1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
4 #include <linux/sched.h>
8 #include <linux/major.h>
9 #include <linux/genhd.h>
10 #include <linux/list.h>
11 #include <linux/llist.h>
12 #include <linux/timer.h>
13 #include <linux/workqueue.h>
14 #include <linux/pagemap.h>
15 #include <linux/backing-dev-defs.h>
16 #include <linux/wait.h>
17 #include <linux/mempool.h>
18 #include <linux/pfn.h>
19 #include <linux/bio.h>
20 #include <linux/stringify.h>
21 #include <linux/gfp.h>
22 #include <linux/bsg.h>
23 #include <linux/smp.h>
24 #include <linux/rcupdate.h>
25 #include <linux/percpu-refcount.h>
26 #include <linux/scatterlist.h>
27 #include <linux/blkzoned.h>
30 struct scsi_ioctl_command;
33 struct elevator_queue;
39 struct blk_flush_queue;
43 #define BLKDEV_MIN_RQ 4
44 #define BLKDEV_MAX_RQ 128 /* Default maximum */
47 * Maximum number of blkcg policies allowed to be registered concurrently.
48 * Defined here to simplify include dependency.
50 #define BLKCG_MAX_POLS 2
52 typedef void (rq_end_io_fn)(struct request *, int);
54 #define BLK_RL_SYNCFULL (1U << 0)
55 #define BLK_RL_ASYNCFULL (1U << 1)
58 struct request_queue *q; /* the queue this rl belongs to */
59 #ifdef CONFIG_BLK_CGROUP
60 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
63 * count[], starved[], and wait[] are indexed by
64 * BLK_RW_SYNC/BLK_RW_ASYNC
69 wait_queue_head_t wait[2];
74 * request command types
76 enum rq_cmd_type_bits {
77 REQ_TYPE_FS = 1, /* fs request */
78 REQ_TYPE_BLOCK_PC, /* scsi command */
79 REQ_TYPE_DRV_PRIV, /* driver defined types from here */
84 typedef __u32 __bitwise req_flags_t;
86 /* elevator knows about this request */
87 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
88 /* drive already may have started this one */
89 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
90 /* uses tagged queueing */
91 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
92 /* may not be passed by ioscheduler */
93 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
94 /* request for flush sequence */
95 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
96 /* merge of different types, fail separately */
97 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
98 /* track inflight for MQ */
99 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
100 /* don't call prep for this one */
101 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
102 /* set for "ide_preempt" requests and also for requests for which the SCSI
103 "quiesce" state must be ignored. */
104 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
105 /* contains copies of user pages */
106 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
107 /* vaguely specified driver internal error. Ignored by the block layer */
108 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
109 /* don't warn about errors */
110 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
111 /* elevator private data attached */
112 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
113 /* account I/O stat */
114 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
115 /* request came from our alloc pool */
116 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
117 /* runtime pm request */
118 #define RQF_PM ((__force req_flags_t)(1 << 15))
119 /* on IO scheduler merge hash */
120 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
121 /* IO stats tracking on */
122 #define RQF_STATS ((__force req_flags_t)(1 << 17))
123 /* Look at ->special_vec for the actual data payload instead of the
125 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
127 /* flags that prevent us from merging requests: */
128 #define RQF_NOMERGE_FLAGS \
129 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
131 #define BLK_MAX_CDB 16
134 * Try to put the fields that are referenced together in the same cacheline.
136 * If you modify this structure, make sure to update blk_rq_init() and
137 * especially blk_mq_rq_ctx_init() to take care of the added fields.
140 struct list_head queuelist;
142 struct call_single_data csd;
146 struct request_queue *q;
147 struct blk_mq_ctx *mq_ctx;
151 unsigned int cmd_flags; /* op and common flags */
152 req_flags_t rq_flags;
153 unsigned long atomic_flags;
155 /* the following two fields are internal, NEVER access directly */
156 unsigned int __data_len; /* total data len */
157 sector_t __sector; /* sector cursor */
163 * The hash is used inside the scheduler, and killed once the
164 * request reaches the dispatch list. The ipi_list is only used
165 * to queue the request for softirq completion, which is long
166 * after the request has been unhashed (and even removed from
167 * the dispatch list).
170 struct hlist_node hash; /* merge hash */
171 struct list_head ipi_list;
175 * The rb_node is only used inside the io scheduler, requests
176 * are pruned when moved to the dispatch queue. So let the
177 * completion_data share space with the rb_node.
180 struct rb_node rb_node; /* sort/lookup */
181 struct bio_vec special_vec;
182 void *completion_data;
186 * Three pointers are available for the IO schedulers, if they need
187 * more they have to dynamically allocate it. Flush requests are
188 * never put on the IO scheduler. So let the flush fields share
189 * space with the elevator data.
