xref: /linux-6.15/include/linux/blkdev.h (revision cc19ada7)
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 
39 #define BLKDEV_MIN_RQ	4
40 #define BLKDEV_MAX_RQ	128	/* Default maximum */
41 
42 /*
43  * Maximum number of blkcg policies allowed to be registered concurrently.
44  * Defined here to simplify include dependency.
45  */
46 #define BLKCG_MAX_POLS		2
47 
48 struct request;
49 typedef void (rq_end_io_fn)(struct request *, int);
50 
51 #define BLK_RL_SYNCFULL		(1U << 0)
52 #define BLK_RL_ASYNCFULL	(1U << 1)
53 
54 struct request_list {
55 	struct request_queue	*q;	/* the queue this rl belongs to */
56 #ifdef CONFIG_BLK_CGROUP
57 	struct blkcg_gq		*blkg;	/* blkg this request pool belongs to */
58 #endif
59 	/*
60 	 * count[], starved[], and wait[] are indexed by
61 	 * BLK_RW_SYNC/BLK_RW_ASYNC
62 	 */
63 	int			count[2];
64 	int			starved[2];
65 	mempool_t		*rq_pool;
66 	wait_queue_head_t	wait[2];
67 	unsigned int		flags;
68 };
69 
70 /*
71  * request command types
72  */
73 enum rq_cmd_type_bits {
74 	REQ_TYPE_FS		= 1,	/* fs request */
75 	REQ_TYPE_BLOCK_PC,		/* scsi command */
76 	REQ_TYPE_DRV_PRIV,		/* driver defined types from here */
77 };
78 
79 #define BLK_MAX_CDB	16
80 
81 /*
82  * Try to put the fields that are referenced together in the same cacheline.
83  *
84  * If you modify this structure, make sure to update blk_rq_init() and
85  * especially blk_mq_rq_ctx_init() to take care of the added fields.
86  */
87 struct request {
88 	struct list_head queuelist;
89 	union {
90 		struct call_single_data csd;
91 		unsigned long fifo_time;
92 	};
93 
94 	struct request_queue *q;
95 	struct blk_mq_ctx *mq_ctx;
96 
97 	u64 cmd_flags;
98 	unsigned cmd_type;
99 	unsigned long atomic_flags;
100 
101 	int cpu;
102 
103 	/* the following two fields are internal, NEVER access directly */
104 	unsigned int __data_len;	/* total data len */
105 	sector_t __sector;		/* sector cursor */
106 
107 	struct bio *bio;
108 	struct bio *biotail;
109 
110 	/*
111 	 * The hash is used inside the scheduler, and killed once the
112 	 * request reaches the dispatch list. The ipi_list is only used
113 	 * to queue the request for softirq completion, which is long
114 	 * after the request has been unhashed (and even removed from
115 	 * the dispatch list).
116 	 */
117 	union {
118 		struct hlist_node hash;	/* merge hash */
119 		struct list_head ipi_list;
120 	};
121 
122 	/*
123 	 * The rb_node is only used inside the io scheduler, requests
124 	 * are pruned when moved to the dispatch queue. So let the
125 	 * completion_data share space with the rb_node.
126 	 */
127 	union {
128 		struct rb_node rb_node;	/* sort/lookup */
129 		void *completion_data;
130 	};
131 
132 	/*
133 	 * Three pointers are available for the IO schedulers, if they need
134 	 * more they have to dynamically allocate it.  Flush requests are
135 	 * never put on the IO scheduler. So let the flush fields share
136 	 * space with the elevator data.
137 	 */
138 	union {
139 		struct {
140 			struct io_cq		*icq;
141 			void			*priv[2];
142 		} elv;
143 
144 		struct {
145 			unsigned int		seq;
146 			struct list_head	list;
147 			rq_end_io_fn		*saved_end_io;
148 		} flush;
149 	};
150 
151 	struct gendisk *rq_disk;
152 	struct hd_struct *part;
153 	unsigned long start_time;
154 #ifdef CONFIG_BLK_CGROUP
155 	struct request_list *rl;		/* rl this rq is alloced from */
156 	unsigned long long start_time_ns;
157 	unsigned long long io_start_time_ns;    /* when passed to hardware */
158 #endif
159 	/* Number of scatter-gather DMA addr+len pairs after
160 	 * physical address coalescing is performed.
161 	 */
162 	unsigned short nr_phys_segments;
163 #if defined(CONFIG_BLK_DEV_INTEGRITY)
164 	unsigned short nr_integrity_segments;
165 #endif
166 
167 	unsigned short ioprio;
168 
169 	void *special;		/* opaque pointer available for LLD use */
170 
171 	int tag;
172 	int errors;
173 
174 	/*
175 	 * when request is used as a packet command carrier
176 	 */
177 	unsigned char __cmd[BLK_MAX_CDB];
178 	unsigned char *cmd;
179 	unsigned short cmd_len;
180 
181 	unsigned int extra_len;	/* length of alignment and padding */
182 	unsigned int sense_len;
183 	unsigned int resid_len;	/* residual count */
184 	void *sense;
185 
186 	unsigned long deadline;
187 	struct list_head timeout_list;
188 	unsigned int timeout;
189 	int retries;
190 
191 	/*
192 	 * completion callback.
