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 call_single_data_t 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;
571 struct bsg_class_device bsg_dev;
574 #ifdef CONFIG_BLK_DEV_THROTTLING
576 struct throtl_data *td;
578 struct rcu_head rcu_head;
579 wait_queue_head_t mq_freeze_wq;
580 struct percpu_ref q_usage_counter;
581 struct list_head all_q_node;
583 struct blk_mq_tag_set *tag_set;
584 struct list_head tag_set_list;
585 struct bio_set *bio_split;
587 #ifdef CONFIG_BLK_DEBUG_FS
588 struct dentry *debugfs_dir;
589 struct dentry *sched_debugfs_dir;
592 bool mq_sysfs_init_done;
597 struct work_struct release_work;
599 #define BLK_MAX_WRITE_HINTS 5
600 u64 write_hints[BLK_MAX_WRITE_HINTS];
603 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
604 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
605 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */
606 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */
607 #define QUEUE_FLAG_DYING 5 /* queue being torn down */
608 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */
609 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */
610 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */
611 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */
612 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */
613 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */
614 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */
615 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
616 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */
617 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */
618 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */
619 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */
620 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */
621 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */
622 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */
623 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */
624 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/
625 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */
626 #define QUEUE_FLAG_WC 23 /* Write back caching */
627 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */
628 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
629 #define QUEUE_FLAG_DAX 26 /* device supports DAX */
630 #define QUEUE_FLAG_STATS 27 /* track rq completion times */
631 #define QUEUE_FLAG_POLL_STATS 28 /* collecting stats for hybrid polling */
632 #define QUEUE_FLAG_REGISTERED 29 /* queue has been registered to a disk */
633 #define QUEUE_FLAG_SCSI_PASSTHROUGH 30 /* queue supports SCSI commands */
634 #define QUEUE_FLAG_QUIESCED 31 /* queue has been quiesced */
636 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
637 (1 << QUEUE_FLAG_STACKABLE) | \
638 (1 << QUEUE_FLAG_SAME_COMP) | \
639 (1 << QUEUE_FLAG_ADD_RANDOM))
641 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
642 (1 << QUEUE_FLAG_STACKABLE) | \
643 (1 << QUEUE_FLAG_SAME_COMP) | \
644 (1 << QUEUE_FLAG_POLL))
647 * @q->queue_lock is set while a queue is being initialized. Since we know
648 * that no other threads access the queue object before @q->queue_lock has
649 * been set, it is safe to manipulate queue flags without holding the
650 * queue_lock if @q->queue_lock == NULL. See also blk_alloc_queue_node() and
651 * blk_init_allocated_queue().
653 static inline void queue_lockdep_assert_held(struct request_queue *q)
656 lockdep_assert_held(q->queue_lock);
659 static inline void queue_flag_set_unlocked(unsigned int flag,
660 struct request_queue *q)
662 __set_bit(flag, &q->queue_flags);
665 static inline int queue_flag_test_and_clear(unsigned int flag,
666 struct request_queue *q)
668 queue_lockdep_assert_held(q);
670 if (test_bit(flag, &q->queue_flags)) {
671 __clear_bit(flag, &q->queue_flags);
678 static inline int queue_flag_test_and_set(unsigned int flag,
679 struct request_queue *q)
681 queue_lockdep_assert_held(q);
683 if (!test_bit(flag, &q->queue_flags)) {
684 __set_bit(flag, &q->queue_flags);
691 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
693 queue_lockdep_assert_held(q);
694 __set_bit(flag, &q->queue_flags);
697 static inline void queue_flag_clear_unlocked(unsigned int flag,
698 struct request_queue *q)
700 __clear_bit(flag, &q->queue_flags);
703 static inline int queue_in_flight(struct request_queue *q)
705 return q->in_flight[0] + q->in_flight[1];
708 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
710 queue_lockdep_assert_held(q);
711 __clear_bit(flag, &q->queue_flags);
714 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
715 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
716 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
717 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
718 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
719 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
720 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
721 #define blk_queue_noxmerges(q) \
722 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
723 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
724 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
725 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
726 #define blk_queue_stackable(q) \
727 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
728 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
729 #define blk_queue_secure_erase(q) \
730 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
731 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
732 #define blk_queue_scsi_passthrough(q) \
733 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
735 #define blk_noretry_request(rq) \
736 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
737 REQ_FAILFAST_DRIVER))
738 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
740 static inline bool blk_account_rq(struct request *rq)
742 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
745 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
746 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
747 /* rq->queuelist of dequeued request must be list_empty() */
748 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
750 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
752 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
755 * Driver can handle struct request, if it either has an old style
756 * request_fn defined, or is blk-mq based.
