xref: /linux-6.15/include/linux/blkdev.h (revision 9be148e4)
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/pagemap.h>
15 #include <linux/backing-dev-defs.h>
16 #include <linux/wait.h>
17 #include <linux/mempool.h>
18 #include <linux/pfn.h>
19 #include <linux/bio.h>
20 #include <linux/stringify.h>
21 #include <linux/gfp.h>
22 #include <linux/bsg.h>
23 #include <linux/smp.h>
24 #include <linux/rcupdate.h>
25 #include <linux/percpu-refcount.h>
26 #include <linux/scatterlist.h>
27 #include <linux/blkzoned.h>
28 #include <linux/pm.h>
29 
30 struct module;
31 struct scsi_ioctl_command;
32 
33 struct request_queue;
34 struct elevator_queue;
35 struct blk_trace;
36 struct request;
37 struct sg_io_hdr;
38 struct bsg_job;
39 struct blkcg_gq;
40 struct blk_flush_queue;
41 struct pr_ops;
42 struct rq_qos;
43 struct blk_queue_stats;
44 struct blk_stat_callback;
45 struct blk_keyslot_manager;
46 
47 #define BLKDEV_MIN_RQ	4
48 #define BLKDEV_MAX_RQ	128	/* Default maximum */
49 
50 /* Must be consistent with blk_mq_poll_stats_bkt() */
51 #define BLK_MQ_POLL_STATS_BKTS 16
52 
53 /* Doing classic polling */
54 #define BLK_MQ_POLL_CLASSIC -1
55 
56 /*
57  * Maximum number of blkcg policies allowed to be registered concurrently.
58  * Defined here to simplify include dependency.
59  */
60 #define BLKCG_MAX_POLS		5
61 
62 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
63 
64 /*
65  * request flags */
66 typedef __u32 __bitwise req_flags_t;
67 
68 /* drive already may have started this one */
69 #define RQF_STARTED		((__force req_flags_t)(1 << 1))
70 /* may not be passed by ioscheduler */
71 #define RQF_SOFTBARRIER		((__force req_flags_t)(1 << 3))
72 /* request for flush sequence */
73 #define RQF_FLUSH_SEQ		((__force req_flags_t)(1 << 4))
74 /* merge of different types, fail separately */
75 #define RQF_MIXED_MERGE		((__force req_flags_t)(1 << 5))
76 /* track inflight for MQ */
77 #define RQF_MQ_INFLIGHT		((__force req_flags_t)(1 << 6))
78 /* don't call prep for this one */
79 #define RQF_DONTPREP		((__force req_flags_t)(1 << 7))
80 /* vaguely specified driver internal error.  Ignored by the block layer */
81 #define RQF_FAILED		((__force req_flags_t)(1 << 10))
82 /* don't warn about errors */
83 #define RQF_QUIET		((__force req_flags_t)(1 << 11))
84 /* elevator private data attached */
85 #define RQF_ELVPRIV		((__force req_flags_t)(1 << 12))
86 /* account into disk and partition IO statistics */
87 #define RQF_IO_STAT		((__force req_flags_t)(1 << 13))
88 /* runtime pm request */
89 #define RQF_PM			((__force req_flags_t)(1 << 15))
90 /* on IO scheduler merge hash */
91 #define RQF_HASHED		((__force req_flags_t)(1 << 16))
92 /* track IO completion time */
93 #define RQF_STATS		((__force req_flags_t)(1 << 17))
94 /* Look at ->special_vec for the actual data payload instead of the
95    bio chain. */
96 #define RQF_SPECIAL_PAYLOAD	((__force req_flags_t)(1 << 18))
97 /* The per-zone write lock is held for this request */
98 #define RQF_ZONE_WRITE_LOCKED	((__force req_flags_t)(1 << 19))
99 /* already slept for hybrid poll */
100 #define RQF_MQ_POLL_SLEPT	((__force req_flags_t)(1 << 20))
101 /* ->timeout has been called, don't expire again */
102 #define RQF_TIMED_OUT		((__force req_flags_t)(1 << 21))
103 
104 /* flags that prevent us from merging requests: */
105 #define RQF_NOMERGE_FLAGS \
106 	(RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
107 
108 /*
109  * Request state for blk-mq.
110  */
111 enum mq_rq_state {
112 	MQ_RQ_IDLE		= 0,
113 	MQ_RQ_IN_FLIGHT		= 1,
114 	MQ_RQ_COMPLETE		= 2,
115 };
116 
117 /*
118  * Try to put the fields that are referenced together in the same cacheline.
119  *
120  * If you modify this structure, make sure to update blk_rq_init() and
121  * especially blk_mq_rq_ctx_init() to take care of the added fields.
122  */
123 struct request {
124 	struct request_queue *q;
125 	struct blk_mq_ctx *mq_ctx;
126 	struct blk_mq_hw_ctx *mq_hctx;
127 
128 	unsigned int cmd_flags;		/* op and common flags */
129 	req_flags_t rq_flags;
130 
131 	int tag;
132 	int internal_tag;
133 
134 	/* the following two fields are internal, NEVER access directly */
135 	unsigned int __data_len;	/* total data len */
136 	sector_t __sector;		/* sector cursor */
137 
138 	struct bio *bio;
139 	struct bio *biotail;
140 
141 	struct list_head queuelist;
142 
143 	/*
144 	 * The hash is used inside the scheduler, and killed once the
145 	 * request reaches the dispatch list. The ipi_list is only used
146 	 * to queue the request for softirq completion, which is long
147 	 * after the request has been unhashed (and even removed from
148 	 * the dispatch list).
149 	 */
150 	union {
151 		struct hlist_node hash;	/* merge hash */
152 		struct llist_node ipi_list;
153 	};
154 
155 	/*
156 	 * The rb_node is only used inside the io scheduler, requests
157 	 * are pruned when moved to the dispatch queue. So let the
158 	 * completion_data share space with the rb_node.
159 	 */
160 	union {
161 		struct rb_node rb_node;	/* sort/lookup */
162 		struct bio_vec special_vec;
163 		void *completion_data;
164 		int error_count; /* for legacy drivers, don't use */
165 	};
166 
167 	/*
168 	 * Three pointers are available for the IO schedulers, if they need
169 	 * more they have to dynamically allocate it.  Flush requests are
170 	 * never put on the IO scheduler. So let the flush fields share
171 	 * space with the elevator data.
172 	 */
173 	union {
174 		struct {
175 			struct io_cq		*icq;
176 			void			*priv[2];
177 		} elv;
178 
179 		struct {
180 			unsigned int		seq;
181 			struct list_head	list;
182 			rq_end_io_fn		*saved_end_io;
183 		} flush;
184 	};
185 
186 	struct gendisk *rq_disk;
187 	struct block_device *part;
188 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
189 	/* Time that the first bio started allocating this request. */
190 	u64 alloc_time_ns;
191 #endif
192 	/* Time that this request was allocated for this IO. */
193 	u64 start_time_ns;
194 	/* Time that I/O was submitted to the device. */
195 	u64 io_start_time_ns;
196 
197 #ifdef CONFIG_BLK_WBT
198 	unsigned short wbt_flags;
199 #endif
200 	/*
201 	 * rq sectors used for blk stats. It has the same value
202 	 * with blk_rq_sectors(rq), except that it never be zeroed
203 	 * by completion.
204 	 */
205 	unsigned short stats_sectors;
206 
207 	/*
208 	 * Number of scatter-gather DMA addr+len pairs after
209 	 * physical address coalescing is performed.
210 	 */
211 	unsigned short nr_phys_segments;
212 
213 #if defined(CONFIG_BLK_DEV_INTEGRITY)
214 	unsigned short nr_integrity_segments;
215 #endif
216 
217 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
218 	struct bio_crypt_ctx *crypt_ctx;
219 	struct blk_ksm_keyslot *crypt_keyslot;
220 #endif
221 
222 	unsigned short write_hint;
223 	unsigned short ioprio;
224 
225 	enum mq_rq_state state;
226 	refcount_t ref;
227 
228 	unsigned int timeout;
229 	unsigned long deadline;
230 
231 	union {
232 		struct __call_single_data csd;
233 		u64 fifo_time;
234 	};
235 
236 	/*
237 	 * completion callback.
