1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
4 #include <linux/sched.h>
5 #include <linux/sched/clock.h>
9 #include <linux/major.h>
10 #include <linux/genhd.h>
11 #include <linux/list.h>
12 #include <linux/llist.h>
13 #include <linux/timer.h>
14 #include <linux/workqueue.h>
15 #include <linux/pagemap.h>
16 #include <linux/backing-dev-defs.h>
17 #include <linux/wait.h>
18 #include <linux/mempool.h>
19 #include <linux/pfn.h>
20 #include <linux/bio.h>
21 #include <linux/stringify.h>
22 #include <linux/gfp.h>
23 #include <linux/bsg.h>
24 #include <linux/smp.h>
25 #include <linux/rcupdate.h>
26 #include <linux/percpu-refcount.h>
27 #include <linux/scatterlist.h>
28 #include <linux/blkzoned.h>
31 struct scsi_ioctl_command;
34 struct elevator_queue;
40 struct blk_flush_queue;
43 struct blk_queue_stats;
44 struct blk_stat_callback;
46 #define BLKDEV_MIN_RQ 4
47 #define BLKDEV_MAX_RQ 128 /* Default maximum */
49 /* Must be consisitent with blk_mq_poll_stats_bkt() */
50 #define BLK_MQ_POLL_STATS_BKTS 16
53 * Maximum number of blkcg policies allowed to be registered concurrently.
54 * Defined here to simplify include dependency.
56 #define BLKCG_MAX_POLS 3
58 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
60 #define BLK_RL_SYNCFULL (1U << 0)
61 #define BLK_RL_ASYNCFULL (1U << 1)
64 struct request_queue *q; /* the queue this rl belongs to */
65 #ifdef CONFIG_BLK_CGROUP
66 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
69 * count[], starved[], and wait[] are indexed by
70 * BLK_RW_SYNC/BLK_RW_ASYNC
75 wait_queue_head_t wait[2];
81 typedef __u32 __bitwise req_flags_t;
83 /* elevator knows about this request */
84 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
85 /* drive already may have started this one */
86 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
87 /* uses tagged queueing */
88 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
89 /* may not be passed by ioscheduler */
90 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
91 /* request for flush sequence */
92 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
93 /* merge of different types, fail separately */
94 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
95 /* track inflight for MQ */
96 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
97 /* don't call prep for this one */
98 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
99 /* set for "ide_preempt" requests and also for requests for which the SCSI
100 "quiesce" state must be ignored. */
101 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
102 /* contains copies of user pages */
103 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
104 /* vaguely specified driver internal error. Ignored by the block layer */
105 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
106 /* don't warn about errors */
107 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
108 /* elevator private data attached */
109 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
110 /* account I/O stat */
111 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
112 /* request came from our alloc pool */
113 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
114 /* runtime pm request */
115 #define RQF_PM ((__force req_flags_t)(1 << 15))
116 /* on IO scheduler merge hash */
117 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
118 /* IO stats tracking on */
119 #define RQF_STATS ((__force req_flags_t)(1 << 17))
120 /* Look at ->special_vec for the actual data payload instead of the
122 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
124 /* flags that prevent us from merging requests: */
125 #define RQF_NOMERGE_FLAGS \
126 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
129 * Try to put the fields that are referenced together in the same cacheline.
131 * If you modify this structure, make sure to update blk_rq_init() and
132 * especially blk_mq_rq_ctx_init() to take care of the added fields.
135 struct list_head queuelist;
137 struct call_single_data csd;
141 struct request_queue *q;
142 struct blk_mq_ctx *mq_ctx;
145 unsigned int cmd_flags; /* op and common flags */
146 req_flags_t rq_flags;
150 unsigned long atomic_flags;
152 /* the following two fields are internal, NEVER access directly */
153 unsigned int __data_len; /* total data len */
155 sector_t __sector; /* sector cursor */
161 * The hash is used inside the scheduler, and killed once the
162 * request reaches the dispatch list. The ipi_list is only used
163 * to queue the request for softirq completion, which is long
164 * after the request has been unhashed (and even removed from
165 * the dispatch list).
168 struct hlist_node hash; /* merge hash */
169 struct list_head ipi_list;
173 * The rb_node is only used inside the io scheduler, requests
174 * are pruned when moved to the dispatch queue. So let the
175 * completion_data share space with the rb_node.
178 struct rb_node rb_node; /* sort/lookup */
179 struct bio_vec special_vec;
180 void *completion_data;
181 int error_count; /* for legacy drivers, don't use */
185 * Three pointers are available for the IO schedulers, if they need
186 * more they have to dynamically allocate it. Flush requests are
187 * never put on the IO scheduler. So let the flush fields share
188 * space with the elevator data.
198 struct list_head list;
199 rq_end_io_fn *saved_end_io;
203 struct gendisk *rq_disk;
204 struct hd_struct *part;
205 unsigned long start_time;
206 struct blk_issue_stat issue_stat;
207 #ifdef CONFIG_BLK_CGROUP
208 struct request_list *rl; /* rl this rq is alloced from */
209 unsigned long long start_time_ns;
210 unsigned long long io_start_time_ns; /* when passed to hardware */
212 /* Number of scatter-gather DMA addr+len pairs after
213 * physical address coalescing is performed.
