xref: /linux-6.15/include/linux/blkdev.h (revision f87deada)
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 
8 #ifdef CONFIG_BLOCK
9 
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
30 #include <linux/seqlock.h>
31 #include <linux/u64_stats_sync.h>
32 
33 struct module;
34 struct scsi_ioctl_command;
35 
36 struct request_queue;
37 struct elevator_queue;
38 struct blk_trace;
39 struct request;
40 struct sg_io_hdr;
41 struct bsg_job;
42 struct blkcg_gq;
43 struct blk_flush_queue;
44 struct pr_ops;
45 struct rq_wb;
46 struct blk_queue_stats;
47 struct blk_stat_callback;
48 
49 #define BLKDEV_MIN_RQ	4
50 #define BLKDEV_MAX_RQ	128	/* Default maximum */
51 
52 /* Must be consistent with blk_mq_poll_stats_bkt() */
53 #define BLK_MQ_POLL_STATS_BKTS 16
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		3
60 
61 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
62 
63 #define BLK_RL_SYNCFULL		(1U << 0)
64 #define BLK_RL_ASYNCFULL	(1U << 1)
65 
66 struct request_list {
67 	struct request_queue	*q;	/* the queue this rl belongs to */
68 #ifdef CONFIG_BLK_CGROUP
69 	struct blkcg_gq		*blkg;	/* blkg this request pool belongs to */
70 #endif
71 	/*
72 	 * count[], starved[], and wait[] are indexed by
73 	 * BLK_RW_SYNC/BLK_RW_ASYNC
74 	 */
75 	int			count[2];
76 	int			starved[2];
77 	mempool_t		*rq_pool;
78 	wait_queue_head_t	wait[2];
79 	unsigned int		flags;
80 };
81 
82 /*
83  * request flags */
84 typedef __u32 __bitwise req_flags_t;
85 
86 /* elevator knows about this request */
87 #define RQF_SORTED		((__force req_flags_t)(1 << 0))
88 /* drive already may have started this one */
89 #define RQF_STARTED		((__force req_flags_t)(1 << 1))
90 /* uses tagged queueing */
91 #define RQF_QUEUED		((__force req_flags_t)(1 << 2))
92 /* may not be passed by ioscheduler */
93 #define RQF_SOFTBARRIER		((__force req_flags_t)(1 << 3))
94 /* request for flush sequence */
95 #define RQF_FLUSH_SEQ		((__force req_flags_t)(1 << 4))
96 /* merge of different types, fail separately */
97 #define RQF_MIXED_MERGE		((__force req_flags_t)(1 << 5))
98 /* track inflight for MQ */
99 #define RQF_MQ_INFLIGHT		((__force req_flags_t)(1 << 6))
100 /* don't call prep for this one */
101 #define RQF_DONTPREP		((__force req_flags_t)(1 << 7))
102 /* set for "ide_preempt" requests and also for requests for which the SCSI
103    "quiesce" state must be ignored. */
104 #define RQF_PREEMPT		((__force req_flags_t)(1 << 8))
105 /* contains copies of user pages */
106 #define RQF_COPY_USER		((__force req_flags_t)(1 << 9))
107 /* vaguely specified driver internal error.  Ignored by the block layer */
108 #define RQF_FAILED		((__force req_flags_t)(1 << 10))
109 /* don't warn about errors */
110 #define RQF_QUIET		((__force req_flags_t)(1 << 11))
111 /* elevator private data attached */
112 #define RQF_ELVPRIV		((__force req_flags_t)(1 << 12))
113 /* account I/O stat */
114 #define RQF_IO_STAT		((__force req_flags_t)(1 << 13))
115 /* request came from our alloc pool */
116 #define RQF_ALLOCED		((__force req_flags_t)(1 << 14))
117 /* runtime pm request */
118 #define RQF_PM			((__force req_flags_t)(1 << 15))
119 /* on IO scheduler merge hash */
120 #define RQF_HASHED		((__force req_flags_t)(1 << 16))
121 /* IO stats tracking on */
122 #define RQF_STATS		((__force req_flags_t)(1 << 17))
123 /* Look at ->special_vec for the actual data payload instead of the
124    bio chain. */
125 #define RQF_SPECIAL_PAYLOAD	((__force req_flags_t)(1 << 18))
126 /* The per-zone write lock is held for this request */
127 #define RQF_ZONE_WRITE_LOCKED	((__force req_flags_t)(1 << 19))
128 /* timeout is expired */
129 #define RQF_MQ_TIMEOUT_EXPIRED	((__force req_flags_t)(1 << 20))
130 /* already slept for hybrid poll */
131 #define RQF_MQ_POLL_SLEPT	((__force req_flags_t)(1 << 21))
132 
133 /* flags that prevent us from merging requests: */
134 #define RQF_NOMERGE_FLAGS \
135 	(RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
136 
137 /*
138  * Try to put the fields that are referenced together in the same cacheline.
139  *
140  * If you modify this structure, make sure to update blk_rq_init() and
141  * especially blk_mq_rq_ctx_init() to take care of the added fields.
142  */
143 struct request {
144 	struct request_queue *q;
145 	struct blk_mq_ctx *mq_ctx;
146 
147 	int cpu;
148 	unsigned int cmd_flags;		/* op and common flags */
149 	req_flags_t rq_flags;
150 
151 	int internal_tag;
152 
153 	/* the following two fields are internal, NEVER access directly */
154 	unsigned int __data_len;	/* total data len */
155 	int tag;
156 	sector_t __sector;		/* sector cursor */
157 
158 	struct bio *bio;
159 	struct bio *biotail;
160 
161 	struct list_head queuelist;
162 
163 	/*
164 	 * The hash is used inside the scheduler, and killed once the
165 	 * request reaches the dispatch list. The ipi_list is only used
166 	 * to queue the request for softirq completion, which is long
167 	 * after the request has been unhashed (and even removed from
168 	 * the dispatch list).
169 	 */
170 	union {
171 		struct hlist_node hash;	/* merge hash */
172 		struct list_head ipi_list;
173 	};
174 
175 	/*
176 	 * The rb_node is only used inside the io scheduler, requests
177 	 * are pruned when moved to the dispatch queue. So let the
178 	 * completion_data share space with the rb_node.
179 	 */
180 	union {
181 		struct rb_node rb_node;	/* sort/lookup */
182 		struct bio_vec special_vec;
183 		void *completion_data;
184 		int error_count; /* for legacy drivers, don't use */
185 	};
186 
187 	/*
188 	 * Three pointers are available for the IO schedulers, if they need
189 	 * more they have to dynamically allocate it.  Flush requests are
190 	 * never put on the IO scheduler. So let the flush fields share
191 	 * space with the elevator data.
192 	 */
193 	union {
194 		struct {
195 			struct io_cq		*icq;
196 			void			*priv[2];
197 		} elv;
198 
199 		struct {
200 			unsigned int		seq;
201 			struct list_head	list;
202 			rq_end_io_fn		*saved_end_io;
203 		} flush;
204 	};
205 
206 	struct gendisk *rq_disk;
207 	struct hd_struct *part;
208 	unsigned long start_time;
209 	struct blk_issue_stat issue_stat;
210 	/* Number of scatter-gather DMA addr+len pairs after
211 	 * physical address coalescing is performed.
212 	 */
213 	unsigned short nr_phys_segments;
214 
215 #if defined(CONFIG_BLK_DEV_INTEGRITY)
216 	unsigned short nr_integrity_segments;
217 #endif
218 
219 	unsigned short write_hint;
220 	unsigned short ioprio;
221 
222 	unsigned int timeout;
223 
224 	void *special;		/* opaque pointer available for LLD use */
225 
226 	unsigned int extra_len;	/* length of alignment and padding */
227 
228 	/*
229 	 * On blk-mq, the lower bits of ->gstate (generation number and
230 	 * state) carry the MQ_RQ_* state value and the upper bits the
231 	 * generation number which is monotonically incremented and used to
232 	 * distinguish the reuse instances.
233 	 *
234 	 * ->gstate_seq allows updates to ->gstate and other fields
235 	 * (currently ->deadline) during request start to be read
236 	 * atomically from the timeout path, so that it can operate on a
237 	 * coherent set of information.