199 struct list_head list;
200 rq_end_io_fn *saved_end_io;
204 struct gendisk *rq_disk;
205 struct hd_struct *part;
206 unsigned long start_time;
207 struct blk_issue_stat issue_stat;
208 #ifdef CONFIG_BLK_CGROUP
209 struct request_list *rl; /* rl this rq is alloced from */
210 unsigned long long start_time_ns;
211 unsigned long long io_start_time_ns; /* when passed to hardware */
213 /* Number of scatter-gather DMA addr+len pairs after
214 * physical address coalescing is performed.
216 unsigned short nr_phys_segments;
217 #if defined(CONFIG_BLK_DEV_INTEGRITY)
218 unsigned short nr_integrity_segments;
221 unsigned short ioprio;
223 void *special; /* opaque pointer available for LLD use */
229 * when request is used as a packet command carrier
231 unsigned char __cmd[BLK_MAX_CDB];
233 unsigned short cmd_len;
235 unsigned int extra_len; /* length of alignment and padding */
236 unsigned int sense_len;
237 unsigned int resid_len; /* residual count */
240 unsigned long deadline;
241 struct list_head timeout_list;
242 unsigned int timeout;
246 * completion callback.
248 rq_end_io_fn *end_io;
252 struct request *next_rq;
255 static inline unsigned short req_get_ioprio(struct request *req)
260 #include <linux/elevator.h>
262 struct blk_queue_ctx;
264 typedef void (request_fn_proc) (struct request_queue *q);
265 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
266 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
267 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
270 typedef void (softirq_done_fn)(struct request *);
271 typedef int (dma_drain_needed_fn)(struct request *);
272 typedef int (lld_busy_fn) (struct request_queue *q);
273 typedef int (bsg_job_fn) (struct bsg_job *);
275 enum blk_eh_timer_return {
281 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
283 enum blk_queue_state {
288 struct blk_queue_tag {
289 struct request **tag_index; /* map of busy tags */
290 unsigned long *tag_map; /* bit map of free/busy tags */
291 int max_depth; /* what we will send to device */
292 int real_max_depth; /* what the array can hold */
293 atomic_t refcnt; /* map can be shared */
294 int alloc_policy; /* tag allocation policy */
295 int next_tag; /* next tag */
297 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
298 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
300 #define BLK_SCSI_MAX_CMDS (256)
301 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
304 * Zoned block device models (zoned limit).
306 enum blk_zoned_model {
307 BLK_ZONED_NONE, /* Regular block device */
308 BLK_ZONED_HA, /* Host-aware zoned block device */
309 BLK_ZONED_HM, /* Host-managed zoned block device */
312 struct queue_limits {
313 unsigned long bounce_pfn;
314 unsigned long seg_boundary_mask;
315 unsigned long virt_boundary_mask;
317 unsigned int max_hw_sectors;
318 unsigned int max_dev_sectors;
319 unsigned int chunk_sectors;
320 unsigned int max_sectors;
321 unsigned int max_segment_size;
322 unsigned int physical_block_size;
323 unsigned int alignment_offset;
326 unsigned int max_discard_sectors;
327 unsigned int max_hw_discard_sectors;
328 unsigned int max_write_same_sectors;
329 unsigned int max_write_zeroes_sectors;
330 unsigned int discard_granularity;
331 unsigned int discard_alignment;
333 unsigned short logical_block_size;
334 unsigned short max_segments;
335 unsigned short max_integrity_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;
410 struct blk_mq_ops *mq_ops;
412 unsigned int *mq_map;
415 struct blk_mq_ctx __percpu *queue_ctx;
416 unsigned int nr_queues;
418 unsigned int queue_depth;
420 /* hw dispatch queues */
421 struct blk_mq_hw_ctx **queue_hw_ctx;
422 unsigned int nr_hw_queues;
425 * Dispatch queue sorting
428 struct request *boundary_rq;
431 * Delayed queue handling
433 struct delayed_work delay_work;
435 struct backing_dev_info backing_dev_info;
438 * The queue owner gets to use this for whatever they like.
439 * ll_rw_blk doesn't touch it.
444 * various queue flags, see QUEUE_* below
446 unsigned long queue_flags;
449 * ida allocated id for this queue. Used to index queues from
455 * queue needs bounce pages for pages above this limit
460 * protects queue structures from reentrancy. ->__queue_lock should
461 * _never_ be used directly, it is queue private. always use
464 spinlock_t __queue_lock;
465 spinlock_t *queue_lock;
475 struct kobject mq_kobj;
477 #ifdef CONFIG_BLK_DEV_INTEGRITY
478 struct blk_integrity integrity;
479 #endif /* CONFIG_BLK_DEV_INTEGRITY */
484 unsigned int nr_pending;
490 unsigned long nr_requests; /* Max # of requests */
491 unsigned int nr_congestion_on;
492 unsigned int nr_congestion_off;
493 unsigned int nr_batching;
495 unsigned int dma_drain_size;
496 void *dma_drain_buffer;
497 unsigned int dma_pad_mask;
498 unsigned int dma_alignment;
500 struct blk_queue_tag *queue_tags;
501 struct list_head tag_busy_list;
503 unsigned int nr_sorted;
504 unsigned int in_flight[2];
506 struct blk_rq_stat rq_stats[2];
509 * Number of active block driver functions for which blk_drain_queue()
510 * must wait. Must be incremented around functions that unlock the
511 * queue_lock internally, e.g. scsi_request_fn().
