xref: /linux-6.15/include/linux/blkdev.h (revision 9376ff9b)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
4 
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
7 
8 #ifdef CONFIG_BLOCK
9 
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
30 #include <linux/seqlock.h>
31 #include <linux/u64_stats_sync.h>
32 
33 struct module;
34 struct scsi_ioctl_command;
35 
36 struct request_queue;
37 struct elevator_queue;
38 struct blk_trace;
39 struct request;
40 struct sg_io_hdr;
41 struct bsg_job;
42 struct blkcg_gq;
43 struct blk_flush_queue;
44 struct pr_ops;
45 struct rq_wb;
46 struct blk_queue_stats;
47 struct blk_stat_callback;
48 
49 #define BLKDEV_MIN_RQ	4
50 #define BLKDEV_MAX_RQ	128	/* Default maximum */
51 
52 /* Must be consistent with blk_mq_poll_stats_bkt() */
53 #define BLK_MQ_POLL_STATS_BKTS 16
54 
55 /*
56  * Maximum number of blkcg policies allowed to be registered concurrently.
57  * Defined here to simplify include dependency.
58  */
59 #define BLKCG_MAX_POLS		3
60 
61 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
62 
63 #define BLK_RL_SYNCFULL		(1U << 0)
64 #define BLK_RL_ASYNCFULL	(1U << 1)
65 
66 struct request_list {
67 	struct request_queue	*q;	/* the queue this rl belongs to */
68 #ifdef CONFIG_BLK_CGROUP
69 	struct blkcg_gq		*blkg;	/* blkg this request pool belongs to */
70 #endif
71 	/*
72 	 * count[], starved[], and wait[] are indexed by
73 	 * BLK_RW_SYNC/BLK_RW_ASYNC
74 	 */
75 	int			count[2];
76 	int			starved[2];
77 	mempool_t		*rq_pool;
78 	wait_queue_head_t	wait[2];
79 	unsigned int		flags;
80 };
81 
82 /*
83  * request flags */
84 typedef __u32 __bitwise req_flags_t;
85 
86 /* elevator knows about this request */
87 #define RQF_SORTED		((__force req_flags_t)(1 << 0))
88 /* drive already may have started this one */
89 #define RQF_STARTED		((__force req_flags_t)(1 << 1))
90 /* uses tagged queueing */
91 #define RQF_QUEUED		((__force req_flags_t)(1 << 2))
92 /* may not be passed by ioscheduler */
93 #define RQF_SOFTBARRIER		((__force req_flags_t)(1 << 3))
94 /* request for flush sequence */
95 #define RQF_FLUSH_SEQ		((__force req_flags_t)(1 << 4))
96 /* merge of different types, fail separately */
97 #define RQF_MIXED_MERGE		((__force req_flags_t)(1 << 5))
98 /* track inflight for MQ */
99 #define RQF_MQ_INFLIGHT		((__force req_flags_t)(1 << 6))
100 /* don't call prep for this one */
101 #define RQF_DONTPREP		((__force req_flags_t)(1 << 7))
102 /* set for "ide_preempt" requests and also for requests for which the SCSI
103    "quiesce" state must be ignored. */
104 #define RQF_PREEMPT		((__force req_flags_t)(1 << 8))
105 /* contains copies of user pages */
106 #define RQF_COPY_USER		((__force req_flags_t)(1 << 9))
107 /* vaguely specified driver internal error.  Ignored by the block layer */
108 #define RQF_FAILED		((__force req_flags_t)(1 << 10))
109 /* don't warn about errors */
110 #define RQF_QUIET		((__force req_flags_t)(1 << 11))
111 /* elevator private data attached */
112 #define RQF_ELVPRIV		((__force req_flags_t)(1 << 12))
113 /* account I/O stat */
114 #define RQF_IO_STAT		((__force req_flags_t)(1 << 13))
115 /* request came from our alloc pool */
116 #define RQF_ALLOCED		((__force req_flags_t)(1 << 14))
117 /* runtime pm request */
118 #define RQF_PM			((__force req_flags_t)(1 << 15))
119 /* on IO scheduler merge hash */
120 #define RQF_HASHED		((__force req_flags_t)(1 << 16))
121 /* IO stats tracking on */
122 #define RQF_STATS		((__force req_flags_t)(1 << 17))
123 /* Look at ->special_vec for the actual data payload instead of the
124    bio chain. */
125 #define RQF_SPECIAL_PAYLOAD	((__force req_flags_t)(1 << 18))
126 /* The per-zone write lock is held for this request */
127 #define RQF_ZONE_WRITE_LOCKED	((__force req_flags_t)(1 << 19))
128 /* timeout is expired */
129 #define RQF_MQ_TIMEOUT_EXPIRED	((__force req_flags_t)(1 << 20))
130 /* already slept for hybrid poll */
131 #define RQF_MQ_POLL_SLEPT	((__force req_flags_t)(1 << 21))
132 
133 /* flags that prevent us from merging requests: */
134 #define RQF_NOMERGE_FLAGS \
135 	(RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
136 
137 /*
138  * Try to put the fields that are referenced together in the same cacheline.
139  *
140  * If you modify this structure, make sure to update blk_rq_init() and
141  * especially blk_mq_rq_ctx_init() to take care of the added fields.
142  */
143 struct request {
144 	struct request_queue *q;
145 	struct blk_mq_ctx *mq_ctx;
146 
147 	int cpu;
148 	unsigned int cmd_flags;		/* op and common flags */
149 	req_flags_t rq_flags;
150 
151 	int internal_tag;
152 
153 	/* the following two fields are internal, NEVER access directly */
154 	unsigned int __data_len;	/* total data len */
155 	int tag;
156 	sector_t __sector;		/* sector cursor */
157 
158 	struct bio *bio;
159 	struct bio *biotail;
160 
161 	struct list_head queuelist;
162 
163 	/*
164 	 * The hash is used inside the scheduler, and killed once the
165 	 * request reaches the dispatch list. The ipi_list is only used
166 	 * to queue the request for softirq completion, which is long
167 	 * after the request has been unhashed (and even removed from
168 	 * the dispatch list).
169 	 */
170 	union {
171 		struct hlist_node hash;	/* merge hash */
172 		struct list_head ipi_list;
173 	};
174 
175 	/*
176 	 * The rb_node is only used inside the io scheduler, requests
177 	 * are pruned when moved to the dispatch queue. So let the
178 	 * completion_data share space with the rb_node.
179 	 */
180 	union {
181 		struct rb_node rb_node;	/* sort/lookup */
182 		struct bio_vec special_vec;
183 		void *completion_data;
184 		int error_count; /* for legacy drivers, don't use */
185 	};
186 
187 	/*
188 	 * Three pointers are available for the IO schedulers, if they need
189 	 * more they have to dynamically allocate it.  Flush requests are
190 	 * never put on the IO scheduler. So let the flush fields share
191 	 * space with the elevator data.
192 	 */
193 	union {
194 		struct {
195 			struct io_cq		*icq;
196 			void			*priv[2];
197 		} elv;
198 
199 		struct {
200 			unsigned int		seq;
201 			struct list_head	list;
202 			rq_end_io_fn		*saved_end_io;
203 		} flush;
204 	};
205 
206 	struct gendisk *rq_disk;
207 	struct hd_struct *part;
208 	unsigned long start_time;
209 	struct blk_issue_stat issue_stat;
210 	/* Number of scatter-gather DMA addr+len pairs after
211 	 * physical address coalescing is performed.
212 	 */
213 	unsigned short nr_phys_segments;
214 
215 #if defined(CONFIG_BLK_DEV_INTEGRITY)
216 	unsigned short nr_integrity_segments;
217 #endif
218 
219 	unsigned short write_hint;
220 	unsigned short ioprio;
221 
222 	unsigned int timeout;
223 
224 	void *special;		/* opaque pointer available for LLD use */
225 
226 	unsigned int extra_len;	/* length of alignment and padding */
227 
228 	/*
229 	 * On blk-mq, the lower bits of ->gstate (generation number and
230 	 * state) carry the MQ_RQ_* state value and the upper bits the
231 	 * generation number which is monotonically incremented and used to
232 	 * distinguish the reuse instances.
233 	 *
234 	 * ->gstate_seq allows updates to ->gstate and other fields
235 	 * (currently ->deadline) during request start to be read
236 	 * atomically from the timeout path, so that it can operate on a
237 	 * coherent set of information.