193 	 */
194 	rq_end_io_fn *end_io;
195 	void *end_io_data;
196 
197 	/* for bidi */
198 	struct request *next_rq;
199 };
200 
201 static inline unsigned short req_get_ioprio(struct request *req)
202 {
203 	return req->ioprio;
204 }
205 
206 #include <linux/elevator.h>
207 
208 struct blk_queue_ctx;
209 
210 typedef void (request_fn_proc) (struct request_queue *q);
211 typedef void (make_request_fn) (struct request_queue *q, struct bio *bio);
212 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
213 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
214 
215 struct bio_vec;
216 typedef void (softirq_done_fn)(struct request *);
217 typedef int (dma_drain_needed_fn)(struct request *);
218 typedef int (lld_busy_fn) (struct request_queue *q);
219 typedef int (bsg_job_fn) (struct bsg_job *);
220 
221 enum blk_eh_timer_return {
222 	BLK_EH_NOT_HANDLED,
223 	BLK_EH_HANDLED,
224 	BLK_EH_RESET_TIMER,
225 };
226 
227 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
228 
229 enum blk_queue_state {
230 	Queue_down,
231 	Queue_up,
232 };
233 
234 struct blk_queue_tag {
235 	struct request **tag_index;	/* map of busy tags */
236 	unsigned long *tag_map;		/* bit map of free/busy tags */
237 	int busy;			/* current depth */
238 	int max_depth;			/* what we will send to device */
239 	int real_max_depth;		/* what the array can hold */
240 	atomic_t refcnt;		/* map can be shared */
241 	int alloc_policy;		/* tag allocation policy */
242 	int next_tag;			/* next tag */
243 };
244 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
245 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
246 
247 #define BLK_SCSI_MAX_CMDS	(256)
248 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
249 
250 struct queue_limits {
251 	unsigned long		bounce_pfn;
252 	unsigned long		seg_boundary_mask;
253 	unsigned long		virt_boundary_mask;
254 
255 	unsigned int		max_hw_sectors;
256 	unsigned int		chunk_sectors;
257 	unsigned int		max_sectors;
258 	unsigned int		max_segment_size;
259 	unsigned int		physical_block_size;
260 	unsigned int		alignment_offset;
261 	unsigned int		io_min;
262 	unsigned int		io_opt;
263 	unsigned int		max_discard_sectors;
264 	unsigned int		max_hw_discard_sectors;
265 	unsigned int		max_write_same_sectors;
266 	unsigned int		discard_granularity;
267 	unsigned int		discard_alignment;
268 
269 	unsigned short		logical_block_size;
270 	unsigned short		max_segments;
271 	unsigned short		max_integrity_segments;
272 
273 	unsigned char		misaligned;
274 	unsigned char		discard_misaligned;
275 	unsigned char		cluster;
276 	unsigned char		discard_zeroes_data;
277 	unsigned char		raid_partial_stripes_expensive;
278 };
279 
280 struct request_queue {
281 	/*
282 	 * Together with queue_head for cacheline sharing
283 	 */
284 	struct list_head	queue_head;
285 	struct request		*last_merge;
286 	struct elevator_queue	*elevator;
287 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
288 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
289 
290 	/*
291 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
292 	 * is used, root blkg allocates from @q->root_rl and all other
293 	 * blkgs from their own blkg->rl.  Which one to use should be
294 	 * determined using bio_request_list().
295 	 */
296 	struct request_list	root_rl;
297 
298 	request_fn_proc		*request_fn;
299 	make_request_fn		*make_request_fn;
300 	prep_rq_fn		*prep_rq_fn;
301 	unprep_rq_fn		*unprep_rq_fn;
302 	softirq_done_fn		*softirq_done_fn;
303 	rq_timed_out_fn		*rq_timed_out_fn;
304 	dma_drain_needed_fn	*dma_drain_needed;
305 	lld_busy_fn		*lld_busy_fn;
306 
307 	struct blk_mq_ops	*mq_ops;
308 
309 	unsigned int		*mq_map;
310 
311 	/* sw queues */
312 	struct blk_mq_ctx __percpu	*queue_ctx;
313 	unsigned int		nr_queues;
314 
315 	/* hw dispatch queues */
316 	struct blk_mq_hw_ctx	**queue_hw_ctx;
317 	unsigned int		nr_hw_queues;
318 
319 	/*
320 	 * Dispatch queue sorting
321 	 */
322 	sector_t		end_sector;
323 	struct request		*boundary_rq;
324 
325 	/*
326 	 * Delayed queue handling
327 	 */
328 	struct delayed_work	delay_work;
329 
330 	struct backing_dev_info	backing_dev_info;
331 
332 	/*
333 	 * The queue owner gets to use this for whatever they like.
334 	 * ll_rw_blk doesn't touch it.
335 	 */
336 	void			*queuedata;
337 
338 	/*
339 	 * various queue flags, see QUEUE_* below
340 	 */
341 	unsigned long		queue_flags;
342 
343 	/*
344 	 * ida allocated id for this queue.  Used to index queues from
345 	 * ioctx.
346 	 */
347 	int			id;
348 
349 	/*
350 	 * queue needs bounce pages for pages above this limit
351 	 */
352 	gfp_t			bounce_gfp;
353 
354 	/*
355 	 * protects queue structures from reentrancy. ->__queue_lock should
356 	 * _never_ be used directly, it is queue private. always use
357 	 * ->queue_lock.
358 	 */
359 	spinlock_t		__queue_lock;
360 	spinlock_t		*queue_lock;
361 
362 	/*
363 	 * queue kobject
364 	 */
365 	struct kobject kobj;
366 
367 	/*
368 	 * mq queue kobject
369 	 */
370 	struct kobject mq_kobj;
371 
372 #ifdef CONFIG_PM
373 	struct device		*dev;
374 	int			rpm_status;
375 	unsigned int		nr_pending;
376 #endif
377 
378 	/*
379 	 * queue settings
380 	 */
381 	unsigned long		nr_requests;	/* Max # of requests */
382 	unsigned int		nr_congestion_on;
383 	unsigned int		nr_congestion_off;
384 	unsigned int		nr_batching;
385 
386 	unsigned int		dma_drain_size;
387 	void			*dma_drain_buffer;
388 	unsigned int		dma_pad_mask;
389 	unsigned int		dma_alignment;
390 
391 	struct blk_queue_tag	*queue_tags;
392 	struct list_head	tag_busy_list;
393 
394 	unsigned int		nr_sorted;
395 	unsigned int		in_flight[2];
396 	/*
397 	 * Number of active block driver functions for which blk_drain_queue()
398 	 * must wait. Must be incremented around functions that unlock the
399 	 * queue_lock internally, e.g. scsi_request_fn().
