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;
44 #define BLKDEV_MIN_RQ 4
45 #define BLKDEV_MAX_RQ 128 /* Default maximum */
48 * Maximum number of blkcg policies allowed to be registered concurrently.
49 * Defined here to simplify include dependency.
51 #define BLKCG_MAX_POLS 2
53 typedef void (rq_end_io_fn)(struct request *, int);
55 #define BLK_RL_SYNCFULL (1U << 0)
56 #define BLK_RL_ASYNCFULL (1U << 1)
59 struct request_queue *q; /* the queue this rl belongs to */
60 #ifdef CONFIG_BLK_CGROUP
61 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
64 * count[], starved[], and wait[] are indexed by
65 * BLK_RW_SYNC/BLK_RW_ASYNC
70 wait_queue_head_t wait[2];
76 typedef __u32 __bitwise req_flags_t;
78 /* elevator knows about this request */
79 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
80 /* drive already may have started this one */
81 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
82 /* uses tagged queueing */
83 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
84 /* may not be passed by ioscheduler */
85 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
86 /* request for flush sequence */
87 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
88 /* merge of different types, fail separately */
89 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
90 /* track inflight for MQ */
91 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
92 /* don't call prep for this one */
93 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
94 /* set for "ide_preempt" requests and also for requests for which the SCSI
95 "quiesce" state must be ignored. */
96 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
97 /* contains copies of user pages */
98 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
99 /* vaguely specified driver internal error. Ignored by the block layer */
100 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
101 /* don't warn about errors */
102 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
103 /* elevator private data attached */
104 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
105 /* account I/O stat */
106 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
107 /* request came from our alloc pool */
108 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
109 /* runtime pm request */
110 #define RQF_PM ((__force req_flags_t)(1 << 15))
111 /* on IO scheduler merge hash */
112 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
113 /* IO stats tracking on */
114 #define RQF_STATS ((__force req_flags_t)(1 << 17))
115 /* Look at ->special_vec for the actual data payload instead of the
117 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
119 /* flags that prevent us from merging requests: */
120 #define RQF_NOMERGE_FLAGS \
121 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
124 * Try to put the fields that are referenced together in the same cacheline.
126 * If you modify this structure, make sure to update blk_rq_init() and
127 * especially blk_mq_rq_ctx_init() to take care of the added fields.
130 struct list_head queuelist;
132 struct call_single_data csd;
136 struct request_queue *q;
137 struct blk_mq_ctx *mq_ctx;
140 unsigned int cmd_flags; /* op and common flags */
141 req_flags_t rq_flags;
145 unsigned long atomic_flags;
147 /* the following two fields are internal, NEVER access directly */
148 unsigned int __data_len; /* total data len */
150 sector_t __sector; /* sector cursor */
156 * The hash is used inside the scheduler, and killed once the
157 * request reaches the dispatch list. The ipi_list is only used
158 * to queue the request for softirq completion, which is long
159 * after the request has been unhashed (and even removed from
160 * the dispatch list).
163 struct hlist_node hash; /* merge hash */
164 struct list_head ipi_list;
168 * The rb_node is only used inside the io scheduler, requests
169 * are pruned when moved to the dispatch queue. So let the
170 * completion_data share space with the rb_node.
173 struct rb_node rb_node; /* sort/lookup */
174 struct bio_vec special_vec;
175 void *completion_data;
179 * Three pointers are available for the IO schedulers, if they need
180 * more they have to dynamically allocate it. Flush requests are
181 * never put on the IO scheduler. So let the flush fields share
182 * space with the elevator data.
192 struct list_head list;
193 rq_end_io_fn *saved_end_io;
197 struct gendisk *rq_disk;
198 struct hd_struct *part;
199 unsigned long start_time;
200 struct blk_issue_stat issue_stat;
201 #ifdef CONFIG_BLK_CGROUP
202 struct request_list *rl; /* rl this rq is alloced from */
203 unsigned long long start_time_ns;
204 unsigned long long io_start_time_ns; /* when passed to hardware */
206 /* Number of scatter-gather DMA addr+len pairs after
207 * physical address coalescing is performed.
209 unsigned short nr_phys_segments;
210 #if defined(CONFIG_BLK_DEV_INTEGRITY)
211 unsigned short nr_integrity_segments;
214 unsigned short ioprio;
216 void *special; /* opaque pointer available for LLD use */
220 unsigned int extra_len; /* length of alignment and padding */
222 unsigned long deadline;
223 struct list_head timeout_list;
224 unsigned int timeout;
228 * completion callback.
