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;
438 struct disk_devt *disk_devt;
441 * The queue owner gets to use this for whatever they like.
442 * ll_rw_blk doesn't touch it.
447 * various queue flags, see QUEUE_* below
449 unsigned long queue_flags;
452 * ida allocated id for this queue. Used to index queues from
458 * queue needs bounce pages for pages above this limit
463 * protects queue structures from reentrancy. ->__queue_lock should
464 * _never_ be used directly, it is queue private. always use
467 spinlock_t __queue_lock;
468 spinlock_t *queue_lock;
478 struct kobject mq_kobj;
480 #ifdef CONFIG_BLK_DEV_INTEGRITY
481 struct blk_integrity integrity;
482 #endif /* CONFIG_BLK_DEV_INTEGRITY */
487 unsigned int nr_pending;
493 unsigned long nr_requests; /* Max # of requests */
494 unsigned int nr_congestion_on;
495 unsigned int nr_congestion_off;
496 unsigned int nr_batching;
498 unsigned int dma_drain_size;
499 void *dma_drain_buffer;
500 unsigned int dma_pad_mask;
501 unsigned int dma_alignment;
503 struct blk_queue_tag *queue_tags;
504 struct list_head tag_busy_list;
506 unsigned int nr_sorted;
507 unsigned int in_flight[2];
509 struct blk_rq_stat rq_stats[2];
512 * Number of active block driver functions for which blk_drain_queue()
513 * must wait. Must be incremented around functions that unlock the
514 * queue_lock internally, e.g. scsi_request_fn().
516 unsigned int request_fn_active;
518 unsigned int rq_timeout;
520 struct timer_list timeout;
521 struct work_struct timeout_work;
522 struct list_head timeout_list;
524 struct list_head icq_list;
525 #ifdef CONFIG_BLK_CGROUP
526 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
527 struct blkcg_gq *root_blkg;
528 struct list_head blkg_list;
531 struct queue_limits limits;
536 unsigned int sg_timeout;
537 unsigned int sg_reserved_size;
539 #ifdef CONFIG_BLK_DEV_IO_TRACE
540 struct blk_trace *blk_trace;
543 * for flush operations
545 struct blk_flush_queue *fq;
547 struct list_head requeue_list;
548 spinlock_t requeue_lock;
549 struct delayed_work requeue_work;
551 struct mutex sysfs_lock;
554 atomic_t mq_freeze_depth;
556 #if defined(CONFIG_BLK_DEV_BSG)
557 bsg_job_fn *bsg_job_fn;
559 struct bsg_class_device bsg_dev;
562 #ifdef CONFIG_BLK_DEV_THROTTLING
564 struct throtl_data *td;
566 struct rcu_head rcu_head;
567 wait_queue_head_t mq_freeze_wq;
568 struct percpu_ref q_usage_counter;
569 struct list_head all_q_node;
571 struct blk_mq_tag_set *tag_set;
572 struct list_head tag_set_list;
573 struct bio_set *bio_split;
575 #ifdef CONFIG_BLK_DEBUG_FS
576 struct dentry *debugfs_dir;
577 struct dentry *mq_debugfs_dir;
580 bool mq_sysfs_init_done;
586 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
587 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
588 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
589 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
590 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
591 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
592 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
593 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
594 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
595 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
596 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
597 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
598 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
599 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
600 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
601 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
602 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
603 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
604 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
605 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
606 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
607 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
608 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
609 #define QUEUE_FLAG_WC 23 /* Write back caching */
610 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
611 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
612 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
613 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
614 #define QUEUE_FLAG_RESTART 28 /* queue needs restart at completion */
616 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
617 (1 << QUEUE_FLAG_STACKABLE) | \
618 (1 << QUEUE_FLAG_SAME_COMP) | \
619 (1 << QUEUE_FLAG_ADD_RANDOM))
621 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
622 (1 << QUEUE_FLAG_STACKABLE) | \
623 (1 << QUEUE_FLAG_SAME_COMP) | \
624 (1 << QUEUE_FLAG_POLL))
626 static inline void queue_lockdep_assert_held(struct request_queue *q)
629 lockdep_assert_held(q->queue_lock);
632 static inline void queue_flag_set_unlocked(unsigned int flag,
633 struct request_queue *q)
635 __set_bit(flag, &q->queue_flags);
638 static inline int queue_flag_test_and_clear(unsigned int flag,
639 struct request_queue *q)
641 queue_lockdep_assert_held(q);
643 if (test_bit(flag, &q->queue_flags)) {
