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 3
59 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
61 #define BLK_RL_SYNCFULL (1U << 0)
62 #define BLK_RL_ASYNCFULL (1U << 1)
65 struct request_queue *q; /* the queue this rl belongs to */
66 #ifdef CONFIG_BLK_CGROUP
67 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
70 * count[], starved[], and wait[] are indexed by
71 * BLK_RW_SYNC/BLK_RW_ASYNC
76 wait_queue_head_t wait[2];
82 typedef __u32 __bitwise req_flags_t;
84 /* elevator knows about this request */
85 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
86 /* drive already may have started this one */
87 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
88 /* uses tagged queueing */
89 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
90 /* may not be passed by ioscheduler */
91 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
92 /* request for flush sequence */
93 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
94 /* merge of different types, fail separately */
95 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
96 /* track inflight for MQ */
97 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
98 /* don't call prep for this one */
99 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
100 /* set for "ide_preempt" requests and also for requests for which the SCSI
101 "quiesce" state must be ignored. */
102 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
103 /* contains copies of user pages */
104 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
105 /* vaguely specified driver internal error. Ignored by the block layer */
106 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
107 /* don't warn about errors */
108 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
109 /* elevator private data attached */
110 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
111 /* account I/O stat */
112 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
113 /* request came from our alloc pool */
114 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
115 /* runtime pm request */
116 #define RQF_PM ((__force req_flags_t)(1 << 15))
117 /* on IO scheduler merge hash */
118 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
119 /* IO stats tracking on */
120 #define RQF_STATS ((__force req_flags_t)(1 << 17))
121 /* Look at ->special_vec for the actual data payload instead of the
123 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
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 alignment_offset;
378 unsigned int max_discard_sectors;
379 unsigned int max_hw_discard_sectors;
380 unsigned int max_write_same_sectors;
381 unsigned int max_write_zeroes_sectors;
382 unsigned int discard_granularity;
383 unsigned int discard_alignment;
385 unsigned short logical_block_size;
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;
509 * ida allocated id for this queue. Used to index queues from
515 * queue needs bounce pages for pages above this limit
520 * protects queue structures from reentrancy. ->__queue_lock should
521 * _never_ be used directly, it is queue private. always use
524 spinlock_t __queue_lock;
525 spinlock_t *queue_lock;
535 struct kobject mq_kobj;
537 #ifdef CONFIG_BLK_DEV_INTEGRITY
538 struct blk_integrity integrity;
539 #endif /* CONFIG_BLK_DEV_INTEGRITY */
544 unsigned int nr_pending;
550 unsigned long nr_requests; /* Max # of requests */
551 unsigned int nr_congestion_on;
552 unsigned int nr_congestion_off;
553 unsigned int nr_batching;
555 unsigned int dma_drain_size;
556 void *dma_drain_buffer;
557 unsigned int dma_pad_mask;
558 unsigned int dma_alignment;
560 struct blk_queue_tag *queue_tags;
562 unsigned int nr_sorted;
563 unsigned int in_flight[2];
566 * Number of active block driver functions for which blk_drain_queue()
567 * must wait. Must be incremented around functions that unlock the
568 * queue_lock internally, e.g. scsi_request_fn().
570 unsigned int request_fn_active;
572 unsigned int rq_timeout;
575 struct blk_stat_callback *poll_cb;
576 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
578 struct timer_list timeout;
579 struct work_struct timeout_work;
580 struct list_head timeout_list;
582 struct list_head icq_list;
583 #ifdef CONFIG_BLK_CGROUP
584 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
585 struct blkcg_gq *root_blkg;
586 struct list_head blkg_list;
589 struct queue_limits limits;
591 #ifdef CONFIG_BLK_DEV_ZONED
593 * Zoned block device information for request dispatch control.
594 * nr_zones is the total number of zones of the device. This is always
595 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
596 * bits which indicates if a zone is conventional (bit clear) or
597 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
598 * bits which indicates if a zone is write locked, that is, if a write
599 * request targeting the zone was dispatched. All three fields are
600 * initialized by the low level device driver (e.g. scsi/sd.c).
601 * Stacking drivers (device mappers) may or may not initialize
604 * Reads of this information must be protected with blk_queue_enter() /
605 * blk_queue_exit(). Modifying this information is only allowed while
606 * no requests are being processed. See also blk_mq_freeze_queue() and
607 * blk_mq_unfreeze_queue().
