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/timer.h>
12 #include <linux/workqueue.h>
13 #include <linux/pagemap.h>
14 #include <linux/backing-dev.h>
15 #include <linux/wait.h>
16 #include <linux/mempool.h>
17 #include <linux/bio.h>
18 #include <linux/stringify.h>
19 #include <linux/gfp.h>
20 #include <linux/bsg.h>
21 #include <linux/smp.h>
22 #include <linux/rcupdate.h>
24 #include <asm/scatterlist.h>
27 struct scsi_ioctl_command;
30 struct elevator_queue;
31 struct request_pm_state;
38 #define BLKDEV_MIN_RQ 4
39 #define BLKDEV_MAX_RQ 128 /* Default maximum */
42 * Maximum number of blkcg policies allowed to be registered concurrently.
43 * Defined here to simplify include dependency.
45 #define BLKCG_MAX_POLS 2
48 typedef void (rq_end_io_fn)(struct request *, int);
50 #define BLK_RL_SYNCFULL (1U << 0)
51 #define BLK_RL_ASYNCFULL (1U << 1)
54 struct request_queue *q; /* the queue this rl belongs to */
55 #ifdef CONFIG_BLK_CGROUP
56 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
59 * count[], starved[], and wait[] are indexed by
60 * BLK_RW_SYNC/BLK_RW_ASYNC
65 wait_queue_head_t wait[2];
70 * request command types
72 enum rq_cmd_type_bits {
73 REQ_TYPE_FS = 1, /* fs request */
74 REQ_TYPE_BLOCK_PC, /* scsi command */
75 REQ_TYPE_SENSE, /* sense request */
76 REQ_TYPE_PM_SUSPEND, /* suspend request */
77 REQ_TYPE_PM_RESUME, /* resume request */
78 REQ_TYPE_PM_SHUTDOWN, /* shutdown request */
79 REQ_TYPE_SPECIAL, /* driver defined type */
81 * for ATA/ATAPI devices. this really doesn't belong here, ide should
82 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
83 * private REQ_LB opcodes to differentiate what type of request this is
85 REQ_TYPE_ATA_TASKFILE,
89 #define BLK_MAX_CDB 16
92 * try to put the fields that are referenced together in the same cacheline.
93 * if you modify this structure, be sure to check block/blk-core.c:blk_rq_init()
97 struct list_head queuelist;
98 struct call_single_data csd;
100 struct request_queue *q;
102 unsigned int cmd_flags;
103 enum rq_cmd_type_bits cmd_type;
104 unsigned long atomic_flags;
108 /* the following two fields are internal, NEVER access directly */
109 unsigned int __data_len; /* total data len */
110 sector_t __sector; /* sector cursor */
115 struct hlist_node hash; /* merge hash */
117 * The rb_node is only used inside the io scheduler, requests
118 * are pruned when moved to the dispatch queue. So let the
119 * completion_data share space with the rb_node.
122 struct rb_node rb_node; /* sort/lookup */
123 void *completion_data;
127 * Three pointers are available for the IO schedulers, if they need
128 * more they have to dynamically allocate it. Flush requests are
129 * never put on the IO scheduler. So let the flush fields share
130 * space with the elevator data.
140 struct list_head list;
141 rq_end_io_fn *saved_end_io;
145 struct gendisk *rq_disk;
146 struct hd_struct *part;
147 unsigned long start_time;
148 #ifdef CONFIG_BLK_CGROUP
149 struct request_list *rl; /* rl this rq is alloced from */
150 unsigned long long start_time_ns;
151 unsigned long long io_start_time_ns; /* when passed to hardware */
153 /* Number of scatter-gather DMA addr+len pairs after
154 * physical address coalescing is performed.
156 unsigned short nr_phys_segments;
157 #if defined(CONFIG_BLK_DEV_INTEGRITY)
158 unsigned short nr_integrity_segments;
161 unsigned short ioprio;
165 void *special; /* opaque pointer available for LLD use */
166 char *buffer; /* kaddr of the current segment if available */
172 * when request is used as a packet command carrier
174 unsigned char __cmd[BLK_MAX_CDB];
176 unsigned short cmd_len;
178 unsigned int extra_len; /* length of alignment and padding */
179 unsigned int sense_len;
180 unsigned int resid_len; /* residual count */
183 unsigned long deadline;
184 struct list_head timeout_list;
185 unsigned int timeout;
189 * completion callback.