513 unsigned int request_fn_active;
515 unsigned int rq_timeout;
517 struct timer_list timeout;
518 struct work_struct timeout_work;
519 struct list_head timeout_list;
521 struct list_head icq_list;
522 #ifdef CONFIG_BLK_CGROUP
523 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
524 struct blkcg_gq *root_blkg;
525 struct list_head blkg_list;
528 struct queue_limits limits;
533 unsigned int sg_timeout;
534 unsigned int sg_reserved_size;
536 #ifdef CONFIG_BLK_DEV_IO_TRACE
537 struct blk_trace *blk_trace;
540 * for flush operations
542 struct blk_flush_queue *fq;
544 struct list_head requeue_list;
545 spinlock_t requeue_lock;
546 struct delayed_work requeue_work;
548 struct mutex sysfs_lock;
551 atomic_t mq_freeze_depth;
553 #if defined(CONFIG_BLK_DEV_BSG)
554 bsg_job_fn *bsg_job_fn;
556 struct bsg_class_device bsg_dev;
559 #ifdef CONFIG_BLK_DEV_THROTTLING
561 struct throtl_data *td;
563 struct rcu_head rcu_head;
564 wait_queue_head_t mq_freeze_wq;
565 struct percpu_ref q_usage_counter;
566 struct list_head all_q_node;
568 struct blk_mq_tag_set *tag_set;
569 struct list_head tag_set_list;
570 struct bio_set *bio_split;
572 bool mq_sysfs_init_done;
575 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
576 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
577 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
578 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
579 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
580 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
581 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
582 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
583 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
584 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
585 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
586 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
587 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
588 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
589 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
590 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
591 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
592 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
593 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
594 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
595 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
596 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
597 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
598 #define QUEUE_FLAG_WC 23 /* Write back caching */
599 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
600 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
601 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
602 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
604 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
605 (1 << QUEUE_FLAG_STACKABLE) | \
606 (1 << QUEUE_FLAG_SAME_COMP) | \
607 (1 << QUEUE_FLAG_ADD_RANDOM))
609 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
610 (1 << QUEUE_FLAG_STACKABLE) | \
611 (1 << QUEUE_FLAG_SAME_COMP) | \
612 (1 << QUEUE_FLAG_POLL))
614 static inline void queue_lockdep_assert_held(struct request_queue *q)
617 lockdep_assert_held(q->queue_lock);
620 static inline void queue_flag_set_unlocked(unsigned int flag,
621 struct request_queue *q)
623 __set_bit(flag, &q->queue_flags);
626 static inline int queue_flag_test_and_clear(unsigned int flag,
627 struct request_queue *q)
629 queue_lockdep_assert_held(q);
631 if (test_bit(flag, &q->queue_flags)) {
632 __clear_bit(flag, &q->queue_flags);
639 static inline int queue_flag_test_and_set(unsigned int flag,
640 struct request_queue *q)
642 queue_lockdep_assert_held(q);
644 if (!test_bit(flag, &q->queue_flags)) {
645 __set_bit(flag, &q->queue_flags);
652 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
654 queue_lockdep_assert_held(q);
655 __set_bit(flag, &q->queue_flags);
658 static inline void queue_flag_clear_unlocked(unsigned int flag,
659 struct request_queue *q)
661 __clear_bit(flag, &q->queue_flags);
664 static inline int queue_in_flight(struct request_queue *q)
666 return q->in_flight[0] + q->in_flight[1];
669 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
671 queue_lockdep_assert_held(q);
672 __clear_bit(flag, &q->queue_flags);
675 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
676 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
677 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
678 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
679 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
680 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
681 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
682 #define blk_queue_noxmerges(q) \
683 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
684 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
685 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
686 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
687 #define blk_queue_stackable(q) \
688 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
689 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
690 #define blk_queue_secure_erase(q) \
691 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
692 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
694 #define blk_noretry_request(rq) \
695 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
696 REQ_FAILFAST_DRIVER))
698 #define blk_account_rq(rq) \
699 (((rq)->rq_flags & RQF_STARTED) && \
700 ((rq)->cmd_type == REQ_TYPE_FS))
702 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
703 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
704 /* rq->queuelist of dequeued request must be list_empty() */
705 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
707 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
709 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
712 * Driver can handle struct request, if it either has an old style
713 * request_fn defined, or is blk-mq based.
