xref: /linux-6.15/include/linux/blkdev.h (revision 07100877)
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  *
95  * If you modify this structure, make sure to update blk_rq_init() and
96  * especially blk_mq_rq_ctx_init() to take care of the added fields.
97  */
98 struct request {
99 	struct list_head queuelist;
100 	union {
101 		struct call_single_data csd;
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 
182 	int tag;
183 	int errors;
184 
185 	/*
186 	 * when request is used as a packet command carrier
187 	 */
188 	unsigned char __cmd[BLK_MAX_CDB];
189 	unsigned char *cmd;
190 	unsigned short cmd_len;
191 
192 	unsigned int extra_len;	/* length of alignment and padding */
193 	unsigned int sense_len;
194 	unsigned int resid_len;	/* residual count */
195 	void *sense;
196 
197 	unsigned long deadline;
198 	struct list_head timeout_list;
199 	unsigned int timeout;
200 	int retries;
201 
202 	/*
203 	 * completion callback.
204 	 */
205 	rq_end_io_fn *end_io;
206 	void *end_io_data;
207 
208 	/* for bidi */
209 	struct request *next_rq;
210 };
211 
212 static inline unsigned short req_get_ioprio(struct request *req)
213 {
214 	return req->ioprio;
215 }
216 
217 /*
218  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
219  * requests. Some step values could eventually be made generic.
220  */
221 struct request_pm_state
222 {
223 	/* PM state machine step value, currently driver specific */
224 	int	pm_step;
225 	/* requested PM state value (S1, S2, S3, S4, ...) */
226 	u32	pm_state;
227 	void*	data;		/* for driver use */
228 };
229 
230 #include <linux/elevator.h>
231 
232 struct blk_queue_ctx;
233 
234 typedef void (request_fn_proc) (struct request_queue *q);
235 typedef void (make_request_fn) (struct request_queue *q, struct bio *bio);
236 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
237 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
238 
239 struct bio_vec;
240 struct bvec_merge_data {
241 	struct block_device *bi_bdev;
242 	sector_t bi_sector;
243 	unsigned bi_size;
244 	unsigned long bi_rw;
245 };
246 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
247 			     struct bio_vec *);
248 typedef void (softirq_done_fn)(struct request *);
249 typedef int (dma_drain_needed_fn)(struct request *);
250 typedef int (lld_busy_fn) (struct request_queue *q);
251 typedef int (bsg_job_fn) (struct bsg_job *);
252 
253 enum blk_eh_timer_return {
254 	BLK_EH_NOT_HANDLED,
255 	BLK_EH_HANDLED,
256 	BLK_EH_RESET_TIMER,
257 };
258 
259 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
260 
261 enum blk_queue_state {
262 	Queue_down,
263 	Queue_up,
264 };
265 
266 struct blk_queue_tag {
267 	struct request **tag_index;	/* map of busy tags */
268 	unsigned long *tag_map;		/* bit map of free/busy tags */
269 	int busy;			/* current depth */
270 	int max_depth;			/* what we will send to device */
271 	int real_max_depth;		/* what the array can hold */
272 	atomic_t refcnt;		/* map can be shared */
273 };
274 
275 #define BLK_SCSI_MAX_CMDS	(256)
276 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
277 
278 struct queue_limits {
279 	unsigned long		bounce_pfn;
280 	unsigned long		seg_boundary_mask;
281 
282 	unsigned int		max_hw_sectors;
283 	unsigned int		chunk_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 __percpu	*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 list_head	requeue_list;
467 	spinlock_t		requeue_lock;
468 	struct work_struct	requeue_work;
469 
470 	struct mutex		sysfs_lock;
471 
472 	int			bypass_depth;
473 
474 #if defined(CONFIG_BLK_DEV_BSG)
475 	bsg_job_fn		*bsg_job_fn;
476 	int			bsg_job_size;
477 	struct bsg_class_device bsg_dev;
478 #endif
479 
480 #ifdef CONFIG_BLK_DEV_THROTTLING
481 	/* Throttle data */
482 	struct throtl_data *td;
483 #endif
484 	struct rcu_head		rcu_head;
485 	wait_queue_head_t	mq_freeze_wq;
486 	struct percpu_counter	mq_usage_counter;
487 	struct list_head	all_q_node;
488 
489 	struct blk_mq_tag_set	*tag_set;
490 	struct list_head	tag_set_list;
491 };
492 
493 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
494 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
495 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
496 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
497 #define QUEUE_FLAG_DYING	5	/* queue being torn down */
498 #define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
499 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
500 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