215 unsigned short nr_phys_segments;
216 #if defined(CONFIG_BLK_DEV_INTEGRITY)
217 unsigned short nr_integrity_segments;
220 unsigned short ioprio;
222 unsigned int timeout;
224 void *special; /* opaque pointer available for LLD use */
226 unsigned int extra_len; /* length of alignment and padding */
228 unsigned short write_hint;
230 unsigned long deadline;
231 struct list_head timeout_list;
234 * completion callback.
236 rq_end_io_fn *end_io;
240 struct request *next_rq;
243 static inline bool blk_rq_is_scsi(struct request *rq)
245 return req_op(rq) == REQ_OP_SCSI_IN || req_op(rq) == REQ_OP_SCSI_OUT;
248 static inline bool blk_rq_is_private(struct request *rq)
250 return req_op(rq) == REQ_OP_DRV_IN || req_op(rq) == REQ_OP_DRV_OUT;
253 static inline bool blk_rq_is_passthrough(struct request *rq)
255 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
258 static inline unsigned short req_get_ioprio(struct request *req)
263 #include <linux/elevator.h>
265 struct blk_queue_ctx;
267 typedef void (request_fn_proc) (struct request_queue *q);
268 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
269 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
270 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
273 typedef void (softirq_done_fn)(struct request *);
274 typedef int (dma_drain_needed_fn)(struct request *);
275 typedef int (lld_busy_fn) (struct request_queue *q);
276 typedef int (bsg_job_fn) (struct bsg_job *);
277 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
278 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
280 enum blk_eh_timer_return {
286 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
288 enum blk_queue_state {
293 struct blk_queue_tag {
294 struct request **tag_index; /* map of busy tags */
295 unsigned long *tag_map; /* bit map of free/busy tags */
296 int max_depth; /* what we will send to device */
297 int real_max_depth; /* what the array can hold */
298 atomic_t refcnt; /* map can be shared */
299 int alloc_policy; /* tag allocation policy */
300 int next_tag; /* next tag */
302 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
303 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
305 #define BLK_SCSI_MAX_CMDS (256)
306 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
309 * Zoned block device models (zoned limit).
311 enum blk_zoned_model {
312 BLK_ZONED_NONE, /* Regular block device */
313 BLK_ZONED_HA, /* Host-aware zoned block device */
314 BLK_ZONED_HM, /* Host-managed zoned block device */
317 struct queue_limits {
318 unsigned long bounce_pfn;
319 unsigned long seg_boundary_mask;
320 unsigned long virt_boundary_mask;
322 unsigned int max_hw_sectors;
323 unsigned int max_dev_sectors;
324 unsigned int chunk_sectors;
325 unsigned int max_sectors;
326 unsigned int max_segment_size;
327 unsigned int physical_block_size;
328 unsigned int alignment_offset;
331 unsigned int max_discard_sectors;
332 unsigned int max_hw_discard_sectors;
333 unsigned int max_write_same_sectors;
334 unsigned int max_write_zeroes_sectors;
335 unsigned int discard_granularity;
336 unsigned int discard_alignment;
338 unsigned short logical_block_size;
339 unsigned short max_segments;
340 unsigned short max_integrity_segments;
341 unsigned short max_discard_segments;
343 unsigned char misaligned;
344 unsigned char discard_misaligned;
345 unsigned char cluster;
346 unsigned char raid_partial_stripes_expensive;
347 enum blk_zoned_model zoned;
350 #ifdef CONFIG_BLK_DEV_ZONED
352 struct blk_zone_report_hdr {
353 unsigned int nr_zones;
357 extern int blkdev_report_zones(struct block_device *bdev,
358 sector_t sector, struct blk_zone *zones,
359 unsigned int *nr_zones, gfp_t gfp_mask);
360 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
361 sector_t nr_sectors, gfp_t gfp_mask);
363 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
364 unsigned int cmd, unsigned long arg);
365 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
366 unsigned int cmd, unsigned long arg);
368 #else /* CONFIG_BLK_DEV_ZONED */
370 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
371 fmode_t mode, unsigned int cmd,
377 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
378 fmode_t mode, unsigned int cmd,
384 #endif /* CONFIG_BLK_DEV_ZONED */
386 struct request_queue {
388 * Together with queue_head for cacheline sharing
390 struct list_head queue_head;
391 struct request *last_merge;
392 struct elevator_queue *elevator;
393 int nr_rqs[2]; /* # allocated [a]sync rqs */
394 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
396 atomic_t shared_hctx_restart;
398 struct blk_queue_stats *stats;
402 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
403 * is used, root blkg allocates from @q->root_rl and all other
404 * blkgs from their own blkg->rl. Which one to use should be
405 * determined using bio_request_list().
407 struct request_list root_rl;
409 request_fn_proc *request_fn;
410 make_request_fn *make_request_fn;
411 prep_rq_fn *prep_rq_fn;
412 unprep_rq_fn *unprep_rq_fn;
413 softirq_done_fn *softirq_done_fn;
414 rq_timed_out_fn *rq_timed_out_fn;
415 dma_drain_needed_fn *dma_drain_needed;
416 lld_busy_fn *lld_busy_fn;
417 /* Called just after a request is allocated */
418 init_rq_fn *init_rq_fn;
419 /* Called just before a request is freed */
420 exit_rq_fn *exit_rq_fn;
421 /* Called from inside blk_get_request() */
422 void (*initialize_rq_fn)(struct request *rq);
424 const struct blk_mq_ops *mq_ops;
426 unsigned int *mq_map;
429 struct blk_mq_ctx __percpu *queue_ctx;
430 unsigned int nr_queues;
432 unsigned int queue_depth;
434 /* hw dispatch queues */
435 struct blk_mq_hw_ctx **queue_hw_ctx;
436 unsigned int nr_hw_queues;
439 * Dispatch queue sorting
442 struct request *boundary_rq;
445 * Delayed queue handling
447 struct delayed_work delay_work;
449 struct backing_dev_info *backing_dev_info;
452 * The queue owner gets to use this for whatever they like.
