1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
4 #include <linux/sched.h>
5 #include <linux/sched/clock.h>
9 #include <linux/major.h>
10 #include <linux/genhd.h>
11 #include <linux/list.h>
12 #include <linux/llist.h>
13 #include <linux/timer.h>
14 #include <linux/workqueue.h>
15 #include <linux/pagemap.h>
16 #include <linux/backing-dev-defs.h>
17 #include <linux/wait.h>
18 #include <linux/mempool.h>
19 #include <linux/pfn.h>
20 #include <linux/bio.h>
21 #include <linux/stringify.h>
22 #include <linux/gfp.h>
23 #include <linux/bsg.h>
24 #include <linux/smp.h>
25 #include <linux/rcupdate.h>
26 #include <linux/percpu-refcount.h>
27 #include <linux/scatterlist.h>
28 #include <linux/blkzoned.h>
31 struct scsi_ioctl_command;
34 struct elevator_queue;
40 struct blk_flush_queue;
43 struct blk_queue_stats;
44 struct blk_stat_callback;
46 #define BLKDEV_MIN_RQ 4
47 #define BLKDEV_MAX_RQ 128 /* Default maximum */
49 /* Must be consisitent with blk_mq_poll_stats_bkt() */
50 #define BLK_MQ_POLL_STATS_BKTS 16
53 * Maximum number of blkcg policies allowed to be registered concurrently.
54 * Defined here to simplify include dependency.
56 #define BLKCG_MAX_POLS 3
58 typedef void (rq_end_io_fn)(struct request *, int);
60 #define BLK_RL_SYNCFULL (1U << 0)
61 #define BLK_RL_ASYNCFULL (1U << 1)
64 struct request_queue *q; /* the queue this rl belongs to */
65 #ifdef CONFIG_BLK_CGROUP
66 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
69 * count[], starved[], and wait[] are indexed by
70 * BLK_RW_SYNC/BLK_RW_ASYNC
75 wait_queue_head_t wait[2];
81 typedef __u32 __bitwise req_flags_t;
83 /* elevator knows about this request */
84 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
85 /* drive already may have started this one */
86 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
87 /* uses tagged queueing */
88 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
89 /* may not be passed by ioscheduler */
90 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
91 /* request for flush sequence */
92 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
93 /* merge of different types, fail separately */
94 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
95 /* track inflight for MQ */
96 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
97 /* don't call prep for this one */
98 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
99 /* set for "ide_preempt" requests and also for requests for which the SCSI
100 "quiesce" state must be ignored. */
101 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
102 /* contains copies of user pages */
103 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
104 /* vaguely specified driver internal error. Ignored by the block layer */
105 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
106 /* don't warn about errors */
107 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
108 /* elevator private data attached */
109 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
110 /* account I/O stat */
111 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
112 /* request came from our alloc pool */
113 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
114 /* runtime pm request */
115 #define RQF_PM ((__force req_flags_t)(1 << 15))
116 /* on IO scheduler merge hash */
117 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
118 /* IO stats tracking on */
119 #define RQF_STATS ((__force req_flags_t)(1 << 17))
120 /* Look at ->special_vec for the actual data payload instead of the
122 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
124 /* flags that prevent us from merging requests: */
125 #define RQF_NOMERGE_FLAGS \
126 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
129 * Try to put the fields that are referenced together in the same cacheline.
131 * If you modify this structure, make sure to update blk_rq_init() and
132 * especially blk_mq_rq_ctx_init() to take care of the added fields.
135 struct list_head queuelist;
137 struct call_single_data csd;
141 struct request_queue *q;
142 struct blk_mq_ctx *mq_ctx;
145 unsigned int cmd_flags; /* op and common flags */
146 req_flags_t rq_flags;
150 unsigned long atomic_flags;
152 /* the following two fields are internal, NEVER access directly */
153 unsigned int __data_len; /* total data len */
155 sector_t __sector; /* sector cursor */
161 * The hash is used inside the scheduler, and killed once the
162 * request reaches the dispatch list. The ipi_list is only used
163 * to queue the request for softirq completion, which is long
164 * after the request has been unhashed (and even removed from
165 * the dispatch list).
168 struct hlist_node hash; /* merge hash */
169 struct list_head ipi_list;
173 * The rb_node is only used inside the io scheduler, requests
174 * are pruned when moved to the dispatch queue. So let the
175 * completion_data share space with the rb_node.
178 struct rb_node rb_node; /* sort/lookup */
179 struct bio_vec special_vec;
180 void *completion_data;
181 int error_count; /* for legacy drivers, don't use */
185 * Three pointers are available for the IO schedulers, if they need
186 * more they have to dynamically allocate it. Flush requests are
187 * never put on the IO scheduler. So let the flush fields share
188 * space with the elevator data.
198 struct list_head list;
199 rq_end_io_fn *saved_end_io;
203 struct gendisk *rq_disk;
204 struct hd_struct *part;
205 unsigned long start_time;
206 struct blk_issue_stat issue_stat;
207 #ifdef CONFIG_BLK_CGROUP
208 struct request_list *rl; /* rl this rq is alloced from */
209 unsigned long long start_time_ns;
210 unsigned long long io_start_time_ns; /* when passed to hardware */
212 /* Number of scatter-gather DMA addr+len pairs after
213 * physical address coalescing is performed.