501 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
502 #define QUEUE_FLAG_FAIL_IO     10	/* fake timeout */
503 #define QUEUE_FLAG_STACKABLE   11	/* supports request stacking */
504 #define QUEUE_FLAG_NONROT      12	/* non-rotational device (SSD) */
505 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
506 #define QUEUE_FLAG_IO_STAT     13	/* do IO stats */
507 #define QUEUE_FLAG_DISCARD     14	/* supports DISCARD */
508 #define QUEUE_FLAG_NOXMERGES   15	/* No extended merges */
509 #define QUEUE_FLAG_ADD_RANDOM  16	/* Contributes to random pool */
510 #define QUEUE_FLAG_SECDISCARD  17	/* supports SECDISCARD */
511 #define QUEUE_FLAG_SAME_FORCE  18	/* force complete on same CPU */
512 #define QUEUE_FLAG_DEAD        19	/* queue tear-down finished */
513 #define QUEUE_FLAG_INIT_DONE   20	/* queue is initialized */
514 #define QUEUE_FLAG_NO_SG_MERGE 21	/* don't attempt to merge SG segments*/
515 
516 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
517 				 (1 << QUEUE_FLAG_STACKABLE)	|	\
518 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
519 				 (1 << QUEUE_FLAG_ADD_RANDOM))
520 
521 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
522 				 (1 << QUEUE_FLAG_SAME_COMP))
523 
524 static inline void queue_lockdep_assert_held(struct request_queue *q)
525 {
526 	if (q->queue_lock)
527 		lockdep_assert_held(q->queue_lock);
528 }
529 
530 static inline void queue_flag_set_unlocked(unsigned int flag,
531 					   struct request_queue *q)
532 {
533 	__set_bit(flag, &q->queue_flags);
534 }
535 
536 static inline int queue_flag_test_and_clear(unsigned int flag,
537 					    struct request_queue *q)
538 {
539 	queue_lockdep_assert_held(q);
540 
541 	if (test_bit(flag, &q->queue_flags)) {
542 		__clear_bit(flag, &q->queue_flags);
543 		return 1;
544 	}
545 
546 	return 0;
547 }
548 
549 static inline int queue_flag_test_and_set(unsigned int flag,
550 					  struct request_queue *q)
551 {
552 	queue_lockdep_assert_held(q);
553 
554 	if (!test_bit(flag, &q->queue_flags)) {
555 		__set_bit(flag, &q->queue_flags);
556 		return 0;
557 	}
558 
559 	return 1;
560 }
561 
562 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
563 {
564 	queue_lockdep_assert_held(q);
565 	__set_bit(flag, &q->queue_flags);
566 }
567 
568 static inline void queue_flag_clear_unlocked(unsigned int flag,
569 					     struct request_queue *q)
570 {
571 	__clear_bit(flag, &q->queue_flags);
572 }
573 
574 static inline int queue_in_flight(struct request_queue *q)
575 {
576 	return q->in_flight[0] + q->in_flight[1];
577 }
578 
579 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
580 {
581 	queue_lockdep_assert_held(q);
582 	__clear_bit(flag, &q->queue_flags);
583 }
584 
585 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
586 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
587 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
588 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
589 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
590 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
591 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
592 #define blk_queue_noxmerges(q)	\
593 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
594 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
595 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
596 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
597 #define blk_queue_stackable(q)	\
598 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
599 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
600 #define blk_queue_secdiscard(q)	(blk_queue_discard(q) && \
601 	test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
602 
603 #define blk_noretry_request(rq) \
604 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
605 			     REQ_FAILFAST_DRIVER))
606 
607 #define blk_account_rq(rq) \
608 	(((rq)->cmd_flags & REQ_STARTED) && \
609 	 ((rq)->cmd_type == REQ_TYPE_FS))
610 
611 #define blk_pm_request(rq)	\
612 	((rq)->cmd_type == REQ_TYPE_PM_SUSPEND || \
613 	 (rq)->cmd_type == REQ_TYPE_PM_RESUME)
614 
615 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
616 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
617 /* rq->queuelist of dequeued request must be list_empty() */
618 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
619 
620 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
621 
622 #define rq_data_dir(rq)		(((rq)->cmd_flags & 1) != 0)
623 
624 /*
625  * Driver can handle struct request, if it either has an old style
626  * request_fn defined, or is blk-mq based.