715 static inline bool queue_is_rq_based(struct request_queue *q)
717 return q->request_fn || q->mq_ops;
720 static inline unsigned int blk_queue_cluster(struct request_queue *q)
722 return q->limits.cluster;
725 static inline enum blk_zoned_model
726 blk_queue_zoned_model(struct request_queue *q)
728 return q->limits.zoned;
731 static inline bool blk_queue_is_zoned(struct request_queue *q)
733 switch (blk_queue_zoned_model(q)) {
742 static inline unsigned int blk_queue_zone_size(struct request_queue *q)
744 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
747 static inline bool rq_is_sync(struct request *rq)
749 return op_is_sync(rq->cmd_flags);
752 static inline bool blk_rl_full(struct request_list *rl, bool sync)
754 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
756 return rl->flags & flag;
759 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
761 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
766 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
768 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
773 static inline bool rq_mergeable(struct request *rq)
775 if (rq->cmd_type != REQ_TYPE_FS)
778 if (req_op(rq) == REQ_OP_FLUSH)
781 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
784 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
786 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
792 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
794 if (bio_data(a) == bio_data(b))
800 static inline unsigned int blk_queue_depth(struct request_queue *q)
803 return q->queue_depth;
805 return q->nr_requests;
809 * q->prep_rq_fn return values
812 BLKPREP_OK, /* serve it */
813 BLKPREP_KILL, /* fatal error, kill, return -EIO */
814 BLKPREP_DEFER, /* leave on queue */
815 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
818 extern unsigned long blk_max_low_pfn, blk_max_pfn;
821 * standard bounce addresses:
823 * BLK_BOUNCE_HIGH : bounce all highmem pages
824 * BLK_BOUNCE_ANY : don't bounce anything
825 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
828 #if BITS_PER_LONG == 32
829 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
831 #define BLK_BOUNCE_HIGH -1ULL
833 #define BLK_BOUNCE_ANY (-1ULL)
834 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
837 * default timeout for SG_IO if none specified
839 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
840 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
843 extern int init_emergency_isa_pool(void);
844 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
846 static inline int init_emergency_isa_pool(void)
850 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
853 #endif /* CONFIG_MMU */
859 unsigned long offset;
864 struct req_iterator {
865 struct bvec_iter iter;
869 /* This should not be used directly - use rq_for_each_segment */
870 #define for_each_bio(_bio) \
871 for (; _bio; _bio = _bio->bi_next)
872 #define __rq_for_each_bio(_bio, rq) \
874 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
876 #define rq_for_each_segment(bvl, _rq, _iter) \
877 __rq_for_each_bio(_iter.bio, _rq) \
878 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
880 #define rq_iter_last(bvec, _iter) \
881 (_iter.bio->bi_next == NULL && \
882 bio_iter_last(bvec, _iter.iter))
884 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
885 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
887 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
888 extern void rq_flush_dcache_pages(struct request *rq);
890 static inline void rq_flush_dcache_pages(struct request *rq)
896 #define vfs_msg(sb, level, fmt, ...) \
897 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
899 #define vfs_msg(sb, level, fmt, ...) \
901 no_printk(fmt, ##__VA_ARGS__); \
902 __vfs_msg(sb, "", " "); \
906 extern int blk_register_queue(struct gendisk *disk);
907 extern void blk_unregister_queue(struct gendisk *disk);
908 extern blk_qc_t generic_make_request(struct bio *bio);
909 extern void blk_rq_init(struct request_queue *q, struct request *rq);
910 extern void blk_put_request(struct request *);
911 extern void __blk_put_request(struct request_queue *, struct request *);
912 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
913 extern void blk_rq_set_block_pc(struct request *);
914 extern void blk_requeue_request(struct request_queue *, struct request *);
915 extern int blk_lld_busy(struct request_queue *q);
916 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
917 struct bio_set *bs, gfp_t gfp_mask,
918 int (*bio_ctr)(struct bio *, struct bio *, void *),
920 extern void blk_rq_unprep_clone(struct request *rq);
921 extern int blk_insert_cloned_request(struct request_queue *q,
923 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
924 extern void blk_delay_queue(struct request_queue *, unsigned long);
925 extern void blk_queue_split(struct request_queue *, struct bio **,
927 extern void blk_recount_segments(struct request_queue *, struct bio *);
928 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
929 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
930 unsigned int, void __user *);
931 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
932 unsigned int, void __user *);
933 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
934 struct scsi_ioctl_command __user *);
936 extern int blk_queue_enter(struct request_queue *q, bool nowait);
937 extern void blk_queue_exit(struct request_queue *q);
938 extern void blk_start_queue(struct request_queue *q);
939 extern void blk_start_queue_async(struct request_queue *q);
940 extern void blk_stop_queue(struct request_queue *q);
941 extern void blk_sync_queue(struct request_queue *q);
942 extern void __blk_stop_queue(struct request_queue *q);
943 extern void __blk_run_queue(struct request_queue *q);
944 extern void __blk_run_queue_uncond(struct request_queue *q);
945 extern void blk_run_queue(struct request_queue *);
946 extern void blk_run_queue_async(struct request_queue *q);
947 extern void blk_mq_quiesce_queue(struct request_queue *q);
948 extern int blk_rq_map_user(struct request_queue *, struct request *,
949 struct rq_map_data *, void __user *, unsigned long,
951 extern int blk_rq_unmap_user(struct bio *);
952 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
953 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
954 struct rq_map_data *, const struct iov_iter *,
956 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
957 struct request *, int);
958 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
959 struct request *, int, rq_end_io_fn *);
961 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
963 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
965 return bdev->bd_disk->queue; /* this is never NULL */
969 * blk_rq_pos() : the current sector
970 * blk_rq_bytes() : bytes left in the entire request
971 * blk_rq_cur_bytes() : bytes left in the current segment
972 * blk_rq_err_bytes() : bytes left till the next error boundary
973 * blk_rq_sectors() : sectors left in the entire request
974 * blk_rq_cur_sectors() : sectors left in the current segment
976 static inline sector_t blk_rq_pos(const struct request *rq)
981 static inline unsigned int blk_rq_bytes(const struct request *rq)
983 return rq->__data_len;
986 static inline int blk_rq_cur_bytes(const struct request *rq)
988 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
991 extern unsigned int blk_rq_err_bytes(const struct request *rq);
993 static inline unsigned int blk_rq_sectors(const struct request *rq)
995 return blk_rq_bytes(rq) >> 9;
998 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1000 return blk_rq_cur_bytes(rq) >> 9;
1003 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1006 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1007 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1009 if (unlikely(op == REQ_OP_WRITE_SAME))
1010 return q->limits.max_write_same_sectors;
1012 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1013 return q->limits.max_write_zeroes_sectors;
1015 return q->limits.max_sectors;
1019 * Return maximum size of a request at given offset. Only valid for
1020 * file system requests.
