xref: /linux-6.15/include/linux/blkdev.h (revision fc4fa6e1)
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 void (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		chunk_sectors;
258 	unsigned int		max_sectors;
259 	unsigned int		max_segment_size;
260 	unsigned int		physical_block_size;
261 	unsigned int		alignment_offset;
262 	unsigned int		io_min;
263 	unsigned int		io_opt;
264 	unsigned int		max_discard_sectors;
265 	unsigned int		max_hw_discard_sectors;
266 	unsigned int		max_write_same_sectors;
267 	unsigned int		discard_granularity;
268 	unsigned int		discard_alignment;
269 
270 	unsigned short		logical_block_size;
271 	unsigned short		max_segments;
272 	unsigned short		max_integrity_segments;
273 
274 	unsigned char		misaligned;
275 	unsigned char		discard_misaligned;
276 	unsigned char		cluster;
277 	unsigned char		discard_zeroes_data;
278 	unsigned char		raid_partial_stripes_expensive;
279 };
280 
281 struct request_queue {
282 	/*
283 	 * Together with queue_head for cacheline sharing
284 	 */
285 	struct list_head	queue_head;
286 	struct request		*last_merge;
287 	struct elevator_queue	*elevator;
288 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
289 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
290 
291 	/*
292 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
293 	 * is used, root blkg allocates from @q->root_rl and all other
294 	 * blkgs from their own blkg->rl.  Which one to use should be
295 	 * determined using bio_request_list().
296 	 */
297 	struct request_list	root_rl;
298 
299 	request_fn_proc		*request_fn;
300 	make_request_fn		*make_request_fn;
301 	prep_rq_fn		*prep_rq_fn;
302 	unprep_rq_fn		*unprep_rq_fn;
303 	softirq_done_fn		*softirq_done_fn;
304 	rq_timed_out_fn		*rq_timed_out_fn;
305 	dma_drain_needed_fn	*dma_drain_needed;
306 	lld_busy_fn		*lld_busy_fn;
307 
308 	struct blk_mq_ops	*mq_ops;
309 
310 	unsigned int		*mq_map;
311 
312 	/* sw queues */
313 	struct blk_mq_ctx __percpu	*queue_ctx;
314 	unsigned int		nr_queues;
315 
316 	/* hw dispatch queues */
317 	struct blk_mq_hw_ctx	**queue_hw_ctx;
318 	unsigned int		nr_hw_queues;
319 
320 	/*
321 	 * Dispatch queue sorting
322 	 */
323 	sector_t		end_sector;
324 	struct request		*boundary_rq;
325 
326 	/*
327 	 * Delayed queue handling
328 	 */
329 	struct delayed_work	delay_work;
330 
331 	struct backing_dev_info	backing_dev_info;
332 
333 	/*
334 	 * The queue owner gets to use this for whatever they like.
335 	 * ll_rw_blk doesn't touch it.
336 	 */
337 	void			*queuedata;
338 
339 	/*
340 	 * various queue flags, see QUEUE_* below
341 	 */
342 	unsigned long		queue_flags;
343 
344 	/*
345 	 * ida allocated id for this queue.  Used to index queues from
346 	 * ioctx.
347 	 */
348 	int			id;
349 
350 	/*
351 	 * queue needs bounce pages for pages above this limit
352 	 */
353 	gfp_t			bounce_gfp;
354 
355 	/*
356 	 * protects queue structures from reentrancy. ->__queue_lock should
357 	 * _never_ be used directly, it is queue private. always use
358 	 * ->queue_lock.
359 	 */
360 	spinlock_t		__queue_lock;
361 	spinlock_t		*queue_lock;
362 
363 	/*
364 	 * queue kobject
365 	 */
366 	struct kobject kobj;
367 
368 	/*
369 	 * mq queue kobject
370 	 */
371 	struct kobject mq_kobj;
372 
373 #ifdef  CONFIG_BLK_DEV_INTEGRITY
374 	struct blk_integrity integrity;
375 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
376 
377 #ifdef CONFIG_PM
378 	struct device		*dev;
379 	int			rpm_status;
380 	unsigned int		nr_pending;
381 #endif
382 
383 	/*
384 	 * queue settings
385 	 */
386 	unsigned long		nr_requests;	/* Max # of requests */
387 	unsigned int		nr_congestion_on;
388 	unsigned int		nr_congestion_off;
389 	unsigned int		nr_batching;
390 
391 	unsigned int		dma_drain_size;
392 	void			*dma_drain_buffer;
393 	unsigned int		dma_pad_mask;
394 	unsigned int		dma_alignment;
395 
396 	struct blk_queue_tag	*queue_tags;
397 	struct list_head	tag_busy_list;
398 
399 	unsigned int		nr_sorted;
400 	unsigned int		in_flight[2];
401 	/*
402 	 * Number of active block driver functions for which blk_drain_queue()
403 	 * must wait. Must be incremented around functions that unlock the
404 	 * queue_lock internally, e.g. scsi_request_fn().