238 	 */
239 	seqcount_t gstate_seq;
240 	u64 gstate;
241 
242 	/*
243 	 * ->aborted_gstate is used by the timeout to claim a specific
244 	 * recycle instance of this request.  See blk_mq_timeout_work().
245 	 */
246 	struct u64_stats_sync aborted_gstate_sync;
247 	u64 aborted_gstate;
248 
249 	/* access through blk_rq_set_deadline, blk_rq_deadline */
250 	unsigned long __deadline;
251 
252 	struct list_head timeout_list;
253 
254 	union {
255 		struct __call_single_data csd;
256 		u64 fifo_time;
257 	};
258 
259 	/*
260 	 * completion callback.
261 	 */
262 	rq_end_io_fn *end_io;
263 	void *end_io_data;
264 
265 	/* for bidi */
266 	struct request *next_rq;
267 
268 #ifdef CONFIG_BLK_CGROUP
269 	struct request_list *rl;		/* rl this rq is alloced from */
270 	unsigned long long start_time_ns;
271 	unsigned long long io_start_time_ns;    /* when passed to hardware */
272 #endif
273 };
274 
275 static inline bool blk_op_is_scsi(unsigned int op)
276 {
277 	return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
278 }
279 
280 static inline bool blk_op_is_private(unsigned int op)
281 {
282 	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
283 }
284 
285 static inline bool blk_rq_is_scsi(struct request *rq)
286 {
287 	return blk_op_is_scsi(req_op(rq));
288 }
289 
290 static inline bool blk_rq_is_private(struct request *rq)
291 {
292 	return blk_op_is_private(req_op(rq));
293 }
294 
295 static inline bool blk_rq_is_passthrough(struct request *rq)
296 {
297 	return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
298 }
299 
300 static inline bool bio_is_passthrough(struct bio *bio)
301 {
302 	unsigned op = bio_op(bio);
303 
304 	return blk_op_is_scsi(op) || blk_op_is_private(op);
305 }
306 
307 static inline unsigned short req_get_ioprio(struct request *req)
308 {
309 	return req->ioprio;
310 }
311 
312 #include <linux/elevator.h>
313 
314 struct blk_queue_ctx;
315 
316 typedef void (request_fn_proc) (struct request_queue *q);
317 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
318 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
319 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
320 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
321 
322 struct bio_vec;
323 typedef void (softirq_done_fn)(struct request *);
324 typedef int (dma_drain_needed_fn)(struct request *);
325 typedef int (lld_busy_fn) (struct request_queue *q);
326 typedef int (bsg_job_fn) (struct bsg_job *);
327 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
328 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
329 
330 enum blk_eh_timer_return {
331 	BLK_EH_NOT_HANDLED,
332 	BLK_EH_HANDLED,
333 	BLK_EH_RESET_TIMER,
334 };
335 
336 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
337 
338 enum blk_queue_state {
339 	Queue_down,
340 	Queue_up,
341 };
342 
343 struct blk_queue_tag {
344 	struct request **tag_index;	/* map of busy tags */
345 	unsigned long *tag_map;		/* bit map of free/busy tags */
346 	int max_depth;			/* what we will send to device */
347 	int real_max_depth;		/* what the array can hold */
348 	atomic_t refcnt;		/* map can be shared */
349 	int alloc_policy;		/* tag allocation policy */
350 	int next_tag;			/* next tag */
351 };
352 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
353 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
354 
355 #define BLK_SCSI_MAX_CMDS	(256)
356 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
357 
358 /*
359  * Zoned block device models (zoned limit).
360  */
361 enum blk_zoned_model {
362 	BLK_ZONED_NONE,	/* Regular block device */
363 	BLK_ZONED_HA,	/* Host-aware zoned block device */
364 	BLK_ZONED_HM,	/* Host-managed zoned block device */
365 };
366 
367 struct queue_limits {
368 	unsigned long		bounce_pfn;
369 	unsigned long		seg_boundary_mask;
370 	unsigned long		virt_boundary_mask;
371 
372 	unsigned int		max_hw_sectors;
373 	unsigned int		max_dev_sectors;
374 	unsigned int		chunk_sectors;
375 	unsigned int		max_sectors;
376 	unsigned int		max_segment_size;
377 	unsigned int		physical_block_size;
378 	unsigned int		alignment_offset;
379 	unsigned int		io_min;
380 	unsigned int		io_opt;
381 	unsigned int		max_discard_sectors;
382 	unsigned int		max_hw_discard_sectors;
383 	unsigned int		max_write_same_sectors;
384 	unsigned int		max_write_zeroes_sectors;
385 	unsigned int		discard_granularity;
386 	unsigned int		discard_alignment;
387 
388 	unsigned short		logical_block_size;
389 	unsigned short		max_segments;
390 	unsigned short		max_integrity_segments;
391 	unsigned short		max_discard_segments;
392 
393 	unsigned char		misaligned;
394 	unsigned char		discard_misaligned;
395 	unsigned char		cluster;
396 	unsigned char		raid_partial_stripes_expensive;
397 	enum blk_zoned_model	zoned;
398 };
399 
400 #ifdef CONFIG_BLK_DEV_ZONED
401 
402 struct blk_zone_report_hdr {
403 	unsigned int	nr_zones;
404 	u8		padding[60];
405 };
406 
407 extern int blkdev_report_zones(struct block_device *bdev,
408 			       sector_t sector, struct blk_zone *zones,
409 			       unsigned int *nr_zones, gfp_t gfp_mask);
410 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
411 			      sector_t nr_sectors, gfp_t gfp_mask);
412 
413 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
414 				     unsigned int cmd, unsigned long arg);
415 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
416 				    unsigned int cmd, unsigned long arg);
417 
418 #else /* CONFIG_BLK_DEV_ZONED */
419 
420 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
421 					    fmode_t mode, unsigned int cmd,
422 					    unsigned long arg)
423 {
424 	return -ENOTTY;
425 }
426 
427 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
428 					   fmode_t mode, unsigned int cmd,
429 					   unsigned long arg)
430 {
431 	return -ENOTTY;
432 }
433 
434 #endif /* CONFIG_BLK_DEV_ZONED */
435 
436 struct request_queue {
437 	/*
438 	 * Together with queue_head for cacheline sharing
439 	 */
440 	struct list_head	queue_head;
441 	struct request		*last_merge;
442 	struct elevator_queue	*elevator;
443 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
444 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
445 
446 	atomic_t		shared_hctx_restart;
447 
448 	struct blk_queue_stats	*stats;
449 	struct rq_wb		*rq_wb;
450 
451 	/*
452 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
453 	 * is used, root blkg allocates from @q->root_rl and all other
454 	 * blkgs from their own blkg->rl.  Which one to use should be
455 	 * determined using bio_request_list().
456 	 */
457 	struct request_list	root_rl;
458 
459 	request_fn_proc		*request_fn;
460 	make_request_fn		*make_request_fn;
461 	poll_q_fn		*poll_fn;
462 	prep_rq_fn		*prep_rq_fn;
463 	unprep_rq_fn		*unprep_rq_fn;
464 	softirq_done_fn		*softirq_done_fn;
465 	rq_timed_out_fn		*rq_timed_out_fn;
466 	dma_drain_needed_fn	*dma_drain_needed;
467 	lld_busy_fn		*lld_busy_fn;
468 	/* Called just after a request is allocated */
469 	init_rq_fn		*init_rq_fn;
470 	/* Called just before a request is freed */
471 	exit_rq_fn		*exit_rq_fn;
472 	/* Called from inside blk_get_request() */
473 	void (*initialize_rq_fn)(struct request *rq);
474 
475 	const struct blk_mq_ops	*mq_ops;
476 
477 	unsigned int		*mq_map;
478 
479 	/* sw queues */
480 	struct blk_mq_ctx __percpu	*queue_ctx;
481 	unsigned int		nr_queues;
482 
483 	unsigned int		queue_depth;
484 
485 	/* hw dispatch queues */
486 	struct blk_mq_hw_ctx	**queue_hw_ctx;
487 	unsigned int		nr_hw_queues;
488 
489 	/*
490 	 * Dispatch queue sorting
491 	 */
492 	sector_t		end_sector;
493 	struct request		*boundary_rq;
494 
495 	/*
496 	 * Delayed queue handling
497 	 */
498 	struct delayed_work	delay_work;
499 
500 	struct backing_dev_info	*backing_dev_info;
501 
502 	/*
503 	 * The queue owner gets to use this for whatever they like.