400 	 */
401 	unsigned int		request_fn_active;
402 
403 	unsigned int		rq_timeout;
404 	struct timer_list	timeout;
405 	struct list_head	timeout_list;
406 
407 	struct list_head	icq_list;
408 #ifdef CONFIG_BLK_CGROUP
409 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
410 	struct blkcg_gq		*root_blkg;
411 	struct list_head	blkg_list;
412 #endif
413 
414 	struct queue_limits	limits;
415 
416 	/*
417 	 * sg stuff
418 	 */
419 	unsigned int		sg_timeout;
420 	unsigned int		sg_reserved_size;
421 	int			node;
422 #ifdef CONFIG_BLK_DEV_IO_TRACE
423 	struct blk_trace	*blk_trace;
424 #endif
425 	/*
426 	 * for flush operations
427 	 */
428 	unsigned int		flush_flags;
429 	unsigned int		flush_not_queueable:1;
430 	struct blk_flush_queue	*fq;
431 
432 	struct list_head	requeue_list;
433 	spinlock_t		requeue_lock;
434 	struct work_struct	requeue_work;
435 
436 	struct mutex		sysfs_lock;
437 
438 	int			bypass_depth;
439 	atomic_t		mq_freeze_depth;
440 
441 #if defined(CONFIG_BLK_DEV_BSG)
442 	bsg_job_fn		*bsg_job_fn;
443 	int			bsg_job_size;
444 	struct bsg_class_device bsg_dev;
445 #endif
446 
447 #ifdef CONFIG_BLK_DEV_THROTTLING
448 	/* Throttle data */
449 	struct throtl_data *td;
450 #endif
451 	struct rcu_head		rcu_head;
452 	wait_queue_head_t	mq_freeze_wq;
453 	struct percpu_ref	mq_usage_counter;
454 	struct list_head	all_q_node;
455 
456 	struct blk_mq_tag_set	*tag_set;
457 	struct list_head	tag_set_list;
458 	struct bio_set		*bio_split;
459 };
460 
461 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
462 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
463 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
464 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
465 #define QUEUE_FLAG_DYING	5	/* queue being torn down */
466 #define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
467 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
468 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
469 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
470 #define QUEUE_FLAG_FAIL_IO     10	/* fake timeout */
471 #define QUEUE_FLAG_STACKABLE   11	/* supports request stacking */
472 #define QUEUE_FLAG_NONROT      12	/* non-rotational device (SSD) */
473 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
474 #define QUEUE_FLAG_IO_STAT     13	/* do IO stats */
475 #define QUEUE_FLAG_DISCARD     14	/* supports DISCARD */
476 #define QUEUE_FLAG_NOXMERGES   15	/* No extended merges */
477 #define QUEUE_FLAG_ADD_RANDOM  16	/* Contributes to random pool */
478 #define QUEUE_FLAG_SECDISCARD  17	/* supports SECDISCARD */
479 #define QUEUE_FLAG_SAME_FORCE  18	/* force complete on same CPU */
480 #define QUEUE_FLAG_DEAD        19	/* queue tear-down finished */
481 #define QUEUE_FLAG_INIT_DONE   20	/* queue is initialized */
482 #define QUEUE_FLAG_NO_SG_MERGE 21	/* don't attempt to merge SG segments*/
483 
484 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
485 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
486 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
487 				 (1 << QUEUE_FLAG_ADD_RANDOM))
488 
489 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
490 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
491 				 (1 << QUEUE_FLAG_SAME_COMP))
492 
493 static inline void queue_lockdep_assert_held(struct request_queue *q)
494 {
495 	if (q->queue_lock)
496 		lockdep_assert_held(q->queue_lock);
497 }
498 
499 static inline void queue_flag_set_unlocked(unsigned int flag,
500 					   struct request_queue *q)
501 {
502 	__set_bit(flag, &q->queue_flags);
503 }
504 
505 static inline int queue_flag_test_and_clear(unsigned int flag,
506 					    struct request_queue *q)
507 {
508 	queue_lockdep_assert_held(q);
509 
510 	if (test_bit(flag, &q->queue_flags)) {
511 		__clear_bit(flag, &q->queue_flags);
512 		return 1;
513 	}
514 
515 	return 0;
516 }
517 
518 static inline int queue_flag_test_and_set(unsigned int flag,
519 					  struct request_queue *q)
520 {
521 	queue_lockdep_assert_held(q);
522 
523 	if (!test_bit(flag, &q->queue_flags)) {
524 		__set_bit(flag, &q->queue_flags);
525 		return 0;
526 	}
527 
528 	return 1;
529 }
530 
531 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
532 {
533 	queue_lockdep_assert_held(q);
534 	__set_bit(flag, &q->queue_flags);
535 }
536 
537 static inline void queue_flag_clear_unlocked(unsigned int flag,
538 					     struct request_queue *q)
539 {
540 	__clear_bit(flag, &q->queue_flags);
541 }
542 
543 static inline int queue_in_flight(struct request_queue *q)
544 {
545 	return q->in_flight[0] + q->in_flight[1];
546 }
547 
548 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
549 {
550 	queue_lockdep_assert_held(q);
551 	__clear_bit(flag, &q->queue_flags);
552 }
553 
554 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
555 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
556 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
557 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
558 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
559 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
560 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
561 #define blk_queue_noxmerges(q)	\
562 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
563 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
564 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
565 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
566 #define blk_queue_stackable(q)	\
567 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
568 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
569 #define blk_queue_secdiscard(q)	(blk_queue_discard(q) && \
570 	test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
571 
572 #define blk_noretry_request(rq) \
573 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
574 			     REQ_FAILFAST_DRIVER))
575 
576 #define blk_account_rq(rq) \
577 	(((rq)->cmd_flags & REQ_STARTED) && \
578 	 ((rq)->cmd_type == REQ_TYPE_FS))
579 
580 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
581 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
582 /* rq->queuelist of dequeued request must be list_empty() */
583 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
584 
585 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
586 
587 #define rq_data_dir(rq)		((int)((rq)->cmd_flags & 1))
588 
589 /*
590  * Driver can handle struct request, if it either has an old style
591  * request_fn defined, or is blk-mq based.
