xref: /linux-6.15/include/linux/blkdev.h (revision a4cdb556)
1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
3 
4 #include <linux/sched.h>
5 
6 #ifdef CONFIG_BLOCK
7 
8 #include <linux/major.h>
9 #include <linux/genhd.h>
10 #include <linux/list.h>
11 #include <linux/llist.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/bio.h>
19 #include <linux/stringify.h>
20 #include <linux/gfp.h>
21 #include <linux/bsg.h>
22 #include <linux/smp.h>
23 #include <linux/rcupdate.h>
24 #include <linux/percpu-refcount.h>
25 #include <linux/scatterlist.h>
26 
27 struct module;
28 struct scsi_ioctl_command;
29 
30 struct request_queue;
31 struct elevator_queue;
32 struct blk_trace;
33 struct request;
34 struct sg_io_hdr;
35 struct bsg_job;
36 struct blkcg_gq;
37 struct blk_flush_queue;
38 struct pr_ops;
39 
40 #define BLKDEV_MIN_RQ	4
41 #define BLKDEV_MAX_RQ	128	/* Default maximum */
42 
43 /*
44  * Maximum number of blkcg policies allowed to be registered concurrently.
45  * Defined here to simplify include dependency.
46  */
47 #define BLKCG_MAX_POLS		2
48 
49 struct request;
50 typedef void (rq_end_io_fn)(struct request *, int);
51 
52 #define BLK_RL_SYNCFULL		(1U << 0)
53 #define BLK_RL_ASYNCFULL	(1U << 1)
54 
55 struct request_list {
56 	struct request_queue	*q;	/* the queue this rl belongs to */
57 #ifdef CONFIG_BLK_CGROUP
58 	struct blkcg_gq		*blkg;	/* blkg this request pool belongs to */
59 #endif
60 	/*
61 	 * count[], starved[], and wait[] are indexed by
62 	 * BLK_RW_SYNC/BLK_RW_ASYNC
63 	 */
64 	int			count[2];
65 	int			starved[2];
66 	mempool_t		*rq_pool;
67 	wait_queue_head_t	wait[2];
68 	unsigned int		flags;
69 };
70 
71 /*
72  * request command types
73  */
74 enum rq_cmd_type_bits {
75 	REQ_TYPE_FS		= 1,	/* fs request */
76 	REQ_TYPE_BLOCK_PC,		/* scsi command */
77 	REQ_TYPE_DRV_PRIV,		/* driver defined types from here */
78 };
79 
80 #define BLK_MAX_CDB	16
81 
82 /*
83  * Try to put the fields that are referenced together in the same cacheline.
84  *
85  * If you modify this structure, make sure to update blk_rq_init() and
86  * especially blk_mq_rq_ctx_init() to take care of the added fields.
87  */
88 struct request {
89 	struct list_head queuelist;
90 	union {
91 		struct call_single_data csd;
92 		unsigned long fifo_time;
93 	};
94 
95 	struct request_queue *q;
96 	struct blk_mq_ctx *mq_ctx;
97 
98 	u64 cmd_flags;
99 	unsigned cmd_type;
100 	unsigned long atomic_flags;
101 
102 	int cpu;
103 
104 	/* the following two fields are internal, NEVER access directly */
105 	unsigned int __data_len;	/* total data len */
106 	sector_t __sector;		/* sector cursor */
107 
108 	struct bio *bio;
109 	struct bio *biotail;
110 
111 	/*
112 	 * The hash is used inside the scheduler, and killed once the
113 	 * request reaches the dispatch list. The ipi_list is only used
114 	 * to queue the request for softirq completion, which is long
115 	 * after the request has been unhashed (and even removed from
116 	 * the dispatch list).
117 	 */
118 	union {
119 		struct hlist_node hash;	/* merge hash */
120 		struct list_head ipi_list;
121 	};
122 
123 	/*
124 	 * The rb_node is only used inside the io scheduler, requests
125 	 * are pruned when moved to the dispatch queue. So let the
126 	 * completion_data share space with the rb_node.
127 	 */
128 	union {
129 		struct rb_node rb_node;	/* sort/lookup */
130 		void *completion_data;
131 	};
132 
133 	/*
134 	 * Three pointers are available for the IO schedulers, if they need
135 	 * more they have to dynamically allocate it.  Flush requests are
136 	 * never put on the IO scheduler. So let the flush fields share
137 	 * space with the elevator data.
138 	 */
139 	union {
140 		struct {
141 			struct io_cq		*icq;
142 			void			*priv[2];
143 		} elv;
144 
145 		struct {
146 			unsigned int		seq;
147 			struct list_head	list;
148 			rq_end_io_fn		*saved_end_io;
149 		} flush;
150 	};
151 
152 	struct gendisk *rq_disk;
153 	struct hd_struct *part;
154 	unsigned long start_time;
155 #ifdef CONFIG_BLK_CGROUP
156 	struct request_list *rl;		/* rl this rq is alloced from */
157 	unsigned long long start_time_ns;
158 	unsigned long long io_start_time_ns;    /* when passed to hardware */
159 #endif
160 	/* Number of scatter-gather DMA addr+len pairs after
161 	 * physical address coalescing is performed.
162 	 */
163 	unsigned short nr_phys_segments;
164 #if defined(CONFIG_BLK_DEV_INTEGRITY)
165 	unsigned short nr_integrity_segments;
166 #endif
167 
168 	unsigned short ioprio;
169 
170 	void *special;		/* opaque pointer available for LLD use */
171 
172 	int tag;
173 	int errors;
174 
175 	/*
176 	 * when request is used as a packet command carrier
177 	 */
178 	unsigned char __cmd[BLK_MAX_CDB];
179 	unsigned char *cmd;
180 	unsigned short cmd_len;
181 
182 	unsigned int extra_len;	/* length of alignment and padding */
183 	unsigned int sense_len;
184 	unsigned int resid_len;	/* residual count */
185 	void *sense;
186 
187 	unsigned long deadline;
188 	struct list_head timeout_list;
189 	unsigned int timeout;
190 	int retries;
191 
192 	/*
193 	 * completion callback.
