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 into disk and partition IO statistics */
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 /* track IO completion time */
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 extern unsigned int blkdev_nr_zones(struct block_device
*bdev
);
400 extern int blkdev_report_zones(struct block_device
*bdev
,
401 sector_t sector
, struct blk_zone
*zones
,
402 unsigned int *nr_zones
, gfp_t gfp_mask
);
403 extern int blkdev_reset_zones(struct block_device
*bdev
, sector_t sectors
,
404 sector_t nr_sectors
, gfp_t gfp_mask
);
405 extern int blk_revalidate_disk_zones(struct gendisk
*disk
);
407 extern int blkdev_report_zones_ioctl(struct block_device
*bdev
, fmode_t mode
,
408 unsigned int cmd
, unsigned long arg
);
409 extern int blkdev_reset_zones_ioctl(struct block_device
*bdev
, fmode_t mode
,
410 unsigned int cmd
, unsigned long arg
);
412 #else /* CONFIG_BLK_DEV_ZONED */
414 static inline unsigned int blkdev_nr_zones(struct block_device
*bdev
)
419 static inline int blk_revalidate_disk_zones(struct gendisk
*disk
)
424 static inline int blkdev_report_zones_ioctl(struct block_device
*bdev
,
425 fmode_t mode
, unsigned int cmd
,
431 static inline int blkdev_reset_zones_ioctl(struct block_device
*bdev
,
432 fmode_t mode
, unsigned int cmd
,
438 #endif /* CONFIG_BLK_DEV_ZONED */
440 struct request_queue
{
442 * Together with queue_head for cacheline sharing
444 struct list_head queue_head
;
445 struct request
*last_merge
;
446 struct elevator_queue
*elevator
;
447 int nr_rqs
[2]; /* # allocated [a]sync rqs */
448 int nr_rqs_elvpriv
; /* # allocated rqs w/ elvpriv */
450 struct blk_queue_stats
*stats
;
451 struct rq_qos
*rq_qos
;
454 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
455 * is used, root blkg allocates from @q->root_rl and all other
456 * blkgs from their own blkg->rl. Which one to use should be
457 * determined using bio_request_list().
459 struct request_list root_rl
;
461 request_fn_proc
*request_fn
;
462 make_request_fn
*make_request_fn
;
464 prep_rq_fn
*prep_rq_fn
;
465 unprep_rq_fn
*unprep_rq_fn
;
466 softirq_done_fn
*softirq_done_fn
;
467 rq_timed_out_fn
*rq_timed_out_fn
;
468 dma_drain_needed_fn
*dma_drain_needed
;
469 lld_busy_fn
*lld_busy_fn
;
470 /* Called just after a request is allocated */
471 init_rq_fn
*init_rq_fn
;
472 /* Called just before a request is freed */
473 exit_rq_fn
*exit_rq_fn
;
474 /* Called from inside blk_get_request() */
475 void (*initialize_rq_fn
)(struct request
*rq
);
477 const struct blk_mq_ops
*mq_ops
;
479 unsigned int *mq_map
;
482 struct blk_mq_ctx __percpu
*queue_ctx
;
483 unsigned int nr_queues
;
485 unsigned int queue_depth
;
487 /* hw dispatch queues */
488 struct blk_mq_hw_ctx
**queue_hw_ctx
;
489 unsigned int nr_hw_queues
;
492 * Dispatch queue sorting
495 struct request
*boundary_rq
;
498 * Delayed queue handling
500 struct delayed_work delay_work
;
502 struct backing_dev_info
*backing_dev_info
;
505 * The queue owner gets to use this for whatever they like.
506 * ll_rw_blk doesn't touch it.
511 * various queue flags, see QUEUE_* below
513 unsigned long queue_flags
;
515 * Number of contexts that have called blk_set_pm_only(). If this
516 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
522 * ida allocated id for this queue. Used to index queues from
528 * queue needs bounce pages for pages above this limit
533 * protects queue structures from reentrancy. ->__queue_lock should
534 * _never_ be used directly, it is queue private. always use
537 spinlock_t __queue_lock
;
538 spinlock_t
*queue_lock
;
548 struct kobject mq_kobj
;
550 #ifdef CONFIG_BLK_DEV_INTEGRITY
551 struct blk_integrity integrity
;
552 #endif /* CONFIG_BLK_DEV_INTEGRITY */
557 unsigned int nr_pending
;
563 unsigned long nr_requests
; /* Max # of requests */
564 unsigned int nr_congestion_on
;
565 unsigned int nr_congestion_off
;
566 unsigned int nr_batching
;
568 unsigned int dma_drain_size
;
569 void *dma_drain_buffer
;
570 unsigned int dma_pad_mask
;
571 unsigned int dma_alignment
;
573 struct blk_queue_tag
*queue_tags
;
575 unsigned int nr_sorted
;
576 unsigned int in_flight
[2];
579 * Number of active block driver functions for which blk_drain_queue()
580 * must wait. Must be incremented around functions that unlock the
581 * queue_lock internally, e.g. scsi_request_fn().
583 unsigned int request_fn_active
;
585 unsigned int rq_timeout
;
588 struct blk_stat_callback
*poll_cb
;
589 struct blk_rq_stat poll_stat
[BLK_MQ_POLL_STATS_BKTS
];
591 struct timer_list timeout
;
592 struct work_struct timeout_work
;
593 struct list_head timeout_list
;
595 struct list_head icq_list
;
596 #ifdef CONFIG_BLK_CGROUP
597 DECLARE_BITMAP (blkcg_pols
, BLKCG_MAX_POLS
);
598 struct blkcg_gq
*root_blkg
;
599 struct list_head blkg_list
;
602 struct queue_limits limits
;
604 #ifdef CONFIG_BLK_DEV_ZONED
606 * Zoned block device information for request dispatch control.
607 * nr_zones is the total number of zones of the device. This is always
608 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
609 * bits which indicates if a zone is conventional (bit clear) or
610 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
611 * bits which indicates if a zone is write locked, that is, if a write
612 * request targeting the zone was dispatched. All three fields are
613 * initialized by the low level device driver (e.g. scsi/sd.c).
614 * Stacking drivers (device mappers) may or may not initialize
617 * Reads of this information must be protected with blk_queue_enter() /
618 * blk_queue_exit(). Modifying this information is only allowed while
619 * no requests are being processed. See also blk_mq_freeze_queue() and
620 * blk_mq_unfreeze_queue().