758 static inline bool queue_is_rq_based(struct request_queue *q)
760 return q->request_fn || q->mq_ops;
763 static inline unsigned int blk_queue_cluster(struct request_queue *q)
765 return q->limits.cluster;
768 static inline enum blk_zoned_model
769 blk_queue_zoned_model(struct request_queue *q)
771 return q->limits.zoned;
774 static inline bool blk_queue_is_zoned(struct request_queue *q)
776 switch (blk_queue_zoned_model(q)) {
785 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
787 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
790 static inline bool rq_is_sync(struct request *rq)
792 return op_is_sync(rq->cmd_flags);
795 static inline bool blk_rl_full(struct request_list *rl, bool sync)
797 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
799 return rl->flags & flag;
802 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
804 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
809 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
811 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
816 static inline bool rq_mergeable(struct request *rq)
818 if (blk_rq_is_passthrough(rq))
821 if (req_op(rq) == REQ_OP_FLUSH)
824 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
827 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
829 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
835 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
837 if (bio_page(a) == bio_page(b) &&
838 bio_offset(a) == bio_offset(b))
844 static inline unsigned int blk_queue_depth(struct request_queue *q)
847 return q->queue_depth;
849 return q->nr_requests;
853 * q->prep_rq_fn return values
856 BLKPREP_OK, /* serve it */
857 BLKPREP_KILL, /* fatal error, kill, return -EIO */
858 BLKPREP_DEFER, /* leave on queue */
859 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
862 extern unsigned long blk_max_low_pfn, blk_max_pfn;
865 * standard bounce addresses:
867 * BLK_BOUNCE_HIGH : bounce all highmem pages
868 * BLK_BOUNCE_ANY : don't bounce anything
869 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
872 #if BITS_PER_LONG == 32
873 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
875 #define BLK_BOUNCE_HIGH -1ULL
877 #define BLK_BOUNCE_ANY (-1ULL)
878 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
881 * default timeout for SG_IO if none specified
883 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
884 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
890 unsigned long offset;
895 struct req_iterator {
896 struct bvec_iter iter;
900 /* This should not be used directly - use rq_for_each_segment */
901 #define for_each_bio(_bio) \
902 for (; _bio; _bio = _bio->bi_next)
903 #define __rq_for_each_bio(_bio, rq) \
905 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
907 #define rq_for_each_segment(bvl, _rq, _iter) \
908 __rq_for_each_bio(_iter.bio, _rq) \
909 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
911 #define rq_iter_last(bvec, _iter) \
912 (_iter.bio->bi_next == NULL && \
913 bio_iter_last(bvec, _iter.iter))
915 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
916 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
918 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
919 extern void rq_flush_dcache_pages(struct request *rq);
921 static inline void rq_flush_dcache_pages(struct request *rq)
927 #define vfs_msg(sb, level, fmt, ...) \
928 __vfs_msg(sb, level, fmt, ##__VA_ARGS__)
930 #define vfs_msg(sb, level, fmt, ...) \
932 no_printk(fmt, ##__VA_ARGS__); \
933 __vfs_msg(sb, "", " "); \
937 extern int blk_register_queue(struct gendisk *disk);
938 extern void blk_unregister_queue(struct gendisk *disk);
939 extern blk_qc_t generic_make_request(struct bio *bio);
940 extern void blk_rq_init(struct request_queue *q, struct request *rq);
941 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
942 extern void blk_put_request(struct request *);
943 extern void __blk_put_request(struct request_queue *, struct request *);
944 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
946 extern void blk_requeue_request(struct request_queue *, struct request *);
947 extern int blk_lld_busy(struct request_queue *q);
948 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
949 struct bio_set *bs, gfp_t gfp_mask,
950 int (*bio_ctr)(struct bio *, struct bio *, void *),
952 extern void blk_rq_unprep_clone(struct request *rq);
953 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
955 extern int blk_rq_append_bio(struct request *rq, struct bio *bio);
956 extern void blk_delay_queue(struct request_queue *, unsigned long);
957 extern void blk_queue_split(struct request_queue *, struct bio **);
958 extern void blk_recount_segments(struct request_queue *, struct bio *);
959 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
960 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
961 unsigned int, void __user *);