194 	 */
195 	rq_end_io_fn *end_io;
196 	void *end_io_data;
197 
198 	/* for bidi */
199 	struct request *next_rq;
200 };
201 
202 static inline unsigned short req_get_ioprio(struct request *req)
203 {
204 	return req->ioprio;
205 }
206 
207 #include <linux/elevator.h>
208 
209 struct blk_queue_ctx;
210 
211 typedef void (request_fn_proc) (struct request_queue *q);
212 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
213 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
214 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
215 
216 struct bio_vec;
217 typedef void (softirq_done_fn)(struct request *);
218 typedef int (dma_drain_needed_fn)(struct request *);
219 typedef int (lld_busy_fn) (struct request_queue *q);
220 typedef int (bsg_job_fn) (struct bsg_job *);
221 
222 enum blk_eh_timer_return {
223 	BLK_EH_NOT_HANDLED,
224 	BLK_EH_HANDLED,
225 	BLK_EH_RESET_TIMER,
226 };
227 
228 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
229 
230 enum blk_queue_state {
231 	Queue_down,
232 	Queue_up,
233 };
234 
235 struct blk_queue_tag {
236 	struct request **tag_index;	/* map of busy tags */
237 	unsigned long *tag_map;		/* bit map of free/busy tags */
238 	int busy;			/* current depth */
239 	int max_depth;			/* what we will send to device */
240 	int real_max_depth;		/* what the array can hold */
241 	atomic_t refcnt;		/* map can be shared */
242 	int alloc_policy;		/* tag allocation policy */
243 	int next_tag;			/* next tag */
244 };
245 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
246 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
247 
248 #define BLK_SCSI_MAX_CMDS	(256)
249 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
250 
251 struct queue_limits {
252 	unsigned long		bounce_pfn;
253 	unsigned long		seg_boundary_mask;
254 	unsigned long		virt_boundary_mask;
255 
256 	unsigned int		max_hw_sectors;
257 	unsigned int		max_dev_sectors;
258 	unsigned int		chunk_sectors;
259 	unsigned int		max_sectors;
260 	unsigned int		max_segment_size;
261 	unsigned int		physical_block_size;
262 	unsigned int		alignment_offset;
263 	unsigned int		io_min;
264 	unsigned int		io_opt;
265 	unsigned int		max_discard_sectors;
266 	unsigned int		max_hw_discard_sectors;
267 	unsigned int		max_write_same_sectors;
268 	unsigned int		discard_granularity;
269 	unsigned int		discard_alignment;
270 
271 	unsigned short		logical_block_size;
272 	unsigned short		max_segments;
273 	unsigned short		max_integrity_segments;
274 
275 	unsigned char		misaligned;
276 	unsigned char		discard_misaligned;
277 	unsigned char		cluster;
278 	unsigned char		discard_zeroes_data;
279 	unsigned char		raid_partial_stripes_expensive;
280 };
281 
282 struct request_queue {
283 	/*
284 	 * Together with queue_head for cacheline sharing
285 	 */
286 	struct list_head	queue_head;
287 	struct request		*last_merge;
288 	struct elevator_queue	*elevator;
289 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
290 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
291 
292 	/*
293 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
294 	 * is used, root blkg allocates from @q->root_rl and all other
295 	 * blkgs from their own blkg->rl.  Which one to use should be
296 	 * determined using bio_request_list().
297 	 */
298 	struct request_list	root_rl;
299 
300 	request_fn_proc		*request_fn;
301 	make_request_fn		*make_request_fn;
302 	prep_rq_fn		*prep_rq_fn;
303 	unprep_rq_fn		*unprep_rq_fn;
304 	softirq_done_fn		*softirq_done_fn;
305 	rq_timed_out_fn		*rq_timed_out_fn;
306 	dma_drain_needed_fn	*dma_drain_needed;
307 	lld_busy_fn		*lld_busy_fn;
308 
309 	struct blk_mq_ops	*mq_ops;
310 
311 	unsigned int		*mq_map;
312 
313 	/* sw queues */
314 	struct blk_mq_ctx __percpu	*queue_ctx;
315 	unsigned int		nr_queues;
316 
317 	/* hw dispatch queues */
318 	struct blk_mq_hw_ctx	**queue_hw_ctx;
319 	unsigned int		nr_hw_queues;
320 
321 	/*
322 	 * Dispatch queue sorting
323 	 */
324 	sector_t		end_sector;
325 	struct request		*boundary_rq;
326 
327 	/*
328 	 * Delayed queue handling
329 	 */
330 	struct delayed_work	delay_work;
331 
332 	struct backing_dev_info	backing_dev_info;
333 
334 	/*
335 	 * The queue owner gets to use this for whatever they like.
336 	 * ll_rw_blk doesn't touch it.
337 	 */
338 	void			*queuedata;
339 
340 	/*
341 	 * various queue flags, see QUEUE_* below
342 	 */
343 	unsigned long		queue_flags;
344 
345 	/*
346 	 * ida allocated id for this queue.  Used to index queues from
347 	 * ioctx.
348 	 */
349 	int			id;
350 
351 	/*
352 	 * queue needs bounce pages for pages above this limit
353 	 */
354 	gfp_t			bounce_gfp;
355 
356 	/*
357 	 * protects queue structures from reentrancy. ->__queue_lock should
358 	 * _never_ be used directly, it is queue private. always use
359 	 * ->queue_lock.
360 	 */
361 	spinlock_t		__queue_lock;
362 	spinlock_t		*queue_lock;
363 
364 	/*
365 	 * queue kobject
366 	 */
367 	struct kobject kobj;
368 
369 	/*
370 	 * mq queue kobject
371 	 */
372 	struct kobject mq_kobj;
373 
374 #ifdef  CONFIG_BLK_DEV_INTEGRITY
375 	struct blk_integrity integrity;
376 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
377 
378 #ifdef CONFIG_PM
379 	struct device		*dev;
380 	int			rpm_status;
381 	unsigned int		nr_pending;
382 #endif
383 
384 	/*
385 	 * queue settings
386 	 */
387 	unsigned long		nr_requests;	/* Max # of requests */
388 	unsigned int		nr_congestion_on;
389 	unsigned int		nr_congestion_off;
390 	unsigned int		nr_batching;
391 
392 	unsigned int		dma_drain_size;
393 	void			*dma_drain_buffer;
394 	unsigned int		dma_pad_mask;
395 	unsigned int		dma_alignment;
396 
397 	struct blk_queue_tag	*queue_tags;
398 	struct list_head	tag_busy_list;
399 
400 	unsigned int		nr_sorted;
401 	unsigned int		in_flight[2];
402 	/*
403 	 * Number of active block driver functions for which blk_drain_queue()
404 	 * must wait. Must be incremented around functions that unlock the
405 	 * queue_lock internally, e.g. scsi_request_fn().