609 unsigned int nr_zones;
610 unsigned long *seq_zones_bitmap;
611 unsigned long *seq_zones_wlock;
612 #endif /* CONFIG_BLK_DEV_ZONED */
617 unsigned int sg_timeout;
618 unsigned int sg_reserved_size;
620 #ifdef CONFIG_BLK_DEV_IO_TRACE
621 struct blk_trace *blk_trace;
622 struct mutex blk_trace_mutex;
625 * for flush operations
627 struct blk_flush_queue *fq;
629 struct list_head requeue_list;
630 spinlock_t requeue_lock;
631 struct delayed_work requeue_work;
633 struct mutex sysfs_lock;
636 atomic_t mq_freeze_depth;
638 #if defined(CONFIG_BLK_DEV_BSG)
639 bsg_job_fn *bsg_job_fn;
640 struct bsg_class_device bsg_dev;
643 #ifdef CONFIG_BLK_DEV_THROTTLING
645 struct throtl_data *td;
647 struct rcu_head rcu_head;
648 wait_queue_head_t mq_freeze_wq;
649 struct percpu_ref q_usage_counter;
650 struct list_head all_q_node;
652 struct blk_mq_tag_set *tag_set;
653 struct list_head tag_set_list;
654 struct bio_set bio_split;
656 #ifdef CONFIG_BLK_DEBUG_FS
657 struct dentry *debugfs_dir;
658 struct dentry *sched_debugfs_dir;
661 bool mq_sysfs_init_done;
666 struct work_struct release_work;
668 #define BLK_MAX_WRITE_HINTS 5
669 u64 write_hints[BLK_MAX_WRITE_HINTS];
672 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
673 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
674 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
675 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
676 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
677 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
678 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
679 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
680 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
681 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
682 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
683 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
684 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
685 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
686 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
687 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
688 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
689 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
690 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
691 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
692 #define QUEUE_FLAG_WC 20 /* Write back caching */
693 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
694 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
695 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
696 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
697 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
698 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
699 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
700 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
701 #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */
703 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
704 (1 << QUEUE_FLAG_SAME_COMP) | \
705 (1 << QUEUE_FLAG_ADD_RANDOM))
707 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
708 (1 << QUEUE_FLAG_SAME_COMP) | \
709 (1 << QUEUE_FLAG_POLL))
711 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
712 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
713 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
714 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
716 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
717 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
718 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
719 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
720 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
721 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
722 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
723 #define blk_queue_noxmerges(q) \
724 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
725 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
726 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
727 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
728 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
729 #define blk_queue_secure_erase(q) \
730 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
731 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
732 #define blk_queue_scsi_passthrough(q) \
733 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
735 #define blk_noretry_request(rq) \
736 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
737 REQ_FAILFAST_DRIVER))
738 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
739 #define blk_queue_preempt_only(q) \
740 test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
741 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
743 extern int blk_set_preempt_only(struct request_queue *q);
744 extern void blk_clear_preempt_only(struct request_queue *q);
746 static inline int queue_in_flight(struct request_queue *q)
748 return q->in_flight[0] + q->in_flight[1];
751 static inline bool blk_account_rq(struct request *rq)
753 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
756 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
757 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
758 /* rq->queuelist of dequeued request must be list_empty() */
759 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
761 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
763 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
766 * Driver can handle struct request, if it either has an old style
767 * request_fn defined, or is blk-mq based.