215 unsigned short nr_phys_segments;
216 #if defined(CONFIG_BLK_DEV_INTEGRITY)
217 unsigned short nr_integrity_segments;
220 unsigned short ioprio;
222 unsigned int timeout;
224 void *special; /* opaque pointer available for LLD use */
226 unsigned int extra_len; /* length of alignment and padding */
228 unsigned long deadline;
229 struct list_head timeout_list;
232 * completion callback.
234 rq_end_io_fn *end_io;
238 struct request *next_rq;
241 static inline bool blk_rq_is_scsi(struct request *rq)
243 return req_op(rq) == REQ_OP_SCSI_IN || req_op(rq) == REQ_OP_SCSI_OUT;
246 static inline bool blk_rq_is_private(struct request *rq)
248 return req_op(rq) == REQ_OP_DRV_IN || req_op(rq) == REQ_OP_DRV_OUT;
251 static inline bool blk_rq_is_passthrough(struct request *rq)
253 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
256 static inline unsigned short req_get_ioprio(struct request *req)
261 #include <linux/elevator.h>
263 struct blk_queue_ctx;
265 typedef void (request_fn_proc) (struct request_queue *q);
266 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
267 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
268 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
271 typedef void (softirq_done_fn)(struct request *);
272 typedef int (dma_drain_needed_fn)(struct request *);
273 typedef int (lld_busy_fn) (struct request_queue *q);
274 typedef int (bsg_job_fn) (struct bsg_job *);
275 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
276 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
278 enum blk_eh_timer_return {
284 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
286 enum blk_queue_state {
291 struct blk_queue_tag {
292 struct request **tag_index; /* map of busy tags */
293 unsigned long *tag_map; /* bit map of free/busy tags */
294 int max_depth; /* what we will send to device */
295 int real_max_depth; /* what the array can hold */
296 atomic_t refcnt; /* map can be shared */
297 int alloc_policy; /* tag allocation policy */
298 int next_tag; /* next tag */
300 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
301 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
303 #define BLK_SCSI_MAX_CMDS (256)
304 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
307 * Zoned block device models (zoned limit).
309 enum blk_zoned_model {
310 BLK_ZONED_NONE, /* Regular block device */
311 BLK_ZONED_HA, /* Host-aware zoned block device */
312 BLK_ZONED_HM, /* Host-managed zoned block device */
315 struct queue_limits {
316 unsigned long bounce_pfn;
317 unsigned long seg_boundary_mask;
318 unsigned long virt_boundary_mask;
320 unsigned int max_hw_sectors;
321 unsigned int max_dev_sectors;
322 unsigned int chunk_sectors;
323 unsigned int max_sectors;
324 unsigned int max_segment_size;
325 unsigned int physical_block_size;
326 unsigned int alignment_offset;
329 unsigned int max_discard_sectors;
330 unsigned int max_hw_discard_sectors;
331 unsigned int max_write_same_sectors;
332 unsigned int max_write_zeroes_sectors;
333 unsigned int discard_granularity;
334 unsigned int discard_alignment;
336 unsigned short logical_block_size;
337 unsigned short max_segments;
338 unsigned short max_integrity_segments;
339 unsigned short max_discard_segments;
341 unsigned char misaligned;
342 unsigned char discard_misaligned;
343 unsigned char cluster;
344 unsigned char raid_partial_stripes_expensive;
345 enum blk_zoned_model zoned;
348 #ifdef CONFIG_BLK_DEV_ZONED
350 struct blk_zone_report_hdr {
351 unsigned int nr_zones;
355 extern int blkdev_report_zones(struct block_device *bdev,
356 sector_t sector, struct blk_zone *zones,
357 unsigned int *nr_zones, gfp_t gfp_mask);
358 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
359 sector_t nr_sectors, gfp_t gfp_mask);
361 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
362 unsigned int cmd, unsigned long arg);
363 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
364 unsigned int cmd, unsigned long arg);
366 #else /* CONFIG_BLK_DEV_ZONED */
368 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
369 fmode_t mode, unsigned int cmd,
375 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
376 fmode_t mode, unsigned int cmd,
382 #endif /* CONFIG_BLK_DEV_ZONED */
384 struct request_queue {
386 * Together with queue_head for cacheline sharing
388 struct list_head queue_head;
389 struct request *last_merge;
390 struct elevator_queue *elevator;
391 int nr_rqs[2]; /* # allocated [a]sync rqs */
392 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
394 struct blk_queue_stats *stats;
398 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
399 * is used, root blkg allocates from @q->root_rl and all other
400 * blkgs from their own blkg->rl. Which one to use should be
401 * determined using bio_request_list().
403 struct request_list root_rl;
405 request_fn_proc *request_fn;
406 make_request_fn *make_request_fn;
407 prep_rq_fn *prep_rq_fn;
408 unprep_rq_fn *unprep_rq_fn;
409 softirq_done_fn *softirq_done_fn;
410 rq_timed_out_fn *rq_timed_out_fn;
411 dma_drain_needed_fn *dma_drain_needed;
412 lld_busy_fn *lld_busy_fn;
413 init_rq_fn *init_rq_fn;
414 exit_rq_fn *exit_rq_fn;
416 const struct blk_mq_ops *mq_ops;
418 unsigned int *mq_map;
421 struct blk_mq_ctx __percpu *queue_ctx;
422 unsigned int nr_queues;
424 unsigned int queue_depth;
426 /* hw dispatch queues */
427 struct blk_mq_hw_ctx **queue_hw_ctx;
428 unsigned int nr_hw_queues;
431 * Dispatch queue sorting
434 struct request *boundary_rq;
437 * Delayed queue handling
439 struct delayed_work delay_work;
441 struct backing_dev_info *backing_dev_info;
444 * The queue owner gets to use this for whatever they like.