406 	 */
407 	unsigned int		request_fn_active;
408 
409 	unsigned int		rq_timeout;
410 	struct timer_list	timeout;
411 	struct list_head	timeout_list;
412 
413 	struct list_head	icq_list;
414 #ifdef CONFIG_BLK_CGROUP
415 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
416 	struct blkcg_gq		*root_blkg;
417 	struct list_head	blkg_list;
418 #endif
419 
420 	struct queue_limits	limits;
421 
422 	/*
423 	 * sg stuff
424 	 */
425 	unsigned int		sg_timeout;
426 	unsigned int		sg_reserved_size;
427 	int			node;
428 #ifdef CONFIG_BLK_DEV_IO_TRACE
429 	struct blk_trace	*blk_trace;
430 #endif
431 	/*
432 	 * for flush operations
433 	 */
434 	unsigned int		flush_flags;
435 	unsigned int		flush_not_queueable:1;
436 	struct blk_flush_queue	*fq;
437 
438 	struct list_head	requeue_list;
439 	spinlock_t		requeue_lock;
440 	struct work_struct	requeue_work;
441 
442 	struct mutex		sysfs_lock;
443 
444 	int			bypass_depth;
445 	atomic_t		mq_freeze_depth;
446 
447 #if defined(CONFIG_BLK_DEV_BSG)
448 	bsg_job_fn		*bsg_job_fn;
449 	int			bsg_job_size;
450 	struct bsg_class_device bsg_dev;
451 #endif
452 
453 #ifdef CONFIG_BLK_DEV_THROTTLING
454 	/* Throttle data */
455 	struct throtl_data *td;
456 #endif
457 	struct rcu_head		rcu_head;
458 	wait_queue_head_t	mq_freeze_wq;
459 	struct percpu_ref	q_usage_counter;
460 	struct list_head	all_q_node;
461 
462 	struct blk_mq_tag_set	*tag_set;
463 	struct list_head	tag_set_list;
464 	struct bio_set		*bio_split;
465 
466 	bool			mq_sysfs_init_done;
467 };
468 
469 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
470 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
471 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
472 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
473 #define QUEUE_FLAG_DYING	5	/* queue being torn down */
474 #define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
475 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
476 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
477 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
478 #define QUEUE_FLAG_FAIL_IO     10	/* fake timeout */
479 #define QUEUE_FLAG_STACKABLE   11	/* supports request stacking */
480 #define QUEUE_FLAG_NONROT      12	/* non-rotational device (SSD) */
481 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
482 #define QUEUE_FLAG_IO_STAT     13	/* do IO stats */
483 #define QUEUE_FLAG_DISCARD     14	/* supports DISCARD */
484 #define QUEUE_FLAG_NOXMERGES   15	/* No extended merges */
485 #define QUEUE_FLAG_ADD_RANDOM  16	/* Contributes to random pool */
486 #define QUEUE_FLAG_SECDISCARD  17	/* supports SECDISCARD */
487 #define QUEUE_FLAG_SAME_FORCE  18	/* force complete on same CPU */
488 #define QUEUE_FLAG_DEAD        19	/* queue tear-down finished */
489 #define QUEUE_FLAG_INIT_DONE   20	/* queue is initialized */
490 #define QUEUE_FLAG_NO_SG_MERGE 21	/* don't attempt to merge SG segments*/
491 #define QUEUE_FLAG_POLL	       22	/* IO polling enabled if set */
492 
493 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
494 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
495 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
496 				 (1 << QUEUE_FLAG_ADD_RANDOM))
497 
498 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
499 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
500 				 (1 << QUEUE_FLAG_SAME_COMP))
501 
502 static inline void queue_lockdep_assert_held(struct request_queue *q)
503 {
504 	if (q->queue_lock)
505 		lockdep_assert_held(q->queue_lock);
506 }
507 
508 static inline void queue_flag_set_unlocked(unsigned int flag,
509 					   struct request_queue *q)
510 {
511 	__set_bit(flag, &q->queue_flags);
512 }
513 
514 static inline int queue_flag_test_and_clear(unsigned int flag,
515 					    struct request_queue *q)
516 {
517 	queue_lockdep_assert_held(q);
518 
519 	if (test_bit(flag, &q->queue_flags)) {
520 		__clear_bit(flag, &q->queue_flags);
521 		return 1;
522 	}
523 
524 	return 0;
525 }
526 
527 static inline int queue_flag_test_and_set(unsigned int flag,
528 					  struct request_queue *q)
529 {
530 	queue_lockdep_assert_held(q);
531 
532 	if (!test_bit(flag, &q->queue_flags)) {
533 		__set_bit(flag, &q->queue_flags);
534 		return 0;
535 	}
536 
537 	return 1;
538 }
539 
540 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
541 {
542 	queue_lockdep_assert_held(q);
543 	__set_bit(flag, &q->queue_flags);
544 }
545 
546 static inline void queue_flag_clear_unlocked(unsigned int flag,
547 					     struct request_queue *q)
548 {
549 	__clear_bit(flag, &q->queue_flags);
550 }
551 
552 static inline int queue_in_flight(struct request_queue *q)
553 {
554 	return q->in_flight[0] + q->in_flight[1];
555 }
556 
557 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
558 {
559 	queue_lockdep_assert_held(q);
560 	__clear_bit(flag, &q->queue_flags);
561 }
562 
563 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
564 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
565 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
566 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
567 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
568 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
569 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
570 #define blk_queue_noxmerges(q)	\
571 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
572 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
573 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
574 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
575 #define blk_queue_stackable(q)	\
576 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
577 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
578 #define blk_queue_secdiscard(q)	(blk_queue_discard(q) && \
579 	test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
580 
581 #define blk_noretry_request(rq) \
582 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
583 			     REQ_FAILFAST_DRIVER))
584 
585 #define blk_account_rq(rq) \
586 	(((rq)->cmd_flags & REQ_STARTED) && \
587 	 ((rq)->cmd_type == REQ_TYPE_FS))
588 
589 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
590 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
591 /* rq->queuelist of dequeued request must be list_empty() */
592 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
593 
594 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
595 
596 #define rq_data_dir(rq)		((int)((rq)->cmd_flags & 1))
597 
598 /*
599  * Driver can handle struct request, if it either has an old style
600  * request_fn defined, or is blk-mq based.
