1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
32 struct scsi_ioctl_command;
35 struct elevator_queue;
41 struct blk_flush_queue;
44 struct blk_queue_stats;
45 struct blk_stat_callback;
47 #define BLKDEV_MIN_RQ 4
48 #define BLKDEV_MAX_RQ 128 /* Default maximum */
50 /* Must be consisitent with blk_mq_poll_stats_bkt() */
51 #define BLK_MQ_POLL_STATS_BKTS 16
54 * Maximum number of blkcg policies allowed to be registered concurrently.
55 * Defined here to simplify include dependency.
57 #define BLKCG_MAX_POLS 3
59 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
61 #define BLK_RL_SYNCFULL (1U << 0)
62 #define BLK_RL_ASYNCFULL (1U << 1)
65 struct request_queue *q; /* the queue this rl belongs to */
66 #ifdef CONFIG_BLK_CGROUP
67 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
70 * count[], starved[], and wait[] are indexed by
71 * BLK_RW_SYNC/BLK_RW_ASYNC
76 wait_queue_head_t wait[2];
82 typedef __u32 __bitwise req_flags_t;
84 /* elevator knows about this request */
85 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
86 /* drive already may have started this one */
87 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
88 /* uses tagged queueing */
89 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
90 /* may not be passed by ioscheduler */
91 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
92 /* request for flush sequence */
93 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
94 /* merge of different types, fail separately */
95 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
96 /* track inflight for MQ */
97 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
98 /* don't call prep for this one */
99 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
100 /* set for "ide_preempt" requests and also for requests for which the SCSI
101 "quiesce" state must be ignored. */
102 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
103 /* contains copies of user pages */
104 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
105 /* vaguely specified driver internal error. Ignored by the block layer */
106 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
107 /* don't warn about errors */
108 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
109 /* elevator private data attached */
110 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
111 /* account I/O stat */
112 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
113 /* request came from our alloc pool */
114 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
115 /* runtime pm request */
116 #define RQF_PM ((__force req_flags_t)(1 << 15))
117 /* on IO scheduler merge hash */
118 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
119 /* IO stats tracking on */
120 #define RQF_STATS ((__force req_flags_t)(1 << 17))
121 /* Look at ->special_vec for the actual data payload instead of the
123 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
125 /* flags that prevent us from merging requests: */
126 #define RQF_NOMERGE_FLAGS \
127 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
130 * Try to put the fields that are referenced together in the same cacheline.
132 * If you modify this structure, make sure to update blk_rq_init() and
133 * especially blk_mq_rq_ctx_init() to take care of the added fields.
136 struct list_head queuelist;
138 struct __call_single_data csd;
142 struct request_queue *q;
143 struct blk_mq_ctx *mq_ctx;
146 unsigned int cmd_flags; /* op and common flags */
147 req_flags_t rq_flags;
151 unsigned long atomic_flags;
153 /* the following two fields are internal, NEVER access directly */
154 unsigned int __data_len; /* total data len */
156 sector_t __sector; /* sector cursor */
162 * The hash is used inside the scheduler, and killed once the
163 * request reaches the dispatch list. The ipi_list is only used
164 * to queue the request for softirq completion, which is long
165 * after the request has been unhashed (and even removed from
166 * the dispatch list).
169 struct hlist_node hash; /* merge hash */
170 struct list_head ipi_list;
174 * The rb_node is only used inside the io scheduler, requests
175 * are pruned when moved to the dispatch queue. So let the
176 * completion_data share space with the rb_node.
179 struct rb_node rb_node; /* sort/lookup */
180 struct bio_vec special_vec;
181 void *completion_data;
182 int error_count; /* for legacy drivers, don't use */
186 * Three pointers are available for the IO schedulers, if they need
187 * more they have to dynamically allocate it. Flush requests are
188 * never put on the IO scheduler. So let the flush fields share
189 * space with the elevator data.
199 struct list_head list;
200 rq_end_io_fn *saved_end_io;
204 struct gendisk *rq_disk;
205 struct hd_struct *part;
206 unsigned long start_time;
207 struct blk_issue_stat issue_stat;
208 #ifdef CONFIG_BLK_CGROUP
209 struct request_list *rl; /* rl this rq is alloced from */
210 unsigned long long start_time_ns;
211 unsigned long long io_start_time_ns; /* when passed to hardware */
213 /* Number of scatter-gather DMA addr+len pairs after
214 * physical address coalescing is performed.
216 unsigned short nr_phys_segments;
217 #if defined(CONFIG_BLK_DEV_INTEGRITY)
218 unsigned short nr_integrity_segments;
221 unsigned short ioprio;
223 unsigned int timeout;
225 void *special; /* opaque pointer available for LLD use */
227 unsigned int extra_len; /* length of alignment and padding */
229 unsigned short write_hint;
231 unsigned long deadline;
232 struct list_head timeout_list;
235 * completion callback.
