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 consistent 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 5
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))
124 /* The per-zone write lock is held for this request */
125 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
126 /* already slept for hybrid poll */
127 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20))
128 /* ->timeout has been called, don't expire again */
129 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21))
131 /* flags that prevent us from merging requests: */
132 #define RQF_NOMERGE_FLAGS \
133 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
136 * Request state for blk-mq.
145 * Try to put the fields that are referenced together in the same cacheline.
147 * If you modify this structure, make sure to update blk_rq_init() and
148 * especially blk_mq_rq_ctx_init() to take care of the added fields.
151 struct request_queue *q;
152 struct blk_mq_ctx *mq_ctx;
155 unsigned int cmd_flags; /* op and common flags */
156 req_flags_t rq_flags;
160 /* the following two fields are internal, NEVER access directly */
161 unsigned int __data_len; /* total data len */
163 sector_t __sector; /* sector cursor */
168 struct list_head queuelist;
171 * The hash is used inside the scheduler, and killed once the
172 * request reaches the dispatch list. The ipi_list is only used
173 * to queue the request for softirq completion, which is long
174 * after the request has been unhashed (and even removed from
175 * the dispatch list).
178 struct hlist_node hash; /* merge hash */
179 struct list_head ipi_list;
183 * The rb_node is only used inside the io scheduler, requests
184 * are pruned when moved to the dispatch queue. So let the
185 * completion_data share space with the rb_node.
188 struct rb_node rb_node; /* sort/lookup */
189 struct bio_vec special_vec;
190 void *completion_data;
191 int error_count; /* for legacy drivers, don't use */
195 * Three pointers are available for the IO schedulers, if they need
196 * more they have to dynamically allocate it. Flush requests are
197 * never put on the IO scheduler. So let the flush fields share
198 * space with the elevator data.
208 struct list_head list;
209 rq_end_io_fn *saved_end_io;
213 struct gendisk *rq_disk;
214 struct hd_struct *part;
215 /* Time that I/O was submitted to the kernel. */
217 /* Time that I/O was submitted to the device. */
218 u64 io_start_time_ns;
220 #ifdef CONFIG_BLK_WBT
221 unsigned short wbt_flags;
223 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
224 unsigned short throtl_size;
228 * Number of scatter-gather DMA addr+len pairs after
229 * physical address coalescing is performed.
231 unsigned short nr_phys_segments;
233 #if defined(CONFIG_BLK_DEV_INTEGRITY)
234 unsigned short nr_integrity_segments;
237 unsigned short write_hint;
238 unsigned short ioprio;
240 void *special; /* opaque pointer available for LLD use */
242 unsigned int extra_len; /* length of alignment and padding */
244 enum mq_rq_state state;
247 unsigned int timeout;
249 /* access through blk_rq_set_deadline, blk_rq_deadline */
250 unsigned long __deadline;
252 struct list_head timeout_list;
255 struct __call_single_data csd;
260 * completion callback.
262 rq_end_io_fn *end_io;
266 struct request *next_rq;
268 #ifdef CONFIG_BLK_CGROUP
269 struct request_list *rl; /* rl this rq is alloced from */
273 static inline bool blk_op_is_scsi(unsigned int op)
275 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
278 static inline bool blk_op_is_private(unsigned int op)
280 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
283 static inline bool blk_rq_is_scsi(struct request *rq)
285 return blk_op_is_scsi(req_op(rq));
288 static inline bool blk_rq_is_private(struct request *rq)
290 return blk_op_is_private(req_op(rq));
293 static inline bool blk_rq_is_passthrough(struct request *rq)
295 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
298 static inline bool bio_is_passthrough(struct bio *bio)
300 unsigned op = bio_op(bio);
302 return blk_op_is_scsi(op) || blk_op_is_private(op);
305 static inline unsigned short req_get_ioprio(struct request *req)
310 #include <linux/elevator.h>
312 struct blk_queue_ctx;
314 typedef void (request_fn_proc) (struct request_queue *q);
315 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
316 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
317 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
318 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
321 typedef void (softirq_done_fn)(struct request *);
322 typedef int (dma_drain_needed_fn)(struct request *);
323 typedef int (lld_busy_fn) (struct request_queue *q);
324 typedef int (bsg_job_fn) (struct bsg_job *);
325 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
326 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
328 enum blk_eh_timer_return {
329 BLK_EH_DONE, /* drivers has completed the command */
330 BLK_EH_RESET_TIMER, /* reset timer and try again */
333 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
335 enum blk_queue_state {
340 struct blk_queue_tag {
341 struct request **tag_index; /* map of busy tags */
342 unsigned long *tag_map; /* bit map of free/busy tags */
343 int max_depth; /* what we will send to device */
344 int real_max_depth; /* what the array can hold */
345 atomic_t refcnt; /* map can be shared */
346 int alloc_policy; /* tag allocation policy */
347 int next_tag; /* next tag */
349 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
350 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
352 #define BLK_SCSI_MAX_CMDS (256)
353 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
356 * Zoned block device models (zoned limit).