622 unsigned int nr_zones
;
623 unsigned long *seq_zones_bitmap
;
624 unsigned long *seq_zones_wlock
;
625 #endif /* CONFIG_BLK_DEV_ZONED */
630 unsigned int sg_timeout
;
631 unsigned int sg_reserved_size
;
633 #ifdef CONFIG_BLK_DEV_IO_TRACE
634 struct blk_trace
*blk_trace
;
635 struct mutex blk_trace_mutex
;
638 * for flush operations
640 struct blk_flush_queue
*fq
;
642 struct list_head requeue_list
;
643 spinlock_t requeue_lock
;
644 struct delayed_work requeue_work
;
646 struct mutex sysfs_lock
;
649 atomic_t mq_freeze_depth
;
651 #if defined(CONFIG_BLK_DEV_BSG)
652 bsg_job_fn
*bsg_job_fn
;
653 struct bsg_class_device bsg_dev
;
656 #ifdef CONFIG_BLK_DEV_THROTTLING
658 struct throtl_data
*td
;
660 struct rcu_head rcu_head
;
661 wait_queue_head_t mq_freeze_wq
;
662 struct percpu_ref q_usage_counter
;
663 struct list_head all_q_node
;
665 struct blk_mq_tag_set
*tag_set
;
666 struct list_head tag_set_list
;
667 struct bio_set bio_split
;
669 #ifdef CONFIG_BLK_DEBUG_FS
670 struct dentry
*debugfs_dir
;
671 struct dentry
*sched_debugfs_dir
;
674 bool mq_sysfs_init_done
;
679 struct work_struct release_work
;
681 #define BLK_MAX_WRITE_HINTS 5
682 u64 write_hints
[BLK_MAX_WRITE_HINTS
];
685 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
686 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
687 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
688 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
689 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
690 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
691 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
692 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
693 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
694 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
695 #define QUEUE_FLAG_IO_STAT 10 /* do disk/partitions IO accounting */
696 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
697 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
698 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
699 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
700 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
701 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
702 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
703 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
704 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
705 #define QUEUE_FLAG_WC 20 /* Write back caching */
706 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
707 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
708 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
709 #define QUEUE_FLAG_STATS 24 /* track IO start and completion times */
710 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
711 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
712 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
713 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
714 #define QUEUE_FLAG_PCI_P2PDMA 29 /* device supports PCI p2p requests */
716 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
717 (1 << QUEUE_FLAG_SAME_COMP) | \
718 (1 << QUEUE_FLAG_ADD_RANDOM))
720 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
721 (1 << QUEUE_FLAG_SAME_COMP) | \
722 (1 << QUEUE_FLAG_POLL))
724 void blk_queue_flag_set(unsigned int flag
, struct request_queue
*q
);
725 void blk_queue_flag_clear(unsigned int flag
, struct request_queue
*q
);
726 bool blk_queue_flag_test_and_set(unsigned int flag
, struct request_queue
*q
);
727 bool blk_queue_flag_test_and_clear(unsigned int flag
, struct request_queue
*q
);
729 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
730 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
731 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
732 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
733 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
734 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
735 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
736 #define blk_queue_noxmerges(q) \
737 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
738 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
739 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
740 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
741 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
742 #define blk_queue_secure_erase(q) \
743 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
744 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
745 #define blk_queue_scsi_passthrough(q) \
746 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
747 #define blk_queue_pci_p2pdma(q) \
748 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
750 #define blk_noretry_request(rq) \
751 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
752 REQ_FAILFAST_DRIVER))
753 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
754 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
755 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
757 extern void blk_set_pm_only(struct request_queue
*q
);
758 extern void blk_clear_pm_only(struct request_queue
*q
);
760 static inline int queue_in_flight(struct request_queue
*q
)
762 return q
->in_flight
[0] + q
->in_flight
[1];
765 static inline bool blk_account_rq(struct request
*rq
)
767 return (rq
->rq_flags
& RQF_STARTED
) && !blk_rq_is_passthrough(rq
);
770 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
771 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
772 /* rq->queuelist of dequeued request must be list_empty() */
773 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
775 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
777 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
780 * Driver can handle struct request, if it either has an old style
781 * request_fn defined, or is blk-mq based.
783 static inline bool queue_is_rq_based(struct request_queue
*q
)
785 return q
->request_fn
|| q
->mq_ops
;
788 static inline unsigned int blk_queue_cluster(struct request_queue
*q
)
790 return q
->limits
.cluster
;
793 static inline enum blk_zoned_model
794 blk_queue_zoned_model(struct request_queue
*q
)
796 return q
->limits
.zoned
;
799 static inline bool blk_queue_is_zoned(struct request_queue
*q
)
801 switch (blk_queue_zoned_model(q
)) {
810 static inline unsigned int blk_queue_zone_sectors(struct request_queue
*q
)
812 return blk_queue_is_zoned(q
) ? q
->limits
.chunk_sectors
: 0;
815 #ifdef CONFIG_BLK_DEV_ZONED
816 static inline unsigned int blk_queue_nr_zones(struct request_queue
*q
)
818 return blk_queue_is_zoned(q
) ? q
->nr_zones
: 0;
821 static inline unsigned int blk_queue_zone_no(struct request_queue
*q
,
824 if (!blk_queue_is_zoned(q
))
826 return sector
>> ilog2(q
->limits
.chunk_sectors
);
829 static inline bool blk_queue_zone_is_seq(struct request_queue
*q
,
832 if (!blk_queue_is_zoned(q
) || !q
->seq_zones_bitmap