405 	 */
406 	unsigned int		request_fn_active;
407 
408 	unsigned int		rq_timeout;
409 	struct timer_list	timeout;
410 	struct list_head	timeout_list;
411 
412 	struct list_head	icq_list;
413 #ifdef CONFIG_BLK_CGROUP
414 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
415 	struct blkcg_gq		*root_blkg;
416 	struct list_head	blkg_list;
417 #endif
418 
419 	struct queue_limits	limits;
420 
421 	/*
422 	 * sg stuff
423 	 */
424 	unsigned int		sg_timeout;
425 	unsigned int		sg_reserved_size;
426 	int			node;
427 #ifdef CONFIG_BLK_DEV_IO_TRACE
428 	struct blk_trace	*blk_trace;
429 #endif
430 	/*
431 	 * for flush operations
432 	 */
433 	unsigned int		flush_flags;
434 	unsigned int		flush_not_queueable:1;
435 	struct blk_flush_queue	*fq;
436 
437 	struct list_head	requeue_list;
438 	spinlock_t		requeue_lock;
439 	struct work_struct	requeue_work;
440 
441 	struct mutex		sysfs_lock;
442 
443 	int			bypass_depth;
444 	atomic_t		mq_freeze_depth;
445 
446 #if defined(CONFIG_BLK_DEV_BSG)
447 	bsg_job_fn		*bsg_job_fn;
448 	int			bsg_job_size;
449 	struct bsg_class_device bsg_dev;
450 #endif
451 
452 #ifdef CONFIG_BLK_DEV_THROTTLING
453 	/* Throttle data */
454 	struct throtl_data *td;
455 #endif
456 	struct rcu_head		rcu_head;
457 	wait_queue_head_t	mq_freeze_wq;
458 	struct percpu_ref	q_usage_counter;
459 	struct list_head	all_q_node;
460 
461 	struct blk_mq_tag_set	*tag_set;
462 	struct list_head	tag_set_list;
463 	struct bio_set		*bio_split;
464 
465 	bool			mq_sysfs_init_done;
466 };
467 
468 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
469 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
470 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
471 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
472 #define QUEUE_FLAG_DYING	5	/* queue being torn down */
473 #define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
474 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
475 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
476 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
477 #define QUEUE_FLAG_FAIL_IO     10	/* fake timeout */
478 #define QUEUE_FLAG_STACKABLE   11	/* supports request stacking */
479 #define QUEUE_FLAG_NONROT      12	/* non-rotational device (SSD) */
480 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
481 #define QUEUE_FLAG_IO_STAT     13	/* do IO stats */
482 #define QUEUE_FLAG_DISCARD     14	/* supports DISCARD */
483 #define QUEUE_FLAG_NOXMERGES   15	/* No extended merges */
484 #define QUEUE_FLAG_ADD_RANDOM  16	/* Contributes to random pool */
485 #define QUEUE_FLAG_SECDISCARD  17	/* supports SECDISCARD */
486 #define QUEUE_FLAG_SAME_FORCE  18	/* force complete on same CPU */
487 #define QUEUE_FLAG_DEAD        19	/* queue tear-down finished */
488 #define QUEUE_FLAG_INIT_DONE   20	/* queue is initialized */
489 #define QUEUE_FLAG_NO_SG_MERGE 21	/* don't attempt to merge SG segments*/
490 
491 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
492 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
493 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
494 				 (1 << QUEUE_FLAG_ADD_RANDOM))
495 
496 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
497 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
498 				 (1 << QUEUE_FLAG_SAME_COMP))
499 
500 static inline void queue_lockdep_assert_held(struct request_queue *q)
501 {
502 	if (q->queue_lock)
503 		lockdep_assert_held(q->queue_lock);
504 }
505 
506 static inline void queue_flag_set_unlocked(unsigned int flag,
507 					   struct request_queue *q)
508 {
509 	__set_bit(flag, &q->queue_flags);
510 }
511 
512 static inline int queue_flag_test_and_clear(unsigned int flag,
513 					    struct request_queue *q)
514 {
515 	queue_lockdep_assert_held(q);
516 
517 	if (test_bit(flag, &q->queue_flags)) {
518 		__clear_bit(flag, &q->queue_flags);
519 		return 1;
520 	}
521 
522 	return 0;
523 }
524 
525 static inline int queue_flag_test_and_set(unsigned int flag,
526 					  struct request_queue *q)
527 {
528 	queue_lockdep_assert_held(q);
529 
530 	if (!test_bit(flag, &q->queue_flags)) {
531 		__set_bit(flag, &q->queue_flags);
532 		return 0;
533 	}
534 
535 	return 1;
536 }
537 
538 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
539 {
540 	queue_lockdep_assert_held(q);
541 	__set_bit(flag, &q->queue_flags);
542 }
543 
544 static inline void queue_flag_clear_unlocked(unsigned int flag,
545 					     struct request_queue *q)
546 {
547 	__clear_bit(flag, &q->queue_flags);
548 }
549 
550 static inline int queue_in_flight(struct request_queue *q)
551 {
552 	return q->in_flight[0] + q->in_flight[1];
553 }
554 
555 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
556 {
557 	queue_lockdep_assert_held(q);
558 	__clear_bit(flag, &q->queue_flags);
559 }
560 
561 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
562 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
563 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
564 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
565 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
566 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
567 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
568 #define blk_queue_noxmerges(q)	\
569 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
570 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
571 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
572 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
573 #define blk_queue_stackable(q)	\
574 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
575 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
576 #define blk_queue_secdiscard(q)	(blk_queue_discard(q) && \
577 	test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
578 
579 #define blk_noretry_request(rq) \
580 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
581 			     REQ_FAILFAST_DRIVER))
582 
583 #define blk_account_rq(rq) \
584 	(((rq)->cmd_flags & REQ_STARTED) && \
585 	 ((rq)->cmd_type == REQ_TYPE_FS))
586 
587 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
588 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
589 /* rq->queuelist of dequeued request must be list_empty() */
590 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
591 
592 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
593 
594 #define rq_data_dir(rq)		((int)((rq)->cmd_flags & 1))
595 
596 /*
597  * Driver can handle struct request, if it either has an old style
598  * request_fn defined, or is blk-mq based.