504 	 * ll_rw_blk doesn't touch it.
505 	 */
506 	void			*queuedata;
507 
508 	/*
509 	 * various queue flags, see QUEUE_* below
510 	 */
511 	unsigned long		queue_flags;
512 
513 	/*
514 	 * ida allocated id for this queue.  Used to index queues from
515 	 * ioctx.
516 	 */
517 	int			id;
518 
519 	/*
520 	 * queue needs bounce pages for pages above this limit
521 	 */
522 	gfp_t			bounce_gfp;
523 
524 	/*
525 	 * protects queue structures from reentrancy. ->__queue_lock should
526 	 * _never_ be used directly, it is queue private. always use
527 	 * ->queue_lock.
528 	 */
529 	spinlock_t		__queue_lock;
530 	spinlock_t		*queue_lock;
531 
532 	/*
533 	 * queue kobject
534 	 */
535 	struct kobject kobj;
536 
537 	/*
538 	 * mq queue kobject
539 	 */
540 	struct kobject mq_kobj;
541 
542 #ifdef  CONFIG_BLK_DEV_INTEGRITY
543 	struct blk_integrity integrity;
544 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
545 
546 #ifdef CONFIG_PM
547 	struct device		*dev;
548 	int			rpm_status;
549 	unsigned int		nr_pending;
550 #endif
551 
552 	/*
553 	 * queue settings
554 	 */
555 	unsigned long		nr_requests;	/* Max # of requests */
556 	unsigned int		nr_congestion_on;
557 	unsigned int		nr_congestion_off;
558 	unsigned int		nr_batching;
559 
560 	unsigned int		dma_drain_size;
561 	void			*dma_drain_buffer;
562 	unsigned int		dma_pad_mask;
563 	unsigned int		dma_alignment;
564 
565 	struct blk_queue_tag	*queue_tags;
566 	struct list_head	tag_busy_list;
567 
568 	unsigned int		nr_sorted;
569 	unsigned int		in_flight[2];
570 
571 	/*
572 	 * Number of active block driver functions for which blk_drain_queue()
573 	 * must wait. Must be incremented around functions that unlock the
574 	 * queue_lock internally, e.g. scsi_request_fn().
575 	 */
576 	unsigned int		request_fn_active;
577 
578 	unsigned int		rq_timeout;
579 	int			poll_nsec;
580 
581 	struct blk_stat_callback	*poll_cb;
582 	struct blk_rq_stat	poll_stat[BLK_MQ_POLL_STATS_BKTS];
583 
584 	struct timer_list	timeout;
585 	struct work_struct	timeout_work;
586 	struct list_head	timeout_list;
587 
588 	struct list_head	icq_list;
589 #ifdef CONFIG_BLK_CGROUP
590 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
591 	struct blkcg_gq		*root_blkg;
592 	struct list_head	blkg_list;
593 #endif
594 
595 	struct queue_limits	limits;
596 
597 	/*
598 	 * Zoned block device information for request dispatch control.
599 	 * nr_zones is the total number of zones of the device. This is always
600 	 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
601 	 * bits which indicates if a zone is conventional (bit clear) or
602 	 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
603 	 * bits which indicates if a zone is write locked, that is, if a write
604 	 * request targeting the zone was dispatched. All three fields are
605 	 * initialized by the low level device driver (e.g. scsi/sd.c).
606 	 * Stacking drivers (device mappers) may or may not initialize
607 	 * these fields.
608 	 */
609 	unsigned int		nr_zones;
610 	unsigned long		*seq_zones_bitmap;
611 	unsigned long		*seq_zones_wlock;
612 
613 	/*
614 	 * sg stuff
615 	 */
616 	unsigned int		sg_timeout;
617 	unsigned int		sg_reserved_size;
618 	int			node;
619 #ifdef CONFIG_BLK_DEV_IO_TRACE
620 	struct blk_trace	*blk_trace;
621 	struct mutex		blk_trace_mutex;
622 #endif
623 	/*
624 	 * for flush operations
625 	 */
626 	struct blk_flush_queue	*fq;
627 
628 	struct list_head	requeue_list;
629 	spinlock_t		requeue_lock;
630 	struct delayed_work	requeue_work;
631 
632 	struct mutex		sysfs_lock;
633 
634 	int			bypass_depth;
635 	atomic_t		mq_freeze_depth;
636 
637 #if defined(CONFIG_BLK_DEV_BSG)
638 	bsg_job_fn		*bsg_job_fn;
639 	struct bsg_class_device bsg_dev;
640 #endif
641 
642 #ifdef CONFIG_BLK_DEV_THROTTLING
643 	/* Throttle data */
644 	struct throtl_data *td;
645 #endif
646 	struct rcu_head		rcu_head;
647 	wait_queue_head_t	mq_freeze_wq;
648 	struct percpu_ref	q_usage_counter;
649 	struct list_head	all_q_node;
650 
651 	struct blk_mq_tag_set	*tag_set;
652 	struct list_head	tag_set_list;
653 	struct bio_set		*bio_split;
654 
655 #ifdef CONFIG_BLK_DEBUG_FS
656 	struct dentry		*debugfs_dir;
657 	struct dentry		*sched_debugfs_dir;
658 #endif
659 
660 	bool			mq_sysfs_init_done;
661 
662 	size_t			cmd_size;
663 	void			*rq_alloc_data;
664 
665 	struct work_struct	release_work;
666 
667 #define BLK_MAX_WRITE_HINTS	5
668 	u64			write_hints[BLK_MAX_WRITE_HINTS];
669 };
670 
671 #define QUEUE_FLAG_QUEUED	0	/* uses generic tag queueing */
672 #define QUEUE_FLAG_STOPPED	1	/* queue is stopped */
673 #define QUEUE_FLAG_DYING	2	/* queue being torn down */
674 #define QUEUE_FLAG_BYPASS	3	/* act as dumb FIFO queue */
675 #define QUEUE_FLAG_BIDI		4	/* queue supports bidi requests */
676 #define QUEUE_FLAG_NOMERGES     5	/* disable merge attempts */
677 #define QUEUE_FLAG_SAME_COMP	6	/* complete on same CPU-group */
678 #define QUEUE_FLAG_FAIL_IO	7	/* fake timeout */
679 #define QUEUE_FLAG_NONROT	9	/* non-rotational device (SSD) */
680 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
681 #define QUEUE_FLAG_IO_STAT     10	/* do IO stats */
682 #define QUEUE_FLAG_DISCARD     11	/* supports DISCARD */
683 #define QUEUE_FLAG_NOXMERGES   12	/* No extended merges */
684 #define QUEUE_FLAG_ADD_RANDOM  13	/* Contributes to random pool */
685 #define QUEUE_FLAG_SECERASE    14	/* supports secure erase */
686 #define QUEUE_FLAG_SAME_FORCE  15	/* force complete on same CPU */
687 #define QUEUE_FLAG_DEAD        16	/* queue tear-down finished */
688 #define QUEUE_FLAG_INIT_DONE   17	/* queue is initialized */
689 #define QUEUE_FLAG_NO_SG_MERGE 18	/* don't attempt to merge SG segments*/
690 #define QUEUE_FLAG_POLL	       19	/* IO polling enabled if set */
691 #define QUEUE_FLAG_WC	       20	/* Write back caching */
692 #define QUEUE_FLAG_FUA	       21	/* device supports FUA writes */
693 #define QUEUE_FLAG_FLUSH_NQ    22	/* flush not queueuable */
694 #define QUEUE_FLAG_DAX         23	/* device supports DAX */
695 #define QUEUE_FLAG_STATS       24	/* track rq completion times */
696 #define QUEUE_FLAG_POLL_STATS  25	/* collecting stats for hybrid polling */
697 #define QUEUE_FLAG_REGISTERED  26	/* queue has been registered to a disk */
698 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27	/* queue supports SCSI commands */
699 #define QUEUE_FLAG_QUIESCED    28	/* queue has been quiesced */
700 #define QUEUE_FLAG_PREEMPT_ONLY	29	/* only process REQ_PREEMPT requests */
701 
702 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
703 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
704 				 (1 << QUEUE_FLAG_ADD_RANDOM))
705 
706 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
707 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
708 				 (1 << QUEUE_FLAG_POLL))
709 
710 /*
711  * @q->queue_lock is set while a queue is being initialized. Since we know
712  * that no other threads access the queue object before @q->queue_lock has
713  * been set, it is safe to manipulate queue flags without holding the
714  * queue_lock if @q->queue_lock == NULL. See also blk_alloc_queue_node() and
715  * blk_init_allocated_queue().