238 	 */
239 	rq_end_io_fn *end_io;
240 	void *end_io_data;
241 };
242 
243 static inline bool blk_op_is_scsi(unsigned int op)
244 {
245 	return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
246 }
247 
248 static inline bool blk_op_is_private(unsigned int op)
249 {
250 	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
251 }
252 
253 static inline bool blk_rq_is_scsi(struct request *rq)
254 {
255 	return blk_op_is_scsi(req_op(rq));
256 }
257 
258 static inline bool blk_rq_is_private(struct request *rq)
259 {
260 	return blk_op_is_private(req_op(rq));
261 }
262 
263 static inline bool blk_rq_is_passthrough(struct request *rq)
264 {
265 	return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
266 }
267 
268 static inline bool bio_is_passthrough(struct bio *bio)
269 {
270 	unsigned op = bio_op(bio);
271 
272 	return blk_op_is_scsi(op) || blk_op_is_private(op);
273 }
274 
275 static inline bool blk_op_is_passthrough(unsigned int op)
276 {
277 	return (blk_op_is_scsi(op & REQ_OP_MASK) ||
278 			blk_op_is_private(op & REQ_OP_MASK));
279 }
280 
281 static inline unsigned short req_get_ioprio(struct request *req)
282 {
283 	return req->ioprio;
284 }
285 
286 #include <linux/elevator.h>
287 
288 struct blk_queue_ctx;
289 
290 struct bio_vec;
291 
292 enum blk_eh_timer_return {
293 	BLK_EH_DONE,		/* drivers has completed the command */
294 	BLK_EH_RESET_TIMER,	/* reset timer and try again */
295 };
296 
297 enum blk_queue_state {
298 	Queue_down,
299 	Queue_up,
300 };
301 
302 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
303 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
304 
305 #define BLK_SCSI_MAX_CMDS	(256)
306 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
307 
308 /*
309  * Zoned block device models (zoned limit).
310  *
311  * Note: This needs to be ordered from the least to the most severe
312  * restrictions for the inheritance in blk_stack_limits() to work.
313  */
314 enum blk_zoned_model {
315 	BLK_ZONED_NONE = 0,	/* Regular block device */
316 	BLK_ZONED_HA,		/* Host-aware zoned block device */
317 	BLK_ZONED_HM,		/* Host-managed zoned block device */
318 };
319 
320 /*
321  * BLK_BOUNCE_NONE:	never bounce (default)
322  * BLK_BOUNCE_HIGH:	bounce all highmem pages
323  */
324 enum blk_bounce {
325 	BLK_BOUNCE_NONE,
326 	BLK_BOUNCE_HIGH,
327 };
328 
329 struct queue_limits {
330 	enum blk_bounce		bounce;
331 	unsigned long		seg_boundary_mask;
332 	unsigned long		virt_boundary_mask;
333 
334 	unsigned int		max_hw_sectors;
335 	unsigned int		max_dev_sectors;
336 	unsigned int		chunk_sectors;
337 	unsigned int		max_sectors;
338 	unsigned int		max_segment_size;
339 	unsigned int		physical_block_size;
340 	unsigned int		logical_block_size;
341 	unsigned int		alignment_offset;
342 	unsigned int		io_min;
343 	unsigned int		io_opt;
344 	unsigned int		max_discard_sectors;
345 	unsigned int		max_hw_discard_sectors;
346 	unsigned int		max_write_same_sectors;
347 	unsigned int		max_write_zeroes_sectors;
348 	unsigned int		max_zone_append_sectors;
349 	unsigned int		discard_granularity;
350 	unsigned int		discard_alignment;
351 	unsigned int		zone_write_granularity;
352 
353 	unsigned short		max_segments;
354 	unsigned short		max_integrity_segments;
355 	unsigned short		max_discard_segments;
356 
357 	unsigned char		misaligned;
358 	unsigned char		discard_misaligned;
359 	unsigned char		raid_partial_stripes_expensive;
360 	enum blk_zoned_model	zoned;
361 };
362 
363 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
364 			       void *data);
365 
366 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
367 
368 #ifdef CONFIG_BLK_DEV_ZONED
369 
370 #define BLK_ALL_ZONES  ((unsigned int)-1)
371 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
372 			unsigned int nr_zones, report_zones_cb cb, void *data);
373 unsigned int blkdev_nr_zones(struct gendisk *disk);
374 extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
375 			    sector_t sectors, sector_t nr_sectors,
376 			    gfp_t gfp_mask);
377 int blk_revalidate_disk_zones(struct gendisk *disk,
378 			      void (*update_driver_data)(struct gendisk *disk));
379 
380 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
381 				     unsigned int cmd, unsigned long arg);
382 extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
383 				  unsigned int cmd, unsigned long arg);
384 
385 #else /* CONFIG_BLK_DEV_ZONED */
386 
387 static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
388 {
389 	return 0;
390 }
391 
392 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
393 					    fmode_t mode, unsigned int cmd,
394 					    unsigned long arg)
395 {
396 	return -ENOTTY;
397 }
398 
399 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
400 					 fmode_t mode, unsigned int cmd,
401 					 unsigned long arg)
402 {
403 	return -ENOTTY;
404 }
405 
406 #endif /* CONFIG_BLK_DEV_ZONED */
407 
408 struct request_queue {
409 	struct request		*last_merge;
410 	struct elevator_queue	*elevator;
411 
412 	struct percpu_ref	q_usage_counter;
413 
414 	struct blk_queue_stats	*stats;
415 	struct rq_qos		*rq_qos;
416 
417 	const struct blk_mq_ops	*mq_ops;
418 
419 	/* sw queues */
420 	struct blk_mq_ctx __percpu	*queue_ctx;
421 
422 	unsigned int		queue_depth;
423 
424 	/* hw dispatch queues */
425 	struct blk_mq_hw_ctx	**queue_hw_ctx;
426 	unsigned int		nr_hw_queues;
427 
428 	struct backing_dev_info	*backing_dev_info;
429 
430 	/*
431 	 * The queue owner gets to use this for whatever they like.
432 	 * ll_rw_blk doesn't touch it.
433 	 */
434 	void			*queuedata;
435 
436 	/*
437 	 * various queue flags, see QUEUE_* below
438 	 */
439 	unsigned long		queue_flags;
440 	/*
441 	 * Number of contexts that have called blk_set_pm_only(). If this
442 	 * counter is above zero then only RQF_PM requests are processed.
443 	 */
444 	atomic_t		pm_only;
445 
446 	/*
447 	 * ida allocated id for this queue.  Used to index queues from
448 	 * ioctx.
449 	 */
450 	int			id;
451 
452 	spinlock_t		queue_lock;
453 
454 	/*
455 	 * queue kobject
456 	 */
457 	struct kobject kobj;
458 
459 	/*
460 	 * mq queue kobject
461 	 */
462 	struct kobject *mq_kobj;
463 
464 #ifdef  CONFIG_BLK_DEV_INTEGRITY
465 	struct blk_integrity integrity;
466 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
467 
468 #ifdef CONFIG_PM
469 	struct device		*dev;
470 	enum rpm_status		rpm_status;
471 #endif
472 
473 	/*
474 	 * queue settings
475 	 */
476 	unsigned long		nr_requests;	/* Max # of requests */
477 
478 	unsigned int		dma_pad_mask;
479 	unsigned int		dma_alignment;
480 
481 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
482 	/* Inline crypto capabilities */
483 	struct blk_keyslot_manager *ksm;
484 #endif
485 
486 	unsigned int		rq_timeout;
487 	int			poll_nsec;
488 
489 	struct blk_stat_callback	*poll_cb;
490 	struct blk_rq_stat	poll_stat[BLK_MQ_POLL_STATS_BKTS];
491 
492 	struct timer_list	timeout;
493 	struct work_struct	timeout_work;
494 
495 	atomic_t		nr_active_requests_shared_sbitmap;
496 
497 	struct list_head	icq_list;
498 #ifdef CONFIG_BLK_CGROUP
499 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
500 	struct blkcg_gq		*root_blkg;
501 	struct list_head	blkg_list;
502 #endif
503 
504 	struct queue_limits	limits;
505 
506 	unsigned int		required_elevator_features;
507 
508 #ifdef CONFIG_BLK_DEV_ZONED
509 	/*
510 	 * Zoned block device information for request dispatch control.
