xref: /linux-6.15/include/linux/blkdev.h (revision 7eb6e340)
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.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 
25 #include <asm/scatterlist.h>
26 
27 struct module;
28 struct scsi_ioctl_command;
29 
30 struct request_queue;
31 struct elevator_queue;
32 struct request_pm_state;
33 struct blk_trace;
34 struct request;
35 struct sg_io_hdr;
36 struct bsg_job;
37 struct blkcg_gq;
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_SENSE,			/* sense request */
77 	REQ_TYPE_PM_SUSPEND,		/* suspend request */
78 	REQ_TYPE_PM_RESUME,		/* resume request */
79 	REQ_TYPE_PM_SHUTDOWN,		/* shutdown request */
80 	REQ_TYPE_SPECIAL,		/* driver defined type */
81 	/*
82 	 * for ATA/ATAPI devices. this really doesn't belong here, ide should
83 	 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
84 	 * private REQ_LB opcodes to differentiate what type of request this is
85 	 */
86 	REQ_TYPE_ATA_TASKFILE,
87 	REQ_TYPE_ATA_PC,
88 };
89 
90 #define BLK_MAX_CDB	16
91 
92 /*
93  * try to put the fields that are referenced together in the same cacheline.
94  * if you modify this structure, be sure to check block/blk-core.c:blk_rq_init()
95  * as well!
96  */
97 struct request {
98 	struct list_head queuelist;
99 	union {
100 		struct call_single_data csd;
101 		struct work_struct mq_flush_work;
102 		unsigned long fifo_time;
103 	};
104 
105 	struct request_queue *q;
106 	struct blk_mq_ctx *mq_ctx;
107 
108 	u64 cmd_flags;
109 	enum rq_cmd_type_bits cmd_type;
110 	unsigned long atomic_flags;
111 
112 	int cpu;
113 
114 	/* the following two fields are internal, NEVER access directly */
115 	unsigned int __data_len;	/* total data len */
116 	sector_t __sector;		/* sector cursor */
117 
118 	struct bio *bio;
119 	struct bio *biotail;
120 
121 	/*
122 	 * The hash is used inside the scheduler, and killed once the
123 	 * request reaches the dispatch list. The ipi_list is only used
124 	 * to queue the request for softirq completion, which is long
125 	 * after the request has been unhashed (and even removed from
126 	 * the dispatch list).
127 	 */
128 	union {
129 		struct hlist_node hash;	/* merge hash */
130 		struct list_head ipi_list;
131 	};
132 
133 	/*
134 	 * The rb_node is only used inside the io scheduler, requests
135 	 * are pruned when moved to the dispatch queue. So let the
136 	 * completion_data share space with the rb_node.
137 	 */
138 	union {
139 		struct rb_node rb_node;	/* sort/lookup */
140 		void *completion_data;
141 	};
142 
143 	/*
144 	 * Three pointers are available for the IO schedulers, if they need
145 	 * more they have to dynamically allocate it.  Flush requests are
146 	 * never put on the IO scheduler. So let the flush fields share
147 	 * space with the elevator data.
148 	 */
149 	union {
150 		struct {
151 			struct io_cq		*icq;
152 			void			*priv[2];
153 		} elv;
154 
155 		struct {
156 			unsigned int		seq;
157 			struct list_head	list;
158 			rq_end_io_fn		*saved_end_io;
159 		} flush;
160 	};
161 
162 	struct gendisk *rq_disk;
163 	struct hd_struct *part;
164 	unsigned long start_time;
165 #ifdef CONFIG_BLK_CGROUP
166 	struct request_list *rl;		/* rl this rq is alloced from */
167 	unsigned long long start_time_ns;
168 	unsigned long long io_start_time_ns;    /* when passed to hardware */
169 #endif
170 	/* Number of scatter-gather DMA addr+len pairs after
171 	 * physical address coalescing is performed.
172 	 */
173 	unsigned short nr_phys_segments;
174 #if defined(CONFIG_BLK_DEV_INTEGRITY)
175 	unsigned short nr_integrity_segments;
176 #endif
177 
178 	unsigned short ioprio;
179 
180 	void *special;		/* opaque pointer available for LLD use */
181 	char *buffer;		/* kaddr of the current segment if available */
182 
183 	int tag;
184 	int errors;
185 
186 	/*
187 	 * when request is used as a packet command carrier
188 	 */
189 	unsigned char __cmd[BLK_MAX_CDB];
190 	unsigned char *cmd;
191 	unsigned short cmd_len;
192 
193 	unsigned int extra_len;	/* length of alignment and padding */
194 	unsigned int sense_len;
195 	unsigned int resid_len;	/* residual count */
196 	void *sense;
197 
198 	unsigned long deadline;
199 	struct list_head timeout_list;
200 	unsigned int timeout;
201 	int retries;
202 
203 	/*
204 	 * completion callback.
205 	 */
206 	rq_end_io_fn *end_io;
207 	void *end_io_data;
208 
209 	/* for bidi */
210 	struct request *next_rq;
211 };
212 
213 static inline unsigned short req_get_ioprio(struct request *req)
214 {
215 	return req->ioprio;
216 }
217 
218 /*
219  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
220  * requests. Some step values could eventually be made generic.