237 rq_end_io_fn *end_io;
241 struct request *next_rq;
244 static inline bool blk_op_is_scsi(unsigned int op)
246 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
249 static inline bool blk_op_is_private(unsigned int op)
251 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
254 static inline bool blk_rq_is_scsi(struct request *rq)
256 return blk_op_is_scsi(req_op(rq));
259 static inline bool blk_rq_is_private(struct request *rq)
261 return blk_op_is_private(req_op(rq));
264 static inline bool blk_rq_is_passthrough(struct request *rq)
266 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
269 static inline bool bio_is_passthrough(struct bio *bio)
271 unsigned op = bio_op(bio);
273 return blk_op_is_scsi(op) || blk_op_is_private(op);
276 static inline unsigned short req_get_ioprio(struct request *req)
281 #include <linux/elevator.h>
283 struct blk_queue_ctx;
285 typedef void (request_fn_proc) (struct request_queue *q);
286 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
287 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
288 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
289 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
292 typedef void (softirq_done_fn)(struct request *);
293 typedef int (dma_drain_needed_fn)(struct request *);
294 typedef int (lld_busy_fn) (struct request_queue *q);
295 typedef int (bsg_job_fn) (struct bsg_job *);
296 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
297 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
299 enum blk_eh_timer_return {
305 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
307 enum blk_queue_state {
312 struct blk_queue_tag {
313 struct request **tag_index; /* map of busy tags */
314 unsigned long *tag_map; /* bit map of free/busy tags */
315 int max_depth; /* what we will send to device */
316 int real_max_depth; /* what the array can hold */
317 atomic_t refcnt; /* map can be shared */
318 int alloc_policy; /* tag allocation policy */
319 int next_tag; /* next tag */
321 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
322 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
324 #define BLK_SCSI_MAX_CMDS (256)
325 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
328 * Zoned block device models (zoned limit).
330 enum blk_zoned_model {
331 BLK_ZONED_NONE, /* Regular block device */
332 BLK_ZONED_HA, /* Host-aware zoned block device */
333 BLK_ZONED_HM, /* Host-managed zoned block device */
336 struct queue_limits {
337 unsigned long bounce_pfn;
338 unsigned long seg_boundary_mask;
339 unsigned long virt_boundary_mask;
341 unsigned int max_hw_sectors;
342 unsigned int max_dev_sectors;
343 unsigned int chunk_sectors;
344 unsigned int max_sectors;
345 unsigned int max_segment_size;
346 unsigned int physical_block_size;
347 unsigned int alignment_offset;
350 unsigned int max_discard_sectors;
351 unsigned int max_hw_discard_sectors;
352 unsigned int max_write_same_sectors;
353 unsigned int max_write_zeroes_sectors;
354 unsigned int discard_granularity;
355 unsigned int discard_alignment;
357 unsigned short logical_block_size;
358 unsigned short max_segments;
359 unsigned short max_integrity_segments;
360 unsigned short max_discard_segments;
362 unsigned char misaligned;
363 unsigned char discard_misaligned;
364 unsigned char cluster;
365 unsigned char raid_partial_stripes_expensive;
366 enum blk_zoned_model zoned;
369 #ifdef CONFIG_BLK_DEV_ZONED
371 struct blk_zone_report_hdr {
372 unsigned int nr_zones;
376 extern int blkdev_report_zones(struct block_device *bdev,
377 sector_t sector, struct blk_zone *zones,
378 unsigned int *nr_zones, gfp_t gfp_mask);
379 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
380 sector_t nr_sectors, gfp_t gfp_mask);
382 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
383 unsigned int cmd, unsigned long arg);
384 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
385 unsigned int cmd, unsigned long arg);
387 #else /* CONFIG_BLK_DEV_ZONED */
389 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
390 fmode_t mode, unsigned int cmd,
396 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
397 fmode_t mode, unsigned int cmd,
403 #endif /* CONFIG_BLK_DEV_ZONED */
405 struct request_queue {
407 * Together with queue_head for cacheline sharing
409 struct list_head queue_head;
410 struct request *last_merge;
411 struct elevator_queue *elevator;
412 int nr_rqs[2]; /* # allocated [a]sync rqs */
413 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
415 atomic_t shared_hctx_restart;
417 struct blk_queue_stats *stats;
421 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
422 * is used, root blkg allocates from @q->root_rl and all other
423 * blkgs from their own blkg->rl. Which one to use should be
424 * determined using bio_request_list().
426 struct request_list root_rl;
428 request_fn_proc *request_fn;
429 make_request_fn *make_request_fn;
431 prep_rq_fn *prep_rq_fn;
432 unprep_rq_fn *unprep_rq_fn;
433 softirq_done_fn *softirq_done_fn;
434 rq_timed_out_fn *rq_timed_out_fn;
435 dma_drain_needed_fn *dma_drain_needed;
436 lld_busy_fn *lld_busy_fn;
437 /* Called just after a request is allocated */
438 init_rq_fn *init_rq_fn;
439 /* Called just before a request is freed */
440 exit_rq_fn *exit_rq_fn;
441 /* Called from inside blk_get_request() */
442 void (*initialize_rq_fn)(struct request *rq);
444 const struct blk_mq_ops *mq_ops;
446 unsigned int *mq_map;
449 struct blk_mq_ctx __percpu *queue_ctx;
450 unsigned int nr_queues;
452 unsigned int queue_depth;
454 /* hw dispatch queues */
455 struct blk_mq_hw_ctx **queue_hw_ctx;
456 unsigned int nr_hw_queues;
459 * Dispatch queue sorting
462 struct request *boundary_rq;
465 * Delayed queue handling
467 struct delayed_work delay_work;
469 struct backing_dev_info *backing_dev_info;
472 * The queue owner gets to use this for whatever they like.
473 * ll_rw_blk doesn't touch it.
