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 busy; /* current depth */
292 int max_depth; /* what we will send to device */
293 int real_max_depth; /* what the array can hold */
294 atomic_t refcnt; /* map can be shared */
295 int alloc_policy; /* tag allocation policy */
296 int next_tag; /* next tag */
298 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
299 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
301 #define BLK_SCSI_MAX_CMDS (256)
302 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
305 * Zoned block device models (zoned limit).
307 enum blk_zoned_model {
308 BLK_ZONED_NONE, /* Regular block device */
309 BLK_ZONED_HA, /* Host-aware zoned block device */
310 BLK_ZONED_HM, /* Host-managed zoned block device */
313 struct queue_limits {
314 unsigned long bounce_pfn;
315 unsigned long seg_boundary_mask;
316 unsigned long virt_boundary_mask;
318 unsigned int max_hw_sectors;
319 unsigned int max_dev_sectors;
320 unsigned int chunk_sectors;
321 unsigned int max_sectors;
322 unsigned int max_segment_size;
323 unsigned int physical_block_size;
324 unsigned int alignment_offset;
327 unsigned int max_discard_sectors;
328 unsigned int max_hw_discard_sectors;
329 unsigned int max_write_same_sectors;
330 unsigned int max_write_zeroes_sectors;
331 unsigned int discard_granularity;
332 unsigned int discard_alignment;
334 unsigned short logical_block_size;
335 unsigned short max_segments;
336 unsigned short max_integrity_segments;
338 unsigned char misaligned;
339 unsigned char discard_misaligned;
340 unsigned char cluster;
341 unsigned char discard_zeroes_data;
342 unsigned char raid_partial_stripes_expensive;
343 enum blk_zoned_model zoned;
346 #ifdef CONFIG_BLK_DEV_ZONED
348 struct blk_zone_report_hdr {
349 unsigned int nr_zones;
353 extern int blkdev_report_zones(struct block_device *bdev,
354 sector_t sector, struct blk_zone *zones,
355 unsigned int *nr_zones, gfp_t gfp_mask);
356 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
357 sector_t nr_sectors, gfp_t gfp_mask);
359 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
360 unsigned int cmd, unsigned long arg);
361 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
362 unsigned int cmd, unsigned long arg);
364 #else /* CONFIG_BLK_DEV_ZONED */
366 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
367 fmode_t mode, unsigned int cmd,
373 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
374 fmode_t mode, unsigned int cmd,
380 #endif /* CONFIG_BLK_DEV_ZONED */
382 struct request_queue {
384 * Together with queue_head for cacheline sharing
386 struct list_head queue_head;
387 struct request *last_merge;
388 struct elevator_queue *elevator;
389 int nr_rqs[2]; /* # allocated [a]sync rqs */
390 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
395 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
396 * is used, root blkg allocates from @q->root_rl and all other
397 * blkgs from their own blkg->rl. Which one to use should be
398 * determined using bio_request_list().
400 struct request_list root_rl;
402 request_fn_proc *request_fn;
403 make_request_fn *make_request_fn;
404 prep_rq_fn *prep_rq_fn;
405 unprep_rq_fn *unprep_rq_fn;
406 softirq_done_fn *softirq_done_fn;
407 rq_timed_out_fn *rq_timed_out_fn;
408 dma_drain_needed_fn *dma_drain_needed;
409 lld_busy_fn *lld_busy_fn;
411 struct blk_mq_ops *mq_ops;
413 unsigned int *mq_map;
416 struct blk_mq_ctx __percpu *queue_ctx;
417 unsigned int nr_queues;
419 unsigned int queue_depth;
421 /* hw dispatch queues */
422 struct blk_mq_hw_ctx **queue_hw_ctx;
423 unsigned int nr_hw_queues;
426 * Dispatch queue sorting
429 struct request *boundary_rq;
432 * Delayed queue handling
434 struct delayed_work delay_work;
436 struct backing_dev_info backing_dev_info;
439 * The queue owner gets to use this for whatever they like.
440 * ll_rw_blk doesn't touch it.