221  */
222 struct request_pm_state
223 {
224 	/* PM state machine step value, currently driver specific */
225 	int	pm_step;
226 	/* requested PM state value (S1, S2, S3, S4, ...) */
227 	u32	pm_state;
228 	void*	data;		/* for driver use */
229 };
230 
231 #include <linux/elevator.h>
232 
233 struct blk_queue_ctx;
234 
235 typedef void (request_fn_proc) (struct request_queue *q);
236 typedef void (make_request_fn) (struct request_queue *q, struct bio *bio);
237 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
238 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
239 
240 struct bio_vec;
241 struct bvec_merge_data {
242 	struct block_device *bi_bdev;
243 	sector_t bi_sector;
244 	unsigned bi_size;
245 	unsigned long bi_rw;
246 };
247 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
248 			     struct bio_vec *);
249 typedef void (softirq_done_fn)(struct request *);
250 typedef int (dma_drain_needed_fn)(struct request *);
251 typedef int (lld_busy_fn) (struct request_queue *q);
252 typedef int (bsg_job_fn) (struct bsg_job *);
253 
254 enum blk_eh_timer_return {
255 	BLK_EH_NOT_HANDLED,
256 	BLK_EH_HANDLED,
257 	BLK_EH_RESET_TIMER,
258 };
259 
260 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
261 
262 enum blk_queue_state {
263 	Queue_down,
264 	Queue_up,
265 };
266 
267 struct blk_queue_tag {
268 	struct request **tag_index;	/* map of busy tags */
269 	unsigned long *tag_map;		/* bit map of free/busy tags */
270 	int busy;			/* current depth */
271 	int max_depth;			/* what we will send to device */
272 	int real_max_depth;		/* what the array can hold */
273 	atomic_t refcnt;		/* map can be shared */
274 };
275 
276 #define BLK_SCSI_MAX_CMDS	(256)
277 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
278 
279 struct queue_limits {
280 	unsigned long		bounce_pfn;
281 	unsigned long		seg_boundary_mask;
282 
283 	unsigned int		max_hw_sectors;
284 	unsigned int		max_sectors;
285 	unsigned int		max_segment_size;
286 	unsigned int		physical_block_size;
287 	unsigned int		alignment_offset;
288 	unsigned int		io_min;
289 	unsigned int		io_opt;
290 	unsigned int		max_discard_sectors;
291 	unsigned int		max_write_same_sectors;
292 	unsigned int		discard_granularity;
293 	unsigned int		discard_alignment;
294 
295 	unsigned short		logical_block_size;
296 	unsigned short		max_segments;
297 	unsigned short		max_integrity_segments;
298 
299 	unsigned char		misaligned;
300 	unsigned char		discard_misaligned;
301 	unsigned char		cluster;
302 	unsigned char		discard_zeroes_data;
303 	unsigned char		raid_partial_stripes_expensive;
304 };
305 
306 struct request_queue {
307 	/*
308 	 * Together with queue_head for cacheline sharing
309 	 */
310 	struct list_head	queue_head;
311 	struct request		*last_merge;
312 	struct elevator_queue	*elevator;
313 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
314 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
315 
316 	/*
317 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
318 	 * is used, root blkg allocates from @q->root_rl and all other
319 	 * blkgs from their own blkg->rl.  Which one to use should be
320 	 * determined using bio_request_list().
321 	 */
322 	struct request_list	root_rl;
323 
324 	request_fn_proc		*request_fn;
325 	make_request_fn		*make_request_fn;
326 	prep_rq_fn		*prep_rq_fn;
327 	unprep_rq_fn		*unprep_rq_fn;
328 	merge_bvec_fn		*merge_bvec_fn;
329 	softirq_done_fn		*softirq_done_fn;
330 	rq_timed_out_fn		*rq_timed_out_fn;
331 	dma_drain_needed_fn	*dma_drain_needed;
332 	lld_busy_fn		*lld_busy_fn;
333 
334 	struct blk_mq_ops	*mq_ops;
335 
336 	unsigned int		*mq_map;
337 
338 	/* sw queues */
339 	struct blk_mq_ctx	*queue_ctx;
340 	unsigned int		nr_queues;
341 
342 	/* hw dispatch queues */
343 	struct blk_mq_hw_ctx	**queue_hw_ctx;
344 	unsigned int		nr_hw_queues;
345 
346 	/*
347 	 * Dispatch queue sorting
348 	 */
349 	sector_t		end_sector;
350 	struct request		*boundary_rq;
351 
352 	/*
353 	 * Delayed queue handling
354 	 */
355 	struct delayed_work	delay_work;
356 
357 	struct backing_dev_info	backing_dev_info;
358 
359 	/*
360 	 * The queue owner gets to use this for whatever they like.
361 	 * ll_rw_blk doesn't touch it.
362 	 */
363 	void			*queuedata;
364 
365 	/*
366 	 * various queue flags, see QUEUE_* below
367 	 */
368 	unsigned long		queue_flags;
369 
370 	/*
371 	 * ida allocated id for this queue.  Used to index queues from
372 	 * ioctx.
373 	 */
374 	int			id;
375 
376 	/*
377 	 * queue needs bounce pages for pages above this limit
378 	 */
379 	gfp_t			bounce_gfp;
380 
381 	/*
382 	 * protects queue structures from reentrancy. ->__queue_lock should
383 	 * _never_ be used directly, it is queue private. always use
384 	 * ->queue_lock.
385 	 */
386 	spinlock_t		__queue_lock;
387 	spinlock_t		*queue_lock;
388 
389 	/*
390 	 * queue kobject
391 	 */
392 	struct kobject kobj;
393 
394 	/*
395 	 * mq queue kobject
396 	 */
397 	struct kobject mq_kobj;
398 
399 #ifdef CONFIG_PM_RUNTIME
400 	struct device		*dev;
401 	int			rpm_status;
402 	unsigned int		nr_pending;
403 #endif
404 
405 	/*
406 	 * queue settings
407 	 */
408 	unsigned long		nr_requests;	/* Max # of requests */
409 	unsigned int		nr_congestion_on;
410 	unsigned int		nr_congestion_off;
411 	unsigned int		nr_batching;
412 
413 	unsigned int		dma_drain_size;
414 	void			*dma_drain_buffer;
415 	unsigned int		dma_pad_mask;
416 	unsigned int		dma_alignment;
417 
418 	struct blk_queue_tag	*queue_tags;
419 	struct list_head	tag_busy_list;
420 
421 	unsigned int		nr_sorted;
422 	unsigned int		in_flight[2];
423 	/*
424 	 * Number of active block driver functions for which blk_drain_queue()
425 	 * must wait. Must be incremented around functions that unlock the
426 	 * queue_lock internally, e.g. scsi_request_fn().