478 * various queue flags, see QUEUE_* below
480 unsigned long queue_flags;
483 * ida allocated id for this queue. Used to index queues from
489 * queue needs bounce pages for pages above this limit
494 * protects queue structures from reentrancy. ->__queue_lock should
495 * _never_ be used directly, it is queue private. always use
498 spinlock_t __queue_lock;
499 spinlock_t *queue_lock;
509 struct kobject mq_kobj;
511 #ifdef CONFIG_BLK_DEV_INTEGRITY
512 struct blk_integrity integrity;
513 #endif /* CONFIG_BLK_DEV_INTEGRITY */
518 unsigned int nr_pending;
524 unsigned long nr_requests; /* Max # of requests */
525 unsigned int nr_congestion_on;
526 unsigned int nr_congestion_off;
527 unsigned int nr_batching;
529 unsigned int dma_drain_size;
530 void *dma_drain_buffer;
531 unsigned int dma_pad_mask;
532 unsigned int dma_alignment;
534 struct blk_queue_tag *queue_tags;
535 struct list_head tag_busy_list;
537 unsigned int nr_sorted;
538 unsigned int in_flight[2];
541 * Number of active block driver functions for which blk_drain_queue()
542 * must wait. Must be incremented around functions that unlock the
543 * queue_lock internally, e.g. scsi_request_fn().
545 unsigned int request_fn_active;
547 unsigned int rq_timeout;
550 struct blk_stat_callback *poll_cb;
551 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
553 struct timer_list timeout;
554 struct work_struct timeout_work;
555 struct list_head timeout_list;
557 struct list_head icq_list;
558 #ifdef CONFIG_BLK_CGROUP
559 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
560 struct blkcg_gq *root_blkg;
561 struct list_head blkg_list;
564 struct queue_limits limits;
569 unsigned int sg_timeout;
570 unsigned int sg_reserved_size;
572 #ifdef CONFIG_BLK_DEV_IO_TRACE
573 struct blk_trace *blk_trace;
574 struct mutex blk_trace_mutex;
577 * for flush operations
579 struct blk_flush_queue *fq;
581 struct list_head requeue_list;
582 spinlock_t requeue_lock;
583 struct delayed_work requeue_work;
585 struct mutex sysfs_lock;
588 atomic_t mq_freeze_depth;
590 #if defined(CONFIG_BLK_DEV_BSG)
591 bsg_job_fn *bsg_job_fn;
592 struct bsg_class_device bsg_dev;
595 #ifdef CONFIG_BLK_DEV_THROTTLING
597 struct throtl_data *td;
599 struct rcu_head rcu_head;
600 wait_queue_head_t mq_freeze_wq;
601 struct percpu_ref q_usage_counter;
602 struct list_head all_q_node;
604 struct blk_mq_tag_set *tag_set;
605 struct list_head tag_set_list;
606 struct bio_set *bio_split;
608 #ifdef CONFIG_BLK_DEBUG_FS
609 struct dentry *debugfs_dir;
610 struct dentry *sched_debugfs_dir;
613 bool mq_sysfs_init_done;
618 struct work_struct release_work;
620 #define BLK_MAX_WRITE_HINTS 5
621 u64 write_hints[BLK_MAX_WRITE_HINTS];
624 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
625 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
626 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
627 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
628 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
629 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
630 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
631 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
632 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
633 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
634 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
635 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
636 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
637 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
638 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
639 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
640 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
641 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
642 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
643 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
644 #define QUEUE_FLAG_WC 20 /* Write back caching */
645 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
646 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
647 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
648 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
649 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
650 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
651 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
652 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
653 #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */
655 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
656 (1 << QUEUE_FLAG_SAME_COMP) | \
657 (1 << QUEUE_FLAG_ADD_RANDOM))
659 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
660 (1 << QUEUE_FLAG_SAME_COMP) | \
661 (1 << QUEUE_FLAG_POLL))
664 * @q->queue_lock is set while a queue is being initialized. Since we know
665 * that no other threads access the queue object before @q->queue_lock has
666 * been set, it is safe to manipulate queue flags without holding the
667 * queue_lock if @q->queue_lock == NULL. See also blk_alloc_queue_node() and
668 * blk_init_allocated_queue().
670 static inline void queue_lockdep_assert_held(struct request_queue *q)
673 lockdep_assert_held(q->queue_lock);
676 static inline void queue_flag_set_unlocked(unsigned int flag,
677 struct request_queue *q)
679 __set_bit(flag, &q->queue_flags);
682 static inline int queue_flag_test_and_clear(unsigned int flag,
683 struct request_queue *q)
685 queue_lockdep_assert_held(q);
687 if (test_bit(flag, &q->queue_flags)) {
688 __clear_bit(flag, &q->queue_flags);
695 static inline int queue_flag_test_and_set(unsigned int flag,
696 struct request_queue *q)
698 queue_lockdep_assert_held(q);
700 if (!test_bit(flag, &q->queue_flags)) {
701 __set_bit(flag, &q->queue_flags);
708 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
710 queue_lockdep_assert_held(q);
711 __set_bit(flag, &q->queue_flags);
714 static inline void queue_flag_clear_unlocked(unsigned int flag,
715 struct request_queue *q)
717 __clear_bit(flag, &q->queue_flags);
720 static inline int queue_in_flight(struct request_queue *q)
722 return q->in_flight[0] + q->in_flight[1];
725 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
727 queue_lockdep_assert_held(q);
728 __clear_bit(flag, &q->queue_flags);
731 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
732 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
733 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
734 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
735 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
736 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
737 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
738 #define blk_queue_noxmerges(q) \
739 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
740 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
741 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
742 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
743 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
744 #define blk_queue_secure_erase(q) \
745 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
746 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
747 #define blk_queue_scsi_passthrough(q) \
748 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
750 #define blk_noretry_request(rq) \
751 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
752 REQ_FAILFAST_DRIVER))
753 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
754 #define blk_queue_preempt_only(q) \
755 test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
757 extern int blk_set_preempt_only(struct request_queue *q);
758 extern void blk_clear_preempt_only(struct request_queue *q);
760 static inline bool blk_account_rq(struct request *rq)
762 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
765 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
766 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
767 /* rq->queuelist of dequeued request must be list_empty() */
768 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
770 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
772 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
775 * Driver can handle struct request, if it either has an old style
776 * request_fn defined, or is blk-mq based.
