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>
30 #include <linux/seqlock.h>
31 #include <linux/u64_stats_sync.h>
34 struct scsi_ioctl_command;
37 struct elevator_queue;
43 struct blk_flush_queue;
46 struct blk_queue_stats;
47 struct blk_stat_callback;
49 #define BLKDEV_MIN_RQ 4
50 #define BLKDEV_MAX_RQ 128 /* Default maximum */
52 /* Must be consistent with blk_mq_poll_stats_bkt() */
53 #define BLK_MQ_POLL_STATS_BKTS 16
56 * Maximum number of blkcg policies allowed to be registered concurrently.
57 * Defined here to simplify include dependency.
59 #define BLKCG_MAX_POLS 3
61 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
63 #define BLK_RL_SYNCFULL (1U << 0)
64 #define BLK_RL_ASYNCFULL (1U << 1)
67 struct request_queue *q; /* the queue this rl belongs to */
68 #ifdef CONFIG_BLK_CGROUP
69 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
72 * count[], starved[], and wait[] are indexed by
73 * BLK_RW_SYNC/BLK_RW_ASYNC
78 wait_queue_head_t wait[2];
84 typedef __u32 __bitwise req_flags_t;
86 /* elevator knows about this request */
87 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
88 /* drive already may have started this one */
89 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
90 /* uses tagged queueing */
91 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
92 /* may not be passed by ioscheduler */
93 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
94 /* request for flush sequence */
95 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
96 /* merge of different types, fail separately */
97 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
98 /* track inflight for MQ */
99 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
100 /* don't call prep for this one */
101 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
102 /* set for "ide_preempt" requests and also for requests for which the SCSI
103 "quiesce" state must be ignored. */
104 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
105 /* contains copies of user pages */
106 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
107 /* vaguely specified driver internal error. Ignored by the block layer */
108 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
109 /* don't warn about errors */
110 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
111 /* elevator private data attached */
112 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
113 /* account I/O stat */
114 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
115 /* request came from our alloc pool */
116 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
117 /* runtime pm request */
118 #define RQF_PM ((__force req_flags_t)(1 << 15))
119 /* on IO scheduler merge hash */
120 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
121 /* IO stats tracking on */
122 #define RQF_STATS ((__force req_flags_t)(1 << 17))
123 /* Look at ->special_vec for the actual data payload instead of the
125 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
126 /* The per-zone write lock is held for this request */
127 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
128 /* timeout is expired */
129 #define RQF_MQ_TIMEOUT_EXPIRED ((__force req_flags_t)(1 << 20))
130 /* already slept for hybrid poll */
131 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 21))
133 /* flags that prevent us from merging requests: */
134 #define RQF_NOMERGE_FLAGS \
135 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
138 * Try to put the fields that are referenced together in the same cacheline.
140 * If you modify this structure, make sure to update blk_rq_init() and
141 * especially blk_mq_rq_ctx_init() to take care of the added fields.
144 struct request_queue *q;
145 struct blk_mq_ctx *mq_ctx;
148 unsigned int cmd_flags; /* op and common flags */
149 req_flags_t rq_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 */
161 struct list_head queuelist;
164 * The hash is used inside the scheduler, and killed once the
165 * request reaches the dispatch list. The ipi_list is only used
166 * to queue the request for softirq completion, which is long
167 * after the request has been unhashed (and even removed from
168 * the dispatch list).
171 struct hlist_node hash; /* merge hash */
172 struct list_head ipi_list;
176 * The rb_node is only used inside the io scheduler, requests
177 * are pruned when moved to the dispatch queue. So let the
178 * completion_data share space with the rb_node.
181 struct rb_node rb_node; /* sort/lookup */
182 struct bio_vec special_vec;
183 void *completion_data;
184 int error_count; /* for legacy drivers, don't use */
188 * Three pointers are available for the IO schedulers, if they need
189 * more they have to dynamically allocate it. Flush requests are
190 * never put on the IO scheduler. So let the flush fields share
191 * space with the elevator data.
201 struct list_head list;
202 rq_end_io_fn *saved_end_io;
206 struct gendisk *rq_disk;
207 struct hd_struct *part;
208 unsigned long start_time;
209 /* Time that I/O was submitted to the device. */
210 u64 io_start_time_ns;
212 #ifdef CONFIG_BLK_WBT
213 unsigned short wbt_flags;
215 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
216 unsigned short throtl_size;
220 * Number of scatter-gather DMA addr+len pairs after
221 * physical address coalescing is performed.
223 unsigned short nr_phys_segments;
225 #if defined(CONFIG_BLK_DEV_INTEGRITY)
226 unsigned short nr_integrity_segments;
229 unsigned short write_hint;
230 unsigned short ioprio;
232 unsigned int timeout;
234 void *special; /* opaque pointer available for LLD use */
236 unsigned int extra_len; /* length of alignment and padding */
239 * On blk-mq, the lower bits of ->gstate (generation number and
240 * state) carry the MQ_RQ_* state value and the upper bits the
241 * generation number which is monotonically incremented and used to
242 * distinguish the reuse instances.
244 * ->gstate_seq allows updates to ->gstate and other fields
245 * (currently ->deadline) during request start to be read
246 * atomically from the timeout path, so that it can operate on a
247 * coherent set of information.
249 seqcount_t gstate_seq;
253 * ->aborted_gstate is used by the timeout to claim a specific
254 * recycle instance of this request. See blk_mq_timeout_work().
256 struct u64_stats_sync aborted_gstate_sync;
259 /* access through blk_rq_set_deadline, blk_rq_deadline */
260 unsigned long __deadline;
262 struct list_head timeout_list;
265 struct __call_single_data csd;
270 * completion callback.