)
834 return test_bit(blk_queue_zone_no(q
, sector
), q
->seq_zones_bitmap
);
836 #else /* CONFIG_BLK_DEV_ZONED */
837 static inline unsigned int blk_queue_nr_zones(struct request_queue
*q
)
841 #endif /* CONFIG_BLK_DEV_ZONED */
843 static inline bool rq_is_sync(struct request
*rq
)
845 return op_is_sync(rq
->cmd_flags
);
848 static inline bool blk_rl_full(struct request_list
*rl
, bool sync
)
850 unsigned int flag
= sync
? BLK_RL_SYNCFULL
: BLK_RL_ASYNCFULL
;
852 return rl
->flags
& flag
;
855 static inline void blk_set_rl_full(struct request_list
*rl
, bool sync
)
857 unsigned int flag
= sync
? BLK_RL_SYNCFULL
: BLK_RL_ASYNCFULL
;
862 static inline void blk_clear_rl_full(struct request_list
*rl
, bool sync
)
864 unsigned int flag
= sync
? BLK_RL_SYNCFULL
: BLK_RL_ASYNCFULL
;
869 static inline bool rq_mergeable(struct request
*rq
)
871 if (blk_rq_is_passthrough(rq
))
874 if (req_op(rq
) == REQ_OP_FLUSH
)
877 if (req_op(rq
) == REQ_OP_WRITE_ZEROES
)
880 if (rq
->cmd_flags
& REQ_NOMERGE_FLAGS
)
882 if (rq
->rq_flags
& RQF_NOMERGE_FLAGS
)
888 static inline bool blk_write_same_mergeable(struct bio
*a
, struct bio
*b
)
890 if (bio_page(a
) == bio_page(b
) &&
891 bio_offset(a
) == bio_offset(b
))
897 static inline unsigned int blk_queue_depth(struct request_queue
*q
)
900 return q
->queue_depth
;
902 return q
->nr_requests
;
906 * q->prep_rq_fn return values
909 BLKPREP_OK
, /* serve it */
910 BLKPREP_KILL
, /* fatal error, kill, return -EIO */
911 BLKPREP_DEFER
, /* leave on queue */
912 BLKPREP_INVALID
, /* invalid command, kill, return -EREMOTEIO */
915 extern unsigned long blk_max_low_pfn
, blk_max_pfn
;
918 * standard bounce addresses:
920 * BLK_BOUNCE_HIGH : bounce all highmem pages
921 * BLK_BOUNCE_ANY : don't bounce anything
922 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
925 #if BITS_PER_LONG == 32
926 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
928 #define BLK_BOUNCE_HIGH -1ULL
930 #define BLK_BOUNCE_ANY (-1ULL)
931 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
934 * default timeout for SG_IO if none specified
936 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
937 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
943 unsigned long offset
;
948 struct req_iterator
{
949 struct bvec_iter iter
;
953 /* This should not be used directly - use rq_for_each_segment */
954 #define for_each_bio(_bio) \
955 for (; _bio; _bio = _bio->bi_next)
956 #define __rq_for_each_bio(_bio, rq) \
958 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
960 #define rq_for_each_segment(bvl, _rq, _iter) \
961 __rq_for_each_bio(_iter.bio, _rq) \
962 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
964 #define rq_iter_last(bvec, _iter) \
965 (_iter.bio->bi_next == NULL && \
966 bio_iter_last(bvec, _iter.iter))
968 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
969 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
971 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
972 extern void rq_flush_dcache_pages(struct request
*rq
);
974 static inline void rq_flush_dcache_pages(struct request
*rq
)
979 extern int blk_register_queue(struct gendisk
*disk
);
980 extern void blk_unregister_queue(struct gendisk
*disk
);
981 extern blk_qc_t
generic_make_request(struct bio
*bio
);
982 extern blk_qc_t
direct_make_request(struct bio
*bio
);
983 extern void blk_rq_init(struct request_queue
*q
, struct request
*rq
);
984 extern void blk_init_request_from_bio(struct request
*req
, struct bio
*bio
);
985 extern void blk_put_request(struct request
*);
986 extern void __blk_put_request(struct request_queue
*, struct request
*);
987 extern struct request
*blk_get_request(struct request_queue
*, unsigned int op
,
988 blk_mq_req_flags_t flags
);
989 extern void blk_requeue_request(struct request_queue
*, struct request
*);
990 extern int blk_lld_busy(struct request_queue
*q
);
991 extern int blk_rq_prep_clone(struct request
*rq
, struct request
*rq_src
,
992 struct bio_set
*bs
, gfp_t gfp_mask
,
993 int (*bio_ctr
)(struct bio
*, struct bio
*, void *),
995 extern void blk_rq_unprep_clone(struct request
*rq
);
996 extern blk_status_t
blk_insert_cloned_request(struct request_queue
*q
,
998 extern int blk_rq_append_bio(struct request
*rq
, struct bio
**bio
);
999 extern void blk_delay_queue(struct request_queue
*, unsigned long);
1000 extern void blk_queue_split(struct request_queue
*, struct bio
**);
1001 extern void blk_recount_segments(struct request_queue
*, struct bio
*);
1002 extern int scsi_verify_blk_ioctl(struct block_device
*, unsigned int);
1003 extern int scsi_cmd_blk_ioctl(struct block_device
*, fmode_t
,
1004 unsigned int, void __user
*);
1005 extern int scsi_cmd_ioctl(struct request_queue
*, struct gendisk
*, fmode_t
,
1006 unsigned int, void __user
*);
1007 extern int sg_scsi_ioctl(struct request_queue
*, struct gendisk
*, fmode_t
,
1008 struct scsi_ioctl_command __user
*);
1010 extern int blk_queue_enter(struct request_queue
*q
, blk_mq_req_flags_t flags
);
1011 extern void blk_queue_exit(struct request_queue
*q
);
1012 extern void blk_start_queue(struct request_queue
*q
);
1013 extern void blk_start_queue_async(struct request_queue
*q
);
1014 extern void blk_stop_queue(struct request_queue
*q
);
1015 extern void blk_sync_queue(struct request_queue
*q
);
1016 extern void __blk_stop_queue(struct request_queue
*q
);
1017 extern void __blk_run_queue(struct request_queue
*q
);
1018 extern void __blk_run_queue_uncond(struct request_queue
*q
);
1019 extern void blk_run_queue(struct request_queue
*);
1020 extern void blk_run_queue_async(struct request_queue
*q
);
1021 extern int blk_rq_map_user(struct request_queue
*, struct request
*,
1022 struct rq_map_data
*, void __user
*, unsigned long,
1024 extern int blk_rq_unmap_user(struct bio
*);
1025 extern int blk_rq_map_kern(struct request_queue
*, struct request
*, void *, unsigned int, gfp_t
);
1026 extern int blk_rq_map_user_iov(struct request_queue
*, struct request
*,
1027 struct rq_map_data
*, const struct iov_iter
*,
1029 extern void blk_execute_rq(struct request_queue
*, struct gendisk
*,
1030 struct request
*, int);
1031 extern void blk_execute_rq_nowait(struct request_queue
*, struct gendisk
*,
1032 struct request
*, int, rq_end_io_fn
*);
1034 int blk_status_to_errno(blk_status_t status
);
1035 blk_status_t
errno_to_blk_status(int errno
);
1037 bool blk_poll(struct request_queue
*q
, blk_qc_t cookie
);
1039 static inline struct request_queue
*bdev_get_queue(struct block_device
*bdev
)
1041 return bdev
->bd_disk
->queue
; /* this is never NULL */
1045 * The basic unit of block I/O is a sector. It is used in a number of contexts
1046 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1047 * bytes. Variables of type sector_t represent an offset or size that is a
1048 * multiple of 512 bytes. Hence these two constants.