238 	 */
239 	seqcount_t gstate_seq;
240 	u64 gstate;
241 
242 	/*
243 	 * ->aborted_gstate is used by the timeout to claim a specific
244 	 * recycle instance of this request.  See blk_mq_timeout_work().
245 	 */
246 	struct u64_stats_sync aborted_gstate_sync;
247 	u64 aborted_gstate;
248 
249 	/* access through blk_rq_set_deadline, blk_rq_deadline */
250 	unsigned long __deadline;
251 
252 	struct list_head timeout_list;
253 
254 	union {
255 		struct __call_single_data csd;
256 		u64 fifo_time;
257 	};
258 
259 	/*
260 	 * completion callback.
261 	 */
262 	rq_end_io_fn *end_io;
263 	void *end_io_data;
264 
265 	/* for bidi */
266 	struct request *next_rq;
267 
268 #ifdef CONFIG_BLK_CGROUP
269 	struct request_list *rl;		/* rl this rq is alloced from */
270 	unsigned long long start_time_ns;
271 	unsigned long long io_start_time_ns;    /* when passed to hardware */
272 #endif
273 };
274 
275 static inline bool blk_op_is_scsi(unsigned int op)
276 {
277 	return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
278 }
279 
280 static inline bool blk_op_is_private(unsigned int op)
281 {
282 	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
283 }
284 
285 static inline bool blk_rq_is_scsi(struct request *rq)
286 {
287 	return blk_op_is_scsi(req_op(rq));
288 }
289 
290 static inline bool blk_rq_is_private(struct request *rq)
291 {
292 	return blk_op_is_private(req_op(rq));
293 }
294 
295 static inline bool blk_rq_is_passthrough(struct request *rq)
296 {
297 	return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
298 }
299 
300 static inline bool bio_is_passthrough(struct bio *bio)
301 {
302 	unsigned op = bio_op(bio);
303 
304 	return blk_op_is_scsi(op) || blk_op_is_private(op);
305 }
306 
307 static inline unsigned short req_get_ioprio(struct request *req)
308 {
309 	return req->ioprio;
310 }
311 
312 #include <linux/elevator.h>
313 
314 struct blk_queue_ctx;
315 
316 typedef void (request_fn_proc) (struct request_queue *q);
317 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
318 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
319 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
320 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
321 
322 struct bio_vec;
323 typedef void (softirq_done_fn)(struct request *);
324 typedef int (dma_drain_needed_fn)(struct request *);
325 typedef int (lld_busy_fn) (struct request_queue *q);
326 typedef int (bsg_job_fn) (struct bsg_job *);
327 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
328 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
329 
330 enum blk_eh_timer_return {
331 	BLK_EH_NOT_HANDLED,
332 	BLK_EH_HANDLED,
333 	BLK_EH_RESET_TIMER,
334 };
335 
336 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
337 
338 enum blk_queue_state {
339 	Queue_down,
340 	Queue_up,
341 };
342 
343 struct blk_queue_tag {
344 	struct request **tag_index;	/* map of busy tags */
345 	unsigned long *tag_map;		/* bit map of free/busy tags */
346 	int max_depth;			/* what we will send to device */
347 	int real_max_depth;		/* what the array can hold */
348 	atomic_t refcnt;		/* map can be shared */
349 	int alloc_policy;		/* tag allocation policy */
350 	int next_tag;			/* next tag */
351 };
352 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
353 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
354 
355 #define BLK_SCSI_MAX_CMDS	(256)
356 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
357 
358 /*
359  * Zoned block device models (zoned limit).
360  */
361 enum blk_zoned_model {
362 	BLK_ZONED_NONE,	/* Regular block device */
363 	BLK_ZONED_HA,	/* Host-aware zoned block device */
364 	BLK_ZONED_HM,	/* Host-managed zoned block device */
365 };
366 
367 struct queue_limits {
368 	unsigned long		bounce_pfn;
369 	unsigned long		seg_boundary_mask;
370 	unsigned long		virt_boundary_mask;
371 
372 	unsigned int		max_hw_sectors;
373 	unsigned int		max_dev_sectors;
374 	unsigned int		chunk_sectors;
375 	unsigned int		max_sectors;
376 	unsigned int		max_segment_size;
377 	unsigned int		physical_block_size;
378 	unsigned int		alignment_offset;
379 	unsigned int		io_min;
380 	unsigned int		io_opt;
381 	unsigned int		max_discard_sectors;
382 	unsigned int		max_hw_discard_sectors;
383 	unsigned int		max_write_same_sectors;
384 	unsigned int		max_write_zeroes_sectors;
385 	unsigned int		discard_granularity;
386 	unsigned int		discard_alignment;
387 
388 	unsigned short		logical_block_size;
389 	unsigned short		max_segments;
390 	unsigned short		max_integrity_segments;
391 	unsigned short		max_discard_segments;
392 
393 	unsigned char		misaligned;
394 	unsigned char		discard_misaligned;
395 	unsigned char		cluster;
396 	unsigned char		raid_partial_stripes_expensive;
397 	enum blk_zoned_model	zoned;
398 };
399 
400 #ifdef CONFIG_BLK_DEV_ZONED
401 
402 struct blk_zone_report_hdr {
403 	unsigned int	nr_zones;
404 	u8		padding[60];
405 };
406 
407 extern int blkdev_report_zones(struct block_device *bdev,
408 			       sector_t sector, struct blk_zone *zones,
409 			       unsigned int *nr_zones, gfp_t gfp_mask);
410 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
411 			      sector_t nr_sectors, gfp_t gfp_mask);
412 
413 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
414 				     unsigned int cmd, unsigned long arg);
415 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
416 				    unsigned int cmd, unsigned long arg);
417 
418 #else /* CONFIG_BLK_DEV_ZONED */
419 
420 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
421 					    fmode_t mode, unsigned int cmd,
422 					    unsigned long arg)
423 {
424 	return -ENOTTY;
425 }
426 
427 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
428 					   fmode_t mode, unsigned int cmd,
429 					   unsigned long arg)
430 {
431 	return -ENOTTY;
432 }
433 
434 #endif /* CONFIG_BLK_DEV_ZONED */
435 
436 struct request_queue {
437 	/*
438 	 * Together with queue_head for cacheline sharing
439 	 */
440 	struct list_head	queue_head;
441 	struct request		*last_merge;
442 	struct elevator_queue	*elevator;
443 	int			nr_rqs[2];	/* # allocated [a]sync rqs */
444 	int			nr_rqs_elvpriv;	/* # allocated rqs w/ elvpriv */
445 
446 	atomic_t		shared_hctx_restart;
447 
448 	struct blk_queue_stats	*stats;
449 	struct rq_wb		*rq_wb;
450 
451 	/*
452 	 * If blkcg is not used, @q->root_rl serves all requests.  If blkcg
453 	 * is used, root blkg allocates from @q->root_rl and all other
454 	 * blkgs from their own blkg->rl.  Which one to use should be
455 	 * determined using bio_request_list().
456 	 */
457 	struct request_list	root_rl;
458 
459 	request_fn_proc		*request_fn;
460 	make_request_fn		*make_request_fn;
461 	poll_q_fn		*poll_fn;
462 	prep_rq_fn		*prep_rq_fn;
463 	unprep_rq_fn		*unprep_rq_fn;
464 	softirq_done_fn		*softirq_done_fn;
465 	rq_timed_out_fn		*rq_timed_out_fn;
466 	dma_drain_needed_fn	*dma_drain_needed;
467 	lld_busy_fn		*lld_busy_fn;
468 	/* Called just after a request is allocated */
469 	init_rq_fn		*init_rq_fn;
470 	/* Called just before a request is freed */
471 	exit_rq_fn		*exit_rq_fn;
472 	/* Called from inside blk_get_request() */
473 	void (*initialize_rq_fn)(struct request *rq);
474 
475 	const struct blk_mq_ops	*mq_ops;
476 
477 	unsigned int		*mq_map;
478 
479 	/* sw queues */
480 	struct blk_mq_ctx __percpu	*queue_ctx;
481 	unsigned int		nr_queues;
482 
483 	unsigned int		queue_depth;
484 
485 	/* hw dispatch queues */
486 	struct blk_mq_hw_ctx	**queue_hw_ctx;
487 	unsigned int		nr_hw_queues;
488 
489 	/*
490 	 * Dispatch queue sorting
491 	 */
492 	sector_t		end_sector;
493 	struct request		*boundary_rq;
494 
495 	/*
496 	 * Delayed queue handling
497 	 */
498 	struct delayed_work	delay_work;
499 
500 	struct backing_dev_info	*backing_dev_info;
501 
502 	/*
503 	 * The queue owner gets to use this for whatever they like.