511 	 * nr_zones is the total number of zones of the device. This is always
512 	 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
513 	 * bits which indicates if a zone is conventional (bit set) or
514 	 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
515 	 * bits which indicates if a zone is write locked, that is, if a write
516 	 * request targeting the zone was dispatched. All three fields are
517 	 * initialized by the low level device driver (e.g. scsi/sd.c).
518 	 * Stacking drivers (device mappers) may or may not initialize
519 	 * these fields.
520 	 *
521 	 * Reads of this information must be protected with blk_queue_enter() /
522 	 * blk_queue_exit(). Modifying this information is only allowed while
523 	 * no requests are being processed. See also blk_mq_freeze_queue() and
524 	 * blk_mq_unfreeze_queue().
525 	 */
526 	unsigned int		nr_zones;
527 	unsigned long		*conv_zones_bitmap;
528 	unsigned long		*seq_zones_wlock;
529 	unsigned int		max_open_zones;
530 	unsigned int		max_active_zones;
531 #endif /* CONFIG_BLK_DEV_ZONED */
532 
533 	/*
534 	 * sg stuff
535 	 */
536 	unsigned int		sg_timeout;
537 	unsigned int		sg_reserved_size;
538 	int			node;
539 	struct mutex		debugfs_mutex;
540 #ifdef CONFIG_BLK_DEV_IO_TRACE
541 	struct blk_trace __rcu	*blk_trace;
542 #endif
543 	/*
544 	 * for flush operations
545 	 */
546 	struct blk_flush_queue	*fq;
547 
548 	struct list_head	requeue_list;
549 	spinlock_t		requeue_lock;
550 	struct delayed_work	requeue_work;
551 
552 	struct mutex		sysfs_lock;
553 	struct mutex		sysfs_dir_lock;
554 
555 	/*
556 	 * for reusing dead hctx instance in case of updating
557 	 * nr_hw_queues
558 	 */
559 	struct list_head	unused_hctx_list;
560 	spinlock_t		unused_hctx_lock;
561 
562 	int			mq_freeze_depth;
563 
564 #if defined(CONFIG_BLK_DEV_BSG)
565 	struct bsg_class_device bsg_dev;
566 #endif
567 
568 #ifdef CONFIG_BLK_DEV_THROTTLING
569 	/* Throttle data */
570 	struct throtl_data *td;
571 #endif
572 	struct rcu_head		rcu_head;
573 	wait_queue_head_t	mq_freeze_wq;
574 	/*
575 	 * Protect concurrent access to q_usage_counter by
576 	 * percpu_ref_kill() and percpu_ref_reinit().
577 	 */
578 	struct mutex		mq_freeze_lock;
579 
580 	struct blk_mq_tag_set	*tag_set;
581 	struct list_head	tag_set_list;
582 	struct bio_set		bio_split;
583 
584 	struct dentry		*debugfs_dir;
585 
586 #ifdef CONFIG_BLK_DEBUG_FS
587 	struct dentry		*sched_debugfs_dir;
588 	struct dentry		*rqos_debugfs_dir;
589 #endif
590 
591 	bool			mq_sysfs_init_done;
592 
593 	size_t			cmd_size;
594 
595 #define BLK_MAX_WRITE_HINTS	5
596 	u64			write_hints[BLK_MAX_WRITE_HINTS];
597 };
598 
599 /* Keep blk_queue_flag_name[] in sync with the definitions below */
600 #define QUEUE_FLAG_STOPPED	0	/* queue is stopped */
601 #define QUEUE_FLAG_DYING	1	/* queue being torn down */
602 #define QUEUE_FLAG_NOMERGES     3	/* disable merge attempts */
603 #define QUEUE_FLAG_SAME_COMP	4	/* complete on same CPU-group */
604 #define QUEUE_FLAG_FAIL_IO	5	/* fake timeout */
605 #define QUEUE_FLAG_NONROT	6	/* non-rotational device (SSD) */
606 #define QUEUE_FLAG_VIRT		QUEUE_FLAG_NONROT /* paravirt device */
607 #define QUEUE_FLAG_IO_STAT	7	/* do disk/partitions IO accounting */
608 #define QUEUE_FLAG_DISCARD	8	/* supports DISCARD */
609 #define QUEUE_FLAG_NOXMERGES	9	/* No extended merges */
610 #define QUEUE_FLAG_ADD_RANDOM	10	/* Contributes to random pool */
611 #define QUEUE_FLAG_SECERASE	11	/* supports secure erase */
612 #define QUEUE_FLAG_SAME_FORCE	12	/* force complete on same CPU */
613 #define QUEUE_FLAG_DEAD		13	/* queue tear-down finished */
614 #define QUEUE_FLAG_INIT_DONE	14	/* queue is initialized */
615 #define QUEUE_FLAG_STABLE_WRITES 15	/* don't modify blks until WB is done */
616 #define QUEUE_FLAG_POLL		16	/* IO polling enabled if set */
617 #define QUEUE_FLAG_WC		17	/* Write back caching */
618 #define QUEUE_FLAG_FUA		18	/* device supports FUA writes */
619 #define QUEUE_FLAG_DAX		19	/* device supports DAX */
620 #define QUEUE_FLAG_STATS	20	/* track IO start and completion times */
621 #define QUEUE_FLAG_POLL_STATS	21	/* collecting stats for hybrid polling */
622 #define QUEUE_FLAG_REGISTERED	22	/* queue has been registered to a disk */
623 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23	/* queue supports SCSI commands */
624 #define QUEUE_FLAG_QUIESCED	24	/* queue has been quiesced */
625 #define QUEUE_FLAG_PCI_P2PDMA	25	/* device supports PCI p2p requests */
626 #define QUEUE_FLAG_ZONE_RESETALL 26	/* supports Zone Reset All */
627 #define QUEUE_FLAG_RQ_ALLOC_TIME 27	/* record rq->alloc_time_ns */
628 #define QUEUE_FLAG_HCTX_ACTIVE	28	/* at least one blk-mq hctx is active */
629 #define QUEUE_FLAG_NOWAIT       29	/* device supports NOWAIT */
630 
631 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
632 				 (1 << QUEUE_FLAG_SAME_COMP) |		\
633 				 (1 << QUEUE_FLAG_NOWAIT))
634 
635 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
636 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
637 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
638 
639 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
640 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
641 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
642 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
643 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
644 #define blk_queue_noxmerges(q)	\
645 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
646 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
647 #define blk_queue_stable_writes(q) \
648 	test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
649 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
650 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
651 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
652 #define blk_queue_zone_resetall(q)	\
653 	test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
654 #define blk_queue_secure_erase(q) \
655 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
656 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
657 #define blk_queue_scsi_passthrough(q)	\
658 	test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
659 #define blk_queue_pci_p2pdma(q)	\
660 	test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
661 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
662 #define blk_queue_rq_alloc_time(q)	\
663 	test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
664 #else
665 #define blk_queue_rq_alloc_time(q)	false
666 #endif
667 
668 #define blk_noretry_request(rq) \
669 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
670 			     REQ_FAILFAST_DRIVER))
671 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
672 #define blk_queue_pm_only(q)	atomic_read(&(q)->pm_only)
673 #define blk_queue_fua(q)	test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
674 #define blk_queue_registered(q)	test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
675 #define blk_queue_nowait(q)	test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
676 
677 extern void blk_set_pm_only(struct request_queue *q);
678 extern void blk_clear_pm_only(struct request_queue *q);
679 
680 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
681 
682 #define rq_data_dir(rq)		(op_is_write(req_op(rq)) ? WRITE : READ)
683 
684 #define rq_dma_dir(rq) \
685 	(op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
686 
687 #define dma_map_bvec(dev, bv, dir, attrs) \
688 	dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
689 	(dir), (attrs))
690 
691 #define queue_to_disk(q)	(dev_to_disk(kobj_to_dev((q)->kobj.parent)))
692 
693 static inline bool queue_is_mq(struct request_queue *q)
694 {
695 	return q->mq_ops;
696 }
697 
698 #ifdef CONFIG_PM
699 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
700 {
701 	return q->rpm_status;
702 }
703 #else