644 __clear_bit(flag, &q->queue_flags);
651 static inline int queue_flag_test_and_set(unsigned int flag,
652 struct request_queue *q)
654 queue_lockdep_assert_held(q);
656 if (!test_bit(flag, &q->queue_flags)) {
657 __set_bit(flag, &q->queue_flags);
664 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
666 queue_lockdep_assert_held(q);
667 __set_bit(flag, &q->queue_flags);
670 static inline void queue_flag_clear_unlocked(unsigned int flag,
671 struct request_queue *q)
673 __clear_bit(flag, &q->queue_flags);
676 static inline int queue_in_flight(struct request_queue *q)
678 return q->in_flight[0] + q->in_flight[1];
681 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
683 queue_lockdep_assert_held(q);
684 __clear_bit(flag, &q->queue_flags);
687 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
688 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
689 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
690 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
691 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
692 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
693 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
694 #define blk_queue_noxmerges(q) \
695 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
696 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
697 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
698 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
699 #define blk_queue_stackable(q) \
700 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
701 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
702 #define blk_queue_secure_erase(q) \
703 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
704 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
706 #define blk_noretry_request(rq) \
707 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
708 REQ_FAILFAST_DRIVER))
710 static inline bool blk_account_rq(struct request *rq)
712 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
715 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
716 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
717 /* rq->queuelist of dequeued request must be list_empty() */
718 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
720 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
722 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
725 * Driver can handle struct request, if it either has an old style
726 * request_fn defined, or is blk-mq based.
728 static inline bool queue_is_rq_based(struct request_queue *q)
730 return q->request_fn || q->mq_ops;
733 static inline unsigned int blk_queue_cluster(struct request_queue *q)
735 return q->limits.cluster;
738 static inline enum blk_zoned_model
739 blk_queue_zoned_model(struct request_queue *q)
741 return q->limits.zoned;
744 static inline bool blk_queue_is_zoned(struct request_queue *q)
746 switch (blk_queue_zoned_model(q)) {
755 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
757 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
760 static inline bool rq_is_sync(struct request *rq)
762 return op_is_sync(rq->cmd_flags);
765 static inline bool blk_rl_full(struct request_list *rl, bool sync)
767 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
769 return rl->flags & flag;
772 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
774 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
779 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
781 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
786 static inline bool rq_mergeable(struct request *rq)
788 if (blk_rq_is_passthrough(rq))
791 if (req_op(rq) == REQ_OP_FLUSH)
794 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
797 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
799 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
805 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
807 if (bio_data(a) == bio_data(b))
813 static inline unsigned int blk_queue_depth(struct request_queue *q)
816 return q->queue_depth;
818 return q->nr_requests;
822 * q->prep_rq_fn return values
825 BLKPREP_OK, /* serve it */
826 BLKPREP_KILL, /* fatal error, kill, return -EIO */
827 BLKPREP_DEFER, /* leave on queue */
828 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
831 extern unsigned long blk_max_low_pfn, blk_max_pfn;
834 * standard bounce addresses:
836 * BLK_BOUNCE_HIGH : bounce all highmem pages
837 * BLK_BOUNCE_ANY : don't bounce anything
838 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
841 #if BITS_PER_LONG == 32
842 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
844 #define BLK_BOUNCE_HIGH -1ULL
846 #define BLK_BOUNCE_ANY (-1ULL)
847 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
850 * default timeout for SG_IO if none specified
852 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
853 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
856 extern int init_emergency_isa_pool(void);
857 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
859 static inline int init_emergency_isa_pool(void)
863 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
866 #endif /* CONFIG_MMU */
872 unsigned long offset;