962 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
963 unsigned int, void __user *);
964 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
965 struct scsi_ioctl_command __user *);
967 extern int blk_queue_enter(struct request_queue *q, bool nowait);
968 extern void blk_queue_exit(struct request_queue *q);
969 extern void blk_start_queue(struct request_queue *q);
970 extern void blk_start_queue_async(struct request_queue *q);
971 extern void blk_stop_queue(struct request_queue *q);
972 extern void blk_sync_queue(struct request_queue *q);
973 extern void __blk_stop_queue(struct request_queue *q);
974 extern void __blk_run_queue(struct request_queue *q);
975 extern void __blk_run_queue_uncond(struct request_queue *q);
976 extern void blk_run_queue(struct request_queue *);
977 extern void blk_run_queue_async(struct request_queue *q);
978 extern int blk_rq_map_user(struct request_queue *, struct request *,
979 struct rq_map_data *, void __user *, unsigned long,
981 extern int blk_rq_unmap_user(struct bio *);
982 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
983 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
984 struct rq_map_data *, const struct iov_iter *,
986 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
987 struct request *, int);
988 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
989 struct request *, int, rq_end_io_fn *);
991 int blk_status_to_errno(blk_status_t status);
992 blk_status_t errno_to_blk_status(int errno);
994 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie);
996 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
998 return bdev->bd_disk->queue; /* this is never NULL */
1002 * blk_rq_pos() : the current sector
1003 * blk_rq_bytes() : bytes left in the entire request
1004 * blk_rq_cur_bytes() : bytes left in the current segment
1005 * blk_rq_err_bytes() : bytes left till the next error boundary
1006 * blk_rq_sectors() : sectors left in the entire request
1007 * blk_rq_cur_sectors() : sectors left in the current segment
1009 static inline sector_t blk_rq_pos(const struct request *rq)
1011 return rq->__sector;
1014 static inline unsigned int blk_rq_bytes(const struct request *rq)
1016 return rq->__data_len;
1019 static inline int blk_rq_cur_bytes(const struct request *rq)
1021 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1024 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1026 static inline unsigned int blk_rq_sectors(const struct request *rq)
1028 return blk_rq_bytes(rq) >> 9;
1031 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1033 return blk_rq_cur_bytes(rq) >> 9;
1037 * Some commands like WRITE SAME have a payload or data transfer size which
1038 * is different from the size of the request. Any driver that supports such
1039 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1040 * calculate the data transfer size.
1042 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1044 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1045 return rq->special_vec.bv_len;
1046 return blk_rq_bytes(rq);
1049 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1052 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1053 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1055 if (unlikely(op == REQ_OP_WRITE_SAME))
1056 return q->limits.max_write_same_sectors;
1058 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1059 return q->limits.max_write_zeroes_sectors;
1061 return q->limits.max_sectors;
1065 * Return maximum size of a request at given offset. Only valid for
1066 * file system requests.
1068 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1071 if (!q->limits.chunk_sectors)
1072 return q->limits.max_sectors;
1074 return q->limits.chunk_sectors -
1075 (offset & (q->limits.chunk_sectors - 1));
1078 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1081 struct request_queue *q = rq->q;
1083 if (blk_rq_is_passthrough(rq))
1084 return q->limits.max_hw_sectors;
1086 if (!q->limits.chunk_sectors ||
1087 req_op(rq) == REQ_OP_DISCARD ||
1088 req_op(rq) == REQ_OP_SECURE_ERASE)
1089 return blk_queue_get_max_sectors(q, req_op(rq));
1091 return min(blk_max_size_offset(q, offset),
1092 blk_queue_get_max_sectors(q, req_op(rq)));
1095 static inline unsigned int blk_rq_count_bios(struct request *rq)
1097 unsigned int nr_bios = 0;
1100 __rq_for_each_bio(bio, rq)
1107 * Request issue related functions.
1109 extern struct request *blk_peek_request(struct request_queue *q);
1110 extern void blk_start_request(struct request *rq);
1111 extern struct request *blk_fetch_request(struct request_queue *q);
1114 * Request completion related functions.
1116 * blk_update_request() completes given number of bytes and updates
1117 * the request without completing it.
1119 * blk_end_request() and friends. __blk_end_request() must be called
1120 * with the request queue spinlock acquired.