704 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
705 {
706 	return RPM_ACTIVE;
707 }
708 #endif
709 
710 static inline enum blk_zoned_model
711 blk_queue_zoned_model(struct request_queue *q)
712 {
713 	if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
714 		return q->limits.zoned;
715 	return BLK_ZONED_NONE;
716 }
717 
718 static inline bool blk_queue_is_zoned(struct request_queue *q)
719 {
720 	switch (blk_queue_zoned_model(q)) {
721 	case BLK_ZONED_HA:
722 	case BLK_ZONED_HM:
723 		return true;
724 	default:
725 		return false;
726 	}
727 }
728 
729 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
730 {
731 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
732 }
733 
734 #ifdef CONFIG_BLK_DEV_ZONED
735 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
736 {
737 	return blk_queue_is_zoned(q) ? q->nr_zones : 0;
738 }
739 
740 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
741 					     sector_t sector)
742 {
743 	if (!blk_queue_is_zoned(q))
744 		return 0;
745 	return sector >> ilog2(q->limits.chunk_sectors);
746 }
747 
748 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
749 					 sector_t sector)
750 {
751 	if (!blk_queue_is_zoned(q))
752 		return false;
753 	if (!q->conv_zones_bitmap)
754 		return true;
755 	return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
756 }
757 
758 static inline void blk_queue_max_open_zones(struct request_queue *q,
759 		unsigned int max_open_zones)
760 {
761 	q->max_open_zones = max_open_zones;
762 }
763 
764 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
765 {
766 	return q->max_open_zones;
767 }
768 
769 static inline void blk_queue_max_active_zones(struct request_queue *q,
770 		unsigned int max_active_zones)
771 {
772 	q->max_active_zones = max_active_zones;
773 }
774 
775 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
776 {
777 	return q->max_active_zones;
778 }
779 #else /* CONFIG_BLK_DEV_ZONED */
780 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
781 {
782 	return 0;
783 }
784 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
785 					 sector_t sector)
786 {
787 	return false;
788 }
789 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
790 					     sector_t sector)
791 {
792 	return 0;
793 }
794 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
795 {
796 	return 0;
797 }
798 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
799 {
800 	return 0;
801 }
802 #endif /* CONFIG_BLK_DEV_ZONED */
803 
804 static inline bool rq_is_sync(struct request *rq)
805 {
806 	return op_is_sync(rq->cmd_flags);
807 }
808 
809 static inline bool rq_mergeable(struct request *rq)
810 {
811 	if (blk_rq_is_passthrough(rq))
812 		return false;
813 
814 	if (req_op(rq) == REQ_OP_FLUSH)
815 		return false;
816 
817 	if (req_op(rq) == REQ_OP_WRITE_ZEROES)
818 		return false;
819 
820 	if (req_op(rq) == REQ_OP_ZONE_APPEND)
821 		return false;
822 
823 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
824 		return false;
825 	if (rq->rq_flags & RQF_NOMERGE_FLAGS)
826 		return false;
827 
828 	return true;
829 }
830 
831 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
832 {
833 	if (bio_page(a) == bio_page(b) &&
834 	    bio_offset(a) == bio_offset(b))
835 		return true;
836 
837 	return false;
838 }
839 
840 static inline unsigned int blk_queue_depth(struct request_queue *q)
841 {
842 	if (q->queue_depth)
843 		return q->queue_depth;
844 
845 	return q->nr_requests;
846 }
847 
848 /*
849  * default timeout for SG_IO if none specified
850  */
851 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
852 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
853 
854 struct rq_map_data {
855 	struct page **pages;
856 	int page_order;
857 	int nr_entries;
858 	unsigned long offset;
859 	int null_mapped;
860 	int from_user;
861 };
862 
863 struct req_iterator {
864 	struct bvec_iter iter;
865 	struct bio *bio;
866 };
867 
868 /* This should not be used directly - use rq_for_each_segment */
869 #define for_each_bio(_bio)		\
870 	for (; _bio; _bio = _bio->bi_next)
871 #define __rq_for_each_bio(_bio, rq)	\
872 	if ((rq->bio))			\
873 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
874 
875 #define rq_for_each_segment(bvl, _rq, _iter)			\
876 	__rq_for_each_bio(_iter.bio, _rq)			\
877 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
878 
879 #define rq_for_each_bvec(bvl, _rq, _iter)			\
880 	__rq_for_each_bio(_iter.bio, _rq)			\
881 		bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
882 
883 #define rq_iter_last(bvec, _iter)				\
884 		(_iter.bio->bi_next == NULL &&			\
885 		 bio_iter_last(bvec, _iter.iter))
886 
887 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
888 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
889 #endif
890 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
891 extern void rq_flush_dcache_pages(struct request *rq);
892 #else
893 static inline void rq_flush_dcache_pages(struct request *rq)
894 {
895 }
896 #endif
897 
898 extern int blk_register_queue(struct gendisk *disk);
899 extern void blk_unregister_queue(struct gendisk *disk);
900 blk_qc_t submit_bio_noacct(struct bio *bio);
901 extern void blk_rq_init(struct request_queue *q, struct request *rq);
902 extern void blk_put_request(struct request *);
903 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
904 				       blk_mq_req_flags_t flags);
905 extern int blk_lld_busy(struct request_queue *q);
906 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
907 			     struct bio_set *bs, gfp_t gfp_mask,
908 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
909 			     void *data);
910 extern void blk_rq_unprep_clone(struct request *rq);
911 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
912 				     struct request *rq);
913 int blk_rq_append_bio(struct request *rq, struct bio *bio);
914 extern void blk_queue_split(struct bio **);
915 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
916 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
917 			      unsigned int, void __user *);
918 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
919 			  unsigned int, void __user *);
920 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
921 			 struct scsi_ioctl_command __user *);
922 extern int get_sg_io_hdr(struct sg_io_hdr *hdr, const void __user *argp);
923 extern int put_sg_io_hdr(const struct sg_io_hdr *hdr, void __user *argp);
924 
925 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
926 extern void blk_queue_exit(struct request_queue *q);
927 extern void blk_sync_queue(struct request_queue *q);
928 extern int blk_rq_map_user(struct request_queue *, struct request *,
929 			   struct rq_map_data *, void __user *, unsigned long,
930 			   gfp_t);
931 extern int blk_rq_unmap_user(struct bio *);
932 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
933 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
934 			       struct rq_map_data *, const struct iov_iter *,
935 			       gfp_t);
936 extern void blk_execute_rq(struct gendisk *, struct request *, int);
937 extern void blk_execute_rq_nowait(struct gendisk *,
938 				  struct request *, int, rq_end_io_fn *);
939 
940 /* Helper to convert REQ_OP_XXX to its string format XXX */
941 extern const char *blk_op_str(unsigned int op);
942 
943 int blk_status_to_errno(blk_status_t status);
944 blk_status_t errno_to_blk_status(int errno);
945 
946 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
947 
948 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
949 {
950 	return bdev->bd_disk->queue;	/* this is never NULL */
951 }
952 
953 /*
954  * The basic unit of block I/O is a sector. It is used in a number of contexts
955  * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
956  * bytes. Variables of type sector_t represent an offset or size that is a
957  * multiple of 512 bytes. Hence these two constants.