272 rq_end_io_fn *end_io;
276 struct request *next_rq;
278 #ifdef CONFIG_BLK_CGROUP
279 struct request_list *rl; /* rl this rq is alloced from */
280 unsigned long long cgroup_start_time_ns;
281 unsigned long long cgroup_io_start_time_ns; /* when passed to hardware */
285 static inline bool blk_op_is_scsi(unsigned int op)
287 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
290 static inline bool blk_op_is_private(unsigned int op)
292 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
295 static inline bool blk_rq_is_scsi(struct request *rq)
297 return blk_op_is_scsi(req_op(rq));
300 static inline bool blk_rq_is_private(struct request *rq)
302 return blk_op_is_private(req_op(rq));
305 static inline bool blk_rq_is_passthrough(struct request *rq)
307 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
310 static inline bool bio_is_passthrough(struct bio *bio)
312 unsigned op = bio_op(bio);
314 return blk_op_is_scsi(op) || blk_op_is_private(op);
317 static inline unsigned short req_get_ioprio(struct request *req)
322 #include <linux/elevator.h>
324 struct blk_queue_ctx;
326 typedef void (request_fn_proc) (struct request_queue *q);
327 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
328 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
329 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
330 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
333 typedef void (softirq_done_fn)(struct request *);
334 typedef int (dma_drain_needed_fn)(struct request *);
335 typedef int (lld_busy_fn) (struct request_queue *q);
336 typedef int (bsg_job_fn) (struct bsg_job *);
337 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
338 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
340 enum blk_eh_timer_return {
346 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
348 enum blk_queue_state {
353 struct blk_queue_tag {
354 struct request **tag_index; /* map of busy tags */
355 unsigned long *tag_map; /* bit map of free/busy tags */
356 int max_depth; /* what we will send to device */
357 int real_max_depth; /* what the array can hold */
358 atomic_t refcnt; /* map can be shared */
359 int alloc_policy; /* tag allocation policy */
360 int next_tag; /* next tag */
362 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
363 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
365 #define BLK_SCSI_MAX_CMDS (256)
366 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
369 * Zoned block device models (zoned limit).
371 enum blk_zoned_model {
372 BLK_ZONED_NONE, /* Regular block device */
373 BLK_ZONED_HA, /* Host-aware zoned block device */
374 BLK_ZONED_HM, /* Host-managed zoned block device */
377 struct queue_limits {
378 unsigned long bounce_pfn;
379 unsigned long seg_boundary_mask;
380 unsigned long virt_boundary_mask;
382 unsigned int max_hw_sectors;
383 unsigned int max_dev_sectors;
384 unsigned int chunk_sectors;
385 unsigned int max_sectors;
386 unsigned int max_segment_size;
387 unsigned int physical_block_size;
388 unsigned int alignment_offset;
391 unsigned int max_discard_sectors;
392 unsigned int max_hw_discard_sectors;
393 unsigned int max_write_same_sectors;
394 unsigned int max_write_zeroes_sectors;
395 unsigned int discard_granularity;
396 unsigned int discard_alignment;
398 unsigned short logical_block_size;
399 unsigned short max_segments;
400 unsigned short max_integrity_segments;
401 unsigned short max_discard_segments;
403 unsigned char misaligned;
404 unsigned char discard_misaligned;
405 unsigned char cluster;
406 unsigned char raid_partial_stripes_expensive;
407 enum blk_zoned_model zoned;
410 #ifdef CONFIG_BLK_DEV_ZONED
412 struct blk_zone_report_hdr {
413 unsigned int nr_zones;
417 extern int blkdev_report_zones(struct block_device *bdev,
418 sector_t sector, struct blk_zone *zones,
419 unsigned int *nr_zones, gfp_t gfp_mask);
420 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
421 sector_t nr_sectors, gfp_t gfp_mask);
423 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
424 unsigned int cmd, unsigned long arg);
425 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
426 unsigned int cmd, unsigned long arg);
428 #else /* CONFIG_BLK_DEV_ZONED */
430 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
431 fmode_t mode, unsigned int cmd,
437 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
438 fmode_t mode, unsigned int cmd,
444 #endif /* CONFIG_BLK_DEV_ZONED */
446 struct request_queue {
448 * Together with queue_head for cacheline sharing
450 struct list_head queue_head;
451 struct request *last_merge;
452 struct elevator_queue *elevator;
453 int nr_rqs[2]; /* # allocated [a]sync rqs */
454 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
456 atomic_t shared_hctx_restart;
458 struct blk_queue_stats *stats;
462 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
463 * is used, root blkg allocates from @q->root_rl and all other
464 * blkgs from their own blkg->rl. Which one to use should be
465 * determined using bio_request_list().
467 struct request_list root_rl;
469 request_fn_proc *request_fn;
470 make_request_fn *make_request_fn;
472 prep_rq_fn *prep_rq_fn;
473 unprep_rq_fn *unprep_rq_fn;
474 softirq_done_fn *softirq_done_fn;
475 rq_timed_out_fn *rq_timed_out_fn;
476 dma_drain_needed_fn *dma_drain_needed;
477 lld_busy_fn *lld_busy_fn;
478 /* Called just after a request is allocated */
479 init_rq_fn *init_rq_fn;
480 /* Called just before a request is freed */
481 exit_rq_fn *exit_rq_fn;
482 /* Called from inside blk_get_request() */
483 void (*initialize_rq_fn)(struct request *rq);
485 const struct blk_mq_ops *mq_ops;
487 unsigned int *mq_map;
490 struct blk_mq_ctx __percpu *queue_ctx;
491 unsigned int nr_queues;
493 unsigned int queue_depth;
495 /* hw dispatch queues */
496 struct blk_mq_hw_ctx **queue_hw_ctx;
497 unsigned int nr_hw_queues;
500 * Dispatch queue sorting
503 struct request *boundary_rq;
506 * Delayed queue handling
508 struct delayed_work delay_work;
510 struct backing_dev_info *backing_dev_info;
513 * The queue owner gets to use this for whatever they like.