601  */
602 static inline bool queue_is_rq_based(struct request_queue *q)
603 {
604 	return q->request_fn || q->mq_ops;
605 }
606 
607 static inline unsigned int blk_queue_cluster(struct request_queue *q)
608 {
609 	return q->limits.cluster;
610 }
611 
612 /*
613  * We regard a request as sync, if either a read or a sync write
614  */
615 static inline bool rw_is_sync(unsigned int rw_flags)
616 {
617 	return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
618 }
619 
620 static inline bool rq_is_sync(struct request *rq)
621 {
622 	return rw_is_sync(rq->cmd_flags);
623 }
624 
625 static inline bool blk_rl_full(struct request_list *rl, bool sync)
626 {
627 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
628 
629 	return rl->flags & flag;
630 }
631 
632 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
633 {
634 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
635 
636 	rl->flags |= flag;
637 }
638 
639 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
640 {
641 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
642 
643 	rl->flags &= ~flag;
644 }
645 
646 static inline bool rq_mergeable(struct request *rq)
647 {
648 	if (rq->cmd_type != REQ_TYPE_FS)
649 		return false;
650 
651 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
652 		return false;
653 
654 	return true;
655 }
656 
657 static inline bool blk_check_merge_flags(unsigned int flags1,
658 					 unsigned int flags2)
659 {
660 	if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
661 		return false;
662 
663 	if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
664 		return false;
665 
666 	if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
667 		return false;
668 
669 	return true;
670 }
671 
672 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
673 {
674 	if (bio_data(a) == bio_data(b))
675 		return true;
676 
677 	return false;
678 }
679 
680 /*
681  * q->prep_rq_fn return values
682  */
683 #define BLKPREP_OK		0	/* serve it */
684 #define BLKPREP_KILL		1	/* fatal error, kill */
685 #define BLKPREP_DEFER		2	/* leave on queue */
686 
687 extern unsigned long blk_max_low_pfn, blk_max_pfn;
688 
689 /*
690  * standard bounce addresses:
691  *
692  * BLK_BOUNCE_HIGH	: bounce all highmem pages
693  * BLK_BOUNCE_ANY	: don't bounce anything
694  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
695  */
696 
697 #if BITS_PER_LONG == 32
698 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
699 #else
700 #define BLK_BOUNCE_HIGH		-1ULL
701 #endif
702 #define BLK_BOUNCE_ANY		(-1ULL)
703 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
704 
705 /*
706  * default timeout for SG_IO if none specified
707  */
708 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
709 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
710 
711 #ifdef CONFIG_BOUNCE
712 extern int init_emergency_isa_pool(void);
713 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
714 #else
715 static inline int init_emergency_isa_pool(void)
716 {
717 	return 0;
718 }
719 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
720 {
721 }
722 #endif /* CONFIG_MMU */
723 
724 struct rq_map_data {
725 	struct page **pages;
726 	int page_order;
727 	int nr_entries;
728 	unsigned long offset;
729 	int null_mapped;
730 	int from_user;
731 };
732 
733 struct req_iterator {
734 	struct bvec_iter iter;
735 	struct bio *bio;
736 };
737 
738 /* This should not be used directly - use rq_for_each_segment */
739 #define for_each_bio(_bio)		\
740 	for (; _bio; _bio = _bio->bi_next)
741 #define __rq_for_each_bio(_bio, rq)	\
742 	if ((rq->bio))			\
743 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
744 
745 #define rq_for_each_segment(bvl, _rq, _iter)			\
746 	__rq_for_each_bio(_iter.bio, _rq)			\
747 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
748 
749 #define rq_iter_last(bvec, _iter)				\
750 		(_iter.bio->bi_next == NULL &&			\
751 		 bio_iter_last(bvec, _iter.iter))
752 
753 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
754 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
755 #endif
756 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
757 extern void rq_flush_dcache_pages(struct request *rq);
758 #else
759 static inline void rq_flush_dcache_pages(struct request *rq)
760 {
761 }
762 #endif
763 
764 extern int blk_register_queue(struct gendisk *disk);
765 extern void blk_unregister_queue(struct gendisk *disk);
766 extern blk_qc_t generic_make_request(struct bio *bio);
767 extern void blk_rq_init(struct request_queue *q, struct request *rq);
768 extern void blk_put_request(struct request *);
769 extern void __blk_put_request(struct request_queue *, struct request *);
770 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
771 extern struct request *blk_make_request(struct request_queue *, struct bio *,
772 					gfp_t);
773 extern void blk_rq_set_block_pc(struct request *);
774 extern void blk_requeue_request(struct request_queue *, struct request *);
775 extern void blk_add_request_payload(struct request *rq, struct page *page,
776 		unsigned int len);
777 extern int blk_lld_busy(struct request_queue *q);
778 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
779 			     struct bio_set *bs, gfp_t gfp_mask,
780 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
781 			     void *data);
782 extern void blk_rq_unprep_clone(struct request *rq);
783 extern int blk_insert_cloned_request(struct request_queue *q,
784 				     struct request *rq);
785 extern void blk_delay_queue(struct request_queue *, unsigned long);
786 extern void blk_queue_split(struct request_queue *, struct bio **,
787 			    struct bio_set *);
788 extern void blk_recount_segments(struct request_queue *, struct bio *);
789 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
790 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
791 			      unsigned int, void __user *);
792 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
793 			  unsigned int, void __user *);
794 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
795 			 struct scsi_ioctl_command __user *);
796 
797 extern int blk_queue_enter(struct request_queue *q, gfp_t gfp);
798 extern void blk_queue_exit(struct request_queue *q);
799 extern void blk_start_queue(struct request_queue *q);
800 extern void blk_stop_queue(struct request_queue *q);
801 extern void blk_sync_queue(struct request_queue *q);
802 extern void __blk_stop_queue(struct request_queue *q);
803 extern void __blk_run_queue(struct request_queue *q);
804 extern void __blk_run_queue_uncond(struct request_queue *q);
805 extern void blk_run_queue(struct request_queue *);
806 extern void blk_run_queue_async(struct request_queue *q);
807 extern int blk_rq_map_user(struct request_queue *, struct request *,
808 			   struct rq_map_data *, void __user *, unsigned long,
809 			   gfp_t);
810 extern int blk_rq_unmap_user(struct bio *);
811 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
812 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
813 			       struct rq_map_data *, const struct iov_iter *,
814 			       gfp_t);
815 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
816 			  struct request *, int);
817 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
818 				  struct request *, int, rq_end_io_fn *);
819 
820 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
821 
822 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
823 {
824 	return bdev->bd_disk->queue;	/* this is never NULL */
825 }
826 
827 /*
828  * blk_rq_pos()			: the current sector
829  * blk_rq_bytes()		: bytes left in the entire request
830  * blk_rq_cur_bytes()		: bytes left in the current segment
831  * blk_rq_err_bytes()		: bytes left till the next error boundary
832  * blk_rq_sectors()		: sectors left in the entire request
833  * blk_rq_cur_sectors()		: sectors left in the current segment
834  */
835 static inline sector_t blk_rq_pos(const struct request *rq)
836 {
837 	return rq->__sector;
838 }
839 
840 static inline unsigned int blk_rq_bytes(const struct request *rq)
841 {
842 	return rq->__data_len;
843 }
844 
845 static inline int blk_rq_cur_bytes(const struct request *rq)
846 {