453 * ll_rw_blk doesn't touch it.
458 * various queue flags, see QUEUE_* below
460 unsigned long queue_flags;
463 * ida allocated id for this queue. Used to index queues from
469 * queue needs bounce pages for pages above this limit
474 * protects queue structures from reentrancy. ->__queue_lock should
475 * _never_ be used directly, it is queue private. always use
478 spinlock_t __queue_lock;
479 spinlock_t *queue_lock;
489 struct kobject mq_kobj;
491 #ifdef CONFIG_BLK_DEV_INTEGRITY
492 struct blk_integrity integrity;
493 #endif /* CONFIG_BLK_DEV_INTEGRITY */
498 unsigned int nr_pending;
504 unsigned long nr_requests; /* Max # of requests */
505 unsigned int nr_congestion_on;
506 unsigned int nr_congestion_off;
507 unsigned int nr_batching;
509 unsigned int dma_drain_size;
510 void *dma_drain_buffer;
511 unsigned int dma_pad_mask;
512 unsigned int dma_alignment;
514 struct blk_queue_tag *queue_tags;
515 struct list_head tag_busy_list;
517 unsigned int nr_sorted;
518 unsigned int in_flight[2];
521 * Number of active block driver functions for which blk_drain_queue()
522 * must wait. Must be incremented around functions that unlock the
523 * queue_lock internally, e.g. scsi_request_fn().
525 unsigned int request_fn_active;
527 unsigned int rq_timeout;
530 struct blk_stat_callback *poll_cb;
531 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
533 struct timer_list timeout;
534 struct work_struct timeout_work;
535 struct list_head timeout_list;
537 struct list_head icq_list;
538 #ifdef CONFIG_BLK_CGROUP
539 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
540 struct blkcg_gq *root_blkg;
541 struct list_head blkg_list;
544 struct queue_limits limits;
549 unsigned int sg_timeout;
550 unsigned int sg_reserved_size;
552 #ifdef CONFIG_BLK_DEV_IO_TRACE
553 struct blk_trace *blk_trace;
556 * for flush operations
558 struct blk_flush_queue *fq;
560 struct list_head requeue_list;
561 spinlock_t requeue_lock;
562 struct delayed_work requeue_work;
564 struct mutex sysfs_lock;
567 atomic_t mq_freeze_depth;
569 #if defined(CONFIG_BLK_DEV_BSG)
570 bsg_job_fn *bsg_job_fn;
572 struct bsg_class_device bsg_dev;
575 #ifdef CONFIG_BLK_DEV_THROTTLING
577 struct throtl_data *td;
579 struct rcu_head rcu_head;
580 wait_queue_head_t mq_freeze_wq;
581 struct percpu_ref q_usage_counter;
582 struct list_head all_q_node;
584 struct blk_mq_tag_set *tag_set;
585 struct list_head tag_set_list;
586 struct bio_set *bio_split;
588 #ifdef CONFIG_BLK_DEBUG_FS
589 struct dentry *debugfs_dir;
590 struct dentry *sched_debugfs_dir;
593 bool mq_sysfs_init_done;
598 struct work_struct release_work;
600 #define BLK_MAX_WRITE_HINTS 5
601 u64 write_hints[BLK_MAX_WRITE_HINTS];
604 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
605 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
606 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
607 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
608 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
609 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
610 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
611 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
612 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
613 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
614 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
615 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
616 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
617 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
618 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
619 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
620 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
621 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
622 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
623 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
624 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
625 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
626 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
627 #define QUEUE_FLAG_WC 23 /* Write back caching */
628 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
629 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
630 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
631 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
632 #define QUEUE_FLAG_POLL_STATS 28 /* collecting stats for hybrid polling */
633 #define QUEUE_FLAG_REGISTERED 29 /* queue has been registered to a disk */
634 #define QUEUE_FLAG_SCSI_PASSTHROUGH 30 /* queue supports SCSI commands */
635 #define QUEUE_FLAG_QUIESCED 31 /* queue has been quiesced */
637 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
638 (1 << QUEUE_FLAG_STACKABLE) | \
639 (1 << QUEUE_FLAG_SAME_COMP) | \
640 (1 << QUEUE_FLAG_ADD_RANDOM))
642 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
643 (1 << QUEUE_FLAG_STACKABLE) | \
644 (1 << QUEUE_FLAG_SAME_COMP) | \
645 (1 << QUEUE_FLAG_POLL))
648 * @q->queue_lock is set while a queue is being initialized. Since we know
649 * that no other threads access the queue object before @q->queue_lock has
650 * been set, it is safe to manipulate queue flags without holding the
651 * queue_lock if @q->queue_lock == NULL. See also blk_alloc_queue_node() and
652 * blk_init_allocated_queue().