716  */
717 static inline void queue_lockdep_assert_held(struct request_queue *q)
718 {
719 	if (q->queue_lock)
720 		lockdep_assert_held(q->queue_lock);
721 }
722 
723 static inline void queue_flag_set_unlocked(unsigned int flag,
724 					   struct request_queue *q)
725 {
726 	__set_bit(flag, &q->queue_flags);
727 }
728 
729 static inline int queue_flag_test_and_clear(unsigned int flag,
730 					    struct request_queue *q)
731 {
732 	queue_lockdep_assert_held(q);
733 
734 	if (test_bit(flag, &q->queue_flags)) {
735 		__clear_bit(flag, &q->queue_flags);
736 		return 1;
737 	}
738 
739 	return 0;
740 }
741 
742 static inline int queue_flag_test_and_set(unsigned int flag,
743 					  struct request_queue *q)
744 {
745 	queue_lockdep_assert_held(q);
746 
747 	if (!test_bit(flag, &q->queue_flags)) {
748 		__set_bit(flag, &q->queue_flags);
749 		return 0;
750 	}
751 
752 	return 1;
753 }
754 
755 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
756 {
757 	queue_lockdep_assert_held(q);
758 	__set_bit(flag, &q->queue_flags);
759 }
760 
761 static inline void queue_flag_clear_unlocked(unsigned int flag,
762 					     struct request_queue *q)
763 {
764 	__clear_bit(flag, &q->queue_flags);
765 }
766 
767 static inline int queue_in_flight(struct request_queue *q)
768 {
769 	return q->in_flight[0] + q->in_flight[1];
770 }
771 
772 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
773 {
774 	queue_lockdep_assert_held(q);
775 	__clear_bit(flag, &q->queue_flags);
776 }
777 
778 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
779 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
780 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
781 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
782 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
783 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
784 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
785 #define blk_queue_noxmerges(q)	\
786 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
787 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
788 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
789 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
790 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
791 #define blk_queue_secure_erase(q) \
792 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
793 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
794 #define blk_queue_scsi_passthrough(q)	\
795 	test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
796 
797 #define blk_noretry_request(rq) \
798 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
799 			     REQ_FAILFAST_DRIVER))
800 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
801 #define blk_queue_preempt_only(q)				\
802 	test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
803 
804 extern int blk_set_preempt_only(struct request_queue *q);
805 extern void blk_clear_preempt_only(struct request_queue *q);
806 
807 static inline bool blk_account_rq(struct request *rq)
808 {
809 	return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
810 }
811 
812 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
813 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
814 /* rq->queuelist of dequeued request must be list_empty() */
815 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
816 
817 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
818 
819 #define rq_data_dir(rq)		(op_is_write(req_op(rq)) ? WRITE : READ)
820 
821 /*
822  * Driver can handle struct request, if it either has an old style
823  * request_fn defined, or is blk-mq based.
824  */
825 static inline bool queue_is_rq_based(struct request_queue *q)
826 {
827 	return q->request_fn || q->mq_ops;
828 }
829 
830 static inline unsigned int blk_queue_cluster(struct request_queue *q)
831 {
832 	return q->limits.cluster;
833 }
834 
835 static inline enum blk_zoned_model
836 blk_queue_zoned_model(struct request_queue *q)
837 {
838 	return q->limits.zoned;
839 }
840 
841 static inline bool blk_queue_is_zoned(struct request_queue *q)
842 {
843 	switch (blk_queue_zoned_model(q)) {
844 	case BLK_ZONED_HA:
845 	case BLK_ZONED_HM:
846 		return true;
847 	default:
848 		return false;
849 	}
850 }
851 
852 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
853 {
854 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
855 }
856 
857 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
858 {
859 	return q->nr_zones;
860 }
861 
862 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
863 					     sector_t sector)
864 {
865 	if (!blk_queue_is_zoned(q))
866 		return 0;
867 	return sector >> ilog2(q->limits.chunk_sectors);
868 }
869 
870 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
871 					 sector_t sector)
872 {
873 	if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
874 		return false;
875 	return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
876 }
877 
878 static inline bool rq_is_sync(struct request *rq)
879 {
880 	return op_is_sync(rq->cmd_flags);
881 }
882 
883 static inline bool blk_rl_full(struct request_list *rl, bool sync)
884 {
885 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
886 
887 	return rl->flags & flag;
888 }
889 
890 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
891 {
892 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
893 
894 	rl->flags |= flag;
895 }
896 
897 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
898 {
899 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
900 
901 	rl->flags &= ~flag;
902 }
903 
904 static inline bool rq_mergeable(struct request *rq)
905 {
906 	if (blk_rq_is_passthrough(rq))
907 		return false;
908 
909 	if (req_op(rq) == REQ_OP_FLUSH)
910 		return false;
911 
912 	if (req_op(rq) == REQ_OP_WRITE_ZEROES)
913 		return false;
914 
915 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
916 		return false;
917 	if (rq->rq_flags & RQF_NOMERGE_FLAGS)
918 		return false;
919 
920 	return true;
921 }
922 
923 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
924 {
925 	if (bio_page(a) == bio_page(b) &&
926 	    bio_offset(a) == bio_offset(b))
927 		return true;
928 
929 	return false;
930 }
931 
932 static inline unsigned int blk_queue_depth(struct request_queue *q)
933 {
934 	if (q->queue_depth)
935 		return q->queue_depth;
936 
937 	return q->nr_requests;
938 }
939 
940 /*
941  * q->prep_rq_fn return values
942  */
943 enum {
944 	BLKPREP_OK,		/* serve it */
945 	BLKPREP_KILL,		/* fatal error, kill, return -EIO */
946 	BLKPREP_DEFER,		/* leave on queue */
947 	BLKPREP_INVALID,	/* invalid command, kill, return -EREMOTEIO */
948 };
949 
950 extern unsigned long blk_max_low_pfn, blk_max_pfn;
951 
952 /*
953  * standard bounce addresses:
954  *
955  * BLK_BOUNCE_HIGH	: bounce all highmem pages
956  * BLK_BOUNCE_ANY	: don't bounce anything
957  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
958  */
959 
960 #if BITS_PER_LONG == 32
961 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
962 #else
963 #define BLK_BOUNCE_HIGH		-1ULL
964 #endif
965 #define BLK_BOUNCE_ANY		(-1ULL)
966 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
967 
968 /*
969  * default timeout for SG_IO if none specified
970  */
971 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
972 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
973 
974 struct rq_map_data {
975 	struct page **pages;
976 	int page_order;
977 	int nr_entries;
978 	unsigned long offset;
979 	int null_mapped;
980 	int from_user;
981 };
982 
983 struct req_iterator {
984 	struct bvec_iter iter;
985 	struct bio *bio;
986 };
987 
988 /* This should not be used directly - use rq_for_each_segment */
989 #define for_each_bio(_bio)		\
990 	for (; _bio; _bio = _bio->bi_next)
991 #define __rq_for_each_bio(_bio, rq)	\
992 	if ((rq->bio))			\
993 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
994 
995 #define rq_for_each_segment(bvl, _rq, _iter)			\
996 	__rq_for_each_bio(_iter.bio, _rq)			\
997 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
998 
999 #define rq_iter_last(bvec, _iter)				\
1000 		(_iter.bio->bi_next == NULL &&			\
1001 		 bio_iter_last(bvec, _iter.iter))
1002 
1003 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1004 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
1005 #endif
1006 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1007 extern void rq_flush_dcache_pages(struct request *rq);
1008 #else
1009 static inline void rq_flush_dcache_pages(struct request *rq)
1010 {
1011 }
1012 #endif
1013 
1014 extern int blk_register_queue(struct gendisk *disk);
1015 extern void blk_unregister_queue(struct gendisk *disk);
1016 extern blk_qc_t generic_make_request(struct bio *bio);
1017 extern blk_qc_t direct_make_request(struct bio *bio);
1018 extern void blk_rq_init(struct request_queue *q, struct request *rq);
1019 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
1020 extern void blk_put_request(struct request *);
1021 extern void __blk_put_request(struct request_queue *, struct request *);
1022 extern struct request *blk_get_request_flags(struct request_queue *,
1023 					     unsigned int op,
1024 					     blk_mq_req_flags_t flags);
1025 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
1026 				       gfp_t gfp_mask);
1027 extern void blk_requeue_request(struct request_queue *, struct request *);
1028 extern int blk_lld_busy(struct request_queue *q);
1029 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
1030 			     struct bio_set *bs, gfp_t gfp_mask,
1031 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
1032 			     void *data);
1033 extern void blk_rq_unprep_clone(struct request *rq);
1034 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
1035 				     struct request *rq);
1036 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
1037 extern void blk_delay_queue(struct request_queue *, unsigned long);
1038 extern void blk_queue_split(struct request_queue *, struct bio **);
1039 extern void blk_recount_segments(struct request_queue *, struct bio *);
1040 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
1041 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