592  */
593 static inline bool queue_is_rq_based(struct request_queue *q)
594 {
595 	return q->request_fn || q->mq_ops;
596 }
597 
598 static inline unsigned int blk_queue_cluster(struct request_queue *q)
599 {
600 	return q->limits.cluster;
601 }
602 
603 /*
604  * We regard a request as sync, if either a read or a sync write
605  */
606 static inline bool rw_is_sync(unsigned int rw_flags)
607 {
608 	return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
609 }
610 
611 static inline bool rq_is_sync(struct request *rq)
612 {
613 	return rw_is_sync(rq->cmd_flags);
614 }
615 
616 static inline bool blk_rl_full(struct request_list *rl, bool sync)
617 {
618 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
619 
620 	return rl->flags & flag;
621 }
622 
623 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
624 {
625 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
626 
627 	rl->flags |= flag;
628 }
629 
630 static inline void blk_clear_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 bool rq_mergeable(struct request *rq)
638 {
639 	if (rq->cmd_type != REQ_TYPE_FS)
640 		return false;
641 
642 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
643 		return false;
644 
645 	return true;
646 }
647 
648 static inline bool blk_check_merge_flags(unsigned int flags1,
649 					 unsigned int flags2)
650 {
651 	if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
652 		return false;
653 
654 	if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
655 		return false;
656 
657 	if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
658 		return false;
659 
660 	return true;
661 }
662 
663 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
664 {
665 	if (bio_data(a) == bio_data(b))
666 		return true;
667 
668 	return false;
669 }
670 
671 /*
672  * q->prep_rq_fn return values
673  */
674 #define BLKPREP_OK		0	/* serve it */
675 #define BLKPREP_KILL		1	/* fatal error, kill */
676 #define BLKPREP_DEFER		2	/* leave on queue */
677 
678 extern unsigned long blk_max_low_pfn, blk_max_pfn;
679 
680 /*
681  * standard bounce addresses:
682  *
683  * BLK_BOUNCE_HIGH	: bounce all highmem pages
684  * BLK_BOUNCE_ANY	: don't bounce anything
685  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
686  */
687 
688 #if BITS_PER_LONG == 32
689 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
690 #else
691 #define BLK_BOUNCE_HIGH		-1ULL
692 #endif
693 #define BLK_BOUNCE_ANY		(-1ULL)
694 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
695 
696 /*
697  * default timeout for SG_IO if none specified
698  */
699 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
700 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
701 
702 #ifdef CONFIG_BOUNCE
703 extern int init_emergency_isa_pool(void);
704 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
705 #else
706 static inline int init_emergency_isa_pool(void)
707 {
708 	return 0;
709 }
710 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
711 {
712 }
713 #endif /* CONFIG_MMU */
714 
715 struct rq_map_data {
716 	struct page **pages;
717 	int page_order;
718 	int nr_entries;
719 	unsigned long offset;
720 	int null_mapped;
721 	int from_user;
722 };
723 
724 struct req_iterator {
725 	struct bvec_iter iter;
726 	struct bio *bio;
727 };
728 
729 /* This should not be used directly - use rq_for_each_segment */
730 #define for_each_bio(_bio)		\
731 	for (; _bio; _bio = _bio->bi_next)
732 #define __rq_for_each_bio(_bio, rq)	\
733 	if ((rq->bio))			\
734 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
735 
736 #define rq_for_each_segment(bvl, _rq, _iter)			\
737 	__rq_for_each_bio(_iter.bio, _rq)			\
738 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
739 
740 #define rq_iter_last(bvec, _iter)				\
741 		(_iter.bio->bi_next == NULL &&			\
742 		 bio_iter_last(bvec, _iter.iter))
743 
744 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
745 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
746 #endif
747 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
748 extern void rq_flush_dcache_pages(struct request *rq);
749 #else
750 static inline void rq_flush_dcache_pages(struct request *rq)
751 {
752 }
753 #endif
754 
755 extern int blk_register_queue(struct gendisk *disk);
756 extern void blk_unregister_queue(struct gendisk *disk);
757 extern void generic_make_request(struct bio *bio);
758 extern void blk_rq_init(struct request_queue *q, struct request *rq);
759 extern void blk_put_request(struct request *);
760 extern void __blk_put_request(struct request_queue *, struct request *);
761 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
762 extern struct request *blk_make_request(struct request_queue *, struct bio *,
763 					gfp_t);
764 extern void blk_rq_set_block_pc(struct request *);
765 extern void blk_requeue_request(struct request_queue *, struct request *);
766 extern void blk_add_request_payload(struct request *rq, struct page *page,
767 		unsigned int len);
768 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
769 extern int blk_lld_busy(struct request_queue *q);
770 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
771 			     struct bio_set *bs, gfp_t gfp_mask,
772 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
773 			     void *data);
774 extern void blk_rq_unprep_clone(struct request *rq);
775 extern int blk_insert_cloned_request(struct request_queue *q,
776 				     struct request *rq);
777 extern void blk_delay_queue(struct request_queue *, unsigned long);
778 extern void blk_queue_split(struct request_queue *, struct bio **,
779 			    struct bio_set *);
780 extern void blk_recount_segments(struct request_queue *, struct bio *);
781 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
782 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
783 			      unsigned int, void __user *);
784 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
785 			  unsigned int, void __user *);
786 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
787 			 struct scsi_ioctl_command __user *);
788 
789 extern void blk_start_queue(struct request_queue *q);
790 extern void blk_stop_queue(struct request_queue *q);
791 extern void blk_sync_queue(struct request_queue *q);
792 extern void __blk_stop_queue(struct request_queue *q);
793 extern void __blk_run_queue(struct request_queue *q);
794 extern void __blk_run_queue_uncond(struct request_queue *q);
795 extern void blk_run_queue(struct request_queue *);
796 extern void blk_run_queue_async(struct request_queue *q);
797 extern int blk_rq_map_user(struct request_queue *, struct request *,
798 			   struct rq_map_data *, void __user *, unsigned long,
799 			   gfp_t);
800 extern int blk_rq_unmap_user(struct bio *);
801 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
802 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
803 			       struct rq_map_data *, const struct iov_iter *,
804 			       gfp_t);
805 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
806 			  struct request *, int);
807 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
808 				  struct request *, int, rq_end_io_fn *);
809 
810 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
811 {
812 	return bdev->bd_disk->queue;	/* this is never NULL */
813 }
814 
815 /*
816  * blk_rq_pos()			: the current sector
817  * blk_rq_bytes()		: bytes left in the entire request
818  * blk_rq_cur_bytes()		: bytes left in the current segment
819  * blk_rq_err_bytes()		: bytes left till the next error boundary
820  * blk_rq_sectors()		: sectors left in the entire request
821  * blk_rq_cur_sectors()		: sectors left in the current segment
822  */
823 static inline sector_t blk_rq_pos(const struct request *rq)
824 {
825 	return rq->__sector;
826 }
827 
828 static inline unsigned int blk_rq_bytes(const struct request *rq)
829 {
830 	return rq->__data_len;
831 }
832 
833 static inline int blk_rq_cur_bytes(const struct request *rq)
834 {
835 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
836 }
837 
838 extern unsigned int blk_rq_err_bytes(const struct request *rq);
839 