191 rq_end_io_fn *end_io;
195 struct request *next_rq;
198 static inline unsigned short req_get_ioprio(struct request *req)
204 * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
205 * requests. Some step values could eventually be made generic.
207 struct request_pm_state
209 /* PM state machine step value, currently driver specific */
211 /* requested PM state value (S1, S2, S3, S4, ...) */
213 void* data; /* for driver use */
216 #include <linux/elevator.h>
218 typedef void (request_fn_proc) (struct request_queue *q);
219 typedef void (make_request_fn) (struct request_queue *q, struct bio *bio);
220 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
221 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
224 struct bvec_merge_data {
225 struct block_device *bi_bdev;
230 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
232 typedef void (softirq_done_fn)(struct request *);
233 typedef int (dma_drain_needed_fn)(struct request *);
234 typedef int (lld_busy_fn) (struct request_queue *q);
235 typedef int (bsg_job_fn) (struct bsg_job *);
237 enum blk_eh_timer_return {
243 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
245 enum blk_queue_state {
250 struct blk_queue_tag {
251 struct request **tag_index; /* map of busy tags */
252 unsigned long *tag_map; /* bit map of free/busy tags */
253 int busy; /* current depth */
254 int max_depth; /* what we will send to device */
255 int real_max_depth; /* what the array can hold */
256 atomic_t refcnt; /* map can be shared */
259 #define BLK_SCSI_MAX_CMDS (256)
260 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
262 struct queue_limits {
263 unsigned long bounce_pfn;
264 unsigned long seg_boundary_mask;
266 unsigned int max_hw_sectors;
267 unsigned int max_sectors;
268 unsigned int max_segment_size;
269 unsigned int physical_block_size;
270 unsigned int alignment_offset;
273 unsigned int max_discard_sectors;
274 unsigned int max_write_same_sectors;
275 unsigned int discard_granularity;
276 unsigned int discard_alignment;
278 unsigned short logical_block_size;
279 unsigned short max_segments;
280 unsigned short max_integrity_segments;
282 unsigned char misaligned;
283 unsigned char discard_misaligned;
284 unsigned char cluster;
285 unsigned char discard_zeroes_data;
288 struct request_queue {
290 * Together with queue_head for cacheline sharing
292 struct list_head queue_head;
293 struct request *last_merge;
294 struct elevator_queue *elevator;
295 int nr_rqs[2]; /* # allocated [a]sync rqs */
296 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
299 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
300 * is used, root blkg allocates from @q->root_rl and all other
301 * blkgs from their own blkg->rl. Which one to use should be
302 * determined using bio_request_list().
304 struct request_list root_rl;
306 request_fn_proc *request_fn;
307 make_request_fn *make_request_fn;
308 prep_rq_fn *prep_rq_fn;
309 unprep_rq_fn *unprep_rq_fn;
310 merge_bvec_fn *merge_bvec_fn;
311 softirq_done_fn *softirq_done_fn;
312 rq_timed_out_fn *rq_timed_out_fn;
313 dma_drain_needed_fn *dma_drain_needed;
314 lld_busy_fn *lld_busy_fn;
317 * Dispatch queue sorting
320 struct request *boundary_rq;
323 * Delayed queue handling
325 struct delayed_work delay_work;
327 struct backing_dev_info backing_dev_info;
330 * The queue owner gets to use this for whatever they like.
331 * ll_rw_blk doesn't touch it.
336 * various queue flags, see QUEUE_* below
338 unsigned long queue_flags;
341 * ida allocated id for this queue. Used to index queues from
347 * queue needs bounce pages for pages above this limit
352 * protects queue structures from reentrancy. ->__queue_lock should
353 * _never_ be used directly, it is queue private. always use
356 spinlock_t __queue_lock;
357 spinlock_t *queue_lock;
367 unsigned long nr_requests; /* Max # of requests */
368 unsigned int nr_congestion_on;
369 unsigned int nr_congestion_off;
370 unsigned int nr_batching;
372 unsigned int dma_drain_size;
373 void *dma_drain_buffer;
374 unsigned int dma_pad_mask;
375 unsigned int dma_alignment;
377 struct blk_queue_tag *queue_tags;
378 struct list_head tag_busy_list;
380 unsigned int nr_sorted;
381 unsigned int in_flight[2];
383 * Number of active block driver functions for which blk_drain_queue()
384 * must wait. Must be incremented around functions that unlock the
385 * queue_lock internally, e.g. scsi_request_fn().
