xref: /linux-6.15/include/linux/blkdev.h (revision bb970707)
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/pfn.h>
19 #include <linux/bio.h>
20 #include <linux/stringify.h>
21 #include <linux/gfp.h>
22 #include <linux/bsg.h>
23 #include <linux/smp.h>
24 #include <linux/rcupdate.h>
25 #include <linux/percpu-refcount.h>
26 #include <linux/scatterlist.h>
27 
28 struct module;
29 struct scsi_ioctl_command;
30 
31 struct request_queue;
32 struct elevator_queue;
33 struct blk_trace;
34 struct request;
35 struct sg_io_hdr;
36 struct bsg_job;
37 struct blkcg_gq;
38 struct blk_flush_queue;
39 struct pr_ops;
40 
41 #define BLKDEV_MIN_RQ	4
42 #define BLKDEV_MAX_RQ	128	/* Default maximum */
43 
44 /*
45  * Maximum number of blkcg policies allowed to be registered concurrently.
46  * Defined here to simplify include dependency.
47  */
48 #define BLKCG_MAX_POLS		2
49 
50 struct request;
51 typedef void (rq_end_io_fn)(struct request *, int);
52 
53 #define BLK_RL_SYNCFULL		(1U << 0)
54 #define BLK_RL_ASYNCFULL	(1U << 1)
55 
56 struct request_list {
57 	struct request_queue	*q;	/* the queue this rl belongs to */
58 #ifdef CONFIG_BLK_CGROUP
59 	struct blkcg_gq		*blkg;	/* blkg this request pool belongs to */
60 #endif
61 	/*
62 	 * count[], starved[], and wait[] are indexed by
63 	 * BLK_RW_SYNC/BLK_RW_ASYNC
64 	 */
65 	int			count[2];
66 	int			starved[2];
67 	mempool_t		*rq_pool;
68 	wait_queue_head_t	wait[2];
69 	unsigned int		flags;
70 };
71 
72 /*
73  * request command types
74  */
75 enum rq_cmd_type_bits {
76 	REQ_TYPE_FS		= 1,	/* fs request */
77 	REQ_TYPE_BLOCK_PC,		/* scsi command */
78 	REQ_TYPE_DRV_PRIV,		/* driver defined types from here */
79 };
80 
81 #define BLK_MAX_CDB	16
82 
83 /*
84  * Try to put the fields that are referenced together in the same cacheline.
85  *
86  * If you modify this structure, make sure to update blk_rq_init() and
87  * especially blk_mq_rq_ctx_init() to take care of the added fields.
88  */
89 struct request {
90 	struct list_head queuelist;
91 	union {
92 		struct call_single_data csd;
93 		unsigned long fifo_time;
94 	};
95 
96 	struct request_queue *q;
97 	struct blk_mq_ctx *mq_ctx;
98 
99 	u64 cmd_flags;
100 	unsigned cmd_type;
101 	unsigned long atomic_flags;
102 
103 	int cpu;
104 
105 	/* the following two fields are internal, NEVER access directly */
106 	unsigned int __data_len;	/* total data len */
107 	sector_t __sector;		/* sector cursor */
108 
109 	struct bio *bio;
110 	struct bio *biotail;
111 
112 	/*
113 	 * The hash is used inside the scheduler, and killed once the
114 	 * request reaches the dispatch list. The ipi_list is only used
115 	 * to queue the request for softirq completion, which is long
116 	 * after the request has been unhashed (and even removed from
117 	 * the dispatch list).
118 	 */
119 	union {
120 		struct hlist_node hash;	/* merge hash */
121 		struct list_head ipi_list;
122 	};
123 
124 	/*
125 	 * The rb_node is only used inside the io scheduler, requests
126 	 * are pruned when moved to the dispatch queue. So let the
127 	 * completion_data share space with the rb_node.
128 	 */
129 	union {
130 		struct rb_node rb_node;	/* sort/lookup */
131 		void *completion_data;
132 	};
133 
134 	/*
135 	 * Three pointers are available for the IO schedulers, if they need
136 	 * more they have to dynamically allocate it.  Flush requests are
137 	 * never put on the IO scheduler. So let the flush fields share
138 	 * space with the elevator data.
139 	 */
140 	union {
141 		struct {
142 			struct io_cq		*icq;
143 			void			*priv[2];
144 		} elv;
145 
146 		struct {
147 			unsigned int		seq;
148 			struct list_head	list;
149 			rq_end_io_fn		*saved_end_io;
150 		} flush;
151 	};
152 
153 	struct gendisk *rq_disk;
154 	struct hd_struct *part;
155 	unsigned long start_time;
156 #ifdef CONFIG_BLK_CGROUP
157 	struct request_list *rl;		/* rl this rq is alloced from */
158 	unsigned long long start_time_ns;
159 	unsigned long long io_start_time_ns;    /* when passed to hardware */
160 #endif
161 	/* Number of scatter-gather DMA addr+len pairs after
162 	 * physical address coalescing is performed.
163 	 */
164 	unsigned short nr_phys_segments;
165 #if defined(CONFIG_BLK_DEV_INTEGRITY)
166 	unsigned short nr_integrity_segments;
167 #endif
168 
169 	unsigned short ioprio;
170 
171 	void *special;		/* opaque pointer available for LLD use */
172 
173 	int tag;
174 	int errors;
175 
176 	/*
177 	 * when request is used as a packet command carrier
178 	 */
179 	unsigned char __cmd[BLK_MAX_CDB];
180 	unsigned char *cmd;
181 	unsigned short cmd_len;
182 
183 	unsigned int extra_len;	/* length of alignment and padding */
184 	unsigned int sense_len;
185 	unsigned int resid_len;	/* residual count */
186 	void *sense;
187 
188 	unsigned long deadline;
189 	struct list_head timeout_list;
190 	unsigned int timeout;
191 	int retries;
192 
193 	/*
194 	 * completion callback.