627  */
628 static inline bool queue_is_rq_based(struct request_queue *q)
629 {
630 	return q->request_fn || q->mq_ops;
631 }
632 
633 static inline unsigned int blk_queue_cluster(struct request_queue *q)
634 {
635 	return q->limits.cluster;
636 }
637 
638 /*
639  * We regard a request as sync, if either a read or a sync write
640  */
641 static inline bool rw_is_sync(unsigned int rw_flags)
642 {
643 	return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
644 }
645 
646 static inline bool rq_is_sync(struct request *rq)
647 {
648 	return rw_is_sync(rq->cmd_flags);
649 }
650 
651 static inline bool blk_rl_full(struct request_list *rl, bool sync)
652 {
653 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
654 
655 	return rl->flags & flag;
656 }
657 
658 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
659 {
660 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
661 
662 	rl->flags |= flag;
663 }
664 
665 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
666 {
667 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
668 
669 	rl->flags &= ~flag;
670 }
671 
672 static inline bool rq_mergeable(struct request *rq)
673 {
674 	if (rq->cmd_type != REQ_TYPE_FS)
675 		return false;
676 
677 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
678 		return false;
679 
680 	return true;
681 }
682 
683 static inline bool blk_check_merge_flags(unsigned int flags1,
684 					 unsigned int flags2)
685 {
686 	if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
687 		return false;
688 
689 	if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
690 		return false;
691 
692 	if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
693 		return false;
694 
695 	return true;
696 }
697 
698 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
699 {
700 	if (bio_data(a) == bio_data(b))
701 		return true;
702 
703 	return false;
704 }
705 
706 /*
707  * q->prep_rq_fn return values
708  */
709 #define BLKPREP_OK		0	/* serve it */
710 #define BLKPREP_KILL		1	/* fatal error, kill */
711 #define BLKPREP_DEFER		2	/* leave on queue */
712 
713 extern unsigned long blk_max_low_pfn, blk_max_pfn;
714 
715 /*
716  * standard bounce addresses:
717  *
718  * BLK_BOUNCE_HIGH	: bounce all highmem pages
719  * BLK_BOUNCE_ANY	: don't bounce anything
720  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
721  */
722 
723 #if BITS_PER_LONG == 32
724 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
725 #else
726 #define BLK_BOUNCE_HIGH		-1ULL
727 #endif
728 #define BLK_BOUNCE_ANY		(-1ULL)
729 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
730 
731 /*
732  * default timeout for SG_IO if none specified
733  */
734 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
735 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
736 
737 #ifdef CONFIG_BOUNCE
738 extern int init_emergency_isa_pool(void);
739 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
740 #else
741 static inline int init_emergency_isa_pool(void)
742 {
743 	return 0;
744 }
745 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
746 {
747 }
748 #endif /* CONFIG_MMU */
749 
750 struct rq_map_data {
751 	struct page **pages;
752 	int page_order;
753 	int nr_entries;
754 	unsigned long offset;
755 	int null_mapped;
756 	int from_user;
757 };
758 
759 struct req_iterator {
760 	struct bvec_iter iter;
761 	struct bio *bio;
762 };
763 
764 /* This should not be used directly - use rq_for_each_segment */
765 #define for_each_bio(_bio)		\
766 	for (; _bio; _bio = _bio->bi_next)
767 #define __rq_for_each_bio(_bio, rq)	\
768 	if ((rq->bio))			\
769 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
770 
771 #define rq_for_each_segment(bvl, _rq, _iter)			\
772 	__rq_for_each_bio(_iter.bio, _rq)			\
773 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
774 
775 #define rq_iter_last(bvec, _iter)				\
776 		(_iter.bio->bi_next == NULL &&			\
777 		 bio_iter_last(bvec, _iter.iter))
778 
779 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
780 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
781 #endif
782 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
783 extern void rq_flush_dcache_pages(struct request *rq);
784 #else
785 static inline void rq_flush_dcache_pages(struct request *rq)
786 {
787 }
788 #endif
789 
790 extern int blk_register_queue(struct gendisk *disk);
791 extern void blk_unregister_queue(struct gendisk *disk);
792 extern void generic_make_request(struct bio *bio);
793 extern void blk_rq_init(struct request_queue *q, struct request *rq);
794 extern void blk_put_request(struct request *);
795 extern void __blk_put_request(struct request_queue *, struct request *);
796 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
797 extern struct request *blk_make_request(struct request_queue *, struct bio *,
798 					gfp_t);
799 extern void blk_rq_set_block_pc(struct request *);
800 extern void blk_requeue_request(struct request_queue *, struct request *);
801 extern void blk_add_request_payload(struct request *rq, struct page *page,
802 		unsigned int len);
803 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
804 extern int blk_lld_busy(struct request_queue *q);
805 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
806 			     struct bio_set *bs, gfp_t gfp_mask,
807 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
808 			     void *data);
809 extern void blk_rq_unprep_clone(struct request *rq);
810 extern int blk_insert_cloned_request(struct request_queue *q,
811 				     struct request *rq);
812 extern void blk_delay_queue(struct request_queue *, unsigned long);
813 extern void blk_recount_segments(struct request_queue *, struct bio *);
814 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
815 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
816 			      unsigned int, void __user *);
817 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
818 			  unsigned int, void __user *);
819 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
820 			 struct scsi_ioctl_command __user *);
821 
822 extern void blk_queue_bio(struct request_queue *q, struct bio *bio);
823 
824 /*
825  * A queue has just exitted congestion.  Note this in the global counter of
826  * congested queues, and wake up anyone who was waiting for requests to be
827  * put back.
828  */
829 static inline void blk_clear_queue_congested(struct request_queue *q, int sync)
830 {
831 	clear_bdi_congested(&q->backing_dev_info, sync);
832 }
833 
834 /*
835  * A queue has just entered congestion.  Flag that in the queue's VM-visible
836  * state flags and increment the global gounter of congested queues.