1042 			      unsigned int, void __user *);
1043 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
1044 			  unsigned int, void __user *);
1045 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
1046 			 struct scsi_ioctl_command __user *);
1047 
1048 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
1049 extern void blk_queue_exit(struct request_queue *q);
1050 extern void blk_start_queue(struct request_queue *q);
1051 extern void blk_start_queue_async(struct request_queue *q);
1052 extern void blk_stop_queue(struct request_queue *q);
1053 extern void blk_sync_queue(struct request_queue *q);
1054 extern void __blk_stop_queue(struct request_queue *q);
1055 extern void __blk_run_queue(struct request_queue *q);
1056 extern void __blk_run_queue_uncond(struct request_queue *q);
1057 extern void blk_run_queue(struct request_queue *);
1058 extern void blk_run_queue_async(struct request_queue *q);
1059 extern int blk_rq_map_user(struct request_queue *, struct request *,
1060 			   struct rq_map_data *, void __user *, unsigned long,
1061 			   gfp_t);
1062 extern int blk_rq_unmap_user(struct bio *);
1063 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1064 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1065 			       struct rq_map_data *, const struct iov_iter *,
1066 			       gfp_t);
1067 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1068 			  struct request *, int);
1069 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1070 				  struct request *, int, rq_end_io_fn *);
1071 
1072 int blk_status_to_errno(blk_status_t status);
1073 blk_status_t errno_to_blk_status(int errno);
1074 
1075 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1076 
1077 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1078 {
1079 	return bdev->bd_disk->queue;	/* this is never NULL */
1080 }
1081 
1082 /*
1083  * blk_rq_pos()			: the current sector
1084  * blk_rq_bytes()		: bytes left in the entire request
1085  * blk_rq_cur_bytes()		: bytes left in the current segment
1086  * blk_rq_err_bytes()		: bytes left till the next error boundary
1087  * blk_rq_sectors()		: sectors left in the entire request
1088  * blk_rq_cur_sectors()		: sectors left in the current segment
1089  */
1090 static inline sector_t blk_rq_pos(const struct request *rq)
1091 {
1092 	return rq->__sector;
1093 }
1094 
1095 static inline unsigned int blk_rq_bytes(const struct request *rq)
1096 {
1097 	return rq->__data_len;
1098 }
1099 
1100 static inline int blk_rq_cur_bytes(const struct request *rq)
1101 {
1102 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1103 }
1104 
1105 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1106 
1107 static inline unsigned int blk_rq_sectors(const struct request *rq)
1108 {
1109 	return blk_rq_bytes(rq) >> 9;
1110 }
1111 
1112 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1113 {
1114 	return blk_rq_cur_bytes(rq) >> 9;
1115 }
1116 
1117 static inline unsigned int blk_rq_zone_no(struct request *rq)
1118 {
1119 	return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1120 }
1121 
1122 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1123 {
1124 	return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1125 }
1126 
1127 /*
1128  * Some commands like WRITE SAME have a payload or data transfer size which
1129  * is different from the size of the request.  Any driver that supports such
1130  * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1131  * calculate the data transfer size.
1132  */
1133 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1134 {
1135 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1136 		return rq->special_vec.bv_len;
1137 	return blk_rq_bytes(rq);
1138 }
1139 
1140 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1141 						     int op)
1142 {
1143 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1144 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1145 
1146 	if (unlikely(op == REQ_OP_WRITE_SAME))
1147 		return q->limits.max_write_same_sectors;
1148 
1149 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
1150 		return q->limits.max_write_zeroes_sectors;
1151 
1152 	return q->limits.max_sectors;
1153 }
1154 
1155 /*
1156  * Return maximum size of a request at given offset. Only valid for
1157  * file system requests.
1158  */
1159 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1160 					       sector_t offset)
1161 {
1162 	if (!q->limits.chunk_sectors)
1163 		return q->limits.max_sectors;
1164 
1165 	return q->limits.chunk_sectors -
1166 			(offset & (q->limits.chunk_sectors - 1));
1167 }
1168 
1169 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1170 						  sector_t offset)
1171 {
1172 	struct request_queue *q = rq->q;
1173 
1174 	if (blk_rq_is_passthrough(rq))
1175 		return q->limits.max_hw_sectors;
1176 
1177 	if (!q->limits.chunk_sectors ||
1178 	    req_op(rq) == REQ_OP_DISCARD ||
1179 	    req_op(rq) == REQ_OP_SECURE_ERASE)
1180 		return blk_queue_get_max_sectors(q, req_op(rq));
1181 
1182 	return min(blk_max_size_offset(q, offset),
1183 			blk_queue_get_max_sectors(q, req_op(rq)));
1184 }
1185 
1186 static inline unsigned int blk_rq_count_bios(struct request *rq)
1187 {
1188 	unsigned int nr_bios = 0;
1189 	struct bio *bio;
1190 
1191 	__rq_for_each_bio(bio, rq)
1192 		nr_bios++;
1193 
1194 	return nr_bios;
1195 }
1196 
1197 /*
1198  * Request issue related functions.
1199  */
1200 extern struct request *blk_peek_request(struct request_queue *q);
1201 extern void blk_start_request(struct request *rq);
1202 extern struct request *blk_fetch_request(struct request_queue *q);
1203 
1204 void blk_steal_bios(struct bio_list *list, struct request *rq);
1205 
1206 /*
1207  * Request completion related functions.
1208  *
1209  * blk_update_request() completes given number of bytes and updates
1210  * the request without completing it.
1211  *
1212  * blk_end_request() and friends.  __blk_end_request() must be called
1213  * with the request queue spinlock acquired.
1214  *
1215  * Several drivers define their own end_request and call
1216  * blk_end_request() for parts of the original function.
1217  * This prevents code duplication in drivers.
1218  */
1219 extern bool blk_update_request(struct request *rq, blk_status_t error,
1220 			       unsigned int nr_bytes);
1221 extern void blk_finish_request(struct request *rq, blk_status_t error);
1222 extern bool blk_end_request(struct request *rq, blk_status_t error,
1223 			    unsigned int nr_bytes);
1224 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1225 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1226 			      unsigned int nr_bytes);
1227 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1228 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1229 
1230 extern void blk_complete_request(struct request *);
1231 extern void __blk_complete_request(struct request *);
1232 extern void blk_abort_request(struct request *);
1233 extern void blk_unprep_request(struct request *);
1234 
1235 /*
1236  * Access functions for manipulating queue properties
1237  */
1238 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1239 					spinlock_t *lock, int node_id);
1240 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1241 extern int blk_init_allocated_queue(struct request_queue *);
1242 extern void blk_cleanup_queue(struct request_queue *);
1243 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1244 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1245 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1246 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1247 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1248 extern void blk_queue_max_discard_segments(struct request_queue *,
1249 		unsigned short);
1250 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1251 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1252 		unsigned int max_discard_sectors);
1253 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1254 		unsigned int max_write_same_sectors);
1255 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1256 		unsigned int max_write_same_sectors);
1257 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1258 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1259 extern void blk_queue_alignment_offset(struct request_queue *q,
1260 				       unsigned int alignment);
1261 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1262 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1263 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1264 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1265 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1266 extern void blk_set_default_limits(struct queue_limits *lim);
1267 extern void blk_set_stacking_limits(struct queue_limits *lim);
1268 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1269 			    sector_t offset);
1270 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1271 			    sector_t offset);
1272 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1273 			      sector_t offset);
1274 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1275 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1276 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1277 extern int blk_queue_dma_drain(struct request_queue *q,
1278 			       dma_drain_needed_fn *dma_drain_needed,
1279 			       void *buf, unsigned int size);
1280 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1281 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1282 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1283 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1284 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1285 extern void blk_queue_dma_alignment(struct request_queue *, int);
1286 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1287 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1288 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1289 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1290 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1291 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1292 
1293 /*
1294  * Number of physical segments as sent to the device.