769 static inline bool queue_is_rq_based(struct request_queue *q)
771 return q->request_fn || q->mq_ops;
774 static inline unsigned int blk_queue_cluster(struct request_queue *q)
776 return q->limits.cluster;
779 static inline enum blk_zoned_model
780 blk_queue_zoned_model(struct request_queue *q)
782 return q->limits.zoned;
785 static inline bool blk_queue_is_zoned(struct request_queue *q)
787 switch (blk_queue_zoned_model(q)) {
796 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
798 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
801 #ifdef CONFIG_BLK_DEV_ZONED
802 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
805 if (!blk_queue_is_zoned(q))
807 return sector >> ilog2(q->limits.chunk_sectors);
810 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
813 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
815 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
817 #endif /* CONFIG_BLK_DEV_ZONED */
819 static inline bool rq_is_sync(struct request *rq)
821 return op_is_sync(rq->cmd_flags);
824 static inline bool blk_rl_full(struct request_list *rl, bool sync)
826 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
828 return rl->flags & flag;
831 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
833 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
838 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
840 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
845 static inline bool rq_mergeable(struct request *rq)
847 if (blk_rq_is_passthrough(rq))
850 if (req_op(rq) == REQ_OP_FLUSH)
853 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
856 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
858 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
864 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
866 if (bio_page(a) == bio_page(b) &&
867 bio_offset(a) == bio_offset(b))
873 static inline unsigned int blk_queue_depth(struct request_queue *q)
876 return q->queue_depth;
878 return q->nr_requests;
882 * q->prep_rq_fn return values
885 BLKPREP_OK, /* serve it */
886 BLKPREP_KILL, /* fatal error, kill, return -EIO */
887 BLKPREP_DEFER, /* leave on queue */
888 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
891 extern unsigned long blk_max_low_pfn, blk_max_pfn;
894 * standard bounce addresses:
896 * BLK_BOUNCE_HIGH : bounce all highmem pages
897 * BLK_BOUNCE_ANY : don't bounce anything
898 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
901 #if BITS_PER_LONG == 32
902 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
904 #define BLK_BOUNCE_HIGH -1ULL
906 #define BLK_BOUNCE_ANY (-1ULL)
907 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
910 * default timeout for SG_IO if none specified
912 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
913 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
919 unsigned long offset;
924 struct req_iterator {
925 struct bvec_iter iter;
929 /* This should not be used directly - use rq_for_each_segment */
930 #define for_each_bio(_bio) \
931 for (; _bio; _bio = _bio->bi_next)
932 #define __rq_for_each_bio(_bio, rq) \
934 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
936 #define rq_for_each_segment(bvl, _rq, _iter) \
937 __rq_for_each_bio(_iter.bio, _rq) \
938 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
940 #define rq_iter_last(bvec, _iter) \
941 (_iter.bio->bi_next == NULL && \
942 bio_iter_last(bvec, _iter.iter))
944 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
945 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
947 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
948 extern void rq_flush_dcache_pages(struct request *rq);
950 static inline void rq_flush_dcache_pages(struct request *rq)
955 extern int blk_register_queue(struct gendisk *disk);
956 extern void blk_unregister_queue(struct gendisk *disk);
957 extern blk_qc_t generic_make_request(struct bio *bio);
958 extern blk_qc_t direct_make_request(struct bio *bio);
959 extern void blk_rq_init(struct request_queue *q, struct request *rq);
960 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
961 extern void blk_put_request(struct request *);
962 extern void __blk_put_request(struct request_queue *, struct request *);
963 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
964 blk_mq_req_flags_t flags);
965 extern void blk_requeue_request(struct request_queue *, struct request *);
966 extern int blk_lld_busy(struct request_queue *q);
967 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
968 struct bio_set *bs, gfp_t gfp_mask,
969 int (*bio_ctr)(struct bio *, struct bio *, void *),
971 extern void blk_rq_unprep_clone(struct request *rq);
972 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
974 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
975 extern void blk_delay_queue(struct request_queue *, unsigned long);
976 extern void blk_queue_split(struct request_queue *, struct bio **);
977 extern void blk_recount_segments(struct request_queue *, struct bio *);
978 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
979 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
980 unsigned int, void __user *);
981 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
982 unsigned int, void __user *);
983 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
984 struct scsi_ioctl_command __user *);
986 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
987 extern void blk_queue_exit(struct request_queue *q);
988 extern void blk_start_queue(struct request_queue *q);
989 extern void blk_start_queue_async(struct request_queue *q);
990 extern void blk_stop_queue(struct request_queue *q);
991 extern void blk_sync_queue(struct request_queue *q);
992 extern void __blk_stop_queue(struct request_queue *q);
993 extern void __blk_run_queue(struct request_queue *q);
994 extern void __blk_run_queue_uncond(struct request_queue *q);
995 extern void blk_run_queue(struct request_queue *);
996 extern void blk_run_queue_async(struct request_queue *q);
997 extern int blk_rq_map_user(struct request_queue *, struct request *,
998 struct rq_map_data *, void __user *, unsigned long,
1000 extern int blk_rq_unmap_user(struct bio *);
1001 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1002 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1003 struct rq_map_data *, const struct iov_iter *,
1005 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1006 struct request *, int);
1007 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1008 struct request *, int, rq_end_io_fn *);
1010 int blk_status_to_errno(blk_status_t status);
1011 blk_status_t errno_to_blk_status(int errno);
1013 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1015 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1017 return bdev->bd_disk->queue; /* this is never NULL */
1021 * The basic unit of block I/O is a sector. It is used in a number of contexts
1022 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1023 * bytes. Variables of type sector_t represent an offset or size that is a
1024 * multiple of 512 bytes. Hence these two constants.
