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 */
158 sector_t __sector; /* sector cursor */
164 * The hash is used inside the scheduler, and killed once the
165 * request reaches the dispatch list. The ipi_list is only used
166 * to queue the request for softirq completion, which is long
167 * after the request has been unhashed (and even removed from
168 * the dispatch list).
171 struct hlist_node hash; /* merge hash */
172 struct list_head ipi_list;
176 * The rb_node is only used inside the io scheduler, requests
177 * are pruned when moved to the dispatch queue. So let the
178 * completion_data share space with the rb_node.
181 struct rb_node rb_node; /* sort/lookup */
182 struct bio_vec special_vec;
183 void *completion_data;
187 * Three pointers are available for the IO schedulers, if they need
188 * more they have to dynamically allocate it. Flush requests are
189 * never put on the IO scheduler. So let the flush fields share
190 * space with the elevator data.
200 struct list_head list;
201 rq_end_io_fn *saved_end_io;
205 struct gendisk *rq_disk;
206 struct hd_struct *part;
207 unsigned long start_time;
208 struct blk_issue_stat issue_stat;
209 #ifdef CONFIG_BLK_CGROUP
210 struct request_list *rl; /* rl this rq is alloced from */
211 unsigned long long start_time_ns;
212 unsigned long long io_start_time_ns; /* when passed to hardware */
214 /* Number of scatter-gather DMA addr+len pairs after
215 * physical address coalescing is performed.
217 unsigned short nr_phys_segments;
218 #if defined(CONFIG_BLK_DEV_INTEGRITY)
219 unsigned short nr_integrity_segments;
222 unsigned short ioprio;
226 void *special; /* opaque pointer available for LLD use */
231 * when request is used as a packet command carrier
233 unsigned char __cmd[BLK_MAX_CDB];
235 unsigned short cmd_len;
237 unsigned int extra_len; /* length of alignment and padding */
238 unsigned int sense_len;
239 unsigned int resid_len; /* residual count */
242 unsigned long deadline;
243 struct list_head timeout_list;
244 unsigned int timeout;
248 * completion callback.
250 rq_end_io_fn *end_io;
254 struct request *next_rq;
257 static inline unsigned short req_get_ioprio(struct request *req)
262 #include <linux/elevator.h>
264 struct blk_queue_ctx;
266 typedef void (request_fn_proc) (struct request_queue *q);
267 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
268 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
269 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
272 typedef void (softirq_done_fn)(struct request *);
273 typedef int (dma_drain_needed_fn)(struct request *);
274 typedef int (lld_busy_fn) (struct request_queue *q);
275 typedef int (bsg_job_fn) (struct bsg_job *);
277 enum blk_eh_timer_return {
283 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
285 enum blk_queue_state {
290 struct blk_queue_tag {
291 struct request **tag_index; /* map of busy tags */
292 unsigned long *tag_map; /* bit map of free/busy tags */
293 int max_depth; /* what we will send to device */
294 int real_max_depth; /* what the array can hold */
295 atomic_t refcnt; /* map can be shared */
296 int alloc_policy; /* tag allocation policy */
297 int next_tag; /* next tag */
299 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
300 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
302 #define BLK_SCSI_MAX_CMDS (256)
303 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
306 * Zoned block device models (zoned limit).
308 enum blk_zoned_model {
309 BLK_ZONED_NONE, /* Regular block device */
310 BLK_ZONED_HA, /* Host-aware zoned block device */
311 BLK_ZONED_HM, /* Host-managed zoned block device */
314 struct queue_limits {
315 unsigned long bounce_pfn;
316 unsigned long seg_boundary_mask;
317 unsigned long virt_boundary_mask;
319 unsigned int max_hw_sectors;
320 unsigned int max_dev_sectors;
321 unsigned int chunk_sectors;
322 unsigned int max_sectors;
323 unsigned int max_segment_size;
324 unsigned int physical_block_size;
325 unsigned int alignment_offset;
328 unsigned int max_discard_sectors;
329 unsigned int max_hw_discard_sectors;
330 unsigned int max_write_same_sectors;
331 unsigned int max_write_zeroes_sectors;
332 unsigned int discard_granularity;
333 unsigned int discard_alignment;
335 unsigned short logical_block_size;
336 unsigned short max_segments;
337 unsigned short max_integrity_segments;
339 unsigned char misaligned;
340 unsigned char discard_misaligned;
341 unsigned char cluster;
342 unsigned char discard_zeroes_data;
343 unsigned char raid_partial_stripes_expensive;
344 enum blk_zoned_model zoned;
347 #ifdef CONFIG_BLK_DEV_ZONED
349 struct blk_zone_report_hdr {
350 unsigned int nr_zones;
354 extern int blkdev_report_zones(struct block_device *bdev,
355 sector_t sector, struct blk_zone *zones,
356 unsigned int *nr_zones, gfp_t gfp_mask);
357 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
358 sector_t nr_sectors, gfp_t gfp_mask);
360 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
361 unsigned int cmd, unsigned long arg);
362 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
363 unsigned int cmd, unsigned long arg);
365 #else /* CONFIG_BLK_DEV_ZONED */
367 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
368 fmode_t mode, unsigned int cmd,
374 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
375 fmode_t mode, unsigned int cmd,
381 #endif /* CONFIG_BLK_DEV_ZONED */
383 struct request_queue {
385 * Together with queue_head for cacheline sharing
387 struct list_head queue_head;
388 struct request *last_merge;
389 struct elevator_queue *elevator;
390 int nr_rqs[2]; /* # allocated [a]sync rqs */
391 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
396 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
397 * is used, root blkg allocates from @q->root_rl and all other
398 * blkgs from their own blkg->rl. Which one to use should be
399 * determined using bio_request_list().