877 struct req_iterator {
878 struct bvec_iter iter;
882 /* This should not be used directly - use rq_for_each_segment */
883 #define for_each_bio(_bio) \
884 for (; _bio; _bio = _bio->bi_next)
885 #define __rq_for_each_bio(_bio, rq) \
887 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
889 #define rq_for_each_segment(bvl, _rq, _iter) \
890 __rq_for_each_bio(_iter.bio, _rq) \
891 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
893 #define rq_iter_last(bvec, _iter) \
894 (_iter.bio->bi_next == NULL && \
895 bio_iter_last(bvec, _iter.iter))
897 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
898 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
900 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
901 extern void rq_flush_dcache_pages(struct request *rq);
903 static inline void rq_flush_dcache_pages(struct request *rq)
909 #define vfs_msg(sb, level, fmt, ...) \
910 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
912 #define vfs_msg(sb, level, fmt, ...) \
914 no_printk(fmt, ##__VA_ARGS__); \
915 __vfs_msg(sb, "", " "); \
919 extern int blk_register_queue(struct gendisk *disk);
920 extern void blk_unregister_queue(struct gendisk *disk);
921 extern blk_qc_t generic_make_request(struct bio *bio);
922 extern void blk_rq_init(struct request_queue *q, struct request *rq);
923 extern void blk_put_request(struct request *);
924 extern void __blk_put_request(struct request_queue *, struct request *);
925 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
926 extern void blk_requeue_request(struct request_queue *, struct request *);
927 extern int blk_lld_busy(struct request_queue *q);
928 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
929 struct bio_set *bs, gfp_t gfp_mask,
930 int (*bio_ctr)(struct bio *, struct bio *, void *),
932 extern void blk_rq_unprep_clone(struct request *rq);
933 extern int blk_insert_cloned_request(struct request_queue *q,
935 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
936 extern void blk_delay_queue(struct request_queue *, unsigned long);
937 extern void blk_queue_split(struct request_queue *, struct bio **,
939 extern void blk_recount_segments(struct request_queue *, struct bio *);
940 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
941 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
942 unsigned int, void __user *);
943 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
944 unsigned int, void __user *);
945 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
946 struct scsi_ioctl_command __user *);
948 extern int blk_queue_enter(struct request_queue *q, bool nowait);
949 extern void blk_queue_exit(struct request_queue *q);
950 extern void blk_start_queue(struct request_queue *q);
951 extern void blk_start_queue_async(struct request_queue *q);
952 extern void blk_stop_queue(struct request_queue *q);
953 extern void blk_sync_queue(struct request_queue *q);
954 extern void __blk_stop_queue(struct request_queue *q);
955 extern void __blk_run_queue(struct request_queue *q);
956 extern void __blk_run_queue_uncond(struct request_queue *q);
957 extern void blk_run_queue(struct request_queue *);
958 extern void blk_run_queue_async(struct request_queue *q);
959 extern void blk_mq_quiesce_queue(struct request_queue *q);
960 extern int blk_rq_map_user(struct request_queue *, struct request *,
961 struct rq_map_data *, void __user *, unsigned long,
963 extern int blk_rq_unmap_user(struct bio *);
964 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
965 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
966 struct rq_map_data *, const struct iov_iter *,
968 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
969 struct request *, int);
970 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
971 struct request *, int, rq_end_io_fn *);
973 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
975 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
977 return bdev->bd_disk->queue; /* this is never NULL */
981 * blk_rq_pos() : the current sector
982 * blk_rq_bytes() : bytes left in the entire request
983 * blk_rq_cur_bytes() : bytes left in the current segment
984 * blk_rq_err_bytes() : bytes left till the next error boundary
985 * blk_rq_sectors() : sectors left in the entire request
986 * blk_rq_cur_sectors() : sectors left in the current segment
988 static inline sector_t blk_rq_pos(const struct request *rq)
993 static inline unsigned int blk_rq_bytes(const struct request *rq)
995 return rq->__data_len;
998 static inline int blk_rq_cur_bytes(const struct request *rq)
1000 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1003 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1005 static inline unsigned int blk_rq_sectors(const struct request *rq)
1007 return blk_rq_bytes(rq) >> 9;
1010 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1012 return blk_rq_cur_bytes(rq) >> 9;
1016 * Some commands like WRITE SAME have a payload or data transfer size which
1017 * is different from the size of the request. Any driver that supports such
1018 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1019 * calculate the data transfer size.