837  */
838 static inline void blk_set_queue_congested(struct request_queue *q, int sync)
839 {
840 	set_bdi_congested(&q->backing_dev_info, sync);
841 }
842 
843 extern void blk_start_queue(struct request_queue *q);
844 extern void blk_stop_queue(struct request_queue *q);
845 extern void blk_sync_queue(struct request_queue *q);
846 extern void __blk_stop_queue(struct request_queue *q);
847 extern void __blk_run_queue(struct request_queue *q);
848 extern void blk_run_queue(struct request_queue *);
849 extern void blk_run_queue_async(struct request_queue *q);
850 extern int blk_rq_map_user(struct request_queue *, struct request *,
851 			   struct rq_map_data *, void __user *, unsigned long,
852 			   gfp_t);
853 extern int blk_rq_unmap_user(struct bio *);
854 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
855 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
856 			       struct rq_map_data *, const struct sg_iovec *,
857 			       int, unsigned int, gfp_t);
858 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
859 			  struct request *, int);
860 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
861 				  struct request *, int, rq_end_io_fn *);
862 
863 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
864 {
865 	return bdev->bd_disk->queue;
866 }
867 
868 /*
869  * blk_rq_pos()			: the current sector
870  * blk_rq_bytes()		: bytes left in the entire request
871  * blk_rq_cur_bytes()		: bytes left in the current segment
872  * blk_rq_err_bytes()		: bytes left till the next error boundary
873  * blk_rq_sectors()		: sectors left in the entire request
874  * blk_rq_cur_sectors()		: sectors left in the current segment
875  */
876 static inline sector_t blk_rq_pos(const struct request *rq)
877 {
878 	return rq->__sector;
879 }
880 
881 static inline unsigned int blk_rq_bytes(const struct request *rq)
882 {
883 	return rq->__data_len;
884 }
885 
886 static inline int blk_rq_cur_bytes(const struct request *rq)
887 {
888 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
889 }
890 
891 extern unsigned int blk_rq_err_bytes(const struct request *rq);
892 
893 static inline unsigned int blk_rq_sectors(const struct request *rq)
894 {
895 	return blk_rq_bytes(rq) >> 9;
896 }
897 
898 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
899 {
900 	return blk_rq_cur_bytes(rq) >> 9;
901 }
902 
903 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
904 						     unsigned int cmd_flags)
905 {
906 	if (unlikely(cmd_flags & REQ_DISCARD))
907 		return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
908 
909 	if (unlikely(cmd_flags & REQ_WRITE_SAME))
910 		return q->limits.max_write_same_sectors;
911 
912 	return q->limits.max_sectors;
913 }
914 
915 /*
916  * Return maximum size of a request at given offset. Only valid for
917  * file system requests.
918  */
919 static inline unsigned int blk_max_size_offset(struct request_queue *q,
920 					       sector_t offset)
921 {
922 	if (!q->limits.chunk_sectors)
923 		return q->limits.max_sectors;
924 
925 	return q->limits.chunk_sectors -
926 			(offset & (q->limits.chunk_sectors - 1));
927 }
928 
929 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
930 {
931 	struct request_queue *q = rq->q;
932 
933 	if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
934 		return q->limits.max_hw_sectors;
935 
936 	if (!q->limits.chunk_sectors)
937 		return blk_queue_get_max_sectors(q, rq->cmd_flags);
938 
939 	return min(blk_max_size_offset(q, blk_rq_pos(rq)),
940 			blk_queue_get_max_sectors(q, rq->cmd_flags));
941 }
942 
943 static inline unsigned int blk_rq_count_bios(struct request *rq)
944 {
945 	unsigned int nr_bios = 0;
946 	struct bio *bio;
947 
948 	__rq_for_each_bio(bio, rq)
949 		nr_bios++;
950 
951 	return nr_bios;
952 }
953 
954 /*
955  * Request issue related functions.
956  */
957 extern struct request *blk_peek_request(struct request_queue *q);
958 extern void blk_start_request(struct request *rq);
959 extern struct request *blk_fetch_request(struct request_queue *q);
960 
961 /*
962  * Request completion related functions.
963  *
964  * blk_update_request() completes given number of bytes and updates
965  * the request without completing it.
966  *
967  * blk_end_request() and friends.  __blk_end_request() must be called
968  * with the request queue spinlock acquired.
969  *
970  * Several drivers define their own end_request and call
971  * blk_end_request() for parts of the original function.
972  * This prevents code duplication in drivers.