401 struct request_list root_rl;
403 request_fn_proc *request_fn;
404 make_request_fn *make_request_fn;
405 prep_rq_fn *prep_rq_fn;
406 unprep_rq_fn *unprep_rq_fn;
407 softirq_done_fn *softirq_done_fn;
408 rq_timed_out_fn *rq_timed_out_fn;
409 dma_drain_needed_fn *dma_drain_needed;
410 lld_busy_fn *lld_busy_fn;
412 const struct blk_mq_ops *mq_ops;
414 unsigned int *mq_map;
417 struct blk_mq_ctx __percpu *queue_ctx;
418 unsigned int nr_queues;
420 unsigned int queue_depth;
422 /* hw dispatch queues */
423 struct blk_mq_hw_ctx **queue_hw_ctx;
424 unsigned int nr_hw_queues;
427 * Dispatch queue sorting
430 struct request *boundary_rq;
433 * Delayed queue handling
435 struct delayed_work delay_work;
437 struct backing_dev_info backing_dev_info;
440 * The queue owner gets to use this for whatever they like.
441 * ll_rw_blk doesn't touch it.
446 * various queue flags, see QUEUE_* below
448 unsigned long queue_flags;
451 * ida allocated id for this queue. Used to index queues from
457 * queue needs bounce pages for pages above this limit
462 * protects queue structures from reentrancy. ->__queue_lock should
463 * _never_ be used directly, it is queue private. always use
466 spinlock_t __queue_lock;
467 spinlock_t *queue_lock;
477 struct kobject mq_kobj;
479 #ifdef CONFIG_BLK_DEV_INTEGRITY
480 struct blk_integrity integrity;
481 #endif /* CONFIG_BLK_DEV_INTEGRITY */
486 unsigned int nr_pending;
492 unsigned long nr_requests; /* Max # of requests */
493 unsigned int nr_congestion_on;
494 unsigned int nr_congestion_off;
495 unsigned int nr_batching;
497 unsigned int dma_drain_size;
498 void *dma_drain_buffer;
499 unsigned int dma_pad_mask;
500 unsigned int dma_alignment;
502 struct blk_queue_tag *queue_tags;
503 struct list_head tag_busy_list;
505 unsigned int nr_sorted;
506 unsigned int in_flight[2];
508 struct blk_rq_stat rq_stats[2];
511 * Number of active block driver functions for which blk_drain_queue()
512 * must wait. Must be incremented around functions that unlock the
513 * queue_lock internally, e.g. scsi_request_fn().