1050 #ifndef SECTOR_SHIFT
1051 #define SECTOR_SHIFT 9
1054 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1058 * blk_rq_pos() : the current sector
1059 * blk_rq_bytes() : bytes left in the entire request
1060 * blk_rq_cur_bytes() : bytes left in the current segment
1061 * blk_rq_err_bytes() : bytes left till the next error boundary
1062 * blk_rq_sectors() : sectors left in the entire request
1063 * blk_rq_cur_sectors() : sectors left in the current segment
1065 static inline sector_t
blk_rq_pos(const struct request
*rq
)
1067 return rq
->__sector
;
1070 static inline unsigned int blk_rq_bytes(const struct request
*rq
)
1072 return rq
->__data_len
;
1075 static inline int blk_rq_cur_bytes(const struct request
*rq
)
1077 return rq
->bio
? bio_cur_bytes(rq
->bio
) : 0;
1080 extern unsigned int blk_rq_err_bytes(const struct request
*rq
);
1082 static inline unsigned int blk_rq_sectors(const struct request
*rq
)
1084 return blk_rq_bytes(rq
) >> SECTOR_SHIFT
;
1087 static inline unsigned int blk_rq_cur_sectors(const struct request
*rq
)
1089 return blk_rq_cur_bytes(rq
) >> SECTOR_SHIFT
;
1092 #ifdef CONFIG_BLK_DEV_ZONED
1093 static inline unsigned int blk_rq_zone_no(struct request
*rq
)
1095 return blk_queue_zone_no(rq
->q
, blk_rq_pos(rq
));
1098 static inline unsigned int blk_rq_zone_is_seq(struct request
*rq
)
1100 return blk_queue_zone_is_seq(rq
->q
, blk_rq_pos(rq
));
1102 #endif /* CONFIG_BLK_DEV_ZONED */
1105 * Some commands like WRITE SAME have a payload or data transfer size which
1106 * is different from the size of the request. Any driver that supports such
1107 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1108 * calculate the data transfer size.
1110 static inline unsigned int blk_rq_payload_bytes(struct request
*rq
)
1112 if (rq
->rq_flags
& RQF_SPECIAL_PAYLOAD
)
1113 return rq
->special_vec
.bv_len
;
1114 return blk_rq_bytes(rq
);
1117 static inline unsigned int blk_queue_get_max_sectors(struct request_queue
*q
,
1120 if (unlikely(op
== REQ_OP_DISCARD
|| op
== REQ_OP_SECURE_ERASE
))
1121 return min(q
->limits
.max_discard_sectors
,
1122 UINT_MAX
>> SECTOR_SHIFT
);
1124 if (unlikely(op
== REQ_OP_WRITE_SAME
))
1125 return q
->limits
.max_write_same_sectors
;
1127 if (unlikely(op
== REQ_OP_WRITE_ZEROES
))
1128 return q
->limits
.max_write_zeroes_sectors
;
1130 return q
->limits
.max_sectors
;
1134 * Return maximum size of a request at given offset. Only valid for
1135 * file system requests.
1137 static inline unsigned int blk_max_size_offset(struct request_queue
*q
,
1140 if (!q
->limits
.chunk_sectors
)
1141 return q
->limits
.max_sectors
;
1143 return min(q
->limits
.max_sectors
, (unsigned int)(q
->limits
.chunk_sectors
-
1144 (offset
& (q
->limits
.chunk_sectors
- 1))));
1147 static inline unsigned int blk_rq_get_max_sectors(struct request
*rq
,
1150 struct request_queue
*q
= rq
->q
;
1152 if (blk_rq_is_passthrough(rq
))
1153 return q
->limits
.max_hw_sectors
;
1155 if (!q
->limits
.chunk_sectors
||
1156 req_op(rq
) == REQ_OP_DISCARD
||
1157 req_op(rq
) == REQ_OP_SECURE_ERASE
)
1158 return blk_queue_get_max_sectors(q
, req_op(rq
));
1160 return min(blk_max_size_offset(q
, offset
),
1161 blk_queue_get_max_sectors(q
, req_op(rq
)));
1164 static inline unsigned int blk_rq_count_bios(struct request
*rq
)
1166 unsigned int nr_bios
= 0;
1169 __rq_for_each_bio(bio
, rq
)
1176 * Request issue related functions.
1178 extern struct request
*blk_peek_request(struct request_queue
*q
);
1179 extern void blk_start_request(struct request
*rq
);
1180 extern struct request
*blk_fetch_request(struct request_queue
*q
);
1182 void blk_steal_bios(struct bio_list
*list
, struct request
*rq
);
1185 * Request completion related functions.
1187 * blk_update_request() completes given number of bytes and updates
1188 * the request without completing it.
1190 * blk_end_request() and friends. __blk_end_request() must be called
1191 * with the request queue spinlock acquired.
1193 * Several drivers define their own end_request and call
1194 * blk_end_request() for parts of the original function.
1195 * This prevents code duplication in drivers.