1295  *
1296  * Normally this is the number of discontiguous data segments sent by the
1297  * submitter.  But for data-less command like discard we might have no
1298  * actual data segments submitted, but the driver might have to add it's
1299  * own special payload.  In that case we still return 1 here so that this
1300  * special payload will be mapped.
1301  */
1302 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1303 {
1304 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1305 		return 1;
1306 	return rq->nr_phys_segments;
1307 }
1308 
1309 /*
1310  * Number of discard segments (or ranges) the driver needs to fill in.
1311  * Each discard bio merged into a request is counted as one segment.
1312  */
1313 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1314 {
1315 	return max_t(unsigned short, rq->nr_phys_segments, 1);
1316 }
1317 
1318 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1319 extern void blk_dump_rq_flags(struct request *, char *);
1320 extern long nr_blockdev_pages(void);
1321 
1322 bool __must_check blk_get_queue(struct request_queue *);
1323 struct request_queue *blk_alloc_queue(gfp_t);
1324 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1325 extern void blk_put_queue(struct request_queue *);
1326 extern void blk_set_queue_dying(struct request_queue *);
1327 
1328 /*
1329  * block layer runtime pm functions
1330  */
1331 #ifdef CONFIG_PM
1332 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1333 extern int blk_pre_runtime_suspend(struct request_queue *q);
1334 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1335 extern void blk_pre_runtime_resume(struct request_queue *q);
1336 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1337 extern void blk_set_runtime_active(struct request_queue *q);
1338 #else
1339 static inline void blk_pm_runtime_init(struct request_queue *q,
1340 	struct device *dev) {}
1341 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1342 {
1343 	return -ENOSYS;
1344 }
1345 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1346 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1347 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1348 static inline void blk_set_runtime_active(struct request_queue *q) {}
1349 #endif
1350 
1351 /*
1352  * blk_plug permits building a queue of related requests by holding the I/O
1353  * fragments for a short period. This allows merging of sequential requests
1354  * into single larger request. As the requests are moved from a per-task list to
1355  * the device's request_queue in a batch, this results in improved scalability
1356  * as the lock contention for request_queue lock is reduced.
1357  *
1358  * It is ok not to disable preemption when adding the request to the plug list
1359  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1360  * the plug list when the task sleeps by itself. For details, please see
1361  * schedule() where blk_schedule_flush_plug() is called.
1362  */
1363 struct blk_plug {
1364 	struct list_head list; /* requests */
1365 	struct list_head mq_list; /* blk-mq requests */
1366 	struct list_head cb_list; /* md requires an unplug callback */
1367 };
1368 #define BLK_MAX_REQUEST_COUNT 16
1369 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1370 
1371 struct blk_plug_cb;
1372 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1373 struct blk_plug_cb {
1374 	struct list_head list;
1375 	blk_plug_cb_fn callback;
1376 	void *data;
1377 };
1378 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1379 					     void *data, int size);
1380 extern void blk_start_plug(struct blk_plug *);
1381 extern void blk_finish_plug(struct blk_plug *);
1382 extern void blk_flush_plug_list(struct blk_plug *, bool);
1383 
1384 static inline void blk_flush_plug(struct task_struct *tsk)
1385 {
1386 	struct blk_plug *plug = tsk->plug;
1387 
1388 	if (plug)
1389 		blk_flush_plug_list(plug, false);
1390 }
1391 
1392 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1393 {
1394 	struct blk_plug *plug = tsk->plug;
1395 
1396 	if (plug)
1397 		blk_flush_plug_list(plug, true);
1398 }
1399 
1400 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1401 {
1402 	struct blk_plug *plug = tsk->plug;
1403 
1404 	return plug &&
1405 		(!list_empty(&plug->list) ||
1406 		 !list_empty(&plug->mq_list) ||
1407 		 !list_empty(&plug->cb_list));
1408 }
1409 
1410 /*
1411  * tag stuff
1412  */
1413 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1414 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1415 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1416 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1417 extern void blk_queue_free_tags(struct request_queue *);
1418 extern int blk_queue_resize_tags(struct request_queue *, int);
1419 extern void blk_queue_invalidate_tags(struct request_queue *);
1420 extern struct blk_queue_tag *blk_init_tags(int, int);
1421 extern void blk_free_tags(struct blk_queue_tag *);
1422 
1423 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1424 						int tag)
1425 {
1426 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1427 		return NULL;
1428 	return bqt->tag_index[tag];
1429 }
1430 
1431 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1432 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1433 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1434 
1435 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
1436 
1437 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1438 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1439 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1440 		sector_t nr_sects, gfp_t gfp_mask, int flags,
1441 		struct bio **biop);
1442 
1443 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1444 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1445 
1446 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1447 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1448 		unsigned flags);
1449 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1450 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1451 
1452 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1453 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1454 {
1455 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1456 				    nr_blocks << (sb->s_blocksize_bits - 9),
1457 				    gfp_mask, flags);
1458 }
1459 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1460 		sector_t nr_blocks, gfp_t gfp_mask)
1461 {
1462 	return blkdev_issue_zeroout(sb->s_bdev,
1463 				    block << (sb->s_blocksize_bits - 9),
1464 				    nr_blocks << (sb->s_blocksize_bits - 9),
1465 				    gfp_mask, 0);
1466 }
1467 
1468 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1469 
1470 enum blk_default_limits {
1471 	BLK_MAX_SEGMENTS	= 128,
1472 	BLK_SAFE_MAX_SECTORS	= 255,
1473 	BLK_DEF_MAX_SECTORS	= 2560,
1474 	BLK_MAX_SEGMENT_SIZE	= 65536,
1475 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1476 };
1477 
1478 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1479 
1480 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1481 {
1482 	return q->limits.seg_boundary_mask;
1483 }
1484 
1485 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1486 {
1487 	return q->limits.virt_boundary_mask;
1488 }
1489 
1490 static inline unsigned int queue_max_sectors(struct request_queue *q)
1491 {
1492 	return q->limits.max_sectors;
1493 }
1494 
1495 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1496 {
1497 	return q->limits.max_hw_sectors;
1498 }
1499 
1500 static inline unsigned short queue_max_segments(struct request_queue *q)
1501 {
1502 	return q->limits.max_segments;
1503 }
1504 
1505 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1506 {
1507 	return q->limits.max_discard_segments;
1508 }
1509 
1510 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1511 {
1512 	return q->limits.max_segment_size;
1513 }
1514 
1515 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1516 {
1517 	int retval = 512;
1518 
1519 	if (q && q->limits.logical_block_size)
1520 		retval = q->limits.logical_block_size;
1521 
1522 	return retval;
1523 }
1524 
1525 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1526 {
1527 	return queue_logical_block_size(bdev_get_queue(bdev));
1528 }
1529 
1530 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1531 {
1532 	return q->limits.physical_block_size;
1533 }
1534 
1535 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1536 {
