 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
633 				 (1 << QUEUE_FLAG_SAME_COMP) |		\
634 				 (1 << QUEUE_FLAG_NOWAIT))
635 
636 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
637 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
638 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
639 
640 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
641 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
642 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
643 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
644 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
645 #define blk_queue_noxmerges(q)	\
646 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
647 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
648 #define blk_queue_stable_writes(q) \
649 	test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
650 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
651 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
652 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
653 #define blk_queue_zone_resetall(q)	\
654 	test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
655 #define blk_queue_secure_erase(q) \
656 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
657 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
658 #define blk_queue_scsi_passthrough(q)	\
659 	test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
660 #define blk_queue_pci_p2pdma(q)	\
661 	test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
662 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
663 #define blk_queue_rq_alloc_time(q)	\
664 	test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
665 #else
666 #define blk_queue_rq_alloc_time(q)	false
667 #endif
668 
669 #define blk_noretry_request(rq) \
670 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
671 			     REQ_FAILFAST_DRIVER))
672 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
673 #define blk_queue_pm_only(q)	atomic_read(&(q)->pm_only)
674 #define blk_queue_fua(q)	test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
675 #define blk_queue_registered(q)	test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
676 #define blk_queue_nowait(q)	test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
677 
678 extern void blk_set_pm_only(struct request_queue *q);
679 extern void blk_clear_pm_only(struct request_queue *q);
680 
681 static inline bool blk_account_rq(struct request *rq)
682 {
683 	return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
684 }
685 
686 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
687 
688 #define rq_data_dir(rq)		(op_is_write(req_op(rq)) ? WRITE : READ)
689 
690 #define rq_dma_dir(rq) \
691 	(op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
692 
693 #define dma_map_bvec(dev, bv, dir, attrs) \
694 	dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
695 	(dir), (attrs))
696 
697 #define queue_to_disk(q)	(dev_to_disk(kobj_to_dev((q)->kobj.parent)))
698 
699 static inline bool queue_is_mq(struct request_queue *q)
700 {
701 	return q->mq_ops;
702 }
703 
704 #ifdef CONFIG_PM
705 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
706 {
707 	return q->rpm_status;
708 }
709 #else
710 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
711 {
712 	return RPM_ACTIVE;
713 }
714 #endif
715 
716 static inline enum blk_zoned_model
717 blk_queue_zoned_model(struct request_queue *q)
718 {
719 	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
720 		return q->limits.zoned;
721 	return BLK_ZONED_NONE;
722 }
723 
724 static inline bool blk_queue_is_zoned(struct request_queue *q)
725 {
726 	switch (blk_queue_zoned_model(q)) {
727 	case BLK_ZONED_HA:
728 	case BLK_ZONED_HM:
729 		return true;
730 	default:
731 		return false;
732 	}
733 }
734 
735 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
736 {
737 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
738 }
739 
740 #ifdef CONFIG_BLK_DEV_ZONED
741 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
742 {
743 	return blk_queue_is_zoned(q) ? q->nr_zones : 0;
744 }
745 
746 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
747 					     sector_t sector)
748 {
749 	if (!blk_queue_is_zoned(q))
750 		return 0;
751 	return sector >> ilog2(q->limits.chunk_sectors);
752 }
753 
754 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
755 					 sector_t sector)
756 {
757 	if (!blk_queue_is_zoned(q))
758 		return false;
759 	if (!q->conv_zones_bitmap)
760 		return true;
761 	return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
762 }
763 
764 static inline void blk_queue_max_open_zones(struct request_queue *q,
765 		unsigned int max_open_zones)
766 {
767 	q->max_open_zones = max_open_zones;
768 }
769 
770 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
771 {
772 	return q->max_open_zones;
773 }
774 
775 static inline void blk_queue_max_active_zones(struct request_queue *q,
776 		unsigned int max_active_zones)
777 {
778 	q->max_active_zones = max_active_zones;
779 }
780 
781 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
782 {
783 	return q->max_active_zones;
784 }
785 #else /* CONFIG_BLK_DEV_ZONED */
786 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
787 {
788 	return 0;
789 }
790 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
791 					 sector_t sector)
792 {
793 	return false;
794 }
795 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
796 					     sector_t sector)
797 {
798 	return 0;
799 }
800 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
801 {
802 	return 0;