515 unsigned int request_fn_active;
517 unsigned int rq_timeout;
519 struct timer_list timeout;
520 struct work_struct timeout_work;
521 struct list_head timeout_list;
523 struct list_head icq_list;
524 #ifdef CONFIG_BLK_CGROUP
525 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
526 struct blkcg_gq *root_blkg;
527 struct list_head blkg_list;
530 struct queue_limits limits;
535 unsigned int sg_timeout;
536 unsigned int sg_reserved_size;
538 #ifdef CONFIG_BLK_DEV_IO_TRACE
539 struct blk_trace *blk_trace;
542 * for flush operations
544 struct blk_flush_queue *fq;
546 struct list_head requeue_list;
547 spinlock_t requeue_lock;
548 struct delayed_work requeue_work;
550 struct mutex sysfs_lock;
553 atomic_t mq_freeze_depth;
555 #if defined(CONFIG_BLK_DEV_BSG)
556 bsg_job_fn *bsg_job_fn;
558 struct bsg_class_device bsg_dev;
561 #ifdef CONFIG_BLK_DEV_THROTTLING
563 struct throtl_data *td;
565 struct rcu_head rcu_head;
566 wait_queue_head_t mq_freeze_wq;
567 struct percpu_ref q_usage_counter;
568 struct list_head all_q_node;
570 struct blk_mq_tag_set *tag_set;
571 struct list_head tag_set_list;
572 struct bio_set *bio_split;
574 bool mq_sysfs_init_done;
577 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
578 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
579 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
580 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
581 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
582 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
583 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
584 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
585 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
586 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
587 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
588 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
589 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
590 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
591 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
592 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
593 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
594 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
595 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
596 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
597 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
598 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
599 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
600 #define QUEUE_FLAG_WC 23 /* Write back caching */
601 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
602 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
603 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
604 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
606 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
607 (1 << QUEUE_FLAG_STACKABLE) | \
608 (1 << QUEUE_FLAG_SAME_COMP) | \
609 (1 << QUEUE_FLAG_ADD_RANDOM))
611 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
612 (1 << QUEUE_FLAG_STACKABLE) | \
613 (1 << QUEUE_FLAG_SAME_COMP) | \
614 (1 << QUEUE_FLAG_POLL))
616 static inline void queue_lockdep_assert_held(struct request_queue *q)
619 lockdep_assert_held(q->queue_lock);
622 static inline void queue_flag_set_unlocked(unsigned int flag,
623 struct request_queue *q)
625 __set_bit(flag, &q->queue_flags);
628 static inline int queue_flag_test_and_clear(unsigned int flag,
629 struct request_queue *q)
631 queue_lockdep_assert_held(q);
633 if (test_bit(flag, &q->queue_flags)) {
634 __clear_bit(flag, &q->queue_flags);
641 static inline int queue_flag_test_and_set(unsigned int flag,
642 struct request_queue *q)
644 queue_lockdep_assert_held(q);
646 if (!test_bit(flag, &q->queue_flags)) {
647 __set_bit(flag, &q->queue_flags);
654 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
656 queue_lockdep_assert_held(q);
657 __set_bit(flag, &q->queue_flags);
660 static inline void queue_flag_clear_unlocked(unsigned int flag,
661 struct request_queue *q)
663 __clear_bit(flag, &q->queue_flags);
666 static inline int queue_in_flight(struct request_queue *q)
668 return q->in_flight[0] + q->in_flight[1];
671 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
673 queue_lockdep_assert_held(q);
674 __clear_bit(flag, &q->queue_flags);
677 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
678 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
679 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
680 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
681 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
682 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
683 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
684 #define blk_queue_noxmerges(q) \
685 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
686 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
687 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
688 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
689 #define blk_queue_stackable(q) \
690 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
691 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
692 #define blk_queue_secure_erase(q) \
693 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
694 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
696 #define blk_noretry_request(rq) \
697 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
698 REQ_FAILFAST_DRIVER))
700 #define blk_account_rq(rq) \
701 (((rq)->rq_flags & RQF_STARTED) && \
702 ((rq)->cmd_type == REQ_TYPE_FS))
704 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
705 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
706 /* rq->queuelist of dequeued request must be list_empty() */
707 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
709 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
711 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
714 * Driver can handle struct request, if it either has an old style
715 * request_fn defined, or is blk-mq based.
717 static inline bool queue_is_rq_based(struct request_queue *q)
719 return q->request_fn || q->mq_ops;
722 static inline unsigned int blk_queue_cluster(struct request_queue *q)
724 return q->limits.cluster;
727 static inline enum blk_zoned_model
728 blk_queue_zoned_model(struct request_queue *q)
730 return q->limits.zoned;
733 static inline bool blk_queue_is_zoned(struct request_queue *q)
735 switch (blk_queue_zoned_model(q)) {
744 static inline unsigned int blk_queue_zone_size(struct request_queue *q)
746 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
749 static inline bool rq_is_sync(struct request *rq)
751 return op_is_sync(rq->cmd_flags);
754 static inline bool blk_rl_full(struct request_list *rl, bool sync)
756 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
758 return rl->flags & flag;
761 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
763 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
768 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
770 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