958  */
959 #ifndef SECTOR_SHIFT
960 #define SECTOR_SHIFT 9
961 #endif
962 #ifndef SECTOR_SIZE
963 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
964 #endif
965 
966 /*
967  * blk_rq_pos()			: the current sector
968  * blk_rq_bytes()		: bytes left in the entire request
969  * blk_rq_cur_bytes()		: bytes left in the current segment
970  * blk_rq_err_bytes()		: bytes left till the next error boundary
971  * blk_rq_sectors()		: sectors left in the entire request
972  * blk_rq_cur_sectors()		: sectors left in the current segment
973  * blk_rq_stats_sectors()	: sectors of the entire request used for stats
974  */
975 static inline sector_t blk_rq_pos(const struct request *rq)
976 {
977 	return rq->__sector;
978 }
979 
980 static inline unsigned int blk_rq_bytes(const struct request *rq)
981 {
982 	return rq->__data_len;
983 }
984 
985 static inline int blk_rq_cur_bytes(const struct request *rq)
986 {
987 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
988 }
989 
990 extern unsigned int blk_rq_err_bytes(const struct request *rq);
991 
992 static inline unsigned int blk_rq_sectors(const struct request *rq)
993 {
994 	return blk_rq_bytes(rq) >> SECTOR_SHIFT;
995 }
996 
997 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
998 {
999 	return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1000 }
1001 
1002 static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
1003 {
1004 	return rq->stats_sectors;
1005 }
1006 
1007 #ifdef CONFIG_BLK_DEV_ZONED
1008 
1009 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
1010 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
1011 
1012 static inline unsigned int blk_rq_zone_no(struct request *rq)
1013 {
1014 	return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1015 }
1016 
1017 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1018 {
1019 	return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1020 }
1021 #endif /* CONFIG_BLK_DEV_ZONED */
1022 
1023 /*
1024  * Some commands like WRITE SAME have a payload or data transfer size which
1025  * is different from the size of the request.  Any driver that supports such
1026  * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1027  * calculate the data transfer size.
1028  */
1029 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1030 {
1031 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1032 		return rq->special_vec.bv_len;
1033 	return blk_rq_bytes(rq);
1034 }
1035 
1036 /*
1037  * Return the first full biovec in the request.  The caller needs to check that
1038  * there are any bvecs before calling this helper.
1039  */
1040 static inline struct bio_vec req_bvec(struct request *rq)
1041 {
1042 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1043 		return rq->special_vec;
1044 	return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
1045 }
1046 
1047 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1048 						     int op)
1049 {
1050 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1051 		return min(q->limits.max_discard_sectors,
1052 			   UINT_MAX >> SECTOR_SHIFT);
1053 
1054 	if (unlikely(op == REQ_OP_WRITE_SAME))
1055 		return q->limits.max_write_same_sectors;
1056 
1057 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
1058 		return q->limits.max_write_zeroes_sectors;
1059 
1060 	return q->limits.max_sectors;
1061 }
1062 
1063 /*
1064  * Return maximum size of a request at given offset. Only valid for
1065  * file system requests.
1066  */
1067 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1068 					       sector_t offset,
1069 					       unsigned int chunk_sectors)
1070 {
1071 	if (!chunk_sectors) {
1072 		if (q->limits.chunk_sectors)
1073 			chunk_sectors = q->limits.chunk_sectors;
1074 		else
1075 			return q->limits.max_sectors;
1076 	}
1077 
1078 	if (likely(is_power_of_2(chunk_sectors)))
1079 		chunk_sectors -= offset & (chunk_sectors - 1);
1080 	else
1081 		chunk_sectors -= sector_div(offset, chunk_sectors);
1082 
1083 	return min(q->limits.max_sectors, chunk_sectors);
1084 }
1085 
1086 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1087 						  sector_t offset)
1088 {
1089 	struct request_queue *q = rq->q;
1090 
1091 	if (blk_rq_is_passthrough(rq))
1092 		return q->limits.max_hw_sectors;
1093 
1094 	if (!q->limits.chunk_sectors ||
1095 	    req_op(rq) == REQ_OP_DISCARD ||
1096 	    req_op(rq) == REQ_OP_SECURE_ERASE)
1097 		return blk_queue_get_max_sectors(q, req_op(rq));
1098 
1099 	return min(blk_max_size_offset(q, offset, 0),
1100 			blk_queue_get_max_sectors(q, req_op(rq)));
1101 }
1102 
1103 static inline unsigned int blk_rq_count_bios(struct request *rq)
1104 {
1105 	unsigned int nr_bios = 0;
1106 	struct bio *bio;
1107 
1108 	__rq_for_each_bio(bio, rq)
1109 		nr_bios++;
1110 
1111 	return nr_bios;
1112 }
1113 
1114 void blk_steal_bios(struct bio_list *list, struct request *rq);
1115 
1116 /*
1117  * Request completion related functions.
1118  *
1119  * blk_update_request() completes given number of bytes and updates
1120  * the request without completing it.
1121  */
1122 extern bool blk_update_request(struct request *rq, blk_status_t error,
1123 			       unsigned int nr_bytes);
1124 
1125 extern void blk_abort_request(struct request *);
1126 
1127 /*
1128  * Access functions for manipulating queue properties
1129  */
1130 extern void blk_cleanup_queue(struct request_queue *);
1131 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
1132 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1133 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1134 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1135 extern void blk_queue_max_discard_segments(struct request_queue *,
1136 		unsigned short);
1137 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1138 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1139 		unsigned int max_discard_sectors);
1140 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1141 		unsigned int max_write_same_sectors);
1142 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1143 		unsigned int max_write_same_sectors);
1144 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1145 extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
1146 		unsigned int max_zone_append_sectors);
1147 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1148 void blk_queue_zone_write_granularity(struct request_queue *q,
1149 				      unsigned int size);
1150 extern void blk_queue_alignment_offset(struct request_queue *q,
1151 				       unsigned int alignment);
1152 void blk_queue_update_readahead(struct request_queue *q);
1153 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1154 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1155 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1156 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1157 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1158 extern void blk_set_default_limits(struct queue_limits *lim);
1159 extern void blk_set_stacking_limits(struct queue_limits *lim);
1160 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1161 			    sector_t offset);
1162 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1163 			      sector_t offset);
1164 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1165 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1166 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1167 extern void blk_queue_dma_alignment(struct request_queue *, int);
1168 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1169 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1170 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1171 extern void blk_queue_required_elevator_features(struct request_queue *q,
1172 						 unsigned int features);
1173 extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
1174 					      struct device *dev);
1175 
1176 /*
1177  * Number of physical segments as sent to the device.
1178  *
1179  * Normally this is the number of discontiguous data segments sent by the
1180  * submitter.  But for data-less command like discard we might have no
1181  * actual data segments submitted, but the driver might have to add it's
1182  * own special payload.  In that case we still return 1 here so that this
1183  * special payload will be mapped.
1184  */
1185 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1186 {
1187 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1188 		return 1;
1189 	return rq->nr_phys_segments;
1190 }
1191 
1192 /*
1193  * Number of discard segments (or ranges) the driver needs to fill in.
1194  * Each discard bio merged into a request is counted as one segment.