387 unsigned int request_fn_active;
389 unsigned int rq_timeout;
390 struct timer_list timeout;
391 struct list_head timeout_list;
393 struct list_head icq_list;
394 #ifdef CONFIG_BLK_CGROUP
395 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
396 struct blkcg_gq *root_blkg;
397 struct list_head blkg_list;
400 struct queue_limits limits;
405 unsigned int sg_timeout;
406 unsigned int sg_reserved_size;
408 #ifdef CONFIG_BLK_DEV_IO_TRACE
409 struct blk_trace *blk_trace;
412 * for flush operations
414 unsigned int flush_flags;
415 unsigned int flush_not_queueable:1;
416 unsigned int flush_queue_delayed:1;
417 unsigned int flush_pending_idx:1;
418 unsigned int flush_running_idx:1;
419 unsigned long flush_pending_since;
420 struct list_head flush_queue[2];
421 struct list_head flush_data_in_flight;
422 struct request flush_rq;
424 struct mutex sysfs_lock;
428 #if defined(CONFIG_BLK_DEV_BSG)
429 bsg_job_fn *bsg_job_fn;
431 struct bsg_class_device bsg_dev;
434 #ifdef CONFIG_BLK_CGROUP
435 struct list_head all_q_node;
437 #ifdef CONFIG_BLK_DEV_THROTTLING
439 struct throtl_data *td;
441 struct rcu_head rcu_head;
444 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
445 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
446 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
447 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
448 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
449 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
450 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
451 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
452 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
453 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
454 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
455 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
456 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
457 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
458 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
459 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
460 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
461 #define QUEUE_FLAG_SECDISCARD 17 /* supports SECDISCARD */
462 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
463 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
465 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
466 (1 << QUEUE_FLAG_STACKABLE) | \
467 (1 << QUEUE_FLAG_SAME_COMP) | \
468 (1 << QUEUE_FLAG_ADD_RANDOM))
470 static inline void queue_lockdep_assert_held(struct request_queue *q)
473 lockdep_assert_held(q->queue_lock);
476 static inline void queue_flag_set_unlocked(unsigned int flag,
477 struct request_queue *q)
479 __set_bit(flag, &q->queue_flags);
482 static inline int queue_flag_test_and_clear(unsigned int flag,
483 struct request_queue *q)
485 queue_lockdep_assert_held(q);
487 if (test_bit(flag, &q->queue_flags)) {
488 __clear_bit(flag, &q->queue_flags);
495 static inline int queue_flag_test_and_set(unsigned int flag,
496 struct request_queue *q)
498 queue_lockdep_assert_held(q);
500 if (!test_bit(flag, &q->queue_flags)) {
501 __set_bit(flag, &q->queue_flags);
508 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
510 queue_lockdep_assert_held(q);
511 __set_bit(flag, &q->queue_flags);
514 static inline void queue_flag_clear_unlocked(unsigned int flag,
515 struct request_queue *q)
517 __clear_bit(flag, &q->queue_flags);
520 static inline int queue_in_flight(struct request_queue *q)
522 return q->in_flight[0] + q->in_flight[1];
525 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
527 queue_lockdep_assert_held(q);
528 __clear_bit(flag, &q->queue_flags);
531 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
532 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
533 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
534 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
535 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
536 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
537 #define blk_queue_noxmerges(q) \
538 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
539 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
540 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
541 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
542 #define blk_queue_stackable(q) \
543 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
544 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
545 #define blk_queue_secdiscard(q) (blk_queue_discard(q) && \
546 test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags))
548 #define blk_noretry_request(rq) \
549 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
550 REQ_FAILFAST_DRIVER))
552 #define blk_account_rq(rq) \
553 (((rq)->cmd_flags & REQ_STARTED) && \
554 ((rq)->cmd_type == REQ_TYPE_FS))
556 #define blk_pm_request(rq) \
557 ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND || \
558 (rq)->cmd_type == REQ_TYPE_PM_RESUME)
560 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
561 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
562 /* rq->queuelist of dequeued request must be list_empty() */
563 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
565 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
567 #define rq_data_dir(rq) ((rq)->cmd_flags & 1)