775 static inline bool rq_mergeable(struct request *rq)
777 if (rq->cmd_type != REQ_TYPE_FS)
780 if (req_op(rq) == REQ_OP_FLUSH)
783 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
786 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
788 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
794 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
796 if (bio_data(a) == bio_data(b))
802 static inline unsigned int blk_queue_depth(struct request_queue *q)
805 return q->queue_depth;
807 return q->nr_requests;
811 * q->prep_rq_fn return values
814 BLKPREP_OK, /* serve it */
815 BLKPREP_KILL, /* fatal error, kill, return -EIO */
816 BLKPREP_DEFER, /* leave on queue */
817 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
820 extern unsigned long blk_max_low_pfn, blk_max_pfn;
823 * standard bounce addresses:
825 * BLK_BOUNCE_HIGH : bounce all highmem pages
826 * BLK_BOUNCE_ANY : don't bounce anything
827 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
830 #if BITS_PER_LONG == 32
831 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
833 #define BLK_BOUNCE_HIGH -1ULL
835 #define BLK_BOUNCE_ANY (-1ULL)
836 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
839 * default timeout for SG_IO if none specified
841 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
842 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
845 extern int init_emergency_isa_pool(void);
846 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
848 static inline int init_emergency_isa_pool(void)
852 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
855 #endif /* CONFIG_MMU */
861 unsigned long offset;
866 struct req_iterator {
867 struct bvec_iter iter;
871 /* This should not be used directly - use rq_for_each_segment */
872 #define for_each_bio(_bio) \
873 for (; _bio; _bio = _bio->bi_next)
874 #define __rq_for_each_bio(_bio, rq) \
876 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
878 #define rq_for_each_segment(bvl, _rq, _iter) \
879 __rq_for_each_bio(_iter.bio, _rq) \
880 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
882 #define rq_iter_last(bvec, _iter) \
883 (_iter.bio->bi_next == NULL && \
884 bio_iter_last(bvec, _iter.iter))
886 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
887 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
889 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
890 extern void rq_flush_dcache_pages(struct request *rq);
892 static inline void rq_flush_dcache_pages(struct request *rq)
898 #define vfs_msg(sb, level, fmt, ...) \
899 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
901 #define vfs_msg(sb, level, fmt, ...) \
903 no_printk(fmt, ##__VA_ARGS__); \
904 __vfs_msg(sb, "", " "); \
908 extern int blk_register_queue(struct gendisk *disk);
909 extern void blk_unregister_queue(struct gendisk *disk);
910 extern blk_qc_t generic_make_request(struct bio *bio);
911 extern void blk_rq_init(struct request_queue *q, struct request *rq);
912 extern void blk_put_request(struct request *);
913 extern void __blk_put_request(struct request_queue *, struct request *);
914 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
915 extern void blk_rq_set_block_pc(struct request *);
916 extern void blk_requeue_request(struct request_queue *, struct request *);
917 extern int blk_lld_busy(struct request_queue *q);
918 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
919 struct bio_set *bs, gfp_t gfp_mask,
920 int (*bio_ctr)(struct bio *, struct bio *, void *),
922 extern void blk_rq_unprep_clone(struct request *rq);
923 extern int blk_insert_cloned_request(struct request_queue *q,
925 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
926 extern void blk_delay_queue(struct request_queue *, unsigned long);
927 extern void blk_queue_split(struct request_queue *, struct bio **,
929 extern void blk_recount_segments(struct request_queue *, struct bio *);
930 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
931 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
932 unsigned int, void __user *);
933 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
934 unsigned int, void __user *);
935 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
936 struct scsi_ioctl_command __user *);
938 extern int blk_queue_enter(struct request_queue *q, bool nowait);
939 extern void blk_queue_exit(struct request_queue *q);
940 extern void blk_start_queue(struct request_queue *q);
941 extern void blk_start_queue_async(struct request_queue *q);
942 extern void blk_stop_queue(struct request_queue *q);
943 extern void blk_sync_queue(struct request_queue *q);
944 extern void __blk_stop_queue(struct request_queue *q);
945 extern void __blk_run_queue(struct request_queue *q);
946 extern void __blk_run_queue_uncond(struct request_queue *q);
947 extern void blk_run_queue(struct request_queue *);
948 extern void blk_run_queue_async(struct request_queue *q);
949 extern void blk_mq_quiesce_queue(struct request_queue *q);
950 extern int blk_rq_map_user(struct request_queue *, struct request *,
951 struct rq_map_data *, void __user *, unsigned long,
953 extern int blk_rq_unmap_user(struct bio *);
954 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
955 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
956 struct rq_map_data *, const struct iov_iter *,
958 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
959 struct request *, int);
960 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
961 struct request *, int, rq_end_io_fn *);
963 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
965 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
967 return bdev->bd_disk->queue; /* this is never NULL */
971 * blk_rq_pos() : the current sector
972 * blk_rq_bytes() : bytes left in the entire request
973 * blk_rq_cur_bytes() : bytes left in the current segment
974 * blk_rq_err_bytes() : bytes left till the next error boundary
975 * blk_rq_sectors() : sectors left in the entire request
976 * blk_rq_cur_sectors() : sectors left in the current segment
978 static inline sector_t blk_rq_pos(const struct request *rq)
983 static inline unsigned int blk_rq_bytes(const struct request *rq)
985 return rq->__data_len;
988 static inline int blk_rq_cur_bytes(const struct request *rq)
990 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
993 extern unsigned int blk_rq_err_bytes(const struct request *rq);
995 static inline unsigned int blk_rq_sectors(const struct request *rq)
997 return blk_rq_bytes(rq) >> 9;
1000 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1002 return blk_rq_cur_bytes(rq) >> 9;
1005 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1008 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1009 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1011 if (unlikely(op == REQ_OP_WRITE_SAME))
1012 return q->limits.max_write_same_sectors;
1014 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1015 return q->limits.max_write_zeroes_sectors;
1017 return q->limits.max_sectors;
1021 * Return maximum size of a request at given offset. Only valid for
1022 * file system requests.