778 static inline bool queue_is_rq_based(struct request_queue *q)
780 return q->request_fn || q->mq_ops;
783 static inline unsigned int blk_queue_cluster(struct request_queue *q)
785 return q->limits.cluster;
788 static inline enum blk_zoned_model
789 blk_queue_zoned_model(struct request_queue *q)
791 return q->limits.zoned;
794 static inline bool blk_queue_is_zoned(struct request_queue *q)
796 switch (blk_queue_zoned_model(q)) {
805 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
807 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
810 static inline bool rq_is_sync(struct request *rq)
812 return op_is_sync(rq->cmd_flags);
815 static inline bool blk_rl_full(struct request_list *rl, bool sync)
817 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
819 return rl->flags & flag;
822 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
824 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
829 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
831 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
836 static inline bool rq_mergeable(struct request *rq)
838 if (blk_rq_is_passthrough(rq))
841 if (req_op(rq) == REQ_OP_FLUSH)
844 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
847 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
849 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
855 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
857 if (bio_page(a) == bio_page(b) &&
858 bio_offset(a) == bio_offset(b))
864 static inline unsigned int blk_queue_depth(struct request_queue *q)
867 return q->queue_depth;
869 return q->nr_requests;
873 * q->prep_rq_fn return values
876 BLKPREP_OK, /* serve it */
877 BLKPREP_KILL, /* fatal error, kill, return -EIO */
878 BLKPREP_DEFER, /* leave on queue */
879 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
882 extern unsigned long blk_max_low_pfn, blk_max_pfn;
885 * standard bounce addresses:
887 * BLK_BOUNCE_HIGH : bounce all highmem pages
888 * BLK_BOUNCE_ANY : don't bounce anything
889 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
892 #if BITS_PER_LONG == 32
893 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
895 #define BLK_BOUNCE_HIGH -1ULL
897 #define BLK_BOUNCE_ANY (-1ULL)
898 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
901 * default timeout for SG_IO if none specified
903 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
904 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
910 unsigned long offset;
915 struct req_iterator {
916 struct bvec_iter iter;
920 /* This should not be used directly - use rq_for_each_segment */
921 #define for_each_bio(_bio) \
922 for (; _bio; _bio = _bio->bi_next)
923 #define __rq_for_each_bio(_bio, rq) \
925 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
927 #define rq_for_each_segment(bvl, _rq, _iter) \
928 __rq_for_each_bio(_iter.bio, _rq) \
929 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
931 #define rq_iter_last(bvec, _iter) \
932 (_iter.bio->bi_next == NULL && \
933 bio_iter_last(bvec, _iter.iter))
935 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
936 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
938 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
939 extern void rq_flush_dcache_pages(struct request *rq);
941 static inline void rq_flush_dcache_pages(struct request *rq)
946 extern int blk_register_queue(struct gendisk *disk);
947 extern void blk_unregister_queue(struct gendisk *disk);
948 extern blk_qc_t generic_make_request(struct bio *bio);
949 extern blk_qc_t direct_make_request(struct bio *bio);
950 extern void blk_rq_init(struct request_queue *q, struct request *rq);
951 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
952 extern void blk_put_request(struct request *);
953 extern void __blk_put_request(struct request_queue *, struct request *);
954 extern struct request *blk_get_request_flags(struct request_queue *,
956 blk_mq_req_flags_t flags);
957 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
959 extern void blk_requeue_request(struct request_queue *, struct request *);
960 extern int blk_lld_busy(struct request_queue *q);
961 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
962 struct bio_set *bs, gfp_t gfp_mask,
963 int (*bio_ctr)(struct bio *, struct bio *, void *),
965 extern void blk_rq_unprep_clone(struct request *rq);
966 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
968 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
969 extern void blk_delay_queue(struct request_queue *, unsigned long);
970 extern void blk_queue_split(struct request_queue *, struct bio **);
971 extern void blk_recount_segments(struct request_queue *, struct bio *);
972 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
973 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
974 unsigned int, void __user *);
975 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
976 unsigned int, void __user *);
977 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
978 struct scsi_ioctl_command __user *);
980 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
981 extern void blk_queue_exit(struct request_queue *q);
982 extern void blk_start_queue(struct request_queue *q);
983 extern void blk_start_queue_async(struct request_queue *q);
984 extern void blk_stop_queue(struct request_queue *q);
985 extern void blk_sync_queue(struct request_queue *q);
986 extern void __blk_stop_queue(struct request_queue *q);
987 extern void __blk_run_queue(struct request_queue *q);
988 extern void __blk_run_queue_uncond(struct request_queue *q);
989 extern void blk_run_queue(struct request_queue *);
990 extern void blk_run_queue_async(struct request_queue *q);
991 extern int blk_rq_map_user(struct request_queue *, struct request *,
992 struct rq_map_data *, void __user *, unsigned long,
994 extern int blk_rq_unmap_user(struct bio *);
995 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
996 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
997 struct rq_map_data *, const struct iov_iter *,
999 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1000 struct request *, int);
1001 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1002 struct request *, int, rq_end_io_fn *);
1004 int blk_status_to_errno(blk_status_t status);
1005 blk_status_t errno_to_blk_status(int errno);
1007 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1009 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1011 return bdev->bd_disk->queue; /* this is never NULL */
1015 * blk_rq_pos() : the current sector
1016 * blk_rq_bytes() : bytes left in the entire request
1017 * blk_rq_cur_bytes() : bytes left in the current segment
1018 * blk_rq_err_bytes() : bytes left till the next error boundary
1019 * blk_rq_sectors() : sectors left in the entire request
1020 * blk_rq_cur_sectors() : sectors left in the current segment