230 rq_end_io_fn *end_io;
234 struct request *next_rq;
237 static inline bool blk_rq_is_scsi(struct request *rq)
239 return req_op(rq) == REQ_OP_SCSI_IN || req_op(rq) == REQ_OP_SCSI_OUT;
242 static inline bool blk_rq_is_private(struct request *rq)
244 return req_op(rq) == REQ_OP_DRV_IN || req_op(rq) == REQ_OP_DRV_OUT;
247 static inline bool blk_rq_is_passthrough(struct request *rq)
249 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
252 static inline unsigned short req_get_ioprio(struct request *req)
257 #include <linux/elevator.h>
259 struct blk_queue_ctx;
261 typedef void (request_fn_proc) (struct request_queue *q);
262 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
263 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
264 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
267 typedef void (softirq_done_fn)(struct request *);
268 typedef int (dma_drain_needed_fn)(struct request *);
269 typedef int (lld_busy_fn) (struct request_queue *q);
270 typedef int (bsg_job_fn) (struct bsg_job *);
271 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
272 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
274 enum blk_eh_timer_return {
280 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
282 enum blk_queue_state {
287 struct blk_queue_tag {
288 struct request **tag_index; /* map of busy tags */
289 unsigned long *tag_map; /* bit map of free/busy tags */
290 int max_depth; /* what we will send to device */
291 int real_max_depth; /* what the array can hold */
292 atomic_t refcnt; /* map can be shared */
293 int alloc_policy; /* tag allocation policy */
294 int next_tag; /* next tag */
296 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
297 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
299 #define BLK_SCSI_MAX_CMDS (256)
300 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
303 * Zoned block device models (zoned limit).
305 enum blk_zoned_model {
306 BLK_ZONED_NONE, /* Regular block device */
307 BLK_ZONED_HA, /* Host-aware zoned block device */
308 BLK_ZONED_HM, /* Host-managed zoned block device */
311 struct queue_limits {
312 unsigned long bounce_pfn;
313 unsigned long seg_boundary_mask;
314 unsigned long virt_boundary_mask;
316 unsigned int max_hw_sectors;
317 unsigned int max_dev_sectors;
318 unsigned int chunk_sectors;
319 unsigned int max_sectors;
320 unsigned int max_segment_size;
321 unsigned int physical_block_size;
322 unsigned int alignment_offset;
325 unsigned int max_discard_sectors;
326 unsigned int max_hw_discard_sectors;
327 unsigned int max_write_same_sectors;
328 unsigned int max_write_zeroes_sectors;
329 unsigned int discard_granularity;
330 unsigned int discard_alignment;
332 unsigned short logical_block_size;
333 unsigned short max_segments;
334 unsigned short max_integrity_segments;
335 unsigned short max_discard_segments;
337 unsigned char misaligned;
338 unsigned char discard_misaligned;
339 unsigned char cluster;
340 unsigned char discard_zeroes_data;
341 unsigned char raid_partial_stripes_expensive;
342 enum blk_zoned_model zoned;
345 #ifdef CONFIG_BLK_DEV_ZONED
347 struct blk_zone_report_hdr {
348 unsigned int nr_zones;
352 extern int blkdev_report_zones(struct block_device *bdev,
353 sector_t sector, struct blk_zone *zones,
354 unsigned int *nr_zones, gfp_t gfp_mask);
355 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
356 sector_t nr_sectors, gfp_t gfp_mask);
358 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
359 unsigned int cmd, unsigned long arg);
360 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
361 unsigned int cmd, unsigned long arg);
363 #else /* CONFIG_BLK_DEV_ZONED */
365 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
366 fmode_t mode, unsigned int cmd,
372 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
373 fmode_t mode, unsigned int cmd,
379 #endif /* CONFIG_BLK_DEV_ZONED */
381 struct request_queue {
383 * Together with queue_head for cacheline sharing
385 struct list_head queue_head;
386 struct request *last_merge;
387 struct elevator_queue *elevator;
388 int nr_rqs[2]; /* # allocated [a]sync rqs */
389 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
394 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
395 * is used, root blkg allocates from @q->root_rl and all other
396 * blkgs from their own blkg->rl. Which one to use should be
397 * determined using bio_request_list().
399 struct request_list root_rl;
401 request_fn_proc *request_fn;
402 make_request_fn *make_request_fn;
403 prep_rq_fn *prep_rq_fn;
404 unprep_rq_fn *unprep_rq_fn;
405 softirq_done_fn *softirq_done_fn;
406 rq_timed_out_fn *rq_timed_out_fn;
407 dma_drain_needed_fn *dma_drain_needed;
408 lld_busy_fn *lld_busy_fn;
409 init_rq_fn *init_rq_fn;
410 exit_rq_fn *exit_rq_fn;
412 const struct blk_mq_ops *mq_ops;
414 unsigned int *mq_map;
417 struct blk_mq_ctx __percpu *queue_ctx;
418 unsigned int nr_queues;
420 unsigned int queue_depth;
422 /* hw dispatch queues */
423 struct blk_mq_hw_ctx **queue_hw_ctx;
424 unsigned int nr_hw_queues;
427 * Dispatch queue sorting
430 struct request *boundary_rq;
433 * Delayed queue handling
435 struct delayed_work delay_work;
437 struct backing_dev_info *backing_dev_info;
440 * The queue owner gets to use this for whatever they like.
441 * ll_rw_blk doesn't touch it.