445 * ll_rw_blk doesn't touch it.
450 * various queue flags, see QUEUE_* below
452 unsigned long queue_flags;
455 * ida allocated id for this queue. Used to index queues from
461 * queue needs bounce pages for pages above this limit
466 * protects queue structures from reentrancy. ->__queue_lock should
467 * _never_ be used directly, it is queue private. always use
470 spinlock_t __queue_lock;
471 spinlock_t *queue_lock;
481 struct kobject mq_kobj;
483 #ifdef CONFIG_BLK_DEV_INTEGRITY
484 struct blk_integrity integrity;
485 #endif /* CONFIG_BLK_DEV_INTEGRITY */
490 unsigned int nr_pending;
496 unsigned long nr_requests; /* Max # of requests */
497 unsigned int nr_congestion_on;
498 unsigned int nr_congestion_off;
499 unsigned int nr_batching;
501 unsigned int dma_drain_size;
502 void *dma_drain_buffer;
503 unsigned int dma_pad_mask;
504 unsigned int dma_alignment;
506 struct blk_queue_tag *queue_tags;
507 struct list_head tag_busy_list;
509 unsigned int nr_sorted;
510 unsigned int in_flight[2];
513 * Number of active block driver functions for which blk_drain_queue()
514 * must wait. Must be incremented around functions that unlock the
515 * queue_lock internally, e.g. scsi_request_fn().
517 unsigned int request_fn_active;
519 unsigned int rq_timeout;
522 struct blk_stat_callback *poll_cb;
523 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
525 struct timer_list timeout;
526 struct work_struct timeout_work;
527 struct list_head timeout_list;
529 struct list_head icq_list;
530 #ifdef CONFIG_BLK_CGROUP
531 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
532 struct blkcg_gq *root_blkg;
533 struct list_head blkg_list;
536 struct queue_limits limits;
541 unsigned int sg_timeout;
542 unsigned int sg_reserved_size;
544 #ifdef CONFIG_BLK_DEV_IO_TRACE
545 struct blk_trace *blk_trace;
548 * for flush operations
550 struct blk_flush_queue *fq;
552 struct list_head requeue_list;
553 spinlock_t requeue_lock;
554 struct delayed_work requeue_work;
556 struct mutex sysfs_lock;
559 atomic_t mq_freeze_depth;
561 #if defined(CONFIG_BLK_DEV_BSG)
562 bsg_job_fn *bsg_job_fn;
564 struct bsg_class_device bsg_dev;
567 #ifdef CONFIG_BLK_DEV_THROTTLING
569 struct throtl_data *td;
571 struct rcu_head rcu_head;
572 wait_queue_head_t mq_freeze_wq;
573 struct percpu_ref q_usage_counter;
574 struct list_head all_q_node;
576 struct blk_mq_tag_set *tag_set;
577 struct list_head tag_set_list;
578 struct bio_set *bio_split;
580 #ifdef CONFIG_BLK_DEBUG_FS
581 struct dentry *debugfs_dir;
582 struct dentry *sched_debugfs_dir;
585 bool mq_sysfs_init_done;
591 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
592 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
593 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
594 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
595 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
596 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
597 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
598 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
599 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
600 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
601 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
602 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
603 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
604 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
605 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
606 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
607 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
608 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
609 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
610 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
611 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
612 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
613 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
614 #define QUEUE_FLAG_WC 23 /* Write back caching */
615 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
616 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
617 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
618 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
619 #define QUEUE_FLAG_POLL_STATS 28 /* collecting stats for hybrid polling */
620 #define QUEUE_FLAG_REGISTERED 29 /* queue has been registered to a disk */
622 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
623 (1 << QUEUE_FLAG_STACKABLE) | \
624 (1 << QUEUE_FLAG_SAME_COMP) | \
625 (1 << QUEUE_FLAG_ADD_RANDOM))
627 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
628 (1 << QUEUE_FLAG_STACKABLE) | \
629 (1 << QUEUE_FLAG_SAME_COMP) | \
630 (1 << QUEUE_FLAG_POLL))
632 static inline void queue_lockdep_assert_held(struct request_queue *q)
635 lockdep_assert_held(q->queue_lock);
638 static inline void queue_flag_set_unlocked(unsigned int flag,
639 struct request_queue *q)
641 __set_bit(flag, &q->queue_flags);
644 static inline int queue_flag_test_and_clear(unsigned int flag,
645 struct request_queue *q)
647 queue_lockdep_assert_held(q);
649 if (test_bit(flag, &q->queue_flags)) {
650 __clear_bit(flag, &q->queue_flags);
657 static inline int queue_flag_test_and_set(unsigned int flag,
658 struct request_queue *q)
660 queue_lockdep_assert_held(q);
662 if (!test_bit(flag, &q->queue_flags)) {
663 __set_bit(flag, &q->queue_flags);
670 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
672 queue_lockdep_assert_held(q);
673 __set_bit(flag, &q->queue_flags);
676 static inline void queue_flag_clear_unlocked(unsigned int flag,
677 struct request_queue *q)
679 __clear_bit(flag, &q->queue_flags);
682 static inline int queue_in_flight(struct request_queue *q)
684 return q->in_flight[0] + q->in_flight[1];
687 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
689 queue_lockdep_assert_held(q);
690 __clear_bit(flag, &q->queue_flags);
693 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
694 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
695 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
696 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
697 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
698 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
699 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
700 #define blk_queue_noxmerges(q) \
701 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
702 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
703 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
704 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
705 #define blk_queue_stackable(q) \
706 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
707 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
708 #define blk_queue_secure_erase(q) \
709 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
710 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
712 #define blk_noretry_request(rq) \
713 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
714 REQ_FAILFAST_DRIVER))
716 static inline bool blk_account_rq(struct request *rq)
718 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
721 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
722 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
723 /* rq->queuelist of dequeued request must be list_empty() */
724 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
726 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
728 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
731 * Driver can handle struct request, if it either has an old style
732 * request_fn defined, or is blk-mq based.