1024 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1027 if (!q->limits.chunk_sectors)
1028 return q->limits.max_sectors;
1030 return q->limits.chunk_sectors -
1031 (offset & (q->limits.chunk_sectors - 1));
1034 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1037 struct request_queue *q = rq->q;
1039 if (unlikely(rq->cmd_type != REQ_TYPE_FS))
1040 return q->limits.max_hw_sectors;
1042 if (!q->limits.chunk_sectors ||
1043 req_op(rq) == REQ_OP_DISCARD ||
1044 req_op(rq) == REQ_OP_SECURE_ERASE)
1045 return blk_queue_get_max_sectors(q, req_op(rq));
1047 return min(blk_max_size_offset(q, offset),
1048 blk_queue_get_max_sectors(q, req_op(rq)));
1051 static inline unsigned int blk_rq_count_bios(struct request *rq)
1053 unsigned int nr_bios = 0;
1056 __rq_for_each_bio(bio, rq)
1063 * blk_rq_set_prio - associate a request with prio from ioc
1064 * @rq: request of interest
1065 * @ioc: target iocontext
1067 * Assocate request prio with ioc prio so request based drivers
1068 * can leverage priority information.
1070 static inline void blk_rq_set_prio(struct request *rq, struct io_context *ioc)
1073 rq->ioprio = ioc->ioprio;
1077 * Request issue related functions.
1079 extern struct request *blk_peek_request(struct request_queue *q);
1080 extern void blk_start_request(struct request *rq);
1081 extern struct request *blk_fetch_request(struct request_queue *q);
1084 * Request completion related functions.
1086 * blk_update_request() completes given number of bytes and updates
1087 * the request without completing it.
1089 * blk_end_request() and friends. __blk_end_request() must be called
1090 * with the request queue spinlock acquired.
1092 * Several drivers define their own end_request and call
1093 * blk_end_request() for parts of the original function.
1094 * This prevents code duplication in drivers.
1096 extern bool blk_update_request(struct request *rq, int error,
1097 unsigned int nr_bytes);
1098 extern void blk_finish_request(struct request *rq, int error);
1099 extern bool blk_end_request(struct request *rq, int error,
1100 unsigned int nr_bytes);
1101 extern void blk_end_request_all(struct request *rq, int error);
1102 extern bool blk_end_request_cur(struct request *rq, int error);
1103 extern bool blk_end_request_err(struct request *rq, int error);
1104 extern bool __blk_end_request(struct request *rq, int error,
1105 unsigned int nr_bytes);
1106 extern void __blk_end_request_all(struct request *rq, int error);
1107 extern bool __blk_end_request_cur(struct request *rq, int error);
1108 extern bool __blk_end_request_err(struct request *rq, int error);
1110 extern void blk_complete_request(struct request *);
1111 extern void __blk_complete_request(struct request *);
1112 extern void blk_abort_request(struct request *);
1113 extern void blk_unprep_request(struct request *);
1116 * Access functions for manipulating queue properties
1118 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1119 spinlock_t *lock, int node_id);
1120 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1121 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
1122 request_fn_proc *, spinlock_t *);
1123 extern void blk_cleanup_queue(struct request_queue *);
1124 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1125 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1126 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1127 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1128 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1129 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1130 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1131 unsigned int max_discard_sectors);
1132 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1133 unsigned int max_write_same_sectors);
1134 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1135 unsigned int max_write_same_sectors);
1136 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1137 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1138 extern void blk_queue_alignment_offset(struct request_queue *q,
1139 unsigned int alignment);
1140 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1141 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1142 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1143 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1144 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1145 extern void blk_set_default_limits(struct queue_limits *lim);
1146 extern void blk_set_stacking_limits(struct queue_limits *lim);
1147 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1149 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1151 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1153 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1154 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1155 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1156 extern int blk_queue_dma_drain(struct request_queue *q,
1157 dma_drain_needed_fn *dma_drain_needed,
1158 void *buf, unsigned int size);
1159 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1160 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1161 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1162 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1163 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1164 extern void blk_queue_dma_alignment(struct request_queue *, int);
1165 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1166 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1167 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1168 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1169 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1170 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1171 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
1173 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1175 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1177 return rq->nr_phys_segments;
1180 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1181 extern void blk_dump_rq_flags(struct request *, char *);
1182 extern long nr_blockdev_pages(void);
1184 bool __must_check blk_get_queue(struct request_queue *);
1185 struct request_queue *blk_alloc_queue(gfp_t);
1186 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1187 extern void blk_put_queue(struct request_queue *);
1188 extern void blk_set_queue_dying(struct request_queue *);
1191 * block layer runtime pm functions
1194 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1195 extern int blk_pre_runtime_suspend(struct request_queue *q);
1196 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1197 extern void blk_pre_runtime_resume(struct request_queue *q);
1198 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1199 extern void blk_set_runtime_active(struct request_queue *q);
1201 static inline void blk_pm_runtime_init(struct request_queue *q,
1202 struct device *dev) {}
1203 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1207 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1208 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1209 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1210 static inline void blk_set_runtime_active(struct request_queue *q) {}
1214 * blk_plug permits building a queue of related requests by holding the I/O
1215 * fragments for a short period. This allows merging of sequential requests
1216 * into single larger request. As the requests are moved from a per-task list to
1217 * the device's request_queue in a batch, this results in improved scalability
1218 * as the lock contention for request_queue lock is reduced.