445 * various queue flags, see QUEUE_* below
447 unsigned long queue_flags;
450 * ida allocated id for this queue. Used to index queues from
456 * queue needs bounce pages for pages above this limit
461 * protects queue structures from reentrancy. ->__queue_lock should
462 * _never_ be used directly, it is queue private. always use
465 spinlock_t __queue_lock;
466 spinlock_t *queue_lock;
476 struct kobject mq_kobj;
478 #ifdef CONFIG_BLK_DEV_INTEGRITY
479 struct blk_integrity integrity;
480 #endif /* CONFIG_BLK_DEV_INTEGRITY */
485 unsigned int nr_pending;
491 unsigned long nr_requests; /* Max # of requests */
492 unsigned int nr_congestion_on;
493 unsigned int nr_congestion_off;
494 unsigned int nr_batching;
496 unsigned int dma_drain_size;
497 void *dma_drain_buffer;
498 unsigned int dma_pad_mask;
499 unsigned int dma_alignment;
501 struct blk_queue_tag *queue_tags;
502 struct list_head tag_busy_list;
504 unsigned int nr_sorted;
505 unsigned int in_flight[2];
507 struct blk_rq_stat rq_stats[2];
510 * Number of active block driver functions for which blk_drain_queue()
511 * must wait. Must be incremented around functions that unlock the
512 * queue_lock internally, e.g. scsi_request_fn().
514 unsigned int request_fn_active;
516 unsigned int rq_timeout;
518 struct timer_list timeout;
519 struct work_struct timeout_work;
520 struct list_head timeout_list;
522 struct list_head icq_list;
523 #ifdef CONFIG_BLK_CGROUP
524 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
525 struct blkcg_gq *root_blkg;
526 struct list_head blkg_list;
529 struct queue_limits limits;
534 unsigned int sg_timeout;
535 unsigned int sg_reserved_size;
537 #ifdef CONFIG_BLK_DEV_IO_TRACE
538 struct blk_trace *blk_trace;
541 * for flush operations
543 struct blk_flush_queue *fq;
545 struct list_head requeue_list;
546 spinlock_t requeue_lock;
547 struct delayed_work requeue_work;
549 struct mutex sysfs_lock;
552 atomic_t mq_freeze_depth;
554 #if defined(CONFIG_BLK_DEV_BSG)
555 bsg_job_fn *bsg_job_fn;
557 struct bsg_class_device bsg_dev;
560 #ifdef CONFIG_BLK_DEV_THROTTLING
562 struct throtl_data *td;
564 struct rcu_head rcu_head;
565 wait_queue_head_t mq_freeze_wq;
566 struct percpu_ref q_usage_counter;
567 struct list_head all_q_node;
569 struct blk_mq_tag_set *tag_set;
570 struct list_head tag_set_list;
571 struct bio_set *bio_split;
573 bool mq_sysfs_init_done;
576 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
577 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
578 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
579 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
580 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
581 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
582 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
583 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
584 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
585 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
586 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
587 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
588 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
589 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
590 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
591 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
592 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
593 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
594 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
595 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
596 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
597 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
598 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
599 #define QUEUE_FLAG_WC 23 /* Write back caching */
600 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
601 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
602 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
603 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
605 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
606 (1 << QUEUE_FLAG_STACKABLE) | \
607 (1 << QUEUE_FLAG_SAME_COMP) | \
608 (1 << QUEUE_FLAG_ADD_RANDOM))
610 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
611 (1 << QUEUE_FLAG_STACKABLE) | \
612 (1 << QUEUE_FLAG_SAME_COMP) | \
613 (1 << QUEUE_FLAG_POLL))
615 static inline void queue_lockdep_assert_held(struct request_queue *q)
618 lockdep_assert_held(q->queue_lock);
621 static inline void queue_flag_set_unlocked(unsigned int flag,
622 struct request_queue *q)
624 __set_bit(flag, &q->queue_flags);
627 static inline int queue_flag_test_and_clear(unsigned int flag,
628 struct request_queue *q)
630 queue_lockdep_assert_held(q);
632 if (test_bit(flag, &q->queue_flags)) {
633 __clear_bit(flag, &q->queue_flags);
640 static inline int queue_flag_test_and_set(unsigned int flag,
641 struct request_queue *q)
643 queue_lockdep_assert_held(q);
645 if (!test_bit(flag, &q->queue_flags)) {
646 __set_bit(flag, &q->queue_flags);
653 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
655 queue_lockdep_assert_held(q);
656 __set_bit(flag, &q->queue_flags);
659 static inline void queue_flag_clear_unlocked(unsigned int flag,
660 struct request_queue *q)
662 __clear_bit(flag, &q->queue_flags);
665 static inline int queue_in_flight(struct request_queue *q)
667 return q->in_flight[0] + q->in_flight[1];
670 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
672 queue_lockdep_assert_held(q);
673 __clear_bit(flag, &q->queue_flags);
676 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
677 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
678 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
679 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
680 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
681 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
682 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
683 #define blk_queue_noxmerges(q) \
684 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
685 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
686 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
687 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
688 #define blk_queue_stackable(q) \
689 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
690 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
691 #define blk_queue_secure_erase(q) \
692 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
693 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
695 #define blk_noretry_request(rq) \
696 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
697 REQ_FAILFAST_DRIVER))
699 #define blk_account_rq(rq) \
700 (((rq)->rq_flags & RQF_STARTED) && \
701 ((rq)->cmd_type == REQ_TYPE_FS))
703 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
704 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
705 /* rq->queuelist of dequeued request must be list_empty() */
706 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
708 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
710 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
713 * Driver can handle struct request, if it either has an old style
714 * request_fn defined, or is blk-mq based.