973  */
974 extern bool blk_update_request(struct request *rq, int error,
975 			       unsigned int nr_bytes);
976 extern void blk_finish_request(struct request *rq, int error);
977 extern bool blk_end_request(struct request *rq, int error,
978 			    unsigned int nr_bytes);
979 extern void blk_end_request_all(struct request *rq, int error);
980 extern bool blk_end_request_cur(struct request *rq, int error);
981 extern bool blk_end_request_err(struct request *rq, int error);
982 extern bool __blk_end_request(struct request *rq, int error,
983 			      unsigned int nr_bytes);
984 extern void __blk_end_request_all(struct request *rq, int error);
985 extern bool __blk_end_request_cur(struct request *rq, int error);
986 extern bool __blk_end_request_err(struct request *rq, int error);
987 
988 extern void blk_complete_request(struct request *);
989 extern void __blk_complete_request(struct request *);
990 extern void blk_abort_request(struct request *);
991 extern void blk_unprep_request(struct request *);
992 
993 /*
994  * Access functions for manipulating queue properties
995  */
996 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
997 					spinlock_t *lock, int node_id);
998 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
999 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
1000 						      request_fn_proc *, spinlock_t *);
1001 extern void blk_cleanup_queue(struct request_queue *);
1002 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1003 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1004 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
1005 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1006 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1007 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1008 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1009 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1010 		unsigned int max_discard_sectors);
1011 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1012 		unsigned int max_write_same_sectors);
1013 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1014 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1015 extern void blk_queue_alignment_offset(struct request_queue *q,
1016 				       unsigned int alignment);
1017 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1018 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1019 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1020 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1021 extern void blk_set_default_limits(struct queue_limits *lim);
1022 extern void blk_set_stacking_limits(struct queue_limits *lim);
1023 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1024 			    sector_t offset);
1025 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1026 			    sector_t offset);
1027 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1028 			      sector_t offset);
1029 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1030 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1031 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1032 extern int blk_queue_dma_drain(struct request_queue *q,
1033 			       dma_drain_needed_fn *dma_drain_needed,
1034 			       void *buf, unsigned int size);
1035 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1036 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1037 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1038 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1039 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
1040 extern void blk_queue_dma_alignment(struct request_queue *, int);
1041 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1042 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1043 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1044 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1045 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
1046 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1047 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1048 
1049 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1050 extern int blk_bio_map_sg(struct request_queue *q, struct bio *bio,
1051 			  struct scatterlist *sglist);
1052 extern void blk_dump_rq_flags(struct request *, char *);
1053 extern long nr_blockdev_pages(void);
1054 
1055 bool __must_check blk_get_queue(struct request_queue *);
1056 struct request_queue *blk_alloc_queue(gfp_t);
1057 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1058 extern void blk_put_queue(struct request_queue *);
1059 
1060 /*
1061  * block layer runtime pm functions
1062  */
1063 #ifdef CONFIG_PM_RUNTIME
1064 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1065 extern int blk_pre_runtime_suspend(struct request_queue *q);
1066 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1067 extern void blk_pre_runtime_resume(struct request_queue *q);
1068 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1069 #else
1070 static inline void blk_pm_runtime_init(struct request_queue *q,
1071 	struct device *dev) {}
1072 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1073 {
1074 	return -ENOSYS;
1075 }
1076 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1077 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1078 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1079 #endif
1080 
1081 /*
1082  * blk_plug permits building a queue of related requests by holding the I/O
1083  * fragments for a short period. This allows merging of sequential requests
1084  * into single larger request. As the requests are moved from a per-task list to
1085  * the device's request_queue in a batch, this results in improved scalability
1086  * as the lock contention for request_queue lock is reduced.
1087  *
1088  * It is ok not to disable preemption when adding the request to the plug list