1022 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1025 if (!q->limits.chunk_sectors)
1026 return q->limits.max_sectors;
1028 return q->limits.chunk_sectors -
1029 (offset & (q->limits.chunk_sectors - 1));
1032 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1035 struct request_queue *q = rq->q;
1037 if (unlikely(rq->cmd_type != REQ_TYPE_FS))
1038 return q->limits.max_hw_sectors;
1040 if (!q->limits.chunk_sectors ||
1041 req_op(rq) == REQ_OP_DISCARD ||
1042 req_op(rq) == REQ_OP_SECURE_ERASE)
1043 return blk_queue_get_max_sectors(q, req_op(rq));
1045 return min(blk_max_size_offset(q, offset),
1046 blk_queue_get_max_sectors(q, req_op(rq)));
1049 static inline unsigned int blk_rq_count_bios(struct request *rq)
1051 unsigned int nr_bios = 0;
1054 __rq_for_each_bio(bio, rq)
1061 * blk_rq_set_prio - associate a request with prio from ioc
1062 * @rq: request of interest
1063 * @ioc: target iocontext
1065 * Assocate request prio with ioc prio so request based drivers
1066 * can leverage priority information.
1068 static inline void blk_rq_set_prio(struct request *rq, struct io_context *ioc)
1071 rq->ioprio = ioc->ioprio;
1075 * Request issue related functions.
1077 extern struct request *blk_peek_request(struct request_queue *q);
1078 extern void blk_start_request(struct request *rq);
1079 extern struct request *blk_fetch_request(struct request_queue *q);
1082 * Request completion related functions.
1084 * blk_update_request() completes given number of bytes and updates
1085 * the request without completing it.
1087 * blk_end_request() and friends. __blk_end_request() must be called
1088 * with the request queue spinlock acquired.
1090 * Several drivers define their own end_request and call
1091 * blk_end_request() for parts of the original function.
1092 * This prevents code duplication in drivers.
1094 extern bool blk_update_request(struct request *rq, int error,
1095 unsigned int nr_bytes);
1096 extern void blk_finish_request(struct request *rq, int error);
1097 extern bool blk_end_request(struct request *rq, int error,
1098 unsigned int nr_bytes);
1099 extern void blk_end_request_all(struct request *rq, int error);
1100 extern bool blk_end_request_cur(struct request *rq, int error);
1101 extern bool blk_end_request_err(struct request *rq, int error);
1102 extern bool __blk_end_request(struct request *rq, int error,
1103 unsigned int nr_bytes);
1104 extern void __blk_end_request_all(struct request *rq, int error);
1105 extern bool __blk_end_request_cur(struct request *rq, int error);
1106 extern bool __blk_end_request_err(struct request *rq, int error);
1108 extern void blk_complete_request(struct request *);
1109 extern void __blk_complete_request(struct request *);
1110 extern void blk_abort_request(struct request *);
1111 extern void blk_unprep_request(struct request *);
1114 * Access functions for manipulating queue properties
1116 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1117 spinlock_t *lock, int node_id);
1118 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1119 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
1120 request_fn_proc *, spinlock_t *);
1121 extern void blk_cleanup_queue(struct request_queue *);
1122 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1123 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1124 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1125 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1126 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1127 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1128 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1129 unsigned int max_discard_sectors);
1130 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1131 unsigned int max_write_same_sectors);
1132 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1133 unsigned int max_write_same_sectors);
1134 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1135 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1136 extern void blk_queue_alignment_offset(struct request_queue *q,
1137 unsigned int alignment);
1138 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1139 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1140 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1141 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1142 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1143 extern void blk_set_default_limits(struct queue_limits *lim);
1144 extern void blk_set_stacking_limits(struct queue_limits *lim);
1145 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1147 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1149 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1151 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1152 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1153 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1154 extern int blk_queue_dma_drain(struct request_queue *q,
1155 dma_drain_needed_fn *dma_drain_needed,
1156 void *buf, unsigned int size);
1157 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1158 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1159 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1160 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1161 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1162 extern void blk_queue_dma_alignment(struct request_queue *, int);
1163 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1164 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1165 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1166 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1167 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1168 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1169 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1171 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1173 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1175 return rq->nr_phys_segments;