803 }
804 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
805 {
806 	return 0;
807 }
808 #endif /* CONFIG_BLK_DEV_ZONED */
809 
810 static inline bool rq_is_sync(struct request *rq)
811 {
812 	return op_is_sync(rq->cmd_flags);
813 }
814 
815 static inline bool rq_mergeable(struct request *rq)
816 {
817 	if (blk_rq_is_passthrough(rq))
818 		return false;
819 
820 	if (req_op(rq) == REQ_OP_FLUSH)
821 		return false;
822 
823 	if (req_op(rq) == REQ_OP_WRITE_ZEROES)
824 		return false;
825 
826 	if (req_op(rq) == REQ_OP_ZONE_APPEND)
827 		return false;
828 
829 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
830 		return false;
831 	if (rq->rq_flags & RQF_NOMERGE_FLAGS)
832 		return false;
833 
834 	return true;
835 }
836 
837 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
838 {
839 	if (bio_page(a) == bio_page(b) &&
840 	    bio_offset(a) == bio_offset(b))
841 		return true;
842 
843 	return false;
844 }
845 
846 static inline unsigned int blk_queue_depth(struct request_queue *q)
847 {
848 	if (q->queue_depth)
849 		return q->queue_depth;
850 
851 	return q->nr_requests;
852 }
853 
854 /*
855  * default timeout for SG_IO if none specified
856  */
857 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
858 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
859 
860 struct rq_map_data {
861 	struct page **pages;
862 	int page_order;
863 	int nr_entries;
864 	unsigned long offset;
865 	int null_mapped;
866 	int from_user;
867 };
868 
869 struct req_iterator {
870 	struct bvec_iter iter;
871 	struct bio *bio;
872 };
873 
874 /* This should not be used directly - use rq_for_each_segment */
875 #define for_each_bio(_bio)		\
876 	for (; _bio; _bio = _bio->bi_next)
877 #define __rq_for_each_bio(_bio, rq)	\
878 	if ((rq->bio))			\
879 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
880 
881 #define rq_for_each_segment(bvl, _rq, _iter)			\
882 	__rq_for_each_bio(_iter.bio, _rq)			\
883 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
884 
885 #define rq_for_each_bvec(bvl, _rq, _iter)			\
886 	__rq_for_each_bio(_iter.bio, _rq)			\
887 		bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
888 
889 #define rq_iter_last(bvec, _iter)				\
890 		(_iter.bio->bi_next == NULL &&			\
891 		 bio_iter_last(bvec, _iter.iter))
892 
893 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
894 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
895 #endif
896 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
897 extern void rq_flush_dcache_pages(struct request *rq);
898 #else
899 static inline void rq_flush_dcache_pages(struct request *rq)
900 {
901 }
902 #endif
903 
904 extern int blk_register_queue(struct gendisk *disk);
905 extern void blk_unregister_queue(struct gendisk *disk);
906 blk_qc_t submit_bio_noacct(struct bio *bio);
907 extern void blk_rq_init(struct request_queue *q, struct request *rq);
908 extern void blk_put_request(struct request *);
909 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
910 				       blk_mq_req_flags_t flags);
911 extern int blk_lld_busy(struct request_queue *q);
912 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
913 			     struct bio_set *bs, gfp_t gfp_mask,
914 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
915 			     void *data);
916 extern void blk_rq_unprep_clone(struct request *rq);
917 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
918 				     struct request *rq);
919 int blk_rq_append_bio(struct request *rq, struct bio *bio);
920 extern void blk_queue_split(struct bio **);
921 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
922 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
923 			      unsigned int, void __user *);
924 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
925 			  unsigned int, void __user *);
926 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
927 			 struct scsi_ioctl_command __user *);
928 extern int get_sg_io_hdr(struct sg_io_hdr *hdr, const void __user *argp);
929 extern int put_sg_io_hdr(const struct sg_io_hdr *hdr, void __user *argp);
930 
931 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
932 extern void blk_queue_exit(struct request_queue *q);
933 extern void blk_sync_queue(struct request_queue *q);
934 extern int blk_rq_map_user(struct request_queue *, struct request *,
935 			   struct rq_map_data *, void __user *, unsigned long,
936 			   gfp_t);
937 extern int blk_rq_unmap_user(struct bio *);
938 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
939 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
940 			       struct rq_map_data *, const struct iov_iter *,
941 			       gfp_t);
942 extern void blk_execute_rq(struct gendisk *, struct request *, int);
943 extern void blk_execute_rq_nowait(struct gendisk *,
944 				  struct request *, int, rq_end_io_fn *);
945 
946 /* Helper to convert REQ_OP_XXX to its string format XXX */
947 extern const char *blk_op_str(unsigned int op);
948 
949 int blk_status_to_errno(blk_status_t status);
950 blk_status_t errno_to_blk_status(int errno);
951 
952 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
953 
954 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
955 {
956 	return bdev->bd_disk->queue;	/* this is never NULL */
957 }
958 
959 /*
960  * The basic unit of block I/O is a sector. It is used in a number of contexts
961  * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
962  * bytes. Variables of type sector_t represent an offset or size that is a
963  * multiple of 512 bytes. Hence these two constants.