1122 * Several drivers define their own end_request and call
1123 * blk_end_request() for parts of the original function.
1124 * This prevents code duplication in drivers.
1126 extern bool blk_update_request(struct request *rq, blk_status_t error,
1127 unsigned int nr_bytes);
1128 extern void blk_finish_request(struct request *rq, blk_status_t error);
1129 extern bool blk_end_request(struct request *rq, blk_status_t error,
1130 unsigned int nr_bytes);
1131 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1132 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1133 unsigned int nr_bytes);
1134 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1135 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1137 extern void blk_complete_request(struct request *);
1138 extern void __blk_complete_request(struct request *);
1139 extern void blk_abort_request(struct request *);
1140 extern void blk_unprep_request(struct request *);
1143 * Access functions for manipulating queue properties
1145 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1146 spinlock_t *lock, int node_id);
1147 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1148 extern int blk_init_allocated_queue(struct request_queue *);
1149 extern void blk_cleanup_queue(struct request_queue *);
1150 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1151 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1152 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1153 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1154 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1155 extern void blk_queue_max_discard_segments(struct request_queue *,
1157 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1158 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1159 unsigned int max_discard_sectors);
1160 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1161 unsigned int max_write_same_sectors);
1162 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1163 unsigned int max_write_same_sectors);
1164 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1165 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1166 extern void blk_queue_alignment_offset(struct request_queue *q,
1167 unsigned int alignment);
1168 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1169 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1170 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1171 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1172 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1173 extern void blk_set_default_limits(struct queue_limits *lim);
1174 extern void blk_set_stacking_limits(struct queue_limits *lim);
1175 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1177 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1179 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1181 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1182 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1183 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1184 extern int blk_queue_dma_drain(struct request_queue *q,
1185 dma_drain_needed_fn *dma_drain_needed,
1186 void *buf, unsigned int size);
1187 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1188 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1189 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1190 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1191 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1192 extern void blk_queue_dma_alignment(struct request_queue *, int);
1193 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1194 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1195 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1196 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1197 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1198 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1201 * Number of physical segments as sent to the device.
1203 * Normally this is the number of discontiguous data segments sent by the
1204 * submitter. But for data-less command like discard we might have no
1205 * actual data segments submitted, but the driver might have to add it's
1206 * own special payload. In that case we still return 1 here so that this
1207 * special payload will be mapped.
1209 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1211 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1213 return rq->nr_phys_segments;
1217 * Number of discard segments (or ranges) the driver needs to fill in.
1218 * Each discard bio merged into a request is counted as one segment.
1220 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1222 return max_t(unsigned short, rq->nr_phys_segments, 1);
1225 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1226 extern void blk_dump_rq_flags(struct request *, char *);
1227 extern long nr_blockdev_pages(void);
1229 bool __must_check blk_get_queue(struct request_queue *);
1230 struct request_queue *blk_alloc_queue(gfp_t);
1231 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1232 extern void blk_put_queue(struct request_queue *);
1233 extern void blk_set_queue_dying(struct request_queue *);
1236 * block layer runtime pm functions
1239 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1240 extern int blk_pre_runtime_suspend(struct request_queue *q);
1241 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1242 extern void blk_pre_runtime_resume(struct request_queue *q);
1243 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1244 extern void blk_set_runtime_active(struct request_queue *q);
1246 static inline void blk_pm_runtime_init(struct request_queue *q,
1247 struct device *dev) {}
1248 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1252 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1253 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1254 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1255 static inline void blk_set_runtime_active(struct request_queue *q) {}
1259 * blk_plug permits building a queue of related requests by holding the I/O
1260 * fragments for a short period. This allows merging of sequential requests
1261 * into single larger request. As the requests are moved from a per-task list to
1262 * the device's request_queue in a batch, this results in improved scalability
1263 * as the lock contention for request_queue lock is reduced.
1265 * It is ok not to disable preemption when adding the request to the plug list
1266 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1267 * the plug list when the task sleeps by itself. For details, please see
1268 * schedule() where blk_schedule_flush_plug() is called.