654 static inline void queue_lockdep_assert_held(struct request_queue *q)
657 lockdep_assert_held(q->queue_lock);
660 static inline void queue_flag_set_unlocked(unsigned int flag,
661 struct request_queue *q)
663 __set_bit(flag, &q->queue_flags);
666 static inline int queue_flag_test_and_clear(unsigned int flag,
667 struct request_queue *q)
669 queue_lockdep_assert_held(q);
671 if (test_bit(flag, &q->queue_flags)) {
672 __clear_bit(flag, &q->queue_flags);
679 static inline int queue_flag_test_and_set(unsigned int flag,
680 struct request_queue *q)
682 queue_lockdep_assert_held(q);
684 if (!test_bit(flag, &q->queue_flags)) {
685 __set_bit(flag, &q->queue_flags);
692 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
694 queue_lockdep_assert_held(q);
695 __set_bit(flag, &q->queue_flags);
698 static inline void queue_flag_clear_unlocked(unsigned int flag,
699 struct request_queue *q)
701 __clear_bit(flag, &q->queue_flags);
704 static inline int queue_in_flight(struct request_queue *q)
706 return q->in_flight[0] + q->in_flight[1];
709 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
711 queue_lockdep_assert_held(q);
712 __clear_bit(flag, &q->queue_flags);
715 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
716 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
717 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
718 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
719 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
720 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
721 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
722 #define blk_queue_noxmerges(q) \
723 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
724 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
725 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
726 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
727 #define blk_queue_stackable(q) \
728 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
729 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
730 #define blk_queue_secure_erase(q) \
731 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
732 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
733 #define blk_queue_scsi_passthrough(q) \
734 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
736 #define blk_noretry_request(rq) \
737 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
738 REQ_FAILFAST_DRIVER))
739 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
741 static inline bool blk_account_rq(struct request *rq)
743 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
746 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
747 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
748 /* rq->queuelist of dequeued request must be list_empty() */
749 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
751 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
753 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
756 * Driver can handle struct request, if it either has an old style
757 * request_fn defined, or is blk-mq based.
759 static inline bool queue_is_rq_based(struct request_queue *q)
761 return q->request_fn || q->mq_ops;
764 static inline unsigned int blk_queue_cluster(struct request_queue *q)
766 return q->limits.cluster;
769 static inline enum blk_zoned_model
770 blk_queue_zoned_model(struct request_queue *q)
772 return q->limits.zoned;
775 static inline bool blk_queue_is_zoned(struct request_queue *q)
777 switch (blk_queue_zoned_model(q)) {
786 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
788 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
791 static inline bool rq_is_sync(struct request *rq)
793 return op_is_sync(rq->cmd_flags);
796 static inline bool blk_rl_full(struct request_list *rl, bool sync)
798 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
800 return rl->flags & flag;
803 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
805 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
810 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
812 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
817 static inline bool rq_mergeable(struct request *rq)
819 if (blk_rq_is_passthrough(rq))
822 if (req_op(rq) == REQ_OP_FLUSH)
825 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
828 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
830 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
836 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
838 if (bio_page(a) == bio_page(b) &&
839 bio_offset(a) == bio_offset(b))
845 static inline unsigned int blk_queue_depth(struct request_queue *q)
848 return q->queue_depth;
850 return q->nr_requests;
854 * q->prep_rq_fn return values
857 BLKPREP_OK, /* serve it */
858 BLKPREP_KILL, /* fatal error, kill, return -EIO */
859 BLKPREP_DEFER, /* leave on queue */
860 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
863 extern unsigned long blk_max_low_pfn, blk_max_pfn;
866 * standard bounce addresses:
868 * BLK_BOUNCE_HIGH : bounce all highmem pages
869 * BLK_BOUNCE_ANY : don't bounce anything
870 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
873 #if BITS_PER_LONG == 32
874 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
876 #define BLK_BOUNCE_HIGH -1ULL
878 #define BLK_BOUNCE_ANY (-1ULL)
879 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
882 * default timeout for SG_IO if none specified
884 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
885 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
891 unsigned long offset;
896 struct req_iterator {
897 struct bvec_iter iter;
901 /* This should not be used directly - use rq_for_each_segment */
902 #define for_each_bio(_bio) \
903 for (; _bio; _bio = _bio->bi_next)
904 #define __rq_for_each_bio(_bio, rq) \
906 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
908 #define rq_for_each_segment(bvl, _rq, _iter) \
909 __rq_for_each_bio(_iter.bio, _rq) \
910 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
912 #define rq_iter_last(bvec, _iter) \
913 (_iter.bio->bi_next == NULL && \
914 bio_iter_last(bvec, _iter.iter))
916 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
917 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
919 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
920 extern void rq_flush_dcache_pages(struct request *rq);
922 static inline void rq_flush_dcache_pages(struct request *rq)
928 #define vfs_msg(sb, level, fmt, ...) \
929 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