716 static inline bool queue_is_rq_based(struct request_queue *q)
718 return q->request_fn || q->mq_ops;
721 static inline unsigned int blk_queue_cluster(struct request_queue *q)
723 return q->limits.cluster;
726 static inline enum blk_zoned_model
727 blk_queue_zoned_model(struct request_queue *q)
729 return q->limits.zoned;
732 static inline bool blk_queue_is_zoned(struct request_queue *q)
734 switch (blk_queue_zoned_model(q)) {
743 static inline unsigned int blk_queue_zone_size(struct request_queue *q)
745 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
748 static inline bool rq_is_sync(struct request *rq)
750 return op_is_sync(rq->cmd_flags);
753 static inline bool blk_rl_full(struct request_list *rl, bool sync)
755 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
757 return rl->flags & flag;
760 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
762 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
767 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
769 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
774 static inline bool rq_mergeable(struct request *rq)
776 if (rq->cmd_type != REQ_TYPE_FS)
779 if (req_op(rq) == REQ_OP_FLUSH)
782 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
785 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
787 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
793 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
795 if (bio_data(a) == bio_data(b))
801 static inline unsigned int blk_queue_depth(struct request_queue *q)
804 return q->queue_depth;
806 return q->nr_requests;
810 * q->prep_rq_fn return values
813 BLKPREP_OK, /* serve it */
814 BLKPREP_KILL, /* fatal error, kill, return -EIO */
815 BLKPREP_DEFER, /* leave on queue */
816 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
819 extern unsigned long blk_max_low_pfn, blk_max_pfn;
822 * standard bounce addresses:
824 * BLK_BOUNCE_HIGH : bounce all highmem pages
825 * BLK_BOUNCE_ANY : don't bounce anything
826 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
829 #if BITS_PER_LONG == 32
830 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
832 #define BLK_BOUNCE_HIGH -1ULL
834 #define BLK_BOUNCE_ANY (-1ULL)
835 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
838 * default timeout for SG_IO if none specified
840 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
841 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
844 extern int init_emergency_isa_pool(void);
845 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
847 static inline int init_emergency_isa_pool(void)
851 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
854 #endif /* CONFIG_MMU */
860 unsigned long offset;
865 struct req_iterator {
866 struct bvec_iter iter;
870 /* This should not be used directly - use rq_for_each_segment */
871 #define for_each_bio(_bio) \
872 for (; _bio; _bio = _bio->bi_next)
873 #define __rq_for_each_bio(_bio, rq) \
875 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
877 #define rq_for_each_segment(bvl, _rq, _iter) \
878 __rq_for_each_bio(_iter.bio, _rq) \
879 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
881 #define rq_iter_last(bvec, _iter) \
882 (_iter.bio->bi_next == NULL && \
883 bio_iter_last(bvec, _iter.iter))
885 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
886 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
888 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
889 extern void rq_flush_dcache_pages(struct request *rq);
891 static inline void rq_flush_dcache_pages(struct request *rq)
897 #define vfs_msg(sb, level, fmt, ...) \
898 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
900 #define vfs_msg(sb, level, fmt, ...) \
902 no_printk(fmt, ##__VA_ARGS__); \
903 __vfs_msg(sb, "", " "); \
907 extern int blk_register_queue(struct gendisk *disk);
908 extern void blk_unregister_queue(struct gendisk *disk);
909 extern blk_qc_t generic_make_request(struct bio *bio);
910 extern void blk_rq_init(struct request_queue *q, struct request *rq);
911 extern void blk_put_request(struct request *);
912 extern void __blk_put_request(struct request_queue *, struct request *);
913 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
914 extern void blk_rq_set_block_pc(struct request *);
915 extern void blk_requeue_request(struct request_queue *, struct request *);
916 extern int blk_lld_busy(struct request_queue *q);
917 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
918 struct bio_set *bs, gfp_t gfp_mask,
919 int (*bio_ctr)(struct bio *, struct bio *, void *),
921 extern void blk_rq_unprep_clone(struct request *rq);