1271 struct list_head list; /* requests */
1272 struct list_head mq_list; /* blk-mq requests */
1273 struct list_head cb_list; /* md requires an unplug callback */
1275 #define BLK_MAX_REQUEST_COUNT 16
1276 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1279 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1280 struct blk_plug_cb {
1281 struct list_head list;
1282 blk_plug_cb_fn callback;
1285 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1286 void *data, int size);
1287 extern void blk_start_plug(struct blk_plug *);
1288 extern void blk_finish_plug(struct blk_plug *);
1289 extern void blk_flush_plug_list(struct blk_plug *, bool);
1291 static inline void blk_flush_plug(struct task_struct *tsk)
1293 struct blk_plug *plug = tsk->plug;
1296 blk_flush_plug_list(plug, false);
1299 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1301 struct blk_plug *plug = tsk->plug;
1304 blk_flush_plug_list(plug, true);
1307 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1309 struct blk_plug *plug = tsk->plug;
1312 (!list_empty(&plug->list) ||
1313 !list_empty(&plug->mq_list) ||
1314 !list_empty(&plug->cb_list));
1320 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1321 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1322 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1323 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1324 extern void blk_queue_free_tags(struct request_queue *);
1325 extern int blk_queue_resize_tags(struct request_queue *, int);
1326 extern void blk_queue_invalidate_tags(struct request_queue *);
1327 extern struct blk_queue_tag *blk_init_tags(int, int);
1328 extern void blk_free_tags(struct blk_queue_tag *);
1330 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1333 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1335 return bqt->tag_index[tag];
1338 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1339 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1340 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1342 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1344 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1345 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1346 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1347 sector_t nr_sects, gfp_t gfp_mask, int flags,
1350 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1351 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1353 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1354 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1356 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1357 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1359 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1360 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1362 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1363 nr_blocks << (sb->s_blocksize_bits - 9),
1366 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1367 sector_t nr_blocks, gfp_t gfp_mask)
1369 return blkdev_issue_zeroout(sb->s_bdev,
1370 block << (sb->s_blocksize_bits - 9),
1371 nr_blocks << (sb->s_blocksize_bits - 9),
1375 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
1377 enum blk_default_limits {
1378 BLK_MAX_SEGMENTS = 128,
1379 BLK_SAFE_MAX_SECTORS = 255,
1380 BLK_DEF_MAX_SECTORS = 2560,
1381 BLK_MAX_SEGMENT_SIZE = 65536,
1382 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1385 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1387 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1389 return q->limits.seg_boundary_mask;
1392 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1394 return q->limits.virt_boundary_mask;
1397 static inline unsigned int queue_max_sectors(struct request_queue *q)
1399 return q->limits.max_sectors;
1402 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1404 return q->limits.max_hw_sectors;
1407 static inline unsigned short queue_max_segments(struct request_queue *q)
1409 return q->limits.max_segments;
1412 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1414 return q->limits.max_discard_segments;
1417 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1419 return q->limits.max_segment_size;
1422 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1426 if (q && q->limits.logical_block_size)
1427 retval = q->limits.logical_block_size;
1432 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1434 return queue_logical_block_size(bdev_get_queue(bdev));
1437 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1439 return q->limits.physical_block_size;
1442 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1444 return queue_physical_block_size(bdev_get_queue(bdev));
1447 static inline unsigned int queue_io_min(struct request_queue *q)
1449 return q->limits.io_min;
1452 static inline int bdev_io_min(struct block_device *bdev)
1454 return queue_io_min(bdev_get_queue(bdev));
1457 static inline unsigned int queue_io_opt(struct request_queue *q)
1459 return q->limits.io_opt;
1462 static inline int bdev_io_opt(struct block_device *bdev)
1464 return queue_io_opt(bdev_get_queue(bdev));
1467 static inline int queue_alignment_offset(struct request_queue *q)
1469 if (q->limits.misaligned)
1472 return q->limits.alignment_offset;
1475 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1477 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1478 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1480 return (granularity + lim->alignment_offset - alignment) % granularity;
1483 static inline int bdev_alignment_offset(struct block_device *bdev)
1485 struct request_queue *q = bdev_get_queue(bdev);