1220 * It is ok not to disable preemption when adding the request to the plug list
1221 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1222 * the plug list when the task sleeps by itself. For details, please see
1223 * schedule() where blk_schedule_flush_plug() is called.
1226 struct list_head list; /* requests */
1227 struct list_head mq_list; /* blk-mq requests */
1228 struct list_head cb_list; /* md requires an unplug callback */
1230 #define BLK_MAX_REQUEST_COUNT 16
1231 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1234 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1235 struct blk_plug_cb {
1236 struct list_head list;
1237 blk_plug_cb_fn callback;
1240 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1241 void *data, int size);
1242 extern void blk_start_plug(struct blk_plug *);
1243 extern void blk_finish_plug(struct blk_plug *);
1244 extern void blk_flush_plug_list(struct blk_plug *, bool);
1246 static inline void blk_flush_plug(struct task_struct *tsk)
1248 struct blk_plug *plug = tsk->plug;
1251 blk_flush_plug_list(plug, false);
1254 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1256 struct blk_plug *plug = tsk->plug;
1259 blk_flush_plug_list(plug, true);
1262 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1264 struct blk_plug *plug = tsk->plug;
1267 (!list_empty(&plug->list) ||
1268 !list_empty(&plug->mq_list) ||
1269 !list_empty(&plug->cb_list));
1275 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1276 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1277 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1278 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1279 extern void blk_queue_free_tags(struct request_queue *);
1280 extern int blk_queue_resize_tags(struct request_queue *, int);
1281 extern void blk_queue_invalidate_tags(struct request_queue *);
1282 extern struct blk_queue_tag *blk_init_tags(int, int);
1283 extern void blk_free_tags(struct blk_queue_tag *);
1285 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1288 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1290 return bqt->tag_index[tag];
1294 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1295 #define BLKDEV_DISCARD_ZERO (1 << 1) /* must reliably zero data */
1297 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1298 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1299 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1300 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1301 sector_t nr_sects, gfp_t gfp_mask, int flags,
1303 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1304 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1305 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1306 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1308 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1309 sector_t nr_sects, gfp_t gfp_mask, bool discard);
1310 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1311 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1313 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1314 nr_blocks << (sb->s_blocksize_bits - 9),
1317 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1318 sector_t nr_blocks, gfp_t gfp_mask)
1320 return blkdev_issue_zeroout(sb->s_bdev,
1321 block << (sb->s_blocksize_bits - 9),
1322 nr_blocks << (sb->s_blocksize_bits - 9),
1326 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1328 enum blk_default_limits {
1329 BLK_MAX_SEGMENTS = 128,
1330 BLK_SAFE_MAX_SECTORS = 255,
1331 BLK_DEF_MAX_SECTORS = 2560,
1332 BLK_MAX_SEGMENT_SIZE = 65536,
1333 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1336 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1338 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1340 return q->limits.bounce_pfn;
1343 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1345 return q->limits.seg_boundary_mask;
1348 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1350 return q->limits.virt_boundary_mask;
1353 static inline unsigned int queue_max_sectors(struct request_queue *q)
1355 return q->limits.max_sectors;
1358 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1360 return q->limits.max_hw_sectors;
1363 static inline unsigned short queue_max_segments(struct request_queue *q)
1365 return q->limits.max_segments;
1368 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1370 return q->limits.max_segment_size;
1373 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1377 if (q && q->limits.logical_block_size)
1378 retval = q->limits.logical_block_size;
1383 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1385 return queue_logical_block_size(bdev_get_queue(bdev));
1388 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1390 return q->limits.physical_block_size;
1393 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1395 return queue_physical_block_size(bdev_get_queue(bdev));
1398 static inline unsigned int queue_io_min(struct request_queue *q)
1400 return q->limits.io_min;
1403 static inline int bdev_io_min(struct block_device *bdev)
1405 return queue_io_min(bdev_get_queue(bdev));
1408 static inline unsigned int queue_io_opt(struct request_queue *q)
1410 return q->limits.io_opt;
1413 static inline int bdev_io_opt(struct block_device *bdev)
1415 return queue_io_opt(bdev_get_queue(bdev));
1418 static inline int queue_alignment_offset(struct request_queue *q)
1420 if (q->limits.misaligned)
1423 return q->limits.alignment_offset;