734 static inline bool queue_is_rq_based(struct request_queue *q)
736 return q->request_fn || q->mq_ops;
739 static inline unsigned int blk_queue_cluster(struct request_queue *q)
741 return q->limits.cluster;
744 static inline enum blk_zoned_model
745 blk_queue_zoned_model(struct request_queue *q)
747 return q->limits.zoned;
750 static inline bool blk_queue_is_zoned(struct request_queue *q)
752 switch (blk_queue_zoned_model(q)) {
761 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
763 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
766 static inline bool rq_is_sync(struct request *rq)
768 return op_is_sync(rq->cmd_flags);
771 static inline bool blk_rl_full(struct request_list *rl, bool sync)
773 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
775 return rl->flags & flag;
778 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
780 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
785 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
787 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
792 static inline bool rq_mergeable(struct request *rq)
794 if (blk_rq_is_passthrough(rq))
797 if (req_op(rq) == REQ_OP_FLUSH)
800 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
803 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
805 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
811 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
813 if (bio_data(a) == bio_data(b))
819 static inline unsigned int blk_queue_depth(struct request_queue *q)
822 return q->queue_depth;
824 return q->nr_requests;
828 * q->prep_rq_fn return values
831 BLKPREP_OK, /* serve it */
832 BLKPREP_KILL, /* fatal error, kill, return -EIO */
833 BLKPREP_DEFER, /* leave on queue */
834 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
837 extern unsigned long blk_max_low_pfn, blk_max_pfn;
840 * standard bounce addresses:
842 * BLK_BOUNCE_HIGH : bounce all highmem pages
843 * BLK_BOUNCE_ANY : don't bounce anything
844 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
847 #if BITS_PER_LONG == 32
848 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
850 #define BLK_BOUNCE_HIGH -1ULL
852 #define BLK_BOUNCE_ANY (-1ULL)
853 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
856 * default timeout for SG_IO if none specified
858 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
859 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
862 extern int init_emergency_isa_pool(void);
863 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
865 static inline int init_emergency_isa_pool(void)
869 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
872 #endif /* CONFIG_MMU */
878 unsigned long offset;
883 struct req_iterator {
884 struct bvec_iter iter;
888 /* This should not be used directly - use rq_for_each_segment */
889 #define for_each_bio(_bio) \
890 for (; _bio; _bio = _bio->bi_next)
891 #define __rq_for_each_bio(_bio, rq) \
893 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
895 #define rq_for_each_segment(bvl, _rq, _iter) \
896 __rq_for_each_bio(_iter.bio, _rq) \
897 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
899 #define rq_iter_last(bvec, _iter) \
900 (_iter.bio->bi_next == NULL && \
901 bio_iter_last(bvec, _iter.iter))
903 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
904 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
906 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
907 extern void rq_flush_dcache_pages(struct request *rq);
909 static inline void rq_flush_dcache_pages(struct request *rq)
915 #define vfs_msg(sb, level, fmt, ...) \
916 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
918 #define vfs_msg(sb, level, fmt, ...) \
920 no_printk(fmt, ##__VA_ARGS__); \
921 __vfs_msg(sb, "", " "); \
925 extern int blk_register_queue(struct gendisk *disk);
926 extern void blk_unregister_queue(struct gendisk *disk);
927 extern blk_qc_t generic_make_request(struct bio *bio);
928 extern void blk_rq_init(struct request_queue *q, struct request *rq);
929 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
930 extern void blk_put_request(struct request *);
931 extern void __blk_put_request(struct request_queue *, struct request *);
932 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
933 extern void blk_requeue_request(struct request_queue *, struct request *);
934 extern int blk_lld_busy(struct request_queue *q);
935 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
936 struct bio_set *bs, gfp_t gfp_mask,
937 int (*bio_ctr)(struct bio *, struct bio *, void *),
939 extern void blk_rq_unprep_clone(struct request *rq);
940 extern int blk_insert_cloned_request(struct request_queue *q,
942 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
943 extern void blk_delay_queue(struct request_queue *, unsigned long);
944 extern void blk_queue_split(struct request_queue *, struct bio **,
946 extern void blk_recount_segments(struct request_queue *, struct bio *);
947 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
948 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
949 unsigned int, void __user *);
950 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
951 unsigned int, void __user *);
952 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
953 struct scsi_ioctl_command __user *);
955 extern int blk_queue_enter(struct request_queue *q, bool nowait);
956 extern void blk_queue_exit(struct request_queue *q);
957 extern void blk_start_queue(struct request_queue *q);
958 extern void blk_start_queue_async(struct request_queue *q);
959 extern void blk_stop_queue(struct request_queue *q);
960 extern void blk_sync_queue(struct request_queue *q);
961 extern void __blk_stop_queue(struct request_queue *q);
962 extern void __blk_run_queue(struct request_queue *q);
963 extern void __blk_run_queue_uncond(struct request_queue *q);
964 extern void blk_run_queue(struct request_queue *);
965 extern void blk_run_queue_async(struct request_queue *q);
966 extern void blk_mq_quiesce_queue(struct request_queue *q);
967 extern int blk_rq_map_user(struct request_queue *, struct request *,
968 struct rq_map_data *, void __user *, unsigned long,
970 extern int blk_rq_unmap_user(struct bio *);
971 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
972 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
973 struct rq_map_data *, const struct iov_iter *,
975 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
976 struct request *, int);
977 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
978 struct request *, int, rq_end_io_fn *);
980 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
982 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
984 return bdev->bd_disk->queue; /* this is never NULL */
988 * blk_rq_pos() : the current sector
989 * blk_rq_bytes() : bytes left in the entire request
990 * blk_rq_cur_bytes() : bytes left in the current segment
991 * blk_rq_err_bytes() : bytes left till the next error boundary