1178 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1179 extern void blk_dump_rq_flags(struct request *, char *);
1180 extern long nr_blockdev_pages(void);
1182 bool __must_check blk_get_queue(struct request_queue *);
1183 struct request_queue *blk_alloc_queue(gfp_t);
1184 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1185 extern void blk_put_queue(struct request_queue *);
1186 extern void blk_set_queue_dying(struct request_queue *);
1189 * block layer runtime pm functions
1192 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1193 extern int blk_pre_runtime_suspend(struct request_queue *q);
1194 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1195 extern void blk_pre_runtime_resume(struct request_queue *q);
1196 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1197 extern void blk_set_runtime_active(struct request_queue *q);
1199 static inline void blk_pm_runtime_init(struct request_queue *q,
1200 struct device *dev) {}
1201 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1205 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1206 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1207 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1208 static inline void blk_set_runtime_active(struct request_queue *q) {}
1212 * blk_plug permits building a queue of related requests by holding the I/O
1213 * fragments for a short period. This allows merging of sequential requests
1214 * into single larger request. As the requests are moved from a per-task list to
1215 * the device's request_queue in a batch, this results in improved scalability
1216 * as the lock contention for request_queue lock is reduced.
1218 * It is ok not to disable preemption when adding the request to the plug list
1219 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1220 * the plug list when the task sleeps by itself. For details, please see
1221 * schedule() where blk_schedule_flush_plug() is called.
1224 struct list_head list; /* requests */
1225 struct list_head mq_list; /* blk-mq requests */
1226 struct list_head cb_list; /* md requires an unplug callback */
1228 #define BLK_MAX_REQUEST_COUNT 16
1229 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1232 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1233 struct blk_plug_cb {
1234 struct list_head list;
1235 blk_plug_cb_fn callback;
1238 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1239 void *data, int size);
1240 extern void blk_start_plug(struct blk_plug *);
1241 extern void blk_finish_plug(struct blk_plug *);
1242 extern void blk_flush_plug_list(struct blk_plug *, bool);
1244 static inline void blk_flush_plug(struct task_struct *tsk)
1246 struct blk_plug *plug = tsk->plug;
1249 blk_flush_plug_list(plug, false);
1252 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1254 struct blk_plug *plug = tsk->plug;
1257 blk_flush_plug_list(plug, true);
1260 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1262 struct blk_plug *plug = tsk->plug;
1265 (!list_empty(&plug->list) ||
1266 !list_empty(&plug->mq_list) ||
1267 !list_empty(&plug->cb_list));
1273 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1274 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1275 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1276 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1277 extern void blk_queue_free_tags(struct request_queue *);
1278 extern int blk_queue_resize_tags(struct request_queue *, int);
1279 extern void blk_queue_invalidate_tags(struct request_queue *);
1280 extern struct blk_queue_tag *blk_init_tags(int, int);
1281 extern void blk_free_tags(struct blk_queue_tag *);
1283 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1286 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1288 return bqt->tag_index[tag];
1292 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1293 #define BLKDEV_DISCARD_ZERO (1 << 1) /* must reliably zero data */
1295 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1296 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1297 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1298 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1299 sector_t nr_sects, gfp_t gfp_mask, int flags,
1301 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1302 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1303 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1304 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1306 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1307 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1308 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1309 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1311 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1312 nr_blocks << (sb->s_blocksize_bits - 9),
1315 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1316 sector_t nr_blocks, gfp_t gfp_mask)
1318 return blkdev_issue_zeroout(sb->s_bdev,
1319 block << (sb->s_blocksize_bits - 9),
1320 nr_blocks << (sb->s_blocksize_bits - 9),
1324 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1326 enum blk_default_limits {
1327 BLK_MAX_SEGMENTS = 128,
1328 BLK_SAFE_MAX_SECTORS = 255,
1329 BLK_DEF_MAX_SECTORS = 2560,
1330 BLK_MAX_SEGMENT_SIZE = 65536,
1331 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1334 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1336 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1338 return q->limits.bounce_pfn;
1341 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1343 return q->limits.seg_boundary_mask;
1346 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1348 return q->limits.virt_boundary_mask;
1351 static inline unsigned int queue_max_sectors(struct request_queue *q)
1353 return q->limits.max_sectors;
1356 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1358 return q->limits.max_hw_sectors;
1361 static inline unsigned short queue_max_segments(struct request_queue *q)
1363 return q->limits.max_segments;
1366 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1368 return q->limits.max_segment_size;