195 	 */
196 	rq_end_io_fn *end_io;
197 	void *end_io_data;
198 
199 	/* for bidi */
200 	struct request *next_rq;
201 };
202 
203 static inline unsigned short req_get_ioprio(struct request *req)
204 {
205 	return req->ioprio;
206 }
207 
208 #include <linux/elevator.h>
209 
210 struct blk_queue_ctx;
211 
212 typedef void (request_fn_proc) (struct request_queue *q);
213 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
214 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
215 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
216 
217 struct bio_vec;
218 typedef void (softirq_done_fn)(struct request *);
219 typedef int (dma_drain_needed_fn)(struct request *);
220 typedef int (lld_busy_fn) (struct request_queue *q);
221 typedef int (bsg_job_fn) (struct bsg_job *);
222 
223 enum blk_eh_timer_return {
224 	BLK_EH_NOT_HANDLED,
225 	BLK_EH_HANDLED,
226 	BLK_EH_RESET_TIMER,
227 };
228 
229 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
230 
231 enum blk_queue_state {
232 	Queue_down,
233 	Queue_up,
234 };
235 
236 struct blk_queue_tag {
237 	struct request **tag_index;	/* map of busy tags */
238 	unsigned long *tag_map;		/* bit map of free/busy tags */
239 	int busy;			/* current depth */
240 	int max_depth;			/* what we will send to device */
241 	int real_max_depth;		/* what the array can hold */
242 	atomic_t refcnt;		/* map can be shared */
243 	int alloc_policy;		/* tag allocation policy */
244 	int next_tag;			/* next tag */
245 };
246 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
247 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
248 
249 #define BLK_SCSI_MAX_CMDS	(256)
250 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
251 
252 struct queue_limits {
253 	unsigned long		bounce_pfn;
254 	unsigned long		seg_boundary_mask;
255 	unsigned long		virt_boundary_mask;
256 
257 	unsigned int		max_hw_sectors;
258 	unsigned int		max_dev_sectors;
259 	unsigned int		chunk_sectors;
260 	unsigned int		max_sectors;
261 	unsigned int		max_segment_size;
262 	unsigned int		physical_block_size;
263 	unsigned int		alignment_offset;
264 	unsigned int		io_min;
265 	unsigned int		io_opt;
266 	unsigned int		max_discard_sectors;
267 	unsigned int		max_hw_discard_sectors;
268 	unsigned int		max_write_same_sectors;
269 	unsigned int		discard_granularity;
270 	unsigned int		discard_alignment;
271 
272 	unsigned short		logical_block_size;
273 	unsigned short		max_segments;
274 	unsigned short		max_integrity_segments;
275 
276 	unsigned char		misaligned;
277 	unsigned char		discard_misaligned;
278 	unsigned char		cluster;
279 	unsigned char		discard_zeroes_data;
280 	unsigned char		raid_partial_stripes_expensive;
281 };
282 
283 struct request_queue {
284 	/*
285 	 * Together with queue_head for cacheline sharing
286 	 */
287 	struct list_head	queue_head;
288 	struct request		*last_merge;
289 	struct elevator_queue	*elevator;
290 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
291 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
292 
293 	/*
294 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
295 	 * is used, root blkg allocates from @q->root_rl and all other
296 	 * blkgs from their own blkg->rl.  Which one to use should be
297 	 * determined using bio_request_list().
298 	 */
299 	struct request_list	root_rl;
300 
301 	request_fn_proc		*request_fn;
302 	make_request_fn		*make_request_fn;
303 	prep_rq_fn		*prep_rq_fn;
304 	unprep_rq_fn		*unprep_rq_fn;
305 	softirq_done_fn		*softirq_done_fn;
306 	rq_timed_out_fn		*rq_timed_out_fn;
307 	dma_drain_needed_fn	*dma_drain_needed;
308 	lld_busy_fn		*lld_busy_fn;
309 
310 	struct blk_mq_ops	*mq_ops;
311 
312 	unsigned int		*mq_map;
313 
314 	/* sw queues */
315 	struct blk_mq_ctx __percpu	*queue_ctx;
316 	unsigned int		nr_queues;
317 
318 	/* hw dispatch queues */
319 	struct blk_mq_hw_ctx	**queue_hw_ctx;
320 	unsigned int		nr_hw_queues;
321 
322 	/*
323 	 * Dispatch queue sorting
324 	 */
325 	sector_t		end_sector;
326 	struct request		*boundary_rq;
327 
328 	/*
329 	 * Delayed queue handling
330 	 */
331 	struct delayed_work	delay_work;
332 
333 	struct backing_dev_info	backing_dev_info;
334 
335 	/*
336 	 * The queue owner gets to use this for whatever they like.
337 	 * ll_rw_blk doesn't touch it.
338 	 */
339 	void			*queuedata;
340 
341 	/*
342 	 * various queue flags, see QUEUE_* below
343 	 */
344 	unsigned long		queue_flags;
345 
346 	/*
347 	 * ida allocated id for this queue.  Used to index queues from
348 	 * ioctx.
349 	 */
350 	int			id;
351 
352 	/*
353 	 * queue needs bounce pages for pages above this limit
354 	 */
355 	gfp_t			bounce_gfp;
356 
357 	/*
358 	 * protects queue structures from reentrancy. ->__queue_lock should
359 	 * _never_ be used directly, it is queue private. always use
360 	 * ->queue_lock.
361 	 */
362 	spinlock_t		__queue_lock;
363 	spinlock_t		*queue_lock;
364 
365 	/*
366 	 * queue kobject
367 	 */
368 	struct kobject kobj;
369 
370 	/*
371 	 * mq queue kobject
372 	 */
373 	struct kobject mq_kobj;
374 
375 #ifdef  CONFIG_BLK_DEV_INTEGRITY
376 	struct blk_integrity integrity;
377 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
378 
379 #ifdef CONFIG_PM
380 	struct device		*dev;
381 	int			rpm_status;
382 	unsigned int		nr_pending;
383 #endif
384 
385 	/*
386 	 * queue settings
387 	 */
388 	unsigned long		nr_requests;	/* Max # of requests */
389 	unsigned int		nr_congestion_on;
390 	unsigned int		nr_congestion_off;
391 	unsigned int		nr_batching;
392 
393 	unsigned int		dma_drain_size;
394 	void			*dma_drain_buffer;
395 	unsigned int		dma_pad_mask;
396 	unsigned int		dma_alignment;
397 
398 	struct blk_queue_tag	*queue_tags;
399 	struct list_head	tag_busy_list;
400 
401 	unsigned int		nr_sorted;
402 	unsigned int		in_flight[2];
403 	/*
404 	 * Number of active block driver functions for which blk_drain_queue()
405 	 * must wait. Must be incremented around functions that unlock the
406 	 * queue_lock internally, e.g. scsi_request_fn().