1197 extern bool blk_update_request(struct request
*rq
, blk_status_t error
,
1198 unsigned int nr_bytes
);
1199 extern void blk_finish_request(struct request
*rq
, blk_status_t error
);
1200 extern bool blk_end_request(struct request
*rq
, blk_status_t error
,
1201 unsigned int nr_bytes
);
1202 extern void blk_end_request_all(struct request
*rq
, blk_status_t error
);
1203 extern bool __blk_end_request(struct request
*rq
, blk_status_t error
,
1204 unsigned int nr_bytes
);
1205 extern void __blk_end_request_all(struct request
*rq
, blk_status_t error
);
1206 extern bool __blk_end_request_cur(struct request
*rq
, blk_status_t error
);
1208 extern void blk_complete_request(struct request
*);
1209 extern void __blk_complete_request(struct request
*);
1210 extern void blk_abort_request(struct request
*);
1211 extern void blk_unprep_request(struct request
*);
1214 * Access functions for manipulating queue properties
1216 extern struct request_queue
*blk_init_queue_node(request_fn_proc
*rfn
,
1217 spinlock_t
*lock
, int node_id
);
1218 extern struct request_queue
*blk_init_queue(request_fn_proc
*, spinlock_t
*);
1219 extern int blk_init_allocated_queue(struct request_queue
*);
1220 extern void blk_cleanup_queue(struct request_queue
*);
1221 extern void blk_queue_make_request(struct request_queue
*, make_request_fn
*);
1222 extern void blk_queue_bounce_limit(struct request_queue
*, u64
);
1223 extern void blk_queue_max_hw_sectors(struct request_queue
*, unsigned int);
1224 extern void blk_queue_chunk_sectors(struct request_queue
*, unsigned int);
1225 extern void blk_queue_max_segments(struct request_queue
*, unsigned short);
1226 extern void blk_queue_max_discard_segments(struct request_queue
*,
1228 extern void blk_queue_max_segment_size(struct request_queue
*, unsigned int);
1229 extern void blk_queue_max_discard_sectors(struct request_queue
*q
,
1230 unsigned int max_discard_sectors
);
1231 extern void blk_queue_max_write_same_sectors(struct request_queue
*q
,
1232 unsigned int max_write_same_sectors
);
1233 extern void blk_queue_max_write_zeroes_sectors(struct request_queue
*q
,
1234 unsigned int max_write_same_sectors
);
1235 extern void blk_queue_logical_block_size(struct request_queue
*, unsigned short);
1236 extern void blk_queue_physical_block_size(struct request_queue
*, unsigned int);
1237 extern void blk_queue_alignment_offset(struct request_queue
*q
,
1238 unsigned int alignment
);
1239 extern void blk_limits_io_min(struct queue_limits
*limits
, unsigned int min
);
1240 extern void blk_queue_io_min(struct request_queue
*q
, unsigned int min
);
1241 extern void blk_limits_io_opt(struct queue_limits
*limits
, unsigned int opt
);
1242 extern void blk_queue_io_opt(struct request_queue
*q
, unsigned int opt
);
1243 extern void blk_set_queue_depth(struct request_queue
*q
, unsigned int depth
);
1244 extern void blk_set_default_limits(struct queue_limits
*lim
);
1245 extern void blk_set_stacking_limits(struct queue_limits
*lim
);
1246 extern int blk_stack_limits(struct queue_limits
*t
, struct queue_limits
*b
,
1248 extern int bdev_stack_limits(struct queue_limits
*t
, struct block_device
*bdev
,
1250 extern void disk_stack_limits(struct gendisk
*disk
, struct block_device
*bdev
,
1252 extern void blk_queue_stack_limits(struct request_queue
*t
, struct request_queue
*b
);
1253 extern void blk_queue_dma_pad(struct request_queue
*, unsigned int);
1254 extern void blk_queue_update_dma_pad(struct request_queue
*, unsigned int);
1255 extern int blk_queue_dma_drain(struct request_queue
*q
,
1256 dma_drain_needed_fn
*dma_drain_needed
,
1257 void *buf
, unsigned int size
);
1258 extern void blk_queue_lld_busy(struct request_queue
*q
, lld_busy_fn
*fn
);
1259 extern void blk_queue_segment_boundary(struct request_queue
*, unsigned long);
1260 extern void blk_queue_virt_boundary(struct request_queue
*, unsigned long);
1261 extern void blk_queue_prep_rq(struct request_queue
*, prep_rq_fn
*pfn
);
1262 extern void blk_queue_unprep_rq(struct request_queue
*, unprep_rq_fn
*ufn
);
1263 extern void blk_queue_dma_alignment(struct request_queue
*, int);
1264 extern void blk_queue_update_dma_alignment(struct request_queue
*, int);
1265 extern void blk_queue_softirq_done(struct request_queue
*, softirq_done_fn
*);
1266 extern void blk_queue_rq_timed_out(struct request_queue
*, rq_timed_out_fn
*);
1267 extern void blk_queue_rq_timeout(struct request_queue
*, unsigned int);
1268 extern void blk_queue_flush_queueable(struct request_queue
*q
, bool queueable
);
1269 extern void blk_queue_write_cache(struct request_queue
*q
, bool enabled
, bool fua
);
1272 * Number of physical segments as sent to the device.
1274 * Normally this is the number of discontiguous data segments sent by the
1275 * submitter. But for data-less command like discard we might have no
1276 * actual data segments submitted, but the driver might have to add it's
1277 * own special payload. In that case we still return 1 here so that this
1278 * special payload will be mapped.
1280 static inline unsigned short blk_rq_nr_phys_segments(struct request
*rq
)
1282 if (rq
->rq_flags
& RQF_SPECIAL_PAYLOAD
)
1284 return rq
->nr_phys_segments
;
1288 * Number of discard segments (or ranges) the driver needs to fill in.
1289 * Each discard bio merged into a request is counted as one segment.
1291 static inline unsigned short blk_rq_nr_discard_segments(struct request
*rq
)
1293 return max_t(unsigned short, rq
->nr_phys_segments
, 1);
1296 extern int blk_rq_map_sg(struct request_queue
*, struct request
*, struct scatterlist
*);
1297 extern void blk_dump_rq_flags(struct request
*, char *);
1298 extern long nr_blockdev_pages(void);
1300 bool __must_check
blk_get_queue(struct request_queue
*);
1301 struct request_queue
*blk_alloc_queue(gfp_t
);
1302 struct request_queue
*blk_alloc_queue_node(gfp_t gfp_mask
, int node_id
,
1304 extern void blk_put_queue(struct request_queue
*);
1305 extern void blk_set_queue_dying(struct request_queue
*);
1308 * blk_plug permits building a queue of related requests by holding the I/O
1309 * fragments for a short period. This allows merging of sequential requests
1310 * into single larger request. As the requests are moved from a per-task list to
1311 * the device's request_queue in a batch, this results in improved scalability
1312 * as the lock contention for request_queue lock is reduced.
1314 * It is ok not to disable preemption when adding the request to the plug list
1315 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1316 * the plug list when the task sleeps by itself. For details, please see
1317 * schedule() where blk_schedule_flush_plug() is called.