1195  */
1196 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1197 {
1198 	return max_t(unsigned short, rq->nr_phys_segments, 1);
1199 }
1200 
1201 int __blk_rq_map_sg(struct request_queue *q, struct request *rq,
1202 		struct scatterlist *sglist, struct scatterlist **last_sg);
1203 static inline int blk_rq_map_sg(struct request_queue *q, struct request *rq,
1204 		struct scatterlist *sglist)
1205 {
1206 	struct scatterlist *last_sg = NULL;
1207 
1208 	return __blk_rq_map_sg(q, rq, sglist, &last_sg);
1209 }
1210 extern void blk_dump_rq_flags(struct request *, char *);
1211 
1212 bool __must_check blk_get_queue(struct request_queue *);
1213 struct request_queue *blk_alloc_queue(int node_id);
1214 extern void blk_put_queue(struct request_queue *);
1215 extern void blk_set_queue_dying(struct request_queue *);
1216 
1217 #ifdef CONFIG_BLOCK
1218 /*
1219  * blk_plug permits building a queue of related requests by holding the I/O
1220  * fragments for a short period. This allows merging of sequential requests
1221  * into single larger request. As the requests are moved from a per-task list to
1222  * the device's request_queue in a batch, this results in improved scalability
1223  * as the lock contention for request_queue lock is reduced.
1224  *
1225  * It is ok not to disable preemption when adding the request to the plug list
1226  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1227  * the plug list when the task sleeps by itself. For details, please see
1228  * schedule() where blk_schedule_flush_plug() is called.
1229  */
1230 struct blk_plug {
1231 	struct list_head mq_list; /* blk-mq requests */
1232 	struct list_head cb_list; /* md requires an unplug callback */
1233 	unsigned short rq_count;
1234 	bool multiple_queues;
1235 	bool nowait;
1236 };
1237 #define BLK_MAX_REQUEST_COUNT 16
1238 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1239 
1240 struct blk_plug_cb;
1241 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1242 struct blk_plug_cb {
1243 	struct list_head list;
1244 	blk_plug_cb_fn callback;
1245 	void *data;
1246 };
1247 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1248 					     void *data, int size);
1249 extern void blk_start_plug(struct blk_plug *);
1250 extern void blk_finish_plug(struct blk_plug *);
1251 extern void blk_flush_plug_list(struct blk_plug *, bool);
1252 
1253 static inline void blk_flush_plug(struct task_struct *tsk)
1254 {
1255 	struct blk_plug *plug = tsk->plug;
1256 
1257 	if (plug)
1258 		blk_flush_plug_list(plug, false);
1259 }
1260 
1261 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1262 {
1263 	struct blk_plug *plug = tsk->plug;
1264 
1265 	if (plug)
1266 		blk_flush_plug_list(plug, true);
1267 }
1268 
1269 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1270 {
1271 	struct blk_plug *plug = tsk->plug;
1272 
1273 	return plug &&
1274 		 (!list_empty(&plug->mq_list) ||
1275 		 !list_empty(&plug->cb_list));
1276 }
1277 
1278 int blkdev_issue_flush(struct block_device *bdev);
1279 long nr_blockdev_pages(void);
1280 #else /* CONFIG_BLOCK */
1281 struct blk_plug {
1282 };
1283 
1284 static inline void blk_start_plug(struct blk_plug *plug)
1285 {
1286 }
1287 
1288 static inline void blk_finish_plug(struct blk_plug *plug)
1289 {
1290 }
1291 
1292 static inline void blk_flush_plug(struct task_struct *task)
1293 {
1294 }
1295 
1296 static inline void blk_schedule_flush_plug(struct task_struct *task)
1297 {
1298 }
1299 
1300 
1301 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1302 {
1303 	return false;
1304 }
1305 
1306 static inline int blkdev_issue_flush(struct block_device *bdev)
1307 {
1308 	return 0;
1309 }
1310 
1311 static inline long nr_blockdev_pages(void)
1312 {
1313 	return 0;
1314 }
1315 #endif /* CONFIG_BLOCK */
1316 
1317 extern void blk_io_schedule(void);
1318 
1319 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1320 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1321 
1322 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
1323 
1324 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1325 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1326 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1327 		sector_t nr_sects, gfp_t gfp_mask, int flags,
1328 		struct bio **biop);
1329 
1330 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1331 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1332 
1333 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1334 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1335 		unsigned flags);
1336 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1337 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1338 
1339 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1340 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1341 {
1342 	return blkdev_issue_discard(sb->s_bdev,
1343 				    block << (sb->s_blocksize_bits -
1344 					      SECTOR_SHIFT),
1345 				    nr_blocks << (sb->s_blocksize_bits -
1346 						  SECTOR_SHIFT),
1347 				    gfp_mask, flags);
1348 }
1349 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1350 		sector_t nr_blocks, gfp_t gfp_mask)
1351 {
1352 	return blkdev_issue_zeroout(sb->s_bdev,
1353 				    block << (sb->s_blocksize_bits -
1354 					      SECTOR_SHIFT),
1355 				    nr_blocks << (sb->s_blocksize_bits -
1356 						  SECTOR_SHIFT),
1357 				    gfp_mask, 0);
1358 }
1359 
1360 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1361 
1362 static inline bool bdev_is_partition(struct block_device *bdev)
1363 {
1364 	return bdev->bd_partno;
1365 }
1366 
1367 enum blk_default_limits {
1368 	BLK_MAX_SEGMENTS	= 128,
1369 	BLK_SAFE_MAX_SECTORS	= 255,
1370 	BLK_DEF_MAX_SECTORS	= 2560,
1371 	BLK_MAX_SEGMENT_SIZE	= 65536,
1372 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1373 };
1374 
1375 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1376 {
1377 	return q->limits.seg_boundary_mask;
1378 }
1379 
1380 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1381 {
1382 	return q->limits.virt_boundary_mask;
1383 }
1384 
1385 static inline unsigned int queue_max_sectors(const struct request_queue *q)
1386 {
1387 	return q->limits.max_sectors;
1388 }
1389 
1390 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1391 {
1392 	return q->limits.max_hw_sectors;
1393 }
1394 
1395 static inline unsigned short queue_max_segments(const struct request_queue *q)
1396 {
1397 	return q->limits.max_segments;
1398 }
1399 
1400 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1401 {
1402 	return q->limits.max_discard_segments;
1403 }
1404 
1405 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1406 {
1407 	return q->limits.max_segment_size;
1408 }
1409 
1410 static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
1411 {
1412 
1413 	const struct queue_limits *l = &q->limits;
1414 
1415 	return min(l->max_zone_append_sectors, l->max_sectors);
1416 }
1417 
1418 static inline unsigned queue_logical_block_size(const struct request_queue *q)
1419 {
1420 	int retval = 512;
1421 
1422 	if (q && q->limits.logical_block_size)
1423 		retval = q->limits.logical_block_size;
1424 
1425 	return retval;
1426 }
1427 
1428 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1429 {
1430 	return queue_logical_block_size(bdev_get_queue(bdev));
1431 }
1432 
1433 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1434 {
1435 	return q->limits.physical_block_size;
1436 }
1437 
1438 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1439 {
1440 	return queue_physical_block_size(bdev_get_queue(bdev));
1441 }
1442 
1443 static inline unsigned int queue_io_min(const struct request_queue *q)
1444 {
1445 	return q->limits.io_min;
1446 }
1447 
1448 static inline int bdev_io_min(struct block_device *bdev)
1449 {
1450 	return queue_io_min(bdev_get_queue(bdev));
1451 }
1452 
1453 static inline unsigned int queue_io_opt(const struct request_queue *q)
1454 {
1455 	return q->limits.io_opt;
1456 }
1457 
1458 static inline int bdev_io_opt(struct block_device *bdev)
1459 {
1460 	return queue_io_opt(bdev_get_queue(bdev));
1461 }
1462 
1463 static inline unsigned int
1464 queue_zone_write_granularity(const struct request_queue *q)
1465 {
1466 	return q->limits.zone_write_granularity;
1467 }
1468 