358 enum blk_zoned_model {
359 BLK_ZONED_NONE, /* Regular block device */
360 BLK_ZONED_HA, /* Host-aware zoned block device */
361 BLK_ZONED_HM, /* Host-managed zoned block device */
364 struct queue_limits {
365 unsigned long bounce_pfn;
366 unsigned long seg_boundary_mask;
367 unsigned long virt_boundary_mask;
369 unsigned int max_hw_sectors;
370 unsigned int max_dev_sectors;
371 unsigned int chunk_sectors;
372 unsigned int max_sectors;
373 unsigned int max_segment_size;
374 unsigned int physical_block_size;
375 unsigned int logical_block_size;
376 unsigned int alignment_offset;
379 unsigned int max_discard_sectors;
380 unsigned int max_hw_discard_sectors;
381 unsigned int max_write_same_sectors;
382 unsigned int max_write_zeroes_sectors;
383 unsigned int discard_granularity;
384 unsigned int discard_alignment;
386 unsigned short max_segments;
387 unsigned short max_integrity_segments;
388 unsigned short max_discard_segments;
390 unsigned char misaligned;
391 unsigned char discard_misaligned;
392 unsigned char cluster;
393 unsigned char raid_partial_stripes_expensive;
394 enum blk_zoned_model zoned;
397 #ifdef CONFIG_BLK_DEV_ZONED
399 struct blk_zone_report_hdr {
400 unsigned int nr_zones;
404 extern int blkdev_report_zones(struct block_device *bdev,
405 sector_t sector, struct blk_zone *zones,
406 unsigned int *nr_zones, gfp_t gfp_mask);
407 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
408 sector_t nr_sectors, gfp_t gfp_mask);
410 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
411 unsigned int cmd, unsigned long arg);
412 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
413 unsigned int cmd, unsigned long arg);
415 #else /* CONFIG_BLK_DEV_ZONED */
417 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
418 fmode_t mode, unsigned int cmd,
424 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
425 fmode_t mode, unsigned int cmd,
431 #endif /* CONFIG_BLK_DEV_ZONED */
433 struct request_queue {
435 * Together with queue_head for cacheline sharing
437 struct list_head queue_head;
438 struct request *last_merge;
439 struct elevator_queue *elevator;
440 int nr_rqs[2]; /* # allocated [a]sync rqs */
441 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
443 struct blk_queue_stats *stats;
444 struct rq_qos *rq_qos;
447 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
448 * is used, root blkg allocates from @q->root_rl and all other
449 * blkgs from their own blkg->rl. Which one to use should be
450 * determined using bio_request_list().
452 struct request_list root_rl;
454 request_fn_proc *request_fn;
455 make_request_fn *make_request_fn;
457 prep_rq_fn *prep_rq_fn;
458 unprep_rq_fn *unprep_rq_fn;
459 softirq_done_fn *softirq_done_fn;
460 rq_timed_out_fn *rq_timed_out_fn;
461 dma_drain_needed_fn *dma_drain_needed;
462 lld_busy_fn *lld_busy_fn;
463 /* Called just after a request is allocated */
464 init_rq_fn *init_rq_fn;
465 /* Called just before a request is freed */
466 exit_rq_fn *exit_rq_fn;
467 /* Called from inside blk_get_request() */
468 void (*initialize_rq_fn)(struct request *rq);
470 const struct blk_mq_ops *mq_ops;
472 unsigned int *mq_map;
475 struct blk_mq_ctx __percpu *queue_ctx;
476 unsigned int nr_queues;
478 unsigned int queue_depth;
480 /* hw dispatch queues */
481 struct blk_mq_hw_ctx **queue_hw_ctx;
482 unsigned int nr_hw_queues;
485 * Dispatch queue sorting
488 struct request *boundary_rq;
491 * Delayed queue handling
493 struct delayed_work delay_work;
495 struct backing_dev_info *backing_dev_info;
498 * The queue owner gets to use this for whatever they like.
499 * ll_rw_blk doesn't touch it.
504 * various queue flags, see QUEUE_* below
506 unsigned long queue_flags;
508 * Number of contexts that have called blk_set_pm_only(). If this
509 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
515 * ida allocated id for this queue. Used to index queues from
521 * queue needs bounce pages for pages above this limit
526 * protects queue structures from reentrancy. ->__queue_lock should
527 * _never_ be used directly, it is queue private. always use
530 spinlock_t __queue_lock;
531 spinlock_t *queue_lock;
541 struct kobject mq_kobj;
543 #ifdef CONFIG_BLK_DEV_INTEGRITY
544 struct blk_integrity integrity;
545 #endif /* CONFIG_BLK_DEV_INTEGRITY */
550 unsigned int nr_pending;
556 unsigned long nr_requests; /* Max # of requests */
557 unsigned int nr_congestion_on;
558 unsigned int nr_congestion_off;
559 unsigned int nr_batching;
561 unsigned int dma_drain_size;
562 void *dma_drain_buffer;
563 unsigned int dma_pad_mask;
564 unsigned int dma_alignment;
566 struct blk_queue_tag *queue_tags;
568 unsigned int nr_sorted;
569 unsigned int in_flight[2];
572 * Number of active block driver functions for which blk_drain_queue()
573 * must wait. Must be incremented around functions that unlock the
574 * queue_lock internally, e.g. scsi_request_fn().