992 * blk_rq_sectors() : sectors left in the entire request
993 * blk_rq_cur_sectors() : sectors left in the current segment
995 static inline sector_t blk_rq_pos(const struct request *rq)
1000 static inline unsigned int blk_rq_bytes(const struct request *rq)
1002 return rq->__data_len;
1005 static inline int blk_rq_cur_bytes(const struct request *rq)
1007 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1010 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1012 static inline unsigned int blk_rq_sectors(const struct request *rq)
1014 return blk_rq_bytes(rq) >> 9;
1017 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1019 return blk_rq_cur_bytes(rq) >> 9;
1023 * Some commands like WRITE SAME have a payload or data transfer size which
1024 * is different from the size of the request. Any driver that supports such
1025 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1026 * calculate the data transfer size.
1028 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1030 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1031 return rq->special_vec.bv_len;
1032 return blk_rq_bytes(rq);
1035 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1038 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1039 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1041 if (unlikely(op == REQ_OP_WRITE_SAME))
1042 return q->limits.max_write_same_sectors;
1044 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1045 return q->limits.max_write_zeroes_sectors;
1047 return q->limits.max_sectors;
1051 * Return maximum size of a request at given offset. Only valid for
1052 * file system requests.
1054 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1057 if (!q->limits.chunk_sectors)
1058 return q->limits.max_sectors;
1060 return q->limits.chunk_sectors -
1061 (offset & (q->limits.chunk_sectors - 1));
1064 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1067 struct request_queue *q = rq->q;
1069 if (blk_rq_is_passthrough(rq))
1070 return q->limits.max_hw_sectors;
1072 if (!q->limits.chunk_sectors ||
1073 req_op(rq) == REQ_OP_DISCARD ||
1074 req_op(rq) == REQ_OP_SECURE_ERASE)
1075 return blk_queue_get_max_sectors(q, req_op(rq));
1077 return min(blk_max_size_offset(q, offset),
1078 blk_queue_get_max_sectors(q, req_op(rq)));
1081 static inline unsigned int blk_rq_count_bios(struct request *rq)
1083 unsigned int nr_bios = 0;
1086 __rq_for_each_bio(bio, rq)
1093 * Request issue related functions.
1095 extern struct request *blk_peek_request(struct request_queue *q);
1096 extern void blk_start_request(struct request *rq);
1097 extern struct request *blk_fetch_request(struct request_queue *q);
1100 * Request completion related functions.
1102 * blk_update_request() completes given number of bytes and updates
1103 * the request without completing it.
1105 * blk_end_request() and friends. __blk_end_request() must be called
1106 * with the request queue spinlock acquired.
1108 * Several drivers define their own end_request and call
1109 * blk_end_request() for parts of the original function.
1110 * This prevents code duplication in drivers.
1112 extern bool blk_update_request(struct request *rq, int error,
1113 unsigned int nr_bytes);
1114 extern void blk_finish_request(struct request *rq, int error);
1115 extern bool blk_end_request(struct request *rq, int error,
1116 unsigned int nr_bytes);
1117 extern void blk_end_request_all(struct request *rq, int error);
1118 extern bool __blk_end_request(struct request *rq, int error,
1119 unsigned int nr_bytes);
1120 extern void __blk_end_request_all(struct request *rq, int error);
1121 extern bool __blk_end_request_cur(struct request *rq, int error);
1123 extern void blk_complete_request(struct request *);
1124 extern void __blk_complete_request(struct request *);
1125 extern void blk_abort_request(struct request *);
1126 extern void blk_unprep_request(struct request *);
1129 * Access functions for manipulating queue properties
1131 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1132 spinlock_t *lock, int node_id);
1133 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1134 extern int blk_init_allocated_queue(struct request_queue *);
1135 extern void blk_cleanup_queue(struct request_queue *);
1136 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1137 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1138 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1139 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1140 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1141 extern void blk_queue_max_discard_segments(struct request_queue *,
1143 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1144 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1145 unsigned int max_discard_sectors);
1146 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1147 unsigned int max_write_same_sectors);
1148 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1149 unsigned int max_write_same_sectors);
1150 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1151 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1152 extern void blk_queue_alignment_offset(struct request_queue *q,
1153 unsigned int alignment);
1154 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1155 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1156 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1157 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1158 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1159 extern void blk_set_default_limits(struct queue_limits *lim);
1160 extern void blk_set_stacking_limits(struct queue_limits *lim);
1161 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1163 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1165 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1167 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1168 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1169 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1170 extern int blk_queue_dma_drain(struct request_queue *q,
1171 dma_drain_needed_fn *dma_drain_needed,
1172 void *buf, unsigned int size);
1173 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1174 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1175 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1176 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1177 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1178 extern void blk_queue_dma_alignment(struct request_queue *, int);
1179 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1180 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1181 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1182 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1183 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1184 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1187 * Number of physical segments as sent to the device.