1487 if (q->limits.misaligned)
1490 if (bdev != bdev->bd_contains)
1491 return bdev->bd_part->alignment_offset;
1493 return q->limits.alignment_offset;
1496 static inline int queue_discard_alignment(struct request_queue *q)
1498 if (q->limits.discard_misaligned)
1501 return q->limits.discard_alignment;
1504 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1506 unsigned int alignment, granularity, offset;
1508 if (!lim->max_discard_sectors)
1511 /* Why are these in bytes, not sectors? */
1512 alignment = lim->discard_alignment >> 9;
1513 granularity = lim->discard_granularity >> 9;
1517 /* Offset of the partition start in 'granularity' sectors */
1518 offset = sector_div(sector, granularity);
1520 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1521 offset = (granularity + alignment - offset) % granularity;
1523 /* Turn it back into bytes, gaah */
1527 static inline int bdev_discard_alignment(struct block_device *bdev)
1529 struct request_queue *q = bdev_get_queue(bdev);
1531 if (bdev != bdev->bd_contains)
1532 return bdev->bd_part->discard_alignment;
1534 return q->limits.discard_alignment;
1537 static inline unsigned int bdev_write_same(struct block_device *bdev)
1539 struct request_queue *q = bdev_get_queue(bdev);
1542 return q->limits.max_write_same_sectors;
1547 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1549 struct request_queue *q = bdev_get_queue(bdev);
1552 return q->limits.max_write_zeroes_sectors;
1557 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1559 struct request_queue *q = bdev_get_queue(bdev);
1562 return blk_queue_zoned_model(q);
1564 return BLK_ZONED_NONE;
1567 static inline bool bdev_is_zoned(struct block_device *bdev)
1569 struct request_queue *q = bdev_get_queue(bdev);
1572 return blk_queue_is_zoned(q);
1577 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1579 struct request_queue *q = bdev_get_queue(bdev);
1582 return blk_queue_zone_sectors(q);
1587 static inline int queue_dma_alignment(struct request_queue *q)
1589 return q ? q->dma_alignment : 511;
1592 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1595 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1596 return !(addr & alignment) && !(len & alignment);
1599 /* assumes size > 256 */
1600 static inline unsigned int blksize_bits(unsigned int size)
1602 unsigned int bits = 8;
1606 } while (size > 256);
1610 static inline unsigned int block_size(struct block_device *bdev)
1612 return bdev->bd_block_size;
1615 static inline bool queue_flush_queueable(struct request_queue *q)
1617 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1620 typedef struct {struct page *v;} Sector;
1622 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1624 static inline void put_dev_sector(Sector p)
1629 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1630 struct bio_vec *bprv, unsigned int offset)
1633 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1637 * Check if adding a bio_vec after bprv with offset would create a gap in
1638 * the SG list. Most drivers don't care about this, but some do.
1640 static inline bool bvec_gap_to_prev(struct request_queue *q,
1641 struct bio_vec *bprv, unsigned int offset)
1643 if (!queue_virt_boundary(q))
1645 return __bvec_gap_to_prev(q, bprv, offset);
1649 * Check if the two bvecs from two bios can be merged to one segment.
1650 * If yes, no need to check gap between the two bios since the 1st bio
1651 * and the 1st bvec in the 2nd bio can be handled in one segment.
1653 static inline bool bios_segs_mergeable(struct request_queue *q,
1654 struct bio *prev, struct bio_vec *prev_last_bv,
1655 struct bio_vec *next_first_bv)
1657 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1659 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1661 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1662 queue_max_segment_size(q))
1667 static inline bool bio_will_gap(struct request_queue *q,
1668 struct request *prev_rq,
1672 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1673 struct bio_vec pb, nb;
1676 * don't merge if the 1st bio starts with non-zero
1677 * offset, otherwise it is quite difficult to respect
1678 * sg gap limit. We work hard to merge a huge number of small
1679 * single bios in case of mkfs.
1682 bio_get_first_bvec(prev_rq->bio, &pb);
1684 bio_get_first_bvec(prev, &pb);
1689 * We don't need to worry about the situation that the
1690 * merged segment ends in unaligned virt boundary:
1692 * - if 'pb' ends aligned, the merged segment ends aligned
1693 * - if 'pb' ends unaligned, the next bio must include
1694 * one single bvec of 'nb', otherwise the 'nb' can't
1697 bio_get_last_bvec(prev, &pb);
1698 bio_get_first_bvec(next, &nb);
1700 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1701 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1707 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1709 return bio_will_gap(req->q, req, req->biotail, bio);
1712 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1714 return bio_will_gap(req->q, NULL, bio, req->bio);
1717 int kblockd_schedule_work(struct work_struct *work);
1718 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1719 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1720 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1721 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1723 #ifdef CONFIG_BLK_CGROUP
1725 * This should not be using sched_clock(). A real patch is in progress
1726 * to fix this up, until that is in place we need to disable preemption
1727 * around sched_clock() in this function and set_io_start_time_ns().