576 unsigned int request_fn_active;
578 unsigned int rq_timeout;
581 struct blk_stat_callback *poll_cb;
582 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
584 struct timer_list timeout;
585 struct work_struct timeout_work;
586 struct list_head timeout_list;
588 struct list_head icq_list;
589 #ifdef CONFIG_BLK_CGROUP
590 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
591 struct blkcg_gq *root_blkg;
592 struct list_head blkg_list;
595 struct queue_limits limits;
597 #ifdef CONFIG_BLK_DEV_ZONED
599 * Zoned block device information for request dispatch control.
600 * nr_zones is the total number of zones of the device. This is always
601 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
602 * bits which indicates if a zone is conventional (bit clear) or
603 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
604 * bits which indicates if a zone is write locked, that is, if a write
605 * request targeting the zone was dispatched. All three fields are
606 * initialized by the low level device driver (e.g. scsi/sd.c).
607 * Stacking drivers (device mappers) may or may not initialize
610 * Reads of this information must be protected with blk_queue_enter() /
611 * blk_queue_exit(). Modifying this information is only allowed while
612 * no requests are being processed. See also blk_mq_freeze_queue() and
613 * blk_mq_unfreeze_queue().
615 unsigned int nr_zones;
616 unsigned long *seq_zones_bitmap;
617 unsigned long *seq_zones_wlock;
618 #endif /* CONFIG_BLK_DEV_ZONED */
623 unsigned int sg_timeout;
624 unsigned int sg_reserved_size;
626 #ifdef CONFIG_BLK_DEV_IO_TRACE
627 struct blk_trace __rcu *blk_trace;
628 struct mutex blk_trace_mutex;
631 * for flush operations
633 struct blk_flush_queue *fq;
635 struct list_head requeue_list;
636 spinlock_t requeue_lock;
637 struct delayed_work requeue_work;
639 struct mutex sysfs_lock;
640 struct mutex sysfs_dir_lock;
643 atomic_t mq_freeze_depth;
645 #if defined(CONFIG_BLK_DEV_BSG)
646 bsg_job_fn *bsg_job_fn;
647 struct bsg_class_device bsg_dev;
650 #ifdef CONFIG_BLK_DEV_THROTTLING
652 struct throtl_data *td;
654 struct rcu_head rcu_head;
655 wait_queue_head_t mq_freeze_wq;
656 struct percpu_ref q_usage_counter;
657 struct list_head all_q_node;
659 struct blk_mq_tag_set *tag_set;
660 struct list_head tag_set_list;
661 struct bio_set bio_split;
663 #ifdef CONFIG_BLK_DEBUG_FS
664 struct dentry *debugfs_dir;
665 struct dentry *sched_debugfs_dir;
668 bool mq_sysfs_init_done;
673 struct work_struct release_work;
675 #define BLK_MAX_WRITE_HINTS 5
676 u64 write_hints[BLK_MAX_WRITE_HINTS];
679 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
680 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
681 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
682 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
683 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
684 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
685 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
686 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
687 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
688 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
689 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
690 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
691 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
692 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
693 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
694 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
695 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
696 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
697 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
698 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
699 #define QUEUE_FLAG_WC 20 /* Write back caching */
700 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
701 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
702 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
703 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
704 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
705 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
706 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
707 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
709 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
710 (1 << QUEUE_FLAG_SAME_COMP) | \
711 (1 << QUEUE_FLAG_ADD_RANDOM))
713 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
714 (1 << QUEUE_FLAG_SAME_COMP) | \
715 (1 << QUEUE_FLAG_POLL))
717 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
718 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
719 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
720 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
722 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
723 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
724 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
725 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
726 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
727 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
728 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
729 #define blk_queue_noxmerges(q) \
730 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
731 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
732 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
733 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
734 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
735 #define blk_queue_secure_erase(q) \
736 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
737 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
738 #define blk_queue_scsi_passthrough(q) \
739 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
741 #define blk_noretry_request(rq) \
742 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
743 REQ_FAILFAST_DRIVER))
744 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
745 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
746 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
747 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
749 extern void blk_set_pm_only(struct request_queue *q);
750 extern void blk_clear_pm_only(struct request_queue *q);
752 static inline int queue_in_flight(struct request_queue *q)
754 return q->in_flight[0] + q->in_flight[1];
757 static inline bool blk_account_rq(struct request *rq)
759 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
762 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
763 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
764 /* rq->queuelist of dequeued request must be list_empty() */
765 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
767 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
769 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
772 * Driver can handle struct request, if it either has an old style
773 * request_fn defined, or is blk-mq based.