847 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
848 }
849 
850 extern unsigned int blk_rq_err_bytes(const struct request *rq);
851 
852 static inline unsigned int blk_rq_sectors(const struct request *rq)
853 {
854 	return blk_rq_bytes(rq) >> 9;
855 }
856 
857 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
858 {
859 	return blk_rq_cur_bytes(rq) >> 9;
860 }
861 
862 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
863 						     unsigned int cmd_flags)
864 {
865 	if (unlikely(cmd_flags & REQ_DISCARD))
866 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
867 
868 	if (unlikely(cmd_flags & REQ_WRITE_SAME))
869 		return q->limits.max_write_same_sectors;
870 
871 	return q->limits.max_sectors;
872 }
873 
874 /*
875  * Return maximum size of a request at given offset. Only valid for
876  * file system requests.
877  */
878 static inline unsigned int blk_max_size_offset(struct request_queue *q,
879 					       sector_t offset)
880 {
881 	if (!q->limits.chunk_sectors)
882 		return q->limits.max_sectors;
883 
884 	return q->limits.chunk_sectors -
885 			(offset & (q->limits.chunk_sectors - 1));
886 }
887 
888 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
889 {
890 	struct request_queue *q = rq->q;
891 
892 	if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
893 		return q->limits.max_hw_sectors;
894 
895 	if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
896 		return blk_queue_get_max_sectors(q, rq->cmd_flags);
897 
898 	return min(blk_max_size_offset(q, blk_rq_pos(rq)),
899 			blk_queue_get_max_sectors(q, rq->cmd_flags));
900 }
901 
902 static inline unsigned int blk_rq_count_bios(struct request *rq)
903 {
904 	unsigned int nr_bios = 0;
905 	struct bio *bio;
906 
907 	__rq_for_each_bio(bio, rq)
908 		nr_bios++;
909 
910 	return nr_bios;
911 }
912 
913 /*
914  * Request issue related functions.
915  */
916 extern struct request *blk_peek_request(struct request_queue *q);
917 extern void blk_start_request(struct request *rq);
918 extern struct request *blk_fetch_request(struct request_queue *q);
919 
920 /*
921  * Request completion related functions.
922  *
923  * blk_update_request() completes given number of bytes and updates
924  * the request without completing it.
925  *
926  * blk_end_request() and friends.  __blk_end_request() must be called
927  * with the request queue spinlock acquired.
928  *
929  * Several drivers define their own end_request and call
930  * blk_end_request() for parts of the original function.
931  * This prevents code duplication in drivers.
932  */
933 extern bool blk_update_request(struct request *rq, int error,
934 			       unsigned int nr_bytes);
935 extern void blk_finish_request(struct request *rq, int error);
936 extern bool blk_end_request(struct request *rq, int error,
937 			    unsigned int nr_bytes);
938 extern void blk_end_request_all(struct request *rq, int error);
939 extern bool blk_end_request_cur(struct request *rq, int error);
940 extern bool blk_end_request_err(struct request *rq, int error);
941 extern bool __blk_end_request(struct request *rq, int error,
942 			      unsigned int nr_bytes);
943 extern void __blk_end_request_all(struct request *rq, int error);
944 extern bool __blk_end_request_cur(struct request *rq, int error);
945 extern bool __blk_end_request_err(struct request *rq, int error);
946 
947 extern void blk_complete_request(struct request *);
948 extern void __blk_complete_request(struct request *);
949 extern void blk_abort_request(struct request *);
950 extern void blk_unprep_request(struct request *);
951 
952 /*
953  * Access functions for manipulating queue properties
954  */
955 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
956 					spinlock_t *lock, int node_id);
957 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
958 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
959 						      request_fn_proc *, spinlock_t *);
960 extern void blk_cleanup_queue(struct request_queue *);
961 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
962 extern void blk_queue_bounce_limit(struct request_queue *, u64);
963 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
964 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
965 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
966 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
967 extern void blk_queue_max_discard_sectors(struct request_queue *q,
968 		unsigned int max_discard_sectors);
969 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
970 		unsigned int max_write_same_sectors);
971 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
972 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
973 extern void blk_queue_alignment_offset(struct request_queue *q,
974 				       unsigned int alignment);
975 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
976 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
977 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
978 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
979 extern void blk_set_default_limits(struct queue_limits *lim);
980 extern void blk_set_stacking_limits(struct queue_limits *lim);
981 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
982 			    sector_t offset);
983 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
984 			    sector_t offset);
985 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
986 			      sector_t offset);
987 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
988 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
989 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
990 extern int blk_queue_dma_drain(struct request_queue *q,
991 			       dma_drain_needed_fn *dma_drain_needed,
992 			       void *buf, unsigned int size);
993 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
994 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
995 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
996 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
997 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
998 extern void blk_queue_dma_alignment(struct request_queue *, int);
999 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1000 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1001 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1002 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1003 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1004 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1005 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1006 
1007 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1008 extern void blk_dump_rq_flags(struct request *, char *);
1009 extern long nr_blockdev_pages(void);
1010 
1011 bool __must_check blk_get_queue(struct request_queue *);
1012 struct request_queue *blk_alloc_queue(gfp_t);
1013 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1014 extern void blk_put_queue(struct request_queue *);
1015 extern void blk_set_queue_dying(struct request_queue *);
1016 
1017 /*
1018  * block layer runtime pm functions
1019  */
1020 #ifdef CONFIG_PM
1021 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1022 extern int blk_pre_runtime_suspend(struct request_queue *q);
1023 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1024 extern void blk_pre_runtime_resume(struct request_queue *q);
1025 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1026 #else
1027 static inline void blk_pm_runtime_init(struct request_queue *q,
1028 	struct device *dev) {}
1029 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1030 {
1031 	return -ENOSYS;
1032 }
1033 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1034 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1035 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1036 #endif
1037 
1038 /*
1039  * blk_plug permits building a queue of related requests by holding the I/O
1040  * fragments for a short period. This allows merging of sequential requests
1041  * into single larger request. As the requests are moved from a per-task list to
1042  * the device's request_queue in a batch, this results in improved scalability
1043  * as the lock contention for request_queue lock is reduced.