504 	 * ll_rw_blk doesn't touch it.
505 	 */
506 	void			*queuedata;
507 
508 	/*
509 	 * various queue flags, see QUEUE_* below
510 	 */
511 	unsigned long		queue_flags;
512 
513 	/*
514 	 * ida allocated id for this queue.  Used to index queues from
515 	 * ioctx.
516 	 */
517 	int			id;
518 
519 	/*
520 	 * queue needs bounce pages for pages above this limit
521 	 */
522 	gfp_t			bounce_gfp;
523 
524 	/*
525 	 * protects queue structures from reentrancy. ->__queue_lock should
526 	 * _never_ be used directly, it is queue private. always use
527 	 * ->queue_lock.
528 	 */
529 	spinlock_t		__queue_lock;
530 	spinlock_t		*queue_lock;
531 
532 	/*
533 	 * queue kobject
534 	 */
535 	struct kobject kobj;
536 
537 	/*
538 	 * mq queue kobject
539 	 */
540 	struct kobject mq_kobj;
541 
542 #ifdef  CONFIG_BLK_DEV_INTEGRITY
543 	struct blk_integrity integrity;
544 #endif	/* CONFIG_BLK_DEV_INTEGRITY */
545 
546 #ifdef CONFIG_PM
547 	struct device		*dev;
548 	int			rpm_status;
549 	unsigned int		nr_pending;
550 #endif
551 
552 	/*
553 	 * queue settings
554 	 */
555 	unsigned long		nr_requests;	/* Max # of requests */
556 	unsigned int		nr_congestion_on;
557 	unsigned int		nr_congestion_off;
558 	unsigned int		nr_batching;
559 
560 	unsigned int		dma_drain_size;
561 	void			*dma_drain_buffer;
562 	unsigned int		dma_pad_mask;
563 	unsigned int		dma_alignment;
564 
565 	struct blk_queue_tag	*queue_tags;
566 	struct list_head	tag_busy_list;
567 
568 	unsigned int		nr_sorted;
569 	unsigned int		in_flight[2];
570 
571 	/*
572 	 * Number of active block driver functions for which blk_drain_queue()
573 	 * must wait. Must be incremented around functions that unlock the
574 	 * queue_lock internally, e.g. scsi_request_fn().
575 	 */
576 	unsigned int		request_fn_active;
577 
578 	unsigned int		rq_timeout;
579 	int			poll_nsec;
580 
581 	struct blk_stat_callback	*poll_cb;
582 	struct blk_rq_stat	poll_stat[BLK_MQ_POLL_STATS_BKTS];
583 
584 	struct timer_list	timeout;
585 	struct work_struct	timeout_work;
586 	struct list_head	timeout_list;
587 
588 	struct list_head	icq_list;
589 #ifdef CONFIG_BLK_CGROUP
590 	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
591 	struct blkcg_gq		*root_blkg;
592 	struct list_head	blkg_list;
593 #endif
594 
595 	struct queue_limits	limits;
596 
597 	/*
598 	 * Zoned block device information for request dispatch control.
599 	 * nr_zones is the total number of zones of the device. This is always
600 	 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
601 	 * bits which indicates if a zone is conventional (bit clear) or
602 	 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
603 	 * bits which indicates if a zone is write locked, that is, if a write
604 	 * request targeting the zone was dispatched. All three fields are
605 	 * initialized by the low level device driver (e.g. scsi/sd.c).
606 	 * Stacking drivers (device mappers) may or may not initialize
607 	 * these fields.
608 	 */
609 	unsigned int		nr_zones;
610 	unsigned long		*seq_zones_bitmap;
611 	unsigned long		*seq_zones_wlock;
612 
613 	/*
614 	 * sg stuff
615 	 */
616 	unsigned int		sg_timeout;
617 	unsigned int		sg_reserved_size;
618 	int			node;
619 #ifdef CONFIG_BLK_DEV_IO_TRACE
620 	struct blk_trace	*blk_trace;
621 	struct mutex		blk_trace_mutex;
622 #endif
623 	/*
624 	 * for flush operations
625 	 */
626 	struct blk_flush_queue	*fq;
627 
628 	struct list_head	requeue_list;
629 	spinlock_t		requeue_lock;
630 	struct delayed_work	requeue_work;
631 
632 	struct mutex		sysfs_lock;
633 
634 	int			bypass_depth;
635 	atomic_t		mq_freeze_depth;
636 
637 #if defined(CONFIG_BLK_DEV_BSG)
638 	bsg_job_fn		*bsg_job_fn;
639 	struct bsg_class_device bsg_dev;
640 #endif
641 
642 #ifdef CONFIG_BLK_DEV_THROTTLING
643 	/* Throttle data */
644 	struct throtl_data *td;
645 #endif
646 	struct rcu_head		rcu_head;
647 	wait_queue_head_t	mq_freeze_wq;
648 	struct percpu_ref	q_usage_counter;
649 	struct list_head	all_q_node;
650 
651 	struct blk_mq_tag_set	*tag_set;
652 	struct list_head	tag_set_list;
653 	struct bio_set		*bio_split;
654 
655 #ifdef CONFIG_BLK_DEBUG_FS
656 	struct dentry		*debugfs_dir;
657 	struct dentry		*sched_debugfs_dir;
658 #endif
659 
660 	bool			mq_sysfs_init_done;
661 
662 	size_t			cmd_size;
663 	void			*rq_alloc_data;
664 
665 	struct work_struct	release_work;
666 
667 #define BLK_MAX_WRITE_HINTS	5
668 	u64			write_hints[BLK_MAX_WRITE_HINTS];
669 };
670 
671 #define QUEUE_FLAG_QUEUED	0	/* uses generic tag queueing */
672 #define QUEUE_FLAG_STOPPED	1	/* queue is stopped */
673 #define QUEUE_FLAG_DYING	2	/* queue being torn down */
674 #define QUEUE_FLAG_BYPASS	3	/* act as dumb FIFO queue */
675 #define QUEUE_FLAG_BIDI		4	/* queue supports bidi requests */
676 #define QUEUE_FLAG_NOMERGES     5	/* disable merge attempts */
677 #define QUEUE_FLAG_SAME_COMP	6	/* complete on same CPU-group */
678 #define QUEUE_FLAG_FAIL_IO	7	/* fake timeout */
679 #define QUEUE_FLAG_NONROT	9	/* non-rotational device (SSD) */
680 #define QUEUE_FLAG_VIRT        QUEUE_FLAG_NONROT /* paravirt device */
681 #define QUEUE_FLAG_IO_STAT     10	/* do IO stats */
682 #define QUEUE_FLAG_DISCARD     11	/* supports DISCARD */
683 #define QUEUE_FLAG_NOXMERGES   12	/* No extended merges */
684 #define QUEUE_FLAG_ADD_RANDOM  13	/* Contributes to random pool */
685 #define QUEUE_FLAG_SECERASE    14	/* supports secure erase */
686 #define QUEUE_FLAG_SAME_FORCE  15	/* force complete on same CPU */
687 #define QUEUE_FLAG_DEAD        16	/* queue tear-down finished */
688 #define QUEUE_FLAG_INIT_DONE   17	/* queue is initialized */
689 #define QUEUE_FLAG_NO_SG_MERGE 18	/* don't attempt to merge SG segments*/
690 #define QUEUE_FLAG_POLL	       19	/* IO polling enabled if set */
691 #define QUEUE_FLAG_WC	       20	/* Write back caching */
692 #define QUEUE_FLAG_FUA	       21	/* device supports FUA writes */
693 #define QUEUE_FLAG_FLUSH_NQ    22	/* flush not queueuable */
694 #define QUEUE_FLAG_DAX         23	/* device supports DAX */
695 #define QUEUE_FLAG_STATS       24	/* track rq completion times */
696 #define QUEUE_FLAG_POLL_STATS  25	/* collecting stats for hybrid polling */
697 #define QUEUE_FLAG_REGISTERED  26	/* queue has been registered to a disk */
698 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27	/* queue supports SCSI commands */
699 #define QUEUE_FLAG_QUIESCED    28	/* queue has been quiesced */
700 #define QUEUE_FLAG_PREEMPT_ONLY	29	/* only process REQ_PREEMPT requests */
701 
702 #define QUEUE_FLAG_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
703 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
704 				 (1 << QUEUE_FLAG_ADD_RANDOM))
705 
706 #define QUEUE_FLAG_MQ_DEFAULT	((1 << QUEUE_FLAG_IO_STAT) |		\
707 				 (1 << QUEUE_FLAG_SAME_COMP)	|	\
708 				 (1 << QUEUE_FLAG_POLL))
709 
710 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
711 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
712 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
713 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
714 
715 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
716 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
717 #define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
718 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
719 #define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
720 #define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
721 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
722 #define blk_queue_noxmerges(q)	\
723 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
724 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
725 #define blk_queue_io_stat(q)	test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
726 #define blk_queue_add_random(q)	test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
727 #define blk_queue_discard(q)	test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
728 #define blk_queue_secure_erase(q) \
729 	(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
730 #define blk_queue_dax(q)	test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
731 #define blk_queue_scsi_passthrough(q)	\
732 	test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
733 
734 #define blk_noretry_request(rq) \
735 	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
736 			     REQ_FAILFAST_DRIVER))
737 #define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
738 #define blk_queue_preempt_only(q)				\
739 	test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
740 
741 extern int blk_set_preempt_only(struct request_queue *q);
742 extern void blk_clear_preempt_only(struct request_queue *q);
743 
744 static inline int queue_in_flight(struct request_queue *q)
745 {
746 	return q->in_flight[0] + q->in_flight[1];
747 }
748 
749 static inline bool blk_account_rq(struct request *rq)
750 {
751 	return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
752 }
753 
754 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
755 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
756 /* rq->queuelist of dequeued request must be list_empty() */
757 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
758 
759 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
760 
761 #define rq_data_dir(rq)		(op_is_write(req_op(rq)) ? WRITE : READ)
762 
763 /*
764  * Driver can handle struct request, if it either has an old style
765  * request_fn defined, or is blk-mq based.