1189 * Normally this is the number of discontiguous data segments sent by the
1190 * submitter. But for data-less command like discard we might have no
1191 * actual data segments submitted, but the driver might have to add it's
1192 * own special payload. In that case we still return 1 here so that this
1193 * special payload will be mapped.
1195 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1197 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1199 return rq->nr_phys_segments;
1203 * Number of discard segments (or ranges) the driver needs to fill in.
1204 * Each discard bio merged into a request is counted as one segment.
1206 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1208 return max_t(unsigned short, rq->nr_phys_segments, 1);
1211 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1212 extern void blk_dump_rq_flags(struct request *, char *);
1213 extern long nr_blockdev_pages(void);
1215 bool __must_check blk_get_queue(struct request_queue *);
1216 struct request_queue *blk_alloc_queue(gfp_t);
1217 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1218 extern void blk_put_queue(struct request_queue *);
1219 extern void blk_set_queue_dying(struct request_queue *);
1222 * block layer runtime pm functions
1225 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1226 extern int blk_pre_runtime_suspend(struct request_queue *q);
1227 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1228 extern void blk_pre_runtime_resume(struct request_queue *q);
1229 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1230 extern void blk_set_runtime_active(struct request_queue *q);
1232 static inline void blk_pm_runtime_init(struct request_queue *q,
1233 struct device *dev) {}
1234 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1238 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1239 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1240 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1241 static inline void blk_set_runtime_active(struct request_queue *q) {}
1245 * blk_plug permits building a queue of related requests by holding the I/O
1246 * fragments for a short period. This allows merging of sequential requests
1247 * into single larger request. As the requests are moved from a per-task list to
1248 * the device's request_queue in a batch, this results in improved scalability
1249 * as the lock contention for request_queue lock is reduced.
1251 * It is ok not to disable preemption when adding the request to the plug list
1252 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1253 * the plug list when the task sleeps by itself. For details, please see
1254 * schedule() where blk_schedule_flush_plug() is called.
1257 struct list_head list; /* requests */
1258 struct list_head mq_list; /* blk-mq requests */
1259 struct list_head cb_list; /* md requires an unplug callback */
1261 #define BLK_MAX_REQUEST_COUNT 16
1262 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1265 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1266 struct blk_plug_cb {
1267 struct list_head list;
1268 blk_plug_cb_fn callback;
1271 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1272 void *data, int size);
1273 extern void blk_start_plug(struct blk_plug *);
1274 extern void blk_finish_plug(struct blk_plug *);
1275 extern void blk_flush_plug_list(struct blk_plug *, bool);
1277 static inline void blk_flush_plug(struct task_struct *tsk)
1279 struct blk_plug *plug = tsk->plug;
1282 blk_flush_plug_list(plug, false);
1285 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1287 struct blk_plug *plug = tsk->plug;
1290 blk_flush_plug_list(plug, true);
1293 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1295 struct blk_plug *plug = tsk->plug;
1298 (!list_empty(&plug->list) ||
1299 !list_empty(&plug->mq_list) ||
1300 !list_empty(&plug->cb_list));
1306 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1307 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1308 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1309 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1310 extern void blk_queue_free_tags(struct request_queue *);
1311 extern int blk_queue_resize_tags(struct request_queue *, int);
1312 extern void blk_queue_invalidate_tags(struct request_queue *);
1313 extern struct blk_queue_tag *blk_init_tags(int, int);
1314 extern void blk_free_tags(struct blk_queue_tag *);
1316 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1319 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1321 return bqt->tag_index[tag];
1324 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1325 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1326 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1328 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1330 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1331 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1332 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1333 sector_t nr_sects, gfp_t gfp_mask, int flags,
1336 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1337 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1339 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1340 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1342 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1343 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1345 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1346 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1348 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1349 nr_blocks << (sb->s_blocksize_bits - 9),
1352 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1353 sector_t nr_blocks, gfp_t gfp_mask)
1355 return blkdev_issue_zeroout(sb->s_bdev,
1356 block << (sb->s_blocksize_bits - 9),
1357 nr_blocks << (sb->s_blocksize_bits - 9),
1361 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1363 enum blk_default_limits {
1364 BLK_MAX_SEGMENTS = 128,
1365 BLK_SAFE_MAX_SECTORS = 255,
1366 BLK_DEF_MAX_SECTORS = 2560,
1367 BLK_MAX_SEGMENT_SIZE = 65536,
1368 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1371 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1373 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1375 return q->limits.bounce_pfn;
1378 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1380 return q->limits.seg_boundary_mask;