775 static inline bool queue_is_rq_based(struct request_queue *q)
777 return q->request_fn || q->mq_ops;
780 static inline unsigned int blk_queue_cluster(struct request_queue *q)
782 return q->limits.cluster;
785 static inline enum blk_zoned_model
786 blk_queue_zoned_model(struct request_queue *q)
788 return q->limits.zoned;
791 static inline bool blk_queue_is_zoned(struct request_queue *q)
793 switch (blk_queue_zoned_model(q)) {
802 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
804 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
807 #ifdef CONFIG_BLK_DEV_ZONED
808 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
811 if (!blk_queue_is_zoned(q))
813 return sector >> ilog2(q->limits.chunk_sectors);
816 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
819 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
821 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
823 #endif /* CONFIG_BLK_DEV_ZONED */
825 static inline bool rq_is_sync(struct request *rq)
827 return op_is_sync(rq->cmd_flags);
830 static inline bool blk_rl_full(struct request_list *rl, bool sync)
832 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
834 return rl->flags & flag;
837 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
839 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
844 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
846 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
851 static inline bool rq_mergeable(struct request *rq)
853 if (blk_rq_is_passthrough(rq))
856 if (req_op(rq) == REQ_OP_FLUSH)
859 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
862 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
864 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
870 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
872 if (bio_page(a) == bio_page(b) &&
873 bio_offset(a) == bio_offset(b))
879 static inline unsigned int blk_queue_depth(struct request_queue *q)
882 return q->queue_depth;
884 return q->nr_requests;
888 * q->prep_rq_fn return values
891 BLKPREP_OK, /* serve it */
892 BLKPREP_KILL, /* fatal error, kill, return -EIO */
893 BLKPREP_DEFER, /* leave on queue */
894 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
897 extern unsigned long blk_max_low_pfn, blk_max_pfn;
900 * standard bounce addresses:
902 * BLK_BOUNCE_HIGH : bounce all highmem pages
903 * BLK_BOUNCE_ANY : don't bounce anything
904 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
907 #if BITS_PER_LONG == 32
908 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
910 #define BLK_BOUNCE_HIGH -1ULL
912 #define BLK_BOUNCE_ANY (-1ULL)
913 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
916 * default timeout for SG_IO if none specified
918 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
919 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
925 unsigned long offset;
930 struct req_iterator {
931 struct bvec_iter iter;
935 /* This should not be used directly - use rq_for_each_segment */
936 #define for_each_bio(_bio) \
937 for (; _bio; _bio = _bio->bi_next)
938 #define __rq_for_each_bio(_bio, rq) \
940 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
942 #define rq_for_each_segment(bvl, _rq, _iter) \
943 __rq_for_each_bio(_iter.bio, _rq) \
944 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
946 #define rq_iter_last(bvec, _iter) \
947 (_iter.bio->bi_next == NULL && \
948 bio_iter_last(bvec, _iter.iter))
950 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
951 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
953 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
954 extern void rq_flush_dcache_pages(struct request *rq);
956 static inline void rq_flush_dcache_pages(struct request *rq)
961 extern int blk_register_queue(struct gendisk *disk);
962 extern void blk_unregister_queue(struct gendisk *disk);
963 extern blk_qc_t generic_make_request(struct bio *bio);
964 extern blk_qc_t direct_make_request(struct bio *bio);
965 extern void blk_rq_init(struct request_queue *q, struct request *rq);
966 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
967 extern void blk_put_request(struct request *);
968 extern void __blk_put_request(struct request_queue *, struct request *);
969 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
970 blk_mq_req_flags_t flags);
971 extern void blk_requeue_request(struct request_queue *, struct request *);
972 extern int blk_lld_busy(struct request_queue *q);
973 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
974 struct bio_set *bs, gfp_t gfp_mask,
975 int (*bio_ctr)(struct bio *, struct bio *, void *),
977 extern void blk_rq_unprep_clone(struct request *rq);
978 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
980 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
981 extern void blk_delay_queue(struct request_queue *, unsigned long);
982 extern void blk_queue_split(struct request_queue *, struct bio **);
983 extern void blk_recount_segments(struct request_queue *, struct bio *);
984 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
985 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
986 unsigned int, void __user *);
987 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
988 unsigned int, void __user *);
989 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
990 struct scsi_ioctl_command __user *);
992 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
993 extern void blk_queue_exit(struct request_queue *q);
994 extern void blk_start_queue(struct request_queue *q);
995 extern void blk_start_queue_async(struct request_queue *q);
996 extern void blk_stop_queue(struct request_queue *q);
997 extern void blk_sync_queue(struct request_queue *q);
998 extern void __blk_stop_queue(struct request_queue *q);
999 extern void __blk_run_queue(struct request_queue *q);
1000 extern void __blk_run_queue_uncond(struct request_queue *q);
1001 extern void blk_run_queue(struct request_queue *);
1002 extern void blk_run_queue_async(struct request_queue *q);
1003 extern int blk_rq_map_user(struct request_queue *, struct request *,
1004 struct rq_map_data *, void __user *, unsigned long,
1006 extern int blk_rq_unmap_user(struct bio *);
1007 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1008 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1009 struct rq_map_data *, const struct iov_iter *,
1011 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1012 struct request *, int);
1013 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1014 struct request *, int, rq_end_io_fn *);
1016 int blk_status_to_errno(blk_status_t status);
1017 blk_status_t errno_to_blk_status(int errno);
1019 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1021 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1023 return bdev->bd_disk->queue; /* this is never NULL */
1027 * The basic unit of block I/O is a sector. It is used in a number of contexts
1028 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1029 * bytes. Variables of type sector_t represent an offset or size that is a
1030 * multiple of 512 bytes. Hence these two constants.