1089  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1090  * the plug list when the task sleeps by itself. For details, please see
1091  * schedule() where blk_schedule_flush_plug() is called.
1092  */
1093 struct blk_plug {
1094 	struct list_head list; /* requests */
1095 	struct list_head mq_list; /* blk-mq requests */
1096 	struct list_head cb_list; /* md requires an unplug callback */
1097 };
1098 #define BLK_MAX_REQUEST_COUNT 16
1099 
1100 struct blk_plug_cb;
1101 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1102 struct blk_plug_cb {
1103 	struct list_head list;
1104 	blk_plug_cb_fn callback;
1105 	void *data;
1106 };
1107 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1108 					     void *data, int size);
1109 extern void blk_start_plug(struct blk_plug *);
1110 extern void blk_finish_plug(struct blk_plug *);
1111 extern void blk_flush_plug_list(struct blk_plug *, bool);
1112 
1113 static inline void blk_flush_plug(struct task_struct *tsk)
1114 {
1115 	struct blk_plug *plug = tsk->plug;
1116 
1117 	if (plug)
1118 		blk_flush_plug_list(plug, false);
1119 }
1120 
1121 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1122 {
1123 	struct blk_plug *plug = tsk->plug;
1124 
1125 	if (plug)
1126 		blk_flush_plug_list(plug, true);
1127 }
1128 
1129 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1130 {
1131 	struct blk_plug *plug = tsk->plug;
1132 
1133 	return plug &&
1134 		(!list_empty(&plug->list) ||
1135 		 !list_empty(&plug->mq_list) ||
1136 		 !list_empty(&plug->cb_list));
1137 }
1138 
1139 /*
1140  * tag stuff
1141  */
1142 #define blk_rq_tagged(rq) \
1143 	((rq)->mq_ctx || ((rq)->cmd_flags & REQ_QUEUED))
1144 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1145 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1146 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1147 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
1148 extern void blk_queue_free_tags(struct request_queue *);
1149 extern int blk_queue_resize_tags(struct request_queue *, int);
1150 extern void blk_queue_invalidate_tags(struct request_queue *);
1151 extern struct blk_queue_tag *blk_init_tags(int);
1152 extern void blk_free_tags(struct blk_queue_tag *);
1153 
1154 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1155 						int tag)
1156 {
1157 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1158 		return NULL;
1159 	return bqt->tag_index[tag];
1160 }
1161 
1162 #define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
1163 
1164 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1165 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1166 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1167 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1168 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1169 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1170 			sector_t nr_sects, gfp_t gfp_mask);
1171 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1172 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1173 {
1174 	return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1175 				    nr_blocks << (sb->s_blocksize_bits - 9),
1176 				    gfp_mask, flags);
1177 }
1178 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1179 		sector_t nr_blocks, gfp_t gfp_mask)
1180 {
1181 	return blkdev_issue_zeroout(sb->s_bdev,
1182 				    block << (sb->s_blocksize_bits - 9),
1183 				    nr_blocks << (sb->s_blocksize_bits - 9),
1184 				    gfp_mask);
1185 }
1186 
1187 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1188 
1189 enum blk_default_limits {
1190 	BLK_MAX_SEGMENTS	= 128,
1191 	BLK_SAFE_MAX_SECTORS	= 255,
1192 	BLK_DEF_MAX_SECTORS	= 1024,
1193 	BLK_MAX_SEGMENT_SIZE	= 65536,
1194 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1195 };
1196 
1197 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1198 
1199 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1200 {
1201 	return q->limits.bounce_pfn;
1202 }
1203 
1204 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1205 {
1206 	return q->limits.seg_boundary_mask;
1207 }
1208 
1209 static inline unsigned int queue_max_sectors(struct request_queue *q)
1210 {
1211 	return q->limits.max_sectors;
1212 }
1213 
1214 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1215 {
1216 	return q->limits.max_hw_sectors;
1217 }
1218 
1219 static inline unsigned short queue_max_segments(struct request_queue *q)
1220 {
1221 	return q->limits.max_segments;
1222 }
1223 
1224 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1225 {
1226 	return q->limits.max_segment_size;
1227 }
1228 
1229 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1230 {
1231 	int retval = 512;
1232 
1233 	if (q && q->limits.logical_block_size)
1234 		retval = q->limits.logical_block_size;
1235 
1236 	return retval;
1237 }
1238 
1239 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1240 {
1241 	return queue_logical_block_size(bdev_get_queue(bdev));
1242 }
1243 
1244 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1245 {
1246 	return q->limits.physical_block_size;
1247 }
1248 
1249 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1250 {
1251 	return queue_physical_block_size(bdev_get_queue(bdev));
1252 }
1253 
1254 static inline unsigned int queue_io_min(struct request_queue *q)
1255 {
1256 	return q->limits.io_min;
1257 }
1258 
1259 static inline int bdev_io_min(struct block_device *bdev)
1260 {
1261 	return queue_io_min(bdev_get_queue(bdev));
1262 }
1263 
1264 static inline unsigned int queue_io_opt(struct request_queue *q)
1265 {
1266 	return q->limits.io_opt;
1267 }
1268 
1269 static inline int bdev_io_opt(struct block_device *bdev)
1270 {
1271 	return queue_io_opt(bdev_get_queue(bdev));
1272 }
1273 