599  */
600 static inline bool queue_is_rq_based(struct request_queue *q)
601 {
602 	return q->request_fn || q->mq_ops;
603 }
604 
605 static inline unsigned int blk_queue_cluster(struct request_queue *q)
606 {
607 	return q->limits.cluster;
608 }
609 
610 /*
611  * We regard a request as sync, if either a read or a sync write
612  */
613 static inline bool rw_is_sync(unsigned int rw_flags)
614 {
615 	return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
616 }
617 
618 static inline bool rq_is_sync(struct request *rq)
619 {
620 	return rw_is_sync(rq->cmd_flags);
621 }
622 
623 static inline bool blk_rl_full(struct request_list *rl, bool sync)
624 {
625 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
626 
627 	return rl->flags & flag;
628 }
629 
630 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
631 {
632 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
633 
634 	rl->flags |= flag;
635 }
636 
637 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
638 {
639 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
640 
641 	rl->flags &= ~flag;
642 }
643 
644 static inline bool rq_mergeable(struct request *rq)
645 {
646 	if (rq->cmd_type != REQ_TYPE_FS)
647 		return false;
648 
649 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
650 		return false;
651 
652 	return true;
653 }
654 
655 static inline bool blk_check_merge_flags(unsigned int flags1,
656 					 unsigned int flags2)
657 {
658 	if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
659 		return false;
660 
661 	if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
662 		return false;
663 
664 	if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
665 		return false;
666 
667 	return true;
668 }
669 
670 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
671 {
672 	if (bio_data(a) == bio_data(b))
673 		return true;
674 
675 	return false;
676 }
677 
678 /*
679  * q->prep_rq_fn return values
680  */
681 #define BLKPREP_OK		0	/* serve it */
682 #define BLKPREP_KILL		1	/* fatal error, kill */
683 #define BLKPREP_DEFER		2	/* leave on queue */
684 
685 extern unsigned long blk_max_low_pfn, blk_max_pfn;
686 
687 /*
688  * standard bounce addresses:
689  *
690  * BLK_BOUNCE_HIGH	: bounce all highmem pages
691  * BLK_BOUNCE_ANY	: don't bounce anything
692  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
693  */
694 
695 #if BITS_PER_LONG == 32
696 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
697 #else
698 #define BLK_BOUNCE_HIGH		-1ULL
699 #endif
700 #define BLK_BOUNCE_ANY		(-1ULL)
701 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
702 
703 /*
704  * default timeout for SG_IO if none specified
705  */
706 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
707 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
708 
709 #ifdef CONFIG_BOUNCE
710 extern int init_emergency_isa_pool(void);
711 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
712 #else
713 static inline int init_emergency_isa_pool(void)
714 {
715 	return 0;
716 }
717 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
718 {
719 }
720 #endif /* CONFIG_MMU */
721 
722 struct rq_map_data {
723 	struct page **pages;
724 	int page_order;
725 	int nr_entries;
726 	unsigned long offset;
727 	int null_mapped;
728 	int from_user;
729 };
730 
731 struct req_iterator {
732 	struct bvec_iter iter;
733 	struct bio *bio;
734 };
735 
736 /* This should not be used directly - use rq_for_each_segment */
737 #define for_each_bio(_bio)		\
738 	for (; _bio; _bio = _bio->bi_next)
739 #define __rq_for_each_bio(_bio, rq)	\
740 	if ((rq->bio))			\
741 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
742 
743 #define rq_for_each_segment(bvl, _rq, _iter)			\
744 	__rq_for_each_bio(_iter.bio, _rq)			\
745 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
746 
747 #define rq_iter_last(bvec, _iter)				\
748 		(_iter.bio->bi_next == NULL &&			\
749 		 bio_iter_last(bvec, _iter.iter))
750 
751 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
752 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
753 #endif
754 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
755 extern void rq_flush_dcache_pages(struct request *rq);
756 #else
757 static inline void rq_flush_dcache_pages(struct request *rq)
758 {
759 }
760 #endif
761 
762 extern int blk_register_queue(struct gendisk *disk);
763 extern void blk_unregister_queue(struct gendisk *disk);
764 extern void generic_make_request(struct bio *bio);
765 extern void blk_rq_init(struct request_queue *q, struct request *rq);
766 extern void blk_put_request(struct request *);
767 extern void __blk_put_request(struct request_queue *, struct request *);
768 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
769 extern struct request *blk_make_request(struct request_queue *, struct bio *,
770 					gfp_t);
771 extern void blk_rq_set_block_pc(struct request *);
772 extern void blk_requeue_request(struct request_queue *, struct request *);
773 extern void blk_add_request_payload(struct request *rq, struct page *page,
774 		unsigned int len);
775 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
776 extern int blk_lld_busy(struct request_queue *q);
777 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
778 			     struct bio_set *bs, gfp_t gfp_mask,