569 static inline unsigned int blk_queue_cluster(struct request_queue *q)
571 return q->limits.cluster;
575 * We regard a request as sync, if either a read or a sync write
577 static inline bool rw_is_sync(unsigned int rw_flags)
579 return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC);
582 static inline bool rq_is_sync(struct request *rq)
584 return rw_is_sync(rq->cmd_flags);
587 static inline bool blk_rl_full(struct request_list *rl, bool sync)
589 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
591 return rl->flags & flag;
594 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
596 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
601 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
603 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
608 static inline bool rq_mergeable(struct request *rq)
610 if (rq->cmd_type != REQ_TYPE_FS)
613 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
619 static inline bool blk_check_merge_flags(unsigned int flags1,
622 if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD))
625 if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE))
628 if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME))
634 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
636 if (bio_data(a) == bio_data(b))
643 * q->prep_rq_fn return values
645 #define BLKPREP_OK 0 /* serve it */
646 #define BLKPREP_KILL 1 /* fatal error, kill */
647 #define BLKPREP_DEFER 2 /* leave on queue */
649 extern unsigned long blk_max_low_pfn, blk_max_pfn;
652 * standard bounce addresses:
654 * BLK_BOUNCE_HIGH : bounce all highmem pages
655 * BLK_BOUNCE_ANY : don't bounce anything
656 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
659 #if BITS_PER_LONG == 32
660 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
662 #define BLK_BOUNCE_HIGH -1ULL
664 #define BLK_BOUNCE_ANY (-1ULL)
665 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
668 * default timeout for SG_IO if none specified
670 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
671 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
674 extern int init_emergency_isa_pool(void);
675 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
677 static inline int init_emergency_isa_pool(void)
681 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
684 #endif /* CONFIG_MMU */
690 unsigned long offset;
695 struct req_iterator {
700 /* This should not be used directly - use rq_for_each_segment */
701 #define for_each_bio(_bio) \
702 for (; _bio; _bio = _bio->bi_next)
703 #define __rq_for_each_bio(_bio, rq) \
705 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
707 #define rq_for_each_segment(bvl, _rq, _iter) \
708 __rq_for_each_bio(_iter.bio, _rq) \
709 bio_for_each_segment(bvl, _iter.bio, _iter.i)
711 #define rq_iter_last(rq, _iter) \
712 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
714 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
715 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
717 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
718 extern void rq_flush_dcache_pages(struct request *rq);
720 static inline void rq_flush_dcache_pages(struct request *rq)
725 extern int blk_register_queue(struct gendisk *disk);
726 extern void blk_unregister_queue(struct gendisk *disk);
727 extern void generic_make_request(struct bio *bio);
728 extern void blk_rq_init(struct request_queue *q, struct request *rq);
729 extern void blk_put_request(struct request *);
730 extern void __blk_put_request(struct request_queue *, struct request *);
731 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
732 extern struct request *blk_make_request(struct request_queue *, struct bio *,
734 extern void blk_requeue_request(struct request_queue *, struct request *);
735 extern void blk_add_request_payload(struct request *rq, struct page *page,
737 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
738 extern int blk_lld_busy(struct request_queue *q);
739 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
740 struct bio_set *bs, gfp_t gfp_mask,
741 int (*bio_ctr)(struct bio *, struct bio *, void *),
743 extern void blk_rq_unprep_clone(struct request *rq);
744 extern int blk_insert_cloned_request(struct request_queue *q,
746 extern void blk_delay_queue(struct request_queue *, unsigned long);
747 extern void blk_recount_segments(struct request_queue *, struct bio *);
748 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
749 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
750 unsigned int, void __user *);
751 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
752 unsigned int, void __user *);
753 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
754 struct scsi_ioctl_command __user *);
756 extern void blk_queue_bio(struct request_queue *q, struct bio *bio);
759 * A queue has just exitted congestion. Note this in the global counter of
760 * congested queues, and wake up anyone who was waiting for requests to be
763 static inline void blk_clear_queue_congested(struct request_queue *q, int sync)
765 clear_bdi_congested(&q->backing_dev_info, sync);
769 * A queue has just entered congestion. Flag that in the queue's VM-visible
770 * state flags and increment the global gounter of congested queues.