964  */
965 #ifndef SECTOR_SHIFT
966 #define SECTOR_SHIFT 9
967 #endif
968 #ifndef SECTOR_SIZE
969 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
970 #endif
971 
972 /*
973  * blk_rq_pos()			: the current sector
974  * blk_rq_bytes()		: bytes left in the entire request
975  * blk_rq_cur_bytes()		: bytes left in the current segment
976  * blk_rq_err_bytes()		: bytes left till the next error boundary
977  * blk_rq_sectors()		: sectors left in the entire request
978  * blk_rq_cur_sectors()		: sectors left in the current segment
979  * blk_rq_stats_sectors()	: sectors of the entire request used for stats
980  */
981 static inline sector_t blk_rq_pos(const struct request *rq)
982 {
983 	return rq->__sector;
984 }
985 
986 static inline unsigned int blk_rq_bytes(const struct request *rq)
987 {
988 	return rq->__data_len;
989 }
990 
991 static inline int blk_rq_cur_bytes(const struct request *rq)
992 {
993 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
994 }
995 
996 extern unsigned int blk_rq_err_bytes(const struct request *rq);
997 
998 static inline unsigned int blk_rq_sectors(const struct request *rq)
999 {
1000 	return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1001 }
1002 
1003 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1004 {
1005 	return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1006 }
1007 
1008 static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
1009 {
1010 	return rq->stats_sectors;
1011 }
1012 
1013 #ifdef CONFIG_BLK_DEV_ZONED
1014 
1015 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
1016 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
1017 
1018 static inline unsigned int blk_rq_zone_no(struct request *rq)
1019 {
1020 	return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1021 }
1022 
1023 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1024 {
1025 	return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1026 }
1027 #endif /* CONFIG_BLK_DEV_ZONED */
1028 
1029 /*
1030  * Some commands like WRITE SAME have a payload or data transfer size which
1031  * is different from the size of the request.  Any driver that supports such
1032  * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1033  * calculate the data transfer size.
1034  */
1035 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1036 {
1037 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1038 		return rq->special_vec.bv_len;
1039 	return blk_rq_bytes(rq);
1040 }
1041 
1042 /*
1043  * Return the first full biovec in the request.  The caller needs to check that
1044  * there are any bvecs before calling this helper.
1045  */
1046 static inline struct bio_vec req_bvec(struct request *rq)
1047 {
1048 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1049 		return rq->special_vec;
1050 	return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
1051 }
1052 
1053 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1054 						     int op)
1055 {
1056 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1057 		return min(q->limits.max_discard_sectors,
1058 			   UINT_MAX >> SECTOR_SHIFT);
1059 
1060 	if (unlikely(op == REQ_OP_WRITE_SAME))
1061 		return q->limits.max_write_same_sectors;
1062 
1063 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
1064 		return q->limits.max_write_zeroes_sectors;
1065 
1066 	return q->limits.max_sectors;
1067 }
1068 
1069 /*
1070  * Return maximum size of a request at given offset. Only valid for
1071  * file system requests.
1072  */
1073 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1074 					       sector_t offset,
1075 					       unsigned int chunk_sectors)
1076 {
1077 	if (!chunk_sectors) {
1078 		if (q->limits.chunk_sectors)
1079 			chunk_sectors = q->limits.chunk_sectors;
1080 		else
1081 			return q->limits.max_sectors;
1082 	}
1083 
1084 	if (likely(is_power_of_2(chunk_sectors)))
1085 		chunk_sectors -= offset & (chunk_sectors - 1);
1086 	else
1087 		chunk_sectors -= sector_div(offset, chunk_sectors);
1088 
1089 	return min(q->limits.max_sectors, chunk_sectors);
1090 }
1091 
1092 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1093 						  sector_t offset)
1094 {
1095 	struct request_queue *q = rq->q;
1096 
1097 	if (blk_rq_is_passthrough(rq))
1098 		return q->limits.max_hw_sectors;
1099 
1100 	if (!q->limits.chunk_sectors ||
1101 	    req_op(rq) == REQ_OP_DISCARD ||
1102 	    req_op(rq) == REQ_OP_SECURE_ERASE)
1103 		return blk_queue_get_max_sectors(q, req_op(rq));
1104 
1105 	return min(blk_max_size_offset(q, offset, 0),
1106 			blk_queue_get_max_sectors(q, req_op(rq)));
1107 }
1108 
1109 static inline unsigned int blk_rq_count_bios(struct request *rq)
1110 {
1111 	unsigned int nr_bios = 0;
1112 	struct bio *bio;
1113 
1114 	__rq_for_each_bio(bio, rq)
1115 		nr_bios++;
1116 
1117 	return nr_bios;
1118 }
1119 
1120 void blk_steal_bios(struct bio_list *list, struct request *rq);
1121 
1122 /*
1123  * Request completion related functions.
1124  *
1125  * blk_update_request() completes given number of bytes and updates
1126  * the request without completing it.
1127  */
1128 extern bool blk_update_request(struct request *rq, blk_status_t error,
1129 			       unsigned int nr_bytes);
1130 
1131 extern void blk_abort_request(struct request *);
1132 
1133 /*
1134  * Access functions for manipulating queue properties
1135  */
1136 extern void blk_cleanup_queue(struct request_queue *);
1137 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
1138 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1139 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1140 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1141 extern void blk_queue_max_discard_segments(struct request_queue *,
1142 		unsigned short);
1143 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1144 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1145 		unsigned int max_discard_sectors);
1146 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1147 		unsigned int max_write_same_sectors);
1148 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1149 		unsigned int max_write_same_sectors);
1150 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1151 extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
1152 		unsigned int max_zone_append_sectors);
1153 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1154 void blk_queue_zone_write_granularity(struct request_queue *q,
1155 				      unsigned int size);
1156 extern void blk_queue_alignment_offset(struct request_queue *q,
1157 				       unsigned int alignment);
1158 void blk_queue_update_readahead(struct request_queue *q);
1159 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1160 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1161 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1162 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1163 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1164 extern void blk_set_default_limits(struct queue_limits *lim);
1165 extern void blk_set_stacking_limits(struct queue_limits *lim);
1166 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1167 			    sector_t offset);
1168 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1169 			      sector_t offset);
1170 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1171 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1172 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1173 extern void blk_queue_dma_alignment(struct request_queue *, int);
1174 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1175 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1176 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1177 extern void blk_queue_required_elevator_features(struct request_queue *q,
1178 						 unsigned int features);
1179 extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
1180 					      struct device *dev);
1181 
1182 /*
1183  * Number of physical segments as sent to the device.