931 #define vfs_msg(sb, level, fmt, ...) \
933 no_printk(fmt, ##__VA_ARGS__); \
934 __vfs_msg(sb, "", " "); \
938 extern int blk_register_queue(struct gendisk *disk);
939 extern void blk_unregister_queue(struct gendisk *disk);
940 extern blk_qc_t generic_make_request(struct bio *bio);
941 extern void blk_rq_init(struct request_queue *q, struct request *rq);
942 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
943 extern void blk_put_request(struct request *);
944 extern void __blk_put_request(struct request_queue *, struct request *);
945 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
947 extern void blk_requeue_request(struct request_queue *, struct request *);
948 extern int blk_lld_busy(struct request_queue *q);
949 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
950 struct bio_set *bs, gfp_t gfp_mask,
951 int (*bio_ctr)(struct bio *, struct bio *, void *),
953 extern void blk_rq_unprep_clone(struct request *rq);
954 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
956 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
957 extern void blk_delay_queue(struct request_queue *, unsigned long);
958 extern void blk_queue_split(struct request_queue *, struct bio **);
959 extern void blk_recount_segments(struct request_queue *, struct bio *);
960 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
961 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
962 unsigned int, void __user *);
963 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
964 unsigned int, void __user *);
965 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
966 struct scsi_ioctl_command __user *);
968 extern int blk_queue_enter(struct request_queue *q, bool nowait);
969 extern void blk_queue_exit(struct request_queue *q);
970 extern void blk_start_queue(struct request_queue *q);
971 extern void blk_start_queue_async(struct request_queue *q);
972 extern void blk_stop_queue(struct request_queue *q);
973 extern void blk_sync_queue(struct request_queue *q);
974 extern void __blk_stop_queue(struct request_queue *q);
975 extern void __blk_run_queue(struct request_queue *q);
976 extern void __blk_run_queue_uncond(struct request_queue *q);
977 extern void blk_run_queue(struct request_queue *);
978 extern void blk_run_queue_async(struct request_queue *q);
979 extern int blk_rq_map_user(struct request_queue *, struct request *,
980 struct rq_map_data *, void __user *, unsigned long,
982 extern int blk_rq_unmap_user(struct bio *);
983 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
984 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
985 struct rq_map_data *, const struct iov_iter *,
987 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
988 struct request *, int);
989 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
990 struct request *, int, rq_end_io_fn *);
992 int blk_status_to_errno(blk_status_t status);
993 blk_status_t errno_to_blk_status(int errno);
995 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
997 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
999 return bdev->bd_disk->queue; /* this is never NULL */
1003 * blk_rq_pos() : the current sector
1004 * blk_rq_bytes() : bytes left in the entire request
1005 * blk_rq_cur_bytes() : bytes left in the current segment
1006 * blk_rq_err_bytes() : bytes left till the next error boundary
1007 * blk_rq_sectors() : sectors left in the entire request
1008 * blk_rq_cur_sectors() : sectors left in the current segment
1010 static inline sector_t blk_rq_pos(const struct request *rq)
1012 return rq->__sector;
1015 static inline unsigned int blk_rq_bytes(const struct request *rq)
1017 return rq->__data_len;
1020 static inline int blk_rq_cur_bytes(const struct request *rq)
1022 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1025 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1027 static inline unsigned int blk_rq_sectors(const struct request *rq)
1029 return blk_rq_bytes(rq) >> 9;
1032 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1034 return blk_rq_cur_bytes(rq) >> 9;
1038 * Some commands like WRITE SAME have a payload or data transfer size which
1039 * is different from the size of the request. Any driver that supports such
1040 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1041 * calculate the data transfer size.
1043 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1045 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1046 return rq->special_vec.bv_len;
1047 return blk_rq_bytes(rq);
1050 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1053 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1054 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1056 if (unlikely(op == REQ_OP_WRITE_SAME))
1057 return q->limits.max_write_same_sectors;
1059 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1060 return q->limits.max_write_zeroes_sectors;
1062 return q->limits.max_sectors;
1066 * Return maximum size of a request at given offset. Only valid for
1067 * file system requests.
1069 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1072 if (!q->limits.chunk_sectors)
1073 return q->limits.max_sectors;
1075 return q->limits.chunk_sectors -
1076 (offset & (q->limits.chunk_sectors - 1));
1079 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1082 struct request_queue *q = rq->q;
1084 if (blk_rq_is_passthrough(rq))
1085 return q->limits.max_hw_sectors;
1087 if (!q->limits.chunk_sectors ||
1088 req_op(rq) == REQ_OP_DISCARD ||
1089 req_op(rq) == REQ_OP_SECURE_ERASE)
1090 return blk_queue_get_max_sectors(q, req_op(rq));
1092 return min(blk_max_size_offset(q, offset),
1093 blk_queue_get_max_sectors(q, req_op(rq)));
1096 static inline unsigned int blk_rq_count_bios(struct request *rq)
1098 unsigned int nr_bios = 0;
1101 __rq_for_each_bio(bio, rq)
1108 * Request issue related functions.
1110 extern struct request *blk_peek_request(struct request_queue *q);
1111 extern void blk_start_request(struct request *rq);
1112 extern struct request *blk_fetch_request(struct request_queue *q);
1115 * Request completion related functions.
1117 * blk_update_request() completes given number of bytes and updates
1118 * the request without completing it.
1120 * blk_end_request() and friends. __blk_end_request() must be called
1121 * with the request queue spinlock acquired.
1123 * Several drivers define their own end_request and call
1124 * blk_end_request() for parts of the original function.
1125 * This prevents code duplication in drivers.