1044  *
1045  * It is ok not to disable preemption when adding the request to the plug list
1046  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1047  * the plug list when the task sleeps by itself. For details, please see
1048  * schedule() where blk_schedule_flush_plug() is called.
1049  */
1050 struct blk_plug {
1051 	struct list_head list; /* requests */
1052 	struct list_head mq_list; /* blk-mq requests */
1053 	struct list_head cb_list; /* md requires an unplug callback */
1054 };
1055 #define BLK_MAX_REQUEST_COUNT 16
1056 
1057 struct blk_plug_cb;
1058 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1059 struct blk_plug_cb {
1060 	struct list_head list;
1061 	blk_plug_cb_fn callback;
1062 	void *data;
1063 };
1064 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1065 					     void *data, int size);
1066 extern void blk_start_plug(struct blk_plug *);
1067 extern void blk_finish_plug(struct blk_plug *);
1068 extern void blk_flush_plug_list(struct blk_plug *, bool);
1069 
1070 static inline void blk_flush_plug(struct task_struct *tsk)
1071 {
1072 	struct blk_plug *plug = tsk->plug;
1073 
1074 	if (plug)
1075 		blk_flush_plug_list(plug, false);
1076 }
1077 
1078 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1079 {
1080 	struct blk_plug *plug = tsk->plug;
1081 
1082 	if (plug)
1083 		blk_flush_plug_list(plug, true);
1084 }
1085 
1086 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1087 {
1088 	struct blk_plug *plug = tsk->plug;
1089 
1090 	return plug &&
1091 		(!list_empty(&plug->list) ||
1092 		 !list_empty(&plug->mq_list) ||
1093 		 !list_empty(&plug->cb_list));
1094 }
1095 
1096 /*
1097  * tag stuff
1098  */
1099 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1100 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1101 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1102 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1103 extern void blk_queue_free_tags(struct request_queue *);
1104 extern int blk_queue_resize_tags(struct request_queue *, int);
1105 extern void blk_queue_invalidate_tags(struct request_queue *);
1106 extern struct blk_queue_tag *blk_init_tags(int, int);
1107 extern void blk_free_tags(struct blk_queue_tag *);
1108 
1109 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1110 						int tag)
1111 {
1112 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1113 		return NULL;
1114 	return bqt->tag_index[tag];
1115 }
1116 
1117 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1118 
1119 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1120 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1121 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1122 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1123 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1124 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1125 		sector_t nr_sects, gfp_t gfp_mask, bool discard);
1126 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1127 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1128 {
1129 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1130 				    nr_blocks << (sb->s_blocksize_bits - 9),
1131 				    gfp_mask, flags);
1132 }
1133 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1134 		sector_t nr_blocks, gfp_t gfp_mask)
1135 {
1136 	return blkdev_issue_zeroout(sb->s_bdev,
1137 				    block << (sb->s_blocksize_bits - 9),
1138 				    nr_blocks << (sb->s_blocksize_bits - 9),
1139 				    gfp_mask, true);
1140 }
1141 
1142 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1143 
1144 enum blk_default_limits {
1145 	BLK_MAX_SEGMENTS	= 128,
1146 	BLK_SAFE_MAX_SECTORS	= 255,
1147 	BLK_DEF_MAX_SECTORS	= 2560,
1148 	BLK_MAX_SEGMENT_SIZE	= 65536,
1149 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1150 };
1151 
1152 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1153 
1154 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1155 {
1156 	return q->limits.bounce_pfn;
1157 }
1158 
1159 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1160 {
1161 	return q->limits.seg_boundary_mask;
1162 }
1163 
1164 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1165 {
1166 	return q->limits.virt_boundary_mask;
1167 }
1168 
1169 static inline unsigned int queue_max_sectors(struct request_queue *q)
1170 {
1171 	return q->limits.max_sectors;
1172 }
1173 
1174 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1175 {
1176 	return q->limits.max_hw_sectors;
1177 }
1178 
1179 static inline unsigned short queue_max_segments(struct request_queue *q)
1180 {
1181 	return q->limits.max_segments;
1182 }
1183 
1184 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1185 {
1186 	return q->limits.max_segment_size;
1187 }
1188 
1189 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1190 {
1191 	int retval = 512;
1192 
1193 	if (q && q->limits.logical_block_size)
1194 		retval = q->limits.logical_block_size;
1195 
1196 	return retval;
1197 }
1198 
1199 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1200 {
1201 	return queue_logical_block_size(bdev_get_queue(bdev));
1202 }
1203 
1204 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1205 {
1206 	return q->limits.physical_block_size;
1207 }
1208 
1209 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1210 {
1211 	return queue_physical_block_size(bdev_get_queue(bdev));
1212 }
1213 
1214 static inline unsigned int queue_io_min(struct request_queue *q)
1215 {
1216 	return q->limits.io_min;
1217 }