922 extern int blk_insert_cloned_request(struct request_queue *q,
924 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
925 extern void blk_delay_queue(struct request_queue *, unsigned long);
926 extern void blk_queue_split(struct request_queue *, struct bio **,
928 extern void blk_recount_segments(struct request_queue *, struct bio *);
929 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
930 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
931 unsigned int, void __user *);
932 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
933 unsigned int, void __user *);
934 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
935 struct scsi_ioctl_command __user *);
937 extern int blk_queue_enter(struct request_queue *q, bool nowait);
938 extern void blk_queue_exit(struct request_queue *q);
939 extern void blk_start_queue(struct request_queue *q);
940 extern void blk_start_queue_async(struct request_queue *q);
941 extern void blk_stop_queue(struct request_queue *q);
942 extern void blk_sync_queue(struct request_queue *q);
943 extern void __blk_stop_queue(struct request_queue *q);
944 extern void __blk_run_queue(struct request_queue *q);
945 extern void __blk_run_queue_uncond(struct request_queue *q);
946 extern void blk_run_queue(struct request_queue *);
947 extern void blk_run_queue_async(struct request_queue *q);
948 extern void blk_mq_quiesce_queue(struct request_queue *q);
949 extern int blk_rq_map_user(struct request_queue *, struct request *,
950 struct rq_map_data *, void __user *, unsigned long,
952 extern int blk_rq_unmap_user(struct bio *);
953 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
954 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
955 struct rq_map_data *, const struct iov_iter *,
957 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
958 struct request *, int);
959 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
960 struct request *, int, rq_end_io_fn *);
962 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
964 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
966 return bdev->bd_disk->queue; /* this is never NULL */
970 * blk_rq_pos() : the current sector
971 * blk_rq_bytes() : bytes left in the entire request
972 * blk_rq_cur_bytes() : bytes left in the current segment
973 * blk_rq_err_bytes() : bytes left till the next error boundary
974 * blk_rq_sectors() : sectors left in the entire request
975 * blk_rq_cur_sectors() : sectors left in the current segment
977 static inline sector_t blk_rq_pos(const struct request *rq)
982 static inline unsigned int blk_rq_bytes(const struct request *rq)
984 return rq->__data_len;
987 static inline int blk_rq_cur_bytes(const struct request *rq)
989 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
992 extern unsigned int blk_rq_err_bytes(const struct request *rq);
994 static inline unsigned int blk_rq_sectors(const struct request *rq)
996 return blk_rq_bytes(rq) >> 9;
999 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1001 return blk_rq_cur_bytes(rq) >> 9;
1004 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1007 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1008 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1010 if (unlikely(op == REQ_OP_WRITE_SAME))
1011 return q->limits.max_write_same_sectors;
1013 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1014 return q->limits.max_write_zeroes_sectors;
1016 return q->limits.max_sectors;
1020 * Return maximum size of a request at given offset. Only valid for
1021 * file system requests.
1023 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1026 if (!q->limits.chunk_sectors)
1027 return q->limits.max_sectors;
1029 return q->limits.chunk_sectors -
1030 (offset & (q->limits.chunk_sectors - 1));
1033 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1036 struct request_queue *q = rq->q;
1038 if (unlikely(rq->cmd_type != REQ_TYPE_FS))
1039 return q->limits.max_hw_sectors;
1041 if (!q->limits.chunk_sectors ||
1042 req_op(rq) == REQ_OP_DISCARD ||
1043 req_op(rq) == REQ_OP_SECURE_ERASE)
1044 return blk_queue_get_max_sectors(q, req_op(rq));
1046 return min(blk_max_size_offset(q, offset),
1047 blk_queue_get_max_sectors(q, req_op(rq)));
1050 static inline unsigned int blk_rq_count_bios(struct request *rq)
1052 unsigned int nr_bios = 0;
1055 __rq_for_each_bio(bio, rq)
1062 * blk_rq_set_prio - associate a request with prio from ioc
1063 * @rq: request of interest
1064 * @ioc: target iocontext
1066 * Assocate request prio with ioc prio so request based drivers
1067 * can leverage priority information.