1426 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1428 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1429 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1431 return (granularity + lim->alignment_offset - alignment) % granularity;
1434 static inline int bdev_alignment_offset(struct block_device *bdev)
1436 struct request_queue *q = bdev_get_queue(bdev);
1438 if (q->limits.misaligned)
1441 if (bdev != bdev->bd_contains)
1442 return bdev->bd_part->alignment_offset;
1444 return q->limits.alignment_offset;
1447 static inline int queue_discard_alignment(struct request_queue *q)
1449 if (q->limits.discard_misaligned)
1452 return q->limits.discard_alignment;
1455 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1457 unsigned int alignment, granularity, offset;
1459 if (!lim->max_discard_sectors)
1462 /* Why are these in bytes, not sectors? */
1463 alignment = lim->discard_alignment >> 9;
1464 granularity = lim->discard_granularity >> 9;
1468 /* Offset of the partition start in 'granularity' sectors */
1469 offset = sector_div(sector, granularity);
1471 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1472 offset = (granularity + alignment - offset) % granularity;
1474 /* Turn it back into bytes, gaah */
1478 static inline int bdev_discard_alignment(struct block_device *bdev)
1480 struct request_queue *q = bdev_get_queue(bdev);
1482 if (bdev != bdev->bd_contains)
1483 return bdev->bd_part->discard_alignment;
1485 return q->limits.discard_alignment;
1488 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1490 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1496 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1498 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1501 static inline unsigned int bdev_write_same(struct block_device *bdev)
1503 struct request_queue *q = bdev_get_queue(bdev);
1506 return q->limits.max_write_same_sectors;
1511 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1513 struct request_queue *q = bdev_get_queue(bdev);
1516 return q->limits.max_write_zeroes_sectors;
1521 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1523 struct request_queue *q = bdev_get_queue(bdev);
1526 return blk_queue_zoned_model(q);
1528 return BLK_ZONED_NONE;
1531 static inline bool bdev_is_zoned(struct block_device *bdev)
1533 struct request_queue *q = bdev_get_queue(bdev);
1536 return blk_queue_is_zoned(q);
1541 static inline unsigned int bdev_zone_size(struct block_device *bdev)
1543 struct request_queue *q = bdev_get_queue(bdev);
1546 return blk_queue_zone_size(q);
1551 static inline int queue_dma_alignment(struct request_queue *q)
1553 return q ? q->dma_alignment : 511;
1556 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1559 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1560 return !(addr & alignment) && !(len & alignment);
1563 /* assumes size > 256 */
1564 static inline unsigned int blksize_bits(unsigned int size)
1566 unsigned int bits = 8;
1570 } while (size > 256);
1574 static inline unsigned int block_size(struct block_device *bdev)
1576 return bdev->bd_block_size;
1579 static inline bool queue_flush_queueable(struct request_queue *q)
1581 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1584 typedef struct {struct page *v;} Sector;
1586 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1588 static inline void put_dev_sector(Sector p)
1593 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1594 struct bio_vec *bprv, unsigned int offset)
1597 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1601 * Check if adding a bio_vec after bprv with offset would create a gap in
1602 * the SG list. Most drivers don't care about this, but some do.
1604 static inline bool bvec_gap_to_prev(struct request_queue *q,
1605 struct bio_vec *bprv, unsigned int offset)
1607 if (!queue_virt_boundary(q))
1609 return __bvec_gap_to_prev(q, bprv, offset);
1613 * Check if the two bvecs from two bios can be merged to one segment.
1614 * If yes, no need to check gap between the two bios since the 1st bio
1615 * and the 1st bvec in the 2nd bio can be handled in one segment.
1617 static inline bool bios_segs_mergeable(struct request_queue *q,
1618 struct bio *prev, struct bio_vec *prev_last_bv,
1619 struct bio_vec *next_first_bv)
1621 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1623 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1625 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1626 queue_max_segment_size(q))
1631 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev,
1634 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1635 struct bio_vec pb, nb;
1637 bio_get_last_bvec(prev, &pb);
1638 bio_get_first_bvec(next, &nb);
1640 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1641 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1647 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1649 return bio_will_gap(req->q, req->biotail, bio);
1652 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1654 return bio_will_gap(req->q, bio, req->bio);
1657 int kblockd_schedule_work(struct work_struct *work);
1658 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1659 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1660 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1662 #ifdef CONFIG_BLK_CGROUP
1664 * This should not be using sched_clock(). A real patch is in progress
1665 * to fix this up, until that is in place we need to disable preemption
1666 * around sched_clock() in this function and set_io_start_time_ns().