1021 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1023 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1024 return rq->special_vec.bv_len;
1025 return blk_rq_bytes(rq);
1028 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1031 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1032 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1034 if (unlikely(op == REQ_OP_WRITE_SAME))
1035 return q->limits.max_write_same_sectors;
1037 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1038 return q->limits.max_write_zeroes_sectors;
1040 return q->limits.max_sectors;
1044 * Return maximum size of a request at given offset. Only valid for
1045 * file system requests.
1047 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1050 if (!q->limits.chunk_sectors)
1051 return q->limits.max_sectors;
1053 return q->limits.chunk_sectors -
1054 (offset & (q->limits.chunk_sectors - 1));
1057 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1060 struct request_queue *q = rq->q;
1062 if (blk_rq_is_passthrough(rq))
1063 return q->limits.max_hw_sectors;
1065 if (!q->limits.chunk_sectors ||
1066 req_op(rq) == REQ_OP_DISCARD ||
1067 req_op(rq) == REQ_OP_SECURE_ERASE)
1068 return blk_queue_get_max_sectors(q, req_op(rq));
1070 return min(blk_max_size_offset(q, offset),
1071 blk_queue_get_max_sectors(q, req_op(rq)));
1074 static inline unsigned int blk_rq_count_bios(struct request *rq)
1076 unsigned int nr_bios = 0;
1079 __rq_for_each_bio(bio, rq)
1086 * blk_rq_set_prio - associate a request with prio from ioc
1087 * @rq: request of interest
1088 * @ioc: target iocontext
1090 * Assocate request prio with ioc prio so request based drivers
1091 * can leverage priority information.
1093 static inline void blk_rq_set_prio(struct request *rq, struct io_context *ioc)
1096 rq->ioprio = ioc->ioprio;
1100 * Request issue related functions.
1102 extern struct request *blk_peek_request(struct request_queue *q);
1103 extern void blk_start_request(struct request *rq);
1104 extern struct request *blk_fetch_request(struct request_queue *q);
1107 * Request completion related functions.
1109 * blk_update_request() completes given number of bytes and updates
1110 * the request without completing it.
1112 * blk_end_request() and friends. __blk_end_request() must be called
1113 * with the request queue spinlock acquired.
1115 * Several drivers define their own end_request and call
1116 * blk_end_request() for parts of the original function.
1117 * This prevents code duplication in drivers.
1119 extern bool blk_update_request(struct request *rq, int error,
1120 unsigned int nr_bytes);
1121 extern void blk_finish_request(struct request *rq, int error);
1122 extern bool blk_end_request(struct request *rq, int error,
1123 unsigned int nr_bytes);
1124 extern void blk_end_request_all(struct request *rq, int error);
1125 extern bool blk_end_request_cur(struct request *rq, int error);
1126 extern bool blk_end_request_err(struct request *rq, int error);
1127 extern bool __blk_end_request(struct request *rq, int error,
1128 unsigned int nr_bytes);
1129 extern void __blk_end_request_all(struct request *rq, int error);
1130 extern bool __blk_end_request_cur(struct request *rq, int error);
1131 extern bool __blk_end_request_err(struct request *rq, int error);
1133 extern void blk_complete_request(struct request *);
1134 extern void __blk_complete_request(struct request *);
1135 extern void blk_abort_request(struct request *);
1136 extern void blk_unprep_request(struct request *);
1139 * Access functions for manipulating queue properties
1141 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1142 spinlock_t *lock, int node_id);
1143 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1144 extern int blk_init_allocated_queue(struct request_queue *);
1145 extern void blk_cleanup_queue(struct request_queue *);
1146 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1147 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1148 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1149 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1150 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1151 extern void blk_queue_max_discard_segments(struct request_queue *,
1153 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1154 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1155 unsigned int max_discard_sectors);
1156 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1157 unsigned int max_write_same_sectors);
1158 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1159 unsigned int max_write_same_sectors);
1160 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1161 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1162 extern void blk_queue_alignment_offset(struct request_queue *q,
1163 unsigned int alignment);
1164 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1165 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1166 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1167 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1168 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1169 extern void blk_set_default_limits(struct queue_limits *lim);
1170 extern void blk_set_stacking_limits(struct queue_limits *lim);
1171 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1173 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1175 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1177 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1178 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1179 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1180 extern int blk_queue_dma_drain(struct request_queue *q,
1181 dma_drain_needed_fn *dma_drain_needed,
1182 void *buf, unsigned int size);
1183 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1184 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1185 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1186 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1187 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1188 extern void blk_queue_dma_alignment(struct request_queue *, int);
1189 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1190 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1191 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1192 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1193 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1194 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1197 * Number of physical segments as sent to the device.