446 * various queue flags, see QUEUE_* below
448 unsigned long queue_flags;
451 * ida allocated id for this queue. Used to index queues from
457 * queue needs bounce pages for pages above this limit
462 * protects queue structures from reentrancy. ->__queue_lock should
463 * _never_ be used directly, it is queue private. always use
466 spinlock_t __queue_lock;
467 spinlock_t *queue_lock;
477 struct kobject mq_kobj;
479 #ifdef CONFIG_BLK_DEV_INTEGRITY
480 struct blk_integrity integrity;
481 #endif /* CONFIG_BLK_DEV_INTEGRITY */
486 unsigned int nr_pending;
492 unsigned long nr_requests; /* Max # of requests */
493 unsigned int nr_congestion_on;
494 unsigned int nr_congestion_off;
495 unsigned int nr_batching;
497 unsigned int dma_drain_size;
498 void *dma_drain_buffer;
499 unsigned int dma_pad_mask;
500 unsigned int dma_alignment;
502 struct blk_queue_tag *queue_tags;
503 struct list_head tag_busy_list;
505 unsigned int nr_sorted;
506 unsigned int in_flight[2];
508 struct blk_rq_stat rq_stats[2];
511 * Number of active block driver functions for which blk_drain_queue()
512 * must wait. Must be incremented around functions that unlock the
513 * queue_lock internally, e.g. scsi_request_fn().
515 unsigned int request_fn_active;
517 unsigned int rq_timeout;
519 struct timer_list timeout;
520 struct work_struct timeout_work;
521 struct list_head timeout_list;
523 struct list_head icq_list;
524 #ifdef CONFIG_BLK_CGROUP
525 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
526 struct blkcg_gq *root_blkg;
527 struct list_head blkg_list;
530 struct queue_limits limits;
535 unsigned int sg_timeout;
536 unsigned int sg_reserved_size;
538 #ifdef CONFIG_BLK_DEV_IO_TRACE
539 struct blk_trace *blk_trace;
542 * for flush operations
544 struct blk_flush_queue *fq;
546 struct list_head requeue_list;
547 spinlock_t requeue_lock;
548 struct delayed_work requeue_work;
550 struct mutex sysfs_lock;
553 atomic_t mq_freeze_depth;
555 #if defined(CONFIG_BLK_DEV_BSG)
556 bsg_job_fn *bsg_job_fn;
558 struct bsg_class_device bsg_dev;
561 #ifdef CONFIG_BLK_DEV_THROTTLING
563 struct throtl_data *td;
565 struct rcu_head rcu_head;
566 wait_queue_head_t mq_freeze_wq;
567 struct percpu_ref q_usage_counter;
568 struct list_head all_q_node;
570 struct blk_mq_tag_set *tag_set;
571 struct list_head tag_set_list;
572 struct bio_set *bio_split;
574 #ifdef CONFIG_BLK_DEBUG_FS
575 struct dentry *debugfs_dir;
576 struct dentry *mq_debugfs_dir;
579 bool mq_sysfs_init_done;
585 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
586 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
587 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
588 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
589 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
590 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
591 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
592 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
593 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
594 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
595 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
596 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
597 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
598 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
599 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
600 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
601 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
602 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
603 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
604 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
605 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
606 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
607 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
608 #define QUEUE_FLAG_WC 23 /* Write back caching */
609 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
610 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
611 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
612 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
613 #define QUEUE_FLAG_RESTART 28 /* queue needs restart at completion */
615 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
616 (1 << QUEUE_FLAG_STACKABLE) | \
617 (1 << QUEUE_FLAG_SAME_COMP) | \
618 (1 << QUEUE_FLAG_ADD_RANDOM))
620 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
621 (1 << QUEUE_FLAG_STACKABLE) | \
622 (1 << QUEUE_FLAG_SAME_COMP) | \
623 (1 << QUEUE_FLAG_POLL))
625 static inline void queue_lockdep_assert_held(struct request_queue *q)
628 lockdep_assert_held(q->queue_lock);
631 static inline void queue_flag_set_unlocked(unsigned int flag,
632 struct request_queue *q)
634 __set_bit(flag, &q->queue_flags);
637 static inline int queue_flag_test_and_clear(unsigned int flag,
638 struct request_queue *q)
640 queue_lockdep_assert_held(q);
642 if (test_bit(flag, &q->queue_flags)) {
643 __clear_bit(flag, &q->queue_flags);
650 static inline int queue_flag_test_and_set(unsigned int flag,
651 struct request_queue *q)
653 queue_lockdep_assert_held(q);
655 if (!test_bit(flag, &q->queue_flags)) {
656 __set_bit(flag, &q->queue_flags);
663 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
665 queue_lockdep_assert_held(q);
666 __set_bit(flag, &q->queue_flags);
669 static inline void queue_flag_clear_unlocked(unsigned int flag,
670 struct request_queue *q)
672 __clear_bit(flag, &q->queue_flags);
675 static inline int queue_in_flight(struct request_queue *q)
677 return q->in_flight[0] + q->in_flight[1];
680 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
682 queue_lockdep_assert_held(q);
683 __clear_bit(flag, &q->queue_flags);
686 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
687 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
688 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
689 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
690 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
691 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
692 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
693 #define blk_queue_noxmerges(q) \
694 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
695 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
696 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
697 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
698 #define blk_queue_stackable(q) \
699 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
700 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
701 #define blk_queue_secure_erase(q) \
702 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
703 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
705 #define blk_noretry_request(rq) \
706 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
707 REQ_FAILFAST_DRIVER))
709 static inline bool blk_account_rq(struct request *rq)
711 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
714 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
715 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
716 /* rq->queuelist of dequeued request must be list_empty() */
717 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
719 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
721 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
724 * Driver can handle struct request, if it either has an old style
725 * request_fn defined, or is blk-mq based.