1069 static inline void blk_rq_set_prio(struct request *rq, struct io_context *ioc)
1072 rq->ioprio = ioc->ioprio;
1076 * Request issue related functions.
1078 extern struct request *blk_peek_request(struct request_queue *q);
1079 extern void blk_start_request(struct request *rq);
1080 extern struct request *blk_fetch_request(struct request_queue *q);
1083 * Request completion related functions.
1085 * blk_update_request() completes given number of bytes and updates
1086 * the request without completing it.
1088 * blk_end_request() and friends. __blk_end_request() must be called
1089 * with the request queue spinlock acquired.
1091 * Several drivers define their own end_request and call
1092 * blk_end_request() for parts of the original function.
1093 * This prevents code duplication in drivers.
1095 extern bool blk_update_request(struct request *rq, int error,
1096 unsigned int nr_bytes);
1097 extern void blk_finish_request(struct request *rq, int error);
1098 extern bool blk_end_request(struct request *rq, int error,
1099 unsigned int nr_bytes);
1100 extern void blk_end_request_all(struct request *rq, int error);
1101 extern bool blk_end_request_cur(struct request *rq, int error);
1102 extern bool blk_end_request_err(struct request *rq, int error);
1103 extern bool __blk_end_request(struct request *rq, int error,
1104 unsigned int nr_bytes);
1105 extern void __blk_end_request_all(struct request *rq, int error);
1106 extern bool __blk_end_request_cur(struct request *rq, int error);
1107 extern bool __blk_end_request_err(struct request *rq, int error);
1109 extern void blk_complete_request(struct request *);
1110 extern void __blk_complete_request(struct request *);
1111 extern void blk_abort_request(struct request *);
1112 extern void blk_unprep_request(struct request *);
1115 * Access functions for manipulating queue properties
1117 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1118 spinlock_t *lock, int node_id);
1119 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1120 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
1121 request_fn_proc *, spinlock_t *);
1122 extern void blk_cleanup_queue(struct request_queue *);
1123 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1124 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1125 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1126 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1127 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1128 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1129 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1130 unsigned int max_discard_sectors);
1131 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1132 unsigned int max_write_same_sectors);
1133 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1134 unsigned int max_write_same_sectors);
1135 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1136 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1137 extern void blk_queue_alignment_offset(struct request_queue *q,
1138 unsigned int alignment);
1139 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1140 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1141 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1142 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1143 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1144 extern void blk_set_default_limits(struct queue_limits *lim);
1145 extern void blk_set_stacking_limits(struct queue_limits *lim);
1146 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1148 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1150 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1152 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1153 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1154 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1155 extern int blk_queue_dma_drain(struct request_queue *q,
1156 dma_drain_needed_fn *dma_drain_needed,
1157 void *buf, unsigned int size);
1158 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1159 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1160 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1161 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1162 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1163 extern void blk_queue_dma_alignment(struct request_queue *, int);
1164 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1165 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1166 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1167 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1168 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1169 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1170 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1172 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1174 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1176 return rq->nr_phys_segments;
1179 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1180 extern void blk_dump_rq_flags(struct request *, char *);
1181 extern long nr_blockdev_pages(void);
1183 bool __must_check blk_get_queue(struct request_queue *);
1184 struct request_queue *blk_alloc_queue(gfp_t);
1185 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1186 extern void blk_put_queue(struct request_queue *);
1187 extern void blk_set_queue_dying(struct request_queue *);
1190 * block layer runtime pm functions
1193 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1194 extern int blk_pre_runtime_suspend(struct request_queue *q);
1195 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1196 extern void blk_pre_runtime_resume(struct request_queue *q);
1197 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1198 extern void blk_set_runtime_active(struct request_queue *q);
1200 static inline void blk_pm_runtime_init(struct request_queue *q,
1201 struct device *dev) {}
1202 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1206 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1207 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1208 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1209 static inline void blk_set_runtime_active(struct request_queue *q) {}
1213 * blk_plug permits building a queue of related requests by holding the I/O
1214 * fragments for a short period. This allows merging of sequential requests
1215 * into single larger request. As the requests are moved from a per-task list to
1216 * the device's request_queue in a batch, this results in improved scalability
1217 * as the lock contention for request_queue lock is reduced.