1199 * Normally this is the number of discontiguous data segments sent by the
1200 * submitter. But for data-less command like discard we might have no
1201 * actual data segments submitted, but the driver might have to add it's
1202 * own special payload. In that case we still return 1 here so that this
1203 * special payload will be mapped.
1205 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1207 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1209 return rq->nr_phys_segments;
1213 * Number of discard segments (or ranges) the driver needs to fill in.
1214 * Each discard bio merged into a request is counted as one segment.
1216 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1218 return max_t(unsigned short, rq->nr_phys_segments, 1);
1221 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1222 extern void blk_dump_rq_flags(struct request *, char *);
1223 extern long nr_blockdev_pages(void);
1225 bool __must_check blk_get_queue(struct request_queue *);
1226 struct request_queue *blk_alloc_queue(gfp_t);
1227 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1228 extern void blk_put_queue(struct request_queue *);
1229 extern void blk_set_queue_dying(struct request_queue *);
1232 * block layer runtime pm functions
1235 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1236 extern int blk_pre_runtime_suspend(struct request_queue *q);
1237 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1238 extern void blk_pre_runtime_resume(struct request_queue *q);
1239 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1240 extern void blk_set_runtime_active(struct request_queue *q);
1242 static inline void blk_pm_runtime_init(struct request_queue *q,
1243 struct device *dev) {}
1244 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1248 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1249 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1250 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1251 static inline void blk_set_runtime_active(struct request_queue *q) {}
1255 * blk_plug permits building a queue of related requests by holding the I/O
1256 * fragments for a short period. This allows merging of sequential requests
1257 * into single larger request. As the requests are moved from a per-task list to
1258 * the device's request_queue in a batch, this results in improved scalability
1259 * as the lock contention for request_queue lock is reduced.
1261 * It is ok not to disable preemption when adding the request to the plug list
1262 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1263 * the plug list when the task sleeps by itself. For details, please see
1264 * schedule() where blk_schedule_flush_plug() is called.
1267 struct list_head list; /* requests */
1268 struct list_head mq_list; /* blk-mq requests */
1269 struct list_head cb_list; /* md requires an unplug callback */
1271 #define BLK_MAX_REQUEST_COUNT 16
1272 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1275 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1276 struct blk_plug_cb {
1277 struct list_head list;
1278 blk_plug_cb_fn callback;
1281 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1282 void *data, int size);
1283 extern void blk_start_plug(struct blk_plug *);
1284 extern void blk_finish_plug(struct blk_plug *);
1285 extern void blk_flush_plug_list(struct blk_plug *, bool);
1287 static inline void blk_flush_plug(struct task_struct *tsk)
1289 struct blk_plug *plug = tsk->plug;
1292 blk_flush_plug_list(plug, false);
1295 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1297 struct blk_plug *plug = tsk->plug;
1300 blk_flush_plug_list(plug, true);
1303 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1305 struct blk_plug *plug = tsk->plug;
1308 (!list_empty(&plug->list) ||
1309 !list_empty(&plug->mq_list) ||
1310 !list_empty(&plug->cb_list));
1316 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1317 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1318 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1319 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1320 extern void blk_queue_free_tags(struct request_queue *);
1321 extern int blk_queue_resize_tags(struct request_queue *, int);
1322 extern void blk_queue_invalidate_tags(struct request_queue *);
1323 extern struct blk_queue_tag *blk_init_tags(int, int);
1324 extern void blk_free_tags(struct blk_queue_tag *);
1326 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1329 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1331 return bqt->tag_index[tag];