1371 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1375 if (q && q->limits.logical_block_size)
1376 retval = q->limits.logical_block_size;
1381 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1383 return queue_logical_block_size(bdev_get_queue(bdev));
1386 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1388 return q->limits.physical_block_size;
1391 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1393 return queue_physical_block_size(bdev_get_queue(bdev));
1396 static inline unsigned int queue_io_min(struct request_queue *q)
1398 return q->limits.io_min;
1401 static inline int bdev_io_min(struct block_device *bdev)
1403 return queue_io_min(bdev_get_queue(bdev));
1406 static inline unsigned int queue_io_opt(struct request_queue *q)
1408 return q->limits.io_opt;
1411 static inline int bdev_io_opt(struct block_device *bdev)
1413 return queue_io_opt(bdev_get_queue(bdev));
1416 static inline int queue_alignment_offset(struct request_queue *q)
1418 if (q->limits.misaligned)
1421 return q->limits.alignment_offset;
1424 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1426 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1427 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1429 return (granularity + lim->alignment_offset - alignment) % granularity;
1432 static inline int bdev_alignment_offset(struct block_device *bdev)
1434 struct request_queue *q = bdev_get_queue(bdev);
1436 if (q->limits.misaligned)
1439 if (bdev != bdev->bd_contains)
1440 return bdev->bd_part->alignment_offset;
1442 return q->limits.alignment_offset;
1445 static inline int queue_discard_alignment(struct request_queue *q)
1447 if (q->limits.discard_misaligned)
1450 return q->limits.discard_alignment;
1453 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1455 unsigned int alignment, granularity, offset;
1457 if (!lim->max_discard_sectors)
1460 /* Why are these in bytes, not sectors? */
1461 alignment = lim->discard_alignment >> 9;
1462 granularity = lim->discard_granularity >> 9;
1466 /* Offset of the partition start in 'granularity' sectors */
1467 offset = sector_div(sector, granularity);
1469 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1470 offset = (granularity + alignment - offset) % granularity;
1472 /* Turn it back into bytes, gaah */
1476 static inline int bdev_discard_alignment(struct block_device *bdev)
1478 struct request_queue *q = bdev_get_queue(bdev);
1480 if (bdev != bdev->bd_contains)
1481 return bdev->bd_part->discard_alignment;
1483 return q->limits.discard_alignment;
1486 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1488 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1494 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1496 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1499 static inline unsigned int bdev_write_same(struct block_device *bdev)
1501 struct request_queue *q = bdev_get_queue(bdev);
1504 return q->limits.max_write_same_sectors;
1509 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1511 struct request_queue *q = bdev_get_queue(bdev);
1514 return q->limits.max_write_zeroes_sectors;
1519 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1521 struct request_queue *q = bdev_get_queue(bdev);
1524 return blk_queue_zoned_model(q);
1526 return BLK_ZONED_NONE;
1529 static inline bool bdev_is_zoned(struct block_device *bdev)
1531 struct request_queue *q = bdev_get_queue(bdev);
1534 return blk_queue_is_zoned(q);
1539 static inline unsigned int bdev_zone_size(struct block_device *bdev)
1541 struct request_queue *q = bdev_get_queue(bdev);
1544 return blk_queue_zone_size(q);
1549 static inline int queue_dma_alignment(struct request_queue *q)
1551 return q ? q->dma_alignment : 511;
1554 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1557 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1558 return !(addr & alignment) && !(len & alignment);
1561 /* assumes size > 256 */
1562 static inline unsigned int blksize_bits(unsigned int size)
1564 unsigned int bits = 8;
1568 } while (size > 256);
1572 static inline unsigned int block_size(struct block_device *bdev)
1574 return bdev->bd_block_size;
1577 static inline bool queue_flush_queueable(struct request_queue *q)
1579 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1582 typedef struct {struct page *v;} Sector;
1584 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1586 static inline void put_dev_sector(Sector p)
1591 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1592 struct bio_vec *bprv, unsigned int offset)
1595 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1599 * Check if adding a bio_vec after bprv with offset would create a gap in
1600 * the SG list. Most drivers don't care about this, but some do.
1602 static inline bool bvec_gap_to_prev(struct request_queue *q,
1603 struct bio_vec *bprv, unsigned int offset)
1605 if (!queue_virt_boundary(q))
1607 return __bvec_gap_to_prev(q, bprv, offset);
1610 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1613 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1614 struct bio_vec pb, nb;
1616 bio_get_last_bvec(prev, &pb);
1617 bio_get_first_bvec(next, &nb);
1619 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1625 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1627 return bio_will_gap(req->q, req->biotail, bio);
1630 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1632 return bio_will_gap(req->q, bio, req->bio);
1635 int kblockd_schedule_work(struct work_struct *work);
1636 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1637 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1638 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1640 #ifdef CONFIG_BLK_CGROUP
1642 * This should not be using sched_clock(). A real patch is in progress
1643 * to fix this up, until that is in place we need to disable preemption
1644 * around sched_clock() in this function and set_io_start_time_ns().