1022 static inline sector_t blk_rq_pos(const struct request *rq)
1024 return rq->__sector;
1027 static inline unsigned int blk_rq_bytes(const struct request *rq)
1029 return rq->__data_len;
1032 static inline int blk_rq_cur_bytes(const struct request *rq)
1034 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1037 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1039 static inline unsigned int blk_rq_sectors(const struct request *rq)
1041 return blk_rq_bytes(rq) >> 9;
1044 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1046 return blk_rq_cur_bytes(rq) >> 9;
1050 * Some commands like WRITE SAME have a payload or data transfer size which
1051 * is different from the size of the request. Any driver that supports such
1052 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1053 * calculate the data transfer size.
1055 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1057 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1058 return rq->special_vec.bv_len;
1059 return blk_rq_bytes(rq);
1062 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1065 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1066 return min(q->limits.max_discard_sectors, UINT_MAX >> 9);
1068 if (unlikely(op == REQ_OP_WRITE_SAME))
1069 return q->limits.max_write_same_sectors;
1071 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1072 return q->limits.max_write_zeroes_sectors;
1074 return q->limits.max_sectors;
1078 * Return maximum size of a request at given offset. Only valid for
1079 * file system requests.
1081 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1084 if (!q->limits.chunk_sectors)
1085 return q->limits.max_sectors;
1087 return q->limits.chunk_sectors -
1088 (offset & (q->limits.chunk_sectors - 1));
1091 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1094 struct request_queue *q = rq->q;
1096 if (blk_rq_is_passthrough(rq))
1097 return q->limits.max_hw_sectors;
1099 if (!q->limits.chunk_sectors ||
1100 req_op(rq) == REQ_OP_DISCARD ||
1101 req_op(rq) == REQ_OP_SECURE_ERASE)
1102 return blk_queue_get_max_sectors(q, req_op(rq));
1104 return min(blk_max_size_offset(q, offset),
1105 blk_queue_get_max_sectors(q, req_op(rq)));
1108 static inline unsigned int blk_rq_count_bios(struct request *rq)
1110 unsigned int nr_bios = 0;
1113 __rq_for_each_bio(bio, rq)
1120 * Request issue related functions.
1122 extern struct request *blk_peek_request(struct request_queue *q);
1123 extern void blk_start_request(struct request *rq);
1124 extern struct request *blk_fetch_request(struct request_queue *q);
1126 void blk_steal_bios(struct bio_list *list, struct request *rq);
1129 * Request completion related functions.
1131 * blk_update_request() completes given number of bytes and updates
1132 * the request without completing it.
1134 * blk_end_request() and friends. __blk_end_request() must be called
1135 * with the request queue spinlock acquired.
1137 * Several drivers define their own end_request and call
1138 * blk_end_request() for parts of the original function.
1139 * This prevents code duplication in drivers.
1141 extern bool blk_update_request(struct request *rq, blk_status_t error,
1142 unsigned int nr_bytes);
1143 extern void blk_finish_request(struct request *rq, blk_status_t error);
1144 extern bool blk_end_request(struct request *rq, blk_status_t error,
1145 unsigned int nr_bytes);
1146 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1147 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1148 unsigned int nr_bytes);
1149 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1150 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1152 extern void blk_complete_request(struct request *);
1153 extern void __blk_complete_request(struct request *);
1154 extern void blk_abort_request(struct request *);
1155 extern void blk_unprep_request(struct request *);
1158 * Access functions for manipulating queue properties
1160 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1161 spinlock_t *lock, int node_id);
1162 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1163 extern int blk_init_allocated_queue(struct request_queue *);
1164 extern void blk_cleanup_queue(struct request_queue *);
1165 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1166 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1167 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1168 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1169 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1170 extern void blk_queue_max_discard_segments(struct request_queue *,
1172 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1173 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1174 unsigned int max_discard_sectors);
1175 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1176 unsigned int max_write_same_sectors);
1177 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1178 unsigned int max_write_same_sectors);
1179 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1180 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1181 extern void blk_queue_alignment_offset(struct request_queue *q,
1182 unsigned int alignment);
1183 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1184 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1185 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1186 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1187 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1188 extern void blk_set_default_limits(struct queue_limits *lim);
1189 extern void blk_set_stacking_limits(struct queue_limits *lim);
1190 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1192 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1194 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1196 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1197 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1198 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1199 extern int blk_queue_dma_drain(struct request_queue *q,
1200 dma_drain_needed_fn *dma_drain_needed,
1201 void *buf, unsigned int size);
1202 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1203 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1204 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1205 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1206 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1207 extern void blk_queue_dma_alignment(struct request_queue *, int);
1208 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1209 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1210 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1211 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1212 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1213 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1216 * Number of physical segments as sent to the device.