1218 
1219 static inline int bdev_io_min(struct block_device *bdev)
1220 {
1221 	return queue_io_min(bdev_get_queue(bdev));
1222 }
1223 
1224 static inline unsigned int queue_io_opt(struct request_queue *q)
1225 {
1226 	return q->limits.io_opt;
1227 }
1228 
1229 static inline int bdev_io_opt(struct block_device *bdev)
1230 {
1231 	return queue_io_opt(bdev_get_queue(bdev));
1232 }
1233 
1234 static inline int queue_alignment_offset(struct request_queue *q)
1235 {
1236 	if (q->limits.misaligned)
1237 		return -1;
1238 
1239 	return q->limits.alignment_offset;
1240 }
1241 
1242 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1243 {
1244 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1245 	unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1246 
1247 	return (granularity + lim->alignment_offset - alignment) % granularity;
1248 }
1249 
1250 static inline int bdev_alignment_offset(struct block_device *bdev)
1251 {
1252 	struct request_queue *q = bdev_get_queue(bdev);
1253 
1254 	if (q->limits.misaligned)
1255 		return -1;
1256 
1257 	if (bdev != bdev->bd_contains)
1258 		return bdev->bd_part->alignment_offset;
1259 
1260 	return q->limits.alignment_offset;
1261 }
1262 
1263 static inline int queue_discard_alignment(struct request_queue *q)
1264 {
1265 	if (q->limits.discard_misaligned)
1266 		return -1;
1267 
1268 	return q->limits.discard_alignment;
1269 }
1270 
1271 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1272 {
1273 	unsigned int alignment, granularity, offset;
1274 
1275 	if (!lim->max_discard_sectors)
1276 		return 0;
1277 
1278 	/* Why are these in bytes, not sectors? */
1279 	alignment = lim->discard_alignment >> 9;
1280 	granularity = lim->discard_granularity >> 9;
1281 	if (!granularity)
1282 		return 0;
1283 
1284 	/* Offset of the partition start in 'granularity' sectors */
1285 	offset = sector_div(sector, granularity);
1286 
1287 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1288 	offset = (granularity + alignment - offset) % granularity;
1289 
1290 	/* Turn it back into bytes, gaah */
1291 	return offset << 9;
1292 }
1293 
1294 static inline int bdev_discard_alignment(struct block_device *bdev)
1295 {
1296 	struct request_queue *q = bdev_get_queue(bdev);
1297 
1298 	if (bdev != bdev->bd_contains)
1299 		return bdev->bd_part->discard_alignment;
1300 
1301 	return q->limits.discard_alignment;
1302 }
1303 
1304 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1305 {
1306 	if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1307 		return 1;
1308 
1309 	return 0;
1310 }
1311 
1312 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1313 {
1314 	return queue_discard_zeroes_data(bdev_get_queue(bdev));
1315 }
1316 
1317 static inline unsigned int bdev_write_same(struct block_device *bdev)
1318 {
1319 	struct request_queue *q = bdev_get_queue(bdev);
1320 
1321 	if (q)
1322 		return q->limits.max_write_same_sectors;
1323 
1324 	return 0;
1325 }
1326 
1327 static inline int queue_dma_alignment(struct request_queue *q)
1328 {
1329 	return q ? q->dma_alignment : 511;
1330 }
1331 
1332 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1333 				 unsigned int len)
1334 {
1335 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1336 	return !(addr & alignment) && !(len & alignment);
1337 }
1338 
1339 /* assumes size > 256 */
1340 static inline unsigned int blksize_bits(unsigned int size)
1341 {
1342 	unsigned int bits = 8;
1343 	do {
1344 		bits++;
1345 		size >>= 1;
1346 	} while (size > 256);
1347 	return bits;
1348 }
1349 
1350 static inline unsigned int block_size(struct block_device *bdev)
1351 {
1352 	return bdev->bd_block_size;
1353 }
1354 
1355 static inline bool queue_flush_queueable(struct request_queue *q)
1356 {
1357 	return !q->flush_not_queueable;
1358 }
1359 
1360 typedef struct {struct page *v;} Sector;
1361 
1362 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1363 
1364 static inline void put_dev_sector(Sector p)
1365 {
1366 	page_cache_release(p.v);
1367 }
1368 
1369 /*
1370  * Check if adding a bio_vec after bprv with offset would create a gap in
1371  * the SG list. Most drivers don't care about this, but some do.
1372  */
1373 static inline bool bvec_gap_to_prev(struct request_queue *q,
1374 				struct bio_vec *bprv, unsigned int offset)
1375 {
1376 	if (!queue_virt_boundary(q))
1377 		return false;
1378 	return offset ||
1379 		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1380 }
1381 
1382 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1383 			 struct bio *next)
1384 {
1385 	if (!bio_has_data(prev))
1386 		return false;
1387 
1388 	return bvec_gap_to_prev(q, &prev->bi_io_vec[prev->bi_vcnt - 1],
1389 				next->bi_io_vec[0].bv_offset);
1390 }
1391 
1392 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1393 {
1394 	return bio_will_gap(req->q, req->biotail, bio);
1395 }
1396 
1397 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1398 {
1399 	return bio_will_gap(req->q, bio, req->bio);
1400 }
1401 
1402 struct work_struct;
1403 int kblockd_schedule_work(struct work_struct *work);
1404 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1405 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1406 
1407 #ifdef CONFIG_BLK_CGROUP
1408 /*
1409  * This should not be using sched_clock(). A real patch is in progress
1410  * to fix this up, until that is in place we need to disable preemption
1411  * around sched_clock() in this function and set_io_start_time_ns().