1219 * It is ok not to disable preemption when adding the request to the plug list
1220 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1221 * the plug list when the task sleeps by itself. For details, please see
1222 * schedule() where blk_schedule_flush_plug() is called.
1225 struct list_head list; /* requests */
1226 struct list_head mq_list; /* blk-mq requests */
1227 struct list_head cb_list; /* md requires an unplug callback */
1229 #define BLK_MAX_REQUEST_COUNT 16
1230 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1233 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1234 struct blk_plug_cb {
1235 struct list_head list;
1236 blk_plug_cb_fn callback;
1239 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1240 void *data, int size);
1241 extern void blk_start_plug(struct blk_plug *);
1242 extern void blk_finish_plug(struct blk_plug *);
1243 extern void blk_flush_plug_list(struct blk_plug *, bool);
1245 static inline void blk_flush_plug(struct task_struct *tsk)
1247 struct blk_plug *plug = tsk->plug;
1250 blk_flush_plug_list(plug, false);
1253 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1255 struct blk_plug *plug = tsk->plug;
1258 blk_flush_plug_list(plug, true);
1261 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1263 struct blk_plug *plug = tsk->plug;
1266 (!list_empty(&plug->list) ||
1267 !list_empty(&plug->mq_list) ||
1268 !list_empty(&plug->cb_list));
1274 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1275 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1276 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1277 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1278 extern void blk_queue_free_tags(struct request_queue *);
1279 extern int blk_queue_resize_tags(struct request_queue *, int);
1280 extern void blk_queue_invalidate_tags(struct request_queue *);
1281 extern struct blk_queue_tag *blk_init_tags(int, int);
1282 extern void blk_free_tags(struct blk_queue_tag *);
1284 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1287 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1289 return bqt->tag_index[tag];
1293 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1294 #define BLKDEV_DISCARD_ZERO (1 << 1) /* must reliably zero data */
1296 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1297 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1298 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1299 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1300 sector_t nr_sects, gfp_t gfp_mask, int flags,
1302 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1303 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1304 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1305 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1307 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1308 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1309 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1310 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1312 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1313 nr_blocks << (sb->s_blocksize_bits - 9),
1316 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1317 sector_t nr_blocks, gfp_t gfp_mask)
1319 return blkdev_issue_zeroout(sb->s_bdev,
1320 block << (sb->s_blocksize_bits - 9),
1321 nr_blocks << (sb->s_blocksize_bits - 9),
1325 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1327 enum blk_default_limits {
1328 BLK_MAX_SEGMENTS = 128,
1329 BLK_SAFE_MAX_SECTORS = 255,
1330 BLK_DEF_MAX_SECTORS = 2560,
1331 BLK_MAX_SEGMENT_SIZE = 65536,
1332 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1335 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1337 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1339 return q->limits.bounce_pfn;
1342 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1344 return q->limits.seg_boundary_mask;
1347 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1349 return q->limits.virt_boundary_mask;
1352 static inline unsigned int queue_max_sectors(struct request_queue *q)
1354 return q->limits.max_sectors;
1357 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1359 return q->limits.max_hw_sectors;
1362 static inline unsigned short queue_max_segments(struct request_queue *q)
1364 return q->limits.max_segments;
1367 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1369 return q->limits.max_segment_size;
1372 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1376 if (q && q->limits.logical_block_size)
1377 retval = q->limits.logical_block_size;
1382 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1384 return queue_logical_block_size(bdev_get_queue(bdev));
1387 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1389 return q->limits.physical_block_size;
1392 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1394 return queue_physical_block_size(bdev_get_queue(bdev));
1397 static inline unsigned int queue_io_min(struct request_queue *q)
1399 return q->limits.io_min;
1402 static inline int bdev_io_min(struct block_device *bdev)
1404 return queue_io_min(bdev_get_queue(bdev));
1407 static inline unsigned int queue_io_opt(struct request_queue *q)