407 	 */
408 	unsigned int		request_fn_active;
409 
410 	unsigned int		rq_timeout;
411 	struct timer_list	timeout;
412 	struct work_struct	timeout_work;
413 	struct list_head	timeout_list;
414 
415 	struct list_head	icq_list;
416 #ifdef CONFIG_BLK_CGROUP
417 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
418 	struct blkcg_gq		*root_blkg;
419 	struct list_head	blkg_list;
420 #endif
421 
422 	struct queue_limits	limits;
423 
424 	/*
425 	 * sg stuff
426 	 */
427 	unsigned int		sg_timeout;
428 	unsigned int		sg_reserved_size;
429 	int			node;
430 #ifdef CONFIG_BLK_DEV_IO_TRACE
431 	struct blk_trace	*blk_trace;
432 #endif
433 	/*
434 	 * for flush operations
435 	 */
436 	struct blk_flush_queue	*fq;
437 
438 	struct list_head	requeue_list;
439 	spinlock_t		requeue_lock;
440 	struct work_struct	requeue_work;
441 
442 	struct mutex		sysfs_lock;
443 
444 	int			bypass_depth;
445 	atomic_t		mq_freeze_depth;
446 
447 #if defined(CONFIG_BLK_DEV_BSG)
448 	bsg_job_fn		*bsg_job_fn;
449 	int			bsg_job_size;
450 	struct bsg_class_device bsg_dev;
451 #endif
452 
453 #ifdef CONFIG_BLK_DEV_THROTTLING
454 	/* Throttle data */
455 	struct throtl_data *td;
456 #endif
457 	struct rcu_head		rcu_head;
458 	wait_queue_head_t	mq_freeze_wq;
459 	struct percpu_ref	q_usage_counter;
460 	struct list_head	all_q_node;
461 
462 	struct blk_mq_tag_set	*tag_set;
463 	struct list_head	tag_set_list;
464 	struct bio_set		*bio_split;
465 
466 	bool			mq_sysfs_init_done;
467 };
468 
469 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
470 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
471 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
472 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
473 #define QUEUE_FLAG_DYING	5	/* queue being torn down */
474 #define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
475 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
476 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
477 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
478 #define QUEUE_FLAG_FAIL_IO     10	/* fake timeout */
479 #define QUEUE_FLAG_STACKABLE   11	/* supports request stacking */
480 #define QUEUE_FLAG_NONROT      12	/* non-rotational device (SSD) */
481 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
482 #define QUEUE_FLAG_IO_STAT     13	/* do IO stats */
483 #define QUEUE_FLAG_DISCARD     14	/* supports DISCARD */
484 #define QUEUE_FLAG_NOXMERGES   15	/* No extended merges */
485 #define QUEUE_FLAG_ADD_RANDOM  16	/* Contributes to random pool */
486 #define QUEUE_FLAG_SECDISCARD  17	/* supports SECDISCARD */
487 #define QUEUE_FLAG_SAME_FORCE  18	/* force complete on same CPU */
488 #define QUEUE_FLAG_DEAD        19	/* queue tear-down finished */
489 #define QUEUE_FLAG_INIT_DONE   20	/* queue is initialized */
490 #define QUEUE_FLAG_NO_SG_MERGE 21	/* don't attempt to merge SG segments*/
491 #define QUEUE_FLAG_POLL	       22	/* IO polling enabled if set */
492 #define QUEUE_FLAG_WC	       23	/* Write back caching */
493 #define QUEUE_FLAG_FUA	       24	/* device supports FUA writes */
494 #define QUEUE_FLAG_FLUSH_NQ    25	/* flush not queueuable */
495 
496 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
497 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
498 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
499 				 (1 << QUEUE_FLAG_ADD_RANDOM))
500 
501 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
502 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
503 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
504 				 (1 << QUEUE_FLAG_POLL))
505 
506 static inline void queue_lockdep_assert_held(struct request_queue *q)
507 {
508 	if (q->queue_lock)
509 		lockdep_assert_held(q->queue_lock);
510 }
511 
512 static inline void queue_flag_set_unlocked(unsigned int flag,
513 					   struct request_queue *q)
514 {
515 	__set_bit(flag, &q->queue_flags);
516 }
517 
518 static inline int queue_flag_test_and_clear(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 		__clear_bit(flag, &q->queue_flags);
525 		return 1;
526 	}
527 
528 	return 0;
529 }
530 
531 static inline int queue_flag_test_and_set(unsigned int flag,
532 					  struct request_queue *q)
533 {
534 	queue_lockdep_assert_held(q);
535 
536 	if (!test_bit(flag, &q->queue_flags)) {
537 		__set_bit(flag, &q->queue_flags);
538 		return 0;
539 	}
540 
541 	return 1;
542 }
543 
544 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
545 {
546 	queue_lockdep_assert_held(q);
547 	__set_bit(flag, &q->queue_flags);
548 }
549 
550 static inline void queue_flag_clear_unlocked(unsigned int flag,
551 					     struct request_queue *q)
552 {
553 	__clear_bit(flag, &q->queue_flags);
554 }
555 
556 static inline int queue_in_flight(struct request_queue *q)
557 {
558 	return q->in_flight[0] + q->in_flight[1];
559 }
560 
561 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
562 {
563 	queue_lockdep_assert_held(q);
564 	__clear_bit(flag, &q->queue_flags);
565 }
566 
567 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
568 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
569 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
570 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
571 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
572 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
573 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
574 #define blk_queue_noxmerges(q)	\
575 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
576 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
577 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
578 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
579 #define blk_queue_stackable(q)	\
580 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
581 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
582 #define blk_queue_secdiscard(q)	(blk_queue_discard(q) && \
583 	test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
584 
585 #define blk_noretry_request(rq) \
586 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
587 			     REQ_FAILFAST_DRIVER))
588 
589 #define blk_account_rq(rq) \
590 	(((rq)->cmd_flags & REQ_STARTED) && \
591 	 ((rq)->cmd_type == REQ_TYPE_FS))
592 
593 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
594 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
595 /* rq->queuelist of dequeued request must be list_empty() */
596 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
597 
598 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
599 
600 #define rq_data_dir(rq)		((int)((rq)->cmd_flags & 1))
601 
602 /*
603  * Driver can handle struct request, if it either has an old style
604  * request_fn defined, or is blk-mq based.