1184  *
1185  * Normally this is the number of discontiguous data segments sent by the
1186  * submitter.  But for data-less command like discard we might have no
1187  * actual data segments submitted, but the driver might have to add it's
1188  * own special payload.  In that case we still return 1 here so that this
1189  * special payload will be mapped.
1190  */
1191 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1192 {
1193 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1194 		return 1;
1195 	return rq->nr_phys_segments;
1196 }
1197 
1198 /*
1199  * Number of discard segments (or ranges) the driver needs to fill in.
1200  * Each discard bio merged into a request is counted as one segment.
1201  */
1202 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1203 {
1204 	return max_t(unsigned short, rq->nr_phys_segments, 1);
1205 }
1206 
1207 int __blk_rq_map_sg(struct request_queue *q, struct request *rq,
1208 		struct scatterlist *sglist, struct scatterlist **last_sg);
1209 static inline int blk_rq_map_sg(struct request_queue *q, struct request *rq,
1210 		struct scatterlist *sglist)
1211 {
1212 	struct scatterlist *last_sg = NULL;
1213 
1214 	return __blk_rq_map_sg(q, rq, sglist, &last_sg);
1215 }
1216 extern void blk_dump_rq_flags(struct request *, char *);
1217 
1218 bool __must_check blk_get_queue(struct request_queue *);
1219 struct request_queue *blk_alloc_queue(int node_id);
1220 extern void blk_put_queue(struct request_queue *);
1221 extern void blk_set_queue_dying(struct request_queue *);
1222 
1223 #ifdef CONFIG_BLOCK
1224 /*
1225  * blk_plug permits building a queue of related requests by holding the I/O
1226  * fragments for a short period. This allows merging of sequential requests
1227  * into single larger request. As the requests are moved from a per-task list to
1228  * the device's request_queue in a batch, this results in improved scalability
1229  * as the lock contention for request_queue lock is reduced.
1230  *
1231  * It is ok not to disable preemption when adding the request to the plug list
1232  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1233  * the plug list when the task sleeps by itself. For details, please see
1234  * schedule() where blk_schedule_flush_plug() is called.
1235  */
1236 struct blk_plug {
1237 	struct list_head mq_list; /* blk-mq requests */
1238 	struct list_head cb_list; /* md requires an unplug callback */
1239 	unsigned short rq_count;
1240 	bool multiple_queues;
1241 	bool nowait;
1242 };
1243 #define BLK_MAX_REQUEST_COUNT 16
1244 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1245 
1246 struct blk_plug_cb;
1247 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1248 struct blk_plug_cb {
1249 	struct list_head list;
1250 	blk_plug_cb_fn callback;
1251 	void *data;
1252 };
1253 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1254 					     void *data, int size);
1255 extern void blk_start_plug(struct blk_plug *);
1256 extern void blk_finish_plug(struct blk_plug *);
1257 extern void blk_flush_plug_list(struct blk_plug *, bool);
1258 
1259 static inline void blk_flush_plug(struct task_struct *tsk)
1260 {
1261 	struct blk_plug *plug = tsk->plug;
1262 
1263 	if (plug)
1264 		blk_flush_plug_list(plug, false);
1265 }
1266 
1267 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1268 {
1269 	struct blk_plug *plug = tsk->plug;
1270 
1271 	if (plug)
1272 		blk_flush_plug_list(plug, true);
1273 }
1274 
1275 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1276 {
1277 	struct blk_plug *plug = tsk->plug;
1278 
1279 	return plug &&
1280 		 (!list_empty(&plug->mq_list) ||
1281 		 !list_empty(&plug->cb_list));
1282 }
1283 
1284 int blkdev_issue_flush(struct block_device *bdev);
1285 long nr_blockdev_pages(void);
1286 #else /* CONFIG_BLOCK */
1287 struct blk_plug {
1288 };
1289 
1290 static inline void blk_start_plug(struct blk_plug *plug)
1291 {
1292 }
1293 
1294 static inline void blk_finish_plug(struct blk_plug *plug)
1295 {
1296 }
1297 
1298 static inline void blk_flush_plug(struct task_struct *task)
1299 {
1300 }
1301 
1302 static inline void blk_schedule_flush_plug(struct task_struct *task)
1303 {
1304 }
1305 
1306 
1307 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1308 {
1309 	return false;
1310 }
1311 
1312 static inline int blkdev_issue_flush(struct block_device *bdev)
1313 {
1314 	return 0;
1315 }
1316 
1317 static inline long nr_blockdev_pages(void)
1318 {
1319 	return 0;
1320 }
1321 #endif /* CONFIG_BLOCK */
1322 
1323 extern void blk_io_schedule(void);
1324 
1325 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1326 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1327 
1328 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
1329 
1330 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1331 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1332 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1333 		sector_t nr_sects, gfp_t gfp_mask, int flags,
1334 		struct bio **biop);
1335 
1336 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1337 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1338 