1537 	return queue_physical_block_size(bdev_get_queue(bdev));
1538 }
1539 
1540 static inline unsigned int queue_io_min(struct request_queue *q)
1541 {
1542 	return q->limits.io_min;
1543 }
1544 
1545 static inline int bdev_io_min(struct block_device *bdev)
1546 {
1547 	return queue_io_min(bdev_get_queue(bdev));
1548 }
1549 
1550 static inline unsigned int queue_io_opt(struct request_queue *q)
1551 {
1552 	return q->limits.io_opt;
1553 }
1554 
1555 static inline int bdev_io_opt(struct block_device *bdev)
1556 {
1557 	return queue_io_opt(bdev_get_queue(bdev));
1558 }
1559 
1560 static inline int queue_alignment_offset(struct request_queue *q)
1561 {
1562 	if (q->limits.misaligned)
1563 		return -1;
1564 
1565 	return q->limits.alignment_offset;
1566 }
1567 
1568 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1569 {
1570 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1571 	unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1572 
1573 	return (granularity + lim->alignment_offset - alignment) % granularity;
1574 }
1575 
1576 static inline int bdev_alignment_offset(struct block_device *bdev)
1577 {
1578 	struct request_queue *q = bdev_get_queue(bdev);
1579 
1580 	if (q->limits.misaligned)
1581 		return -1;
1582 
1583 	if (bdev != bdev->bd_contains)
1584 		return bdev->bd_part->alignment_offset;
1585 
1586 	return q->limits.alignment_offset;
1587 }
1588 
1589 static inline int queue_discard_alignment(struct request_queue *q)
1590 {
1591 	if (q->limits.discard_misaligned)
1592 		return -1;
1593 
1594 	return q->limits.discard_alignment;
1595 }
1596 
1597 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1598 {
1599 	unsigned int alignment, granularity, offset;
1600 
1601 	if (!lim->max_discard_sectors)
1602 		return 0;
1603 
1604 	/* Why are these in bytes, not sectors? */
1605 	alignment = lim->discard_alignment >> 9;
1606 	granularity = lim->discard_granularity >> 9;
1607 	if (!granularity)
1608 		return 0;
1609 
1610 	/* Offset of the partition start in 'granularity' sectors */
1611 	offset = sector_div(sector, granularity);
1612 
1613 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1614 	offset = (granularity + alignment - offset) % granularity;
1615 
1616 	/* Turn it back into bytes, gaah */
1617 	return offset << 9;
1618 }
1619 
1620 static inline int bdev_discard_alignment(struct block_device *bdev)
1621 {
1622 	struct request_queue *q = bdev_get_queue(bdev);
1623 
1624 	if (bdev != bdev->bd_contains)
1625 		return bdev->bd_part->discard_alignment;
1626 
1627 	return q->limits.discard_alignment;
1628 }
1629 
1630 static inline unsigned int bdev_write_same(struct block_device *bdev)
1631 {
1632 	struct request_queue *q = bdev_get_queue(bdev);
1633 
1634 	if (q)
1635 		return q->limits.max_write_same_sectors;
1636 
1637 	return 0;
1638 }
1639 
1640 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1641 {
1642 	struct request_queue *q = bdev_get_queue(bdev);
1643 
1644 	if (q)
1645 		return q->limits.max_write_zeroes_sectors;
1646 
1647 	return 0;
1648 }
1649 
1650 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1651 {
1652 	struct request_queue *q = bdev_get_queue(bdev);
1653 
1654 	if (q)
1655 		return blk_queue_zoned_model(q);
1656 
1657 	return BLK_ZONED_NONE;
1658 }
1659 
1660 static inline bool bdev_is_zoned(struct block_device *bdev)
1661 {
1662 	struct request_queue *q = bdev_get_queue(bdev);
1663 
1664 	if (q)
1665 		return blk_queue_is_zoned(q);
1666 
1667 	return false;
1668 }
1669 
1670 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1671 {
1672 	struct request_queue *q = bdev_get_queue(bdev);
1673 
1674 	if (q)
1675 		return blk_queue_zone_sectors(q);
1676 	return 0;
1677 }
1678 
1679 static inline unsigned int bdev_nr_zones(struct block_device *bdev)
1680 {
1681 	struct request_queue *q = bdev_get_queue(bdev);
1682 
1683 	if (q)
1684 		return blk_queue_nr_zones(q);
1685 	return 0;
1686 }
1687 
1688 static inline int queue_dma_alignment(struct request_queue *q)
1689 {
1690 	return q ? q->dma_alignment : 511;
1691 }
1692 
1693 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1694 				 unsigned int len)
1695 {
1696 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1697 	return !(addr & alignment) && !(len & alignment);
1698 }
1699 
1700 /* assumes size > 256 */
1701 static inline unsigned int blksize_bits(unsigned int size)
1702 {
1703 	unsigned int bits = 8;
1704 	do {
1705 		bits++;
1706 		size >>= 1;
1707 	} while (size > 256);
1708 	return bits;
1709 }
1710 
1711 static inline unsigned int block_size(struct block_device *bdev)
1712 {
1713 	return bdev->bd_block_size;
1714 }
1715 
1716 static inline bool queue_flush_queueable(struct request_queue *q)
1717 {
1718 	return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1719 }
1720 
1721 typedef struct {struct page *v;} Sector;
1722 
1723 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1724 
1725 static inline void put_dev_sector(Sector p)
1726 {
1727 	put_page(p.v);
1728 }
1729 
1730 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1731 				struct bio_vec *bprv, unsigned int offset)
1732 {
1733 	return offset ||
1734 		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1735 }
1736 
1737 /*
1738  * Check if adding a bio_vec after bprv with offset would create a gap in
1739  * the SG list. Most drivers don't care about this, but some do.
1740  */
1741 static inline bool bvec_gap_to_prev(struct request_queue *q,
1742 				struct bio_vec *bprv, unsigned int offset)
1743 {
1744 	if (!queue_virt_boundary(q))
1745 		return false;
1746 	return __bvec_gap_to_prev(q, bprv, offset);
1747 }
1748 
1749 /*
1750  * Check if the two bvecs from two bios can be merged to one segment.
1751  * If yes, no need to check gap between the two bios since the 1st bio
1752  * and the 1st bvec in the 2nd bio can be handled in one segment.
1753  */
1754 static inline bool bios_segs_mergeable(struct request_queue *q,
1755 		struct bio *prev, struct bio_vec *prev_last_bv,
1756 		struct bio_vec *next_first_bv)
1757 {
1758 	if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1759 		return false;
1760 	if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1761 		return false;
1762 	if (prev->bi_seg_back_size + next_first_bv->bv_len >
1763 			queue_max_segment_size(q))
1764 		return false;
1765 	return true;
1766 }
1767 
1768 static inline bool bio_will_gap(struct request_queue *q,
1769 				struct request *prev_rq,
1770 				struct bio *prev,
1771 				struct bio *next)
1772 {
1773 	if (bio_has_data(prev) && queue_virt_boundary(q)) {
1774 		struct bio_vec pb, nb;
1775 
1776 		/*
1777 		 * don't merge if the 1st bio starts with non-zero
1778 		 * offset, otherwise it is quite difficult to respect
1779 		 * sg gap limit. We work hard to merge a huge number of small
1780 		 * single bios in case of mkfs.
1781 		 */
1782 		if (prev_rq)
1783 			bio_get_first_bvec(prev_rq->bio, &pb);
1784 		else
1785 			bio_get_first_bvec(prev, &pb);
1786 		if (pb.bv_offset)
1787 			return true;
1788 
1789 		/*
1790 		 * We don't need to worry about the situation that the
1791 		 * merged segment ends in unaligned virt boundary:
1792 		 *
1793 		 * - if 'pb' ends aligned, the merged segment ends aligned
1794 		 * - if 'pb' ends unaligned, the next bio must include
1795 		 *   one single bvec of 'nb', otherwise the 'nb' can't
1796 		 *   merge with 'pb'
1797 		 */
1798 		bio_get_last_bvec(prev, &pb);
1799 		bio_get_first_bvec(next, &nb);
1800 
1801 		if (!bios_segs_mergeable(q, prev, &pb, &nb))
1802 			return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1803 	}
1804 
1805 	return false;
1806 }
1807 
1808 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1809 {
1810 	return bio_will_gap(req->q, req, req->biotail, bio);
1811 }
1812 
1813 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1814 {
1815 	return bio_will_gap(req->q, NULL, bio, req->bio);
1816 }
1817 
1818 int kblockd_schedule_work(struct work_struct *work);
1819 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1820 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1821 
1822 #ifdef CONFIG_BLK_CGROUP
1823 /*
1824  * This should not be using sched_clock(). A real patch is in progress
1825  * to fix this up, until that is in place we need to disable preemption
1826  * around sched_clock() in this function and set_io_start_time_ns().