779 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
780 			     void *data);
781 extern void blk_rq_unprep_clone(struct request *rq);
782 extern int blk_insert_cloned_request(struct request_queue *q,
783 				     struct request *rq);
784 extern void blk_delay_queue(struct request_queue *, unsigned long);
785 extern void blk_queue_split(struct request_queue *, struct bio **,
786 			    struct bio_set *);
787 extern void blk_recount_segments(struct request_queue *, struct bio *);
788 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
789 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
790 			      unsigned int, void __user *);
791 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
792 			  unsigned int, void __user *);
793 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
794 			 struct scsi_ioctl_command __user *);
795 
796 extern void blk_start_queue(struct request_queue *q);
797 extern void blk_stop_queue(struct request_queue *q);
798 extern void blk_sync_queue(struct request_queue *q);
799 extern void __blk_stop_queue(struct request_queue *q);
800 extern void __blk_run_queue(struct request_queue *q);
801 extern void __blk_run_queue_uncond(struct request_queue *q);
802 extern void blk_run_queue(struct request_queue *);
803 extern void blk_run_queue_async(struct request_queue *q);
804 extern int blk_rq_map_user(struct request_queue *, struct request *,
805 			   struct rq_map_data *, void __user *, unsigned long,
806 			   gfp_t);
807 extern int blk_rq_unmap_user(struct bio *);
808 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
809 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
810 			       struct rq_map_data *, const struct iov_iter *,
811 			       gfp_t);
812 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
813 			  struct request *, int);
814 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
815 				  struct request *, int, rq_end_io_fn *);
816 
817 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
818 {
819 	return bdev->bd_disk->queue;	/* this is never NULL */
820 }
821 
822 /*
823  * blk_rq_pos()			: the current sector
824  * blk_rq_bytes()		: bytes left in the entire request
825  * blk_rq_cur_bytes()		: bytes left in the current segment
826  * blk_rq_err_bytes()		: bytes left till the next error boundary
827  * blk_rq_sectors()		: sectors left in the entire request
828  * blk_rq_cur_sectors()		: sectors left in the current segment
829  */
830 static inline sector_t blk_rq_pos(const struct request *rq)
831 {
832 	return rq->__sector;
833 }
834 
835 static inline unsigned int blk_rq_bytes(const struct request *rq)
836 {
837 	return rq->__data_len;
838 }
839 
840 static inline int blk_rq_cur_bytes(const struct request *rq)
841 {
842 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
843 }
844 
845 extern unsigned int blk_rq_err_bytes(const struct request *rq);
846 
847 static inline unsigned int blk_rq_sectors(const struct request *rq)
848 {
849 	return blk_rq_bytes(rq) >> 9;
850 }
851 
852 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
853 {
854 	return blk_rq_cur_bytes(rq) >> 9;
855 }
856 
857 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
858 						     unsigned int cmd_flags)
859 {
860 	if (unlikely(cmd_flags & REQ_DISCARD))
861 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
862 
863 	if (unlikely(cmd_flags & REQ_WRITE_SAME))
864 		return q->limits.max_write_same_sectors;
865 
866 	return q->limits.max_sectors;
867 }
868 
869 /*
870  * Return maximum size of a request at given offset. Only valid for
871  * file system requests.
872  */
873 static inline unsigned int blk_max_size_offset(struct request_queue *q,
874 					       sector_t offset)
875 {
876 	if (!q->limits.chunk_sectors)
877 		return q->limits.max_sectors;
878 
879 	return q->limits.chunk_sectors -
880 			(offset & (q->limits.chunk_sectors - 1));
881 }
882 
883 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
884 {
885 	struct request_queue *q = rq->q;
886 
887 	if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
888 		return q->limits.max_hw_sectors;
889 
890 	if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
891 		return blk_queue_get_max_sectors(q, rq->cmd_flags);
892 
893 	return min(blk_max_size_offset(q, blk_rq_pos(rq)),
894 			blk_queue_get_max_sectors(q, rq->cmd_flags));
895 }
896 
897 static inline unsigned int blk_rq_count_bios(struct request *rq)
898 {
899 	unsigned int nr_bios = 0;
900 	struct bio *bio;
901 
902 	__rq_for_each_bio(bio, rq)
903 		nr_bios++;
904 
905 	return nr_bios;
906 }
907 
908 /*
909  * Request issue related functions.
910  */
911 extern struct request *blk_peek_request(struct request_queue *q);
912 extern void blk_start_request(struct request *rq);
913 extern struct request *blk_fetch_request(struct request_queue *q);
914 
915 /*
916  * Request completion related functions.
917  *
918  * blk_update_request() completes given number of bytes and updates
919  * the request without completing it.
920  *
921  * blk_end_request() and friends.  __blk_end_request() must be called
922  * with the request queue spinlock acquired.
923  *
924  * Several drivers define their own end_request and call
925  * blk_end_request() for parts of the original function.
926  * This prevents code duplication in drivers.