727 static inline bool queue_is_rq_based(struct request_queue *q)
729 return q->request_fn || q->mq_ops;
732 static inline unsigned int blk_queue_cluster(struct request_queue *q)
734 return q->limits.cluster;
737 static inline enum blk_zoned_model
738 blk_queue_zoned_model(struct request_queue *q)
740 return q->limits.zoned;
743 static inline bool blk_queue_is_zoned(struct request_queue *q)
745 switch (blk_queue_zoned_model(q)) {
754 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
756 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
759 static inline bool rq_is_sync(struct request *rq)
761 return op_is_sync(rq->cmd_flags);
764 static inline bool blk_rl_full(struct request_list *rl, bool sync)
766 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
768 return rl->flags & flag;
771 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
773 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
778 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
780 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
785 static inline bool rq_mergeable(struct request *rq)
787 if (blk_rq_is_passthrough(rq))
790 if (req_op(rq) == REQ_OP_FLUSH)
793 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
796 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
798 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
804 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
806 if (bio_data(a) == bio_data(b))
812 static inline unsigned int blk_queue_depth(struct request_queue *q)
815 return q->queue_depth;
817 return q->nr_requests;
821 * q->prep_rq_fn return values
824 BLKPREP_OK, /* serve it */
825 BLKPREP_KILL, /* fatal error, kill, return -EIO */
826 BLKPREP_DEFER, /* leave on queue */
827 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
830 extern unsigned long blk_max_low_pfn, blk_max_pfn;
833 * standard bounce addresses:
835 * BLK_BOUNCE_HIGH : bounce all highmem pages
836 * BLK_BOUNCE_ANY : don't bounce anything
837 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
840 #if BITS_PER_LONG == 32
841 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
843 #define BLK_BOUNCE_HIGH -1ULL
845 #define BLK_BOUNCE_ANY (-1ULL)
846 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
849 * default timeout for SG_IO if none specified
851 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
852 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
855 extern int init_emergency_isa_pool(void);
856 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
858 static inline int init_emergency_isa_pool(void)
862 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
865 #endif /* CONFIG_MMU */
871 unsigned long offset;
876 struct req_iterator {
877 struct bvec_iter iter;
881 /* This should not be used directly - use rq_for_each_segment */
882 #define for_each_bio(_bio) \
883 for (; _bio; _bio = _bio->bi_next)
884 #define __rq_for_each_bio(_bio, rq) \
886 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
888 #define rq_for_each_segment(bvl, _rq, _iter) \
889 __rq_for_each_bio(_iter.bio, _rq) \
890 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
892 #define rq_iter_last(bvec, _iter) \
893 (_iter.bio->bi_next == NULL && \
894 bio_iter_last(bvec, _iter.iter))
896 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
897 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
899 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
900 extern void rq_flush_dcache_pages(struct request *rq);
902 static inline void rq_flush_dcache_pages(struct request *rq)
908 #define vfs_msg(sb, level, fmt, ...) \
909 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
911 #define vfs_msg(sb, level, fmt, ...) \
913 no_printk(fmt, ##__VA_ARGS__); \
914 __vfs_msg(sb, "", " "); \
918 extern int blk_register_queue(struct gendisk *disk);
919 extern void blk_unregister_queue(struct gendisk *disk);
920 extern blk_qc_t generic_make_request(struct bio *bio);
921 extern void blk_rq_init(struct request_queue *q, struct request *rq);
922 extern void blk_put_request(struct request *);
923 extern void __blk_put_request(struct request_queue *, struct request *);
924 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
925 extern void blk_requeue_request(struct request_queue *, struct request *);
926 extern int blk_lld_busy(struct request_queue *q);
927 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
928 struct bio_set *bs, gfp_t gfp_mask,
929 int (*bio_ctr)(struct bio *, struct bio *, void *),
931 extern void blk_rq_unprep_clone(struct request *rq);
932 extern int blk_insert_cloned_request(struct request_queue *q,
934 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
935 extern void blk_delay_queue(struct request_queue *, unsigned long);
936 extern void blk_queue_split(struct request_queue *, struct bio **,
938 extern void blk_recount_segments(struct request_queue *, struct bio *);
939 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
940 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
941 unsigned int, void __user *);
942 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
943 unsigned int, void __user *);
944 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
945 struct scsi_ioctl_command __user *);
947 extern int blk_queue_enter(struct request_queue *q, bool nowait);
948 extern void blk_queue_exit(struct request_queue *q);
949 extern void blk_start_queue(struct request_queue *q);
950 extern void blk_start_queue_async(struct request_queue *q);
951 extern void blk_stop_queue(struct request_queue *q);
952 extern void blk_sync_queue(struct request_queue *q);
953 extern void __blk_stop_queue(struct request_queue *q);
954 extern void __blk_run_queue(struct request_queue *q);
955 extern void __blk_run_queue_uncond(struct request_queue *q);
956 extern void blk_run_queue(struct request_queue *);
957 extern void blk_run_queue_async(struct request_queue *q);
958 extern void blk_mq_quiesce_queue(struct request_queue *q);
959 extern int blk_rq_map_user(struct request_queue *, struct request *,
960 struct rq_map_data *, void __user *, unsigned long,
962 extern int blk_rq_unmap_user(struct bio *);
963 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
964 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
965 struct rq_map_data *, const struct iov_iter *,
967 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
968 struct request *, int);
969 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
970 struct request *, int, rq_end_io_fn *);
972 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
974 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
976 return bdev->bd_disk->queue; /* this is never NULL */
980 * blk_rq_pos() : the current sector
981 * blk_rq_bytes() : bytes left in the entire request
982 * blk_rq_cur_bytes() : bytes left in the current segment
983 * blk_rq_err_bytes() : bytes left till the next error boundary
984 * blk_rq_sectors() : sectors left in the entire request
985 * blk_rq_cur_sectors() : sectors left in the current segment
987 static inline sector_t blk_rq_pos(const struct request *rq)
992 static inline unsigned int blk_rq_bytes(const struct request *rq)
994 return rq->__data_len;
997 static inline int blk_rq_cur_bytes(const struct request *rq)
999 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1002 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1004 static inline unsigned int blk_rq_sectors(const struct request *rq)
1006 return blk_rq_bytes(rq) >> 9;
1009 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1011 return blk_rq_cur_bytes(rq) >> 9;
1015 * Some commands like WRITE SAME have a payload or data transfer size which
1016 * is different from the size of the request. Any driver that supports such
1017 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1018 * calculate the data transfer size.