1729 static inline void set_start_time_ns(struct request *req)
1732 req->start_time_ns = sched_clock();
1736 static inline void set_io_start_time_ns(struct request *req)
1739 req->io_start_time_ns = sched_clock();
1743 static inline uint64_t rq_start_time_ns(struct request *req)
1745 return req->start_time_ns;
1748 static inline uint64_t rq_io_start_time_ns(struct request *req)
1750 return req->io_start_time_ns;
1753 static inline void set_start_time_ns(struct request *req) {}
1754 static inline void set_io_start_time_ns(struct request *req) {}
1755 static inline uint64_t rq_start_time_ns(struct request *req)
1759 static inline uint64_t rq_io_start_time_ns(struct request *req)
1765 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1766 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1767 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1768 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1770 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1772 enum blk_integrity_flags {
1773 BLK_INTEGRITY_VERIFY = 1 << 0,
1774 BLK_INTEGRITY_GENERATE = 1 << 1,
1775 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1776 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1779 struct blk_integrity_iter {
1783 unsigned int data_size;
1784 unsigned short interval;
1785 const char *disk_name;
1788 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1790 struct blk_integrity_profile {
1791 integrity_processing_fn *generate_fn;
1792 integrity_processing_fn *verify_fn;
1796 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1797 extern void blk_integrity_unregister(struct gendisk *);
1798 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1799 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1800 struct scatterlist *);
1801 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1802 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1804 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1807 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1809 struct blk_integrity *bi = &disk->queue->integrity;
1818 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1820 return blk_get_integrity(bdev->bd_disk);
1823 static inline bool blk_integrity_rq(struct request *rq)
1825 return rq->cmd_flags & REQ_INTEGRITY;
1828 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1831 q->limits.max_integrity_segments = segs;
1834 static inline unsigned short
1835 queue_max_integrity_segments(struct request_queue *q)
1837 return q->limits.max_integrity_segments;
1840 static inline bool integrity_req_gap_back_merge(struct request *req,
1843 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1844 struct bio_integrity_payload *bip_next = bio_integrity(next);
1846 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1847 bip_next->bip_vec[0].bv_offset);
1850 static inline bool integrity_req_gap_front_merge(struct request *req,
1853 struct bio_integrity_payload *bip = bio_integrity(bio);
1854 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1856 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1857 bip_next->bip_vec[0].bv_offset);
1860 #else /* CONFIG_BLK_DEV_INTEGRITY */
1863 struct block_device;
1865 struct blk_integrity;
1867 static inline int blk_integrity_rq(struct request *rq)
1871 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1876 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1878 struct scatterlist *s)
1882 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1886 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1890 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1894 static inline void blk_integrity_register(struct gendisk *d,
1895 struct blk_integrity *b)
1898 static inline void blk_integrity_unregister(struct gendisk *d)
1901 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1905 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1909 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1915 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1922 static inline bool integrity_req_gap_back_merge(struct request *req,
1927 static inline bool integrity_req_gap_front_merge(struct request *req,
1933 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1935 struct block_device_operations {
1936 int (*open) (struct block_device *, fmode_t);
1937 void (*release) (struct gendisk *, fmode_t);
1938 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1939 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1940 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1941 unsigned int (*check_events) (struct gendisk *disk,
1942 unsigned int clearing);
1943 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1944 int (*media_changed) (struct gendisk *);
1945 void (*unlock_native_capacity) (struct gendisk *);
1946 int (*revalidate_disk) (struct gendisk *);
1947 int (*getgeo)(struct block_device *, struct hd_geometry *);
1948 /* this callback is with swap_lock and sometimes page table lock held */
1949 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1950 struct module *owner;
1951 const struct pr_ops *pr_ops;
1954 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1956 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1957 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1958 struct writeback_control *);
1959 #else /* CONFIG_BLOCK */
1961 struct block_device;
1964 * stubs for when the block layer is configured out
1966 #define buffer_heads_over_limit 0
1968 static inline long nr_blockdev_pages(void)
1976 static inline void blk_start_plug(struct blk_plug *plug)
1980 static inline void blk_finish_plug(struct blk_plug *plug)
1984 static inline void blk_flush_plug(struct task_struct *task)
1988 static inline void blk_schedule_flush_plug(struct task_struct *task)
1993 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1998 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1999 sector_t *error_sector)
2004 #endif /* CONFIG_BLOCK */