1026 #ifndef SECTOR_SHIFT
1027 #define SECTOR_SHIFT 9
1030 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1034 * blk_rq_pos() : the current sector
1035 * blk_rq_bytes() : bytes left in the entire request
1036 * blk_rq_cur_bytes() : bytes left in the current segment
1037 * blk_rq_err_bytes() : bytes left till the next error boundary
1038 * blk_rq_sectors() : sectors left in the entire request
1039 * blk_rq_cur_sectors() : sectors left in the current segment
1041 static inline sector_t blk_rq_pos(const struct request *rq)
1043 return rq->__sector;
1046 static inline unsigned int blk_rq_bytes(const struct request *rq)
1048 return rq->__data_len;
1051 static inline int blk_rq_cur_bytes(const struct request *rq)
1053 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1056 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1058 static inline unsigned int blk_rq_sectors(const struct request *rq)
1060 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1063 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1065 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1068 #ifdef CONFIG_BLK_DEV_ZONED
1069 static inline unsigned int blk_rq_zone_no(struct request *rq)
1071 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1074 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1076 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1078 #endif /* CONFIG_BLK_DEV_ZONED */
1081 * Some commands like WRITE SAME have a payload or data transfer size which
1082 * is different from the size of the request. Any driver that supports such
1083 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1084 * calculate the data transfer size.
1086 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1088 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1089 return rq->special_vec.bv_len;
1090 return blk_rq_bytes(rq);
1093 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1096 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1097 return min(q->limits.max_discard_sectors,
1098 UINT_MAX >> SECTOR_SHIFT);
1100 if (unlikely(op == REQ_OP_WRITE_SAME))
1101 return q->limits.max_write_same_sectors;
1103 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1104 return q->limits.max_write_zeroes_sectors;
1106 return q->limits.max_sectors;
1110 * Return maximum size of a request at given offset. Only valid for
1111 * file system requests.
1113 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1116 if (!q->limits.chunk_sectors)
1117 return q->limits.max_sectors;
1119 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1120 (offset & (q->limits.chunk_sectors - 1))));
1123 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1126 struct request_queue *q = rq->q;
1128 if (blk_rq_is_passthrough(rq))
1129 return q->limits.max_hw_sectors;
1131 if (!q->limits.chunk_sectors ||
1132 req_op(rq) == REQ_OP_DISCARD ||
1133 req_op(rq) == REQ_OP_SECURE_ERASE)
1134 return blk_queue_get_max_sectors(q, req_op(rq));
1136 return min(blk_max_size_offset(q, offset),
1137 blk_queue_get_max_sectors(q, req_op(rq)));
1140 static inline unsigned int blk_rq_count_bios(struct request *rq)
1142 unsigned int nr_bios = 0;
1145 __rq_for_each_bio(bio, rq)
1152 * Request issue related functions.
1154 extern struct request *blk_peek_request(struct request_queue *q);
1155 extern void blk_start_request(struct request *rq);
1156 extern struct request *blk_fetch_request(struct request_queue *q);
1158 void blk_steal_bios(struct bio_list *list, struct request *rq);
1161 * Request completion related functions.
1163 * blk_update_request() completes given number of bytes and updates
1164 * the request without completing it.
1166 * blk_end_request() and friends. __blk_end_request() must be called
1167 * with the request queue spinlock acquired.
1169 * Several drivers define their own end_request and call
1170 * blk_end_request() for parts of the original function.
1171 * This prevents code duplication in drivers.