927  */
928 extern bool blk_update_request(struct request *rq, int error,
929 			       unsigned int nr_bytes);
930 extern void blk_finish_request(struct request *rq, int error);
931 extern bool blk_end_request(struct request *rq, int error,
932 			    unsigned int nr_bytes);
933 extern void blk_end_request_all(struct request *rq, int error);
934 extern bool blk_end_request_cur(struct request *rq, int error);
935 extern bool blk_end_request_err(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 
942 extern void blk_complete_request(struct request *);
943 extern void __blk_complete_request(struct request *);
944 extern void blk_abort_request(struct request *);
945 extern void blk_unprep_request(struct request *);
946 
947 /*
948  * Access functions for manipulating queue properties
949  */
950 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
951 					spinlock_t *lock, int node_id);
952 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
953 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
954 						      request_fn_proc *, spinlock_t *);
955 extern void blk_cleanup_queue(struct request_queue *);
956 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
957 extern void blk_queue_bounce_limit(struct request_queue *, u64);
958 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
959 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
960 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
961 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
962 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
963 extern void blk_queue_max_discard_sectors(struct request_queue *q,
964 		unsigned int max_discard_sectors);
965 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
966 		unsigned int max_write_same_sectors);
967 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
968 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
969 extern void blk_queue_alignment_offset(struct request_queue *q,
970 				       unsigned int alignment);
971 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
972 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
973 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
974 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
975 extern void blk_set_default_limits(struct queue_limits *lim);
976 extern void blk_set_stacking_limits(struct queue_limits *lim);
977 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
978 			    sector_t offset);
979 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
980 			    sector_t offset);
981 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
982 			      sector_t offset);
983 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
984 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
985 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
986 extern int blk_queue_dma_drain(struct request_queue *q,
987 			       dma_drain_needed_fn *dma_drain_needed,
988 			       void *buf, unsigned int size);
989 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
990 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
991 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
992 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
993 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
994 extern void blk_queue_dma_alignment(struct request_queue *, int);
995 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
996 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
997 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
998 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
999 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1000 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1001 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1002 
1003 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1004 extern void blk_dump_rq_flags(struct request *, char *);
1005 extern long nr_blockdev_pages(void);
1006 
1007 bool __must_check blk_get_queue(struct request_queue *);
1008 struct request_queue *blk_alloc_queue(gfp_t);
1009 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1010 extern void blk_put_queue(struct request_queue *);
1011 extern void blk_set_queue_dying(struct request_queue *);
1012 
1013 /*
1014  * block layer runtime pm functions
1015  */
1016 #ifdef CONFIG_PM
1017 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1018 extern int blk_pre_runtime_suspend(struct request_queue *q);
1019 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1020 extern void blk_pre_runtime_resume(struct request_queue *q);
1021 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1022 #else
1023 static inline void blk_pm_runtime_init(struct request_queue *q,
1024 	struct device *dev) {}
1025 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1026 {
1027 	return -ENOSYS;
1028 }
1029 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1030 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1031 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1032 #endif
1033 
1034 /*
1035  * blk_plug permits building a queue of related requests by holding the I/O
1036  * fragments for a short period. This allows merging of sequential requests
1037  * into single larger request. As the requests are moved from a per-task list to
1038  * the device's request_queue in a batch, this results in improved scalability
1039  * as the lock contention for request_queue lock is reduced.
1040  *
1041  * It is ok not to disable preemption when adding the request to the plug list
1042  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1043  * the plug list when the task sleeps by itself. For details, please see
1044  * schedule() where blk_schedule_flush_plug() is called.
1045  */
1046 struct blk_plug {
1047 	struct list_head list; /* requests */
1048 	struct list_head mq_list; /* blk-mq requests */
1049 	struct list_head cb_list; /* md requires an unplug callback */
1050 };
1051 #define BLK_MAX_REQUEST_COUNT 16
1052 
1053 struct blk_plug_cb;
1054 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1055 struct blk_plug_cb {
1056 	struct list_head list;
1057 	blk_plug_cb_fn callback;
1058 	void *data;