1320 struct list_head list
; /* requests */
1321 struct list_head mq_list
; /* blk-mq requests */
1322 struct list_head cb_list
; /* md requires an unplug callback */
1324 #define BLK_MAX_REQUEST_COUNT 16
1325 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1328 typedef void (*blk_plug_cb_fn
)(struct blk_plug_cb
*, bool);
1329 struct blk_plug_cb
{
1330 struct list_head list
;
1331 blk_plug_cb_fn callback
;
1334 extern struct blk_plug_cb
*blk_check_plugged(blk_plug_cb_fn unplug
,
1335 void *data
, int size
);
1336 extern void blk_start_plug(struct blk_plug
*);
1337 extern void blk_finish_plug(struct blk_plug
*);
1338 extern void blk_flush_plug_list(struct blk_plug
*, bool);
1340 static inline void blk_flush_plug(struct task_struct
*tsk
)
1342 struct blk_plug
*plug
= tsk
->plug
;
1345 blk_flush_plug_list(plug
, false);
1348 static inline void blk_schedule_flush_plug(struct task_struct
*tsk
)
1350 struct blk_plug
*plug
= tsk
->plug
;
1353 blk_flush_plug_list(plug
, true);
1356 static inline bool blk_needs_flush_plug(struct task_struct
*tsk
)
1358 struct blk_plug
*plug
= tsk
->plug
;
1361 (!list_empty(&plug
->list
) ||
1362 !list_empty(&plug
->mq_list
) ||
1363 !list_empty(&plug
->cb_list
));
1369 extern int blk_queue_start_tag(struct request_queue
*, struct request
*);
1370 extern struct request
*blk_queue_find_tag(struct request_queue
*, int);
1371 extern void blk_queue_end_tag(struct request_queue
*, struct request
*);
1372 extern int blk_queue_init_tags(struct request_queue
*, int, struct blk_queue_tag
*, int);
1373 extern void blk_queue_free_tags(struct request_queue
*);
1374 extern int blk_queue_resize_tags(struct request_queue
*, int);
1375 extern struct blk_queue_tag
*blk_init_tags(int, int);
1376 extern void blk_free_tags(struct blk_queue_tag
*);
1378 static inline struct request
*blk_map_queue_find_tag(struct blk_queue_tag
*bqt
,
1381 if (unlikely(bqt
== NULL
|| tag
>= bqt
->real_max_depth
))
1383 return bqt
->tag_index
[tag
];
1386 extern int blkdev_issue_flush(struct block_device
*, gfp_t
, sector_t
*);
1387 extern int blkdev_issue_write_same(struct block_device
*bdev
, sector_t sector
,
1388 sector_t nr_sects
, gfp_t gfp_mask
, struct page
*page
);
1390 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1392 extern int blkdev_issue_discard(struct block_device
*bdev
, sector_t sector
,
1393 sector_t nr_sects
, gfp_t gfp_mask
, unsigned long flags
);
1394 extern int __blkdev_issue_discard(struct block_device
*bdev
, sector_t sector
,
1395 sector_t nr_sects
, gfp_t gfp_mask
, int flags
,
1398 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1399 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1401 extern int __blkdev_issue_zeroout(struct block_device
*bdev
, sector_t sector
,
1402 sector_t nr_sects
, gfp_t gfp_mask
, struct bio
**biop
,
1404 extern int blkdev_issue_zeroout(struct block_device
*bdev
, sector_t sector
,
1405 sector_t nr_sects
, gfp_t gfp_mask
, unsigned flags
);
1407 static inline int sb_issue_discard(struct super_block
*sb
, sector_t block
,
1408 sector_t nr_blocks
, gfp_t gfp_mask
, unsigned long flags
)
1410 return blkdev_issue_discard(sb
->s_bdev
,
1411 block
<< (sb
->s_blocksize_bits
-
1413 nr_blocks
<< (sb
->s_blocksize_bits
-
1417 static inline int sb_issue_zeroout(struct super_block
*sb
, sector_t block
,
1418 sector_t nr_blocks
, gfp_t gfp_mask
)
1420 return blkdev_issue_zeroout(sb
->s_bdev
,
1421 block
<< (sb
->s_blocksize_bits
-
1423 nr_blocks
<< (sb
->s_blocksize_bits
-
1428 extern int blk_verify_command(unsigned char *cmd
, fmode_t mode
);
1430 enum blk_default_limits
{
1431 BLK_MAX_SEGMENTS
= 128,
1432 BLK_SAFE_MAX_SECTORS
= 255,
1433 BLK_DEF_MAX_SECTORS
= 2560,
1434 BLK_MAX_SEGMENT_SIZE
= 65536,
1435 BLK_SEG_BOUNDARY_MASK
= 0xFFFFFFFFUL
,
1438 static inline unsigned long queue_segment_boundary(struct request_queue
*q
)
1440 return q
->limits
.seg_boundary_mask
;
1443 static inline unsigned long queue_virt_boundary(struct request_queue
*q
)
1445 return q
->limits
.virt_boundary_mask
;
1448 static inline unsigned int queue_max_sectors(struct request_queue
*q
)
1450 return q
->limits
.max_sectors
;
1453 static inline unsigned int queue_max_hw_sectors(struct request_queue
*q
)
1455 return q
->limits
.max_hw_sectors
;
1458 static inline unsigned short queue_max_segments(struct request_queue
*q
)
1460 return q
->limits
.max_segments
;
1463 static inline unsigned short queue_max_discard_segments(struct request_queue
*q
)
1465 return q
->limits
.max_discard_segments
;
1468 static inline unsigned int queue_max_segment_size(struct request_queue
*q
)
1470 return q
->limits
.max_segment_size
;
1473 static inline unsigned short queue_logical_block_size(struct request_queue
*q
)
1477 if (q
&& q
->limits
.logical_block_size
)
1478 retval
= q
->limits
.logical_block_size
;
1483 static inline unsigned short bdev_logical_block_size(struct block_device
*bdev
)
1485 return queue_logical_block_size(bdev_get_queue(bdev
));
1488 static inline unsigned int queue_physical_block_size(struct request_queue
*q
)
1490 return q
->limits
.physical_block_size
;
1493 static inline unsigned int bdev_physical_block_size(struct block_device
*bdev
)
1495 return queue_physical_block_size(bdev_get_queue(bdev
));
1498 static inline unsigned int queue_io_min(struct request_queue
*q
)
1500 return q
->limits
.io_min
;
1503 static inline int bdev_io_min(struct block_device
*bdev
)
1505 return queue_io_min(bdev_get_queue(bdev
));
1508 static inline unsigned int queue_io_opt(struct request_queue
*q
)
1510 return q
->limits
.io_opt
;
1513 static inline int bdev_io_opt(struct block_device
*bdev
)
1515 return queue_io_opt(bdev_get_queue(bdev
));
1518 static inline int queue_alignment_offset(struct request_queue
*q
)
1520 if (q
->limits
.misaligned
)
1523 return q
->limits
.alignment_offset
;
1526 static inline int queue_limit_alignment_offset(struct queue_limits
*lim
, sector_t sector
)
1528 unsigned int granularity
= max(lim
->physical_block_size
, lim
->io_min
);
1529 unsigned int alignment
= sector_div(sector