1173 extern bool blk_update_request(struct request *rq, blk_status_t error,
1174 unsigned int nr_bytes);
1175 extern void blk_finish_request(struct request *rq, blk_status_t error);
1176 extern bool blk_end_request(struct request *rq, blk_status_t error,
1177 unsigned int nr_bytes);
1178 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1179 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1180 unsigned int nr_bytes);
1181 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1182 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1184 extern void blk_complete_request(struct request *);
1185 extern void __blk_complete_request(struct request *);
1186 extern void blk_abort_request(struct request *);
1187 extern void blk_unprep_request(struct request *);
1190 * Access functions for manipulating queue properties
1192 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1193 spinlock_t *lock, int node_id);
1194 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1195 extern int blk_init_allocated_queue(struct request_queue *);
1196 extern void blk_cleanup_queue(struct request_queue *);
1197 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1198 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1199 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1200 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1201 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1202 extern void blk_queue_max_discard_segments(struct request_queue *,
1204 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1205 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1206 unsigned int max_discard_sectors);
1207 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1208 unsigned int max_write_same_sectors);
1209 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1210 unsigned int max_write_same_sectors);
1211 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1212 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1213 extern void blk_queue_alignment_offset(struct request_queue *q,
1214 unsigned int alignment);
1215 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1216 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1217 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1218 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1219 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1220 extern void blk_set_default_limits(struct queue_limits *lim);
1221 extern void blk_set_stacking_limits(struct queue_limits *lim);
1222 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1224 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1226 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1228 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1229 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1230 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1231 extern int blk_queue_dma_drain(struct request_queue *q,
1232 dma_drain_needed_fn *dma_drain_needed,
1233 void *buf, unsigned int size);
1234 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1235 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1236 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1237 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1238 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1239 extern void blk_queue_dma_alignment(struct request_queue *, int);
1240 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1241 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1242 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1243 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1244 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1245 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1248 * Number of physical segments as sent to the device.
1250 * Normally this is the number of discontiguous data segments sent by the
1251 * submitter. But for data-less command like discard we might have no
1252 * actual data segments submitted, but the driver might have to add it's
1253 * own special payload. In that case we still return 1 here so that this
1254 * special payload will be mapped.
1256 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1258 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1260 return rq->nr_phys_segments;
1264 * Number of discard segments (or ranges) the driver needs to fill in.
1265 * Each discard bio merged into a request is counted as one segment.
1267 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1269 return max_t(unsigned short, rq->nr_phys_segments, 1);
1272 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1273 extern void blk_dump_rq_flags(struct request *, char *);
1274 extern long nr_blockdev_pages(void);
1276 bool __must_check blk_get_queue(struct request_queue *);
1277 struct request_queue *blk_alloc_queue(gfp_t);
1278 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1280 extern void blk_put_queue(struct request_queue *);
1281 extern void blk_set_queue_dying(struct request_queue *);
1284 * block layer runtime pm functions
1287 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1288 extern int blk_pre_runtime_suspend(struct request_queue *q);
1289 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1290 extern void blk_pre_runtime_resume(struct request_queue *q);
1291 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1292 extern void blk_set_runtime_active(struct request_queue *q);
1294 static inline void blk_pm_runtime_init(struct request_queue *q,
1295 struct device *dev) {}
1296 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1300 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1301 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1302 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1303 static inline void blk_set_runtime_active(struct request_queue *q) {}
1307 * blk_plug permits building a queue of related requests by holding the I/O
1308 * fragments for a short period. This allows merging of sequential requests
1309 * into single larger request. As the requests are moved from a per-task list to
1310 * the device's request_queue in a batch, this results in improved scalability
1311 * as the lock contention for request_queue lock is reduced.
1313 * It is ok not to disable preemption when adding the request to the plug list
1314 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1315 * the plug list when the task sleeps by itself. For details, please see
1316 * schedule() where blk_schedule_flush_plug() is called.