1218 * Normally this is the number of discontiguous data segments sent by the
1219 * submitter. But for data-less command like discard we might have no
1220 * actual data segments submitted, but the driver might have to add it's
1221 * own special payload. In that case we still return 1 here so that this
1222 * special payload will be mapped.
1224 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1226 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1228 return rq->nr_phys_segments;
1232 * Number of discard segments (or ranges) the driver needs to fill in.
1233 * Each discard bio merged into a request is counted as one segment.
1235 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1237 return max_t(unsigned short, rq->nr_phys_segments, 1);
1240 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1241 extern void blk_dump_rq_flags(struct request *, char *);
1242 extern long nr_blockdev_pages(void);
1244 bool __must_check blk_get_queue(struct request_queue *);
1245 struct request_queue *blk_alloc_queue(gfp_t);
1246 struct request_queue *blk_alloc_queue_node(gfp_t, int);
1247 extern void blk_put_queue(struct request_queue *);
1248 extern void blk_set_queue_dying(struct request_queue *);
1251 * block layer runtime pm functions
1254 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1255 extern int blk_pre_runtime_suspend(struct request_queue *q);
1256 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1257 extern void blk_pre_runtime_resume(struct request_queue *q);
1258 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1259 extern void blk_set_runtime_active(struct request_queue *q);
1261 static inline void blk_pm_runtime_init(struct request_queue *q,
1262 struct device *dev) {}
1263 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1267 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1268 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1269 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1270 static inline void blk_set_runtime_active(struct request_queue *q) {}
1274 * blk_plug permits building a queue of related requests by holding the I/O
1275 * fragments for a short period. This allows merging of sequential requests
1276 * into single larger request. As the requests are moved from a per-task list to
1277 * the device's request_queue in a batch, this results in improved scalability
1278 * as the lock contention for request_queue lock is reduced.
1280 * It is ok not to disable preemption when adding the request to the plug list
1281 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1282 * the plug list when the task sleeps by itself. For details, please see
1283 * schedule() where blk_schedule_flush_plug() is called.
1286 struct list_head list; /* requests */
1287 struct list_head mq_list; /* blk-mq requests */
1288 struct list_head cb_list; /* md requires an unplug callback */
1290 #define BLK_MAX_REQUEST_COUNT 16
1291 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1294 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1295 struct blk_plug_cb {
1296 struct list_head list;
1297 blk_plug_cb_fn callback;
1300 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1301 void *data, int size);
1302 extern void blk_start_plug(struct blk_plug *);
1303 extern void blk_finish_plug(struct blk_plug *);
1304 extern void blk_flush_plug_list(struct blk_plug *, bool);
1306 static inline void blk_flush_plug(struct task_struct *tsk)
1308 struct blk_plug *plug = tsk->plug;
1311 blk_flush_plug_list(plug, false);
1314 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1316 struct blk_plug *plug = tsk->plug;
1319 blk_flush_plug_list(plug, true);
1322 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1324 struct blk_plug *plug = tsk->plug;
1327 (!list_empty(&plug->list) ||
1328 !list_empty(&plug->mq_list) ||
1329 !list_empty(&plug->cb_list));
1335 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1336 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1337 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1338 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1339 extern void blk_queue_free_tags(struct request_queue *);
1340 extern int blk_queue_resize_tags(struct request_queue *, int);
1341 extern void blk_queue_invalidate_tags(struct request_queue *);
1342 extern struct blk_queue_tag *blk_init_tags(int, int);
1343 extern void blk_free_tags(struct blk_queue_tag *);
1345 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1348 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1350 return bqt->tag_index[tag];
1353 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1354 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1355 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1357 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1359 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1360 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1361 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1362 sector_t nr_sects, gfp_t gfp_mask, int flags,
1365 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1366 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1368 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1369 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1371 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1372 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1374 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1375 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1377 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9),
1378 nr_blocks << (sb->s_blocksize_bits - 9),
1381 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1382 sector_t nr_blocks, gfp_t gfp_mask)
1384 return blkdev_issue_zeroout(sb->s_bdev,
1385 block << (sb->s_blocksize_bits - 9),
1386 nr_blocks << (sb->s_blocksize_bits - 9),
1390 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1392 enum blk_default_limits {
1393 BLK_MAX_SEGMENTS = 128,
1394 BLK_SAFE_MAX_SECTORS = 255,
1395 BLK_DEF_MAX_SECTORS = 2560,
1396 BLK_MAX_SEGMENT_SIZE = 65536,
1397 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1400 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1402 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1404 return q->limits.seg_boundary_mask;