1409 return q->limits.io_opt;
1412 static inline int bdev_io_opt(struct block_device *bdev)
1414 return queue_io_opt(bdev_get_queue(bdev));
1417 static inline int queue_alignment_offset(struct request_queue *q)
1419 if (q->limits.misaligned)
1422 return q->limits.alignment_offset;
1425 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1427 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1428 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1430 return (granularity + lim->alignment_offset - alignment) % granularity;
1433 static inline int bdev_alignment_offset(struct block_device *bdev)
1435 struct request_queue *q = bdev_get_queue(bdev);
1437 if (q->limits.misaligned)
1440 if (bdev != bdev->bd_contains)
1441 return bdev->bd_part->alignment_offset;
1443 return q->limits.alignment_offset;
1446 static inline int queue_discard_alignment(struct request_queue *q)
1448 if (q->limits.discard_misaligned)
1451 return q->limits.discard_alignment;
1454 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1456 unsigned int alignment, granularity, offset;
1458 if (!lim->max_discard_sectors)
1461 /* Why are these in bytes, not sectors? */
1462 alignment = lim->discard_alignment >> 9;
1463 granularity = lim->discard_granularity >> 9;
1467 /* Offset of the partition start in 'granularity' sectors */
1468 offset = sector_div(sector, granularity);
1470 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1471 offset = (granularity + alignment - offset) % granularity;
1473 /* Turn it back into bytes, gaah */
1477 static inline int bdev_discard_alignment(struct block_device *bdev)
1479 struct request_queue *q = bdev_get_queue(bdev);
1481 if (bdev != bdev->bd_contains)
1482 return bdev->bd_part->discard_alignment;
1484 return q->limits.discard_alignment;
1487 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1489 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1495 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1497 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1500 static inline unsigned int bdev_write_same(struct block_device *bdev)
1502 struct request_queue *q = bdev_get_queue(bdev);
1505 return q->limits.max_write_same_sectors;
1510 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1512 struct request_queue *q = bdev_get_queue(bdev);
1515 return q->limits.max_write_zeroes_sectors;
1520 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1522 struct request_queue *q = bdev_get_queue(bdev);
1525 return blk_queue_zoned_model(q);
1527 return BLK_ZONED_NONE;
1530 static inline bool bdev_is_zoned(struct block_device *bdev)
1532 struct request_queue *q = bdev_get_queue(bdev);
1535 return blk_queue_is_zoned(q);
1540 static inline unsigned int bdev_zone_size(struct block_device *bdev)
1542 struct request_queue *q = bdev_get_queue(bdev);
1545 return blk_queue_zone_size(q);
1550 static inline int queue_dma_alignment(struct request_queue *q)
1552 return q ? q->dma_alignment : 511;
1555 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1558 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1559 return !(addr & alignment) && !(len & alignment);
1562 /* assumes size > 256 */
1563 static inline unsigned int blksize_bits(unsigned int size)
1565 unsigned int bits = 8;
1569 } while (size > 256);
1573 static inline unsigned int block_size(struct block_device *bdev)
1575 return bdev->bd_block_size;
1578 static inline bool queue_flush_queueable(struct request_queue *q)
1580 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1583 typedef struct {struct page *v;} Sector;
1585 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1587 static inline void put_dev_sector(Sector p)
1592 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1593 struct bio_vec *bprv, unsigned int offset)
1596 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1600 * Check if adding a bio_vec after bprv with offset would create a gap in
1601 * the SG list. Most drivers don't care about this, but some do.
1603 static inline bool bvec_gap_to_prev(struct request_queue *q,
1604 struct bio_vec *bprv, unsigned int offset)
1606 if (!queue_virt_boundary(q))
1608 return __bvec_gap_to_prev(q, bprv, offset);
1611 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1614 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1615 struct bio_vec pb, nb;
1617 bio_get_last_bvec(prev, &pb);
1618 bio_get_first_bvec(next, &nb);
1620 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1626 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1628 return bio_will_gap(req->q, req->biotail, bio);
1631 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1633 return bio_will_gap(req->q, bio, req->bio);
1636 int kblockd_schedule_work(struct work_struct *work);
1637 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1638 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1639 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1641 #ifdef CONFIG_BLK_CGROUP
1643 * This should not be using sched_clock(). A real patch is in progress
1644 * to fix this up, until that is in place we need to disable preemption
1645 * around sched_clock() in this function and set_io_start_time_ns().