1412  */
1413 static inline void set_start_time_ns(struct request *req)
1414 {
1415 	preempt_disable();
1416 	req->start_time_ns = sched_clock();
1417 	preempt_enable();
1418 }
1419 
1420 static inline void set_io_start_time_ns(struct request *req)
1421 {
1422 	preempt_disable();
1423 	req->io_start_time_ns = sched_clock();
1424 	preempt_enable();
1425 }
1426 
1427 static inline uint64_t rq_start_time_ns(struct request *req)
1428 {
1429         return req->start_time_ns;
1430 }
1431 
1432 static inline uint64_t rq_io_start_time_ns(struct request *req)
1433 {
1434         return req->io_start_time_ns;
1435 }
1436 #else
1437 static inline void set_start_time_ns(struct request *req) {}
1438 static inline void set_io_start_time_ns(struct request *req) {}
1439 static inline uint64_t rq_start_time_ns(struct request *req)
1440 {
1441 	return 0;
1442 }
1443 static inline uint64_t rq_io_start_time_ns(struct request *req)
1444 {
1445 	return 0;
1446 }
1447 #endif
1448 
1449 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1450 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1451 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1452 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1453 
1454 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1455 
1456 enum blk_integrity_flags {
1457 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1458 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1459 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1460 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1461 };
1462 
1463 struct blk_integrity_iter {
1464 	void			*prot_buf;
1465 	void			*data_buf;
1466 	sector_t		seed;
1467 	unsigned int		data_size;
1468 	unsigned short		interval;
1469 	const char		*disk_name;
1470 };
1471 
1472 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1473 
1474 struct blk_integrity_profile {
1475 	integrity_processing_fn		*generate_fn;
1476 	integrity_processing_fn		*verify_fn;
1477 	const char			*name;
1478 };
1479 
1480 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1481 extern void blk_integrity_unregister(struct gendisk *);
1482 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1483 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1484 				   struct scatterlist *);
1485 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1486 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1487 				   struct request *);
1488 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1489 				    struct bio *);
1490 
1491 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1492 {
1493 	struct blk_integrity *bi = &disk->queue->integrity;
1494 
1495 	if (!bi->profile)
1496 		return NULL;
1497 
1498 	return bi;
1499 }
1500 
1501 static inline
1502 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1503 {
1504 	return blk_get_integrity(bdev->bd_disk);
1505 }
1506 
1507 static inline bool blk_integrity_rq(struct request *rq)
1508 {
1509 	return rq->cmd_flags & REQ_INTEGRITY;
1510 }
1511 
1512 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1513 						    unsigned int segs)
1514 {
1515 	q->limits.max_integrity_segments = segs;
1516 }
1517 
1518 static inline unsigned short
1519 queue_max_integrity_segments(struct request_queue *q)
1520 {
1521 	return q->limits.max_integrity_segments;
1522 }
1523 
1524 static inline bool integrity_req_gap_back_merge(struct request *req,
1525 						struct bio *next)
1526 {
1527 	struct bio_integrity_payload *bip = bio_integrity(req->bio);
1528 	struct bio_integrity_payload *bip_next = bio_integrity(next);
1529 
1530 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1531 				bip_next->bip_vec[0].bv_offset);
1532 }
1533 
1534 static inline bool integrity_req_gap_front_merge(struct request *req,
1535 						 struct bio *bio)
1536 {
1537 	struct bio_integrity_payload *bip = bio_integrity(bio);
1538 	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1539 
1540 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1541 				bip_next->bip_vec[0].bv_offset);
1542 }
1543 
1544 #else /* CONFIG_BLK_DEV_INTEGRITY */
1545 
1546 struct bio;
1547 struct block_device;
1548 struct gendisk;
1549 struct blk_integrity;
1550 
1551 static inline int blk_integrity_rq(struct request *rq)
1552 {
1553 	return 0;
1554 }
1555 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1556 					    struct bio *b)
1557 {
1558 	return 0;
1559 }
1560 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1561 					  struct bio *b,
1562 					  struct scatterlist *s)
1563 {
1564 	return 0;
1565 }
1566 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1567 {
1568 	return NULL;
1569 }
1570 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1571 {
1572 	return NULL;
1573 }
1574 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1575 {
1576 	return 0;
1577 }
1578 static inline void blk_integrity_register(struct gendisk *d,
1579 					 struct blk_integrity *b)
1580 {
1581 }
1582 static inline void blk_integrity_unregister(struct gendisk *d)
1583 {
1584 }
1585 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1586 						    unsigned int segs)
1587 {
1588 }
1589 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1590 {
1591 	return 0;
1592 }
1593 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1594 					  struct request *r1,
1595 					  struct request *r2)
1596 {
1597 	return true;
1598 }
1599 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1600 					   struct request *r,
1601 					   struct bio *b)
1602 {
1603 	return true;
1604 }
1605 
1606 static inline bool integrity_req_gap_back_merge(struct request *req,
1607 						struct bio *next)
1608 {
1609 	return false;
1610 }
1611 static inline bool integrity_req_gap_front_merge(struct request *req,
1612 						 struct bio *bio)
1613 {
1614 	return false;
1615 }
1616 
1617 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1618 
1619 struct block_device_operations {
1620 	int (*open) (struct block_device *, fmode_t);
1621 	void (*release) (struct gendisk *, fmode_t);
1622 	int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1623 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1624 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1625 	long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1626 			unsigned long *pfn);
1627 	unsigned int (*check_events) (struct gendisk *disk,
1628 				      unsigned int clearing);
1629 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
1630 	int (*media_changed) (struct gendisk *);
1631 	void (*unlock_native_capacity) (struct gendisk *);
1632 	int (*revalidate_disk) (struct gendisk *);
1633 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1634 	/* this callback is with swap_lock and sometimes page table lock held */
1635 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1636 	struct module *owner;
1637 	const struct pr_ops *pr_ops;
1638 };
1639 
1640 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1641 				 unsigned long);
1642 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1643 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1644 						struct writeback_control *);
1645 extern long bdev_direct_access(struct block_device *, sector_t,
1646 		void __pmem **addr, unsigned long *pfn, long size);
1647 #else /* CONFIG_BLOCK */
1648 
1649 struct block_device;
1650 
1651 /*
1652  * stubs for when the block layer is configured out
1653  */
1654 #define buffer_heads_over_limit 0
1655 
1656 static inline long nr_blockdev_pages(void)
1657 {
1658 	return 0;
1659 }
1660 
1661 struct blk_plug {
1662 };
1663 
1664 static inline void blk_start_plug(struct blk_plug *plug)
1665 {
1666 }
1667 
1668 static inline void blk_finish_plug(struct blk_plug *plug)
1669 {
1670 }
1671 
1672 static inline void blk_flush_plug(struct task_struct *task)
1673 {
1674 }
1675 
1676 static inline void blk_schedule_flush_plug(struct task_struct *task)
1677 {
1678 }
1679 
1680 
1681 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1682 {
1683 	return false;
1684 }
1685 
1686 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1687 				     sector_t *error_sector)
1688 {
1689 	return 0;
1690 }
1691 
1692 #endif /* CONFIG_BLOCK */
1693 
1694 #endif
1695