1407 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1409 return q->limits.virt_boundary_mask;
1412 static inline unsigned int queue_max_sectors(struct request_queue *q)
1414 return q->limits.max_sectors;
1417 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1419 return q->limits.max_hw_sectors;
1422 static inline unsigned short queue_max_segments(struct request_queue *q)
1424 return q->limits.max_segments;
1427 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1429 return q->limits.max_discard_segments;
1432 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1434 return q->limits.max_segment_size;
1437 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1441 if (q && q->limits.logical_block_size)
1442 retval = q->limits.logical_block_size;
1447 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1449 return queue_logical_block_size(bdev_get_queue(bdev));
1452 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1454 return q->limits.physical_block_size;
1457 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1459 return queue_physical_block_size(bdev_get_queue(bdev));
1462 static inline unsigned int queue_io_min(struct request_queue *q)
1464 return q->limits.io_min;
1467 static inline int bdev_io_min(struct block_device *bdev)
1469 return queue_io_min(bdev_get_queue(bdev));
1472 static inline unsigned int queue_io_opt(struct request_queue *q)
1474 return q->limits.io_opt;
1477 static inline int bdev_io_opt(struct block_device *bdev)
1479 return queue_io_opt(bdev_get_queue(bdev));
1482 static inline int queue_alignment_offset(struct request_queue *q)
1484 if (q->limits.misaligned)
1487 return q->limits.alignment_offset;
1490 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1492 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1493 unsigned int alignment = sector_div(sector, granularity >> 9) << 9;
1495 return (granularity + lim->alignment_offset - alignment) % granularity;
1498 static inline int bdev_alignment_offset(struct block_device *bdev)
1500 struct request_queue *q = bdev_get_queue(bdev);
1502 if (q->limits.misaligned)
1505 if (bdev != bdev->bd_contains)
1506 return bdev->bd_part->alignment_offset;
1508 return q->limits.alignment_offset;
1511 static inline int queue_discard_alignment(struct request_queue *q)
1513 if (q->limits.discard_misaligned)
1516 return q->limits.discard_alignment;
1519 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1521 unsigned int alignment, granularity, offset;
1523 if (!lim->max_discard_sectors)
1526 /* Why are these in bytes, not sectors? */
1527 alignment = lim->discard_alignment >> 9;
1528 granularity = lim->discard_granularity >> 9;
1532 /* Offset of the partition start in 'granularity' sectors */
1533 offset = sector_div(sector, granularity);
1535 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1536 offset = (granularity + alignment - offset) % granularity;
1538 /* Turn it back into bytes, gaah */
1542 static inline int bdev_discard_alignment(struct block_device *bdev)
1544 struct request_queue *q = bdev_get_queue(bdev);
1546 if (bdev != bdev->bd_contains)
1547 return bdev->bd_part->discard_alignment;
1549 return q->limits.discard_alignment;
1552 static inline unsigned int bdev_write_same(struct block_device *bdev)
1554 struct request_queue *q = bdev_get_queue(bdev);
1557 return q->limits.max_write_same_sectors;
1562 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1564 struct request_queue *q = bdev_get_queue(bdev);
1567 return q->limits.max_write_zeroes_sectors;
1572 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1574 struct request_queue *q = bdev_get_queue(bdev);
1577 return blk_queue_zoned_model(q);
1579 return BLK_ZONED_NONE;
1582 static inline bool bdev_is_zoned(struct block_device *bdev)
1584 struct request_queue *q = bdev_get_queue(bdev);
1587 return blk_queue_is_zoned(q);
1592 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1594 struct request_queue *q = bdev_get_queue(bdev);
1597 return blk_queue_zone_sectors(q);
1602 static inline int queue_dma_alignment(struct request_queue *q)
1604 return q ? q->dma_alignment : 511;
1607 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1610 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1611 return !(addr & alignment) && !(len & alignment);
1614 /* assumes size > 256 */
1615 static inline unsigned int blksize_bits(unsigned int size)
1617 unsigned int bits = 8;
1621 } while (size > 256);
1625 static inline unsigned int block_size(struct block_device *bdev)
1627 return bdev->bd_block_size;
1630 static inline bool queue_flush_queueable(struct request_queue *q)
1632 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1635 typedef struct {struct page *v;} Sector;
1637 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1639 static inline void put_dev_sector(Sector p)
1644 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1645 struct bio_vec *bprv, unsigned int offset)
1648 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1652 * Check if adding a bio_vec after bprv with offset would create a gap in
1653 * the SG list. Most drivers don't care about this, but some do.
1655 static inline bool bvec_gap_to_prev(struct request_queue *q,
1656 struct bio_vec *bprv, unsigned int offset)
1658 if (!queue_virt_boundary(q))
1660 return __bvec_gap_to_prev(q, bprv, offset);
1664 * Check if the two bvecs from two bios can be merged to one segment.
1665 * If yes, no need to check gap between the two bios since the 1st bio
1666 * and the 1st bvec in the 2nd bio can be handled in one segment.
1668 static inline bool bios_segs_mergeable(struct request_queue *q,
1669 struct bio *prev, struct bio_vec *prev_last_bv,
1670 struct bio_vec *next_first_bv)
1672 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1674 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1676 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1677 queue_max_segment_size(q))
1682 static inline bool bio_will_gap(struct request_queue *q,
1683 struct request *prev_rq,
1687 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1688 struct bio_vec pb, nb;
1691 * don't merge if the 1st bio starts with non-zero
1692 * offset, otherwise it is quite difficult to respect
1693 * sg gap limit. We work hard to merge a huge number of small
1694 * single bios in case of mkfs.