1319 struct list_head list; /* requests */
1320 struct list_head mq_list; /* blk-mq requests */
1321 struct list_head cb_list; /* md requires an unplug callback */
1323 #define BLK_MAX_REQUEST_COUNT 16
1324 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1327 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1328 struct blk_plug_cb {
1329 struct list_head list;
1330 blk_plug_cb_fn callback;
1333 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1334 void *data, int size);
1335 extern void blk_start_plug(struct blk_plug *);
1336 extern void blk_finish_plug(struct blk_plug *);
1337 extern void blk_flush_plug_list(struct blk_plug *, bool);
1339 static inline void blk_flush_plug(struct task_struct *tsk)
1341 struct blk_plug *plug = tsk->plug;
1344 blk_flush_plug_list(plug, false);
1347 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1349 struct blk_plug *plug = tsk->plug;
1352 blk_flush_plug_list(plug, true);
1355 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1357 struct blk_plug *plug = tsk->plug;
1360 (!list_empty(&plug->list) ||
1361 !list_empty(&plug->mq_list) ||
1362 !list_empty(&plug->cb_list));
1368 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1369 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1370 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1371 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1372 extern void blk_queue_free_tags(struct request_queue *);
1373 extern int blk_queue_resize_tags(struct request_queue *, int);
1374 extern struct blk_queue_tag *blk_init_tags(int, int);
1375 extern void blk_free_tags(struct blk_queue_tag *);
1377 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1380 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1382 return bqt->tag_index[tag];
1385 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1386 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1387 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1389 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1391 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1392 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1393 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1394 sector_t nr_sects, gfp_t gfp_mask, int flags,
1397 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1398 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1400 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1401 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1403 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1404 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1406 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1407 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1409 return blkdev_issue_discard(sb->s_bdev,
1410 block << (sb->s_blocksize_bits -
1412 nr_blocks << (sb->s_blocksize_bits -
1416 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1417 sector_t nr_blocks, gfp_t gfp_mask)
1419 return blkdev_issue_zeroout(sb->s_bdev,
1420 block << (sb->s_blocksize_bits -
1422 nr_blocks << (sb->s_blocksize_bits -
1427 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1429 enum blk_default_limits {
1430 BLK_MAX_SEGMENTS = 128,
1431 BLK_SAFE_MAX_SECTORS = 255,
1432 BLK_DEF_MAX_SECTORS = 2560,
1433 BLK_MAX_SEGMENT_SIZE = 65536,
1434 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1437 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1439 return q->limits.seg_boundary_mask;
1442 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1444 return q->limits.virt_boundary_mask;
1447 static inline unsigned int queue_max_sectors(struct request_queue *q)
1449 return q->limits.max_sectors;
1452 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1454 return q->limits.max_hw_sectors;
1457 static inline unsigned short queue_max_segments(struct request_queue *q)
1459 return q->limits.max_segments;
1462 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1464 return q->limits.max_discard_segments;
1467 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1469 return q->limits.max_segment_size;
1472 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1476 if (q && q->limits.logical_block_size)
1477 retval = q->limits.logical_block_size;
1482 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1484 return queue_logical_block_size(bdev_get_queue(bdev));
1487 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1489 return q->limits.physical_block_size;
1492 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1494 return queue_physical_block_size(bdev_get_queue(bdev));
1497 static inline unsigned int queue_io_min(struct request_queue *q)
1499 return q->limits.io_min;
1502 static inline int bdev_io_min(struct block_device *bdev)
1504 return queue_io_min(bdev_get_queue(bdev));
1507 static inline unsigned int queue_io_opt(struct request_queue *q)
1509 return q->limits.io_opt;
1512 static inline int bdev_io_opt(struct block_device *bdev)
1514 return queue_io_opt(bdev_get_queue(bdev));
1517 static inline int queue_alignment_offset(struct request_queue *q)
1519 if (q->limits.misaligned)
1522 return q->limits.alignment_offset;
1525 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1527 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1528 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1531 return (granularity + lim->alignment_offset - alignment) % granularity;
1534 static inline int bdev_alignment_offset(struct block_device *bdev)
1536 struct request_queue *q = bdev_get_queue(bdev);
1538 if (q->limits.misaligned)
1541 if (bdev != bdev->bd_contains)
1542 return bdev->bd_part->alignment_offset;
1544 return q->limits.alignment_offset;
1547 static inline int queue_discard_alignment(struct request_queue *q)
1549 if (q->limits.discard_misaligned)
1552 return q->limits.discard_alignment;
1555 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1557 unsigned int alignment, granularity, offset;
1559 if (!lim->max_discard_sectors)
1562 /* Why are these in bytes, not sectors? */
1563 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1564 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1568 /* Offset of the partition start in 'granularity' sectors */
1569 offset = sector_div(sector, granularity);
1571 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1572 offset = (granularity + alignment - offset) % granularity;
1574 /* Turn it back into bytes, gaah */
1575 return offset << SECTOR_SHIFT;
1578 static inline int bdev_discard_alignment(struct block_device *bdev)
1580 struct request_queue *q = bdev_get_queue(bdev);
1582 if (bdev != bdev->bd_contains)
1583 return bdev->bd_part->discard_alignment;
1585 return q->limits.discard_alignment;
1588 static inline unsigned int bdev_write_same(struct block_device *bdev)
1590 struct request_queue *q = bdev_get_queue(bdev);
1593 return q->limits.max_write_same_sectors;