427 	 */
428 	unsigned int		request_fn_active;
429 
430 	unsigned int		rq_timeout;
431 	struct timer_list	timeout;
432 	struct list_head	timeout_list;
433 
434 	struct list_head	icq_list;
435 #ifdef CONFIG_BLK_CGROUP
436 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
437 	struct blkcg_gq		*root_blkg;
438 	struct list_head	blkg_list;
439 #endif
440 
441 	struct queue_limits	limits;
442 
443 	/*
444 	 * sg stuff
445 	 */
446 	unsigned int		sg_timeout;
447 	unsigned int		sg_reserved_size;
448 	int			node;
449 #ifdef CONFIG_BLK_DEV_IO_TRACE
450 	struct blk_trace	*blk_trace;
451 #endif
452 	/*
453 	 * for flush operations
454 	 */
455 	unsigned int		flush_flags;
456 	unsigned int		flush_not_queueable:1;
457 	unsigned int		flush_queue_delayed:1;
458 	unsigned int		flush_pending_idx:1;
459 	unsigned int		flush_running_idx:1;
460 	unsigned long		flush_pending_since;
461 	struct list_head	flush_queue[2];
462 	struct list_head	flush_data_in_flight;
463 	struct request		*flush_rq;
464 	spinlock_t		mq_flush_lock;
465 
466 	struct mutex		sysfs_lock;
467 
468 	int			bypass_depth;
469 
470 #if defined(CONFIG_BLK_DEV_BSG)
471 	bsg_job_fn		*bsg_job_fn;
472 	int			bsg_job_size;
473 	struct bsg_class_device bsg_dev;
474 #endif
475 
476 #ifdef CONFIG_BLK_DEV_THROTTLING
477 	/* Throttle data */
478 	struct throtl_data *td;
479 #endif
480 	struct rcu_head		rcu_head;
481 	wait_queue_head_t	mq_freeze_wq;
482 	struct percpu_counter	mq_usage_counter;
483 	struct list_head	all_q_node;
484 };
485 
486 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
487 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
488 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
489 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
490 #define QUEUE_FLAG_DYING	5	/* queue being torn down */
491 #define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
492 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
493 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
494 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
495 #define QUEUE_FLAG_FAIL_IO     10	/* fake timeout */
496 #define QUEUE_FLAG_STACKABLE   11	/* supports request stacking */
497 #define QUEUE_FLAG_NONROT      12	/* non-rotational device (SSD) */
498 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
499 #define QUEUE_FLAG_IO_STAT     13	/* do IO stats */
500 #define QUEUE_FLAG_DISCARD     14	/* supports DISCARD */
501 #define QUEUE_FLAG_NOXMERGES   15	/* No extended merges */
502 #define QUEUE_FLAG_ADD_RANDOM  16	/* Contributes to random pool */
503 #define QUEUE_FLAG_SECDISCARD  17	/* supports SECDISCARD */
504 #define QUEUE_FLAG_SAME_FORCE  18	/* force complete on same CPU */
505 #define QUEUE_FLAG_DEAD        19	/* queue tear-down finished */
506 #define QUEUE_FLAG_INIT_DONE   20	/* queue is initialized */
507 
508 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
509 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
510 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
511 				 (1 << QUEUE_FLAG_ADD_RANDOM))
512 
513 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
514 				 (1 << QUEUE_FLAG_SAME_COMP))
515 
516 static inline void queue_lockdep_assert_held(struct request_queue *q)
517 {
518 	if (q->queue_lock)
519 		lockdep_assert_held(q->queue_lock);
520 }
521 
522 static inline void queue_flag_set_unlocked(unsigned int flag,
523 					   struct request_queue *q)
524 {
525 	__set_bit(flag, &q->queue_flags);
526 }
527 
528 static inline int queue_flag_test_and_clear(unsigned int flag,
529 					    struct request_queue *q)
530 {
531 	queue_lockdep_assert_held(q);
532 
533 	if (test_bit(flag, &q->queue_flags)) {
534 		__clear_bit(flag, &q->queue_flags);
535 		return 1;
536 	}
537 
538 	return 0;
539 }
540 
541 static inline int queue_flag_test_and_set(unsigned int flag,
542 					  struct request_queue *q)
543 {
544 	queue_lockdep_assert_held(q);
545 
546 	if (!test_bit(flag, &q->queue_flags)) {
547 		__set_bit(flag, &q->queue_flags);
548 		return 0;
549 	}
550 
551 	return 1;
552 }
553 
554 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
555 {
556 	queue_lockdep_assert_held(q);
557 	__set_bit(flag, &q->queue_flags);
558 }
559 
560 static inline void queue_flag_clear_unlocked(unsigned int flag,
561 					     struct request_queue *q)
562 {
563 	__clear_bit(flag, &q->queue_flags);
564 }
565 
566 static inline int queue_in_flight(struct request_queue *q)
567 {
568 	return q->in_flight[0] + q->in_flight[1];
569 }
570 
571 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
572 {
573 	queue_lockdep_assert_held(q);
574 	__clear_bit(flag, &q->queue_flags);
575 }
576 
577 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
578 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
579 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
580 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
581 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
582 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
583 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
584 #define blk_queue_noxmerges(q)	\
585 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
586 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
587 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
588 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
589 #define blk_queue_stackable(q)	\
590 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
591 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
592 #define blk_queue_secdiscard(q)	(blk_queue_discard(q) && \
593 	test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
594 
595 #define blk_noretry_request(rq) \
596 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
597 			     REQ_FAILFAST_DRIVER))
598 
599 #define blk_account_rq(rq) \
600 	(((rq)->cmd_flags & REQ_STARTED) && \
601 	 ((rq)->cmd_type == REQ_TYPE_FS))
602 
603 #define blk_pm_request(rq)	\
604 	((rq)->cmd_type == REQ_TYPE_PM_SUSPEND || \
605 	 (rq)->cmd_type == REQ_TYPE_PM_RESUME)
606 
607 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
608 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
609 /* rq->queuelist of dequeued request must be list_empty() */
610 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
611 