1339 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1340 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1341 		unsigned flags);
1342 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1343 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1344 
1345 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1346 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1347 {
1348 	return blkdev_issue_discard(sb->s_bdev,
1349 				    block << (sb->s_blocksize_bits -
1350 					      SECTOR_SHIFT),
1351 				    nr_blocks << (sb->s_blocksize_bits -
1352 						  SECTOR_SHIFT),
1353 				    gfp_mask, flags);
1354 }
1355 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1356 		sector_t nr_blocks, gfp_t gfp_mask)
1357 {
1358 	return blkdev_issue_zeroout(sb->s_bdev,
1359 				    block << (sb->s_blocksize_bits -
1360 					      SECTOR_SHIFT),
1361 				    nr_blocks << (sb->s_blocksize_bits -
1362 						  SECTOR_SHIFT),
1363 				    gfp_mask, 0);
1364 }
1365 
1366 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1367 
1368 static inline bool bdev_is_partition(struct block_device *bdev)
1369 {
1370 	return bdev->bd_partno;
1371 }
1372 
1373 enum blk_default_limits {
1374 	BLK_MAX_SEGMENTS	= 128,
1375 	BLK_SAFE_MAX_SECTORS	= 255,
1376 	BLK_DEF_MAX_SECTORS	= 2560,
1377 	BLK_MAX_SEGMENT_SIZE	= 65536,
1378 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1379 };
1380 
1381 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1382 {
1383 	return q->limits.seg_boundary_mask;
1384 }
1385 
1386 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1387 {
1388 	return q->limits.virt_boundary_mask;
1389 }
1390 
1391 static inline unsigned int queue_max_sectors(const struct request_queue *q)
1392 {
1393 	return q->limits.max_sectors;
1394 }
1395 
1396 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1397 {
1398 	return q->limits.max_hw_sectors;
1399 }
1400 
1401 static inline unsigned short queue_max_segments(const struct request_queue *q)
1402 {
1403 	return q->limits.max_segments;
1404 }
1405 
1406 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1407 {
1408 	return q->limits.max_discard_segments;
1409 }
1410 
1411 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1412 {
1413 	return q->limits.max_segment_size;
1414 }
1415 
1416 static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
1417 {
1418 
1419 	const struct queue_limits *l = &q->limits;
1420 
1421 	return min(l->max_zone_append_sectors, l->max_sectors);
1422 }
1423 
1424 static inline unsigned queue_logical_block_size(const struct request_queue *q)
1425 {
1426 	int retval = 512;
1427 
1428 	if (q && q->limits.logical_block_size)
1429 		retval = q->limits.logical_block_size;
1430 
1431 	return retval;
1432 }
1433 
1434 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1435 {
1436 	return queue_logical_block_size(bdev_get_queue(bdev));
1437 }
1438 
1439 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1440 {
1441 	return q->limits.physical_block_size;
1442 }
1443 
1444 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1445 {
1446 	return queue_physical_block_size(bdev_get_queue(bdev));
1447 }
1448 
1449 static inline unsigned int queue_io_min(const struct request_queue *q)
1450 {
1451 	return q->limits.io_min;
1452 }
1453 
1454 static inline int bdev_io_min(struct block_device *bdev)
1455 {
1456 	return queue_io_min(bdev_get_queue(bdev));
1457 }
1458 
1459 static inline unsigned int queue_io_opt(const struct request_queue *q)
1460 {
1461 	return q->limits.io_opt;
1462 }
1463 
1464 static inline int bdev_io_opt(struct block_device *bdev)
1465 {
1466 	return queue_io_opt(bdev_get_queue(bdev));
1467 }
1468 
1469 static inline unsigned int
1470 queue_zone_write_granularity(const struct request_queue *q)
1471 {
1472 	return q->limits.zone_write_granularity;
1473 }
1474 
1475 static inline unsigned int
1476 bdev_zone_write_granularity(struct block_device *bdev)
1477 {
1478 	return queue_zone_write_granularity(bdev_get_queue(bdev));
1479 }
1480 
1481 static inline int queue_alignment_offset(const struct request_queue *q)
1482 {
1483 	if (q->limits.misaligned)
1484 		return -1;
1485 
1486 	return q->limits.alignment_offset;
1487 }
1488 
1489 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1490 {
1491 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1492 	unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1493 		<< SECTOR_SHIFT;
1494 
1495 	return (granularity + lim->alignment_offset - alignment) % granularity;
1496 }
1497 
1498 static inline int bdev_alignment_offset(struct block_device *bdev)
1499 {
1500 	struct request_queue *q = bdev_get_queue(bdev);
1501 
1502 	if (q->limits.misaligned)
1503 		return -1;
1504 	if (bdev_is_partition(bdev))
1505 		return queue_limit_alignment_offset(&q->limits,
1506 				bdev->bd_start_sect);
1507 	return q->limits.alignment_offset;
1508 }
1509 
1510 static inline int queue_discard_alignment(const struct request_queue *q)
1511 {
1512 	if (q->limits.discard_misaligned)
1513 		return -1;
1514 
1515 	return q->limits.discard_alignment;
1516 }
1517 
1518 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1519 {
1520 	unsigned int alignment, granularity, offset;