1032 #ifndef SECTOR_SHIFT
1033 #define SECTOR_SHIFT 9
1036 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1040 * blk_rq_pos() : the current sector
1041 * blk_rq_bytes() : bytes left in the entire request
1042 * blk_rq_cur_bytes() : bytes left in the current segment
1043 * blk_rq_err_bytes() : bytes left till the next error boundary
1044 * blk_rq_sectors() : sectors left in the entire request
1045 * blk_rq_cur_sectors() : sectors left in the current segment
1047 static inline sector_t blk_rq_pos(const struct request *rq)
1049 return rq->__sector;
1052 static inline unsigned int blk_rq_bytes(const struct request *rq)
1054 return rq->__data_len;
1057 static inline int blk_rq_cur_bytes(const struct request *rq)
1059 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1062 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1064 static inline unsigned int blk_rq_sectors(const struct request *rq)
1066 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1069 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1071 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1074 #ifdef CONFIG_BLK_DEV_ZONED
1075 static inline unsigned int blk_rq_zone_no(struct request *rq)
1077 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1080 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1082 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1084 #endif /* CONFIG_BLK_DEV_ZONED */
1087 * Some commands like WRITE SAME have a payload or data transfer size which
1088 * is different from the size of the request. Any driver that supports such
1089 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1090 * calculate the data transfer size.
1092 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1094 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1095 return rq->special_vec.bv_len;
1096 return blk_rq_bytes(rq);
1099 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1102 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1103 return min(q->limits.max_discard_sectors,
1104 UINT_MAX >> SECTOR_SHIFT);
1106 if (unlikely(op == REQ_OP_WRITE_SAME))
1107 return q->limits.max_write_same_sectors;
1109 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1110 return q->limits.max_write_zeroes_sectors;
1112 return q->limits.max_sectors;
1116 * Return maximum size of a request at given offset. Only valid for
1117 * file system requests.
1119 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1122 if (!q->limits.chunk_sectors)
1123 return q->limits.max_sectors;
1125 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1126 (offset & (q->limits.chunk_sectors - 1))));
1129 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1132 struct request_queue *q = rq->q;
1134 if (blk_rq_is_passthrough(rq))
1135 return q->limits.max_hw_sectors;
1137 if (!q->limits.chunk_sectors ||
1138 req_op(rq) == REQ_OP_DISCARD ||
1139 req_op(rq) == REQ_OP_SECURE_ERASE)
1140 return blk_queue_get_max_sectors(q, req_op(rq));
1142 return min(blk_max_size_offset(q, offset),
1143 blk_queue_get_max_sectors(q, req_op(rq)));
1146 static inline unsigned int blk_rq_count_bios(struct request *rq)
1148 unsigned int nr_bios = 0;
1151 __rq_for_each_bio(bio, rq)
1158 * Request issue related functions.
1160 extern struct request *blk_peek_request(struct request_queue *q);
1161 extern void blk_start_request(struct request *rq);
1162 extern struct request *blk_fetch_request(struct request_queue *q);
1164 void blk_steal_bios(struct bio_list *list, struct request *rq);
1167 * Request completion related functions.
1169 * blk_update_request() completes given number of bytes and updates
1170 * the request without completing it.
1172 * blk_end_request() and friends. __blk_end_request() must be called
1173 * with the request queue spinlock acquired.
1175 * Several drivers define their own end_request and call
1176 * blk_end_request() for parts of the original function.
1177 * This prevents code duplication in drivers.
1179 extern bool blk_update_request(struct request *rq, blk_status_t error,
1180 unsigned int nr_bytes);
1181 extern void blk_finish_request(struct request *rq, blk_status_t error);
1182 extern bool blk_end_request(struct request *rq, blk_status_t error,
1183 unsigned int nr_bytes);
1184 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1185 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1186 unsigned int nr_bytes);
1187 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1188 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1190 extern void blk_complete_request(struct request *);
1191 extern void __blk_complete_request(struct request *);
1192 extern void blk_abort_request(struct request *);
1193 extern void blk_unprep_request(struct request *);
1196 * Access functions for manipulating queue properties
1198 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1199 spinlock_t *lock, int node_id);
1200 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1201 extern int blk_init_allocated_queue(struct request_queue *);
1202 extern void blk_cleanup_queue(struct request_queue *);
1203 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1204 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1205 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1206 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1207 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1208 extern void blk_queue_max_discard_segments(struct request_queue *,
1210 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1211 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1212 unsigned int max_discard_sectors);
1213 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1214 unsigned int max_write_same_sectors);
1215 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1216 unsigned int max_write_same_sectors);
1217 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1218 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1219 extern void blk_queue_alignment_offset(struct request_queue *q,
1220 unsigned int alignment);
1221 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1222 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1223 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1224 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1225 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1226 extern void blk_set_default_limits(struct queue_limits *lim);
1227 extern void blk_set_stacking_limits(struct queue_limits *lim);
1228 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1230 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1232 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1234 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1235 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1236 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1237 extern int blk_queue_dma_drain(struct request_queue *q,
1238 dma_drain_needed_fn *dma_drain_needed,
1239 void *buf, unsigned int size);
1240 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1241 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1242 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1243 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1244 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1245 extern void blk_queue_dma_alignment(struct request_queue *, int);
1246 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1247 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1248 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1249 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1250 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1251 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1254 * Number of physical segments as sent to the device.
1256 * Normally this is the number of discontiguous data segments sent by the
1257 * submitter. But for data-less command like discard we might have no
1258 * actual data segments submitted, but the driver might have to add it's
1259 * own special payload. In that case we still return 1 here so that this
1260 * special payload will be mapped.
1262 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1264 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1266 return rq->nr_phys_segments;
1270 * Number of discard segments (or ranges) the driver needs to fill in.