1335 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1336 #define BLKDEV_DISCARD_ZERO (1 << 1) /* must reliably zero data */
1338 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1339 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1340 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1341 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1342 sector_t nr_sects, gfp_t gfp_mask, int flags,
1344 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1345 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1346 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1347 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1349 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1350 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1351 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1352 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1354 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1355 nr_blocks << (sb->s_blocksize_bits - 9),
1358 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1359 sector_t nr_blocks, gfp_t gfp_mask)
1361 return blkdev_issue_zeroout(sb->s_bdev,
1362 block << (sb->s_blocksize_bits - 9),
1363 nr_blocks << (sb->s_blocksize_bits - 9),
1367 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1369 enum blk_default_limits {
1370 BLK_MAX_SEGMENTS = 128,
1371 BLK_SAFE_MAX_SECTORS = 255,
1372 BLK_DEF_MAX_SECTORS = 2560,
1373 BLK_MAX_SEGMENT_SIZE = 65536,
1374 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1377 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1379 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1381 return q->limits.bounce_pfn;
1384 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1386 return q->limits.seg_boundary_mask;
1389 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1391 return q->limits.virt_boundary_mask;
1394 static inline unsigned int queue_max_sectors(struct request_queue *q)
1396 return q->limits.max_sectors;
1399 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1401 return q->limits.max_hw_sectors;
1404 static inline unsigned short queue_max_segments(struct request_queue *q)
1406 return q->limits.max_segments;
1409 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1411 return q->limits.max_discard_segments;
1414 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1416 return q->limits.max_segment_size;
1419 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1423 if (q && q->limits.logical_block_size)
1424 retval = q->limits.logical_block_size;
1429 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1431 return queue_logical_block_size(bdev_get_queue(bdev));
1434 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1436 return q->limits.physical_block_size;
1439 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1441 return queue_physical_block_size(bdev_get_queue(bdev));
1444 static inline unsigned int queue_io_min(struct request_queue *q)
1446 return q->limits.io_min;
1449 static inline int bdev_io_min(struct block_device *bdev)
1451 return queue_io_min(bdev_get_queue(bdev));
1454 static inline unsigned int queue_io_opt(struct request_queue *q)
1456 return q->limits.io_opt;
1459 static inline int bdev_io_opt(struct block_device *bdev)
1461 return queue_io_opt(bdev_get_queue(bdev));
1464 static inline int queue_alignment_offset(struct request_queue *q)
1466 if (q->limits.misaligned)
1469 return q->limits.alignment_offset;
1472 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1474 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1475 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1477 return (granularity + lim->alignment_offset - alignment) % granularity;
1480 static inline int bdev_alignment_offset(struct block_device *bdev)
1482 struct request_queue *q = bdev_get_queue(bdev);
1484 if (q->limits.misaligned)
1487 if (bdev != bdev->bd_contains)
1488 return bdev->bd_part->alignment_offset;
1490 return q->limits.alignment_offset;
1493 static inline int queue_discard_alignment(struct request_queue *q)
1495 if (q->limits.discard_misaligned)
1498 return q->limits.discard_alignment;
1501 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1503 unsigned int alignment, granularity, offset;
1505 if (!lim->max_discard_sectors)
1508 /* Why are these in bytes, not sectors? */
1509 alignment = lim->discard_alignment >> 9;
1510 granularity = lim->discard_granularity >> 9;
1514 /* Offset of the partition start in 'granularity' sectors */