1383 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1385 return q->limits.virt_boundary_mask;
1388 static inline unsigned int queue_max_sectors(struct request_queue *q)
1390 return q->limits.max_sectors;
1393 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1395 return q->limits.max_hw_sectors;
1398 static inline unsigned short queue_max_segments(struct request_queue *q)
1400 return q->limits.max_segments;
1403 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1405 return q->limits.max_discard_segments;
1408 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1410 return q->limits.max_segment_size;
1413 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1417 if (q && q->limits.logical_block_size)
1418 retval = q->limits.logical_block_size;
1423 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1425 return queue_logical_block_size(bdev_get_queue(bdev));
1428 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1430 return q->limits.physical_block_size;
1433 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1435 return queue_physical_block_size(bdev_get_queue(bdev));
1438 static inline unsigned int queue_io_min(struct request_queue *q)
1440 return q->limits.io_min;
1443 static inline int bdev_io_min(struct block_device *bdev)
1445 return queue_io_min(bdev_get_queue(bdev));
1448 static inline unsigned int queue_io_opt(struct request_queue *q)
1450 return q->limits.io_opt;
1453 static inline int bdev_io_opt(struct block_device *bdev)
1455 return queue_io_opt(bdev_get_queue(bdev));
1458 static inline int queue_alignment_offset(struct request_queue *q)
1460 if (q->limits.misaligned)
1463 return q->limits.alignment_offset;
1466 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1468 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1469 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1471 return (granularity + lim->alignment_offset - alignment) % granularity;
1474 static inline int bdev_alignment_offset(struct block_device *bdev)
1476 struct request_queue *q = bdev_get_queue(bdev);
1478 if (q->limits.misaligned)
1481 if (bdev != bdev->bd_contains)
1482 return bdev->bd_part->alignment_offset;
1484 return q->limits.alignment_offset;
1487 static inline int queue_discard_alignment(struct request_queue *q)
1489 if (q->limits.discard_misaligned)
1492 return q->limits.discard_alignment;
1495 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1497 unsigned int alignment, granularity, offset;
1499 if (!lim->max_discard_sectors)
1502 /* Why are these in bytes, not sectors? */
1503 alignment = lim->discard_alignment >> 9;
1504 granularity = lim->discard_granularity >> 9;
1508 /* Offset of the partition start in 'granularity' sectors */
1509 offset = sector_div(sector, granularity);
1511 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1512 offset = (granularity + alignment - offset) % granularity;
1514 /* Turn it back into bytes, gaah */
1518 static inline int bdev_discard_alignment(struct block_device *bdev)
1520 struct request_queue *q = bdev_get_queue(bdev);
1522 if (bdev != bdev->bd_contains)
1523 return bdev->bd_part->discard_alignment;
1525 return q->limits.discard_alignment;
1528 static inline unsigned int bdev_write_same(struct block_device *bdev)
1530 struct request_queue *q = bdev_get_queue(bdev);
1533 return q->limits.max_write_same_sectors;
1538 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1540 struct request_queue *q = bdev_get_queue(bdev);
1543 return q->limits.max_write_zeroes_sectors;
1548 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1550 struct request_queue *q = bdev_get_queue(bdev);
1553 return blk_queue_zoned_model(q);
1555 return BLK_ZONED_NONE;
1558 static inline bool bdev_is_zoned(struct block_device *bdev)
1560 struct request_queue *q = bdev_get_queue(bdev);
1563 return blk_queue_is_zoned(q);
1568 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1570 struct request_queue *q = bdev_get_queue(bdev);
1573 return blk_queue_zone_sectors(q);
1578 static inline int queue_dma_alignment(struct request_queue *q)
1580 return q ? q->dma_alignment : 511;
1583 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1586 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1587 return !(addr & alignment) && !(len & alignment);
1590 /* assumes size > 256 */
1591 static inline unsigned int blksize_bits(unsigned int size)
1593 unsigned int bits = 8;
1597 } while (size > 256);
1601 static inline unsigned int block_size(struct block_device *bdev)
1603 return bdev->bd_block_size;
1606 static inline bool queue_flush_queueable(struct request_queue *q)
1608 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1611 typedef struct {struct page *v;} Sector;
1613 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1615 static inline void put_dev_sector(Sector p)
1620 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1621 struct bio_vec *bprv, unsigned int offset)
1624 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1628 * Check if adding a bio_vec after bprv with offset would create a gap in
1629 * the SG list. Most drivers don't care about this, but some do.
1631 static inline bool bvec_gap_to_prev(struct request_queue *q,
1632 struct bio_vec *bprv, unsigned int offset)
1634 if (!queue_virt_boundary(q))
1636 return __bvec_gap_to_prev(q, bprv, offset);
1640 * Check if the two bvecs from two bios can be merged to one segment.
1641 * If yes, no need to check gap between the two bios since the 1st bio
1642 * and the 1st bvec in the 2nd bio can be handled in one segment.
1644 static inline bool bios_segs_mergeable(struct request_queue *q,
1645 struct bio *prev, struct bio_vec *prev_last_bv,
1646 struct bio_vec *next_first_bv)
1648 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1650 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1652 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1653 queue_max_segment_size(q))
1658 static inline bool bio_will_gap(struct request_queue *q,
1659 struct request *prev_rq,
1663 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1664 struct bio_vec pb, nb;
1667 * don't merge if the 1st bio starts with non-zero
1668 * offset, otherwise it is quite difficult to respect
1669 * sg gap limit. We work hard to merge a huge number of small
1670 * single bios in case of mkfs.