1059 };
1060 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1061 					     void *data, int size);
1062 extern void blk_start_plug(struct blk_plug *);
1063 extern void blk_finish_plug(struct blk_plug *);
1064 extern void blk_flush_plug_list(struct blk_plug *, bool);
1065 
1066 static inline void blk_flush_plug(struct task_struct *tsk)
1067 {
1068 	struct blk_plug *plug = tsk->plug;
1069 
1070 	if (plug)
1071 		blk_flush_plug_list(plug, false);
1072 }
1073 
1074 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1075 {
1076 	struct blk_plug *plug = tsk->plug;
1077 
1078 	if (plug)
1079 		blk_flush_plug_list(plug, true);
1080 }
1081 
1082 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1083 {
1084 	struct blk_plug *plug = tsk->plug;
1085 
1086 	return plug &&
1087 		(!list_empty(&plug->list) ||
1088 		 !list_empty(&plug->mq_list) ||
1089 		 !list_empty(&plug->cb_list));
1090 }
1091 
1092 /*
1093  * tag stuff
1094  */
1095 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1096 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1097 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1098 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1099 extern void blk_queue_free_tags(struct request_queue *);
1100 extern int blk_queue_resize_tags(struct request_queue *, int);
1101 extern void blk_queue_invalidate_tags(struct request_queue *);
1102 extern struct blk_queue_tag *blk_init_tags(int, int);
1103 extern void blk_free_tags(struct blk_queue_tag *);
1104 
1105 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1106 						int tag)
1107 {
1108 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1109 		return NULL;
1110 	return bqt->tag_index[tag];
1111 }
1112 
1113 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1114 
1115 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1116 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1117 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1118 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1119 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1120 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1121 		sector_t nr_sects, gfp_t gfp_mask, bool discard);
1122 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1123 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1124 {
1125 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1126 				    nr_blocks << (sb->s_blocksize_bits - 9),
1127 				    gfp_mask, flags);
1128 }
1129 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1130 		sector_t nr_blocks, gfp_t gfp_mask)
1131 {
1132 	return blkdev_issue_zeroout(sb->s_bdev,
1133 				    block << (sb->s_blocksize_bits - 9),
1134 				    nr_blocks << (sb->s_blocksize_bits - 9),
1135 				    gfp_mask, true);
1136 }
1137 
1138 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1139 
1140 enum blk_default_limits {
1141 	BLK_MAX_SEGMENTS	= 128,
1142 	BLK_SAFE_MAX_SECTORS	= 255,
1143 	BLK_DEF_MAX_SECTORS	= 2560,
1144 	BLK_MAX_SEGMENT_SIZE	= 65536,
1145 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1146 };
1147 
1148 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1149 
1150 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1151 {
1152 	return q->limits.bounce_pfn;
1153 }
1154 
1155 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1156 {
1157 	return q->limits.seg_boundary_mask;
1158 }
1159 
1160 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1161 {
1162 	return q->limits.virt_boundary_mask;
1163 }
1164 
1165 static inline unsigned int queue_max_sectors(struct request_queue *q)
1166 {
1167 	return q->limits.max_sectors;
1168 }
1169 
1170 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1171 {
1172 	return q->limits.max_hw_sectors;
1173 }
1174 
1175 static inline unsigned short queue_max_segments(struct request_queue *q)
1176 {
1177 	return q->limits.max_segments;
1178 }
1179 
1180 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1181 {
1182 	return q->limits.max_segment_size;
1183 }
1184 
1185 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1186 {
1187 	int retval = 512;
1188 
1189 	if (q && q->limits.logical_block_size)
1190 		retval = q->limits.logical_block_size;
1191 
1192 	return retval;
1193 }
1194 
1195 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1196 {
1197 	return queue_logical_block_size(bdev_get_queue(bdev));
1198 }
1199 
1200 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1201 {
1202 	return q->limits.physical_block_size;
1203 }
1204 
1205 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1206 {
1207 	return queue_physical_block_size(bdev_get_queue(bdev));
1208 }
1209 
1210 static inline unsigned int queue_io_min(struct request_queue *q)
1211 {
1212 	return q->limits.io_min;
1213 }
1214 
1215 static inline int bdev_io_min(struct block_device *bdev)
1216 {
1217 	return queue_io_min(bdev_get_queue(bdev));
1218 }
1219 
1220 static inline unsigned int queue_io_opt(struct request_queue *q)
1221 {
1222 	return q->limits.io_opt;
1223 }
1224 
1225 static inline int bdev_io_opt(struct block_device *bdev)
1226 {
1227 	return queue_io_opt(bdev_get_queue(bdev));
1228 }
1229 
1230 static inline int queue_alignment_offset(struct request_queue *q)
1231 {
1232 	if (q->limits.misaligned)
1233 		return -1;
1234 
1235 	return q->limits.alignment_offset;
1236 }
1237 
1238 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1239 {
1240 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1241 	unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1242 
1243 	return (granularity + lim->alignment_offset - alignment) % granularity;
1244 }
1245 
1246 static inline int bdev_alignment_offset(struct block_device *bdev)
1247 {
1248 	struct request_queue *q = bdev_get_queue(bdev);
1249 
1250 	if (q->limits.misaligned)
1251 		return -1;
1252 
1253 	if (bdev != bdev->bd_contains)
1254 		return bdev->bd_part->alignment_offset;
1255 
1256 	return q->limits.alignment_offset;
1257 }
1258 
1259 static inline int queue_discard_alignment(struct request_queue *q)
1260 {
1261 	if (q->limits.discard_misaligned)
1262 		return -1;
1263 
1264 	return q->limits.discard_alignment;
1265 }