840 static inline unsigned int blk_rq_sectors(const struct request *rq)
841 {
842 	return blk_rq_bytes(rq) >> 9;
843 }
844 
845 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
846 {
847 	return blk_rq_cur_bytes(rq) >> 9;
848 }
849 
850 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
851 						     unsigned int cmd_flags)
852 {
853 	if (unlikely(cmd_flags & REQ_DISCARD))
854 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
855 
856 	if (unlikely(cmd_flags & REQ_WRITE_SAME))
857 		return q->limits.max_write_same_sectors;
858 
859 	return q->limits.max_sectors;
860 }
861 
862 /*
863  * Return maximum size of a request at given offset. Only valid for
864  * file system requests.
865  */
866 static inline unsigned int blk_max_size_offset(struct request_queue *q,
867 					       sector_t offset)
868 {
869 	if (!q->limits.chunk_sectors)
870 		return q->limits.max_sectors;
871 
872 	return q->limits.chunk_sectors -
873 			(offset & (q->limits.chunk_sectors - 1));
874 }
875 
876 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
877 {
878 	struct request_queue *q = rq->q;
879 
880 	if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
881 		return q->limits.max_hw_sectors;
882 
883 	if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
884 		return blk_queue_get_max_sectors(q, rq->cmd_flags);
885 
886 	return min(blk_max_size_offset(q, blk_rq_pos(rq)),
887 			blk_queue_get_max_sectors(q, rq->cmd_flags));
888 }
889 
890 static inline unsigned int blk_rq_count_bios(struct request *rq)
891 {
892 	unsigned int nr_bios = 0;
893 	struct bio *bio;
894 
895 	__rq_for_each_bio(bio, rq)
896 		nr_bios++;
897 
898 	return nr_bios;
899 }
900 
901 /*
902  * Request issue related functions.
903  */
904 extern struct request *blk_peek_request(struct request_queue *q);
905 extern void blk_start_request(struct request *rq);
906 extern struct request *blk_fetch_request(struct request_queue *q);
907 
908 /*
909  * Request completion related functions.
910  *
911  * blk_update_request() completes given number of bytes and updates
912  * the request without completing it.
913  *
914  * blk_end_request() and friends.  __blk_end_request() must be called
915  * with the request queue spinlock acquired.
916  *
917  * Several drivers define their own end_request and call
918  * blk_end_request() for parts of the original function.
919  * This prevents code duplication in drivers.
920  */
921 extern bool blk_update_request(struct request *rq, int error,
922 			       unsigned int nr_bytes);
923 extern void blk_finish_request(struct request *rq, int error);
924 extern bool blk_end_request(struct request *rq, int error,
925 			    unsigned int nr_bytes);
926 extern void blk_end_request_all(struct request *rq, int error);
927 extern bool blk_end_request_cur(struct request *rq, int error);
928 extern bool blk_end_request_err(struct request *rq, int error);
929 extern bool __blk_end_request(struct request *rq, int error,
930 			      unsigned int nr_bytes);
931 extern void __blk_end_request_all(struct request *rq, int error);
932 extern bool __blk_end_request_cur(struct request *rq, int error);
933 extern bool __blk_end_request_err(struct request *rq, int error);
934 
935 extern void blk_complete_request(struct request *);
936 extern void __blk_complete_request(struct request *);
937 extern void blk_abort_request(struct request *);
938 extern void blk_unprep_request(struct request *);
939 
940 /*
941  * Access functions for manipulating queue properties
942  */
943 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
944 					spinlock_t *lock, int node_id);
945 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
946 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
947 						      request_fn_proc *, spinlock_t *);
948 extern void blk_cleanup_queue(struct request_queue *);
949 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
950 extern void blk_queue_bounce_limit(struct request_queue *, u64);
951 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
952 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
953 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
954 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
955 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
956 extern void blk_queue_max_discard_sectors(struct request_queue *q,
957 		unsigned int max_discard_sectors);
958 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
959 		unsigned int max_write_same_sectors);
960 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
961 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
962 extern void blk_queue_alignment_offset(struct request_queue *q,
963 				       unsigned int alignment);
964 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
965 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
966 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
967 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
968 extern void blk_set_default_limits(struct queue_limits *lim);
969 extern void blk_set_stacking_limits(struct queue_limits *lim);
970 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
971 			    sector_t offset);
972 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
973 			    sector_t offset);
974 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
975 			      sector_t offset);
976 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
977 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
978 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
979 extern int blk_queue_dma_drain(struct request_queue *q,
980 			       dma_drain_needed_fn *dma_drain_needed,
981 			       void *buf, unsigned int size);
982 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
983 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
984 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
985 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
986 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
987 extern void blk_queue_dma_alignment(struct request_queue *, int);
988 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
989 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
990 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
991 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
992 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
993 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
994 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
995 
996 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
997 extern void blk_dump_rq_flags(struct request *, char *);
998 extern long nr_blockdev_pages(void);
999 
1000 bool __must_check blk_get_queue(struct request_queue *);
1001 struct request_queue *blk_alloc_queue(gfp_t);
1002 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1003 extern void blk_put_queue(struct request_queue *);
1004 extern void blk_set_queue_dying(struct request_queue *);
1005 
1006 /*
1007  * block layer runtime pm functions
1008  */
1009 #ifdef CONFIG_PM
1010 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1011 extern int blk_pre_runtime_suspend(struct request_queue *q);
1012 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1013 extern void blk_pre_runtime_resume(struct request_queue *q);
1014 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1015 #else
1016 static inline void blk_pm_runtime_init(struct request_queue *q,
1017 	struct device *dev) {}
1018 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1019 {
1020 	return -ENOSYS;
1021 }
1022 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1023 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1024 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1025 #endif
1026 
1027 /*
1028  * blk_plug permits building a queue of related requests by holding the I/O
1029  * fragments for a short period. This allows merging of sequential requests
1030  * into single larger request. As the requests are moved from a per-task list to
1031  * the device's request_queue in a batch, this results in improved scalability
1032  * as the lock contention for request_queue lock is reduced.