766  */
767 static inline bool queue_is_rq_based(struct request_queue *q)
768 {
769 	return q->request_fn || q->mq_ops;
770 }
771 
772 static inline unsigned int blk_queue_cluster(struct request_queue *q)
773 {
774 	return q->limits.cluster;
775 }
776 
777 static inline enum blk_zoned_model
778 blk_queue_zoned_model(struct request_queue *q)
779 {
780 	return q->limits.zoned;
781 }
782 
783 static inline bool blk_queue_is_zoned(struct request_queue *q)
784 {
785 	switch (blk_queue_zoned_model(q)) {
786 	case BLK_ZONED_HA:
787 	case BLK_ZONED_HM:
788 		return true;
789 	default:
790 		return false;
791 	}
792 }
793 
794 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
795 {
796 	return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
797 }
798 
799 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
800 {
801 	return q->nr_zones;
802 }
803 
804 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
805 					     sector_t sector)
806 {
807 	if (!blk_queue_is_zoned(q))
808 		return 0;
809 	return sector >> ilog2(q->limits.chunk_sectors);
810 }
811 
812 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
813 					 sector_t sector)
814 {
815 	if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
816 		return false;
817 	return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
818 }
819 
820 static inline bool rq_is_sync(struct request *rq)
821 {
822 	return op_is_sync(rq->cmd_flags);
823 }
824 
825 static inline bool blk_rl_full(struct request_list *rl, bool sync)
826 {
827 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
828 
829 	return rl->flags & flag;
830 }
831 
832 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
833 {
834 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
835 
836 	rl->flags |= flag;
837 }
838 
839 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
840 {
841 	unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
842 
843 	rl->flags &= ~flag;
844 }
845 
846 static inline bool rq_mergeable(struct request *rq)
847 {
848 	if (blk_rq_is_passthrough(rq))
849 		return false;
850 
851 	if (req_op(rq) == REQ_OP_FLUSH)
852 		return false;
853 
854 	if (req_op(rq) == REQ_OP_WRITE_ZEROES)
855 		return false;
856 
857 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
858 		return false;
859 	if (rq->rq_flags & RQF_NOMERGE_FLAGS)
860 		return false;
861 
862 	return true;
863 }
864 
865 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
866 {
867 	if (bio_page(a) == bio_page(b) &&
868 	    bio_offset(a) == bio_offset(b))
869 		return true;
870 
871 	return false;
872 }
873 
874 static inline unsigned int blk_queue_depth(struct request_queue *q)
875 {
876 	if (q->queue_depth)
877 		return q->queue_depth;
878 
879 	return q->nr_requests;
880 }
881 
882 /*
883  * q->prep_rq_fn return values
884  */
885 enum {
886 	BLKPREP_OK,		/* serve it */
887 	BLKPREP_KILL,		/* fatal error, kill, return -EIO */
888 	BLKPREP_DEFER,		/* leave on queue */
889 	BLKPREP_INVALID,	/* invalid command, kill, return -EREMOTEIO */
890 };
891 
892 extern unsigned long blk_max_low_pfn, blk_max_pfn;
893 
894 /*
895  * standard bounce addresses:
896  *
897  * BLK_BOUNCE_HIGH	: bounce all highmem pages
898  * BLK_BOUNCE_ANY	: don't bounce anything
899  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
900  */
901 
902 #if BITS_PER_LONG == 32
903 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
904 #else
905 #define BLK_BOUNCE_HIGH		-1ULL
906 #endif
907 #define BLK_BOUNCE_ANY		(-1ULL)
908 #define BLK_BOUNCE_ISA		(DMA_BIT_MASK(24))
909 
910 /*
911  * default timeout for SG_IO if none specified
912  */
913 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
914 #define BLK_MIN_SG_TIMEOUT	(7 * HZ)
915 
916 struct rq_map_data {
917 	struct page **pages;
918 	int page_order;
919 	int nr_entries;
920 	unsigned long offset;
921 	int null_mapped;
922 	int from_user;
923 };
924 
925 struct req_iterator {
926 	struct bvec_iter iter;
927 	struct bio *bio;
928 };
929 
930 /* This should not be used directly - use rq_for_each_segment */
931 #define for_each_bio(_bio)		\
932 	for (; _bio; _bio = _bio->bi_next)
933 #define __rq_for_each_bio(_bio, rq)	\
934 	if ((rq->bio))			\
935 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
936 
937 #define rq_for_each_segment(bvl, _rq, _iter)			\
938 	__rq_for_each_bio(_iter.bio, _rq)			\
939 		bio_for_each_segment(bvl, _iter.bio, _iter.iter)
940 
941 #define rq_iter_last(bvec, _iter)				\
942 		(_iter.bio->bi_next == NULL &&			\
943 		 bio_iter_last(bvec, _iter.iter))
944 
945 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
946 # error	"You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
947 #endif
948 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
949 extern void rq_flush_dcache_pages(struct request *rq);
950 #else
951 static inline void rq_flush_dcache_pages(struct request *rq)
952 {
953 }
954 #endif
955 
956 extern int blk_register_queue(struct gendisk *disk);
957 extern void blk_unregister_queue(struct gendisk *disk);
958 extern blk_qc_t generic_make_request(struct bio *bio);
959 extern blk_qc_t direct_make_request(struct bio *bio);
960 extern void blk_rq_init(struct request_queue *q, struct request *rq);
961 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
962 extern void blk_put_request(struct request *);
963 extern void __blk_put_request(struct request_queue *, struct request *);
964 extern struct request *blk_get_request_flags(struct request_queue *,
965 					     unsigned int op,
966 					     blk_mq_req_flags_t flags);
967 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
968 				       gfp_t gfp_mask);
969 extern void blk_requeue_request(struct request_queue *, struct request *);
970 extern int blk_lld_busy(struct request_queue *q);
971 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
972 			     struct bio_set *bs, gfp_t gfp_mask,
973 			     int (*bio_ctr)(struct bio *, struct bio *, void *),
974 			     void *data);
975 extern void blk_rq_unprep_clone(struct request *rq);
976 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
977 				     struct request *rq);
978 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
979 extern void blk_delay_queue(struct request_queue *, unsigned long);
980 extern void blk_queue_split(struct request_queue *, struct bio **);
981 extern void blk_recount_segments(struct request_queue *, struct bio *);
982 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
983 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
984 			      unsigned int, void __user *);
985 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
986 			  unsigned int, void __user *);
987 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
988 			 struct scsi_ioctl_command __user *);
989 
990 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
991 extern void blk_queue_exit(struct request_queue *q);
992 extern void blk_start_queue(struct request_queue *q);
993 extern void blk_start_queue_async(struct request_queue *q);
994 extern void blk_stop_queue(struct request_queue *q);
995 extern void blk_sync_queue(struct request_queue *q);
996 extern void __blk_stop_queue(struct request_queue *q);
997 extern void __blk_run_queue(struct request_queue *q);
998 extern void __blk_run_queue_uncond(struct request_queue *q);
999 extern void blk_run_queue(struct request_queue *);
1000 extern void blk_run_queue_async(struct request_queue *q);
1001 extern int blk_rq_map_user(struct request_queue *, struct request *,
1002 			   struct rq_map_data *, void __user *, unsigned long,
1003 			   gfp_t);
1004 extern int blk_rq_unmap_user(struct bio *);
1005 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1006 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1007 			       struct rq_map_data *, const struct iov_iter *,
1008 			       gfp_t);
1009 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1010 			  struct request *, int);
1011 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1012 				  struct request *, int, rq_end_io_fn *);
1013 
1014 int blk_status_to_errno(blk_status_t status);
1015 blk_status_t errno_to_blk_status(int errno);
1016 
1017 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1018 
1019 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1020 {
1021 	return bdev->bd_disk->queue;	/* this is never NULL */
1022 }
1023 
1024 /*
1025  * The basic unit of block I/O is a sector. It is used in a number of contexts
1026  * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1027  * bytes. Variables of type sector_t represent an offset or size that is a
1028  * multiple of 512 bytes. Hence these two constants.