605  */
606 static inline bool queue_is_rq_based(struct request_queue *q)
607 {
608 	return q->request_fn || q->mq_ops;
609 }
610 
611 static inline unsigned int blk_queue_cluster(struct request_queue *q)
612 {
613 	return q->limits.cluster;
614 }
615 
616 /*
617  * We regard a request as sync, if either a read or a sync write
618  */
619 static inline bool rw_is_sync(unsigned int rw_flags)
620 {
621 	return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
622 }
623 
624 static inline bool rq_is_sync(struct request *rq)
625 {
626 	return rw_is_sync(rq->cmd_flags);
627 }
628 
629 static inline bool blk_rl_full(struct request_list *rl, bool sync)
630 {
631 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
632 
633 	return rl->flags & flag;
634 }
635 
636 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
637 {
638 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
639 
640 	rl->flags |= flag;
641 }
642 
643 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
644 {
645 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
646 
647 	rl->flags &= ~flag;
648 }
649 
650 static inline bool rq_mergeable(struct request *rq)
651 {
652 	if (rq->cmd_type != REQ_TYPE_FS)
653 		return false;
654 
655 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
656 		return false;
657 
658 	return true;
659 }
660 
661 static inline bool blk_check_merge_flags(unsigned int flags1,
662 					 unsigned int flags2)
663 {
664 	if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
665 		return false;
666 
667 	if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
668 		return false;
669 
670 	if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
671 		return false;
672 
673 	return true;
674 }
675 
676 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
677 {
678 	if (bio_data(a) == bio_data(b))
679 		return true;
680 
681 	return false;
682 }
683 
684 /*
685  * q->prep_rq_fn return values
686  */
687 enum {
688 	BLKPREP_OK,		/* serve it */
689 	BLKPREP_KILL,		/* fatal error, kill, return -EIO */
690 	BLKPREP_DEFER,		/* leave on queue */
691 	BLKPREP_INVALID,	/* invalid command, kill, return -EREMOTEIO */
692 };
693 
694 extern unsigned long blk_max_low_pfn, blk_max_pfn;
695 
696 /*
697  * standard bounce addresses:
698  *
699  * BLK_BOUNCE_HIGH	: bounce all highmem pages
700  * BLK_BOUNCE_ANY	: don't bounce anything
701  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
702  */
703 
704 #if BITS_PER_LONG == 32
705 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
706 #else
707 #define BLK_BOUNCE_HIGH		-1ULL
708 #endif
709 #define BLK_BOUNCE_ANY		(-1ULL)
710 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
711 
712 /*
713  * default timeout for SG_IO if none specified
714  */
715 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
716 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
717 
718 #ifdef CONFIG_BOUNCE
719 extern int init_emergency_isa_pool(void);
720 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
721 #else
722 static inline int init_emergency_isa_pool(void)
723 {
724 	return 0;
725 }
726 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
727 {
728 }
729 #endif /* CONFIG_MMU */
730 
731 struct rq_map_data {
732 	struct page **pages;
733 	int page_order;
734 	int nr_entries;
735 	unsigned long offset;
736 	int null_mapped;
737 	int from_user;
738 };
739 
740 struct req_iterator {
741 	struct bvec_iter iter;
742 	struct bio *bio;
743 };
744 
745 /* This should not be used directly - use rq_for_each_segment */
746 #define for_each_bio(_bio)		\
747 	for (; _bio; _bio = _bio->bi_next)
748 #define __rq_for_each_bio(_bio, rq)	\
749 	if ((rq->bio))			\
750 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
751 
752 #define rq_for_each_segment(bvl, _rq, _iter)			\
753 	__rq_for_each_bio(_iter.bio, _rq)			\
754 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
755 
756 #define rq_iter_last(bvec, _iter)				\
757 		(_iter.bio->bi_next == NULL &&			\
758 		 bio_iter_last(bvec, _iter.iter))
759 
760 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
761 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
762 #endif
763 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
764 extern void rq_flush_dcache_pages(struct request *rq);
765 #else
766 static inline void rq_flush_dcache_pages(struct request *rq)
767 {
768 }
769 #endif
770 
771 #ifdef CONFIG_PRINTK
772 #define vfs_msg(sb, level, fmt, ...)				\
773 	__vfs_msg(sb, level, fmt, ##__VA_ARGS__)
774 #else
775 #define vfs_msg(sb, level, fmt, ...)				\
776 do {								\
777 	no_printk(fmt, ##__VA_ARGS__);				\
778 	__vfs_msg(sb, "", " ");					\
779 } while (0)
780 #endif
781 
782 extern int blk_register_queue(struct gendisk *disk);
783 extern void blk_unregister_queue(struct gendisk *disk);
784 extern blk_qc_t generic_make_request(struct bio *bio);
785 extern void blk_rq_init(struct request_queue *q, struct request *rq);
786 extern void blk_put_request(struct request *);
787 extern void __blk_put_request(struct request_queue *, struct request *);
788 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
789 extern struct request *blk_make_request(struct request_queue *, struct bio *,
790 					gfp_t);
791 extern void blk_rq_set_block_pc(struct request *);
792 extern void blk_requeue_request(struct request_queue *, struct request *);
793 extern void blk_add_request_payload(struct request *rq, struct page *page,
794 		int offset, unsigned int len);
795 extern int blk_lld_busy(struct request_queue *q);
796 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
797 			     struct bio_set *bs, gfp_t gfp_mask,
798 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
799 			     void *data);
800 extern void blk_rq_unprep_clone(struct request *rq);
801 extern int blk_insert_cloned_request(struct request_queue *q,
802 				     struct request *rq);
803 extern void blk_delay_queue(struct request_queue *, unsigned long);
804 extern void blk_queue_split(struct request_queue *, struct bio **,
805 			    struct bio_set *);
806 extern void blk_recount_segments(struct request_queue *, struct bio *);
807 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
808 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
809 			      unsigned int, void __user *);
810 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
811 			  unsigned int, void __user *);
812 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
813 			 struct scsi_ioctl_command __user *);
814 
815 extern int blk_queue_enter(struct request_queue *q, bool nowait);
816 extern void blk_queue_exit(struct request_queue *q);
817 extern void blk_start_queue(struct request_queue *q);
818 extern void blk_start_queue_async(struct request_queue *q);
819 extern void blk_stop_queue(struct request_queue *q);
820 extern void blk_sync_queue(struct request_queue *q);
821 extern void __blk_stop_queue(struct request_queue *q);
822 extern void __blk_run_queue(struct request_queue *q);
823 extern void __blk_run_queue_uncond(struct request_queue *q);
824 extern void blk_run_queue(struct request_queue *);
825 extern void blk_run_queue_async(struct request_queue *q);
826 extern int blk_rq_map_user(struct request_queue *, struct request *,
827 			   struct rq_map_data *, void __user *, unsigned long,
828 			   gfp_t);
829 extern int blk_rq_unmap_user(struct bio *);
830 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
831 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
832 			       struct rq_map_data *, const struct iov_iter *,
833 			       gfp_t);
834 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
835 			  struct request *, int);
836 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
837 				  struct request *, int, rq_end_io_fn *);
838 
839 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
840 
841 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
842 {
843 	return bdev->bd_disk->queue;	/* this is never NULL */
844 }
845 
846 /*
847  * blk_rq_pos()			: the current sector
848  * blk_rq_bytes()		: bytes left in the entire request
849  * blk_rq_cur_bytes()		: bytes left in the current segment
850  * blk_rq_err_bytes()		: bytes left till the next error boundary
851  * blk_rq_sectors()		: sectors left in the entire request
852  * blk_rq_cur_sectors()		: sectors left in the current segment
853  */
854 static inline sector_t blk_rq_pos(const struct request *rq)
855 {
856 	return rq->__sector;
857 }
858 
859 static inline unsigned int blk_rq_bytes(const struct request *rq)
860 {
861 	return rq->__data_len;
862 }
863 
864 static inline int blk_rq_cur_bytes(const struct request *rq)
865 {
866 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
867 }
868 
869 extern unsigned int blk_rq_err_bytes(const struct request *rq);
870 
871 static inline unsigned int blk_rq_sectors(const struct request *rq)
872 {
873 	return blk_rq_bytes(rq) >> 9;
874 }
875 
876 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
877 {
878 	return blk_rq_cur_bytes(rq) >> 9;
879 }
880 
881 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
882 						     unsigned int cmd_flags)
883 {
884 	if (unlikely(cmd_flags & REQ_DISCARD))
885 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
886 
887 	if (unlikely(cmd_flags & REQ_WRITE_SAME))
888 		return q->limits.max_write_same_sectors;
889 
890 	return q->limits.max_sectors;
891 }
892 
893 /*
894  * Return maximum size of a request at given offset. Only valid for
895  * file system requests.