1673 bio_get_first_bvec(prev_rq->bio, &pb);
1675 bio_get_first_bvec(prev, &pb);
1680 * We don't need to worry about the situation that the
1681 * merged segment ends in unaligned virt boundary:
1683 * - if 'pb' ends aligned, the merged segment ends aligned
1684 * - if 'pb' ends unaligned, the next bio must include
1685 * one single bvec of 'nb', otherwise the 'nb' can't
1688 bio_get_last_bvec(prev, &pb);
1689 bio_get_first_bvec(next, &nb);
1691 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1692 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1698 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1700 return bio_will_gap(req->q, req, req->biotail, bio);
1703 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1705 return bio_will_gap(req->q, NULL, bio, req->bio);
1708 int kblockd_schedule_work(struct work_struct *work);
1709 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1710 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1711 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1712 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1714 #ifdef CONFIG_BLK_CGROUP
1716 * This should not be using sched_clock(). A real patch is in progress
1717 * to fix this up, until that is in place we need to disable preemption
1718 * around sched_clock() in this function and set_io_start_time_ns().
1720 static inline void set_start_time_ns(struct request *req)
1723 req->start_time_ns = sched_clock();
1727 static inline void set_io_start_time_ns(struct request *req)
1730 req->io_start_time_ns = sched_clock();
1734 static inline uint64_t rq_start_time_ns(struct request *req)
1736 return req->start_time_ns;
1739 static inline uint64_t rq_io_start_time_ns(struct request *req)
1741 return req->io_start_time_ns;
1744 static inline void set_start_time_ns(struct request *req) {}
1745 static inline void set_io_start_time_ns(struct request *req) {}
1746 static inline uint64_t rq_start_time_ns(struct request *req)
1750 static inline uint64_t rq_io_start_time_ns(struct request *req)
1756 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1757 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1758 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1759 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1761 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1763 enum blk_integrity_flags {
1764 BLK_INTEGRITY_VERIFY = 1 << 0,
1765 BLK_INTEGRITY_GENERATE = 1 << 1,
1766 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1767 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1770 struct blk_integrity_iter {
1774 unsigned int data_size;
1775 unsigned short interval;
1776 const char *disk_name;
1779 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1781 struct blk_integrity_profile {
1782 integrity_processing_fn *generate_fn;
1783 integrity_processing_fn *verify_fn;
1787 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1788 extern void blk_integrity_unregister(struct gendisk *);
1789 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1790 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1791 struct scatterlist *);
1792 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1793 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1795 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1798 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1800 struct blk_integrity *bi = &disk->queue->integrity;
1809 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1811 return blk_get_integrity(bdev->bd_disk);
1814 static inline bool blk_integrity_rq(struct request *rq)
1816 return rq->cmd_flags & REQ_INTEGRITY;
1819 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1822 q->limits.max_integrity_segments = segs;
1825 static inline unsigned short
1826 queue_max_integrity_segments(struct request_queue *q)
1828 return q->limits.max_integrity_segments;
1831 static inline bool integrity_req_gap_back_merge(struct request *req,
1834 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1835 struct bio_integrity_payload *bip_next = bio_integrity(next);
1837 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1838 bip_next->bip_vec[0].bv_offset);
1841 static inline bool integrity_req_gap_front_merge(struct request *req,
1844 struct bio_integrity_payload *bip = bio_integrity(bio);
1845 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1847 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1848 bip_next->bip_vec[0].bv_offset);
1851 #else /* CONFIG_BLK_DEV_INTEGRITY */
1854 struct block_device;
1856 struct blk_integrity;
1858 static inline int blk_integrity_rq(struct request *rq)
1862 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1867 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1869 struct scatterlist *s)
1873 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1877 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1881 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1885 static inline void blk_integrity_register(struct gendisk *d,
1886 struct blk_integrity *b)
1889 static inline void blk_integrity_unregister(struct gendisk *d)
1892 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1896 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1900 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1906 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1913 static inline bool integrity_req_gap_back_merge(struct request *req,
1918 static inline bool integrity_req_gap_front_merge(struct request *req,
1924 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1926 struct block_device_operations {
1927 int (*open) (struct block_device *, fmode_t);
1928 void (*release) (struct gendisk *, fmode_t);
1929 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1930 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1931 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1932 unsigned int (*check_events) (struct gendisk *disk,
1933 unsigned int clearing);
1934 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1935 int (*media_changed) (struct gendisk *);
1936 void (*unlock_native_capacity) (struct gendisk *);
1937 int (*revalidate_disk) (struct gendisk *);
1938 int (*getgeo)(struct block_device *, struct hd_geometry *);
1939 /* this callback is with swap_lock and sometimes page table lock held */
1940 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1941 struct module *owner;
1942 const struct pr_ops *pr_ops;
1945 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1947 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1948 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1949 struct writeback_control *);
1950 #else /* CONFIG_BLOCK */
1952 struct block_device;
1955 * stubs for when the block layer is configured out
1957 #define buffer_heads_over_limit 0
1959 static inline long nr_blockdev_pages(void)
1967 static inline void blk_start_plug(struct blk_plug *plug)
1971 static inline void blk_finish_plug(struct blk_plug *plug)
1975 static inline void blk_flush_plug(struct task_struct *task)
1979 static inline void blk_schedule_flush_plug(struct task_struct *task)
1984 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1989 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1990 sector_t *error_sector)
1995 #endif /* CONFIG_BLOCK */