1827  */
1828 static inline void set_start_time_ns(struct request *req)
1829 {
1830 	preempt_disable();
1831 	req->start_time_ns = sched_clock();
1832 	preempt_enable();
1833 }
1834 
1835 static inline void set_io_start_time_ns(struct request *req)
1836 {
1837 	preempt_disable();
1838 	req->io_start_time_ns = sched_clock();
1839 	preempt_enable();
1840 }
1841 
1842 static inline uint64_t rq_start_time_ns(struct request *req)
1843 {
1844         return req->start_time_ns;
1845 }
1846 
1847 static inline uint64_t rq_io_start_time_ns(struct request *req)
1848 {
1849         return req->io_start_time_ns;
1850 }
1851 #else
1852 static inline void set_start_time_ns(struct request *req) {}
1853 static inline void set_io_start_time_ns(struct request *req) {}
1854 static inline uint64_t rq_start_time_ns(struct request *req)
1855 {
1856 	return 0;
1857 }
1858 static inline uint64_t rq_io_start_time_ns(struct request *req)
1859 {
1860 	return 0;
1861 }
1862 #endif
1863 
1864 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1865 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1866 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1867 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1868 
1869 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1870 
1871 enum blk_integrity_flags {
1872 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1873 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1874 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1875 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1876 };
1877 
1878 struct blk_integrity_iter {
1879 	void			*prot_buf;
1880 	void			*data_buf;
1881 	sector_t		seed;
1882 	unsigned int		data_size;
1883 	unsigned short		interval;
1884 	const char		*disk_name;
1885 };
1886 
1887 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1888 
1889 struct blk_integrity_profile {
1890 	integrity_processing_fn		*generate_fn;
1891 	integrity_processing_fn		*verify_fn;
1892 	const char			*name;
1893 };
1894 
1895 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1896 extern void blk_integrity_unregister(struct gendisk *);
1897 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1898 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1899 				   struct scatterlist *);
1900 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1901 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1902 				   struct request *);
1903 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1904 				    struct bio *);
1905 
1906 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1907 {
1908 	struct blk_integrity *bi = &disk->queue->integrity;
1909 
1910 	if (!bi->profile)
1911 		return NULL;
1912 
1913 	return bi;
1914 }
1915 
1916 static inline
1917 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1918 {
1919 	return blk_get_integrity(bdev->bd_disk);
1920 }
1921 
1922 static inline bool blk_integrity_rq(struct request *rq)
1923 {
1924 	return rq->cmd_flags & REQ_INTEGRITY;
1925 }
1926 
1927 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1928 						    unsigned int segs)
1929 {
1930 	q->limits.max_integrity_segments = segs;
1931 }
1932 
1933 static inline unsigned short
1934 queue_max_integrity_segments(struct request_queue *q)
1935 {
1936 	return q->limits.max_integrity_segments;
1937 }
1938 
1939 static inline bool integrity_req_gap_back_merge(struct request *req,
1940 						struct bio *next)
1941 {
1942 	struct bio_integrity_payload *bip = bio_integrity(req->bio);
1943 	struct bio_integrity_payload *bip_next = bio_integrity(next);
1944 
1945 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1946 				bip_next->bip_vec[0].bv_offset);
1947 }
1948 
1949 static inline bool integrity_req_gap_front_merge(struct request *req,
1950 						 struct bio *bio)
1951 {
1952 	struct bio_integrity_payload *bip = bio_integrity(bio);
1953 	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1954 
1955 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1956 				bip_next->bip_vec[0].bv_offset);
1957 }
1958 
1959 #else /* CONFIG_BLK_DEV_INTEGRITY */
1960 
1961 struct bio;
1962 struct block_device;
1963 struct gendisk;
1964 struct blk_integrity;
1965 
1966 static inline int blk_integrity_rq(struct request *rq)
1967 {
1968 	return 0;
1969 }
1970 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1971 					    struct bio *b)
1972 {
1973 	return 0;
1974 }
1975 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1976 					  struct bio *b,
1977 					  struct scatterlist *s)
1978 {
1979 	return 0;
1980 }
1981 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1982 {
1983 	return NULL;
1984 }
1985 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1986 {
1987 	return NULL;
1988 }
1989 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1990 {
1991 	return 0;
1992 }
1993 static inline void blk_integrity_register(struct gendisk *d,
1994 					 struct blk_integrity *b)
1995 {
1996 }
1997 static inline void blk_integrity_unregister(struct gendisk *d)
1998 {
1999 }
2000 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
2001 						    unsigned int segs)
2002 {
2003 }
2004 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
2005 {
2006 	return 0;
2007 }
2008 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
2009 					  struct request *r1,
2010 					  struct request *r2)
2011 {
2012 	return true;
2013 }
2014 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
2015 					   struct request *r,
2016 					   struct bio *b)
2017 {
2018 	return true;
2019 }
2020 
2021 static inline bool integrity_req_gap_back_merge(struct request *req,
2022 						struct bio *next)
2023 {
2024 	return false;
2025 }
2026 static inline bool integrity_req_gap_front_merge(struct request *req,
2027 						 struct bio *bio)
2028 {
2029 	return false;
2030 }
2031 
2032 #endif /* CONFIG_BLK_DEV_INTEGRITY */
2033 
2034 struct block_device_operations {
2035 	int (*open) (struct block_device *, fmode_t);
2036 	void (*release) (struct gendisk *, fmode_t);
2037 	int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
2038 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
2039 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
2040 	unsigned int (*check_events) (struct gendisk *disk,
2041 				      unsigned int clearing);
2042 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
2043 	int (*media_changed) (struct gendisk *);
2044 	void (*unlock_native_capacity) (struct gendisk *);
2045 	int (*revalidate_disk) (struct gendisk *);
2046 	int (*getgeo)(struct block_device *, struct hd_geometry *);
2047 	/* this callback is with swap_lock and sometimes page table lock held */
2048 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
2049 	struct module *owner;
2050 	const struct pr_ops *pr_ops;
2051 };
2052 
2053 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
2054 				 unsigned long);
2055 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
2056 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
2057 						struct writeback_control *);
2058 
2059 #ifdef CONFIG_BLK_DEV_ZONED
2060 bool blk_req_needs_zone_write_lock(struct request *rq);
2061 void __blk_req_zone_write_lock(struct request *rq);
2062 void __blk_req_zone_write_unlock(struct request *rq);
2063 
2064 static inline void blk_req_zone_write_lock(struct request *rq)
2065 {
2066 	if (blk_req_needs_zone_write_lock(rq))
2067 		__blk_req_zone_write_lock(rq);
2068 }
2069 
2070 static inline void blk_req_zone_write_unlock(struct request *rq)
2071 {
2072 	if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
2073 		__blk_req_zone_write_unlock(rq);
2074 }
2075 
2076 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2077 {
2078 	return rq->q->seq_zones_wlock &&
2079 		test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
2080 }
2081 
2082 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2083 {
2084 	if (!blk_req_needs_zone_write_lock(rq))
2085 		return true;
2086 	return !blk_req_zone_is_write_locked(rq);
2087 }
2088 #else
2089 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2090 {
2091 	return false;
2092 }
2093 
2094 static inline void blk_req_zone_write_lock(struct request *rq)
2095 {
2096 }
2097 
2098 static inline void blk_req_zone_write_unlock(struct request *rq)
2099 {
2100 }
2101 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2102 {
2103 	return false;
2104 }
2105 
2106 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2107 {
2108 	return true;
2109 }
2110 #endif /* CONFIG_BLK_DEV_ZONED */
2111 
2112 #else /* CONFIG_BLOCK */
2113 
2114 struct block_device;
2115 
2116 /*
2117  * stubs for when the block layer is configured out
2118  */
2119 #define buffer_heads_over_limit 0
2120 
2121 static inline long nr_blockdev_pages(void)
2122 {
2123 	return 0;
2124 }
2125 
2126 struct blk_plug {
2127 };
2128 
2129 static inline void blk_start_plug(struct blk_plug *plug)
2130 {
2131 }
2132 
2133 static inline void blk_finish_plug(struct blk_plug *plug)
2134 {
2135 }
2136 
2137 static inline void blk_flush_plug(struct task_struct *task)
2138 {
2139 }
2140 
2141 static inline void blk_schedule_flush_plug(struct task_struct *task)
2142 {
2143 }
2144 
2145 
2146 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2147 {
2148 	return false;
2149 }
2150 
2151 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2152 				     sector_t *error_sector)
2153 {
2154 	return 0;
2155 }
2156 
2157 #endif /* CONFIG_BLOCK */
2158 
2159 #endif
2160