1020 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1022 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1023 return rq->special_vec.bv_len;
1024 return blk_rq_bytes(rq);
1027 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1030 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1031 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1033 if (unlikely(op == REQ_OP_WRITE_SAME))
1034 return q->limits.max_write_same_sectors;
1036 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1037 return q->limits.max_write_zeroes_sectors;
1039 return q->limits.max_sectors;
1043 * Return maximum size of a request at given offset. Only valid for
1044 * file system requests.
1046 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1049 if (!q->limits.chunk_sectors)
1050 return q->limits.max_sectors;
1052 return q->limits.chunk_sectors -
1053 (offset & (q->limits.chunk_sectors - 1));
1056 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1059 struct request_queue *q = rq->q;
1061 if (blk_rq_is_passthrough(rq))
1062 return q->limits.max_hw_sectors;
1064 if (!q->limits.chunk_sectors ||
1065 req_op(rq) == REQ_OP_DISCARD ||
1066 req_op(rq) == REQ_OP_SECURE_ERASE)
1067 return blk_queue_get_max_sectors(q, req_op(rq));
1069 return min(blk_max_size_offset(q, offset),
1070 blk_queue_get_max_sectors(q, req_op(rq)));
1073 static inline unsigned int blk_rq_count_bios(struct request *rq)
1075 unsigned int nr_bios = 0;
1078 __rq_for_each_bio(bio, rq)
1085 * blk_rq_set_prio - associate a request with prio from ioc
1086 * @rq: request of interest
1087 * @ioc: target iocontext
1089 * Assocate request prio with ioc prio so request based drivers
1090 * can leverage priority information.
1092 static inline void blk_rq_set_prio(struct request *rq, struct io_context *ioc)
1095 rq->ioprio = ioc->ioprio;
1099 * Request issue related functions.
1101 extern struct request *blk_peek_request(struct request_queue *q);
1102 extern void blk_start_request(struct request *rq);
1103 extern struct request *blk_fetch_request(struct request_queue *q);
1106 * Request completion related functions.
1108 * blk_update_request() completes given number of bytes and updates
1109 * the request without completing it.
1111 * blk_end_request() and friends. __blk_end_request() must be called
1112 * with the request queue spinlock acquired.
1114 * Several drivers define their own end_request and call
1115 * blk_end_request() for parts of the original function.
1116 * This prevents code duplication in drivers.
1118 extern bool blk_update_request(struct request *rq, int error,
1119 unsigned int nr_bytes);
1120 extern void blk_finish_request(struct request *rq, int error);
1121 extern bool blk_end_request(struct request *rq, int error,
1122 unsigned int nr_bytes);
1123 extern void blk_end_request_all(struct request *rq, int error);
1124 extern bool blk_end_request_cur(struct request *rq, int error);
1125 extern bool blk_end_request_err(struct request *rq, int error);
1126 extern bool __blk_end_request(struct request *rq, int error,
1127 unsigned int nr_bytes);
1128 extern void __blk_end_request_all(struct request *rq, int error);
1129 extern bool __blk_end_request_cur(struct request *rq, int error);
1130 extern bool __blk_end_request_err(struct request *rq, int error);
1132 extern void blk_complete_request(struct request *);
1133 extern void __blk_complete_request(struct request *);
1134 extern void blk_abort_request(struct request *);
1135 extern void blk_unprep_request(struct request *);
1138 * Access functions for manipulating queue properties
1140 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1141 spinlock_t *lock, int node_id);
1142 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1143 extern int blk_init_allocated_queue(struct request_queue *);
1144 extern void blk_cleanup_queue(struct request_queue *);
1145 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1146 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1147 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1148 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1149 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1150 extern void blk_queue_max_discard_segments(struct request_queue *,
1152 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1153 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1154 unsigned int max_discard_sectors);
1155 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1156 unsigned int max_write_same_sectors);
1157 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1158 unsigned int max_write_same_sectors);
1159 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1160 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1161 extern void blk_queue_alignment_offset(struct request_queue *q,
1162 unsigned int alignment);
1163 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1164 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1165 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1166 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1167 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1168 extern void blk_set_default_limits(struct queue_limits *lim);
1169 extern void blk_set_stacking_limits(struct queue_limits *lim);
1170 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1172 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1174 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1176 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1177 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1178 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1179 extern int blk_queue_dma_drain(struct request_queue *q,
1180 dma_drain_needed_fn *dma_drain_needed,
1181 void *buf, unsigned int size);
1182 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1183 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1184 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1185 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1186 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1187 extern void blk_queue_dma_alignment(struct request_queue *, int);
1188 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1189 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1190 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1191 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1192 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1193 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1196 * Number of physical segments as sent to the device.