514 * ll_rw_blk doesn't touch it.
519 * various queue flags, see QUEUE_* below
521 unsigned long queue_flags;
524 * ida allocated id for this queue. Used to index queues from
530 * queue needs bounce pages for pages above this limit
535 * protects queue structures from reentrancy. ->__queue_lock should
536 * _never_ be used directly, it is queue private. always use
539 spinlock_t __queue_lock;
540 spinlock_t *queue_lock;
550 struct kobject mq_kobj;
552 #ifdef CONFIG_BLK_DEV_INTEGRITY
553 struct blk_integrity integrity;
554 #endif /* CONFIG_BLK_DEV_INTEGRITY */
559 unsigned int nr_pending;
565 unsigned long nr_requests; /* Max # of requests */
566 unsigned int nr_congestion_on;
567 unsigned int nr_congestion_off;
568 unsigned int nr_batching;
570 unsigned int dma_drain_size;
571 void *dma_drain_buffer;
572 unsigned int dma_pad_mask;
573 unsigned int dma_alignment;
575 struct blk_queue_tag *queue_tags;
576 struct list_head tag_busy_list;
578 unsigned int nr_sorted;
579 unsigned int in_flight[2];
582 * Number of active block driver functions for which blk_drain_queue()
583 * must wait. Must be incremented around functions that unlock the
584 * queue_lock internally, e.g. scsi_request_fn().
586 unsigned int request_fn_active;
588 unsigned int rq_timeout;
591 struct blk_stat_callback *poll_cb;
592 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
594 struct timer_list timeout;
595 struct work_struct timeout_work;
596 struct list_head timeout_list;
598 struct list_head icq_list;
599 #ifdef CONFIG_BLK_CGROUP
600 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
601 struct blkcg_gq *root_blkg;
602 struct list_head blkg_list;
605 struct queue_limits limits;
608 * Zoned block device information for request dispatch control.
609 * nr_zones is the total number of zones of the device. This is always
610 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
611 * bits which indicates if a zone is conventional (bit clear) or
612 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
613 * bits which indicates if a zone is write locked, that is, if a write
614 * request targeting the zone was dispatched. All three fields are
615 * initialized by the low level device driver (e.g. scsi/sd.c).
616 * Stacking drivers (device mappers) may or may not initialize
619 * Reads of this information must be protected with blk_queue_enter() /
620 * blk_queue_exit(). Modifying this information is only allowed while
621 * no requests are being processed. See also blk_mq_freeze_queue() and
622 * blk_mq_unfreeze_queue().
624 unsigned int nr_zones;
625 unsigned long *seq_zones_bitmap;
626 unsigned long *seq_zones_wlock;
631 unsigned int sg_timeout;
632 unsigned int sg_reserved_size;
634 #ifdef CONFIG_BLK_DEV_IO_TRACE
635 struct blk_trace *blk_trace;
636 struct mutex blk_trace_mutex;
639 * for flush operations
641 struct blk_flush_queue *fq;
643 struct list_head requeue_list;
644 spinlock_t requeue_lock;
645 struct delayed_work requeue_work;
647 struct mutex sysfs_lock;
650 atomic_t mq_freeze_depth;
652 #if defined(CONFIG_BLK_DEV_BSG)
653 bsg_job_fn *bsg_job_fn;
654 struct bsg_class_device bsg_dev;
657 #ifdef CONFIG_BLK_DEV_THROTTLING
659 struct throtl_data *td;
661 struct rcu_head rcu_head;
662 wait_queue_head_t mq_freeze_wq;
663 struct percpu_ref q_usage_counter;
664 struct list_head all_q_node;
666 struct blk_mq_tag_set *tag_set;
667 struct list_head tag_set_list;
668 struct bio_set *bio_split;
670 #ifdef CONFIG_BLK_DEBUG_FS
671 struct dentry *debugfs_dir;
672 struct dentry *sched_debugfs_dir;
675 bool mq_sysfs_init_done;
680 struct work_struct release_work;
682 #define BLK_MAX_WRITE_HINTS 5
683 u64 write_hints[BLK_MAX_WRITE_HINTS];
686 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
687 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
688 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
689 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
690 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
691 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
692 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
693 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
694 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
695 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
696 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
697 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
698 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
699 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
700 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
701 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
702 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
703 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
704 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
705 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
706 #define QUEUE_FLAG_WC 20 /* Write back caching */
707 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
708 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
709 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
710 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
711 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
712 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
713 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
714 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
715 #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */
717 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
718 (1 << QUEUE_FLAG_SAME_COMP) | \
719 (1 << QUEUE_FLAG_ADD_RANDOM))
721 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
722 (1 << QUEUE_FLAG_SAME_COMP) | \
723 (1 << QUEUE_FLAG_POLL))
725 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
726 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
727 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
728 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
730 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
731 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
732 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
733 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
734 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
735 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
736 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
737 #define blk_queue_noxmerges(q) \
738 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
739 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
740 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
741 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
742 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
743 #define blk_queue_secure_erase(q) \
744 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
745 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
746 #define blk_queue_scsi_passthrough(q) \
747 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
749 #define blk_noretry_request(rq) \
750 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
751 REQ_FAILFAST_DRIVER))
752 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
753 #define blk_queue_preempt_only(q) \
754 test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
755 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(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 int queue_in_flight(struct request_queue *q)
762 return q->in_flight[0] + q->in_flight[1];
765 static inline bool blk_account_rq(struct request *rq)
767 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
770 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
771 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
772 /* rq->queuelist of dequeued request must be list_empty() */
773 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
775 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
777 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
780 * Driver can handle struct request, if it either has an old style
781 * request_fn defined, or is blk-mq based.