896  */
897 static inline unsigned int blk_max_size_offset(struct request_queue *q,
898 					       sector_t offset)
899 {
900 	if (!q->limits.chunk_sectors)
901 		return q->limits.max_sectors;
902 
903 	return q->limits.chunk_sectors -
904 			(offset & (q->limits.chunk_sectors - 1));
905 }
906 
907 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
908 {
909 	struct request_queue *q = rq->q;
910 
911 	if (unlikely(rq->cmd_type != REQ_TYPE_FS))
912 		return q->limits.max_hw_sectors;
913 
914 	if (!q->limits.chunk_sectors || (rq->cmd_flags & REQ_DISCARD))
915 		return blk_queue_get_max_sectors(q, rq->cmd_flags);
916 
917 	return min(blk_max_size_offset(q, blk_rq_pos(rq)),
918 			blk_queue_get_max_sectors(q, rq->cmd_flags));
919 }
920 
921 static inline unsigned int blk_rq_count_bios(struct request *rq)
922 {
923 	unsigned int nr_bios = 0;
924 	struct bio *bio;
925 
926 	__rq_for_each_bio(bio, rq)
927 		nr_bios++;
928 
929 	return nr_bios;
930 }
931 
932 /*
933  * Request issue related functions.
934  */
935 extern struct request *blk_peek_request(struct request_queue *q);
936 extern void blk_start_request(struct request *rq);
937 extern struct request *blk_fetch_request(struct request_queue *q);
938 
939 /*
940  * Request completion related functions.
941  *
942  * blk_update_request() completes given number of bytes and updates
943  * the request without completing it.
944  *
945  * blk_end_request() and friends.  __blk_end_request() must be called
946  * with the request queue spinlock acquired.
947  *
948  * Several drivers define their own end_request and call
949  * blk_end_request() for parts of the original function.
950  * This prevents code duplication in drivers.
951  */
952 extern bool blk_update_request(struct request *rq, int error,
953 			       unsigned int nr_bytes);
954 extern void blk_finish_request(struct request *rq, int error);
955 extern bool blk_end_request(struct request *rq, int error,
956 			    unsigned int nr_bytes);
957 extern void blk_end_request_all(struct request *rq, int error);
958 extern bool blk_end_request_cur(struct request *rq, int error);
959 extern bool blk_end_request_err(struct request *rq, int error);
960 extern bool __blk_end_request(struct request *rq, int error,
961 			      unsigned int nr_bytes);
962 extern void __blk_end_request_all(struct request *rq, int error);
963 extern bool __blk_end_request_cur(struct request *rq, int error);
964 extern bool __blk_end_request_err(struct request *rq, int error);
965 
966 extern void blk_complete_request(struct request *);
967 extern void __blk_complete_request(struct request *);
968 extern void blk_abort_request(struct request *);
969 extern void blk_unprep_request(struct request *);
970 
971 /*
972  * Access functions for manipulating queue properties
973  */
974 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
975 					spinlock_t *lock, int node_id);
976 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
977 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
978 						      request_fn_proc *, spinlock_t *);
979 extern void blk_cleanup_queue(struct request_queue *);
980 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
981 extern void blk_queue_bounce_limit(struct request_queue *, u64);
982 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
983 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
984 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
985 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
986 extern void blk_queue_max_discard_sectors(struct request_queue *q,
987 		unsigned int max_discard_sectors);
988 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
989 		unsigned int max_write_same_sectors);
990 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
991 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
992 extern void blk_queue_alignment_offset(struct request_queue *q,
993 				       unsigned int alignment);
994 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
995 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
996 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
997 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
998 extern void blk_set_default_limits(struct queue_limits *lim);
999 extern void blk_set_stacking_limits(struct queue_limits *lim);
1000 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1001 			    sector_t offset);
1002 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1003 			    sector_t offset);
1004 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1005 			      sector_t offset);
1006 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1007 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1008 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1009 extern int blk_queue_dma_drain(struct request_queue *q,
1010 			       dma_drain_needed_fn *dma_drain_needed,
1011 			       void *buf, unsigned int size);
1012 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1013 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1014 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1015 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1016 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1017 extern void blk_queue_dma_alignment(struct request_queue *, int);
1018 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1019 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1020 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1021 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1022 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1023 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1024 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1025 
1026 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1027 extern void blk_dump_rq_flags(struct request *, char *);
1028 extern long nr_blockdev_pages(void);
1029 
1030 bool __must_check blk_get_queue(struct request_queue *);
1031 struct request_queue *blk_alloc_queue(gfp_t);
1032 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1033 extern void blk_put_queue(struct request_queue *);
1034 extern void blk_set_queue_dying(struct request_queue *);
1035 
1036 /*
1037  * block layer runtime pm functions
1038  */
1039 #ifdef CONFIG_PM
1040 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1041 extern int blk_pre_runtime_suspend(struct request_queue *q);
1042 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1043 extern void blk_pre_runtime_resume(struct request_queue *q);
1044 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1045 extern void blk_set_runtime_active(struct request_queue *q);
1046 #else
1047 static inline void blk_pm_runtime_init(struct request_queue *q,
1048 	struct device *dev) {}
1049 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1050 {
1051 	return -ENOSYS;
1052 }
1053 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1054 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1055 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1056 extern inline void blk_set_runtime_active(struct request_queue *q) {}
1057 #endif
1058 
1059 /*
1060  * blk_plug permits building a queue of related requests by holding the I/O
1061  * fragments for a short period. This allows merging of sequential requests
1062  * into single larger request. As the requests are moved from a per-task list to
1063  * the device's request_queue in a batch, this results in improved scalability
1064  * as the lock contention for request_queue lock is reduced.