1271 * Each discard bio merged into a request is counted as one segment.
1273 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1275 return max_t(unsigned short, rq->nr_phys_segments, 1);
1278 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1279 extern void blk_dump_rq_flags(struct request *, char *);
1280 extern long nr_blockdev_pages(void);
1282 bool __must_check blk_get_queue(struct request_queue *);
1283 struct request_queue *blk_alloc_queue(gfp_t);
1284 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1286 extern void blk_put_queue(struct request_queue *);
1287 extern void blk_set_queue_dying(struct request_queue *);
1290 * block layer runtime pm functions
1293 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1294 extern int blk_pre_runtime_suspend(struct request_queue *q);
1295 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1296 extern void blk_pre_runtime_resume(struct request_queue *q);
1297 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1298 extern void blk_set_runtime_active(struct request_queue *q);
1300 static inline void blk_pm_runtime_init(struct request_queue *q,
1301 struct device *dev) {}
1302 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1306 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1307 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1308 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1309 static inline void blk_set_runtime_active(struct request_queue *q) {}
1313 * blk_plug permits building a queue of related requests by holding the I/O
1314 * fragments for a short period. This allows merging of sequential requests
1315 * into single larger request. As the requests are moved from a per-task list to
1316 * the device's request_queue in a batch, this results in improved scalability
1317 * as the lock contention for request_queue lock is reduced.
1319 * It is ok not to disable preemption when adding the request to the plug list
1320 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1321 * the plug list when the task sleeps by itself. For details, please see
1322 * schedule() where blk_schedule_flush_plug() is called.
1325 struct list_head list; /* requests */
1326 struct list_head mq_list; /* blk-mq requests */
1327 struct list_head cb_list; /* md requires an unplug callback */
1329 #define BLK_MAX_REQUEST_COUNT 16
1330 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1333 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1334 struct blk_plug_cb {
1335 struct list_head list;
1336 blk_plug_cb_fn callback;
1339 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1340 void *data, int size);
1341 extern void blk_start_plug(struct blk_plug *);
1342 extern void blk_finish_plug(struct blk_plug *);
1343 extern void blk_flush_plug_list(struct blk_plug *, bool);
1345 static inline void blk_flush_plug(struct task_struct *tsk)
1347 struct blk_plug *plug = tsk->plug;
1350 blk_flush_plug_list(plug, false);
1353 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1355 struct blk_plug *plug = tsk->plug;
1358 blk_flush_plug_list(plug, true);
1361 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1363 struct blk_plug *plug = tsk->plug;
1366 (!list_empty(&plug->list) ||
1367 !list_empty(&plug->mq_list) ||
1368 !list_empty(&plug->cb_list));
1374 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1375 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1376 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1377 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1378 extern void blk_queue_free_tags(struct request_queue *);
1379 extern int blk_queue_resize_tags(struct request_queue *, int);
1380 extern struct blk_queue_tag *blk_init_tags(int, int);
1381 extern void blk_free_tags(struct blk_queue_tag *);
1383 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1386 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1388 return bqt->tag_index[tag];
1391 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1392 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1393 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1395 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1397 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1398 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1399 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1400 sector_t nr_sects, gfp_t gfp_mask, int flags,
1403 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1404 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1406 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1407 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1409 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1410 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1412 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1413 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1415 return blkdev_issue_discard(sb->s_bdev,
1416 block << (sb->s_blocksize_bits -
1418 nr_blocks << (sb->s_blocksize_bits -
1422 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1423 sector_t nr_blocks, gfp_t gfp_mask)
1425 return blkdev_issue_zeroout(sb->s_bdev,
1426 block << (sb->s_blocksize_bits -
1428 nr_blocks << (sb->s_blocksize_bits -
1433 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1435 enum blk_default_limits {
1436 BLK_MAX_SEGMENTS = 128,
1437 BLK_SAFE_MAX_SECTORS = 255,
1438 BLK_DEF_MAX_SECTORS = 2560,
1439 BLK_MAX_SEGMENT_SIZE = 65536,
1440 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1443 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1445 return q->limits.seg_boundary_mask;
1448 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1450 return q->limits.virt_boundary_mask;
1453 static inline unsigned int queue_max_sectors(struct request_queue *q)
1455 return q->limits.max_sectors;
1458 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1460 return q->limits.max_hw_sectors;
1463 static inline unsigned short queue_max_segments(struct request_queue *q)
1465 return q->limits.max_segments;