1033  *
1034  * It is ok not to disable preemption when adding the request to the plug list
1035  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1036  * the plug list when the task sleeps by itself. For details, please see
1037  * schedule() where blk_schedule_flush_plug() is called.
1038  */
1039 struct blk_plug {
1040 	struct list_head list; /* requests */
1041 	struct list_head mq_list; /* blk-mq requests */
1042 	struct list_head cb_list; /* md requires an unplug callback */
1043 };
1044 #define BLK_MAX_REQUEST_COUNT 16
1045 
1046 struct blk_plug_cb;
1047 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1048 struct blk_plug_cb {
1049 	struct list_head list;
1050 	blk_plug_cb_fn callback;
1051 	void *data;
1052 };
1053 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1054 					     void *data, int size);
1055 extern void blk_start_plug(struct blk_plug *);
1056 extern void blk_finish_plug(struct blk_plug *);
1057 extern void blk_flush_plug_list(struct blk_plug *, bool);
1058 
1059 static inline void blk_flush_plug(struct task_struct *tsk)
1060 {
1061 	struct blk_plug *plug = tsk->plug;
1062 
1063 	if (plug)
1064 		blk_flush_plug_list(plug, false);
1065 }
1066 
1067 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1068 {
1069 	struct blk_plug *plug = tsk->plug;
1070 
1071 	if (plug)
1072 		blk_flush_plug_list(plug, true);
1073 }
1074 
1075 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1076 {
1077 	struct blk_plug *plug = tsk->plug;
1078 
1079 	return plug &&
1080 		(!list_empty(&plug->list) ||
1081 		 !list_empty(&plug->mq_list) ||
1082 		 !list_empty(&plug->cb_list));
1083 }
1084 
1085 /*
1086  * tag stuff
1087  */
1088 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1089 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1090 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1091 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1092 extern void blk_queue_free_tags(struct request_queue *);
1093 extern int blk_queue_resize_tags(struct request_queue *, int);
1094 extern void blk_queue_invalidate_tags(struct request_queue *);
1095 extern struct blk_queue_tag *blk_init_tags(int, int);
1096 extern void blk_free_tags(struct blk_queue_tag *);
1097 
1098 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1099 						int tag)
1100 {
1101 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1102 		return NULL;
1103 	return bqt->tag_index[tag];
1104 }
1105 
1106 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1107 
1108 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1109 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1110 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1111 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1112 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1113 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1114 		sector_t nr_sects, gfp_t gfp_mask, bool discard);
1115 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1116 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1117 {
1118 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1119 				    nr_blocks << (sb->s_blocksize_bits - 9),
1120 				    gfp_mask, flags);
1121 }
1122 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1123 		sector_t nr_blocks, gfp_t gfp_mask)
1124 {
1125 	return blkdev_issue_zeroout(sb->s_bdev,
1126 				    block << (sb->s_blocksize_bits - 9),
1127 				    nr_blocks << (sb->s_blocksize_bits - 9),
1128 				    gfp_mask, true);
1129 }
1130 
1131 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1132 
1133 enum blk_default_limits {
1134 	BLK_MAX_SEGMENTS	= 128,
1135 	BLK_SAFE_MAX_SECTORS	= 255,
1136 	BLK_DEF_MAX_SECTORS	= 2560,
1137 	BLK_MAX_SEGMENT_SIZE	= 65536,
1138 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1139 };
1140 
1141 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1142 
1143 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1144 {
1145 	return q->limits.bounce_pfn;
1146 }
1147 
1148 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1149 {
1150 	return q->limits.seg_boundary_mask;
1151 }
1152 
1153 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1154 {
1155 	return q->limits.virt_boundary_mask;
1156 }
1157 
1158 static inline unsigned int queue_max_sectors(struct request_queue *q)
1159 {
1160 	return q->limits.max_sectors;
1161 }
1162 
1163 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1164 {
1165 	return q->limits.max_hw_sectors;
1166 }
1167 
1168 static inline unsigned short queue_max_segments(struct request_queue *q)
1169 {
1170 	return q->limits.max_segments;
1171 }
1172 
1173 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1174 {
1175 	return q->limits.max_segment_size;
1176 }
1177 
1178 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1179 {
1180 	int retval = 512;
1181 
1182 	if (q && q->limits.logical_block_size)
1183 		retval = q->limits.logical_block_size;
1184 
1185 	return retval;
1186 }
1187 
1188 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1189 {
1190 	return queue_logical_block_size(bdev_get_queue(bdev));
1191 }
1192 
1193 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1194 {
1195 	return q->limits.physical_block_size;
1196 }