, granularity
>> SECTOR_SHIFT
)
1532 return (granularity
+ lim
->alignment_offset
- alignment
) % granularity
;
1535 static inline int bdev_alignment_offset(struct block_device
*bdev
)
1537 struct request_queue
*q
= bdev_get_queue(bdev
);
1539 if (q
->limits
.misaligned
)
1542 if (bdev
!= bdev
->bd_contains
)
1543 return bdev
->bd_part
->alignment_offset
;
1545 return q
->limits
.alignment_offset
;
1548 static inline int queue_discard_alignment(struct request_queue
*q
)
1550 if (q
->limits
.discard_misaligned
)
1553 return q
->limits
.discard_alignment
;
1556 static inline int queue_limit_discard_alignment(struct queue_limits
*lim
, sector_t sector
)
1558 unsigned int alignment
, granularity
, offset
;
1560 if (!lim
->max_discard_sectors
)
1563 /* Why are these in bytes, not sectors? */
1564 alignment
= lim
->discard_alignment
>> SECTOR_SHIFT
;
1565 granularity
= lim
->discard_granularity
>> SECTOR_SHIFT
;
1569 /* Offset of the partition start in 'granularity' sectors */
1570 offset
= sector_div(sector
, granularity
);
1572 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1573 offset
= (granularity
+ alignment
- offset
) % granularity
;
1575 /* Turn it back into bytes, gaah */
1576 return offset
<< SECTOR_SHIFT
;
1579 static inline int bdev_discard_alignment(struct block_device
*bdev
)
1581 struct request_queue
*q
= bdev_get_queue(bdev
);
1583 if (bdev
!= bdev
->bd_contains
)
1584 return bdev
->bd_part
->discard_alignment
;
1586 return q
->limits
.discard_alignment
;
1589 static inline unsigned int bdev_write_same(struct block_device
*bdev
)
1591 struct request_queue
*q
= bdev_get_queue(bdev
);
1594 return q
->limits
.max_write_same_sectors
;
1599 static inline unsigned int bdev_write_zeroes_sectors(struct block_device
*bdev
)
1601 struct request_queue
*q
= bdev_get_queue(bdev
);
1604 return q
->limits
.max_write_zeroes_sectors
;
1609 static inline enum blk_zoned_model
bdev_zoned_model(struct block_device
*bdev
)
1611 struct request_queue
*q
= bdev_get_queue(bdev
);
1614 return blk_queue_zoned_model(q
);
1616 return BLK_ZONED_NONE
;
1619 static inline bool bdev_is_zoned(struct block_device
*bdev
)
1621 struct request_queue
*q
= bdev_get_queue(bdev
);
1624 return blk_queue_is_zoned(q
);
1629 static inline unsigned int bdev_zone_sectors(struct block_device
*bdev
)
1631 struct request_queue
*q
= bdev_get_queue(bdev
);
1634 return blk_queue_zone_sectors(q
);
1638 static inline int queue_dma_alignment(struct request_queue
*q
)
1640 return q
? q
->dma_alignment
: 511;
1643 static inline int blk_rq_aligned(struct request_queue
*q
, unsigned long addr
,
1646 unsigned int alignment
= queue_dma_alignment(q
) | q
->dma_pad_mask
;
1647 return !(addr
& alignment
) && !(len
& alignment
);
1650 /* assumes size > 256 */
1651 static inline unsigned int blksize_bits(unsigned int size
)
1653 unsigned int bits
= 8;
1657 } while (size
> 256);
1661 static inline unsigned int block_size(struct block_device
*bdev
)
1663 return bdev
->bd_block_size
;
1666 static inline bool queue_flush_queueable(struct request_queue
*q
)
1668 return !test_bit(QUEUE_FLAG_FLUSH_NQ
, &q
->queue_flags
);
1671 typedef struct {struct page
*v
;} Sector
;
1673 unsigned char *read_dev_sector(struct block_device
*, sector_t
, Sector
*);
1675 static inline void put_dev_sector(Sector p
)
1680 int kblockd_schedule_work(struct work_struct
*work
);
1681 int kblockd_schedule_work_on(int cpu
, struct work_struct
*work
);
1682 int kblockd_mod_delayed_work_on(int cpu
, struct delayed_work
*dwork
, unsigned long delay
);
1684 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1685 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1686 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1687 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1689 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1691 enum blk_integrity_flags
{
1692 BLK_INTEGRITY_VERIFY
= 1 << 0,
1693 BLK_INTEGRITY_GENERATE
= 1 << 1,
1694 BLK_INTEGRITY_DEVICE_CAPABLE
= 1 << 2,
1695 BLK_INTEGRITY_IP_CHECKSUM
= 1 << 3,
1698 struct blk_integrity_iter
{
1702 unsigned int data_size
;
1703 unsigned short interval
;
1704 const char *disk_name
;
1707 typedef blk_status_t (integrity_processing_fn
) (struct blk_integrity_iter
*);
1709 struct blk_integrity_profile
{
1710 integrity_processing_fn
*generate_fn
;
1711 integrity_processing_fn
*verify_fn
;
1715 extern void blk_integrity_register(struct gendisk
*, struct blk_integrity
*);
1716 extern void blk_integrity_unregister(struct gendisk
*);
1717 extern int blk_integrity_compare(struct gendisk
*, struct gendisk
*);
1718 extern int blk_rq_map_integrity_sg(struct request_queue
*, struct bio
*,
1719 struct scatterlist
*);
1720 extern int blk_rq_count_integrity_sg(struct request_queue
*, struct bio
*);
1721 extern bool blk_integrity_merge_rq(struct request_queue
*, struct request
*,
1723 extern bool blk_integrity_merge_bio(struct request_queue
*, struct request
*,
1726 static inline struct blk_integrity
*blk_get_integrity(struct gendisk
*disk
)
1728 struct blk_integrity
*bi
= &disk
->queue
->integrity
;
1737 struct blk_integrity
*bdev_get_integrity(struct block_device
*bdev
)
1739 return blk_get_integrity(bdev
->bd_disk
);
1742 static inline bool blk_integrity_rq(struct request
*rq
)
1744 return rq
->cmd_flags
& REQ_INTEGRITY
;
1747 static inline void blk_queue_max_integrity_segments(struct request_queue
*q
,
1750 q
->limits
.max_integrity_segments
= segs
;
1753 static inline unsigned short
1754 queue_max_integrity_segments(struct request_queue
*q
)
1756 return q
->limits
.max_integrity_segments
;
1760 * bio_integrity_intervals - Return number of integrity intervals for a bio
1761 * @bi: blk_integrity profile for device
1762 * @sectors: Size of the bio in 512-byte sectors
1764 * Description: The block layer calculates everything in 512 byte
1765 * sectors but integrity metadata is done in terms of the data integrity
1766 * interval size of the storage device. Convert the block layer sectors
1767 * to the appropriate number of integrity intervals.