1646 static inline void set_start_time_ns(struct request *req)
1649 req->start_time_ns = sched_clock();
1653 static inline void set_io_start_time_ns(struct request *req)
1656 req->io_start_time_ns = sched_clock();
1660 static inline uint64_t rq_start_time_ns(struct request *req)
1662 return req->start_time_ns;
1665 static inline uint64_t rq_io_start_time_ns(struct request *req)
1667 return req->io_start_time_ns;
1670 static inline void set_start_time_ns(struct request *req) {}
1671 static inline void set_io_start_time_ns(struct request *req) {}
1672 static inline uint64_t rq_start_time_ns(struct request *req)
1676 static inline uint64_t rq_io_start_time_ns(struct request *req)
1682 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1683 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1684 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1685 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1687 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1689 enum blk_integrity_flags {
1690 BLK_INTEGRITY_VERIFY = 1 << 0,
1691 BLK_INTEGRITY_GENERATE = 1 << 1,
1692 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1693 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1696 struct blk_integrity_iter {
1700 unsigned int data_size;
1701 unsigned short interval;
1702 const char *disk_name;
1705 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1707 struct blk_integrity_profile {
1708 integrity_processing_fn *generate_fn;
1709 integrity_processing_fn *verify_fn;
1713 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1714 extern void blk_integrity_unregister(struct gendisk *);
1715 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1716 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1717 struct scatterlist *);
1718 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1719 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1721 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1724 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1726 struct blk_integrity *bi = &disk->queue->integrity;
1735 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1737 return blk_get_integrity(bdev->bd_disk);
1740 static inline bool blk_integrity_rq(struct request *rq)
1742 return rq->cmd_flags & REQ_INTEGRITY;
1745 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1748 q->limits.max_integrity_segments = segs;
1751 static inline unsigned short
1752 queue_max_integrity_segments(struct request_queue *q)
1754 return q->limits.max_integrity_segments;
1757 static inline bool integrity_req_gap_back_merge(struct request *req,
1760 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1761 struct bio_integrity_payload *bip_next = bio_integrity(next);
1763 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1764 bip_next->bip_vec[0].bv_offset);
1767 static inline bool integrity_req_gap_front_merge(struct request *req,
1770 struct bio_integrity_payload *bip = bio_integrity(bio);
1771 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1773 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1774 bip_next->bip_vec[0].bv_offset);
1777 #else /* CONFIG_BLK_DEV_INTEGRITY */
1780 struct block_device;
1782 struct blk_integrity;
1784 static inline int blk_integrity_rq(struct request *rq)
1788 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1793 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1795 struct scatterlist *s)
1799 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1803 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1807 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1811 static inline void blk_integrity_register(struct gendisk *d,
1812 struct blk_integrity *b)
1815 static inline void blk_integrity_unregister(struct gendisk *d)
1818 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1822 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1826 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1832 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1839 static inline bool integrity_req_gap_back_merge(struct request *req,
1844 static inline bool integrity_req_gap_front_merge(struct request *req,
1850 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1853 * struct blk_dax_ctl - control and output parameters for ->direct_access
1854 * @sector: (input) offset relative to a block_device
1855 * @addr: (output) kernel virtual address for @sector populated by driver
1856 * @pfn: (output) page frame number for @addr populated by driver
1857 * @size: (input) number of bytes requested
1859 struct blk_dax_ctl {
1866 struct block_device_operations {
1867 int (*open) (struct block_device *, fmode_t);
1868 void (*release) (struct gendisk *, fmode_t);
1869 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1870 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1871 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1872 long (*direct_access)(struct block_device *, sector_t, void **, pfn_t *,
1874 unsigned int (*check_events) (struct gendisk *disk,
1875 unsigned int clearing);
1876 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1877 int (*media_changed) (struct gendisk *);
1878 void (*unlock_native_capacity) (struct gendisk *);
1879 int (*revalidate_disk) (struct gendisk *);
1880 int (*getgeo)(struct block_device *, struct hd_geometry *);
1881 /* this callback is with swap_lock and sometimes page table lock held */
1882 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1883 struct module *owner;
1884 const struct pr_ops *pr_ops;
1887 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1889 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1890 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1891 struct writeback_control *);
1892 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1893 extern int bdev_dax_supported(struct super_block *, int);
1894 extern bool bdev_dax_capable(struct block_device *);
1895 #else /* CONFIG_BLOCK */
1897 struct block_device;
1900 * stubs for when the block layer is configured out
1902 #define buffer_heads_over_limit 0
1904 static inline long nr_blockdev_pages(void)
1912 static inline void blk_start_plug(struct blk_plug *plug)
1916 static inline void blk_finish_plug(struct blk_plug *plug)
1920 static inline void blk_flush_plug(struct task_struct *task)
1924 static inline void blk_schedule_flush_plug(struct task_struct *task)
1929 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1934 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1935 sector_t *error_sector)
1940 #endif /* CONFIG_BLOCK */