772 static inline void blk_set_queue_congested(struct request_queue *q, int sync)
774 set_bdi_congested(&q->backing_dev_info, sync);
777 extern void blk_start_queue(struct request_queue *q);
778 extern void blk_stop_queue(struct request_queue *q);
779 extern void blk_sync_queue(struct request_queue *q);
780 extern void __blk_stop_queue(struct request_queue *q);
781 extern void __blk_run_queue(struct request_queue *q);
782 extern void blk_run_queue(struct request_queue *);
783 extern void blk_run_queue_async(struct request_queue *q);
784 extern int blk_rq_map_user(struct request_queue *, struct request *,
785 struct rq_map_data *, void __user *, unsigned long,
787 extern int blk_rq_unmap_user(struct bio *);
788 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
789 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
790 struct rq_map_data *, struct sg_iovec *, int,
791 unsigned int, gfp_t);
792 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
793 struct request *, int);
794 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
795 struct request *, int, rq_end_io_fn *);
797 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
799 return bdev->bd_disk->queue;
803 * blk_rq_pos() : the current sector
804 * blk_rq_bytes() : bytes left in the entire request
805 * blk_rq_cur_bytes() : bytes left in the current segment
806 * blk_rq_err_bytes() : bytes left till the next error boundary
807 * blk_rq_sectors() : sectors left in the entire request
808 * blk_rq_cur_sectors() : sectors left in the current segment
810 static inline sector_t blk_rq_pos(const struct request *rq)
815 static inline unsigned int blk_rq_bytes(const struct request *rq)
817 return rq->__data_len;
820 static inline int blk_rq_cur_bytes(const struct request *rq)
822 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
825 extern unsigned int blk_rq_err_bytes(const struct request *rq);
827 static inline unsigned int blk_rq_sectors(const struct request *rq)
829 return blk_rq_bytes(rq) >> 9;
832 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
834 return blk_rq_cur_bytes(rq) >> 9;
837 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
838 unsigned int cmd_flags)
840 if (unlikely(cmd_flags & REQ_DISCARD))
841 return q->limits.max_discard_sectors;
843 if (unlikely(cmd_flags & REQ_WRITE_SAME))
844 return q->limits.max_write_same_sectors;
846 return q->limits.max_sectors;
849 static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
851 struct request_queue *q = rq->q;
853 if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC))
854 return q->limits.max_hw_sectors;
856 return blk_queue_get_max_sectors(q, rq->cmd_flags);
860 * Request issue related functions.
862 extern struct request *blk_peek_request(struct request_queue *q);
863 extern void blk_start_request(struct request *rq);
864 extern struct request *blk_fetch_request(struct request_queue *q);
867 * Request completion related functions.
869 * blk_update_request() completes given number of bytes and updates
870 * the request without completing it.
872 * blk_end_request() and friends. __blk_end_request() must be called
873 * with the request queue spinlock acquired.
875 * Several drivers define their own end_request and call
876 * blk_end_request() for parts of the original function.
877 * This prevents code duplication in drivers.
879 extern bool blk_update_request(struct request *rq, int error,
880 unsigned int nr_bytes);
881 extern bool blk_end_request(struct request *rq, int error,
882 unsigned int nr_bytes);
883 extern void blk_end_request_all(struct request *rq, int error);
884 extern bool blk_end_request_cur(struct request *rq, int error);
885 extern bool blk_end_request_err(struct request *rq, int error);
886 extern bool __blk_end_request(struct request *rq, int error,
887 unsigned int nr_bytes);
888 extern void __blk_end_request_all(struct request *rq, int error);
889 extern bool __blk_end_request_cur(struct request *rq, int error);
890 extern bool __blk_end_request_err(struct request *rq, int error);
892 extern void blk_complete_request(struct request *);
893 extern void __blk_complete_request(struct request *);
894 extern void blk_abort_request(struct request *);
895 extern void blk_unprep_request(struct request *);
898 * Access functions for manipulating queue properties
900 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
901 spinlock_t *lock, int node_id);
902 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
903 extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
904 request_fn_proc *, spinlock_t *);
905 extern void blk_cleanup_queue(struct request_queue *);
906 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
907 extern void blk_queue_bounce_limit(struct request_queue *, u64);
908 extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int);
909 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
910 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
911 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
912 extern void blk_queue_max_discard_sectors(struct request_queue *q,
913 unsigned int max_discard_sectors);
914 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
915 unsigned int max_write_same_sectors);
916 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
917 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