1668 static inline void set_start_time_ns(struct request *req)
1671 req->start_time_ns = sched_clock();
1675 static inline void set_io_start_time_ns(struct request *req)
1678 req->io_start_time_ns = sched_clock();
1682 static inline uint64_t rq_start_time_ns(struct request *req)
1684 return req->start_time_ns;
1687 static inline uint64_t rq_io_start_time_ns(struct request *req)
1689 return req->io_start_time_ns;
1692 static inline void set_start_time_ns(struct request *req) {}
1693 static inline void set_io_start_time_ns(struct request *req) {}
1694 static inline uint64_t rq_start_time_ns(struct request *req)
1698 static inline uint64_t rq_io_start_time_ns(struct request *req)
1704 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1705 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1706 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1707 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1709 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1711 enum blk_integrity_flags {
1712 BLK_INTEGRITY_VERIFY = 1 << 0,
1713 BLK_INTEGRITY_GENERATE = 1 << 1,
1714 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1715 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1718 struct blk_integrity_iter {
1722 unsigned int data_size;
1723 unsigned short interval;
1724 const char *disk_name;
1727 typedef int (integrity_processing_fn) (struct blk_integrity_iter *);
1729 struct blk_integrity_profile {
1730 integrity_processing_fn *generate_fn;
1731 integrity_processing_fn *verify_fn;
1735 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1736 extern void blk_integrity_unregister(struct gendisk *);
1737 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1738 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1739 struct scatterlist *);
1740 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1741 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1743 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1746 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1748 struct blk_integrity *bi = &disk->queue->integrity;
1757 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1759 return blk_get_integrity(bdev->bd_disk);
1762 static inline bool blk_integrity_rq(struct request *rq)
1764 return rq->cmd_flags & REQ_INTEGRITY;
1767 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1770 q->limits.max_integrity_segments = segs;
1773 static inline unsigned short
1774 queue_max_integrity_segments(struct request_queue *q)
1776 return q->limits.max_integrity_segments;
1779 static inline bool integrity_req_gap_back_merge(struct request *req,
1782 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1783 struct bio_integrity_payload *bip_next = bio_integrity(next);
1785 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1786 bip_next->bip_vec[0].bv_offset);
1789 static inline bool integrity_req_gap_front_merge(struct request *req,
1792 struct bio_integrity_payload *bip = bio_integrity(bio);
1793 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1795 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1796 bip_next->bip_vec[0].bv_offset);
1799 #else /* CONFIG_BLK_DEV_INTEGRITY */
1802 struct block_device;
1804 struct blk_integrity;
1806 static inline int blk_integrity_rq(struct request *rq)
1810 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1815 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1817 struct scatterlist *s)
1821 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1825 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1829 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1833 static inline void blk_integrity_register(struct gendisk *d,
1834 struct blk_integrity *b)
1837 static inline void blk_integrity_unregister(struct gendisk *d)
1840 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1844 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1848 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1854 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1861 static inline bool integrity_req_gap_back_merge(struct request *req,
1866 static inline bool integrity_req_gap_front_merge(struct request *req,
1872 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1875 * struct blk_dax_ctl - control and output parameters for ->direct_access
1876 * @sector: (input) offset relative to a block_device
1877 * @addr: (output) kernel virtual address for @sector populated by driver
1878 * @pfn: (output) page frame number for @addr populated by driver
1879 * @size: (input) number of bytes requested
1881 struct blk_dax_ctl {
1888 struct block_device_operations {
1889 int (*open) (struct block_device *, fmode_t);
1890 void (*release) (struct gendisk *, fmode_t);
1891 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1892 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1893 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1894 long (*direct_access)(struct block_device *, sector_t, void **, pfn_t *,
1896 unsigned int (*check_events) (struct gendisk *disk,
1897 unsigned int clearing);
1898 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1899 int (*media_changed) (struct gendisk *);
1900 void (*unlock_native_capacity) (struct gendisk *);
1901 int (*revalidate_disk) (struct gendisk *);
1902 int (*getgeo)(struct block_device *, struct hd_geometry *);
1903 /* this callback is with swap_lock and sometimes page table lock held */
1904 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1905 struct module *owner;
1906 const struct pr_ops *pr_ops;
1909 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1911 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1912 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1913 struct writeback_control *);
1914 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *);
1915 extern int bdev_dax_supported(struct super_block *, int);
1916 extern bool bdev_dax_capable(struct block_device *);
1917 #else /* CONFIG_BLOCK */
1919 struct block_device;
1922 * stubs for when the block layer is configured out
1924 #define buffer_heads_over_limit 0
1926 static inline long nr_blockdev_pages(void)
1934 static inline void blk_start_plug(struct blk_plug *plug)
1938 static inline void blk_finish_plug(struct blk_plug *plug)
1942 static inline void blk_flush_plug(struct task_struct *task)
1946 static inline void blk_schedule_flush_plug(struct task_struct *task)
1951 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1956 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1957 sector_t *error_sector)
1962 #endif /* CONFIG_BLOCK */