1029  */
1030 #ifndef SECTOR_SHIFT
1031 #define SECTOR_SHIFT 9
1032 #endif
1033 #ifndef SECTOR_SIZE
1034 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1035 #endif
1036 
1037 /*
1038  * blk_rq_pos()			: the current sector
1039  * blk_rq_bytes()		: bytes left in the entire request
1040  * blk_rq_cur_bytes()		: bytes left in the current segment
1041  * blk_rq_err_bytes()		: bytes left till the next error boundary
1042  * blk_rq_sectors()		: sectors left in the entire request
1043  * blk_rq_cur_sectors()		: sectors left in the current segment
1044  */
1045 static inline sector_t blk_rq_pos(const struct request *rq)
1046 {
1047 	return rq->__sector;
1048 }
1049 
1050 static inline unsigned int blk_rq_bytes(const struct request *rq)
1051 {
1052 	return rq->__data_len;
1053 }
1054 
1055 static inline int blk_rq_cur_bytes(const struct request *rq)
1056 {
1057 	return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1058 }
1059 
1060 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1061 
1062 static inline unsigned int blk_rq_sectors(const struct request *rq)
1063 {
1064 	return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1065 }
1066 
1067 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1068 {
1069 	return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1070 }
1071 
1072 static inline unsigned int blk_rq_zone_no(struct request *rq)
1073 {
1074 	return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1075 }
1076 
1077 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1078 {
1079 	return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1080 }
1081 
1082 /*
1083  * Some commands like WRITE SAME have a payload or data transfer size which
1084  * is different from the size of the request.  Any driver that supports such
1085  * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1086  * calculate the data transfer size.
1087  */
1088 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1089 {
1090 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1091 		return rq->special_vec.bv_len;
1092 	return blk_rq_bytes(rq);
1093 }
1094 
1095 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1096 						     int op)
1097 {
1098 	if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1099 		return min(q->limits.max_discard_sectors,
1100 			   UINT_MAX >> SECTOR_SHIFT);
1101 
1102 	if (unlikely(op == REQ_OP_WRITE_SAME))
1103 		return q->limits.max_write_same_sectors;
1104 
1105 	if (unlikely(op == REQ_OP_WRITE_ZEROES))
1106 		return q->limits.max_write_zeroes_sectors;
1107 
1108 	return q->limits.max_sectors;
1109 }
1110 
1111 /*
1112  * Return maximum size of a request at given offset. Only valid for
1113  * file system requests.
1114  */
1115 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1116 					       sector_t offset)
1117 {
1118 	if (!q->limits.chunk_sectors)
1119 		return q->limits.max_sectors;
1120 
1121 	return q->limits.chunk_sectors -
1122 			(offset & (q->limits.chunk_sectors - 1));
1123 }
1124 
1125 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1126 						  sector_t offset)
1127 {
1128 	struct request_queue *q = rq->q;
1129 
1130 	if (blk_rq_is_passthrough(rq))
1131 		return q->limits.max_hw_sectors;
1132 
1133 	if (!q->limits.chunk_sectors ||
1134 	    req_op(rq) == REQ_OP_DISCARD ||
1135 	    req_op(rq) == REQ_OP_SECURE_ERASE)
1136 		return blk_queue_get_max_sectors(q, req_op(rq));
1137 
1138 	return min(blk_max_size_offset(q, offset),
1139 			blk_queue_get_max_sectors(q, req_op(rq)));
1140 }
1141 
1142 static inline unsigned int blk_rq_count_bios(struct request *rq)
1143 {
1144 	unsigned int nr_bios = 0;
1145 	struct bio *bio;
1146 
1147 	__rq_for_each_bio(bio, rq)
1148 		nr_bios++;
1149 
1150 	return nr_bios;
1151 }
1152 
1153 /*
1154  * Request issue related functions.
1155  */
1156 extern struct request *blk_peek_request(struct request_queue *q);
1157 extern void blk_start_request(struct request *rq);
1158 extern struct request *blk_fetch_request(struct request_queue *q);
1159 
1160 void blk_steal_bios(struct bio_list *list, struct request *rq);
1161 
1162 /*
1163  * Request completion related functions.
1164  *
1165  * blk_update_request() completes given number of bytes and updates
1166  * the request without completing it.
1167  *
1168  * blk_end_request() and friends.  __blk_end_request() must be called
1169  * with the request queue spinlock acquired.
1170  *
1171  * Several drivers define their own end_request and call
1172  * blk_end_request() for parts of the original function.
1173  * This prevents code duplication in drivers.
1174  */
1175 extern bool blk_update_request(struct request *rq, blk_status_t error,
1176 			       unsigned int nr_bytes);
1177 extern void blk_finish_request(struct request *rq, blk_status_t error);
1178 extern bool blk_end_request(struct request *rq, blk_status_t error,
1179 			    unsigned int nr_bytes);
1180 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1181 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1182 			      unsigned int nr_bytes);
1183 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1184 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1185 
1186 extern void blk_complete_request(struct request *);
1187 extern void __blk_complete_request(struct request *);
1188 extern void blk_abort_request(struct request *);
1189 extern void blk_unprep_request(struct request *);
1190 
1191 /*
1192  * Access functions for manipulating queue properties
1193  */
1194 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1195 					spinlock_t *lock, int node_id);
1196 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1197 extern int blk_init_allocated_queue(struct request_queue *);
1198 extern void blk_cleanup_queue(struct request_queue *);
1199 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1200 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1201 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1202 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1203 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1204 extern void blk_queue_max_discard_segments(struct request_queue *,
1205 		unsigned short);
1206 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1207 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1208 		unsigned int max_discard_sectors);
1209 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1210 		unsigned int max_write_same_sectors);
1211 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1212 		unsigned int max_write_same_sectors);
1213 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1214 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1215 extern void blk_queue_alignment_offset(struct request_queue *q,
1216 				       unsigned int alignment);
1217 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1218 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1219 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1220 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1221 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1222 extern void blk_set_default_limits(struct queue_limits *lim);
1223 extern void blk_set_stacking_limits(struct queue_limits *lim);
1224 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1225 			    sector_t offset);
1226 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1227 			    sector_t offset);
1228 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1229 			      sector_t offset);
1230 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1231 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1232 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1233 extern int blk_queue_dma_drain(struct request_queue *q,
1234 			       dma_drain_needed_fn *dma_drain_needed,
1235 			       void *buf, unsigned int size);
1236 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1237 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1238 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1239 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1240 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1241 extern void blk_queue_dma_alignment(struct request_queue *, int);
1242 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1243 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1244 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1245 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1246 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1247 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1248 
1249 /*
1250  * Number of physical segments as sent to the device.