1521 
1522 	if (!lim->max_discard_sectors)
1523 		return 0;
1524 
1525 	/* Why are these in bytes, not sectors? */
1526 	alignment = lim->discard_alignment >> SECTOR_SHIFT;
1527 	granularity = lim->discard_granularity >> SECTOR_SHIFT;
1528 	if (!granularity)
1529 		return 0;
1530 
1531 	/* Offset of the partition start in 'granularity' sectors */
1532 	offset = sector_div(sector, granularity);
1533 
1534 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1535 	offset = (granularity + alignment - offset) % granularity;
1536 
1537 	/* Turn it back into bytes, gaah */
1538 	return offset << SECTOR_SHIFT;
1539 }
1540 
1541 static inline int bdev_discard_alignment(struct block_device *bdev)
1542 {
1543 	struct request_queue *q = bdev_get_queue(bdev);
1544 
1545 	if (bdev_is_partition(bdev))
1546 		return queue_limit_discard_alignment(&q->limits,
1547 				bdev->bd_start_sect);
1548 	return q->limits.discard_alignment;
1549 }
1550 
1551 static inline unsigned int bdev_write_same(struct block_device *bdev)
1552 {
1553 	struct request_queue *q = bdev_get_queue(bdev);
1554 
1555 	if (q)
1556 		return q->limits.max_write_same_sectors;
1557 
1558 	return 0;
1559 }
1560 
1561 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1562 {
1563 	struct request_queue *q = bdev_get_queue(bdev);
1564 
1565 	if (q)
1566 		return q->limits.max_write_zeroes_sectors;
1567 
1568 	return 0;
1569 }
1570 
1571 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1572 {
1573 	struct request_queue *q = bdev_get_queue(bdev);
1574 
1575 	if (q)
1576 		return blk_queue_zoned_model(q);
1577 
1578 	return BLK_ZONED_NONE;
1579 }
1580 
1581 static inline bool bdev_is_zoned(struct block_device *bdev)
1582 {
1583 	struct request_queue *q = bdev_get_queue(bdev);
1584 
1585 	if (q)
1586 		return blk_queue_is_zoned(q);
1587 
1588 	return false;
1589 }
1590 
1591 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1592 {
1593 	struct request_queue *q = bdev_get_queue(bdev);
1594 
1595 	if (q)
1596 		return blk_queue_zone_sectors(q);
1597 	return 0;
1598 }
1599 
1600 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1601 {
1602 	struct request_queue *q = bdev_get_queue(bdev);
1603 
1604 	if (q)
1605 		return queue_max_open_zones(q);
1606 	return 0;
1607 }
1608 
1609 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1610 {
1611 	struct request_queue *q = bdev_get_queue(bdev);
1612 
1613 	if (q)
1614 		return queue_max_active_zones(q);
1615 	return 0;
1616 }
1617 
1618 static inline int queue_dma_alignment(const struct request_queue *q)
1619 {
1620 	return q ? q->dma_alignment : 511;
1621 }
1622 
1623 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1624 				 unsigned int len)
1625 {
1626 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1627 	return !(addr & alignment) && !(len & alignment);
1628 }
1629 
1630 /* assumes size > 256 */
1631 static inline unsigned int blksize_bits(unsigned int size)
1632 {
1633 	unsigned int bits = 8;
1634 	do {
1635 		bits++;
1636 		size >>= 1;
1637 	} while (size > 256);
1638 	return bits;
1639 }
1640 
1641 static inline unsigned int block_size(struct block_device *bdev)
1642 {
1643 	return 1 << bdev->bd_inode->i_blkbits;
1644 }
1645 
1646 int kblockd_schedule_work(struct work_struct *work);
1647 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1648 
1649 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1650 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1651 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1652 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1653 
1654 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1655 
1656 enum blk_integrity_flags {
1657 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1658 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1659 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1660 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1661 };
1662 
1663 struct blk_integrity_iter {
1664 	void			*prot_buf;
1665 	void			*data_buf;
1666 	sector_t		seed;
1667 	unsigned int		data_size;
1668 	unsigned short		interval;
1669 	const char		*disk_name;
1670 };
1671 
1672 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1673 typedef void (integrity_prepare_fn) (struct request *);
1674 typedef void (integrity_complete_fn) (struct request *, unsigned int);
1675 
1676 struct blk_integrity_profile {
1677 	integrity_processing_fn		*generate_fn;
1678 	integrity_processing_fn		*verify_fn;
1679 	integrity_prepare_fn		*prepare_fn;
1680 	integrity_complete_fn		*complete_fn;
1681 	const char			*name;
1682 };
1683 
1684 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1685 extern void blk_integrity_unregister(struct gendisk *);
1686 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1687 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1688 				   struct scatterlist *);
1689 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1690 
1691 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1692 {
1693 	struct blk_integrity *bi = &disk->queue->integrity;
1694 
1695 	if (!bi->profile)
1696 		return NULL;
1697 
1698 	return bi;
1699 }
1700 
1701 static inline