1198 * Normally this is the number of discontiguous data segments sent by the
1199 * submitter. But for data-less command like discard we might have no
1200 * actual data segments submitted, but the driver might have to add it's
1201 * own special payload. In that case we still return 1 here so that this
1202 * special payload will be mapped.
1204 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1206 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1208 return rq->nr_phys_segments;
1212 * Number of discard segments (or ranges) the driver needs to fill in.
1213 * Each discard bio merged into a request is counted as one segment.
1215 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1217 return max_t(unsigned short, rq->nr_phys_segments, 1);
1220 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1221 extern void blk_dump_rq_flags(struct request *, char *);
1222 extern long nr_blockdev_pages(void);
1224 bool __must_check blk_get_queue(struct request_queue *);
1225 struct request_queue *blk_alloc_queue(gfp_t);
1226 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1227 extern void blk_put_queue(struct request_queue *);
1228 extern void blk_set_queue_dying(struct request_queue *);
1231 * block layer runtime pm functions
1234 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1235 extern int blk_pre_runtime_suspend(struct request_queue *q);
1236 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1237 extern void blk_pre_runtime_resume(struct request_queue *q);
1238 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1239 extern void blk_set_runtime_active(struct request_queue *q);
1241 static inline void blk_pm_runtime_init(struct request_queue *q,
1242 struct device *dev) {}
1243 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1247 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1248 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1249 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1250 static inline void blk_set_runtime_active(struct request_queue *q) {}
1254 * blk_plug permits building a queue of related requests by holding the I/O
1255 * fragments for a short period. This allows merging of sequential requests
1256 * into single larger request. As the requests are moved from a per-task list to
1257 * the device's request_queue in a batch, this results in improved scalability
1258 * as the lock contention for request_queue lock is reduced.
1260 * It is ok not to disable preemption when adding the request to the plug list
1261 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1262 * the plug list when the task sleeps by itself. For details, please see
1263 * schedule() where blk_schedule_flush_plug() is called.
1266 struct list_head list; /* requests */
1267 struct list_head mq_list; /* blk-mq requests */
1268 struct list_head cb_list; /* md requires an unplug callback */
1270 #define BLK_MAX_REQUEST_COUNT 16
1271 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1274 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1275 struct blk_plug_cb {
1276 struct list_head list;
1277 blk_plug_cb_fn callback;
1280 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1281 void *data, int size);
1282 extern void blk_start_plug(struct blk_plug *);
1283 extern void blk_finish_plug(struct blk_plug *);
1284 extern void blk_flush_plug_list(struct blk_plug *, bool);
1286 static inline void blk_flush_plug(struct task_struct *tsk)
1288 struct blk_plug *plug = tsk->plug;
1291 blk_flush_plug_list(plug, false);
1294 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1296 struct blk_plug *plug = tsk->plug;
1299 blk_flush_plug_list(plug, true);
1302 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1304 struct blk_plug *plug = tsk->plug;
1307 (!list_empty(&plug->list) ||
1308 !list_empty(&plug->mq_list) ||
1309 !list_empty(&plug->cb_list));
1315 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1316 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1317 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1318 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1319 extern void blk_queue_free_tags(struct request_queue *);
1320 extern int blk_queue_resize_tags(struct request_queue *, int);
1321 extern void blk_queue_invalidate_tags(struct request_queue *);
1322 extern struct blk_queue_tag *blk_init_tags(int, int);
1323 extern void blk_free_tags(struct blk_queue_tag *);
1325 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1328 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1330 return bqt->tag_index[tag];
1334 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1335 #define BLKDEV_DISCARD_ZERO (1 << 1) /* must reliably zero data */
1337 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1338 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1339 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1340 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1341 sector_t nr_sects, gfp_t gfp_mask, int flags,
1343 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1344 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1345 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1346 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1348 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1349 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1350 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1351 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1353 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1354 nr_blocks << (sb->s_blocksize_bits - 9),
1357 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1358 sector_t nr_blocks, gfp_t gfp_mask)
1360 return blkdev_issue_zeroout(sb->s_bdev,
1361 block << (sb->s_blocksize_bits - 9),
1362 nr_blocks << (sb->s_blocksize_bits - 9),
1366 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1368 enum blk_default_limits {
1369 BLK_MAX_SEGMENTS = 128,
1370 BLK_SAFE_MAX_SECTORS = 255,
1371 BLK_DEF_MAX_SECTORS = 2560,
1372 BLK_MAX_SEGMENT_SIZE = 65536,
1373 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1376 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1378 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1380 return q->limits.bounce_pfn;
1383 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1385 return q->limits.seg_boundary_mask;
1388 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1390 return q->limits.virt_boundary_mask;
1393 static inline unsigned int queue_max_sectors(struct request_queue *q)
1395 return q->limits.max_sectors;
1398 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1400 return q->limits.max_hw_sectors;
1403 static inline unsigned short queue_max_segments(struct request_queue *q)
1405 return q->limits.max_segments;