783 static inline bool queue_is_rq_based(struct request_queue *q)
785 return q->request_fn || q->mq_ops;
788 static inline unsigned int blk_queue_cluster(struct request_queue *q)
790 return q->limits.cluster;
793 static inline enum blk_zoned_model
794 blk_queue_zoned_model(struct request_queue *q)
796 return q->limits.zoned;
799 static inline bool blk_queue_is_zoned(struct request_queue *q)
801 switch (blk_queue_zoned_model(q)) {
810 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
812 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
815 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
820 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
823 if (!blk_queue_is_zoned(q))
825 return sector >> ilog2(q->limits.chunk_sectors);
828 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
831 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
833 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
836 static inline bool rq_is_sync(struct request *rq)
838 return op_is_sync(rq->cmd_flags);
841 static inline bool blk_rl_full(struct request_list *rl, bool sync)
843 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
845 return rl->flags & flag;
848 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
850 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
855 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
857 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
862 static inline bool rq_mergeable(struct request *rq)
864 if (blk_rq_is_passthrough(rq))
867 if (req_op(rq) == REQ_OP_FLUSH)
870 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
873 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
875 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
881 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
883 if (bio_page(a) == bio_page(b) &&
884 bio_offset(a) == bio_offset(b))
890 static inline unsigned int blk_queue_depth(struct request_queue *q)
893 return q->queue_depth;
895 return q->nr_requests;
899 * q->prep_rq_fn return values
902 BLKPREP_OK, /* serve it */
903 BLKPREP_KILL, /* fatal error, kill, return -EIO */
904 BLKPREP_DEFER, /* leave on queue */
905 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
908 extern unsigned long blk_max_low_pfn, blk_max_pfn;
911 * standard bounce addresses:
913 * BLK_BOUNCE_HIGH : bounce all highmem pages
914 * BLK_BOUNCE_ANY : don't bounce anything
915 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
918 #if BITS_PER_LONG == 32
919 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
921 #define BLK_BOUNCE_HIGH -1ULL
923 #define BLK_BOUNCE_ANY (-1ULL)
924 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
927 * default timeout for SG_IO if none specified
929 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
930 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
936 unsigned long offset;
941 struct req_iterator {
942 struct bvec_iter iter;
946 /* This should not be used directly - use rq_for_each_segment */
947 #define for_each_bio(_bio) \
948 for (; _bio; _bio = _bio->bi_next)
949 #define __rq_for_each_bio(_bio, rq) \
951 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
953 #define rq_for_each_segment(bvl, _rq, _iter) \
954 __rq_for_each_bio(_iter.bio, _rq) \
955 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
957 #define rq_iter_last(bvec, _iter) \
958 (_iter.bio->bi_next == NULL && \
959 bio_iter_last(bvec, _iter.iter))
961 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
962 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
964 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
965 extern void rq_flush_dcache_pages(struct request *rq);
967 static inline void rq_flush_dcache_pages(struct request *rq)
972 extern int blk_register_queue(struct gendisk *disk);
973 extern void blk_unregister_queue(struct gendisk *disk);
974 extern blk_qc_t generic_make_request(struct bio *bio);
975 extern blk_qc_t direct_make_request(struct bio *bio);
976 extern void blk_rq_init(struct request_queue *q, struct request *rq);
977 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
978 extern void blk_put_request(struct request *);
979 extern void __blk_put_request(struct request_queue *, struct request *);
980 extern struct request *blk_get_request_flags(struct request_queue *,
982 blk_mq_req_flags_t flags);
983 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
985 extern void blk_requeue_request(struct request_queue *, struct request *);
986 extern int blk_lld_busy(struct request_queue *q);
987 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
988 struct bio_set *bs, gfp_t gfp_mask,
989 int (*bio_ctr)(struct bio *, struct bio *, void *),
991 extern void blk_rq_unprep_clone(struct request *rq);
992 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
994 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
995 extern void blk_delay_queue(struct request_queue *, unsigned long);
996 extern void blk_queue_split(struct request_queue *, struct bio **);
997 extern void blk_recount_segments(struct request_queue *, struct bio *);
998 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
999 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
1000 unsigned int, void __user *);
1001 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
1002 unsigned int, void __user *);
1003 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
1004 struct scsi_ioctl_command __user *);
1006 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
1007 extern void blk_queue_exit(struct request_queue *q);
1008 extern void blk_start_queue(struct request_queue *q);
1009 extern void blk_start_queue_async(struct request_queue *q);
1010 extern void blk_stop_queue(struct request_queue *q);
1011 extern void blk_sync_queue(struct request_queue *q);
1012 extern void __blk_stop_queue(struct request_queue *q);
1013 extern void __blk_run_queue(struct request_queue *q);
1014 extern void __blk_run_queue_uncond(struct request_queue *q);
1015 extern void blk_run_queue(struct request_queue *);
1016 extern void blk_run_queue_async(struct request_queue *q);
1017 extern int blk_rq_map_user(struct request_queue *, struct request *,
1018 struct rq_map_data *, void __user *, unsigned long,
1020 extern int blk_rq_unmap_user(struct bio *);
1021 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1022 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1023 struct rq_map_data *, const struct iov_iter *,
1025 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1026 struct request *, int);
1027 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1028 struct request *, int, rq_end_io_fn *);
1030 int blk_status_to_errno(blk_status_t status);
1031 blk_status_t errno_to_blk_status(int errno);
1033 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1035 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1037 return bdev->bd_disk->queue; /* this is never NULL */
1041 * The basic unit of block I/O is a sector. It is used in a number of contexts
1042 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1043 * bytes. Variables of type sector_t represent an offset or size that is a
1044 * multiple of 512 bytes. Hence these two constants.