1697 bio_get_first_bvec(prev_rq->bio, &pb);
1699 bio_get_first_bvec(prev, &pb);
1704 * We don't need to worry about the situation that the
1705 * merged segment ends in unaligned virt boundary:
1707 * - if 'pb' ends aligned, the merged segment ends aligned
1708 * - if 'pb' ends unaligned, the next bio must include
1709 * one single bvec of 'nb', otherwise the 'nb' can't
1712 bio_get_last_bvec(prev, &pb);
1713 bio_get_first_bvec(next, &nb);
1715 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1716 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1722 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1724 return bio_will_gap(req->q, req, req->biotail, bio);
1727 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1729 return bio_will_gap(req->q, NULL, bio, req->bio);
1732 int kblockd_schedule_work(struct work_struct *work);
1733 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1734 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay);
1735 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1736 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1738 #ifdef CONFIG_BLK_CGROUP
1740 * This should not be using sched_clock(). A real patch is in progress
1741 * to fix this up, until that is in place we need to disable preemption
1742 * around sched_clock() in this function and set_io_start_time_ns().
1744 static inline void set_start_time_ns(struct request *req)
1747 req->start_time_ns = sched_clock();
1751 static inline void set_io_start_time_ns(struct request *req)
1754 req->io_start_time_ns = sched_clock();
1758 static inline uint64_t rq_start_time_ns(struct request *req)
1760 return req->start_time_ns;
1763 static inline uint64_t rq_io_start_time_ns(struct request *req)
1765 return req->io_start_time_ns;
1768 static inline void set_start_time_ns(struct request *req) {}
1769 static inline void set_io_start_time_ns(struct request *req) {}
1770 static inline uint64_t rq_start_time_ns(struct request *req)
1774 static inline uint64_t rq_io_start_time_ns(struct request *req)
1780 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1781 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1782 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1783 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1785 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1787 enum blk_integrity_flags {
1788 BLK_INTEGRITY_VERIFY = 1 << 0,
1789 BLK_INTEGRITY_GENERATE = 1 << 1,
1790 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1791 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1794 struct blk_integrity_iter {
1798 unsigned int data_size;
1799 unsigned short interval;
1800 const char *disk_name;
1803 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1805 struct blk_integrity_profile {
1806 integrity_processing_fn *generate_fn;
1807 integrity_processing_fn *verify_fn;
1811 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1812 extern void blk_integrity_unregister(struct gendisk *);
1813 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1814 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1815 struct scatterlist *);
1816 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1817 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1819 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1822 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1824 struct blk_integrity *bi = &disk->queue->integrity;
1833 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1835 return blk_get_integrity(bdev->bd_disk);
1838 static inline bool blk_integrity_rq(struct request *rq)
1840 return rq->cmd_flags & REQ_INTEGRITY;
1843 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1846 q->limits.max_integrity_segments = segs;
1849 static inline unsigned short
1850 queue_max_integrity_segments(struct request_queue *q)
1852 return q->limits.max_integrity_segments;
1855 static inline bool integrity_req_gap_back_merge(struct request *req,
1858 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1859 struct bio_integrity_payload *bip_next = bio_integrity(next);
1861 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1862 bip_next->bip_vec[0].bv_offset);
1865 static inline bool integrity_req_gap_front_merge(struct request *req,
1868 struct bio_integrity_payload *bip = bio_integrity(bio);
1869 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1871 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1872 bip_next->bip_vec[0].bv_offset);
1875 #else /* CONFIG_BLK_DEV_INTEGRITY */
1878 struct block_device;
1880 struct blk_integrity;
1882 static inline int blk_integrity_rq(struct request *rq)
1886 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1891 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1893 struct scatterlist *s)
1897 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1901 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1905 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1909 static inline void blk_integrity_register(struct gendisk *d,
1910 struct blk_integrity *b)
1913 static inline void blk_integrity_unregister(struct gendisk *d)
1916 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1920 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1924 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1930 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1937 static inline bool integrity_req_gap_back_merge(struct request *req,
1942 static inline bool integrity_req_gap_front_merge(struct request *req,
1948 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1950 struct block_device_operations {
1951 int (*open) (struct block_device *, fmode_t);
1952 void (*release) (struct gendisk *, fmode_t);
1953 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1954 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1955 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1956 unsigned int (*check_events) (struct gendisk *disk,
1957 unsigned int clearing);
1958 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1959 int (*media_changed) (struct gendisk *);
1960 void (*unlock_native_capacity) (struct gendisk *);
1961 int (*revalidate_disk) (struct gendisk *);
1962 int (*getgeo)(struct block_device *, struct hd_geometry *);
1963 /* this callback is with swap_lock and sometimes page table lock held */
1964 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1965 struct module *owner;
1966 const struct pr_ops *pr_ops;
1969 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1971 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1972 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1973 struct writeback_control *);
1974 #else /* CONFIG_BLOCK */
1976 struct block_device;
1979 * stubs for when the block layer is configured out
1981 #define buffer_heads_over_limit 0
1983 static inline long nr_blockdev_pages(void)
1991 static inline void blk_start_plug(struct blk_plug *plug)
1995 static inline void blk_finish_plug(struct blk_plug *plug)
1999 static inline void blk_flush_plug(struct task_struct *task)
2003 static inline void blk_schedule_flush_plug(struct task_struct *task)
2008 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2013 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2014 sector_t *error_sector)
2019 #endif /* CONFIG_BLOCK */