1515 offset = sector_div(sector, granularity);
1517 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1518 offset = (granularity + alignment - offset) % granularity;
1520 /* Turn it back into bytes, gaah */
1524 static inline int bdev_discard_alignment(struct block_device *bdev)
1526 struct request_queue *q = bdev_get_queue(bdev);
1528 if (bdev != bdev->bd_contains)
1529 return bdev->bd_part->discard_alignment;
1531 return q->limits.discard_alignment;
1534 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1536 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1542 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1544 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1547 static inline unsigned int bdev_write_same(struct block_device *bdev)
1549 struct request_queue *q = bdev_get_queue(bdev);
1552 return q->limits.max_write_same_sectors;
1557 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1559 struct request_queue *q = bdev_get_queue(bdev);
1562 return q->limits.max_write_zeroes_sectors;
1567 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1569 struct request_queue *q = bdev_get_queue(bdev);
1572 return blk_queue_zoned_model(q);
1574 return BLK_ZONED_NONE;
1577 static inline bool bdev_is_zoned(struct block_device *bdev)
1579 struct request_queue *q = bdev_get_queue(bdev);
1582 return blk_queue_is_zoned(q);
1587 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1589 struct request_queue *q = bdev_get_queue(bdev);
1592 return blk_queue_zone_sectors(q);
1597 static inline int queue_dma_alignment(struct request_queue *q)
1599 return q ? q->dma_alignment : 511;
1602 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1605 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1606 return !(addr & alignment) && !(len & alignment);
1609 /* assumes size > 256 */
1610 static inline unsigned int blksize_bits(unsigned int size)
1612 unsigned int bits = 8;
1616 } while (size > 256);
1620 static inline unsigned int block_size(struct block_device *bdev)
1622 return bdev->bd_block_size;
1625 static inline bool queue_flush_queueable(struct request_queue *q)
1627 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1630 typedef struct {struct page *v;} Sector;
1632 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1634 static inline void put_dev_sector(Sector p)
1639 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1640 struct bio_vec *bprv, unsigned int offset)
1643 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1647 * Check if adding a bio_vec after bprv with offset would create a gap in
1648 * the SG list. Most drivers don't care about this, but some do.
1650 static inline bool bvec_gap_to_prev(struct request_queue *q,
1651 struct bio_vec *bprv, unsigned int offset)
1653 if (!queue_virt_boundary(q))
1655 return __bvec_gap_to_prev(q, bprv, offset);
1659 * Check if the two bvecs from two bios can be merged to one segment.
1660 * If yes, no need to check gap between the two bios since the 1st bio
1661 * and the 1st bvec in the 2nd bio can be handled in one segment.
1663 static inline bool bios_segs_mergeable(struct request_queue *q,
1664 struct bio *prev, struct bio_vec *prev_last_bv,
1665 struct bio_vec *next_first_bv)
1667 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1669 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1671 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1672 queue_max_segment_size(q))
1677 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1680 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1681 struct bio_vec pb, nb;
1683 bio_get_last_bvec(prev, &pb);
1684 bio_get_first_bvec(next, &nb);
1686 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1687 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1693 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1695 return bio_will_gap(req->q, req->biotail, bio);
1698 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1700 return bio_will_gap(req->q, bio, req->bio);
1703 int kblockd_schedule_work(struct work_struct *work);
1704 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1705 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1706 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1708 #ifdef CONFIG_BLK_CGROUP
1710 * This should not be using sched_clock(). A real patch is in progress
1711 * to fix this up, until that is in place we need to disable preemption
1712 * around sched_clock() in this function and set_io_start_time_ns().