1274 static inline int queue_alignment_offset(struct request_queue *q)
1275 {
1276 	if (q->limits.misaligned)
1277 		return -1;
1278 
1279 	return q->limits.alignment_offset;
1280 }
1281 
1282 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1283 {
1284 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1285 	unsigned int alignment = (sector << 9) & (granularity - 1);
1286 
1287 	return (granularity + lim->alignment_offset - alignment)
1288 		& (granularity - 1);
1289 }
1290 
1291 static inline int bdev_alignment_offset(struct block_device *bdev)
1292 {
1293 	struct request_queue *q = bdev_get_queue(bdev);
1294 
1295 	if (q->limits.misaligned)
1296 		return -1;
1297 
1298 	if (bdev != bdev->bd_contains)
1299 		return bdev->bd_part->alignment_offset;
1300 
1301 	return q->limits.alignment_offset;
1302 }
1303 
1304 static inline int queue_discard_alignment(struct request_queue *q)
1305 {
1306 	if (q->limits.discard_misaligned)
1307 		return -1;
1308 
1309 	return q->limits.discard_alignment;
1310 }
1311 
1312 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1313 {
1314 	unsigned int alignment, granularity, offset;
1315 
1316 	if (!lim->max_discard_sectors)
1317 		return 0;
1318 
1319 	/* Why are these in bytes, not sectors? */
1320 	alignment = lim->discard_alignment >> 9;
1321 	granularity = lim->discard_granularity >> 9;
1322 	if (!granularity)
1323 		return 0;
1324 
1325 	/* Offset of the partition start in 'granularity' sectors */
1326 	offset = sector_div(sector, granularity);
1327 
1328 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1329 	offset = (granularity + alignment - offset) % granularity;
1330 
1331 	/* Turn it back into bytes, gaah */
1332 	return offset << 9;
1333 }
1334 
1335 static inline int bdev_discard_alignment(struct block_device *bdev)
1336 {
1337 	struct request_queue *q = bdev_get_queue(bdev);
1338 
1339 	if (bdev != bdev->bd_contains)
1340 		return bdev->bd_part->discard_alignment;
1341 
1342 	return q->limits.discard_alignment;
1343 }
1344 
1345 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1346 {
1347 	if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1348 		return 1;
1349 
1350 	return 0;
1351 }
1352 
1353 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1354 {
1355 	return queue_discard_zeroes_data(bdev_get_queue(bdev));
1356 }
1357 
1358 static inline unsigned int bdev_write_same(struct block_device *bdev)
1359 {
1360 	struct request_queue *q = bdev_get_queue(bdev);
1361 
1362 	if (q)
1363 		return q->limits.max_write_same_sectors;
1364 
1365 	return 0;
1366 }
1367 
1368 static inline int queue_dma_alignment(struct request_queue *q)
1369 {
1370 	return q ? q->dma_alignment : 511;
1371 }
1372 
1373 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1374 				 unsigned int len)
1375 {
1376 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1377 	return !(addr & alignment) && !(len & alignment);
1378 }
1379 
1380 /* assumes size > 256 */
1381 static inline unsigned int blksize_bits(unsigned int size)
1382 {
1383 	unsigned int bits = 8;
1384 	do {
1385 		bits++;
1386 		size >>= 1;
1387 	} while (size > 256);
1388 	return bits;
1389 }
1390 
1391 static inline unsigned int block_size(struct block_device *bdev)
1392 {
1393 	return bdev->bd_block_size;
1394 }
1395 
1396 static inline bool queue_flush_queueable(struct request_queue *q)
1397 {
1398 	return !q->flush_not_queueable;
1399 }
1400 
1401 typedef struct {struct page *v;} Sector;
1402 
1403 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1404 
1405 static inline void put_dev_sector(Sector p)
1406 {
1407 	page_cache_release(p.v);
1408 }
1409 
1410 struct work_struct;
1411 int kblockd_schedule_work(struct work_struct *work);
1412 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1413 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1414 
1415 #ifdef CONFIG_BLK_CGROUP
1416 /*
1417  * This should not be using sched_clock(). A real patch is in progress
1418  * to fix this up, until that is in place we need to disable preemption
1419  * around sched_clock() in this function and set_io_start_time_ns().
1420  */
1421 static inline void set_start_time_ns(struct request *req)
1422 {
1423 	preempt_disable();
1424 	req->start_time_ns = sched_clock();
1425 	preempt_enable();
1426 }
1427 
1428 static inline void set_io_start_time_ns(struct request *req)
1429 {
1430 	preempt_disable();
1431 	req->io_start_time_ns = sched_clock();
1432 	preempt_enable();
1433 }
1434 
1435 static inline uint64_t rq_start_time_ns(struct request *req)
1436 {
1437         return req->start_time_ns;
1438 }
1439 
1440 static inline uint64_t rq_io_start_time_ns(struct request *req)
1441 {
1442         return req->io_start_time_ns;
1443 }
1444 #else
1445 static inline void set_start_time_ns(struct request *req) {}
1446 static inline void set_io_start_time_ns(struct request *req) {}
1447 static inline uint64_t rq_start_time_ns(struct request *req)
1448 {
1449 	return 0;
1450 }
1451 static inline uint64_t rq_io_start_time_ns(struct request *req)
1452 {
1453 	return 0;
1454 }
1455 #endif
1456 
1457 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1458 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1459 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1460 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1461 
1462 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1463 
1464 #define INTEGRITY_FLAG_READ	2	/* verify data integrity on read */
1465 #define INTEGRITY_FLAG_WRITE	4	/* generate data integrity on write */
1466 
1467 struct blk_integrity_exchg {
1468 	void			*prot_buf;
1469 	void			*data_buf;
1470 	sector_t		sector;
1471 	unsigned int		data_size;
1472 	unsigned short		sector_size;
1473 	const char		*disk_name;
1474 };
1475 