1127 extern bool blk_update_request(struct request *rq, blk_status_t error,
1128 unsigned int nr_bytes);
1129 extern void blk_finish_request(struct request *rq, blk_status_t error);
1130 extern bool blk_end_request(struct request *rq, blk_status_t error,
1131 unsigned int nr_bytes);
1132 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1133 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1134 unsigned int nr_bytes);
1135 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1136 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1138 extern void blk_complete_request(struct request *);
1139 extern void __blk_complete_request(struct request *);
1140 extern void blk_abort_request(struct request *);
1141 extern void blk_unprep_request(struct request *);
1144 * Access functions for manipulating queue properties
1146 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1147 spinlock_t *lock, int node_id);
1148 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1149 extern int blk_init_allocated_queue(struct request_queue *);
1150 extern void blk_cleanup_queue(struct request_queue *);
1151 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1152 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1153 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1154 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1155 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1156 extern void blk_queue_max_discard_segments(struct request_queue *,
1158 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1159 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1160 unsigned int max_discard_sectors);
1161 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1162 unsigned int max_write_same_sectors);
1163 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1164 unsigned int max_write_same_sectors);
1165 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1166 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1167 extern void blk_queue_alignment_offset(struct request_queue *q,
1168 unsigned int alignment);
1169 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1170 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1171 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1172 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1173 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1174 extern void blk_set_default_limits(struct queue_limits *lim);
1175 extern void blk_set_stacking_limits(struct queue_limits *lim);
1176 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1178 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1180 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1182 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1183 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1184 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1185 extern int blk_queue_dma_drain(struct request_queue *q,
1186 dma_drain_needed_fn *dma_drain_needed,
1187 void *buf, unsigned int size);
1188 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1189 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1190 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1191 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1192 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1193 extern void blk_queue_dma_alignment(struct request_queue *, int);
1194 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1195 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1196 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1197 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1198 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1199 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1202 * Number of physical segments as sent to the device.
1204 * Normally this is the number of discontiguous data segments sent by the
1205 * submitter. But for data-less command like discard we might have no
1206 * actual data segments submitted, but the driver might have to add it's
1207 * own special payload. In that case we still return 1 here so that this
1208 * special payload will be mapped.
1210 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1212 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1214 return rq->nr_phys_segments;
1218 * Number of discard segments (or ranges) the driver needs to fill in.
1219 * Each discard bio merged into a request is counted as one segment.
1221 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1223 return max_t(unsigned short, rq->nr_phys_segments, 1);
1226 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1227 extern void blk_dump_rq_flags(struct request *, char *);
1228 extern long nr_blockdev_pages(void);
1230 bool __must_check blk_get_queue(struct request_queue *);
1231 struct request_queue *blk_alloc_queue(gfp_t);
1232 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1233 extern void blk_put_queue(struct request_queue *);
1234 extern void blk_set_queue_dying(struct request_queue *);
1237 * block layer runtime pm functions
1240 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1241 extern int blk_pre_runtime_suspend(struct request_queue *q);
1242 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1243 extern void blk_pre_runtime_resume(struct request_queue *q);
1244 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1245 extern void blk_set_runtime_active(struct request_queue *q);
1247 static inline void blk_pm_runtime_init(struct request_queue *q,
1248 struct device *dev) {}
1249 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1253 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1254 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1255 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1256 static inline void blk_set_runtime_active(struct request_queue *q) {}
1260 * blk_plug permits building a queue of related requests by holding the I/O
1261 * fragments for a short period. This allows merging of sequential requests
1262 * into single larger request. As the requests are moved from a per-task list to
1263 * the device's request_queue in a batch, this results in improved scalability
1264 * as the lock contention for request_queue lock is reduced.
1266 * It is ok not to disable preemption when adding the request to the plug list
1267 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1268 * the plug list when the task sleeps by itself. For details, please see
1269 * schedule() where blk_schedule_flush_plug() is called.
1272 struct list_head list; /* requests */
1273 struct list_head mq_list; /* blk-mq requests */
1274 struct list_head cb_list; /* md requires an unplug callback */
1276 #define BLK_MAX_REQUEST_COUNT 16
1277 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1280 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1281 struct blk_plug_cb {
1282 struct list_head list;
1283 blk_plug_cb_fn callback;
1286 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1287 void *data, int size);
1288 extern void blk_start_plug(struct blk_plug *);
1289 extern void blk_finish_plug(struct blk_plug *);
1290 extern void blk_flush_plug_list(struct blk_plug *, bool);
1292 static inline void blk_flush_plug(struct task_struct *tsk)
1294 struct blk_plug *plug = tsk->plug;
1297 blk_flush_plug_list(plug, false);
1300 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1302 struct blk_plug *plug = tsk->plug;
1305 blk_flush_plug_list(plug, true);
1308 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1310 struct blk_plug *plug = tsk->plug;
1313 (!list_empty(&plug->list) ||
1314 !list_empty(&plug->mq_list) ||
1315 !list_empty(&plug->cb_list));
1321 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1322 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1323 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1324 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1325 extern void blk_queue_free_tags(struct request_queue *);
1326 extern int blk_queue_resize_tags(struct request_queue *, int);
1327 extern void blk_queue_invalidate_tags(struct request_queue *);
1328 extern struct blk_queue_tag *blk_init_tags(int, int);
1329 extern void blk_free_tags(struct blk_queue_tag *);