1046 #ifndef SECTOR_SHIFT
1047 #define SECTOR_SHIFT 9
1050 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1054 * blk_rq_pos() : the current sector
1055 * blk_rq_bytes() : bytes left in the entire request
1056 * blk_rq_cur_bytes() : bytes left in the current segment
1057 * blk_rq_err_bytes() : bytes left till the next error boundary
1058 * blk_rq_sectors() : sectors left in the entire request
1059 * blk_rq_cur_sectors() : sectors left in the current segment
1061 static inline sector_t blk_rq_pos(const struct request *rq)
1063 return rq->__sector;
1066 static inline unsigned int blk_rq_bytes(const struct request *rq)
1068 return rq->__data_len;
1071 static inline int blk_rq_cur_bytes(const struct request *rq)
1073 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1076 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1078 static inline unsigned int blk_rq_sectors(const struct request *rq)
1080 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1083 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1085 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1088 static inline unsigned int blk_rq_zone_no(struct request *rq)
1090 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1093 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1095 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1099 * Some commands like WRITE SAME have a payload or data transfer size which
1100 * is different from the size of the request. Any driver that supports such
1101 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1102 * calculate the data transfer size.
1104 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1106 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1107 return rq->special_vec.bv_len;
1108 return blk_rq_bytes(rq);
1111 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1114 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1115 return min(q->limits.max_discard_sectors,
1116 UINT_MAX >> SECTOR_SHIFT);
1118 if (unlikely(op == REQ_OP_WRITE_SAME))
1119 return q->limits.max_write_same_sectors;
1121 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1122 return q->limits.max_write_zeroes_sectors;
1124 return q->limits.max_sectors;
1128 * Return maximum size of a request at given offset. Only valid for
1129 * file system requests.
1131 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1134 if (!q->limits.chunk_sectors)
1135 return q->limits.max_sectors;
1137 return q->limits.chunk_sectors -
1138 (offset & (q->limits.chunk_sectors - 1));
1141 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1144 struct request_queue *q = rq->q;
1146 if (blk_rq_is_passthrough(rq))
1147 return q->limits.max_hw_sectors;
1149 if (!q->limits.chunk_sectors ||
1150 req_op(rq) == REQ_OP_DISCARD ||
1151 req_op(rq) == REQ_OP_SECURE_ERASE)
1152 return blk_queue_get_max_sectors(q, req_op(rq));
1154 return min(blk_max_size_offset(q, offset),
1155 blk_queue_get_max_sectors(q, req_op(rq)));
1158 static inline unsigned int blk_rq_count_bios(struct request *rq)
1160 unsigned int nr_bios = 0;
1163 __rq_for_each_bio(bio, rq)
1170 * Request issue related functions.
1172 extern struct request *blk_peek_request(struct request_queue *q);
1173 extern void blk_start_request(struct request *rq);
1174 extern struct request *blk_fetch_request(struct request_queue *q);
1176 void blk_steal_bios(struct bio_list *list, struct request *rq);
1179 * Request completion related functions.
1181 * blk_update_request() completes given number of bytes and updates
1182 * the request without completing it.
1184 * blk_end_request() and friends. __blk_end_request() must be called
1185 * with the request queue spinlock acquired.
1187 * Several drivers define their own end_request and call
1188 * blk_end_request() for parts of the original function.
1189 * This prevents code duplication in drivers.
1191 extern bool blk_update_request(struct request *rq, blk_status_t error,
1192 unsigned int nr_bytes);
1193 extern void blk_finish_request(struct request *rq, blk_status_t error);
1194 extern bool blk_end_request(struct request *rq, blk_status_t error,
1195 unsigned int nr_bytes);
1196 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1197 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1198 unsigned int nr_bytes);
1199 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1200 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1202 extern void blk_complete_request(struct request *);
1203 extern void __blk_complete_request(struct request *);
1204 extern void blk_abort_request(struct request *);
1205 extern void blk_unprep_request(struct request *);
1208 * Access functions for manipulating queue properties
1210 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1211 spinlock_t *lock, int node_id);
1212 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1213 extern int blk_init_allocated_queue(struct request_queue *);
1214 extern void blk_cleanup_queue(struct request_queue *);
1215 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1216 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1217 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1218 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1219 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1220 extern void blk_queue_max_discard_segments(struct request_queue *,
1222 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1223 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1224 unsigned int max_discard_sectors);
1225 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1226 unsigned int max_write_same_sectors);
1227 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1228 unsigned int max_write_same_sectors);
1229 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1230 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1231 extern void blk_queue_alignment_offset(struct request_queue *q,
1232 unsigned int alignment);
1233 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1234 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1235 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1236 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1237 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1238 extern void blk_set_default_limits(struct queue_limits *lim);
1239 extern void blk_set_stacking_limits(struct queue_limits *lim);
1240 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1242 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1244 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1246 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1247 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1248 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1249 extern int blk_queue_dma_drain(struct request_queue *q,
1250 dma_drain_needed_fn *dma_drain_needed,
1251 void *buf, unsigned int size);
1252 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1253 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1254 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1255 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1256 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1257 extern void blk_queue_dma_alignment(struct request_queue *, int);
1258 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1259 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1260 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1261 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1262 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1263 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1266 * Number of physical segments as sent to the device.