1598 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1600 struct request_queue *q = bdev_get_queue(bdev);
1603 return q->limits.max_write_zeroes_sectors;
1608 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1610 struct request_queue *q = bdev_get_queue(bdev);
1613 return blk_queue_zoned_model(q);
1615 return BLK_ZONED_NONE;
1618 static inline bool bdev_is_zoned(struct block_device *bdev)
1620 struct request_queue *q = bdev_get_queue(bdev);
1623 return blk_queue_is_zoned(q);
1628 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1630 struct request_queue *q = bdev_get_queue(bdev);
1633 return blk_queue_zone_sectors(q);
1637 static inline int queue_dma_alignment(struct request_queue *q)
1639 return q ? q->dma_alignment : 511;
1642 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1645 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1646 return !(addr & alignment) && !(len & alignment);
1649 /* assumes size > 256 */
1650 static inline unsigned int blksize_bits(unsigned int size)
1652 unsigned int bits = 8;
1656 } while (size > 256);
1660 static inline unsigned int block_size(struct block_device *bdev)
1662 return bdev->bd_block_size;
1665 static inline bool queue_flush_queueable(struct request_queue *q)
1667 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1670 typedef struct {struct page *v;} Sector;
1672 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1674 static inline void put_dev_sector(Sector p)
1679 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1680 struct bio_vec *bprv, unsigned int offset)
1683 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1687 * Check if adding a bio_vec after bprv with offset would create a gap in
1688 * the SG list. Most drivers don't care about this, but some do.
1690 static inline bool bvec_gap_to_prev(struct request_queue *q,
1691 struct bio_vec *bprv, unsigned int offset)
1693 if (!queue_virt_boundary(q))
1695 return __bvec_gap_to_prev(q, bprv, offset);
1699 * Check if the two bvecs from two bios can be merged to one segment.
1700 * If yes, no need to check gap between the two bios since the 1st bio
1701 * and the 1st bvec in the 2nd bio can be handled in one segment.
1703 static inline bool bios_segs_mergeable(struct request_queue *q,
1704 struct bio *prev, struct bio_vec *prev_last_bv,
1705 struct bio_vec *next_first_bv)
1707 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1709 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1711 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1712 queue_max_segment_size(q))
1717 static inline bool bio_will_gap(struct request_queue *q,
1718 struct request *prev_rq,
1722 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1723 struct bio_vec pb, nb;
1726 * don't merge if the 1st bio starts with non-zero
1727 * offset, otherwise it is quite difficult to respect
1728 * sg gap limit. We work hard to merge a huge number of small
1729 * single bios in case of mkfs.
1732 bio_get_first_bvec(prev_rq->bio, &pb);
1734 bio_get_first_bvec(prev, &pb);
1739 * We don't need to worry about the situation that the
1740 * merged segment ends in unaligned virt boundary:
1742 * - if 'pb' ends aligned, the merged segment ends aligned
1743 * - if 'pb' ends unaligned, the next bio must include
1744 * one single bvec of 'nb', otherwise the 'nb' can't
1747 bio_get_last_bvec(prev, &pb);
1748 bio_get_first_bvec(next, &nb);
1750 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1751 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1757 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1759 return bio_will_gap(req->q, req, req->biotail, bio);
1762 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1764 return bio_will_gap(req->q, NULL, bio, req->bio);
1767 int kblockd_schedule_work(struct work_struct *work);
1768 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1769 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1771 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1772 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1773 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1774 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1776 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1778 enum blk_integrity_flags {
1779 BLK_INTEGRITY_VERIFY = 1 << 0,
1780 BLK_INTEGRITY_GENERATE = 1 << 1,
1781 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1782 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1785 struct blk_integrity_iter {
1789 unsigned int data_size;
1790 unsigned short interval;
1791 const char *disk_name;
1794 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1796 struct blk_integrity_profile {
1797 integrity_processing_fn *generate_fn;
1798 integrity_processing_fn *verify_fn;
1802 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1803 extern void blk_integrity_unregister(struct gendisk *);
1804 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1805 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1806 struct scatterlist *);
1807 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1808 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1810 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1813 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1815 struct blk_integrity *bi = &disk->queue->integrity;
1824 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1826 return blk_get_integrity(bdev->bd_disk);
1829 static inline bool blk_integrity_rq(struct request *rq)
1831 return rq->cmd_flags & REQ_INTEGRITY;
1834 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1837 q->limits.max_integrity_segments = segs;
1840 static inline unsigned short
1841 queue_max_integrity_segments(struct request_queue *q)
1843 return q->limits.max_integrity_segments;
1846 static inline bool integrity_req_gap_back_merge(struct request *req,
1849 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1850 struct bio_integrity_payload *bip_next = bio_integrity(next);
1852 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1853 bip_next->bip_vec[0].bv_offset);
1856 static inline bool integrity_req_gap_front_merge(struct request *req,
1859 struct bio_integrity_payload *bip = bio_integrity(bio);
1860 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1862 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1863 bip_next->bip_vec[0].bv_offset);
1867 * bio_integrity_intervals - Return number of integrity intervals for a bio
1868 * @bi: blk_integrity profile for device
1869 * @sectors: Size of the bio in 512-byte sectors
1871 * Description: The block layer calculates everything in 512 byte
1872 * sectors but integrity metadata is done in terms of the data integrity
1873 * interval size of the storage device. Convert the block layer sectors
1874 * to the appropriate number of integrity intervals.