612 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
613 
614 #define rq_data_dir(rq)		(((rq)->cmd_flags & 1) != 0)
615 
616 static inline unsigned int blk_queue_cluster(struct request_queue *q)
617 {
618 	return q->limits.cluster;
619 }
620 
621 /*
622  * We regard a request as sync, if either a read or a sync write
623  */
624 static inline bool rw_is_sync(unsigned int rw_flags)
625 {
626 	return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
627 }
628 
629 static inline bool rq_is_sync(struct request *rq)
630 {
631 	return rw_is_sync(rq->cmd_flags);
632 }
633 
634 static inline bool blk_rl_full(struct request_list *rl, bool sync)
635 {
636 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
637 
638 	return rl->flags & flag;
639 }
640 
641 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
642 {
643 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
644 
645 	rl->flags |= flag;
646 }
647 
648 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
649 {
650 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
651 
652 	rl->flags &= ~flag;
653 }
654 
655 static inline bool rq_mergeable(struct request *rq)
656 {
657 	if (rq->cmd_type != REQ_TYPE_FS)
658 		return false;
659 
660 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
661 		return false;
662 
663 	return true;
664 }
665 
666 static inline bool blk_check_merge_flags(unsigned int flags1,
667 					 unsigned int flags2)
668 {
669 	if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
670 		return false;
671 
672 	if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
673 		return false;
674 
675 	if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
676 		return false;
677 
678 	return true;
679 }
680 
681 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
682 {
683 	if (bio_data(a) == bio_data(b))
684 		return true;
685 
686 	return false;
687 }
688 
689 /*
690  * q->prep_rq_fn return values
691  */
692 #define BLKPREP_OK		0	/* serve it */
693 #define BLKPREP_KILL		1	/* fatal error, kill */
694 #define BLKPREP_DEFER		2	/* leave on queue */
695 
696 extern unsigned long blk_max_low_pfn, blk_max_pfn;
697 
698 /*
699  * standard bounce addresses:
700  *
701  * BLK_BOUNCE_HIGH	: bounce all highmem pages
702  * BLK_BOUNCE_ANY	: don't bounce anything
703  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
704  */
705 
706 #if BITS_PER_LONG == 32
707 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
708 #else
709 #define BLK_BOUNCE_HIGH		-1ULL
710 #endif
711 #define BLK_BOUNCE_ANY		(-1ULL)
712 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
713 
714 /*
715  * default timeout for SG_IO if none specified
716  */
717 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
718 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
719 
720 #ifdef CONFIG_BOUNCE
721 extern int init_emergency_isa_pool(void);
722 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
723 #else
724 static inline int init_emergency_isa_pool(void)
725 {
726 	return 0;
727 }
728 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
729 {
730 }
731 #endif /* CONFIG_MMU */
732 
733 struct rq_map_data {
734 	struct page **pages;
735 	int page_order;
736 	int nr_entries;
737 	unsigned long offset;
738 	int null_mapped;
739 	int from_user;
740 };
741 
742 struct req_iterator {
743 	struct bvec_iter iter;
744 	struct bio *bio;
745 };
746 
747 /* This should not be used directly - use rq_for_each_segment */
748 #define for_each_bio(_bio)		\
749 	for (; _bio; _bio = _bio->bi_next)
750 #define __rq_for_each_bio(_bio, rq)	\
751 	if ((rq->bio))			\
752 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
753 
754 #define rq_for_each_segment(bvl, _rq, _iter)			\
755 	__rq_for_each_bio(_iter.bio, _rq)			\
756 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
757 
758 #define rq_iter_last(bvec, _iter)				\
759 		(_iter.bio->bi_next == NULL &&			\
760 		 bio_iter_last(bvec, _iter.iter))
761 
762 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
763 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
764 #endif
765 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
766 extern void rq_flush_dcache_pages(struct request *rq);
767 #else
768 static inline void rq_flush_dcache_pages(struct request *rq)
769 {
770 }
771 #endif
772 
773 extern int blk_register_queue(struct gendisk *disk);
774 extern void blk_unregister_queue(struct gendisk *disk);
775 extern void generic_make_request(struct bio *bio);
776 extern void blk_rq_init(struct request_queue *q, struct request *rq);
777 extern void blk_put_request(struct request *);
778 extern void __blk_put_request(struct request_queue *, struct request *);
779 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
780 extern struct request *blk_make_request(struct request_queue *, struct bio *,
781 					gfp_t);
782 extern void blk_requeue_request(struct request_queue *, struct request *);
783 extern void blk_add_request_payload(struct request *rq, struct page *page,
784 		unsigned int len);
785 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
786 extern int blk_lld_busy(struct request_queue *q);
787 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
788 			     struct bio_set *bs, gfp_t gfp_mask,
789 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
790 			     void *data);
791 extern void blk_rq_unprep_clone(struct request *rq);
792 extern int blk_insert_cloned_request(struct request_queue *q,
793 				     struct request *rq);
794 extern void blk_delay_queue(struct request_queue *, unsigned long);
795 extern void blk_recount_segments(struct request_queue *, struct bio *);
796 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
797 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
798 			      unsigned int, void __user *);
799 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
800 			  unsigned int, void __user *);
801 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
802 			 struct scsi_ioctl_command __user *);
803 
804 extern void blk_queue_bio(struct request_queue *q, struct bio *bio);
805 
806 /*
807  * A queue has just exitted congestion.  Note this in the global counter of
808  * congested queues, and wake up anyone who was waiting for requests to be
809  * put back.