1331 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1334 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1336 return bqt->tag_index[tag];
1339 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1340 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1341 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1343 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1345 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1346 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1347 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1348 sector_t nr_sects, gfp_t gfp_mask, int flags,
1351 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1352 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1354 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1355 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1357 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1358 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1360 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1361 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1363 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1364 nr_blocks << (sb->s_blocksize_bits - 9),
1367 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1368 sector_t nr_blocks, gfp_t gfp_mask)
1370 return blkdev_issue_zeroout(sb->s_bdev,
1371 block << (sb->s_blocksize_bits - 9),
1372 nr_blocks << (sb->s_blocksize_bits - 9),
1376 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1378 enum blk_default_limits {
1379 BLK_MAX_SEGMENTS = 128,
1380 BLK_SAFE_MAX_SECTORS = 255,
1381 BLK_DEF_MAX_SECTORS = 2560,
1382 BLK_MAX_SEGMENT_SIZE = 65536,
1383 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1386 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1388 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1390 return q->limits.seg_boundary_mask;
1393 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1395 return q->limits.virt_boundary_mask;
1398 static inline unsigned int queue_max_sectors(struct request_queue *q)
1400 return q->limits.max_sectors;
1403 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1405 return q->limits.max_hw_sectors;
1408 static inline unsigned short queue_max_segments(struct request_queue *q)
1410 return q->limits.max_segments;
1413 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1415 return q->limits.max_discard_segments;
1418 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1420 return q->limits.max_segment_size;
1423 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1427 if (q && q->limits.logical_block_size)
1428 retval = q->limits.logical_block_size;
1433 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1435 return queue_logical_block_size(bdev_get_queue(bdev));
1438 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1440 return q->limits.physical_block_size;
1443 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1445 return queue_physical_block_size(bdev_get_queue(bdev));
1448 static inline unsigned int queue_io_min(struct request_queue *q)
1450 return q->limits.io_min;
1453 static inline int bdev_io_min(struct block_device *bdev)
1455 return queue_io_min(bdev_get_queue(bdev));
1458 static inline unsigned int queue_io_opt(struct request_queue *q)
1460 return q->limits.io_opt;
1463 static inline int bdev_io_opt(struct block_device *bdev)
1465 return queue_io_opt(bdev_get_queue(bdev));
1468 static inline int queue_alignment_offset(struct request_queue *q)
1470 if (q->limits.misaligned)
1473 return q->limits.alignment_offset;
1476 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1478 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1479 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1481 return (granularity + lim->alignment_offset - alignment) % granularity;
1484 static inline int bdev_alignment_offset(struct block_device *bdev)
1486 struct request_queue *q = bdev_get_queue(bdev);
1488 if (q->limits.misaligned)
1491 if (bdev != bdev->bd_contains)
1492 return bdev->bd_part->alignment_offset;
1494 return q->limits.alignment_offset;
1497 static inline int queue_discard_alignment(struct request_queue *q)
1499 if (q->limits.discard_misaligned)
1502 return q->limits.discard_alignment;
1505 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1507 unsigned int alignment, granularity, offset;
1509 if (!lim->max_discard_sectors)
1512 /* Why are these in bytes, not sectors? */
1513 alignment = lim->discard_alignment >> 9;
1514 granularity = lim->discard_granularity >> 9;
1518 /* Offset of the partition start in 'granularity' sectors */
1519 offset = sector_div(sector, granularity);
1521 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1522 offset = (granularity + alignment - offset) % granularity;
1524 /* Turn it back into bytes, gaah */
1528 static inline int bdev_discard_alignment(struct block_device *bdev)
1530 struct request_queue *q = bdev_get_queue(bdev);
1532 if (bdev != bdev->bd_contains)
1533 return bdev->bd_part->discard_alignment;
1535 return q->limits.discard_alignment;
1538 static inline unsigned int bdev_write_same(struct block_device *bdev)
1540 struct request_queue *q = bdev_get_queue(bdev);
1543 return q->limits.max_write_same_sectors;
1548 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1550 struct request_queue *q = bdev_get_queue(bdev);
1553 return q->limits.max_write_zeroes_sectors;
1558 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1560 struct request_queue *q = bdev_get_queue(bdev);
1563 return blk_queue_zoned_model(q);
1565 return BLK_ZONED_NONE;
1568 static inline bool bdev_is_zoned(struct block_device *bdev)
1570 struct request_queue *q = bdev_get_queue(bdev);
1573 return blk_queue_is_zoned(q);
1578 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1580 struct request_queue *q = bdev_get_queue(bdev);
1583 return blk_queue_zone_sectors(q);
1588 static inline int queue_dma_alignment(struct request_queue *q)
1590 return q ? q->dma_alignment : 511;
1593 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1596 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1597 return !(addr & alignment) && !(len & alignment);
1600 /* assumes size > 256 */
1601 static inline unsigned int blksize_bits(unsigned int size)
1603 unsigned int bits = 8;
1607 } while (size > 256);
1611 static inline unsigned int block_size(struct block_device *bdev)
1613 return bdev->bd_block_size;
1616 static inline bool queue_flush_queueable(struct request_queue *q)
1618 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1621 typedef struct {struct page *v;} Sector;
1623 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1625 static inline void put_dev_sector(Sector p)
1630 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1631 struct bio_vec *bprv, unsigned int offset)
1634 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1638 * Check if adding a bio_vec after bprv with offset would create a gap in
1639 * the SG list. Most drivers don't care about this, but some do.
1641 static inline bool bvec_gap_to_prev(struct request_queue *q,
1642 struct bio_vec *bprv, unsigned int offset)
1644 if (!queue_virt_boundary(q))
1646 return __bvec_gap_to_prev(q, bprv, offset);
1650 * Check if the two bvecs from two bios can be merged to one segment.
1651 * If yes, no need to check gap between the two bios since the 1st bio
1652 * and the 1st bvec in the 2nd bio can be handled in one segment.
1654 static inline bool bios_segs_mergeable(struct request_queue *q,
1655 struct bio *prev, struct bio_vec *prev_last_bv,
1656 struct bio_vec *next_first_bv)
1658 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1660 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1662 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1663 queue_max_segment_size(q))
1668 static inline bool bio_will_gap(struct request_queue *q,
1669 struct request *prev_rq,
1673 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1674 struct bio_vec pb, nb;
1677 * don't merge if the 1st bio starts with non-zero
1678 * offset, otherwise it is quite difficult to respect
1679 * sg gap limit. We work hard to merge a huge number of small
1680 * single bios in case of mkfs.