918 extern void blk_queue_alignment_offset(struct request_queue *q,
919 unsigned int alignment);
920 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
921 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
922 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
923 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
924 extern void blk_set_default_limits(struct queue_limits *lim);
925 extern void blk_set_stacking_limits(struct queue_limits *lim);
926 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
928 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
930 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
932 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
933 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
934 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
935 extern int blk_queue_dma_drain(struct request_queue *q,
936 dma_drain_needed_fn *dma_drain_needed,
937 void *buf, unsigned int size);
938 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
939 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
940 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
941 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
942 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
943 extern void blk_queue_dma_alignment(struct request_queue *, int);
944 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
945 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
946 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
947 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
948 extern void blk_queue_flush(struct request_queue *q, unsigned int flush);
949 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
950 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
952 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
953 extern int blk_bio_map_sg(struct request_queue *q, struct bio *bio,
954 struct scatterlist *sglist);
955 extern void blk_dump_rq_flags(struct request *, char *);
956 extern long nr_blockdev_pages(void);
958 bool __must_check blk_get_queue(struct request_queue *);
959 struct request_queue *blk_alloc_queue(gfp_t);
960 struct request_queue *blk_alloc_queue_node(gfp_t, int);
961 extern void blk_put_queue(struct request_queue *);
964 * blk_plug permits building a queue of related requests by holding the I/O
965 * fragments for a short period. This allows merging of sequential requests
966 * into single larger request. As the requests are moved from a per-task list to
967 * the device's request_queue in a batch, this results in improved scalability
968 * as the lock contention for request_queue lock is reduced.
970 * It is ok not to disable preemption when adding the request to the plug list
971 * or when attempting a merge, because blk_schedule_flush_list() will only flush
972 * the plug list when the task sleeps by itself. For details, please see
973 * schedule() where blk_schedule_flush_plug() is called.
976 unsigned long magic; /* detect uninitialized use-cases */
977 struct list_head list; /* requests */
978 struct list_head cb_list; /* md requires an unplug callback */
980 #define BLK_MAX_REQUEST_COUNT 16
983 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
985 struct list_head list;
986 blk_plug_cb_fn callback;
989 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
990 void *data, int size);
991 extern void blk_start_plug(struct blk_plug *);
992 extern void blk_finish_plug(struct blk_plug *);
993 extern void blk_flush_plug_list(struct blk_plug *, bool);
995 static inline void blk_flush_plug(struct task_struct *tsk)
997 struct blk_plug *plug = tsk->plug;
1000 blk_flush_plug_list(plug, false);
1003 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1005 struct blk_plug *plug = tsk->plug;
1008 blk_flush_plug_list(plug, true);
1011 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1013 struct blk_plug *plug = tsk->plug;
1015 return plug && (!list_empty(&plug->list) || !list_empty(&plug->cb_list));
1021 #define blk_rq_tagged(rq) ((rq)->cmd_flags & REQ_QUEUED)
1022 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1023 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1024 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1025 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
1026 extern void blk_queue_free_tags(struct request_queue *);
1027 extern int blk_queue_resize_tags(struct request_queue *, int);
1028 extern void blk_queue_invalidate_tags(struct request_queue *);
1029 extern struct blk_queue_tag *blk_init_tags(int);
1030 extern void blk_free_tags(struct blk_queue_tag *);
1032 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1035 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1037 return bqt->tag_index[tag];
1040 #define BLKDEV_DISCARD_SECURE 0x01 /* secure discard */
1042 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1043 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1044 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1045 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1046 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1047 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1048 sector_t nr_sects, gfp_t gfp_mask);
1049 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1050 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1052 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1053 nr_blocks << (sb->s_blocksize_bits - 9),