1251  *
1252  * Normally this is the number of discontiguous data segments sent by the
1253  * submitter.  But for data-less command like discard we might have no
1254  * actual data segments submitted, but the driver might have to add it's
1255  * own special payload.  In that case we still return 1 here so that this
1256  * special payload will be mapped.
1257  */
1258 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1259 {
1260 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1261 		return 1;
1262 	return rq->nr_phys_segments;
1263 }
1264 
1265 /*
1266  * Number of discard segments (or ranges) the driver needs to fill in.
1267  * Each discard bio merged into a request is counted as one segment.
1268  */
1269 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1270 {
1271 	return max_t(unsigned short, rq->nr_phys_segments, 1);
1272 }
1273 
1274 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1275 extern void blk_dump_rq_flags(struct request *, char *);
1276 extern long nr_blockdev_pages(void);
1277 
1278 bool __must_check blk_get_queue(struct request_queue *);
1279 struct request_queue *blk_alloc_queue(gfp_t);
1280 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1281 					   spinlock_t *lock);
1282 extern void blk_put_queue(struct request_queue *);
1283 extern void blk_set_queue_dying(struct request_queue *);
1284 
1285 /*
1286  * block layer runtime pm functions
1287  */
1288 #ifdef CONFIG_PM
1289 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1290 extern int blk_pre_runtime_suspend(struct request_queue *q);
1291 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1292 extern void blk_pre_runtime_resume(struct request_queue *q);
1293 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1294 extern void blk_set_runtime_active(struct request_queue *q);
1295 #else
1296 static inline void blk_pm_runtime_init(struct request_queue *q,
1297 	struct device *dev) {}
1298 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1299 {
1300 	return -ENOSYS;
1301 }
1302 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1303 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1304 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1305 static inline void blk_set_runtime_active(struct request_queue *q) {}
1306 #endif
1307 
1308 /*
1309  * blk_plug permits building a queue of related requests by holding the I/O
1310  * fragments for a short period. This allows merging of sequential requests
1311  * into single larger request. As the requests are moved from a per-task list to
1312  * the device's request_queue in a batch, this results in improved scalability
1313  * as the lock contention for request_queue lock is reduced.
1314  *
1315  * It is ok not to disable preemption when adding the request to the plug list
1316  * or when attempting a merge, because blk_schedule_flush_list() will only flush
1317  * the plug list when the task sleeps by itself. For details, please see
1318  * schedule() where blk_schedule_flush_plug() is called.
1319  */
1320 struct blk_plug {
1321 	struct list_head list; /* requests */
1322 	struct list_head mq_list; /* blk-mq requests */
1323 	struct list_head cb_list; /* md requires an unplug callback */
1324 };
1325 #define BLK_MAX_REQUEST_COUNT 16
1326 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1327 
1328 struct blk_plug_cb;
1329 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1330 struct blk_plug_cb {
1331 	struct list_head list;
1332 	blk_plug_cb_fn callback;
1333 	void *data;
1334 };
1335 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1336 					     void *data, int size);
1337 extern void blk_start_plug(struct blk_plug *);
1338 extern void blk_finish_plug(struct blk_plug *);
1339 extern void blk_flush_plug_list(struct blk_plug *, bool);
1340 
1341 static inline void blk_flush_plug(struct task_struct *tsk)
1342 {
1343 	struct blk_plug *plug = tsk->plug;
1344 
1345 	if (plug)
1346 		blk_flush_plug_list(plug, false);
1347 }
1348 
1349 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1350 {
1351 	struct blk_plug *plug = tsk->plug;
1352 
1353 	if (plug)
1354 		blk_flush_plug_list(plug, true);
1355 }
1356 
1357 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1358 {
1359 	struct blk_plug *plug = tsk->plug;
1360 
1361 	return plug &&
1362 		(!list_empty(&plug->list) ||
1363 		 !list_empty(&plug->mq_list) ||
1364 		 !list_empty(&plug->cb_list));
1365 }
1366 
1367 /*
1368  * tag stuff
1369  */
1370 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1371 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1372 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1373 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1374 extern void blk_queue_free_tags(struct request_queue *);
1375 extern int blk_queue_resize_tags(struct request_queue *, int);
1376 extern void blk_queue_invalidate_tags(struct request_queue *);
1377 extern struct blk_queue_tag *blk_init_tags(int, int);
1378 extern void blk_free_tags(struct blk_queue_tag *);
1379 
1380 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1381 						int tag)
1382 {
1383 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1384 		return NULL;
1385 	return bqt->tag_index[tag];
1386 }
1387 
1388 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1389 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1390 		sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1391 
1392 #define BLKDEV_DISCARD_SECURE	(1 << 0)	/* issue a secure erase */
1393 
1394 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1395 		sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1396 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1397 		sector_t nr_sects, gfp_t gfp_mask, int flags,
1398 		struct bio **biop);
1399 
1400 #define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1401 #define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1402 
1403 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1404 		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1405 		unsigned flags);
1406 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1407 		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1408 
1409 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1410 		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1411 {
1412 	return blkdev_issue_discard(sb->s_bdev,
1413 				    block << (sb->s_blocksize_bits -
1414 					      SECTOR_SHIFT),
1415 				    nr_blocks << (sb->s_blocksize_bits -
1416 						  SECTOR_SHIFT),
1417 				    gfp_mask, flags);
1418 }
1419 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1420 		sector_t nr_blocks, gfp_t gfp_mask)
1421 {
1422 	return blkdev_issue_zeroout(sb->s_bdev,
1423 				    block << (sb->s_blocksize_bits -
1424 					      SECTOR_SHIFT),
1425 				    nr_blocks << (sb->s_blocksize_bits -
1426 						  SECTOR_SHIFT),
1427 				    gfp_mask, 0);
1428 }
1429 
1430 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1431 
1432 enum blk_default_limits {
1433 	BLK_MAX_SEGMENTS	= 128,
1434 	BLK_SAFE_MAX_SECTORS	= 255,
1435 	BLK_DEF_MAX_SECTORS	= 2560,
1436 	BLK_MAX_SEGMENT_SIZE	= 65536,
1437 	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1438 };
1439 
1440 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1441 
1442 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1443 {
1444 	return q->limits.seg_boundary_mask;
1445 }
1446 
1447 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1448 {
1449 	return q->limits.virt_boundary_mask;
1450 }
1451 
1452 static inline unsigned int queue_max_sectors(struct request_queue *q)
1453 {
1454 	return q->limits.max_sectors;
1455 }
1456 
1457 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1458 {
1459 	return q->limits.max_hw_sectors;
1460 }
1461 
1462 static inline unsigned short queue_max_segments(struct request_queue *q)
1463 {
1464 	return q->limits.max_segments;
1465 }
1466 
1467 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1468 {
1469 	return q->limits.max_discard_segments;
1470 }
1471 
1472 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1473 {
1474 	return q->limits.max_segment_size;
1475 }
1476 
1477 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1478 {
1479 	int retval = 512;
1480 
1481 	if (q && q->limits.logical_block_size)
1482 		retval = q->limits.logical_block_size;
1483 
1484 	return retval;
1485 }
1486 
1487 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1488 {
1489 	return queue_logical_block_size(bdev_get_queue(bdev));
1490 }
1491 
1492 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1493 {