1683 bio_get_first_bvec(prev_rq->bio, &pb);
1685 bio_get_first_bvec(prev, &pb);
1690 * We don't need to worry about the situation that the
1691 * merged segment ends in unaligned virt boundary:
1693 * - if 'pb' ends aligned, the merged segment ends aligned
1694 * - if 'pb' ends unaligned, the next bio must include
1695 * one single bvec of 'nb', otherwise the 'nb' can't
1698 bio_get_last_bvec(prev, &pb);
1699 bio_get_first_bvec(next, &nb);
1701 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1702 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1708 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1710 return bio_will_gap(req->q, req, req->biotail, bio);
1713 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1715 return bio_will_gap(req->q, NULL, bio, req->bio);
1718 int kblockd_schedule_work(struct work_struct *work);
1719 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1720 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1721 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1722 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1724 #ifdef CONFIG_BLK_CGROUP
1726 * This should not be using sched_clock(). A real patch is in progress
1727 * to fix this up, until that is in place we need to disable preemption
1728 * around sched_clock() in this function and set_io_start_time_ns().
1730 static inline void set_start_time_ns(struct request *req)
1733 req->start_time_ns = sched_clock();
1737 static inline void set_io_start_time_ns(struct request *req)
1740 req->io_start_time_ns = sched_clock();
1744 static inline uint64_t rq_start_time_ns(struct request *req)
1746 return req->start_time_ns;
1749 static inline uint64_t rq_io_start_time_ns(struct request *req)
1751 return req->io_start_time_ns;
1754 static inline void set_start_time_ns(struct request *req) {}
1755 static inline void set_io_start_time_ns(struct request *req) {}
1756 static inline uint64_t rq_start_time_ns(struct request *req)
1760 static inline uint64_t rq_io_start_time_ns(struct request *req)
1766 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1767 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1768 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1769 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1771 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1773 enum blk_integrity_flags {
1774 BLK_INTEGRITY_VERIFY = 1 << 0,
1775 BLK_INTEGRITY_GENERATE = 1 << 1,
1776 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1777 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1780 struct blk_integrity_iter {
1784 unsigned int data_size;
1785 unsigned short interval;
1786 const char *disk_name;
1789 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1791 struct blk_integrity_profile {
1792 integrity_processing_fn *generate_fn;
1793 integrity_processing_fn *verify_fn;
1797 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1798 extern void blk_integrity_unregister(struct gendisk *);
1799 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1800 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1801 struct scatterlist *);
1802 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1803 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1805 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1808 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1810 struct blk_integrity *bi = &disk->queue->integrity;
1819 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1821 return blk_get_integrity(bdev->bd_disk);
1824 static inline bool blk_integrity_rq(struct request *rq)
1826 return rq->cmd_flags & REQ_INTEGRITY;
1829 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1832 q->limits.max_integrity_segments = segs;
1835 static inline unsigned short
1836 queue_max_integrity_segments(struct request_queue *q)
1838 return q->limits.max_integrity_segments;
1841 static inline bool integrity_req_gap_back_merge(struct request *req,
1844 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1845 struct bio_integrity_payload *bip_next = bio_integrity(next);
1847 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1848 bip_next->bip_vec[0].bv_offset);
1851 static inline bool integrity_req_gap_front_merge(struct request *req,
1854 struct bio_integrity_payload *bip = bio_integrity(bio);
1855 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1857 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1858 bip_next->bip_vec[0].bv_offset);
1861 #else /* CONFIG_BLK_DEV_INTEGRITY */
1864 struct block_device;
1866 struct blk_integrity;
1868 static inline int blk_integrity_rq(struct request *rq)
1872 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1877 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1879 struct scatterlist *s)
1883 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1887 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1891 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1895 static inline void blk_integrity_register(struct gendisk *d,
1896 struct blk_integrity *b)
1899 static inline void blk_integrity_unregister(struct gendisk *d)
1902 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1906 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1910 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1916 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1923 static inline bool integrity_req_gap_back_merge(struct request *req,
1928 static inline bool integrity_req_gap_front_merge(struct request *req,
1934 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1936 struct block_device_operations {
1937 int (*open) (struct block_device *, fmode_t);
1938 void (*release) (struct gendisk *, fmode_t);
1939 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1940 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1941 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1942 unsigned int (*check_events) (struct gendisk *disk,
1943 unsigned int clearing);
1944 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1945 int (*media_changed) (struct gendisk *);
1946 void (*unlock_native_capacity) (struct gendisk *);
1947 int (*revalidate_disk) (struct gendisk *);
1948 int (*getgeo)(struct block_device *, struct hd_geometry *);
1949 /* this callback is with swap_lock and sometimes page table lock held */
1950 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1951 struct module *owner;
1952 const struct pr_ops *pr_ops;
1955 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1957 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1958 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1959 struct writeback_control *);
1960 #else /* CONFIG_BLOCK */
1962 struct block_device;
1965 * stubs for when the block layer is configured out
1967 #define buffer_heads_over_limit 0
1969 static inline long nr_blockdev_pages(void)
1977 static inline void blk_start_plug(struct blk_plug *plug)
1981 static inline void blk_finish_plug(struct blk_plug *plug)
1985 static inline void blk_flush_plug(struct task_struct *task)
1989 static inline void blk_schedule_flush_plug(struct task_struct *task)
1994 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1999 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2000 sector_t *error_sector)
2005 #endif /* CONFIG_BLOCK */