1876 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1877 unsigned int sectors)
1879 return sectors >> (bi->interval_exp - 9);
1882 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1883 unsigned int sectors)
1885 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1888 #else /* CONFIG_BLK_DEV_INTEGRITY */
1891 struct block_device;
1893 struct blk_integrity;
1895 static inline int blk_integrity_rq(struct request *rq)
1899 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1904 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1906 struct scatterlist *s)
1910 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1914 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1918 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1922 static inline void blk_integrity_register(struct gendisk *d,
1923 struct blk_integrity *b)
1926 static inline void blk_integrity_unregister(struct gendisk *d)
1929 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1933 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1937 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1943 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1950 static inline bool integrity_req_gap_back_merge(struct request *req,
1955 static inline bool integrity_req_gap_front_merge(struct request *req,
1961 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1962 unsigned int sectors)
1967 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1968 unsigned int sectors)
1973 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1975 struct block_device_operations {
1976 int (*open) (struct block_device *, fmode_t);
1977 void (*release) (struct gendisk *, fmode_t);
1978 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1979 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1980 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1981 unsigned int (*check_events) (struct gendisk *disk,
1982 unsigned int clearing);
1983 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1984 int (*media_changed) (struct gendisk *);
1985 void (*unlock_native_capacity) (struct gendisk *);
1986 int (*revalidate_disk) (struct gendisk *);
1987 int (*getgeo)(struct block_device *, struct hd_geometry *);
1988 /* this callback is with swap_lock and sometimes page table lock held */
1989 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1990 struct module *owner;
1991 const struct pr_ops *pr_ops;
1994 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1996 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1997 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1998 struct writeback_control *);
2000 #ifdef CONFIG_BLK_DEV_ZONED
2001 bool blk_req_needs_zone_write_lock(struct request *rq);
2002 void __blk_req_zone_write_lock(struct request *rq);
2003 void __blk_req_zone_write_unlock(struct request *rq);
2005 static inline void blk_req_zone_write_lock(struct request *rq)
2007 if (blk_req_needs_zone_write_lock(rq))
2008 __blk_req_zone_write_lock(rq);
2011 static inline void blk_req_zone_write_unlock(struct request *rq)
2013 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
2014 __blk_req_zone_write_unlock(rq);
2017 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2019 return rq->q->seq_zones_wlock &&
2020 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
2023 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2025 if (!blk_req_needs_zone_write_lock(rq))
2027 return !blk_req_zone_is_write_locked(rq);
2030 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2035 static inline void blk_req_zone_write_lock(struct request *rq)
2039 static inline void blk_req_zone_write_unlock(struct request *rq)
2042 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2047 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2051 #endif /* CONFIG_BLK_DEV_ZONED */
2053 #else /* CONFIG_BLOCK */
2055 struct block_device;
2058 * stubs for when the block layer is configured out
2060 #define buffer_heads_over_limit 0
2062 static inline long nr_blockdev_pages(void)
2070 static inline void blk_start_plug(struct blk_plug *plug)
2074 static inline void blk_finish_plug(struct blk_plug *plug)
2078 static inline void blk_flush_plug(struct task_struct *task)
2082 static inline void blk_schedule_flush_plug(struct task_struct *task)
2087 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2092 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2093 sector_t *error_sector)
2098 #endif /* CONFIG_BLOCK */