1494 	return q->limits.physical_block_size;
1495 }
1496 
1497 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1498 {
1499 	return queue_physical_block_size(bdev_get_queue(bdev));
1500 }
1501 
1502 static inline unsigned int queue_io_min(struct request_queue *q)
1503 {
1504 	return q->limits.io_min;
1505 }
1506 
1507 static inline int bdev_io_min(struct block_device *bdev)
1508 {
1509 	return queue_io_min(bdev_get_queue(bdev));
1510 }
1511 
1512 static inline unsigned int queue_io_opt(struct request_queue *q)
1513 {
1514 	return q->limits.io_opt;
1515 }
1516 
1517 static inline int bdev_io_opt(struct block_device *bdev)
1518 {
1519 	return queue_io_opt(bdev_get_queue(bdev));
1520 }
1521 
1522 static inline int queue_alignment_offset(struct request_queue *q)
1523 {
1524 	if (q->limits.misaligned)
1525 		return -1;
1526 
1527 	return q->limits.alignment_offset;
1528 }
1529 
1530 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1531 {
1532 	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1533 	unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1534 		<< SECTOR_SHIFT;
1535 
1536 	return (granularity + lim->alignment_offset - alignment) % granularity;
1537 }
1538 
1539 static inline int bdev_alignment_offset(struct block_device *bdev)
1540 {
1541 	struct request_queue *q = bdev_get_queue(bdev);
1542 
1543 	if (q->limits.misaligned)
1544 		return -1;
1545 
1546 	if (bdev != bdev->bd_contains)
1547 		return bdev->bd_part->alignment_offset;
1548 
1549 	return q->limits.alignment_offset;
1550 }
1551 
1552 static inline int queue_discard_alignment(struct request_queue *q)
1553 {
1554 	if (q->limits.discard_misaligned)
1555 		return -1;
1556 
1557 	return q->limits.discard_alignment;
1558 }
1559 
1560 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1561 {
1562 	unsigned int alignment, granularity, offset;
1563 
1564 	if (!lim->max_discard_sectors)
1565 		return 0;
1566 
1567 	/* Why are these in bytes, not sectors? */
1568 	alignment = lim->discard_alignment >> SECTOR_SHIFT;
1569 	granularity = lim->discard_granularity >> SECTOR_SHIFT;
1570 	if (!granularity)
1571 		return 0;
1572 
1573 	/* Offset of the partition start in 'granularity' sectors */
1574 	offset = sector_div(sector, granularity);
1575 
1576 	/* And why do we do this modulus *again* in blkdev_issue_discard()? */
1577 	offset = (granularity + alignment - offset) % granularity;
1578 
1579 	/* Turn it back into bytes, gaah */
1580 	return offset << SECTOR_SHIFT;
1581 }
1582 
1583 static inline int bdev_discard_alignment(struct block_device *bdev)
1584 {
1585 	struct request_queue *q = bdev_get_queue(bdev);
1586 
1587 	if (bdev != bdev->bd_contains)
1588 		return bdev->bd_part->discard_alignment;
1589 
1590 	return q->limits.discard_alignment;
1591 }
1592 
1593 static inline unsigned int bdev_write_same(struct block_device *bdev)
1594 {
1595 	struct request_queue *q = bdev_get_queue(bdev);
1596 
1597 	if (q)
1598 		return q->limits.max_write_same_sectors;
1599 
1600 	return 0;
1601 }
1602 
1603 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1604 {
1605 	struct request_queue *q = bdev_get_queue(bdev);
1606 
1607 	if (q)
1608 		return q->limits.max_write_zeroes_sectors;
1609 
1610 	return 0;
1611 }
1612 
1613 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1614 {
1615 	struct request_queue *q = bdev_get_queue(bdev);
1616 
1617 	if (q)
1618 		return blk_queue_zoned_model(q);
1619 
1620 	return BLK_ZONED_NONE;
1621 }
1622 
1623 static inline bool bdev_is_zoned(struct block_device *bdev)
1624 {
1625 	struct request_queue *q = bdev_get_queue(bdev);
1626 
1627 	if (q)
1628 		return blk_queue_is_zoned(q);
1629 
1630 	return false;
1631 }
1632 
1633 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1634 {
1635 	struct request_queue *q = bdev_get_queue(bdev);
1636 
1637 	if (q)
1638 		return blk_queue_zone_sectors(q);
1639 	return 0;
1640 }
1641 
1642 static inline unsigned int bdev_nr_zones(struct block_device *bdev)
1643 {
1644 	struct request_queue *q = bdev_get_queue(bdev);
1645 
1646 	if (q)
1647 		return blk_queue_nr_zones(q);
1648 	return 0;
1649 }
1650 
1651 static inline int queue_dma_alignment(struct request_queue *q)
1652 {
1653 	return q ? q->dma_alignment : 511;
1654 }
1655 
1656 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1657 				 unsigned int len)
1658 {
1659 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1660 	return !(addr & alignment) && !(len & alignment);
1661 }
1662 
1663 /* assumes size > 256 */
1664 static inline unsigned int blksize_bits(unsigned int size)
1665 {
1666 	unsigned int bits = 8;
1667 	do {
1668 		bits++;
1669 		size >>= 1;
1670 	} while (size > 256);
1671 	return bits;
1672 }
1673 
1674 static inline unsigned int block_size(struct block_device *bdev)
1675 {
1676 	return bdev->bd_block_size;
1677 }
1678 
1679 static inline bool queue_flush_queueable(struct request_queue *q)
1680 {
1681 	return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1682 }
1683 
1684 typedef struct {struct page *v;} Sector;
1685 
1686 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1687 
1688 static inline void put_dev_sector(Sector p)
1689 {
1690 	put_page(p.v);
1691 }
1692 
1693 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1694 				struct bio_vec *bprv, unsigned int offset)
1695 {
1696 	return offset ||
1697 		((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1698 }
1699 
1700 /*
1701  * Check if adding a bio_vec after bprv with offset would create a gap in
1702  * the SG list. Most drivers don't care about this, but some do.
1703  */
1704 static inline bool bvec_gap_to_prev(struct request_queue *q,
1705 				struct bio_vec *bprv, unsigned int offset)
1706 {
1707 	if (!queue_virt_boundary(q))
1708 		return false;
1709 	return __bvec_gap_to_prev(q, bprv, offset);
1710 }
1711 
1712 /*
1713  * Check if the two bvecs from two bios can be merged to one segment.
1714  * If yes, no need to check gap between the two bios since the 1st bio
1715  * and the 1st bvec in the 2nd bio can be handled in one segment.
1716  */
1717 static inline bool bios_segs_mergeable(struct request_queue *q,
1718 		struct bio *prev, struct bio_vec *prev_last_bv,
1719 		struct bio_vec *next_first_bv)
1720 {
1721 	if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1722 		return false;
1723 	if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1724 		return false;
1725 	if (prev->bi_seg_back_size + next_first_bv->bv_len >
1726 			queue_max_segment_size(q))
1727 		return false;
1728 	return true;
1729 }
1730 
1731 static inline bool bio_will_gap(struct request_queue *q,
1732 				struct request *prev_rq,
1733 				struct bio *prev,
1734 				struct bio *next)
1735 {
1736 	if (bio_has_data(prev) && queue_virt_boundary(q)) {
1737 		struct bio_vec pb, nb;
1738 
1739 		/*
1740 		 * don't merge if the 1st bio starts with non-zero
1741 		 * offset, otherwise it is quite difficult to respect
1742 		 * sg gap limit. We work hard to merge a huge number of small
1743 		 * single bios in case of mkfs.
1744 		 */
1745 		if (prev_rq)
1746 			bio_get_first_bvec(prev_rq->bio, &pb);
1747 		else
1748 			bio_get_first_bvec(prev, &pb);
1749 		if (pb.bv_offset)
1750 			return true;
1751 
1752 		/*
1753 		 * We don't need to worry about the situation that the
1754 		 * merged segment ends in unaligned virt boundary:
1755 		 *
1756 		 * - if 'pb' ends aligned, the merged segment ends aligned
1757 		 * - if 'pb' ends unaligned, the next bio must include
1758 		 *   one single bvec of 'nb', otherwise the 'nb' can't
1759 		 *   merge with 'pb'
1760 		 */
1761 		bio_get_last_bvec(prev, &pb);
1762 		bio_get_first_bvec(next, &nb);
1763 
1764 		if (!bios_segs_mergeable(q, prev, &pb, &nb))
1765 			return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1766 	}
1767 
1768 	return false;
1769 }
1770 
1771 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1772 {
1773 	return bio_will_gap(req->q, req, req->biotail, bio);
1774 }
1775 
1776 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1777 {
1778 	return bio_will_gap(req->q, NULL, bio, req->bio);
1779 }
1780 
1781 int kblockd_schedule_work(struct work_struct *work);
1782 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1783 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1784 
1785 #ifdef CONFIG_BLK_CGROUP
1786 /*
1787  * This should not be using sched_clock(). A real patch is in progress
1788  * to fix this up, until that is in place we need to disable preemption
1789  * around sched_clock() in this function and set_io_start_time_ns().