1408 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1410 return q->limits.max_discard_segments;
1413 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1415 return q->limits.max_segment_size;
1418 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1422 if (q && q->limits.logical_block_size)
1423 retval = q->limits.logical_block_size;
1428 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1430 return queue_logical_block_size(bdev_get_queue(bdev));
1433 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1435 return q->limits.physical_block_size;
1438 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1440 return queue_physical_block_size(bdev_get_queue(bdev));
1443 static inline unsigned int queue_io_min(struct request_queue *q)
1445 return q->limits.io_min;
1448 static inline int bdev_io_min(struct block_device *bdev)
1450 return queue_io_min(bdev_get_queue(bdev));
1453 static inline unsigned int queue_io_opt(struct request_queue *q)
1455 return q->limits.io_opt;
1458 static inline int bdev_io_opt(struct block_device *bdev)
1460 return queue_io_opt(bdev_get_queue(bdev));
1463 static inline int queue_alignment_offset(struct request_queue *q)
1465 if (q->limits.misaligned)
1468 return q->limits.alignment_offset;
1471 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1473 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1474 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1476 return (granularity + lim->alignment_offset - alignment) % granularity;
1479 static inline int bdev_alignment_offset(struct block_device *bdev)
1481 struct request_queue *q = bdev_get_queue(bdev);
1483 if (q->limits.misaligned)
1486 if (bdev != bdev->bd_contains)
1487 return bdev->bd_part->alignment_offset;
1489 return q->limits.alignment_offset;
1492 static inline int queue_discard_alignment(struct request_queue *q)
1494 if (q->limits.discard_misaligned)
1497 return q->limits.discard_alignment;
1500 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1502 unsigned int alignment, granularity, offset;
1504 if (!lim->max_discard_sectors)
1507 /* Why are these in bytes, not sectors? */
1508 alignment = lim->discard_alignment >> 9;
1509 granularity = lim->discard_granularity >> 9;
1513 /* Offset of the partition start in 'granularity' sectors */
1514 offset = sector_div(sector, granularity);
1516 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1517 offset = (granularity + alignment - offset) % granularity;
1519 /* Turn it back into bytes, gaah */
1523 static inline int bdev_discard_alignment(struct block_device *bdev)
1525 struct request_queue *q = bdev_get_queue(bdev);
1527 if (bdev != bdev->bd_contains)
1528 return bdev->bd_part->discard_alignment;
1530 return q->limits.discard_alignment;
1533 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1535 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1541 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1543 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1546 static inline unsigned int bdev_write_same(struct block_device *bdev)
1548 struct request_queue *q = bdev_get_queue(bdev);
1551 return q->limits.max_write_same_sectors;
1556 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1558 struct request_queue *q = bdev_get_queue(bdev);
1561 return q->limits.max_write_zeroes_sectors;
1566 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1568 struct request_queue *q = bdev_get_queue(bdev);
1571 return blk_queue_zoned_model(q);
1573 return BLK_ZONED_NONE;
1576 static inline bool bdev_is_zoned(struct block_device *bdev)
1578 struct request_queue *q = bdev_get_queue(bdev);
1581 return blk_queue_is_zoned(q);
1586 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1588 struct request_queue *q = bdev_get_queue(bdev);
1591 return blk_queue_zone_sectors(q);
1596 static inline int queue_dma_alignment(struct request_queue *q)
1598 return q ? q->dma_alignment : 511;
1601 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1604 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1605 return !(addr & alignment) && !(len & alignment);
1608 /* assumes size > 256 */
1609 static inline unsigned int blksize_bits(unsigned int size)
1611 unsigned int bits = 8;
1615 } while (size > 256);
1619 static inline unsigned int block_size(struct block_device *bdev)
1621 return bdev->bd_block_size;
1624 static inline bool queue_flush_queueable(struct request_queue *q)
1626 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1629 typedef struct {struct page *v;} Sector;
1631 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1633 static inline void put_dev_sector(Sector p)
1638 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1639 struct bio_vec *bprv, unsigned int offset)
1642 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1646 * Check if adding a bio_vec after bprv with offset would create a gap in
1647 * the SG list. Most drivers don't care about this, but some do.
1649 static inline bool bvec_gap_to_prev(struct request_queue *q,
1650 struct bio_vec *bprv, unsigned int offset)
1652 if (!queue_virt_boundary(q))
1654 return __bvec_gap_to_prev(q, bprv, offset);
1658 * Check if the two bvecs from two bios can be merged to one segment.
1659 * If yes, no need to check gap between the two bios since the 1st bio
1660 * and the 1st bvec in the 2nd bio can be handled in one segment.
1662 static inline bool bios_segs_mergeable(struct request_queue *q,
1663 struct bio *prev, struct bio_vec *prev_last_bv,
1664 struct bio_vec *next_first_bv)
1666 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1668 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1670 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1671 queue_max_segment_size(q))
1676 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1679 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1680 struct bio_vec pb, nb;
1682 bio_get_last_bvec(prev, &pb);
1683 bio_get_first_bvec(next, &nb);
1685 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1686 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1692 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1694 return bio_will_gap(req->q, req->biotail, bio);
1697 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1699 return bio_will_gap(req->q, bio, req->bio);
1702 int kblockd_schedule_work(struct work_struct *work);
1703 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1704 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1705 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1707 #ifdef CONFIG_BLK_CGROUP
1709 * This should not be using sched_clock(). A real patch is in progress
1710 * to fix this up, until that is in place we need to disable preemption
1711 * around sched_clock() in this function and set_io_start_time_ns().