1702 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1703 {
1704 	return blk_get_integrity(bdev->bd_disk);
1705 }
1706 
1707 static inline bool
1708 blk_integrity_queue_supports_integrity(struct request_queue *q)
1709 {
1710 	return q->integrity.profile;
1711 }
1712 
1713 static inline bool blk_integrity_rq(struct request *rq)
1714 {
1715 	return rq->cmd_flags & REQ_INTEGRITY;
1716 }
1717 
1718 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1719 						    unsigned int segs)
1720 {
1721 	q->limits.max_integrity_segments = segs;
1722 }
1723 
1724 static inline unsigned short
1725 queue_max_integrity_segments(const struct request_queue *q)
1726 {
1727 	return q->limits.max_integrity_segments;
1728 }
1729 
1730 /**
1731  * bio_integrity_intervals - Return number of integrity intervals for a bio
1732  * @bi:		blk_integrity profile for device
1733  * @sectors:	Size of the bio in 512-byte sectors
1734  *
1735  * Description: The block layer calculates everything in 512 byte
1736  * sectors but integrity metadata is done in terms of the data integrity
1737  * interval size of the storage device.  Convert the block layer sectors
1738  * to the appropriate number of integrity intervals.
1739  */
1740 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1741 						   unsigned int sectors)
1742 {
1743 	return sectors >> (bi->interval_exp - 9);
1744 }
1745 
1746 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1747 					       unsigned int sectors)
1748 {
1749 	return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1750 }
1751 
1752 /*
1753  * Return the first bvec that contains integrity data.  Only drivers that are
1754  * limited to a single integrity segment should use this helper.
1755  */
1756 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1757 {
1758 	if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1759 		return NULL;
1760 	return rq->bio->bi_integrity->bip_vec;
1761 }
1762 
1763 #else /* CONFIG_BLK_DEV_INTEGRITY */
1764 
1765 struct bio;
1766 struct block_device;
1767 struct gendisk;
1768 struct blk_integrity;
1769 
1770 static inline int blk_integrity_rq(struct request *rq)
1771 {
1772 	return 0;
1773 }
1774 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1775 					    struct bio *b)
1776 {
1777 	return 0;
1778 }
1779 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1780 					  struct bio *b,
1781 					  struct scatterlist *s)
1782 {
1783 	return 0;
1784 }
1785 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1786 {
1787 	return NULL;
1788 }
1789 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1790 {
1791 	return NULL;
1792 }
1793 static inline bool
1794 blk_integrity_queue_supports_integrity(struct request_queue *q)
1795 {
1796 	return false;
1797 }
1798 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1799 {
1800 	return 0;
1801 }
1802 static inline void blk_integrity_register(struct gendisk *d,
1803 					 struct blk_integrity *b)
1804 {
1805 }
1806 static inline void blk_integrity_unregister(struct gendisk *d)
1807 {
1808 }
1809 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1810 						    unsigned int segs)
1811 {
1812 }
1813 static inline unsigned short queue_max_integrity_segments(const struct request_queue *q)
1814 {
1815 	return 0;
1816 }
1817 
1818 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1819 						   unsigned int sectors)
1820 {
1821 	return 0;
1822 }
1823 
1824 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1825 					       unsigned int sectors)
1826 {
1827 	return 0;
1828 }
1829 
1830 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1831 {
1832 	return NULL;
1833 }
1834 
1835 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1836 
1837 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1838 
1839 bool blk_ksm_register(struct blk_keyslot_manager *ksm, struct request_queue *q);
1840 
1841 void blk_ksm_unregister(struct request_queue *q);
1842 
1843 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1844 
1845 static inline bool blk_ksm_register(struct blk_keyslot_manager *ksm,
1846 				    struct request_queue *q)
1847 {
1848 	return true;
1849 }
1850 
1851 static inline void blk_ksm_unregister(struct request_queue *q) { }
1852 
1853 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1854 
1855 
1856 struct block_device_operations {
1857 	blk_qc_t (*submit_bio) (struct bio *bio);
1858 	int (*open) (struct block_device *, fmode_t);
1859 	void (*release) (struct gendisk *, fmode_t);
1860 	int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1861 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1862 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1863 	unsigned int (*check_events) (struct gendisk *disk,
1864 				      unsigned int clearing);
1865 	void (*unlock_native_capacity) (struct gendisk *);
1866 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1867 	int (*set_read_only)(struct block_device *bdev, bool ro);
1868 	/* this callback is with swap_lock and sometimes page table lock held */
1869 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1870 	int (*report_zones)(struct gendisk *, sector_t sector,