1268 * Normally this is the number of discontiguous data segments sent by the
1269 * submitter. But for data-less command like discard we might have no
1270 * actual data segments submitted, but the driver might have to add it's
1271 * own special payload. In that case we still return 1 here so that this
1272 * special payload will be mapped.
1274 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1276 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1278 return rq->nr_phys_segments;
1282 * Number of discard segments (or ranges) the driver needs to fill in.
1283 * Each discard bio merged into a request is counted as one segment.
1285 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1287 return max_t(unsigned short, rq->nr_phys_segments, 1);
1290 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1291 extern void blk_dump_rq_flags(struct request *, char *);
1292 extern long nr_blockdev_pages(void);
1294 bool __must_check blk_get_queue(struct request_queue *);
1295 struct request_queue *blk_alloc_queue(gfp_t);
1296 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1298 extern void blk_put_queue(struct request_queue *);
1299 extern void blk_set_queue_dying(struct request_queue *);
1302 * block layer runtime pm functions
1305 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1306 extern int blk_pre_runtime_suspend(struct request_queue *q);
1307 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1308 extern void blk_pre_runtime_resume(struct request_queue *q);
1309 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1310 extern void blk_set_runtime_active(struct request_queue *q);
1312 static inline void blk_pm_runtime_init(struct request_queue *q,
1313 struct device *dev) {}
1314 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1318 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1319 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1320 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1321 static inline void blk_set_runtime_active(struct request_queue *q) {}
1325 * blk_plug permits building a queue of related requests by holding the I/O
1326 * fragments for a short period. This allows merging of sequential requests
1327 * into single larger request. As the requests are moved from a per-task list to
1328 * the device's request_queue in a batch, this results in improved scalability
1329 * as the lock contention for request_queue lock is reduced.
1331 * It is ok not to disable preemption when adding the request to the plug list
1332 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1333 * the plug list when the task sleeps by itself. For details, please see
1334 * schedule() where blk_schedule_flush_plug() is called.
1337 struct list_head list; /* requests */
1338 struct list_head mq_list; /* blk-mq requests */
1339 struct list_head cb_list; /* md requires an unplug callback */
1341 #define BLK_MAX_REQUEST_COUNT 16
1342 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1345 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1346 struct blk_plug_cb {
1347 struct list_head list;
1348 blk_plug_cb_fn callback;
1351 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1352 void *data, int size);
1353 extern void blk_start_plug(struct blk_plug *);
1354 extern void blk_finish_plug(struct blk_plug *);
1355 extern void blk_flush_plug_list(struct blk_plug *, bool);
1357 static inline void blk_flush_plug(struct task_struct *tsk)
1359 struct blk_plug *plug = tsk->plug;
1362 blk_flush_plug_list(plug, false);
1365 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1367 struct blk_plug *plug = tsk->plug;
1370 blk_flush_plug_list(plug, true);
1373 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1375 struct blk_plug *plug = tsk->plug;
1378 (!list_empty(&plug->list) ||
1379 !list_empty(&plug->mq_list) ||
1380 !list_empty(&plug->cb_list));
1386 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1387 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1388 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1389 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1390 extern void blk_queue_free_tags(struct request_queue *);
1391 extern int blk_queue_resize_tags(struct request_queue *, int);
1392 extern void blk_queue_invalidate_tags(struct request_queue *);
1393 extern struct blk_queue_tag *blk_init_tags(int, int);
1394 extern void blk_free_tags(struct blk_queue_tag *);
1396 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1399 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1401 return bqt->tag_index[tag];
1404 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1405 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1406 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1408 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1410 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1411 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1412 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1413 sector_t nr_sects, gfp_t gfp_mask, int flags,
1416 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1417 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1419 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1420 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1422 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1423 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1425 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1426 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1428 return blkdev_issue_discard(sb->s_bdev,
1429 block << (sb->s_blocksize_bits -
1431 nr_blocks << (sb->s_blocksize_bits -
1435 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1436 sector_t nr_blocks, gfp_t gfp_mask)
1438 return blkdev_issue_zeroout(sb->s_bdev,
1439 block << (sb->s_blocksize_bits -
1441 nr_blocks << (sb->s_blocksize_bits -
1446 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1448 enum blk_default_limits {
1449 BLK_MAX_SEGMENTS = 128,
1450 BLK_SAFE_MAX_SECTORS = 255,
1451 BLK_DEF_MAX_SECTORS = 2560,
1452 BLK_MAX_SEGMENT_SIZE = 65536,
1453 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1456 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1458 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1460 return q->limits.seg_boundary_mask;
1463 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1465 return q->limits.virt_boundary_mask;
1468 static inline unsigned int queue_max_sectors(struct request_queue *q)
1470 return q->limits.max_sectors;
1473 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1475 return q->limits.max_hw_sectors;
1478 static inline unsigned short queue_max_segments(struct request_queue *q)
1480 return q->limits.max_segments;
1483 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1485 return q->limits.max_discard_segments;
1488 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1490 return q->limits.max_segment_size;
1493 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1497 if (q && q->limits.logical_block_size)