1714 static inline void set_start_time_ns(struct request *req)
1717 req->start_time_ns = sched_clock();
1721 static inline void set_io_start_time_ns(struct request *req)
1724 req->io_start_time_ns = sched_clock();
1728 static inline uint64_t rq_start_time_ns(struct request *req)
1730 return req->start_time_ns;
1733 static inline uint64_t rq_io_start_time_ns(struct request *req)
1735 return req->io_start_time_ns;
1738 static inline void set_start_time_ns(struct request *req) {}
1739 static inline void set_io_start_time_ns(struct request *req) {}
1740 static inline uint64_t rq_start_time_ns(struct request *req)
1744 static inline uint64_t rq_io_start_time_ns(struct request *req)
1750 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1751 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1752 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1753 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1755 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1757 enum blk_integrity_flags {
1758 BLK_INTEGRITY_VERIFY = 1 << 0,
1759 BLK_INTEGRITY_GENERATE = 1 << 1,
1760 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1761 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1764 struct blk_integrity_iter {
1768 unsigned int data_size;
1769 unsigned short interval;
1770 const char *disk_name;
1773 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1775 struct blk_integrity_profile {
1776 integrity_processing_fn *generate_fn;
1777 integrity_processing_fn *verify_fn;
1781 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1782 extern void blk_integrity_unregister(struct gendisk *);
1783 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1784 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1785 struct scatterlist *);
1786 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1787 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1789 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1792 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1794 struct blk_integrity *bi = &disk->queue->integrity;
1803 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1805 return blk_get_integrity(bdev->bd_disk);
1808 static inline bool blk_integrity_rq(struct request *rq)
1810 return rq->cmd_flags & REQ_INTEGRITY;
1813 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1816 q->limits.max_integrity_segments = segs;
1819 static inline unsigned short
1820 queue_max_integrity_segments(struct request_queue *q)
1822 return q->limits.max_integrity_segments;
1825 static inline bool integrity_req_gap_back_merge(struct request *req,
1828 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1829 struct bio_integrity_payload *bip_next = bio_integrity(next);
1831 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1832 bip_next->bip_vec[0].bv_offset);
1835 static inline bool integrity_req_gap_front_merge(struct request *req,
1838 struct bio_integrity_payload *bip = bio_integrity(bio);
1839 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1841 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1842 bip_next->bip_vec[0].bv_offset);
1845 #else /* CONFIG_BLK_DEV_INTEGRITY */
1848 struct block_device;
1850 struct blk_integrity;
1852 static inline int blk_integrity_rq(struct request *rq)
1856 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1861 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1863 struct scatterlist *s)
1867 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1871 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1875 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1879 static inline void blk_integrity_register(struct gendisk *d,
1880 struct blk_integrity *b)
1883 static inline void blk_integrity_unregister(struct gendisk *d)
1886 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1890 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1894 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1900 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1907 static inline bool integrity_req_gap_back_merge(struct request *req,
1912 static inline bool integrity_req_gap_front_merge(struct request *req,
1918 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1921 * struct blk_dax_ctl - control and output parameters for ->direct_access
1922 * @sector: (input) offset relative to a block_device
1923 * @addr: (output) kernel virtual address for @sector populated by driver
1924 * @pfn: (output) page frame number for @addr populated by driver
1925 * @size: (input) number of bytes requested
1927 struct blk_dax_ctl {
1934 struct block_device_operations {
1935 int (*open) (struct block_device *, fmode_t);
1936 void (*release) (struct gendisk *, fmode_t);
1937 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1938 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1939 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1940 long (*direct_access)(struct block_device *, sector_t, void **, pfn_t *,
1942 unsigned int (*check_events) (struct gendisk *disk,
1943 unsigned int clearing);
1944 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1945 int (*media_changed) (struct gendisk *);
1946 void (*unlock_native_capacity) (struct gendisk *);
1947 int (*revalidate_disk) (struct gendisk *);
1948 int (*getgeo)(struct block_device *, struct hd_geometry *);
1949 /* this callback is with swap_lock and sometimes page table lock held */
1950 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1951 struct module *owner;
1952 const struct pr_ops *pr_ops;
1955 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1957 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1958 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1959 struct writeback_control *);
1960 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1961 extern int bdev_dax_supported(struct super_block *, int);
1962 extern bool bdev_dax_capable(struct block_device *);
1963 #else /* CONFIG_BLOCK */
1965 struct block_device;
1968 * stubs for when the block layer is configured out
1970 #define buffer_heads_over_limit 0
1972 static inline long nr_blockdev_pages(void)
1980 static inline void blk_start_plug(struct blk_plug *plug)
1984 static inline void blk_finish_plug(struct blk_plug *plug)
1988 static inline void blk_flush_plug(struct task_struct *task)
1992 static inline void blk_schedule_flush_plug(struct task_struct *task)
1997 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2002 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2003 sector_t *error_sector)
2008 #endif /* CONFIG_BLOCK */