1469 static inline unsigned int
1470 bdev_zone_write_granularity(struct block_device *bdev)
1471 {
1472 	return queue_zone_write_granularity(bdev_get_queue(bdev));
1473 }
1474 
1475 static inline int queue_alignment_offset(const struct request_queue *q)
1476 {
1477 	if (q->limits.misaligned)
1478 		return -1;
1479 
1480 	return q->limits.alignment_offset;
1481 }
1482 
1483 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1484 {
1485 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1486 	unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1487 		<< SECTOR_SHIFT;
1488 
1489 	return (granularity + lim->alignment_offset - alignment) % granularity;
1490 }
1491 
1492 static inline int bdev_alignment_offset(struct block_device *bdev)
1493 {
1494 	struct request_queue *q = bdev_get_queue(bdev);
1495 
1496 	if (q->limits.misaligned)
1497 		return -1;
1498 	if (bdev_is_partition(bdev))
1499 		return queue_limit_alignment_offset(&q->limits,
1500 				bdev->bd_start_sect);
1501 	return q->limits.alignment_offset;
1502 }
1503 
1504 static inline int queue_discard_alignment(const struct request_queue *q)
1505 {
1506 	if (q->limits.discard_misaligned)
1507 		return -1;
1508 
1509 	return q->limits.discard_alignment;
1510 }
1511 
1512 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1513 {
1514 	unsigned int alignment, granularity, offset;
1515 
1516 	if (!lim->max_discard_sectors)
1517 		return 0;
1518 
1519 	/* Why are these in bytes, not sectors? */
1520 	alignment = lim->discard_alignment >> SECTOR_SHIFT;
1521 	granularity = lim->discard_granularity >> SECTOR_SHIFT;
1522 	if (!granularity)
1523 		return 0;
1524 
1525 	/* Offset of the partition start in 'granularity' sectors */
1526 	offset = sector_div(sector, granularity);
1527 
1528 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1529 	offset = (granularity + alignment - offset) % granularity;
1530 
1531 	/* Turn it back into bytes, gaah */
1532 	return offset << SECTOR_SHIFT;
1533 }
1534 
1535 static inline int bdev_discard_alignment(struct block_device *bdev)
1536 {
1537 	struct request_queue *q = bdev_get_queue(bdev);
1538 
1539 	if (bdev_is_partition(bdev))
1540 		return queue_limit_discard_alignment(&q->limits,
1541 				bdev->bd_start_sect);
1542 	return q->limits.discard_alignment;
1543 }
1544 
1545 static inline unsigned int bdev_write_same(struct block_device *bdev)
1546 {
1547 	struct request_queue *q = bdev_get_queue(bdev);
1548 
1549 	if (q)
1550 		return q->limits.max_write_same_sectors;
1551 
1552 	return 0;
1553 }
1554 
1555 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1556 {
1557 	struct request_queue *q = bdev_get_queue(bdev);
1558 
1559 	if (q)
1560 		return q->limits.max_write_zeroes_sectors;
1561 
1562 	return 0;
1563 }
1564 
1565 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1566 {
1567 	struct request_queue *q = bdev_get_queue(bdev);
1568 
1569 	if (q)
1570 		return blk_queue_zoned_model(q);
1571 
1572 	return BLK_ZONED_NONE;
1573 }
1574 
1575 static inline bool bdev_is_zoned(struct block_device *bdev)
1576 {
1577 	struct request_queue *q = bdev_get_queue(bdev);
1578 
1579 	if (q)
1580 		return blk_queue_is_zoned(q);
1581 
1582 	return false;
1583 }
1584 
1585 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1586 {
1587 	struct request_queue *q = bdev_get_queue(bdev);
1588 
1589 	if (q)
1590 		return blk_queue_zone_sectors(q);
1591 	return 0;
1592 }
1593 
1594 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1595 {
1596 	struct request_queue *q = bdev_get_queue(bdev);
1597 
1598 	if (q)
1599 		return queue_max_open_zones(q);
1600 	return 0;
1601 }
1602 
1603 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1604 {
1605 	struct request_queue *q = bdev_get_queue(bdev);
1606 
1607 	if (q)
1608 		return queue_max_active_zones(q);
1609 	return 0;
1610 }
1611 
1612 static inline int queue_dma_alignment(const struct request_queue *q)
1613 {
1614 	return q ? q->dma_alignment : 511;
1615 }
1616 
1617 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1618 				 unsigned int len)
1619 {
1620 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1621 	return !(addr & alignment) && !(len & alignment);
1622 }
1623 
1624 /* assumes size > 256 */
1625 static inline unsigned int blksize_bits(unsigned int size)
1626 {
1627 	unsigned int bits = 8;
1628 	do {
1629 		bits++;
1630 		size >>= 1;
1631 	} while (size > 256);
1632 	return bits;
1633 }
1634 
1635 static inline unsigned int block_size(struct block_device *bdev)
1636 {
1637 	return 1 << bdev->bd_inode->i_blkbits;
1638 }
1639 
1640 int kblockd_schedule_work(struct work_struct *work);
1641 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1642 
1643 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1644 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1645 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1646 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1647 
1648 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1649 
1650 enum blk_integrity_flags {
1651 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1652 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1653 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1654 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1655 };
1656 
1657 struct blk_integrity_iter {
1658 	void			*prot_buf;
1659 	void			*data_buf;
1660 	sector_t		seed;
1661 	unsigned int		data_size;
1662 	unsigned short		interval;
1663 	const char		*disk_name;
1664 };
1665 
1666 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1667 typedef void (integrity_prepare_fn) (struct request *);
1668 typedef void (integrity_complete_fn) (struct request *, unsigned int);
1669 
1670 struct blk_integrity_profile {
1671 	integrity_processing_fn		*generate_fn;
1672 	integrity_processing_fn		*verify_fn;
1673 	integrity_prepare_fn		*prepare_fn;
1674 	integrity_complete_fn		*complete_fn;
1675 	const char			*name;
1676 };
1677 
1678 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1679 extern void blk_integrity_unregister(struct gendisk *);
1680 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1681 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1682 				   struct scatterlist *);
1683 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1684 
1685 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1686 {
1687 	struct blk_integrity *bi = &disk->queue->integrity;
1688 
1689 	if (!bi->profile)
1690 		return NULL;
1691 
1692 	return bi;
1693 }
1694 
1695 static inline
1696 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1697 {
1698 	return blk_get_integrity(bdev->bd_disk);
1699 }
1700 
1701 static inline bool
1702 blk_integrity_queue_supports_integrity(struct request_queue *q)
1703 {
1704 	return q->integrity.profile;
1705 }
1706 
1707 static inline bool blk_integrity_rq(struct request *rq)
1708 {
1709 	return rq->cmd_flags & REQ_INTEGRITY;
1710 }
1711 
1712 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1713 						    unsigned int segs)
1714 {
1715 	q->limits.max_integrity_segments = segs;
1716 }
1717 
1718 static inline unsigned short
1719 queue_max_integrity_segments(const struct request_queue *q)
1720 {
1721 	return q->limits.max_integrity_segments;
1722 }
1723 
1724 /**
1725  * bio_integrity_intervals - Return number of integrity intervals for a bio
1726  * @bi:		blk_integrity profile for device
1727  * @sectors:	Size of the bio in 512-byte sectors
1728  *
1729  * Description: The block layer calculates everything in 512 byte
1730  * sectors but integrity metadata is done in terms of the data integrity
1731  * interval size of the storage device.  Convert the block layer sectors
1732  * to the appropriate number of integrity intervals.
1733  */
1734 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1735 						   unsigned int sectors)
1736 {
1737 	return sectors >> (bi->interval_exp - 9);
1738 }
1739 
1740 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1741 					       unsigned int sectors)
1742 {
1743 	return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1744 }
1745 
1746 /*
1747  * Return the first bvec that contains integrity data.  Only drivers that are
1748  * limited to a single integrity segment should use this helper.