810  */
811 static inline void blk_clear_queue_congested(struct request_queue *q, int sync)
812 {
813 	clear_bdi_congested(&q->backing_dev_info, sync);
814 }
815 
816 /*
817  * A queue has just entered congestion.  Flag that in the queue's VM-visible
818  * state flags and increment the global gounter of congested queues.
819  */
820 static inline void blk_set_queue_congested(struct request_queue *q, int sync)
821 {
822 	set_bdi_congested(&q->backing_dev_info, sync);
823 }
824 
825 extern void blk_start_queue(struct request_queue *q);
826 extern void blk_stop_queue(struct request_queue *q);
827 extern void blk_sync_queue(struct request_queue *q);
828 extern void __blk_stop_queue(struct request_queue *q);
829 extern void __blk_run_queue(struct request_queue *q);
830 extern void blk_run_queue(struct request_queue *);
831 extern void blk_run_queue_async(struct request_queue *q);
832 extern int blk_rq_map_user(struct request_queue *, struct request *,
833 			   struct rq_map_data *, void __user *, unsigned long,
834 			   gfp_t);
835 extern int blk_rq_unmap_user(struct bio *);
836 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
837 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
838 			       struct rq_map_data *, const struct sg_iovec *,
839 			       int, unsigned int, gfp_t);
840 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
841 			  struct request *, int);
842 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
843 				  struct request *, int, rq_end_io_fn *);
844 
845 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
846 {
847 	return bdev->bd_disk->queue;
848 }
849 
850 /*
851  * blk_rq_pos()			: the current sector
852  * blk_rq_bytes()		: bytes left in the entire request
853  * blk_rq_cur_bytes()		: bytes left in the current segment
854  * blk_rq_err_bytes()		: bytes left till the next error boundary
855  * blk_rq_sectors()		: sectors left in the entire request
856  * blk_rq_cur_sectors()		: sectors left in the current segment
857  */
858 static inline sector_t blk_rq_pos(const struct request *rq)
859 {
860 	return rq->__sector;
861 }
862 
863 static inline unsigned int blk_rq_bytes(const struct request *rq)
864 {
865 	return rq->__data_len;
866 }
867 
868 static inline int blk_rq_cur_bytes(const struct request *rq)
869 {
870 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
871 }
872 
873 extern unsigned int blk_rq_err_bytes(const struct request *rq);
874 
875 static inline unsigned int blk_rq_sectors(const struct request *rq)
876 {
877 	return blk_rq_bytes(rq) >> 9;
878 }
879 
880 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
881 {
882 	return blk_rq_cur_bytes(rq) >> 9;
883 }
884 
885 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
886 						     unsigned int cmd_flags)
887 {
888 	if (unlikely(cmd_flags & REQ_DISCARD))
889 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
890 
891 	if (unlikely(cmd_flags & REQ_WRITE_SAME))
892 		return q->limits.max_write_same_sectors;
893 
894 	return q->limits.max_sectors;
895 }
896 
897 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
898 {
899 	struct request_queue *q = rq->q;
900 
901 	if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
902 		return q->limits.max_hw_sectors;
903 
904 	return blk_queue_get_max_sectors(q, rq->cmd_flags);
905 }
906 
907 static inline unsigned int blk_rq_count_bios(struct request *rq)
908 {
909 	unsigned int nr_bios = 0;
910 	struct bio *bio;
911 
912 	__rq_for_each_bio(bio, rq)
913 		nr_bios++;
914 
915 	return nr_bios;
916 }
917 
918 /*
919  * Request issue related functions.
920  */
921 extern struct request *blk_peek_request(struct request_queue *q);
922 extern void blk_start_request(struct request *rq);
923 extern struct request *blk_fetch_request(struct request_queue *q);
924 
925 /*
926  * Request completion related functions.
927  *
928  * blk_update_request() completes given number of bytes and updates
929  * the request without completing it.
930  *
931  * blk_end_request() and friends.  __blk_end_request() must be called
932  * with the request queue spinlock acquired.
933  *
934  * Several drivers define their own end_request and call
935  * blk_end_request() for parts of the original function.
936  * This prevents code duplication in drivers.