1476 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1477 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1478 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1479 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1480 
1481 struct blk_integrity {
1482 	integrity_gen_fn	*generate_fn;
1483 	integrity_vrfy_fn	*verify_fn;
1484 	integrity_set_tag_fn	*set_tag_fn;
1485 	integrity_get_tag_fn	*get_tag_fn;
1486 
1487 	unsigned short		flags;
1488 	unsigned short		tuple_size;
1489 	unsigned short		sector_size;
1490 	unsigned short		tag_size;
1491 
1492 	const char		*name;
1493 
1494 	struct kobject		kobj;
1495 };
1496 
1497 extern bool blk_integrity_is_initialized(struct gendisk *);
1498 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1499 extern void blk_integrity_unregister(struct gendisk *);
1500 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1501 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1502 				   struct scatterlist *);
1503 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1504 extern int blk_integrity_merge_rq(struct request_queue *, struct request *,
1505 				  struct request *);
1506 extern int blk_integrity_merge_bio(struct request_queue *, struct request *,
1507 				   struct bio *);
1508 
1509 static inline
1510 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1511 {
1512 	return bdev->bd_disk->integrity;
1513 }
1514 
1515 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1516 {
1517 	return disk->integrity;
1518 }
1519 
1520 static inline int blk_integrity_rq(struct request *rq)
1521 {
1522 	if (rq->bio == NULL)
1523 		return 0;
1524 
1525 	return bio_integrity(rq->bio);
1526 }
1527 
1528 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1529 						    unsigned int segs)
1530 {
1531 	q->limits.max_integrity_segments = segs;
1532 }
1533 
1534 static inline unsigned short
1535 queue_max_integrity_segments(struct request_queue *q)
1536 {
1537 	return q->limits.max_integrity_segments;
1538 }
1539 
1540 #else /* CONFIG_BLK_DEV_INTEGRITY */
1541 
1542 struct bio;
1543 struct block_device;
1544 struct gendisk;
1545 struct blk_integrity;
1546 
1547 static inline int blk_integrity_rq(struct request *rq)
1548 {
1549 	return 0;
1550 }
1551 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1552 					    struct bio *b)
1553 {
1554 	return 0;
1555 }
1556 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1557 					  struct bio *b,
1558 					  struct scatterlist *s)
1559 {
1560 	return 0;
1561 }
1562 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1563 {
1564 	return 0;
1565 }
1566 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1567 {
1568 	return NULL;
1569 }
1570 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1571 {
1572 	return 0;
1573 }
1574 static inline int blk_integrity_register(struct gendisk *d,
1575 					 struct blk_integrity *b)
1576 {
1577 	return 0;
1578 }
1579 static inline void blk_integrity_unregister(struct gendisk *d)
1580 {
1581 }
1582 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1583 						    unsigned int segs)
1584 {
1585 }
1586 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1587 {
1588 	return 0;
1589 }
1590 static inline int blk_integrity_merge_rq(struct request_queue *rq,
1591 					 struct request *r1,
1592 					 struct request *r2)
1593 {
1594 	return 0;
1595 }
1596 static inline int blk_integrity_merge_bio(struct request_queue *rq,
1597 					  struct request *r,
1598 					  struct bio *b)
1599 {
1600 	return 0;
1601 }
1602 static inline bool blk_integrity_is_initialized(struct gendisk *g)
1603 {
1604 	return 0;
1605 }
1606 
1607 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1608 
1609 struct block_device_operations {
1610 	int (*open) (struct block_device *, fmode_t);
1611 	void (*release) (struct gendisk *, fmode_t);
1612 	int (*rw_page)(struct block_device *, sector_t, struct page *, int rw);
1613 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1614 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1615 	int (*direct_access) (struct block_device *, sector_t,
1616 						void **, unsigned long *);
1617 	unsigned int (*check_events) (struct gendisk *disk,
1618 				      unsigned int clearing);
1619 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
1620 	int (*media_changed) (struct gendisk *);
1621 	void (*unlock_native_capacity) (struct gendisk *);
1622 	int (*revalidate_disk) (struct gendisk *);
1623 	int (*getgeo)(struct block_device *, struct hd_geometry *);
1624 	/* this callback is with swap_lock and sometimes page table lock held */
1625 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1626 	struct module *owner;
1627 };
1628 
1629 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1630 				 unsigned long);
1631 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1632 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1633 						struct writeback_control *);
1634 #else /* CONFIG_BLOCK */
1635 
1636 struct block_device;
1637 
1638 /*
1639  * stubs for when the block layer is configured out
1640  */
1641 #define buffer_heads_over_limit 0
1642 
1643 static inline long nr_blockdev_pages(void)
1644 {
1645 	return 0;
1646 }
1647 
1648 struct blk_plug {
1649 };
1650 
1651 static inline void blk_start_plug(struct blk_plug *plug)
1652 {
1653 }
1654 
1655 static inline void blk_finish_plug(struct blk_plug *plug)
1656 {
1657 }
1658 
1659 static inline void blk_flush_plug(struct task_struct *task)
1660 {
1661 }
1662 
1663 static inline void blk_schedule_flush_plug(struct task_struct *task)
1664 {
1665 }
1666 
1667 
1668 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1669 {
1670 	return false;
1671 }
1672 
1673 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1674 				     sector_t *error_sector)
1675 {
1676 	return 0;
1677 }
1678 
1679 #endif /* CONFIG_BLOCK */
1680 
1681 #endif
1682