1647 static inline void set_start_time_ns(struct request *req)
1650 req->start_time_ns = sched_clock();
1654 static inline void set_io_start_time_ns(struct request *req)
1657 req->io_start_time_ns = sched_clock();
1661 static inline uint64_t rq_start_time_ns(struct request *req)
1663 return req->start_time_ns;
1666 static inline uint64_t rq_io_start_time_ns(struct request *req)
1668 return req->io_start_time_ns;
1671 static inline void set_start_time_ns(struct request *req) {}
1672 static inline void set_io_start_time_ns(struct request *req) {}
1673 static inline uint64_t rq_start_time_ns(struct request *req)
1677 static inline uint64_t rq_io_start_time_ns(struct request *req)
1683 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1684 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1685 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1686 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1688 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1690 enum blk_integrity_flags {
1691 BLK_INTEGRITY_VERIFY = 1 << 0,
1692 BLK_INTEGRITY_GENERATE = 1 << 1,
1693 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1694 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1697 struct blk_integrity_iter {
1701 unsigned int data_size;
1702 unsigned short interval;
1703 const char *disk_name;
1706 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1708 struct blk_integrity_profile {
1709 integrity_processing_fn *generate_fn;
1710 integrity_processing_fn *verify_fn;
1714 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1715 extern void blk_integrity_unregister(struct gendisk *);
1716 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1717 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1718 struct scatterlist *);
1719 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1720 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1722 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1725 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1727 struct blk_integrity *bi = &disk->queue->integrity;
1736 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1738 return blk_get_integrity(bdev->bd_disk);
1741 static inline bool blk_integrity_rq(struct request *rq)
1743 return rq->cmd_flags & REQ_INTEGRITY;
1746 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1749 q->limits.max_integrity_segments = segs;
1752 static inline unsigned short
1753 queue_max_integrity_segments(struct request_queue *q)
1755 return q->limits.max_integrity_segments;
1758 static inline bool integrity_req_gap_back_merge(struct request *req,
1761 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1762 struct bio_integrity_payload *bip_next = bio_integrity(next);
1764 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1765 bip_next->bip_vec[0].bv_offset);
1768 static inline bool integrity_req_gap_front_merge(struct request *req,
1771 struct bio_integrity_payload *bip = bio_integrity(bio);
1772 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1774 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1775 bip_next->bip_vec[0].bv_offset);
1778 #else /* CONFIG_BLK_DEV_INTEGRITY */
1781 struct block_device;
1783 struct blk_integrity;
1785 static inline int blk_integrity_rq(struct request *rq)
1789 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1794 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1796 struct scatterlist *s)
1800 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1804 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1808 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1812 static inline void blk_integrity_register(struct gendisk *d,
1813 struct blk_integrity *b)
1816 static inline void blk_integrity_unregister(struct gendisk *d)
1819 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1823 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1827 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1833 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1840 static inline bool integrity_req_gap_back_merge(struct request *req,
1845 static inline bool integrity_req_gap_front_merge(struct request *req,
1851 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1854 * struct blk_dax_ctl - control and output parameters for ->direct_access
1855 * @sector: (input) offset relative to a block_device
1856 * @addr: (output) kernel virtual address for @sector populated by driver
1857 * @pfn: (output) page frame number for @addr populated by driver
1858 * @size: (input) number of bytes requested
1860 struct blk_dax_ctl {
1867 struct block_device_operations {
1868 int (*open) (struct block_device *, fmode_t);
1869 void (*release) (struct gendisk *, fmode_t);
1870 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1871 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1872 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1873 long (*direct_access)(struct block_device *, sector_t, void **, pfn_t *,
1875 unsigned int (*check_events) (struct gendisk *disk,
1876 unsigned int clearing);
1877 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1878 int (*media_changed) (struct gendisk *);
1879 void (*unlock_native_capacity) (struct gendisk *);
1880 int (*revalidate_disk) (struct gendisk *);
1881 int (*getgeo)(struct block_device *, struct hd_geometry *);
1882 /* this callback is with swap_lock and sometimes page table lock held */
1883 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1884 struct module *owner;
1885 const struct pr_ops *pr_ops;
1888 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1890 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1891 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1892 struct writeback_control *);
1893 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1894 extern int bdev_dax_supported(struct super_block *, int);
1895 extern bool bdev_dax_capable(struct block_device *);
1896 #else /* CONFIG_BLOCK */
1898 struct block_device;
1901 * stubs for when the block layer is configured out
1903 #define buffer_heads_over_limit 0
1905 static inline long nr_blockdev_pages(void)
1913 static inline void blk_start_plug(struct blk_plug *plug)
1917 static inline void blk_finish_plug(struct blk_plug *plug)
1921 static inline void blk_flush_plug(struct task_struct *task)
1925 static inline void blk_schedule_flush_plug(struct task_struct *task)
1930 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1935 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1936 sector_t *error_sector)
1941 #endif /* CONFIG_BLOCK */