1498 retval = q->limits.logical_block_size;
1503 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1505 return queue_logical_block_size(bdev_get_queue(bdev));
1508 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1510 return q->limits.physical_block_size;
1513 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1515 return queue_physical_block_size(bdev_get_queue(bdev));
1518 static inline unsigned int queue_io_min(struct request_queue *q)
1520 return q->limits.io_min;
1523 static inline int bdev_io_min(struct block_device *bdev)
1525 return queue_io_min(bdev_get_queue(bdev));
1528 static inline unsigned int queue_io_opt(struct request_queue *q)
1530 return q->limits.io_opt;
1533 static inline int bdev_io_opt(struct block_device *bdev)
1535 return queue_io_opt(bdev_get_queue(bdev));
1538 static inline int queue_alignment_offset(struct request_queue *q)
1540 if (q->limits.misaligned)
1543 return q->limits.alignment_offset;
1546 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1548 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1549 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1552 return (granularity + lim->alignment_offset - alignment) % granularity;
1555 static inline int bdev_alignment_offset(struct block_device *bdev)
1557 struct request_queue *q = bdev_get_queue(bdev);
1559 if (q->limits.misaligned)
1562 if (bdev != bdev->bd_contains)
1563 return bdev->bd_part->alignment_offset;
1565 return q->limits.alignment_offset;
1568 static inline int queue_discard_alignment(struct request_queue *q)
1570 if (q->limits.discard_misaligned)
1573 return q->limits.discard_alignment;
1576 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1578 unsigned int alignment, granularity, offset;
1580 if (!lim->max_discard_sectors)
1583 /* Why are these in bytes, not sectors? */
1584 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1585 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1589 /* Offset of the partition start in 'granularity' sectors */
1590 offset = sector_div(sector, granularity);
1592 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1593 offset = (granularity + alignment - offset) % granularity;
1595 /* Turn it back into bytes, gaah */
1596 return offset << SECTOR_SHIFT;
1599 static inline int bdev_discard_alignment(struct block_device *bdev)
1601 struct request_queue *q = bdev_get_queue(bdev);
1603 if (bdev != bdev->bd_contains)
1604 return bdev->bd_part->discard_alignment;
1606 return q->limits.discard_alignment;
1609 static inline unsigned int bdev_write_same(struct block_device *bdev)
1611 struct request_queue *q = bdev_get_queue(bdev);
1614 return q->limits.max_write_same_sectors;
1619 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1621 struct request_queue *q = bdev_get_queue(bdev);
1624 return q->limits.max_write_zeroes_sectors;
1629 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1631 struct request_queue *q = bdev_get_queue(bdev);
1634 return blk_queue_zoned_model(q);
1636 return BLK_ZONED_NONE;
1639 static inline bool bdev_is_zoned(struct block_device *bdev)
1641 struct request_queue *q = bdev_get_queue(bdev);
1644 return blk_queue_is_zoned(q);
1649 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1651 struct request_queue *q = bdev_get_queue(bdev);
1654 return blk_queue_zone_sectors(q);
1658 static inline unsigned int bdev_nr_zones(struct block_device *bdev)
1660 struct request_queue *q = bdev_get_queue(bdev);
1663 return blk_queue_nr_zones(q);
1667 static inline int queue_dma_alignment(struct request_queue *q)
1669 return q ? q->dma_alignment : 511;
1672 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1675 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1676 return !(addr & alignment) && !(len & alignment);
1679 /* assumes size > 256 */
1680 static inline unsigned int blksize_bits(unsigned int size)
1682 unsigned int bits = 8;
1686 } while (size > 256);
1690 static inline unsigned int block_size(struct block_device *bdev)
1692 return bdev->bd_block_size;
1695 static inline bool queue_flush_queueable(struct request_queue *q)
1697 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1700 typedef struct {struct page *v;} Sector;
1702 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1704 static inline void put_dev_sector(Sector p)
1709 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1710 struct bio_vec *bprv, unsigned int offset)
1713 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1717 * Check if adding a bio_vec after bprv with offset would create a gap in
1718 * the SG list. Most drivers don't care about this, but some do.
1720 static inline bool bvec_gap_to_prev(struct request_queue *q,
1721 struct bio_vec *bprv, unsigned int offset)
1723 if (!queue_virt_boundary(q))
1725 return __bvec_gap_to_prev(q, bprv, offset);
1729 * Check if the two bvecs from two bios can be merged to one segment.
1730 * If yes, no need to check gap between the two bios since the 1st bio
1731 * and the 1st bvec in the 2nd bio can be handled in one segment.
1733 static inline bool bios_segs_mergeable(struct request_queue *q,
1734 struct bio *prev, struct bio_vec *prev_last_bv,
1735 struct bio_vec *next_first_bv)
1737 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1739 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1741 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1742 queue_max_segment_size(q))
1747 static inline bool bio_will_gap(struct request_queue *q,
1748 struct request *prev_rq,
1752 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1753 struct bio_vec pb, nb;
1756 * don't merge if the 1st bio starts with non-zero
1757 * offset, otherwise it is quite difficult to respect
1758 * sg gap limit. We work hard to merge a huge number of small
1759 * single bios in case of mkfs.
1762 bio_get_first_bvec(prev_rq->bio, &pb);
1764 bio_get_first_bvec(prev, &pb);
1769 * We don't need to worry about the situation that the
1770 * merged segment ends in unaligned virt boundary:
1772 * - if 'pb' ends aligned, the merged segment ends aligned
1773 * - if 'pb' ends unaligned, the next bio must include
1774 * one single bvec of 'nb', otherwise the 'nb' can't
1777 bio_get_last_bvec(prev, &pb);
1778 bio_get_first_bvec(next, &nb);
1780 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1781 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1787 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1789 return bio_will_gap(req->q, req, req->biotail, bio);
1792 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1794 return bio_will_gap(req->q, NULL, bio, req->bio);
1797 int kblockd_schedule_work(struct work_struct *work);
1798 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1799 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1801 #ifdef CONFIG_BLK_CGROUP
1803 * This should not be using sched_clock(). A real patch is in progress
1804 * to fix this up, until that is in place we need to disable preemption
1805 * around sched_clock() in this function and set_io_start_time_ns().