1749  */
1750 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1751 {
1752 	if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1753 		return NULL;
1754 	return rq->bio->bi_integrity->bip_vec;
1755 }
1756 
1757 #else /* CONFIG_BLK_DEV_INTEGRITY */
1758 
1759 struct bio;
1760 struct block_device;
1761 struct gendisk;
1762 struct blk_integrity;
1763 
1764 static inline int blk_integrity_rq(struct request *rq)
1765 {
1766 	return 0;
1767 }
1768 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1769 					    struct bio *b)
1770 {
1771 	return 0;
1772 }
1773 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1774 					  struct bio *b,
1775 					  struct scatterlist *s)
1776 {
1777 	return 0;
1778 }
1779 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1780 {
1781 	return NULL;
1782 }
1783 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1784 {
1785 	return NULL;
1786 }
1787 static inline bool
1788 blk_integrity_queue_supports_integrity(struct request_queue *q)
1789 {
1790 	return false;
1791 }
1792 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1793 {
1794 	return 0;
1795 }
1796 static inline void blk_integrity_register(struct gendisk *d,
1797 					 struct blk_integrity *b)
1798 {
1799 }
1800 static inline void blk_integrity_unregister(struct gendisk *d)
1801 {
1802 }
1803 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1804 						    unsigned int segs)
1805 {
1806 }
1807 static inline unsigned short queue_max_integrity_segments(const struct request_queue *q)
1808 {
1809 	return 0;
1810 }
1811 
1812 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1813 						   unsigned int sectors)
1814 {
1815 	return 0;
1816 }
1817 
1818 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1819 					       unsigned int sectors)
1820 {
1821 	return 0;
1822 }
1823 
1824 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1825 {
1826 	return NULL;
1827 }
1828 
1829 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1830 
1831 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1832 
1833 bool blk_ksm_register(struct blk_keyslot_manager *ksm, struct request_queue *q);
1834 
1835 void blk_ksm_unregister(struct request_queue *q);
1836 
1837 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1838 
1839 static inline bool blk_ksm_register(struct blk_keyslot_manager *ksm,
1840 				    struct request_queue *q)
1841 {
1842 	return true;
1843 }
1844 
1845 static inline void blk_ksm_unregister(struct request_queue *q) { }
1846 
1847 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1848 
1849 
1850 struct block_device_operations {
1851 	blk_qc_t (*submit_bio) (struct bio *bio);
1852 	int (*open) (struct block_device *, fmode_t);
1853 	void (*release) (struct gendisk *, fmode_t);
1854 	int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1855 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1856 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1857 	unsigned int (*check_events) (struct gendisk *disk,
1858 				      unsigned int clearing);
1859 	void (*unlock_native_capacity) (struct gendisk *);
1860 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1861 	int (*set_read_only)(struct block_device *bdev, bool ro);
1862 	/* this callback is with swap_lock and sometimes page table lock held */
1863 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1864 	int (*report_zones)(struct gendisk *, sector_t sector,
1865 			unsigned int nr_zones, report_zones_cb cb, void *data);
1866 	char *(*devnode)(struct gendisk *disk, umode_t *mode);
1867 	struct module *owner;
1868 	const struct pr_ops *pr_ops;
1869 };
1870 
1871 #ifdef CONFIG_COMPAT
1872 extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1873 				      unsigned int, unsigned long);
1874 #else
1875 #define blkdev_compat_ptr_ioctl NULL
1876 #endif
1877 
1878 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1879 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1880 						struct writeback_control *);
1881 
1882 #ifdef CONFIG_BLK_DEV_ZONED
1883 bool blk_req_needs_zone_write_lock(struct request *rq);
1884 bool blk_req_zone_write_trylock(struct request *rq);
1885 void __blk_req_zone_write_lock(struct request *rq);
1886 void __blk_req_zone_write_unlock(struct request *rq);
1887 
1888 static inline void blk_req_zone_write_lock(struct request *rq)
1889 {
1890 	if (blk_req_needs_zone_write_lock(rq))
1891 		__blk_req_zone_write_lock(rq);
1892 }
1893 
1894 static inline void blk_req_zone_write_unlock(struct request *rq)
1895 {
1896 	if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1897 		__blk_req_zone_write_unlock(rq);
1898 }
1899 
1900 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1901 {
1902 	return rq->q->seq_zones_wlock &&
1903 		test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1904 }
1905 
1906 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1907 {
1908 	if (!blk_req_needs_zone_write_lock(rq))
1909 		return true;
1910 	return !blk_req_zone_is_write_locked(rq);
1911 }
1912 #else
1913 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1914 {
1915 	return false;
1916 }
1917 
1918 static inline void blk_req_zone_write_lock(struct request *rq)
1919 {
1920 }
1921 
1922 static inline void blk_req_zone_write_unlock(struct request *rq)
1923 {
1924 }
1925 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1926 {
1927 	return false;
1928 }
1929 
1930 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1931 {
1932 	return true;
1933 }
1934 #endif /* CONFIG_BLK_DEV_ZONED */
1935 
1936 static inline void blk_wake_io_task(struct task_struct *waiter)
1937 {
1938 	/*
1939 	 * If we're polling, the task itself is doing the completions. For
1940 	 * that case, we don't need to signal a wakeup, it's enough to just
1941 	 * mark us as RUNNING.
1942 	 */
1943 	if (waiter == current)
1944 		__set_current_state(TASK_RUNNING);
1945 	else
1946 		wake_up_process(waiter);
1947 }
1948 
1949 unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1950 		unsigned int op);
1951 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1952 		unsigned long start_time);
1953 
1954 unsigned long bio_start_io_acct(struct bio *bio);
1955 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1956 		struct block_device *orig_bdev);
1957 
1958 /**
1959  * bio_end_io_acct - end I/O accounting for bio based drivers
1960  * @bio:	bio to end account for
1961  * @start:	start time returned by bio_start_io_acct()
1962  */
1963 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1964 {
1965 	return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1966 }
1967 
1968 int bdev_read_only(struct block_device *bdev);
1969 int set_blocksize(struct block_device *bdev, int size);
1970 
1971 const char *bdevname(struct block_device *bdev, char *buffer);
1972 int lookup_bdev(const char *pathname, dev_t *dev);
1973 
1974 void blkdev_show(struct seq_file *seqf, off_t offset);
1975 
1976 #define BDEVNAME_SIZE	32	/* Largest string for a blockdev identifier */
1977 #define BDEVT_SIZE	10	/* Largest string for MAJ:MIN for blkdev */
1978 #ifdef CONFIG_BLOCK
1979 #define BLKDEV_MAJOR_MAX	512
1980 #else
1981 #define BLKDEV_MAJOR_MAX	0
1982 #endif
1983 
1984 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1985 		void *holder);
1986 struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
1987 int bd_prepare_to_claim(struct block_device *bdev, void *holder);
1988 void bd_abort_claiming(struct block_device *bdev, void *holder);
1989 void blkdev_put(struct block_device *bdev, fmode_t mode);
1990 
1991 /* just for blk-cgroup, don't use elsewhere */
1992 struct block_device *blkdev_get_no_open(dev_t dev);
1993 void blkdev_put_no_open(struct block_device *bdev);
1994 
1995 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
1996 void bdev_add(struct block_device *bdev, dev_t dev);
1997 struct block_device *I_BDEV(struct inode *inode);
1998 struct block_device *bdgrab(struct block_device *bdev);
1999 void bdput(struct block_device *);
2000 int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
2001 		loff_t lend);
2002 
2003 #ifdef CONFIG_BLOCK
2004 void invalidate_bdev(struct block_device *bdev);
2005 int sync_blockdev(struct block_device *bdev);
2006 #else
2007 static inline void invalidate_bdev(struct block_device *bdev)
2008 {
2009 }
2010 static inline int sync_blockdev(struct block_device *bdev)
2011 {
2012 	return 0;
2013 }
2014 #endif
2015 int fsync_bdev(struct block_device *bdev);
2016 
2017 int freeze_bdev(struct block_device *bdev);
2018 int thaw_bdev(struct block_device *bdev);
2019 
2020 #endif /* _LINUX_BLKDEV_H */
2021