1790  */
1791 static inline void set_start_time_ns(struct request *req)
1792 {
1793 	preempt_disable();
1794 	req->start_time_ns = sched_clock();
1795 	preempt_enable();
1796 }
1797 
1798 static inline void set_io_start_time_ns(struct request *req)
1799 {
1800 	preempt_disable();
1801 	req->io_start_time_ns = sched_clock();
1802 	preempt_enable();
1803 }
1804 
1805 static inline uint64_t rq_start_time_ns(struct request *req)
1806 {
1807         return req->start_time_ns;
1808 }
1809 
1810 static inline uint64_t rq_io_start_time_ns(struct request *req)
1811 {
1812         return req->io_start_time_ns;
1813 }
1814 #else
1815 static inline void set_start_time_ns(struct request *req) {}
1816 static inline void set_io_start_time_ns(struct request *req) {}
1817 static inline uint64_t rq_start_time_ns(struct request *req)
1818 {
1819 	return 0;
1820 }
1821 static inline uint64_t rq_io_start_time_ns(struct request *req)
1822 {
1823 	return 0;
1824 }
1825 #endif
1826 
1827 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1828 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1829 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1830 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1831 
1832 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1833 
1834 enum blk_integrity_flags {
1835 	BLK_INTEGRITY_VERIFY		= 1 << 0,
1836 	BLK_INTEGRITY_GENERATE		= 1 << 1,
1837 	BLK_INTEGRITY_DEVICE_CAPABLE	= 1 << 2,
1838 	BLK_INTEGRITY_IP_CHECKSUM	= 1 << 3,
1839 };
1840 
1841 struct blk_integrity_iter {
1842 	void			*prot_buf;
1843 	void			*data_buf;
1844 	sector_t		seed;
1845 	unsigned int		data_size;
1846 	unsigned short		interval;
1847 	const char		*disk_name;
1848 };
1849 
1850 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1851 
1852 struct blk_integrity_profile {
1853 	integrity_processing_fn		*generate_fn;
1854 	integrity_processing_fn		*verify_fn;
1855 	const char			*name;
1856 };
1857 
1858 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1859 extern void blk_integrity_unregister(struct gendisk *);
1860 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1861 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1862 				   struct scatterlist *);
1863 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1864 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1865 				   struct request *);
1866 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1867 				    struct bio *);
1868 
1869 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1870 {
1871 	struct blk_integrity *bi = &disk->queue->integrity;
1872 
1873 	if (!bi->profile)
1874 		return NULL;
1875 
1876 	return bi;
1877 }
1878 
1879 static inline
1880 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1881 {
1882 	return blk_get_integrity(bdev->bd_disk);
1883 }
1884 
1885 static inline bool blk_integrity_rq(struct request *rq)
1886 {
1887 	return rq->cmd_flags & REQ_INTEGRITY;
1888 }
1889 
1890 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1891 						    unsigned int segs)
1892 {
1893 	q->limits.max_integrity_segments = segs;
1894 }
1895 
1896 static inline unsigned short
1897 queue_max_integrity_segments(struct request_queue *q)
1898 {
1899 	return q->limits.max_integrity_segments;
1900 }
1901 
1902 static inline bool integrity_req_gap_back_merge(struct request *req,
1903 						struct bio *next)
1904 {
1905 	struct bio_integrity_payload *bip = bio_integrity(req->bio);
1906 	struct bio_integrity_payload *bip_next = bio_integrity(next);
1907 
1908 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1909 				bip_next->bip_vec[0].bv_offset);
1910 }
1911 
1912 static inline bool integrity_req_gap_front_merge(struct request *req,
1913 						 struct bio *bio)
1914 {
1915 	struct bio_integrity_payload *bip = bio_integrity(bio);
1916 	struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1917 
1918 	return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1919 				bip_next->bip_vec[0].bv_offset);
1920 }
1921 
1922 #else /* CONFIG_BLK_DEV_INTEGRITY */
1923 
1924 struct bio;
1925 struct block_device;
1926 struct gendisk;
1927 struct blk_integrity;
1928 
1929 static inline int blk_integrity_rq(struct request *rq)
1930 {
1931 	return 0;
1932 }
1933 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1934 					    struct bio *b)
1935 {
1936 	return 0;
1937 }
1938 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1939 					  struct bio *b,
1940 					  struct scatterlist *s)
1941 {
1942 	return 0;
1943 }
1944 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1945 {
1946 	return NULL;
1947 }
1948 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1949 {
1950 	return NULL;
1951 }
1952 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1953 {
1954 	return 0;
1955 }
1956 static inline void blk_integrity_register(struct gendisk *d,
1957 					 struct blk_integrity *b)
1958 {
1959 }
1960 static inline void blk_integrity_unregister(struct gendisk *d)
1961 {
1962 }
1963 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1964 						    unsigned int segs)
1965 {
1966 }
1967 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1968 {
1969 	return 0;
1970 }
1971 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1972 					  struct request *r1,
1973 					  struct request *r2)
1974 {
1975 	return true;
1976 }
1977 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1978 					   struct request *r,
1979 					   struct bio *b)
1980 {
1981 	return true;
1982 }
1983 
1984 static inline bool integrity_req_gap_back_merge(struct request *req,
1985 						struct bio *next)
1986 {
1987 	return false;
1988 }
1989 static inline bool integrity_req_gap_front_merge(struct request *req,
1990 						 struct bio *bio)
1991 {
1992 	return false;
1993 }
1994 
1995 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1996 
1997 struct block_device_operations {
1998 	int (*open) (struct block_device *, fmode_t);
1999 	void (*release) (struct gendisk *, fmode_t);
2000 	int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
2001 	int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
2002 	int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
2003 	unsigned int (*check_events) (struct gendisk *disk,
2004 				      unsigned int clearing);
2005 	/* ->media_changed() is DEPRECATED, use ->check_events() instead */
2006 	int (*media_changed) (struct gendisk *);
2007 	void (*unlock_native_capacity) (struct gendisk *);
2008 	int (*revalidate_disk) (struct gendisk *);
2009 	int (*getgeo)(struct block_device *, struct hd_geometry *);
2010 	/* this callback is with swap_lock and sometimes page table lock held */
2011 	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
2012 	struct module *owner;
2013 	const struct pr_ops *pr_ops;
2014 };
2015 
2016 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
2017 				 unsigned long);
2018 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
2019 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
2020 						struct writeback_control *);
2021 
2022 #ifdef CONFIG_BLK_DEV_ZONED
2023 bool blk_req_needs_zone_write_lock(struct request *rq);
2024 void __blk_req_zone_write_lock(struct request *rq);
2025 void __blk_req_zone_write_unlock(struct request *rq);
2026 
2027 static inline void blk_req_zone_write_lock(struct request *rq)
2028 {
2029 	if (blk_req_needs_zone_write_lock(rq))
2030 		__blk_req_zone_write_lock(rq);
2031 }
2032 
2033 static inline void blk_req_zone_write_unlock(struct request *rq)
2034 {
2035 	if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
2036 		__blk_req_zone_write_unlock(rq);
2037 }
2038 
2039 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2040 {
2041 	return rq->q->seq_zones_wlock &&
2042 		test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
2043 }
2044 
2045 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2046 {
2047 	if (!blk_req_needs_zone_write_lock(rq))
2048 		return true;
2049 	return !blk_req_zone_is_write_locked(rq);
2050 }
2051 #else
2052 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2053 {
2054 	return false;
2055 }
2056 
2057 static inline void blk_req_zone_write_lock(struct request *rq)
2058 {
2059 }
2060 
2061 static inline void blk_req_zone_write_unlock(struct request *rq)
2062 {
2063 }
2064 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2065 {
2066 	return false;
2067 }
2068 
2069 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2070 {
2071 	return true;
2072 }
2073 #endif /* CONFIG_BLK_DEV_ZONED */
2074 
2075 #else /* CONFIG_BLOCK */
2076 
2077 struct block_device;
2078 
2079 /*
2080  * stubs for when the block layer is configured out
2081  */
2082 #define buffer_heads_over_limit 0
2083 
2084 static inline long nr_blockdev_pages(void)
2085 {
2086 	return 0;
2087 }
2088 
2089 struct blk_plug {
2090 };
2091 
2092 static inline void blk_start_plug(struct blk_plug *plug)
2093 {
2094 }
2095 
2096 static inline void blk_finish_plug(struct blk_plug *plug)
2097 {
2098 }
2099 
2100 static inline void blk_flush_plug(struct task_struct *task)
2101 {
2102 }
2103 
2104 static inline void blk_schedule_flush_plug(struct task_struct *task)
2105 {
2106 }
2107 
2108 
2109 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2110 {
2111 	return false;
2112 }
2113 
2114 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2115 				     sector_t *error_sector)
2116 {
2117 	return 0;
2118 }
2119 
2120 #endif /* CONFIG_BLOCK */
2121 
2122 #endif
2123