1468 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1470 return q->limits.max_discard_segments;
1473 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1475 return q->limits.max_segment_size;
1478 static inline unsigned queue_logical_block_size(struct request_queue *q)
1482 if (q && q->limits.logical_block_size)
1483 retval = q->limits.logical_block_size;
1488 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1490 return queue_logical_block_size(bdev_get_queue(bdev));
1493 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1495 return q->limits.physical_block_size;
1498 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1500 return queue_physical_block_size(bdev_get_queue(bdev));
1503 static inline unsigned int queue_io_min(struct request_queue *q)
1505 return q->limits.io_min;
1508 static inline int bdev_io_min(struct block_device *bdev)
1510 return queue_io_min(bdev_get_queue(bdev));
1513 static inline unsigned int queue_io_opt(struct request_queue *q)
1515 return q->limits.io_opt;
1518 static inline int bdev_io_opt(struct block_device *bdev)
1520 return queue_io_opt(bdev_get_queue(bdev));
1523 static inline int queue_alignment_offset(struct request_queue *q)
1525 if (q->limits.misaligned)
1528 return q->limits.alignment_offset;
1531 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1533 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1534 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1537 return (granularity + lim->alignment_offset - alignment) % granularity;
1540 static inline int bdev_alignment_offset(struct block_device *bdev)
1542 struct request_queue *q = bdev_get_queue(bdev);
1544 if (q->limits.misaligned)
1547 if (bdev != bdev->bd_contains)
1548 return bdev->bd_part->alignment_offset;
1550 return q->limits.alignment_offset;
1553 static inline int queue_discard_alignment(struct request_queue *q)
1555 if (q->limits.discard_misaligned)
1558 return q->limits.discard_alignment;
1561 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1563 unsigned int alignment, granularity, offset;
1565 if (!lim->max_discard_sectors)
1568 /* Why are these in bytes, not sectors? */
1569 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1570 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1574 /* Offset of the partition start in 'granularity' sectors */
1575 offset = sector_div(sector, granularity);
1577 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1578 offset = (granularity + alignment - offset) % granularity;
1580 /* Turn it back into bytes, gaah */
1581 return offset << SECTOR_SHIFT;
1584 static inline int bdev_discard_alignment(struct block_device *bdev)
1586 struct request_queue *q = bdev_get_queue(bdev);
1588 if (bdev != bdev->bd_contains)
1589 return bdev->bd_part->discard_alignment;
1591 return q->limits.discard_alignment;
1594 static inline unsigned int bdev_write_same(struct block_device *bdev)
1596 struct request_queue *q = bdev_get_queue(bdev);
1599 return q->limits.max_write_same_sectors;
1604 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1606 struct request_queue *q = bdev_get_queue(bdev);
1609 return q->limits.max_write_zeroes_sectors;
1614 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1616 struct request_queue *q = bdev_get_queue(bdev);
1619 return blk_queue_zoned_model(q);
1621 return BLK_ZONED_NONE;
1624 static inline bool bdev_is_zoned(struct block_device *bdev)
1626 struct request_queue *q = bdev_get_queue(bdev);
1629 return blk_queue_is_zoned(q);
1634 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1636 struct request_queue *q = bdev_get_queue(bdev);
1639 return blk_queue_zone_sectors(q);
1643 static inline int queue_dma_alignment(struct request_queue *q)
1645 return q ? q->dma_alignment : 511;
1648 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1651 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1652 return !(addr & alignment) && !(len & alignment);
1655 /* assumes size > 256 */
1656 static inline unsigned int blksize_bits(unsigned int size)
1658 unsigned int bits = 8;
1662 } while (size > 256);
1666 static inline unsigned int block_size(struct block_device *bdev)
1668 return bdev->bd_block_size;
1671 static inline bool queue_flush_queueable(struct request_queue *q)
1673 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1676 typedef struct {struct page *v;} Sector;
1678 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1680 static inline void put_dev_sector(Sector p)
1685 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1686 struct bio_vec *bprv, unsigned int offset)
1689 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1693 * Check if adding a bio_vec after bprv with offset would create a gap in
1694 * the SG list. Most drivers don't care about this, but some do.
1696 static inline bool bvec_gap_to_prev(struct request_queue *q,
1697 struct bio_vec *bprv, unsigned int offset)
1699 if (!queue_virt_boundary(q))
1701 return __bvec_gap_to_prev(q, bprv, offset);
1705 * Check if the two bvecs from two bios can be merged to one segment.
1706 * If yes, no need to check gap between the two bios since the 1st bio
1707 * and the 1st bvec in the 2nd bio can be handled in one segment.
1709 static inline bool bios_segs_mergeable(struct request_queue *q,
1710 struct bio *prev, struct bio_vec *prev_last_bv,
1711 struct bio_vec *next_first_bv)
1713 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1715 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1717 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1718 queue_max_segment_size(q))
1723 static inline bool bio_will_gap(struct request_queue *q,
1724 struct request *prev_rq,
1728 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1729 struct bio_vec pb, nb;
1732 * don't merge if the 1st bio starts with non-zero
1733 * offset, otherwise it is quite difficult to respect
1734 * sg gap limit. We work hard to merge a huge number of small
1735 * single bios in case of mkfs.