1807 static inline void set_start_time_ns(struct request *req)
1810 req->cgroup_start_time_ns = sched_clock();
1814 static inline void set_io_start_time_ns(struct request *req)
1817 req->cgroup_io_start_time_ns = sched_clock();
1821 static inline uint64_t rq_start_time_ns(struct request *req)
1823 return req->cgroup_start_time_ns;
1826 static inline uint64_t rq_io_start_time_ns(struct request *req)
1828 return req->cgroup_io_start_time_ns;
1831 static inline void set_start_time_ns(struct request *req) {}
1832 static inline void set_io_start_time_ns(struct request *req) {}
1833 static inline uint64_t rq_start_time_ns(struct request *req)
1837 static inline uint64_t rq_io_start_time_ns(struct request *req)
1843 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1844 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1845 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1846 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1848 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1850 enum blk_integrity_flags {
1851 BLK_INTEGRITY_VERIFY = 1 << 0,
1852 BLK_INTEGRITY_GENERATE = 1 << 1,
1853 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1854 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1857 struct blk_integrity_iter {
1861 unsigned int data_size;
1862 unsigned short interval;
1863 const char *disk_name;
1866 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1868 struct blk_integrity_profile {
1869 integrity_processing_fn *generate_fn;
1870 integrity_processing_fn *verify_fn;
1874 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1875 extern void blk_integrity_unregister(struct gendisk *);
1876 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1877 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1878 struct scatterlist *);
1879 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1880 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1882 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1885 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1887 struct blk_integrity *bi = &disk->queue->integrity;
1896 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1898 return blk_get_integrity(bdev->bd_disk);
1901 static inline bool blk_integrity_rq(struct request *rq)
1903 return rq->cmd_flags & REQ_INTEGRITY;
1906 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1909 q->limits.max_integrity_segments = segs;
1912 static inline unsigned short
1913 queue_max_integrity_segments(struct request_queue *q)
1915 return q->limits.max_integrity_segments;
1918 static inline bool integrity_req_gap_back_merge(struct request *req,
1921 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1922 struct bio_integrity_payload *bip_next = bio_integrity(next);
1924 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1925 bip_next->bip_vec[0].bv_offset);
1928 static inline bool integrity_req_gap_front_merge(struct request *req,
1931 struct bio_integrity_payload *bip = bio_integrity(bio);
1932 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1934 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1935 bip_next->bip_vec[0].bv_offset);
1938 #else /* CONFIG_BLK_DEV_INTEGRITY */
1941 struct block_device;
1943 struct blk_integrity;
1945 static inline int blk_integrity_rq(struct request *rq)
1949 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1954 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1956 struct scatterlist *s)
1960 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1964 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1968 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1972 static inline void blk_integrity_register(struct gendisk *d,
1973 struct blk_integrity *b)
1976 static inline void blk_integrity_unregister(struct gendisk *d)
1979 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1983 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1987 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1993 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
2000 static inline bool integrity_req_gap_back_merge(struct request *req,
2005 static inline bool integrity_req_gap_front_merge(struct request *req,
2011 #endif /* CONFIG_BLK_DEV_INTEGRITY */
2013 struct block_device_operations {
2014 int (*open) (struct block_device *, fmode_t);
2015 void (*release) (struct gendisk *, fmode_t);
2016 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
2017 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
2018 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
2019 unsigned int (*check_events) (struct gendisk *disk,
2020 unsigned int clearing);
2021 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
2022 int (*media_changed) (struct gendisk *);
2023 void (*unlock_native_capacity) (struct gendisk *);
2024 int (*revalidate_disk) (struct gendisk *);
2025 int (*getgeo)(struct block_device *, struct hd_geometry *);
2026 /* this callback is with swap_lock and sometimes page table lock held */
2027 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
2028 struct module *owner;
2029 const struct pr_ops *pr_ops;
2032 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
2034 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
2035 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
2036 struct writeback_control *);
2038 #ifdef CONFIG_BLK_DEV_ZONED
2039 bool blk_req_needs_zone_write_lock(struct request *rq);
2040 void __blk_req_zone_write_lock(struct request *rq);
2041 void __blk_req_zone_write_unlock(struct request *rq);
2043 static inline void blk_req_zone_write_lock(struct request *rq)
2045 if (blk_req_needs_zone_write_lock(rq))
2046 __blk_req_zone_write_lock(rq);
2049 static inline void blk_req_zone_write_unlock(struct request *rq)
2051 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
2052 __blk_req_zone_write_unlock(rq);
2055 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2057 return rq->q->seq_zones_wlock &&
2058 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
2061 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2063 if (!blk_req_needs_zone_write_lock(rq))
2065 return !blk_req_zone_is_write_locked(rq);
2068 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2073 static inline void blk_req_zone_write_lock(struct request *rq)
2077 static inline void blk_req_zone_write_unlock(struct request *rq)
2080 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2085 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2089 #endif /* CONFIG_BLK_DEV_ZONED */
2091 #else /* CONFIG_BLOCK */
2093 struct block_device;
2096 * stubs for when the block layer is configured out
2098 #define buffer_heads_over_limit 0
2100 static inline long nr_blockdev_pages(void)
2108 static inline void blk_start_plug(struct blk_plug *plug)
2112 static inline void blk_finish_plug(struct blk_plug *plug)
2116 static inline void blk_flush_plug(struct task_struct *task)
2120 static inline void blk_schedule_flush_plug(struct task_struct *task)
2125 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2130 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2131 sector_t *error_sector)
2136 #endif /* CONFIG_BLOCK */