1871 			unsigned int nr_zones, report_zones_cb cb, void *data);
1872 	char *(*devnode)(struct gendisk *disk, umode_t *mode);
1873 	struct module *owner;
1874 	const struct pr_ops *pr_ops;
1875 };
1876 
1877 #ifdef CONFIG_COMPAT
1878 extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1879 				      unsigned int, unsigned long);
1880 #else
1881 #define blkdev_compat_ptr_ioctl NULL
1882 #endif
1883 
1884 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1885 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1886 						struct writeback_control *);
1887 
1888 #ifdef CONFIG_BLK_DEV_ZONED
1889 bool blk_req_needs_zone_write_lock(struct request *rq);
1890 bool blk_req_zone_write_trylock(struct request *rq);
1891 void __blk_req_zone_write_lock(struct request *rq);
1892 void __blk_req_zone_write_unlock(struct request *rq);
1893 
1894 static inline void blk_req_zone_write_lock(struct request *rq)
1895 {
1896 	if (blk_req_needs_zone_write_lock(rq))
1897 		__blk_req_zone_write_lock(rq);
1898 }
1899 
1900 static inline void blk_req_zone_write_unlock(struct request *rq)
1901 {
1902 	if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1903 		__blk_req_zone_write_unlock(rq);
1904 }
1905 
1906 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1907 {
1908 	return rq->q->seq_zones_wlock &&
1909 		test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1910 }
1911 
1912 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1913 {
1914 	if (!blk_req_needs_zone_write_lock(rq))
1915 		return true;
1916 	return !blk_req_zone_is_write_locked(rq);
1917 }
1918 #else
1919 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1920 {
1921 	return false;
1922 }
1923 
1924 static inline void blk_req_zone_write_lock(struct request *rq)
1925 {
1926 }
1927 
1928 static inline void blk_req_zone_write_unlock(struct request *rq)
1929 {
1930 }
1931 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1932 {
1933 	return false;
1934 }
1935 
1936 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1937 {
1938 	return true;
1939 }
1940 #endif /* CONFIG_BLK_DEV_ZONED */
1941 
1942 static inline void blk_wake_io_task(struct task_struct *waiter)
1943 {
1944 	/*
1945 	 * If we're polling, the task itself is doing the completions. For
1946 	 * that case, we don't need to signal a wakeup, it's enough to just
1947 	 * mark us as RUNNING.
1948 	 */
1949 	if (waiter == current)
1950 		__set_current_state(TASK_RUNNING);
1951 	else
1952 		wake_up_process(waiter);
1953 }
1954 
1955 unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1956 		unsigned int op);
1957 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1958 		unsigned long start_time);
1959 
1960 unsigned long bio_start_io_acct(struct bio *bio);
1961 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1962 		struct block_device *orig_bdev);
1963 
1964 /**
1965  * bio_end_io_acct - end I/O accounting for bio based drivers
1966  * @bio:	bio to end account for
1967  * @start:	start time returned by bio_start_io_acct()
1968  */
1969 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1970 {
1971 	return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1972 }
1973 
1974 int bdev_read_only(struct block_device *bdev);
1975 int set_blocksize(struct block_device *bdev, int size);
1976 
1977 const char *bdevname(struct block_device *bdev, char *buffer);
1978 int lookup_bdev(const char *pathname, dev_t *dev);
1979 
1980 void blkdev_show(struct seq_file *seqf, off_t offset);
1981 
1982 #define BDEVNAME_SIZE	32	/* Largest string for a blockdev identifier */
1983 #define BDEVT_SIZE	10	/* Largest string for MAJ:MIN for blkdev */
1984 #ifdef CONFIG_BLOCK
1985 #define BLKDEV_MAJOR_MAX	512
1986 #else
1987 #define BLKDEV_MAJOR_MAX	0
1988 #endif
1989 
1990 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1991 		void *holder);
1992 struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
1993 int bd_prepare_to_claim(struct block_device *bdev, void *holder);
1994 void bd_abort_claiming(struct block_device *bdev, void *holder);
1995 void blkdev_put(struct block_device *bdev, fmode_t mode);
1996 
1997 /* just for blk-cgroup, don't use elsewhere */
1998 struct block_device *blkdev_get_no_open(dev_t dev);
1999 void blkdev_put_no_open(struct block_device *bdev);
2000 
2001 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
2002 void bdev_add(struct block_device *bdev, dev_t dev);
2003 struct block_device *I_BDEV(struct inode *inode);
2004 struct block_device *bdgrab(struct block_device *bdev);
2005 void bdput(struct block_device *);
2006 int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
2007 		loff_t lend);
2008 
2009 #ifdef CONFIG_BLOCK
2010 void invalidate_bdev(struct block_device *bdev);
2011 int sync_blockdev(struct block_device *bdev);
2012 #else
2013 static inline void invalidate_bdev(struct block_device *bdev)
2014 {
2015 }
2016 static inline int sync_blockdev(struct block_device *bdev)
2017 {
2018 	return 0;
2019 }
2020 #endif
2021 int fsync_bdev(struct block_device *bdev);
2022 
2023 int freeze_bdev(struct block_device *bdev);
2024 int thaw_bdev(struct block_device *bdev);
2025 
2026 #endif /* _LINUX_BLKDEV_H */
2027