1065  *
1066  * It is ok not to disable preemption when adding the request to the plug list
1067  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1068  * the plug list when the task sleeps by itself. For details, please see
1069  * schedule() where blk_schedule_flush_plug() is called.
1070  */
1071 struct blk_plug {
1072 	struct list_head list; /* requests */
1073 	struct list_head mq_list; /* blk-mq requests */
1074 	struct list_head cb_list; /* md requires an unplug callback */
1075 };
1076 #define BLK_MAX_REQUEST_COUNT 16
1077 
1078 struct blk_plug_cb;
1079 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1080 struct blk_plug_cb {
1081 	struct list_head list;
1082 	blk_plug_cb_fn callback;
1083 	void *data;
1084 };
1085 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1086 					     void *data, int size);
1087 extern void blk_start_plug(struct blk_plug *);
1088 extern void blk_finish_plug(struct blk_plug *);
1089 extern void blk_flush_plug_list(struct blk_plug *, bool);
1090 
1091 static inline void blk_flush_plug(struct task_struct *tsk)
1092 {
1093 	struct blk_plug *plug = tsk->plug;
1094 
1095 	if (plug)
1096 		blk_flush_plug_list(plug, false);
1097 }
1098 
1099 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1100 {
1101 	struct blk_plug *plug = tsk->plug;
1102 
1103 	if (plug)
1104 		blk_flush_plug_list(plug, true);
1105 }
1106 
1107 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1108 {
1109 	struct blk_plug *plug = tsk->plug;
1110 
1111 	return plug &&
1112 		(!list_empty(&plug->list) ||
1113 		 !list_empty(&plug->mq_list) ||
1114 		 !list_empty(&plug->cb_list));
1115 }
1116 
1117 /*
1118  * tag stuff
1119  */
1120 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1121 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1122 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1123 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1124 extern void blk_queue_free_tags(struct request_queue *);
1125 extern int blk_queue_resize_tags(struct request_queue *, int);
1126 extern void blk_queue_invalidate_tags(struct request_queue *);
1127 extern struct blk_queue_tag *blk_init_tags(int, int);
1128 extern void blk_free_tags(struct blk_queue_tag *);
1129 
1130 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1131 						int tag)
1132 {
1133 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1134 		return NULL;
1135 	return bqt->tag_index[tag];
1136 }
1137 
1138 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1139 
1140 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1141 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1142 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1143 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1144 		sector_t nr_sects, gfp_t gfp_mask, int type, struct bio **biop);
1145 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1146 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1147 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1148 		sector_t nr_sects, gfp_t gfp_mask, bool discard);
1149 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1150 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1151 {
1152 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1153 				    nr_blocks << (sb->s_blocksize_bits - 9),
1154 				    gfp_mask, flags);
1155 }
1156 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1157 		sector_t nr_blocks, gfp_t gfp_mask)
1158 {
1159 	return blkdev_issue_zeroout(sb->s_bdev,
1160 				    block << (sb->s_blocksize_bits - 9),
1161 				    nr_blocks << (sb->s_blocksize_bits - 9),
1162 				    gfp_mask, true);
1163 }
1164 
1165 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1166 
1167 enum blk_default_limits {
1168 	BLK_MAX_SEGMENTS	= 128,
1169 	BLK_SAFE_MAX_SECTORS	= 255,
1170 	BLK_DEF_MAX_SECTORS	= 2560,
1171 	BLK_MAX_SEGMENT_SIZE	= 65536,
1172 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1173 };
1174 
1175 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1176 
1177 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1178 {
1179 	return q->limits.bounce_pfn;
1180 }
1181 
1182 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1183 {
1184 	return q->limits.seg_boundary_mask;
1185 }
1186 
1187 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1188 {
1189 	return q->limits.virt_boundary_mask;
1190 }
1191 
1192 static inline unsigned int queue_max_sectors(struct request_queue *q)
1193 {
1194 	return q->limits.max_sectors;
1195 }
1196 
1197 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1198 {
1199 	return q->limits.max_hw_sectors;
1200 }
1201 
1202 static inline unsigned short queue_max_segments(struct request_queue *q)
1203 {
1204 	return q->limits.max_segments;
1205 }
1206 
1207 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1208 {
1209 	return q->limits.max_segment_size;
1210 }
1211 
1212 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1213 {
1214 	int retval = 512;
1215 
1216 	if (q && q->limits.logical_block_size)
1217 		retval = q->limits.logical_block_size;
1218 
1219 	return retval;
1220 }
1221 
1222 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1223 {
1224 	return queue_logical_block_size(bdev_get_queue(bdev));
1225 }
1226 
1227 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1228 {
1229 	return q->limits.physical_block_size;
1230 }
1231 
1232 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1233 {
1234 	return queue_physical_block_size(bdev_get_queue(bdev));
1235 }
1236 
1237 static inline unsigned int queue_io_min(struct request_queue *q)
1238 {
1239 	return q->limits.io_min;
1240 }
1241 
1242 static inline int bdev_io_min(struct block_device *bdev)
1243 {