937  */
938 extern bool blk_update_request(struct request *rq, int error,
939 			       unsigned int nr_bytes);
940 extern bool blk_end_request(struct request *rq, int error,
941 			    unsigned int nr_bytes);
942 extern void blk_end_request_all(struct request *rq, int error);
943 extern bool blk_end_request_cur(struct request *rq, int error);
944 extern bool blk_end_request_err(struct request *rq, int error);
945 extern bool __blk_end_request(struct request *rq, int error,
946 			      unsigned int nr_bytes);
947 extern void __blk_end_request_all(struct request *rq, int error);
948 extern bool __blk_end_request_cur(struct request *rq, int error);
949 extern bool __blk_end_request_err(struct request *rq, int error);
950 
951 extern void blk_complete_request(struct request *);
952 extern void __blk_complete_request(struct request *);
953 extern void blk_abort_request(struct request *);
954 extern void blk_unprep_request(struct request *);
955 
956 /*
957  * Access functions for manipulating queue properties
958  */
959 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
960 					spinlock_t *lock, int node_id);
961 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
962 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
963 						      request_fn_proc *, spinlock_t *);
964 extern void blk_cleanup_queue(struct request_queue *);
965 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
966 extern void blk_queue_bounce_limit(struct request_queue *, u64);
967 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
968 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
969 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
970 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
971 extern void blk_queue_max_discard_sectors(struct request_queue *q,
972 		unsigned int max_discard_sectors);
973 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
974 		unsigned int max_write_same_sectors);
975 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
976 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
977 extern void blk_queue_alignment_offset(struct request_queue *q,
978 				       unsigned int alignment);
979 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
980 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
981 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
982 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
983 extern void blk_set_default_limits(struct queue_limits *lim);
984 extern void blk_set_stacking_limits(struct queue_limits *lim);
985 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
986 			    sector_t offset);
987 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
988 			    sector_t offset);
989 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
990 			      sector_t offset);
991 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
992 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
993 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
994 extern int blk_queue_dma_drain(struct request_queue *q,
995 			       dma_drain_needed_fn *dma_drain_needed,
996 			       void *buf, unsigned int size);
997 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
998 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
999 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1000 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1001 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
1002 extern void blk_queue_dma_alignment(struct request_queue *, int);
1003 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1004 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1005 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1006 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1007 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1008 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1009 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1010 
1011 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1012 extern int blk_bio_map_sg(struct request_queue *q, struct bio *bio,
1013 			  struct scatterlist *sglist);
1014 extern void blk_dump_rq_flags(struct request *, char *);
1015 extern long nr_blockdev_pages(void);
1016 
1017 bool __must_check blk_get_queue(struct request_queue *);
1018 struct request_queue *blk_alloc_queue(gfp_t);
1019 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1020 extern void blk_put_queue(struct request_queue *);
1021 
1022 /*
1023  * block layer runtime pm functions
1024  */
1025 #ifdef CONFIG_PM_RUNTIME
1026 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1027 extern int blk_pre_runtime_suspend(struct request_queue *q);
1028 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1029 extern void blk_pre_runtime_resume(struct request_queue *q);
1030 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1031 #else
1032 static inline void blk_pm_runtime_init(struct request_queue *q,
1033 	struct device *dev) {}
1034 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1035 {
1036 	return -ENOSYS;
1037 }
1038 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1039 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1040 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1041 #endif
1042 
1043 /*
1044  * blk_plug permits building a queue of related requests by holding the I/O
1045  * fragments for a short period. This allows merging of sequential requests
1046  * into single larger request. As the requests are moved from a per-task list to
1047  * the device's request_queue in a batch, this results in improved scalability
1048  * as the lock contention for request_queue lock is reduced.
1049  *
1050  * It is ok not to disable preemption when adding the request to the plug list
1051  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1052  * the plug list when the task sleeps by itself. For details, please see
1053  * schedule() where blk_schedule_flush_plug() is called.
1054  */
1055 struct blk_plug {
1056 	unsigned long magic; /* detect uninitialized use-cases */
1057 	struct list_head list; /* requests */
1058 	struct list_head mq_list; /* blk-mq requests */
1059 	struct list_head cb_list; /* md requires an unplug callback */
1060 };
1061 #define BLK_MAX_REQUEST_COUNT 16
1062 
1063 struct blk_plug_cb;
1064 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1065 struct blk_plug_cb {
1066 	struct list_head list;
1067 	blk_plug_cb_fn callback;
1068 	void *data;
1069 };
1070 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1071 					     void *data, int size);
1072 extern void blk_start_plug(struct blk_plug *);
1073 extern void blk_finish_plug(struct blk_plug *);
1074 extern void blk_flush_plug_list(struct blk_plug *, bool);
1075 
1076 static inline void blk_flush_plug(struct task_struct *tsk)
1077 {
1078 	struct blk_plug *plug = tsk->plug;
1079 
1080 	if (plug)
1081 		blk_flush_plug_list(plug, false);
1082 }
1083 
1084 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1085 {
1086 	struct blk_plug *plug = tsk->plug;
1087 
1088 	if (plug)
1089 		blk_flush_plug_list(plug, true);