1266 
1267 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1268 {
1269 	unsigned int alignment, granularity, offset;
1270 
1271 	if (!lim->max_discard_sectors)
1272 		return 0;
1273 
1274 	/* Why are these in bytes, not sectors? */
1275 	alignment = lim->discard_alignment >> 9;
1276 	granularity = lim->discard_granularity >> 9;
1277 	if (!granularity)
1278 		return 0;
1279 
1280 	/* Offset of the partition start in 'granularity' sectors */
1281 	offset = sector_div(sector, granularity);
1282 
1283 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1284 	offset = (granularity + alignment - offset) % granularity;
1285 
1286 	/* Turn it back into bytes, gaah */
1287 	return offset << 9;
1288 }
1289 
1290 static inline int bdev_discard_alignment(struct block_device *bdev)
1291 {
1292 	struct request_queue *q = bdev_get_queue(bdev);
1293 
1294 	if (bdev != bdev->bd_contains)
1295 		return bdev->bd_part->discard_alignment;
1296 
1297 	return q->limits.discard_alignment;
1298 }
1299 
1300 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1301 {
1302 	if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1303 		return 1;
1304 
1305 	return 0;
1306 }
1307 
1308 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1309 {
1310 	return queue_discard_zeroes_data(bdev_get_queue(bdev));
1311 }
1312 
1313 static inline unsigned int bdev_write_same(struct block_device *bdev)
1314 {
1315 	struct request_queue *q = bdev_get_queue(bdev);
1316 
1317 	if (q)
1318 		return q->limits.max_write_same_sectors;
1319 
1320 	return 0;
1321 }
1322 
1323 static inline int queue_dma_alignment(struct request_queue *q)
1324 {
1325 	return q ? q->dma_alignment : 511;
1326 }
1327 
1328 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1329 				 unsigned int len)
1330 {
1331 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1332 	return !(addr & alignment) && !(len & alignment);
1333 }
1334 
1335 /* assumes size > 256 */
1336 static inline unsigned int blksize_bits(unsigned int size)
1337 {
1338 	unsigned int bits = 8;
1339 	do {
1340 		bits++;
1341 		size >>= 1;
1342 	} while (size > 256);
1343 	return bits;
1344 }
1345 
1346 static inline unsigned int block_size(struct block_device *bdev)
1347 {
1348 	return bdev->bd_block_size;
1349 }
1350 
1351 static inline bool queue_flush_queueable(struct request_queue *q)
1352 {
1353 	return !q->flush_not_queueable;
1354 }
1355 
1356 typedef struct {struct page *v;} Sector;
1357 
1358 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1359 
1360 static inline void put_dev_sector(Sector p)
1361 {
1362 	page_cache_release(p.v);
1363 }
1364 
1365 /*
1366  * Check if adding a bio_vec after bprv with offset would create a gap in
1367  * the SG list. Most drivers don't care about this, but some do.
1368  */
1369 static inline bool bvec_gap_to_prev(struct request_queue *q,
1370 				struct bio_vec *bprv, unsigned int offset)
1371 {
1372 	if (!queue_virt_boundary(q))
1373 		return false;
1374 	return offset ||
1375 		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1376 }
1377 
1378 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1379 			 struct bio *next)
1380 {
1381 	if (!bio_has_data(prev))
1382 		return false;
1383 
1384 	return bvec_gap_to_prev(q, &prev->bi_io_vec[prev->bi_vcnt - 1],
1385 				next->bi_io_vec[0].bv_offset);
1386 }
1387 
1388 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1389 {
1390 	return bio_will_gap(req->q, req->biotail, bio);
1391 }
1392 
1393 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1394 {
1395 	return bio_will_gap(req->q, bio, req->bio);
1396 }
1397 
1398 struct work_struct;
1399 int kblockd_schedule_work(struct work_struct *work);
1400 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1401 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1402 
1403 #ifdef CONFIG_BLK_CGROUP
1404 /*
1405  * This should not be using sched_clock(). A real patch is in progress
1406  * to fix this up, until that is in place we need to disable preemption
1407  * around sched_clock() in this function and set_io_start_time_ns().
1408  */
1409 static inline void set_start_time_ns(struct request *req)
1410 {
1411 	preempt_disable();
1412 	req->start_time_ns = sched_clock();
1413 	preempt_enable();
1414 }
1415 
1416 static inline void set_io_start_time_ns(struct request *req)
1417 {
1418 	preempt_disable();
1419 	req->io_start_time_ns = sched_clock();
1420 	preempt_enable();
1421 }
1422 
1423 static inline uint64_t rq_start_time_ns(struct request *req)
1424 {
1425         return req->start_time_ns;
1426 }
1427 
1428 static inline uint64_t rq_io_start_time_ns(struct request *req)
1429 {
1430         return req->io_start_time_ns;
1431 }
1432 #else
1433 static inline void set_start_time_ns(struct request *req) {}
1434 static inline void set_io_start_time_ns(struct request *req) {}
1435 static inline uint64_t rq_start_time_ns(struct request *req)
1436 {
1437 	return 0;
1438 }
1439 static inline uint64_t rq_io_start_time_ns(struct request *req)
1440 {
1441 	return 0;
1442 }
1443 #endif
1444 
1445 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1446 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1447 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1448 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1449 
1450 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1451 
1452 enum blk_integrity_flags {
1453 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1454 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1455 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1456 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1457 };
1458 
1459 struct blk_integrity_iter {
1460 	void			*prot_buf;
1461 	void			*data_buf;
1462 	sector_t		seed;
1463 	unsigned int		data_size;
1464 	unsigned short		interval;
1465 	const char		*disk_name;
1466 };
1467 
1468 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1469 
1470 struct blk_integrity_profile {
1471 	integrity_processing_fn		*generate_fn;
1472 	integrity_processing_fn		*verify_fn;
1473 	const char			*name;
1474 };
1475 