1244 	return queue_io_min(bdev_get_queue(bdev));
1245 }
1246 
1247 static inline unsigned int queue_io_opt(struct request_queue *q)
1248 {
1249 	return q->limits.io_opt;
1250 }
1251 
1252 static inline int bdev_io_opt(struct block_device *bdev)
1253 {
1254 	return queue_io_opt(bdev_get_queue(bdev));
1255 }
1256 
1257 static inline int queue_alignment_offset(struct request_queue *q)
1258 {
1259 	if (q->limits.misaligned)
1260 		return -1;
1261 
1262 	return q->limits.alignment_offset;
1263 }
1264 
1265 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1266 {
1267 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1268 	unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1269 
1270 	return (granularity + lim->alignment_offset - alignment) % granularity;
1271 }
1272 
1273 static inline int bdev_alignment_offset(struct block_device *bdev)
1274 {
1275 	struct request_queue *q = bdev_get_queue(bdev);
1276 
1277 	if (q->limits.misaligned)
1278 		return -1;
1279 
1280 	if (bdev != bdev->bd_contains)
1281 		return bdev->bd_part->alignment_offset;
1282 
1283 	return q->limits.alignment_offset;
1284 }
1285 
1286 static inline int queue_discard_alignment(struct request_queue *q)
1287 {
1288 	if (q->limits.discard_misaligned)
1289 		return -1;
1290 
1291 	return q->limits.discard_alignment;
1292 }
1293 
1294 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1295 {
1296 	unsigned int alignment, granularity, offset;
1297 
1298 	if (!lim->max_discard_sectors)
1299 		return 0;
1300 
1301 	/* Why are these in bytes, not sectors? */
1302 	alignment = lim->discard_alignment >> 9;
1303 	granularity = lim->discard_granularity >> 9;
1304 	if (!granularity)
1305 		return 0;
1306 
1307 	/* Offset of the partition start in 'granularity' sectors */
1308 	offset = sector_div(sector, granularity);
1309 
1310 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1311 	offset = (granularity + alignment - offset) % granularity;
1312 
1313 	/* Turn it back into bytes, gaah */
1314 	return offset << 9;
1315 }
1316 
1317 static inline int bdev_discard_alignment(struct block_device *bdev)
1318 {
1319 	struct request_queue *q = bdev_get_queue(bdev);
1320 
1321 	if (bdev != bdev->bd_contains)
1322 		return bdev->bd_part->discard_alignment;
1323 
1324 	return q->limits.discard_alignment;
1325 }
1326 
1327 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1328 {
1329 	if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1330 		return 1;
1331 
1332 	return 0;
1333 }
1334 
1335 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1336 {
1337 	return queue_discard_zeroes_data(bdev_get_queue(bdev));
1338 }
1339 
1340 static inline unsigned int bdev_write_same(struct block_device *bdev)
1341 {
1342 	struct request_queue *q = bdev_get_queue(bdev);
1343 
1344 	if (q)
1345 		return q->limits.max_write_same_sectors;
1346 
1347 	return 0;
1348 }
1349 
1350 static inline int queue_dma_alignment(struct request_queue *q)
1351 {
1352 	return q ? q->dma_alignment : 511;
1353 }
1354 
1355 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1356 				 unsigned int len)
1357 {
1358 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1359 	return !(addr & alignment) && !(len & alignment);
1360 }
1361 
1362 /* assumes size > 256 */
1363 static inline unsigned int blksize_bits(unsigned int size)
1364 {
1365 	unsigned int bits = 8;
1366 	do {
1367 		bits++;
1368 		size >>= 1;
1369 	} while (size > 256);
1370 	return bits;
1371 }
1372 
1373 static inline unsigned int block_size(struct block_device *bdev)
1374 {
1375 	return bdev->bd_block_size;
1376 }
1377 
1378 static inline bool queue_flush_queueable(struct request_queue *q)
1379 {
1380 	return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1381 }
1382 
1383 typedef struct {struct page *v;} Sector;
1384 
1385 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1386 
1387 static inline void put_dev_sector(Sector p)
1388 {
1389 	put_page(p.v);
1390 }
1391 
1392 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1393 				struct bio_vec *bprv, unsigned int offset)
1394 {
1395 	return offset ||
1396 		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1397 }
1398 
1399 /*
1400  * Check if adding a bio_vec after bprv with offset would create a gap in
1401  * the SG list. Most drivers don't care about this, but some do.
1402  */
1403 static inline bool bvec_gap_to_prev(struct request_queue *q,
1404 				struct bio_vec *bprv, unsigned int offset)
1405 {
1406 	if (!queue_virt_boundary(q))
1407 		return false;
1408 	return __bvec_gap_to_prev(q, bprv, offset);
1409 }
1410 
1411 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1412 			 struct bio *next)
1413 {
1414 	if (bio_has_data(prev) && queue_virt_boundary(q)) {
1415 		struct bio_vec pb, nb;
1416 
1417 		bio_get_last_bvec(prev, &pb);
1418 		bio_get_first_bvec(next, &nb);
1419 
1420 		return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1421 	}
1422 
1423 	return false;
1424 }
1425 
1426 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1427 {
1428 	return bio_will_gap(req->q, req->biotail, bio);
1429 }
1430 
1431 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1432 {
1433 	return bio_will_gap(req->q, bio, req->bio);
1434 }
1435 
1436 struct work_struct;
1437 int kblockd_schedule_work(struct work_struct *work);
1438 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1439 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1440 
1441 #ifdef CONFIG_BLK_CGROUP
1442 /*
1443  * This should not be using sched_clock(). A real patch is in progress
1444  * to fix this up, until that is in place we need to disable preemption
1445  * around sched_clock() in this function and set_io_start_time_ns().