1713 static inline void set_start_time_ns(struct request *req)
1716 req->start_time_ns = sched_clock();
1720 static inline void set_io_start_time_ns(struct request *req)
1723 req->io_start_time_ns = sched_clock();
1727 static inline uint64_t rq_start_time_ns(struct request *req)
1729 return req->start_time_ns;
1732 static inline uint64_t rq_io_start_time_ns(struct request *req)
1734 return req->io_start_time_ns;
1737 static inline void set_start_time_ns(struct request *req) {}
1738 static inline void set_io_start_time_ns(struct request *req) {}
1739 static inline uint64_t rq_start_time_ns(struct request *req)
1743 static inline uint64_t rq_io_start_time_ns(struct request *req)
1749 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1750 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1751 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1752 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1754 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1756 enum blk_integrity_flags {
1757 BLK_INTEGRITY_VERIFY = 1 << 0,
1758 BLK_INTEGRITY_GENERATE = 1 << 1,
1759 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1760 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1763 struct blk_integrity_iter {
1767 unsigned int data_size;
1768 unsigned short interval;
1769 const char *disk_name;
1772 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1774 struct blk_integrity_profile {
1775 integrity_processing_fn *generate_fn;
1776 integrity_processing_fn *verify_fn;
1780 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1781 extern void blk_integrity_unregister(struct gendisk *);
1782 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1783 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1784 struct scatterlist *);
1785 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1786 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1788 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1791 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1793 struct blk_integrity *bi = &disk->queue->integrity;
1802 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1804 return blk_get_integrity(bdev->bd_disk);
1807 static inline bool blk_integrity_rq(struct request *rq)
1809 return rq->cmd_flags & REQ_INTEGRITY;
1812 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1815 q->limits.max_integrity_segments = segs;
1818 static inline unsigned short
1819 queue_max_integrity_segments(struct request_queue *q)
1821 return q->limits.max_integrity_segments;
1824 static inline bool integrity_req_gap_back_merge(struct request *req,
1827 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1828 struct bio_integrity_payload *bip_next = bio_integrity(next);
1830 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1831 bip_next->bip_vec[0].bv_offset);
1834 static inline bool integrity_req_gap_front_merge(struct request *req,
1837 struct bio_integrity_payload *bip = bio_integrity(bio);
1838 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1840 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1841 bip_next->bip_vec[0].bv_offset);
1844 #else /* CONFIG_BLK_DEV_INTEGRITY */
1847 struct block_device;
1849 struct blk_integrity;
1851 static inline int blk_integrity_rq(struct request *rq)
1855 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1860 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1862 struct scatterlist *s)
1866 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1870 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1874 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1878 static inline void blk_integrity_register(struct gendisk *d,
1879 struct blk_integrity *b)
1882 static inline void blk_integrity_unregister(struct gendisk *d)
1885 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1889 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1893 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1899 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1906 static inline bool integrity_req_gap_back_merge(struct request *req,
1911 static inline bool integrity_req_gap_front_merge(struct request *req,
1917 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1920 * struct blk_dax_ctl - control and output parameters for ->direct_access
1921 * @sector: (input) offset relative to a block_device
1922 * @addr: (output) kernel virtual address for @sector populated by driver
1923 * @pfn: (output) page frame number for @addr populated by driver
1924 * @size: (input) number of bytes requested
1926 struct blk_dax_ctl {
1933 struct block_device_operations {
1934 int (*open) (struct block_device *, fmode_t);
1935 void (*release) (struct gendisk *, fmode_t);
1936 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1937 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1938 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1939 long (*direct_access)(struct block_device *, sector_t, void **, pfn_t *,
1941 unsigned int (*check_events) (struct gendisk *disk,
1942 unsigned int clearing);
1943 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1944 int (*media_changed) (struct gendisk *);
1945 void (*unlock_native_capacity) (struct gendisk *);
1946 int (*revalidate_disk) (struct gendisk *);
1947 int (*getgeo)(struct block_device *, struct hd_geometry *);
1948 /* this callback is with swap_lock and sometimes page table lock held */
1949 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1950 struct module *owner;
1951 const struct pr_ops *pr_ops;
1954 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1956 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1957 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1958 struct writeback_control *);
1959 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1960 extern int bdev_dax_supported(struct super_block *, int);
1961 extern bool bdev_dax_capable(struct block_device *);
1962 #else /* CONFIG_BLOCK */
1964 struct block_device;
1967 * stubs for when the block layer is configured out
1969 #define buffer_heads_over_limit 0
1971 static inline long nr_blockdev_pages(void)
1979 static inline void blk_start_plug(struct blk_plug *plug)
1983 static inline void blk_finish_plug(struct blk_plug *plug)
1987 static inline void blk_flush_plug(struct task_struct *task)
1991 static inline void blk_schedule_flush_plug(struct task_struct *task)
1996 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2001 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2002 sector_t *error_sector)
2007 #endif /* CONFIG_BLOCK */