1476 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1477 extern void blk_integrity_unregister(struct gendisk *);
1478 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1479 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1480 				   struct scatterlist *);
1481 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1482 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1483 				   struct request *);
1484 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1485 				    struct bio *);
1486 
1487 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1488 {
1489 	struct blk_integrity *bi = &disk->queue->integrity;
1490 
1491 	if (!bi->profile)
1492 		return NULL;
1493 
1494 	return bi;
1495 }
1496 
1497 static inline
1498 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1499 {
1500 	return blk_get_integrity(bdev->bd_disk);
1501 }
1502 
1503 static inline bool blk_integrity_rq(struct request *rq)
1504 {
1505 	return rq->cmd_flags & REQ_INTEGRITY;
1506 }
1507 
1508 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1509 						    unsigned int segs)
1510 {
1511 	q->limits.max_integrity_segments = segs;
1512 }
1513 
1514 static inline unsigned short
1515 queue_max_integrity_segments(struct request_queue *q)
1516 {
1517 	return q->limits.max_integrity_segments;
1518 }
1519 
1520 static inline bool integrity_req_gap_back_merge(struct request *req,
1521 						struct bio *next)
1522 {
1523 	struct bio_integrity_payload *bip = bio_integrity(req->bio);
1524 	struct bio_integrity_payload *bip_next = bio_integrity(next);
1525 
1526 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1527 				bip_next->bip_vec[0].bv_offset);
1528 }
1529 
1530 static inline bool integrity_req_gap_front_merge(struct request *req,
1531 						 struct bio *bio)
1532 {
1533 	struct bio_integrity_payload *bip = bio_integrity(bio);
1534 	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1535 
1536 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1537 				bip_next->bip_vec[0].bv_offset);
1538 }
1539 
1540 #else /* CONFIG_BLK_DEV_INTEGRITY */
1541 
1542 struct bio;
1543 struct block_device;
1544 struct gendisk;
1545 struct blk_integrity;
1546 
1547 static inline int blk_integrity_rq(struct request *rq)
1548 {
1549 	return 0;
1550 }
1551 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1552 					    struct bio *b)
1553 {
1554 	return 0;
1555 }
1556 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1557 					  struct bio *b,
1558 					  struct scatterlist *s)
1559 {
1560 	return 0;
1561 }
1562 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1563 {
1564 	return NULL;
1565 }
1566 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1567 {
1568 	return NULL;
1569 }
1570 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1571 {
1572 	return 0;
1573 }
1574 static inline void blk_integrity_register(struct gendisk *d,
1575 					 struct blk_integrity *b)
1576 {
1577 }
1578 static inline void blk_integrity_unregister(struct gendisk *d)
1579 {
1580 }
1581 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1582 						    unsigned int segs)
1583 {
1584 }
1585 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1586 {
1587 	return 0;
1588 }
1589 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1590 					  struct request *r1,
1591 					  struct request *r2)
1592 {
1593 	return true;
1594 }
1595 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1596 					   struct request *r,
1597 					   struct bio *b)
1598 {
1599 	return true;
1600 }
1601 
1602 static inline bool integrity_req_gap_back_merge(struct request *req,
1603 						struct bio *next)
1604 {
1605 	return false;
1606 }
1607 static inline bool integrity_req_gap_front_merge(struct request *req,
1608 						 struct bio *bio)
1609 {
1610 	return false;
1611 }
1612 
1613 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1614 
1615 struct block_device_operations {
1616 	int (*open) (struct block_device *, fmode_t);
1617 	void (*release) (struct gendisk *, fmode_t);
1618 	int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1619 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1620 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1621 	long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1622 			unsigned long *pfn);
1623 	unsigned int (*check_events) (struct gendisk *disk,
1624 				      unsigned int clearing);
1625 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
1626 	int (*media_changed) (struct gendisk *);
1627 	void (*unlock_native_capacity) (struct gendisk *);
1628 	int (*revalidate_disk) (struct gendisk *);
1629 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1630 	/* this callback is with swap_lock and sometimes page table lock held */
1631 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1632 	struct module *owner;
1633 	const struct pr_ops *pr_ops;
1634 };
1635 
1636 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1637 				 unsigned long);
1638 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1639 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1640 						struct writeback_control *);
1641 extern long bdev_direct_access(struct block_device *, sector_t,
1642 		void __pmem **addr, unsigned long *pfn, long size);
1643 #else /* CONFIG_BLOCK */
1644 
1645 struct block_device;
1646 
1647 /*
1648  * stubs for when the block layer is configured out
1649  */
1650 #define buffer_heads_over_limit 0
1651 
1652 static inline long nr_blockdev_pages(void)
1653 {
1654 	return 0;
1655 }
1656 
1657 struct blk_plug {
1658 };
1659 
1660 static inline void blk_start_plug(struct blk_plug *plug)
1661 {
1662 }
1663 
1664 static inline void blk_finish_plug(struct blk_plug *plug)
1665 {
1666 }
1667 
1668 static inline void blk_flush_plug(struct task_struct *task)
1669 {
1670 }
1671 
1672 static inline void blk_schedule_flush_plug(struct task_struct *task)
1673 {
1674 }
1675 
1676 
1677 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1678 {
1679 	return false;
1680 }
1681 
1682 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1683 				     sector_t *error_sector)
1684 {
1685 	return 0;
1686 }
1687 
1688 #endif /* CONFIG_BLOCK */
1689 
1690 #endif
1691