1446  */
1447 static inline void set_start_time_ns(struct request *req)
1448 {
1449 	preempt_disable();
1450 	req->start_time_ns = sched_clock();
1451 	preempt_enable();
1452 }
1453 
1454 static inline void set_io_start_time_ns(struct request *req)
1455 {
1456 	preempt_disable();
1457 	req->io_start_time_ns = sched_clock();
1458 	preempt_enable();
1459 }
1460 
1461 static inline uint64_t rq_start_time_ns(struct request *req)
1462 {
1463         return req->start_time_ns;
1464 }
1465 
1466 static inline uint64_t rq_io_start_time_ns(struct request *req)
1467 {
1468         return req->io_start_time_ns;
1469 }
1470 #else
1471 static inline void set_start_time_ns(struct request *req) {}
1472 static inline void set_io_start_time_ns(struct request *req) {}
1473 static inline uint64_t rq_start_time_ns(struct request *req)
1474 {
1475 	return 0;
1476 }
1477 static inline uint64_t rq_io_start_time_ns(struct request *req)
1478 {
1479 	return 0;
1480 }
1481 #endif
1482 
1483 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1484 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1485 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1486 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1487 
1488 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1489 
1490 enum blk_integrity_flags {
1491 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1492 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1493 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1494 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1495 };
1496 
1497 struct blk_integrity_iter {
1498 	void			*prot_buf;
1499 	void			*data_buf;
1500 	sector_t		seed;
1501 	unsigned int		data_size;
1502 	unsigned short		interval;
1503 	const char		*disk_name;
1504 };
1505 
1506 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1507 
1508 struct blk_integrity_profile {
1509 	integrity_processing_fn		*generate_fn;
1510 	integrity_processing_fn		*verify_fn;
1511 	const char			*name;
1512 };
1513 
1514 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1515 extern void blk_integrity_unregister(struct gendisk *);
1516 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1517 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1518 				   struct scatterlist *);
1519 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1520 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1521 				   struct request *);
1522 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1523 				    struct bio *);
1524 
1525 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1526 {
1527 	struct blk_integrity *bi = &disk->queue->integrity;
1528 
1529 	if (!bi->profile)
1530 		return NULL;
1531 
1532 	return bi;
1533 }
1534 
1535 static inline
1536 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1537 {
1538 	return blk_get_integrity(bdev->bd_disk);
1539 }
1540 
1541 static inline bool blk_integrity_rq(struct request *rq)
1542 {
1543 	return rq->cmd_flags & REQ_INTEGRITY;
1544 }
1545 
1546 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1547 						    unsigned int segs)
1548 {
1549 	q->limits.max_integrity_segments = segs;
1550 }
1551 
1552 static inline unsigned short
1553 queue_max_integrity_segments(struct request_queue *q)
1554 {
1555 	return q->limits.max_integrity_segments;
1556 }
1557 
1558 static inline bool integrity_req_gap_back_merge(struct request *req,
1559 						struct bio *next)
1560 {
1561 	struct bio_integrity_payload *bip = bio_integrity(req->bio);
1562 	struct bio_integrity_payload *bip_next = bio_integrity(next);
1563 
1564 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1565 				bip_next->bip_vec[0].bv_offset);
1566 }
1567 
1568 static inline bool integrity_req_gap_front_merge(struct request *req,
1569 						 struct bio *bio)
1570 {
1571 	struct bio_integrity_payload *bip = bio_integrity(bio);
1572 	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1573 
1574 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1575 				bip_next->bip_vec[0].bv_offset);
1576 }
1577 
1578 #else /* CONFIG_BLK_DEV_INTEGRITY */
1579 
1580 struct bio;
1581 struct block_device;
1582 struct gendisk;
1583 struct blk_integrity;
1584 
1585 static inline int blk_integrity_rq(struct request *rq)
1586 {
1587 	return 0;
1588 }
1589 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1590 					    struct bio *b)
1591 {
1592 	return 0;
1593 }
1594 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1595 					  struct bio *b,
1596 					  struct scatterlist *s)
1597 {
1598 	return 0;
1599 }
1600 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1601 {
1602 	return NULL;
1603 }
1604 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1605 {
1606 	return NULL;
1607 }
1608 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1609 {
1610 	return 0;
1611 }
1612 static inline void blk_integrity_register(struct gendisk *d,
1613 					 struct blk_integrity *b)
1614 {
1615 }
1616 static inline void blk_integrity_unregister(struct gendisk *d)
1617 {
1618 }
1619 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1620 						    unsigned int segs)
1621 {
1622 }
1623 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1624 {
1625 	return 0;
1626 }
1627 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1628 					  struct request *r1,
1629 					  struct request *r2)
1630 {
1631 	return true;
1632 }
1633 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1634 					   struct request *r,
1635 					   struct bio *b)
1636 {
1637 	return true;
1638 }
1639 
1640 static inline bool integrity_req_gap_back_merge(struct request *req,
1641 						struct bio *next)
1642 {
1643 	return false;
1644 }
1645 static inline bool integrity_req_gap_front_merge(struct request *req,
1646 						 struct bio *bio)
1647 {
1648 	return false;
1649 }
1650 
1651 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1652 
1653 /**
1654  * struct blk_dax_ctl - control and output parameters for ->direct_access
1655  * @sector: (input) offset relative to a block_device
1656  * @addr: (output) kernel virtual address for @sector populated by driver
1657  * @pfn: (output) page frame number for @addr populated by driver
1658  * @size: (input) number of bytes requested
1659  */
1660 struct blk_dax_ctl {
1661 	sector_t sector;
1662 	void __pmem *addr;
1663 	long size;
1664 	pfn_t pfn;
1665 };
1666 
1667 struct block_device_operations {
1668 	int (*open) (struct block_device *, fmode_t);
1669 	void (*release) (struct gendisk *, fmode_t);
1670 	int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1671 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1672 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1673 	long (*direct_access)(struct block_device *, sector_t, void __pmem **,
1674 			pfn_t *, long);
1675 	unsigned int (*check_events) (struct gendisk *disk,
1676 				      unsigned int clearing);
1677 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
1678 	int (*media_changed) (struct gendisk *);
1679 	void (*unlock_native_capacity) (struct gendisk *);
1680 	int (*revalidate_disk) (struct gendisk *);
1681 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1682 	/* this callback is with swap_lock and sometimes page table lock held */
1683 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1684 	struct module *owner;
1685 	const struct pr_ops *pr_ops;
1686 };
1687 
1688 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1689 				 unsigned long);
1690 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1691 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1692 						struct writeback_control *);
1693 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1694 extern int bdev_dax_supported(struct super_block *, int);
1695 extern bool bdev_dax_capable(struct block_device *);
1696 #else /* CONFIG_BLOCK */
1697 
1698 struct block_device;
1699 
1700 /*
1701  * stubs for when the block layer is configured out
1702  */
1703 #define buffer_heads_over_limit 0
1704 
1705 static inline long nr_blockdev_pages(void)
1706 {
1707 	return 0;
1708 }
1709 
1710 struct blk_plug {
1711 };
1712 
1713 static inline void blk_start_plug(struct blk_plug *plug)
1714 {
1715 }
1716 
1717 static inline void blk_finish_plug(struct blk_plug *plug)
1718 {
1719 }
1720 
1721 static inline void blk_flush_plug(struct task_struct *task)
1722 {
1723 }
1724 
1725 static inline void blk_schedule_flush_plug(struct task_struct *task)
1726 {
1727 }
1728 
1729 
1730 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1731 {
1732 	return false;
1733 }
1734 
1735 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1736 				     sector_t *error_sector)
1737 {
1738 	return 0;
1739 }
1740 
1741 #endif /* CONFIG_BLOCK */
1742 
1743 #endif
1744