1738 bio_get_first_bvec(prev_rq->bio, &pb);
1740 bio_get_first_bvec(prev, &pb);
1745 * We don't need to worry about the situation that the
1746 * merged segment ends in unaligned virt boundary:
1748 * - if 'pb' ends aligned, the merged segment ends aligned
1749 * - if 'pb' ends unaligned, the next bio must include
1750 * one single bvec of 'nb', otherwise the 'nb' can't
1753 bio_get_last_bvec(prev, &pb);
1754 bio_get_first_bvec(next, &nb);
1756 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1757 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1763 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1765 return bio_will_gap(req->q, req, req->biotail, bio);
1768 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1770 return bio_will_gap(req->q, NULL, bio, req->bio);
1773 int kblockd_schedule_work(struct work_struct *work);
1774 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1775 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1777 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1778 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1779 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1780 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1782 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1784 enum blk_integrity_flags {
1785 BLK_INTEGRITY_VERIFY = 1 << 0,
1786 BLK_INTEGRITY_GENERATE = 1 << 1,
1787 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1788 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1791 struct blk_integrity_iter {
1795 unsigned int data_size;
1796 unsigned short interval;
1797 const char *disk_name;
1800 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1802 struct blk_integrity_profile {
1803 integrity_processing_fn *generate_fn;
1804 integrity_processing_fn *verify_fn;
1808 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1809 extern void blk_integrity_unregister(struct gendisk *);
1810 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1811 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1812 struct scatterlist *);
1813 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1814 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1816 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1819 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1821 struct blk_integrity *bi = &disk->queue->integrity;
1830 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1832 return blk_get_integrity(bdev->bd_disk);
1835 static inline bool blk_integrity_rq(struct request *rq)
1837 return rq->cmd_flags & REQ_INTEGRITY;
1840 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1843 q->limits.max_integrity_segments = segs;
1846 static inline unsigned short
1847 queue_max_integrity_segments(struct request_queue *q)
1849 return q->limits.max_integrity_segments;
1852 static inline bool integrity_req_gap_back_merge(struct request *req,
1855 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1856 struct bio_integrity_payload *bip_next = bio_integrity(next);
1858 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1859 bip_next->bip_vec[0].bv_offset);
1862 static inline bool integrity_req_gap_front_merge(struct request *req,
1865 struct bio_integrity_payload *bip = bio_integrity(bio);
1866 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1868 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1869 bip_next->bip_vec[0].bv_offset);
1873 * bio_integrity_intervals - Return number of integrity intervals for a bio
1874 * @bi: blk_integrity profile for device
1875 * @sectors: Size of the bio in 512-byte sectors
1877 * Description: The block layer calculates everything in 512 byte
1878 * sectors but integrity metadata is done in terms of the data integrity
1879 * interval size of the storage device. Convert the block layer sectors
1880 * to the appropriate number of integrity intervals.
1882 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1883 unsigned int sectors)
1885 return sectors >> (bi->interval_exp - 9);
1888 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1889 unsigned int sectors)
1891 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1894 #else /* CONFIG_BLK_DEV_INTEGRITY */
1897 struct block_device;
1899 struct blk_integrity;
1901 static inline int blk_integrity_rq(struct request *rq)
1905 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1910 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1912 struct scatterlist *s)
1916 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1920 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1924 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1928 static inline void blk_integrity_register(struct gendisk *d,
1929 struct blk_integrity *b)
1932 static inline void blk_integrity_unregister(struct gendisk *d)
1935 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1939 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1943 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1949 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1956 static inline bool integrity_req_gap_back_merge(struct request *req,
1961 static inline bool integrity_req_gap_front_merge(struct request *req,
1967 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1968 unsigned int sectors)
1973 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1974 unsigned int sectors)
1979 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1981 struct block_device_operations {
1982 int (*open) (struct block_device *, fmode_t);
1983 void (*release) (struct gendisk *, fmode_t);
1984 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1985 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1986 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1987 unsigned int (*check_events) (struct gendisk *disk,
1988 unsigned int clearing);
1989 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1990 int (*media_changed) (struct gendisk *);
1991 void (*unlock_native_capacity) (struct gendisk *);
1992 int (*revalidate_disk) (struct gendisk *);
1993 int (*getgeo)(struct block_device *, struct hd_geometry *);
1994 /* this callback is with swap_lock and sometimes page table lock held */
1995 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1996 struct module *owner;
1997 const struct pr_ops *pr_ops;
2000 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
2002 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
2003 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
2004 struct writeback_control *);
2006 #ifdef CONFIG_BLK_DEV_ZONED
2007 bool blk_req_needs_zone_write_lock(struct request *rq);
2008 void __blk_req_zone_write_lock(struct request *rq);
2009 void __blk_req_zone_write_unlock(struct request *rq);
2011 static inline void blk_req_zone_write_lock(struct request *rq)
2013 if (blk_req_needs_zone_write_lock(rq))
2014 __blk_req_zone_write_lock(rq);
2017 static inline void blk_req_zone_write_unlock(struct request *rq)
2019 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
2020 __blk_req_zone_write_unlock(rq);
2023 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2025 return rq->q->seq_zones_wlock &&
2026 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
2029 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2031 if (!blk_req_needs_zone_write_lock(rq))
2033 return !blk_req_zone_is_write_locked(rq);
2036 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2041 static inline void blk_req_zone_write_lock(struct request *rq)
2045 static inline void blk_req_zone_write_unlock(struct request *rq)
2048 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2053 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2057 #endif /* CONFIG_BLK_DEV_ZONED */
2059 #else /* CONFIG_BLOCK */
2061 struct block_device;
2064 * stubs for when the block layer is configured out
2066 #define buffer_heads_over_limit 0
2068 static inline long nr_blockdev_pages(void)
2076 static inline void blk_start_plug(struct blk_plug *plug)
2080 static inline void blk_finish_plug(struct blk_plug *plug)
2084 static inline void blk_flush_plug(struct task_struct *task)
2088 static inline void blk_schedule_flush_plug(struct task_struct *task)
2093 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2098 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2099 sector_t *error_sector)
2104 #endif /* CONFIG_BLOCK */