1056 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1057 sector_t nr_blocks, gfp_t gfp_mask)
1059 return blkdev_issue_zeroout(sb->s_bdev,
1060 block << (sb->s_blocksize_bits - 9),
1061 nr_blocks << (sb->s_blocksize_bits - 9),
1065 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1067 enum blk_default_limits {
1068 BLK_MAX_SEGMENTS = 128,
1069 BLK_SAFE_MAX_SECTORS = 255,
1070 BLK_DEF_MAX_SECTORS = 1024,
1071 BLK_MAX_SEGMENT_SIZE = 65536,
1072 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1075 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1077 static inline unsigned long queue_bounce_pfn(struct request_queue *q)
1079 return q->limits.bounce_pfn;
1082 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1084 return q->limits.seg_boundary_mask;
1087 static inline unsigned int queue_max_sectors(struct request_queue *q)
1089 return q->limits.max_sectors;
1092 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1094 return q->limits.max_hw_sectors;
1097 static inline unsigned short queue_max_segments(struct request_queue *q)
1099 return q->limits.max_segments;
1102 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1104 return q->limits.max_segment_size;
1107 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1111 if (q && q->limits.logical_block_size)
1112 retval = q->limits.logical_block_size;
1117 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1119 return queue_logical_block_size(bdev_get_queue(bdev));
1122 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1124 return q->limits.physical_block_size;
1127 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1129 return queue_physical_block_size(bdev_get_queue(bdev));
1132 static inline unsigned int queue_io_min(struct request_queue *q)
1134 return q->limits.io_min;
1137 static inline int bdev_io_min(struct block_device *bdev)
1139 return queue_io_min(bdev_get_queue(bdev));
1142 static inline unsigned int queue_io_opt(struct request_queue *q)
1144 return q->limits.io_opt;
1147 static inline int bdev_io_opt(struct block_device *bdev)
1149 return queue_io_opt(bdev_get_queue(bdev));
1152 static inline int queue_alignment_offset(struct request_queue *q)
1154 if (q->limits.misaligned)
1157 return q->limits.alignment_offset;
1160 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1162 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1163 unsigned int alignment = (sector << 9) & (granularity - 1);
1165 return (granularity + lim->alignment_offset - alignment)
1166 & (granularity - 1);
1169 static inline int bdev_alignment_offset(struct block_device *bdev)
1171 struct request_queue *q = bdev_get_queue(bdev);
1173 if (q->limits.misaligned)
1176 if (bdev != bdev->bd_contains)
1177 return bdev->bd_part->alignment_offset;
1179 return q->limits.alignment_offset;
1182 static inline int queue_discard_alignment(struct request_queue *q)
1184 if (q->limits.discard_misaligned)
1187 return q->limits.discard_alignment;
1190 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1192 unsigned int alignment, granularity, offset;
1194 if (!lim->max_discard_sectors)
1197 /* Why are these in bytes, not sectors? */
1198 alignment = lim->discard_alignment >> 9;
1199 granularity = lim->discard_granularity >> 9;
1203 /* Offset of the partition start in 'granularity' sectors */
1204 offset = sector_div(sector, granularity);
1206 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1207 offset = (granularity + alignment - offset) % granularity;
1209 /* Turn it back into bytes, gaah */
1213 static inline int bdev_discard_alignment(struct block_device *bdev)
1215 struct request_queue *q = bdev_get_queue(bdev);
1217 if (bdev != bdev->bd_contains)
1218 return bdev->bd_part->discard_alignment;
1220 return q->limits.discard_alignment;
1223 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1225 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1)
1231 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1233 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1236 static inline unsigned int bdev_write_same(struct block_device *bdev)
1238 struct request_queue *q = bdev_get_queue(bdev);
1241 return q->limits.max_write_same_sectors;
1246 static inline int queue_dma_alignment(struct request_queue *q)
1248 return q ? q->dma_alignment : 511;
1251 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1254 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1255 return !(addr & alignment) && !(len & alignment);
1258 /* assumes size > 256 */
1259 static inline unsigned int blksize_bits(unsigned int size)
1261 unsigned int bits = 8;
1265 } while (size > 256);
1269 static inline unsigned int block_size(struct block_device *bdev)
1271 return bdev->bd_block_size;
1274 static inline bool queue_flush_queueable(struct request_queue *q)
1276 return !q->flush_not_queueable;
1279 typedef struct {struct page *v;} Sector;
1281 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1283 static inline void put_dev_sector(Sector p)
1285 page_cache_release(p.v);
1289 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
1291 #ifdef CONFIG_BLK_CGROUP
1293 * This should not be using sched_clock(). A real patch is in progress
1294 * to fix this up, until that is in place we need to disable preemption
1295 * around sched_clock() in this function and set_io_start_time_ns().