1090 }
1091 
1092 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1093 {
1094 	struct blk_plug *plug = tsk->plug;
1095 
1096 	return plug &&
1097 		(!list_empty(&plug->list) ||
1098 		 !list_empty(&plug->mq_list) ||
1099 		 !list_empty(&plug->cb_list));
1100 }
1101 
1102 /*
1103  * tag stuff
1104  */
1105 #define blk_rq_tagged(rq)		((rq)->cmd_flags & REQ_QUEUED)
1106 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1107 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1108 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1109 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
1110 extern void blk_queue_free_tags(struct request_queue *);
1111 extern int blk_queue_resize_tags(struct request_queue *, int);
1112 extern void blk_queue_invalidate_tags(struct request_queue *);
1113 extern struct blk_queue_tag *blk_init_tags(int);
1114 extern void blk_free_tags(struct blk_queue_tag *);
1115 
1116 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1117 						int tag)
1118 {
1119 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1120 		return NULL;
1121 	return bqt->tag_index[tag];
1122 }
1123 
1124 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1125 
1126 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1127 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1128 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1129 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1130 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1131 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1132 			sector_t nr_sects, gfp_t gfp_mask);
1133 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1134 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1135 {
1136 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1137 				    nr_blocks << (sb->s_blocksize_bits - 9),
1138 				    gfp_mask, flags);
1139 }
1140 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1141 		sector_t nr_blocks, gfp_t gfp_mask)
1142 {
1143 	return blkdev_issue_zeroout(sb->s_bdev,
1144 				    block << (sb->s_blocksize_bits - 9),
1145 				    nr_blocks << (sb->s_blocksize_bits - 9),
1146 				    gfp_mask);
1147 }
1148 
1149 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1150 
1151 enum blk_default_limits {
1152 	BLK_MAX_SEGMENTS	= 128,
1153 	BLK_SAFE_MAX_SECTORS	= 255,
1154 	BLK_DEF_MAX_SECTORS	= 1024,
1155 	BLK_MAX_SEGMENT_SIZE	= 65536,
1156 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1157 };
1158 
1159 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1160 
1161 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1162 {
1163 	return q->limits.bounce_pfn;
1164 }
1165 
1166 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1167 {
1168 	return q->limits.seg_boundary_mask;
1169 }
1170 
1171 static inline unsigned int queue_max_sectors(struct request_queue *q)
1172 {
1173 	return q->limits.max_sectors;
1174 }
1175 
1176 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1177 {
1178 	return q->limits.max_hw_sectors;
1179 }
1180 
1181 static inline unsigned short queue_max_segments(struct request_queue *q)
1182 {
1183 	return q->limits.max_segments;
1184 }
1185 
1186 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1187 {
1188 	return q->limits.max_segment_size;
1189 }
1190 
1191 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1192 {
1193 	int retval = 512;
1194 
1195 	if (q && q->limits.logical_block_size)
1196 		retval = q->limits.logical_block_size;
1197 
1198 	return retval;
1199 }
1200 
1201 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1202 {
1203 	return queue_logical_block_size(bdev_get_queue(bdev));
1204 }
1205 
1206 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1207 {
1208 	return q->limits.physical_block_size;
1209 }
1210 
1211 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1212 {
1213 	return queue_physical_block_size(bdev_get_queue(bdev));
1214 }
1215 
1216 static inline unsigned int queue_io_min(struct request_queue *q)
1217 {
1218 	return q->limits.io_min;
1219 }
1220 
1221 static inline int bdev_io_min(struct block_device *bdev)
1222 {
1223 	return queue_io_min(bdev_get_queue(bdev));
1224 }
1225 
1226 static inline unsigned int queue_io_opt(struct request_queue *q)
1227 {
1228 	return q->limits.io_opt;
1229 }
1230 
1231 static inline int bdev_io_opt(struct block_device *bdev)
1232 {
1233 	return queue_io_opt(bdev_get_queue(bdev));
1234 }
1235 
1236 static inline int queue_alignment_offset(struct request_queue *q)
1237 {
1238 	if (q->limits.misaligned)
1239 		return -1;
1240 
1241 	return q->limits.alignment_offset;
1242 }
1243 
1244 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1245 {
1246 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1247 	unsigned int alignment = (sector << 9) & (granularity - 1);
1248 
1249 	return (granularity + lim->alignment_offset - alignment)
1250 		& (granularity - 1);
1251 }
1252 
1253 static inline int bdev_alignment_offset(struct block_device *bdev)
1254 {
1255 	struct request_queue *q = bdev_get_queue(bdev);
1256 
1257 	if (q->limits.misaligned)
1258 		return -1;
1259 
1260 	if (bdev != bdev->bd_contains)
1261 		return bdev->bd_part->alignment_offset;
1262 
1263 	return q->limits.alignment_offset;
1264 }
1265 
1266 static inline int queue_discard_alignment(struct request_queue *q)
1267 {
1268 	if (q->limits.discard_misaligned)
1269 		return -1;
1270 
1271 	return q->limits.discard_alignment;
1272 }
1273 
1274 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1275 {
1276 	unsigned int alignment, granularity, offset;
1277 
1278 	if (!lim->max_discard_sectors)
1279 		return 0;
1280 
1281 	/* Why are these in bytes, not sectors? */
1282 	alignment = lim->discard_alignment >> 9;
1283 	granularity = lim->discard_granularity >> 9;
1284 	if (!granularity)
1285 		return 0;
1286 
1287 	/* Offset of the partition start in 'granularity' sectors */
1288 	offset = sector_div(sector, granularity);
1289 
1290 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1291 	offset = (granularity + alignment - offset) % granularity;
1292 
1293 	/* Turn it back into bytes, gaah */
1294 	return offset << 9;
1295 }
1296 
1297 static inline int bdev_discard_alignment(struct block_device *bdev)
1298 {
1299 	struct request_queue *q = bdev_get_queue(bdev);
1300 
1301 	if (bdev != bdev->bd_contains)
1302 		return bdev->bd_part->discard_alignment;
1303 
1304 	return q->limits.discard_alignment;
1305 }
1306 
1307 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1308 {
1309 	if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1310 		return 1;
1311 
1312 	return 0;
1313 }
1314 
1315 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1316 {
1317 	return queue_discard_zeroes_data(bdev_get_queue(bdev));
1318 }
1319 
1320 static inline unsigned int bdev_write_same(struct block_device *bdev)
1321 {
1322 	struct request_queue *q = bdev_get_queue(bdev);
1323 
1324 	if (q)
1325 		return q->limits.max_write_same_sectors;
1326 
1327 	return 0;
1328 }
1329 