1197 
1198 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1199 {
1200 	return queue_physical_block_size(bdev_get_queue(bdev));
1201 }
1202 
1203 static inline unsigned int queue_io_min(struct request_queue *q)
1204 {
1205 	return q->limits.io_min;
1206 }
1207 
1208 static inline int bdev_io_min(struct block_device *bdev)
1209 {
1210 	return queue_io_min(bdev_get_queue(bdev));
1211 }
1212 
1213 static inline unsigned int queue_io_opt(struct request_queue *q)
1214 {
1215 	return q->limits.io_opt;
1216 }
1217 
1218 static inline int bdev_io_opt(struct block_device *bdev)
1219 {
1220 	return queue_io_opt(bdev_get_queue(bdev));
1221 }
1222 
1223 static inline int queue_alignment_offset(struct request_queue *q)
1224 {
1225 	if (q->limits.misaligned)
1226 		return -1;
1227 
1228 	return q->limits.alignment_offset;
1229 }
1230 
1231 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1232 {
1233 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1234 	unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1235 
1236 	return (granularity + lim->alignment_offset - alignment) % granularity;
1237 }
1238 
1239 static inline int bdev_alignment_offset(struct block_device *bdev)
1240 {
1241 	struct request_queue *q = bdev_get_queue(bdev);
1242 
1243 	if (q->limits.misaligned)
1244 		return -1;
1245 
1246 	if (bdev != bdev->bd_contains)
1247 		return bdev->bd_part->alignment_offset;
1248 
1249 	return q->limits.alignment_offset;
1250 }
1251 
1252 static inline int queue_discard_alignment(struct request_queue *q)
1253 {
1254 	if (q->limits.discard_misaligned)
1255 		return -1;
1256 
1257 	return q->limits.discard_alignment;
1258 }
1259 
1260 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1261 {
1262 	unsigned int alignment, granularity, offset;
1263 
1264 	if (!lim->max_discard_sectors)
1265 		return 0;
1266 
1267 	/* Why are these in bytes, not sectors? */
1268 	alignment = lim->discard_alignment >> 9;
1269 	granularity = lim->discard_granularity >> 9;
1270 	if (!granularity)
1271 		return 0;
1272 
1273 	/* Offset of the partition start in 'granularity' sectors */
1274 	offset = sector_div(sector, granularity);
1275 
1276 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1277 	offset = (granularity + alignment - offset) % granularity;
1278 
1279 	/* Turn it back into bytes, gaah */
1280 	return offset << 9;
1281 }
1282 
1283 static inline int bdev_discard_alignment(struct block_device *bdev)
1284 {
1285 	struct request_queue *q = bdev_get_queue(bdev);
1286 
1287 	if (bdev != bdev->bd_contains)
1288 		return bdev->bd_part->discard_alignment;
1289 
1290 	return q->limits.discard_alignment;
1291 }
1292 
1293 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1294 {
1295 	if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1296 		return 1;
1297 
1298 	return 0;
1299 }
1300 
1301 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1302 {
1303 	return queue_discard_zeroes_data(bdev_get_queue(bdev));
1304 }
1305 
1306 static inline unsigned int bdev_write_same(struct block_device *bdev)
1307 {
1308 	struct request_queue *q = bdev_get_queue(bdev);
1309 
1310 	if (q)
1311 		return q->limits.max_write_same_sectors;
1312 
1313 	return 0;
1314 }
1315 
1316 static inline int queue_dma_alignment(struct request_queue *q)
1317 {
1318 	return q ? q->dma_alignment : 511;
1319 }
1320 
1321 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1322 				 unsigned int len)
1323 {
1324 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1325 	return !(addr & alignment) && !(len & alignment);
1326 }
1327 
1328 /* assumes size > 256 */
1329 static inline unsigned int blksize_bits(unsigned int size)
1330 {
1331 	unsigned int bits = 8;
1332 	do {
1333 		bits++;
1334 		size >>= 1;
1335 	} while (size > 256);
1336 	return bits;
1337 }
1338 
1339 static inline unsigned int block_size(struct block_device *bdev)
1340 {
1341 	return bdev->bd_block_size;
1342 }
1343 
1344 static inline bool queue_flush_queueable(struct request_queue *q)
1345 {
1346 	return !q->flush_not_queueable;
1347 }
1348 
1349 typedef struct {struct page *v;} Sector;
1350 
1351 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1352 
1353 static inline void put_dev_sector(Sector p)
1354 {
1355 	page_cache_release(p.v);
1356 }
1357 
1358 /*
1359  * Check if adding a bio_vec after bprv with offset would create a gap in
1360  * the SG list. Most drivers don't care about this, but some do.
1361  */
1362 static inline bool bvec_gap_to_prev(struct request_queue *q,
1363 				struct bio_vec *bprv, unsigned int offset)
1364 {
1365 	if (!queue_virt_boundary(q))
1366 		return false;
1367 	return offset ||
1368 		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1369 }
1370 
1371 struct work_struct;
1372 int kblockd_schedule_work(struct work_struct *work);
1373 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1374 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1375 
1376 #ifdef CONFIG_BLK_CGROUP
1377 /*
1378  * This should not be using sched_clock(). A real patch is in progress
1379  * to fix this up, until that is in place we need to disable preemption
1380  * around sched_clock() in this function and set_io_start_time_ns().