1297 static inline void set_start_time_ns(struct request *req)
1300 req->start_time_ns = sched_clock();
1304 static inline void set_io_start_time_ns(struct request *req)
1307 req->io_start_time_ns = sched_clock();
1311 static inline uint64_t rq_start_time_ns(struct request *req)
1313 return req->start_time_ns;
1316 static inline uint64_t rq_io_start_time_ns(struct request *req)
1318 return req->io_start_time_ns;
1321 static inline void set_start_time_ns(struct request *req) {}
1322 static inline void set_io_start_time_ns(struct request *req) {}
1323 static inline uint64_t rq_start_time_ns(struct request *req)
1327 static inline uint64_t rq_io_start_time_ns(struct request *req)
1333 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1334 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1335 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1336 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1338 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1340 #define INTEGRITY_FLAG_READ 2 /* verify data integrity on read */
1341 #define INTEGRITY_FLAG_WRITE 4 /* generate data integrity on write */
1343 struct blk_integrity_exchg {
1347 unsigned int data_size;
1348 unsigned short sector_size;
1349 const char *disk_name;
1352 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1353 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1354 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1355 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1357 struct blk_integrity {
1358 integrity_gen_fn *generate_fn;
1359 integrity_vrfy_fn *verify_fn;
1360 integrity_set_tag_fn *set_tag_fn;
1361 integrity_get_tag_fn *get_tag_fn;
1363 unsigned short flags;
1364 unsigned short tuple_size;
1365 unsigned short sector_size;
1366 unsigned short tag_size;
1370 struct kobject kobj;
1373 extern bool blk_integrity_is_initialized(struct gendisk *);
1374 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1375 extern void blk_integrity_unregister(struct gendisk *);
1376 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1377 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1378 struct scatterlist *);
1379 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1380 extern int blk_integrity_merge_rq(struct request_queue *, struct request *,
1382 extern int blk_integrity_merge_bio(struct request_queue *, struct request *,
1386 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1388 return bdev->bd_disk->integrity;
1391 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1393 return disk->integrity;
1396 static inline int blk_integrity_rq(struct request *rq)
1398 if (rq->bio == NULL)
1401 return bio_integrity(rq->bio);
1404 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1407 q->limits.max_integrity_segments = segs;
1410 static inline unsigned short
1411 queue_max_integrity_segments(struct request_queue *q)
1413 return q->limits.max_integrity_segments;
1416 #else /* CONFIG_BLK_DEV_INTEGRITY */
1419 struct block_device;
1421 struct blk_integrity;
1423 static inline int blk_integrity_rq(struct request *rq)
1427 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1432 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1434 struct scatterlist *s)
1438 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1442 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1446 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1450 static inline int blk_integrity_register(struct gendisk *d,
1451 struct blk_integrity *b)
1455 static inline void blk_integrity_unregister(struct gendisk *d)
1458 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1462 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1466 static inline int blk_integrity_merge_rq(struct request_queue *rq,
1472 static inline int blk_integrity_merge_bio(struct request_queue *rq,
1478 static inline bool blk_integrity_is_initialized(struct gendisk *g)
1483 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1485 struct block_device_operations {
1486 int (*open) (struct block_device *, fmode_t);
1487 void (*release) (struct gendisk *, fmode_t);
1488 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1489 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1490 int (*direct_access) (struct block_device *, sector_t,
1491 void **, unsigned long *);
1492 unsigned int (*check_events) (struct gendisk *disk,
1493 unsigned int clearing);
1494 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1495 int (*media_changed) (struct gendisk *);
1496 void (*unlock_native_capacity) (struct gendisk *);
1497 int (*revalidate_disk) (struct gendisk *);
1498 int (*getgeo)(struct block_device *, struct hd_geometry *);
1499 /* this callback is with swap_lock and sometimes page table lock held */
1500 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1501 struct module *owner;
1504 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1506 #else /* CONFIG_BLOCK */
1508 * stubs for when the block layer is configured out
1510 #define buffer_heads_over_limit 0
1512 static inline long nr_blockdev_pages(void)
1520 static inline void blk_start_plug(struct blk_plug *plug)
1524 static inline void blk_finish_plug(struct blk_plug *plug)
1528 static inline void blk_flush_plug(struct task_struct *task)
1532 static inline void blk_schedule_flush_plug(struct task_struct *task)
1537 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1542 #endif /* CONFIG_BLOCK */