1330 static inline int queue_dma_alignment(struct request_queue *q)
1331 {
1332 	return q ? q->dma_alignment : 511;
1333 }
1334 
1335 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1336 				 unsigned int len)
1337 {
1338 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1339 	return !(addr & alignment) && !(len & alignment);
1340 }
1341 
1342 /* assumes size > 256 */
1343 static inline unsigned int blksize_bits(unsigned int size)
1344 {
1345 	unsigned int bits = 8;
1346 	do {
1347 		bits++;
1348 		size >>= 1;
1349 	} while (size > 256);
1350 	return bits;
1351 }
1352 
1353 static inline unsigned int block_size(struct block_device *bdev)
1354 {
1355 	return bdev->bd_block_size;
1356 }
1357 
1358 static inline bool queue_flush_queueable(struct request_queue *q)
1359 {
1360 	return !q->flush_not_queueable;
1361 }
1362 
1363 typedef struct {struct page *v;} Sector;
1364 
1365 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1366 
1367 static inline void put_dev_sector(Sector p)
1368 {
1369 	page_cache_release(p.v);
1370 }
1371 
1372 struct work_struct;
1373 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
1374 int kblockd_schedule_delayed_work(struct request_queue *q, 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 #define INTEGRITY_FLAG_READ	2	/* verify data integrity on read */
1426 #define INTEGRITY_FLAG_WRITE	4	/* generate data integrity on write */
1427 
1428 struct blk_integrity_exchg {
1429 	void			*prot_buf;
1430 	void			*data_buf;
1431 	sector_t		sector;
1432 	unsigned int		data_size;
1433 	unsigned short		sector_size;
1434 	const char		*disk_name;
1435 };
1436 
1437 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1438 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1439 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1440 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1441 
1442 struct blk_integrity {
1443 	integrity_gen_fn	*generate_fn;
1444 	integrity_vrfy_fn	*verify_fn;
1445 	integrity_set_tag_fn	*set_tag_fn;
1446 	integrity_get_tag_fn	*get_tag_fn;
1447 
1448 	unsigned short		flags;
1449 	unsigned short		tuple_size;
1450 	unsigned short		sector_size;
1451 	unsigned short		tag_size;
1452 
1453 	const char		*name;
1454 
1455 	struct kobject		kobj;
1456 };
1457 
1458 extern bool blk_integrity_is_initialized(struct gendisk *);
1459 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1460 extern void blk_integrity_unregister(struct gendisk *);
1461 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1462 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1463 				   struct scatterlist *);
1464 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1465 extern int blk_integrity_merge_rq(struct request_queue *, struct request *,
1466 				  struct request *);
1467 extern int blk_integrity_merge_bio(struct request_queue *, struct request *,
1468 				   struct bio *);
1469 
1470 static inline
1471 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1472 {
1473 	return bdev->bd_disk->integrity;
1474 }
1475 
1476 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1477 {
1478 	return disk->integrity;
1479 }
1480 
1481 static inline int blk_integrity_rq(struct request *rq)
1482 {
1483 	if (rq->bio == NULL)
1484 		return 0;
1485 
1486 	return bio_integrity(rq->bio);
1487 }
1488 
1489 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1490 						    unsigned int segs)
1491 {
1492 	q->limits.max_integrity_segments = segs;
1493 }
1494 
1495 static inline unsigned short
1496 queue_max_integrity_segments(struct request_queue *q)
1497 {
1498 	return q->limits.max_integrity_segments;
1499 }
1500 
1501 #else /* CONFIG_BLK_DEV_INTEGRITY */
1502 
1503 struct bio;
1504 struct block_device;
1505 struct gendisk;
1506 struct blk_integrity;
1507 
1508 static inline int blk_integrity_rq(struct request *rq)
1509 {
1510 	return 0;
1511 }
1512 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1513 					    struct bio *b)
1514 {
1515 	return 0;
1516 }
1517 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1518 					  struct bio *b,
1519 					  struct scatterlist *s)
1520 {
1521 	return 0;
1522 }
1523 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1524 {
1525 	return 0;
1526 }
1527 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1528 {
1529 	return NULL;
1530 }
1531 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1532 {
1533 	return 0;
1534 }
1535 static inline int blk_integrity_register(struct gendisk *d,
1536 					 struct blk_integrity *b)
1537 {
1538 	return 0;
1539 }
1540 static inline void blk_integrity_unregister(struct gendisk *d)
1541 {
1542 }
1543 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1544 						    unsigned int segs)
1545 {
1546 }
1547 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1548 {
1549 	return 0;
1550 }
1551 static inline int blk_integrity_merge_rq(struct request_queue *rq,
1552 					 struct request *r1,
1553 					 struct request *r2)
1554 {
1555 	return 0;
1556 }
1557 static inline int blk_integrity_merge_bio(struct request_queue *rq,
1558 					  struct request *r,
1559 					  struct bio *b)
1560 {
1561 	return 0;
1562 }
1563 static inline bool blk_integrity_is_initialized(struct gendisk *g)
1564 {
1565 	return 0;
1566 }
1567 
1568 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1569 
1570 struct block_device_operations {
1571 	int (*open) (struct block_device *, fmode_t);
1572 	void (*release) (struct gendisk *, fmode_t);
1573 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1574 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1575 	int (*direct_access) (struct block_device *, sector_t,
1576 						void **, unsigned long *);
1577 	unsigned int (*check_events) (struct gendisk *disk,
1578 				      unsigned int clearing);
1579 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
1580 	int (*media_changed) (struct gendisk *);
1581 	void (*unlock_native_capacity) (struct gendisk *);
1582 	int (*revalidate_disk) (struct gendisk *);
1583 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1584 	/* this callback is with swap_lock and sometimes page table lock held */
1585 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1586 	struct module *owner;
1587 };
1588 
1589 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1590 				 unsigned long);
1591 #else /* CONFIG_BLOCK */
1592 /*
1593  * stubs for when the block layer is configured out
1594  */
1595 #define buffer_heads_over_limit 0
1596 
1597 static inline long nr_blockdev_pages(void)
1598 {
1599 	return 0;
1600 }
1601 
1602 struct blk_plug {
1603 };
1604 
1605 static inline void blk_start_plug(struct blk_plug *plug)
1606 {
1607 }
1608 
1609 static inline void blk_finish_plug(struct blk_plug *plug)
1610 {
1611 }
1612 
1613 static inline void blk_flush_plug(struct task_struct *task)
1614 {
1615 }
1616 
1617 static inline void blk_schedule_flush_plug(struct task_struct *task)
1618 {
1619 }
1620 
1621 
1622 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1623 {
1624 	return false;
1625 }
1626 
1627 #endif /* CONFIG_BLOCK */
1628 
1629 #endif
1630