1381  */
1382 static inline void set_start_time_ns(struct request *req)
1383 {
1384 	preempt_disable();
1385 	req->start_time_ns = sched_clock();
1386 	preempt_enable();
1387 }
1388 
1389 static inline void set_io_start_time_ns(struct request *req)
1390 {
1391 	preempt_disable();
1392 	req->io_start_time_ns = sched_clock();
1393 	preempt_enable();
1394 }
1395 
1396 static inline uint64_t rq_start_time_ns(struct request *req)
1397 {
1398         return req->start_time_ns;
1399 }
1400 
1401 static inline uint64_t rq_io_start_time_ns(struct request *req)
1402 {
1403         return req->io_start_time_ns;
1404 }
1405 #else
1406 static inline void set_start_time_ns(struct request *req) {}
1407 static inline void set_io_start_time_ns(struct request *req) {}
1408 static inline uint64_t rq_start_time_ns(struct request *req)
1409 {
1410 	return 0;
1411 }
1412 static inline uint64_t rq_io_start_time_ns(struct request *req)
1413 {
1414 	return 0;
1415 }
1416 #endif
1417 
1418 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1419 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1420 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1421 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1422 
1423 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1424 
1425 enum blk_integrity_flags {
1426 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1427 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1428 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1429 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1430 };
1431 
1432 struct blk_integrity_iter {
1433 	void			*prot_buf;
1434 	void			*data_buf;
1435 	sector_t		seed;
1436 	unsigned int		data_size;
1437 	unsigned short		interval;
1438 	const char		*disk_name;
1439 };
1440 
1441 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1442 
1443 struct blk_integrity {
1444 	integrity_processing_fn	*generate_fn;
1445 	integrity_processing_fn	*verify_fn;
1446 
1447 	unsigned short		flags;
1448 	unsigned short		tuple_size;
1449 	unsigned short		interval;
1450 	unsigned short		tag_size;
1451 
1452 	const char		*name;
1453 
1454 	struct kobject		kobj;
1455 };
1456 
1457 extern bool blk_integrity_is_initialized(struct gendisk *);
1458 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1459 extern void blk_integrity_unregister(struct gendisk *);
1460 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1461 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1462 				   struct scatterlist *);
1463 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1464 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1465 				   struct request *);
1466 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1467 				    struct bio *);
1468 
1469 static inline
1470 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1471 {
1472 	return bdev->bd_disk->integrity;
1473 }
1474 
1475 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1476 {
1477 	return disk->integrity;
1478 }
1479 
1480 static inline bool blk_integrity_rq(struct request *rq)
1481 {
1482 	return rq->cmd_flags & REQ_INTEGRITY;
1483 }
1484 
1485 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1486 						    unsigned int segs)
1487 {
1488 	q->limits.max_integrity_segments = segs;
1489 }
1490 
1491 static inline unsigned short
1492 queue_max_integrity_segments(struct request_queue *q)
1493 {
1494 	return q->limits.max_integrity_segments;
1495 }
1496 
1497 #else /* CONFIG_BLK_DEV_INTEGRITY */
1498 
1499 struct bio;
1500 struct block_device;
1501 struct gendisk;
1502 struct blk_integrity;
1503 
1504 static inline int blk_integrity_rq(struct request *rq)
1505 {
1506 	return 0;
1507 }
1508 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1509 					    struct bio *b)
1510 {
1511 	return 0;
1512 }
1513 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1514 					  struct bio *b,
1515 					  struct scatterlist *s)
1516 {
1517 	return 0;
1518 }
1519 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1520 {
1521 	return NULL;
1522 }
1523 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1524 {
1525 	return NULL;
1526 }
1527 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1528 {
1529 	return 0;
1530 }
1531 static inline int blk_integrity_register(struct gendisk *d,
1532 					 struct blk_integrity *b)
1533 {
1534 	return 0;
1535 }
1536 static inline void blk_integrity_unregister(struct gendisk *d)
1537 {
1538 }
1539 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1540 						    unsigned int segs)
1541 {
1542 }
1543 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1544 {
1545 	return 0;
1546 }
1547 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1548 					  struct request *r1,
1549 					  struct request *r2)
1550 {
1551 	return true;
1552 }
1553 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1554 					   struct request *r,
1555 					   struct bio *b)
1556 {
1557 	return true;
1558 }
1559 static inline bool blk_integrity_is_initialized(struct gendisk *g)
1560 {
1561 	return 0;
1562 }
1563 
1564 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1565 
1566 struct block_device_operations {
1567 	int (*open) (struct block_device *, fmode_t);
1568 	void (*release) (struct gendisk *, fmode_t);
1569 	int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1570 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1571 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1572 	long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1573 			unsigned long *pfn);
1574 	unsigned int (*check_events) (struct gendisk *disk,
1575 				      unsigned int clearing);
1576 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
1577 	int (*media_changed) (struct gendisk *);
1578 	void (*unlock_native_capacity) (struct gendisk *);
1579 	int (*revalidate_disk) (struct gendisk *);
1580 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1581 	/* this callback is with swap_lock and sometimes page table lock held */
1582 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1583 	struct module *owner;
1584 };
1585 
1586 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1587 				 unsigned long);
1588 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1589 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1590 						struct writeback_control *);
1591 extern long bdev_direct_access(struct block_device *, sector_t,
1592 		void __pmem **addr, unsigned long *pfn, long size);
1593 #else /* CONFIG_BLOCK */
1594 
1595 struct block_device;
1596 
1597 /*
1598  * stubs for when the block layer is configured out
1599  */
1600 #define buffer_heads_over_limit 0
1601 
1602 static inline long nr_blockdev_pages(void)
1603 {
1604 	return 0;
1605 }
1606 
1607 struct blk_plug {
1608 };
1609 
1610 static inline void blk_start_plug(struct blk_plug *plug)
1611 {
1612 }
1613 
1614 static inline void blk_finish_plug(struct blk_plug *plug)
1615 {
1616 }
1617 
1618 static inline void blk_flush_plug(struct task_struct *task)
1619 {
1620 }
1621 
1622 static inline void blk_schedule_flush_plug(struct task_struct *task)
1623 {
1624 }
1625 
1626 
1627 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1628 {
1629 	return false;
1630 }
1631 
1632 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1633 				     sector_t *error_sector)
1634 {
1635 	return 0;
1636 }
1637 
1638 #endif /* CONFIG_BLOCK */
1639 
1640 #endif
1641