1769 static inline unsigned int bio_integrity_intervals(struct blk_integrity
*bi
,
1770 unsigned int sectors
)
1772 return sectors
>> (bi
->interval_exp
- 9);
1775 static inline unsigned int bio_integrity_bytes(struct blk_integrity
*bi
,
1776 unsigned int sectors
)
1778 return bio_integrity_intervals(bi
, sectors
) * bi
->tuple_size
;
1781 #else /* CONFIG_BLK_DEV_INTEGRITY */
1784 struct block_device
;
1786 struct blk_integrity
;
1788 static inline int blk_integrity_rq(struct request
*rq
)
1792 static inline int blk_rq_count_integrity_sg(struct request_queue
*q
,
1797 static inline int blk_rq_map_integrity_sg(struct request_queue
*q
,
1799 struct scatterlist
*s
)
1803 static inline struct blk_integrity
*bdev_get_integrity(struct block_device
*b
)
1807 static inline struct blk_integrity
*blk_get_integrity(struct gendisk
*disk
)
1811 static inline int blk_integrity_compare(struct gendisk
*a
, struct gendisk
*b
)
1815 static inline void blk_integrity_register(struct gendisk
*d
,
1816 struct blk_integrity
*b
)
1819 static inline void blk_integrity_unregister(struct gendisk
*d
)
1822 static inline void blk_queue_max_integrity_segments(struct request_queue
*q
,
1826 static inline unsigned short queue_max_integrity_segments(struct request_queue
*q
)
1830 static inline bool blk_integrity_merge_rq(struct request_queue
*rq
,
1836 static inline bool blk_integrity_merge_bio(struct request_queue
*rq
,
1843 static inline unsigned int bio_integrity_intervals(struct blk_integrity
*bi
,
1844 unsigned int sectors
)
1849 static inline unsigned int bio_integrity_bytes(struct blk_integrity
*bi
,
1850 unsigned int sectors
)
1855 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1857 struct block_device_operations
{
1858 int (*open
) (struct block_device
*, fmode_t
);
1859 void (*release
) (struct gendisk
*, fmode_t
);
1860 int (*rw_page
)(struct block_device
*, sector_t
, struct page
*, unsigned int);
1861 int (*ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1862 int (*compat_ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1863 unsigned int (*check_events
) (struct gendisk
*disk
,
1864 unsigned int clearing
);
1865 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1866 int (*media_changed
) (struct gendisk
*);
1867 void (*unlock_native_capacity
) (struct gendisk
*);
1868 int (*revalidate_disk
) (struct gendisk
*);
1869 int (*getgeo
)(struct block_device
*, struct hd_geometry
*);
1870 /* this callback is with swap_lock and sometimes page table lock held */
1871 void (*swap_slot_free_notify
) (struct block_device
*, unsigned long);
1872 int (*report_zones
)(struct gendisk
*, sector_t sector
,
1873 struct blk_zone
*zones
, unsigned int *nr_zones
,
1875 struct module
*owner
;
1876 const struct pr_ops
*pr_ops
;
1879 extern int __blkdev_driver_ioctl(struct block_device
*, fmode_t
, unsigned int,
1881 extern int bdev_read_page(struct block_device
*, sector_t
, struct page
*);
1882 extern int bdev_write_page(struct block_device
*, sector_t
, struct page
*,
1883 struct writeback_control
*);
1885 #ifdef CONFIG_BLK_DEV_ZONED
1886 bool blk_req_needs_zone_write_lock(struct request
*rq
);
1887 void __blk_req_zone_write_lock(struct request
*rq
);
1888 void __blk_req_zone_write_unlock(struct request
*rq
);
1890 static inline void blk_req_zone_write_lock(struct request
*rq
)
1892 if (blk_req_needs_zone_write_lock(rq
))
1893 __blk_req_zone_write_lock(rq
);
1896 static inline void blk_req_zone_write_unlock(struct request
*rq
)
1898 if (rq
->rq_flags
& RQF_ZONE_WRITE_LOCKED
)
1899 __blk_req_zone_write_unlock(rq
);
1902 static inline bool blk_req_zone_is_write_locked(struct request
*rq
)
1904 return rq
->q
->seq_zones_wlock
&&
1905 test_bit(blk_rq_zone_no(rq
), rq
->q
->seq_zones_wlock
);
1908 static inline bool blk_req_can_dispatch_to_zone(struct request
*rq
)
1910 if (!blk_req_needs_zone_write_lock(rq
))
1912 return !blk_req_zone_is_write_locked(rq
);
1915 static inline bool blk_req_needs_zone_write_lock(struct request
*rq
)
1920 static inline void blk_req_zone_write_lock(struct request
*rq
)
1924 static inline void blk_req_zone_write_unlock(struct request
*rq
)
1927 static inline bool blk_req_zone_is_write_locked(struct request
*rq
)
1932 static inline bool blk_req_can_dispatch_to_zone(struct request
*rq
)
1936 #endif /* CONFIG_BLK_DEV_ZONED */
1938 #else /* CONFIG_BLOCK */
1940 struct block_device
;
1943 * stubs for when the block layer is configured out
1945 #define buffer_heads_over_limit 0
1947 static inline long nr_blockdev_pages(void)
1955 static inline void blk_start_plug(struct blk_plug
*plug
)
1959 static inline void blk_finish_plug(struct blk_plug
*plug
)
1963 static inline void blk_flush_plug(struct task_struct
*task
)
1967 static inline void blk_schedule_flush_plug(struct task_struct
*task
)
1972 static inline bool blk_needs_flush_plug(struct task_struct
*tsk
)
1977 static inline int blkdev_issue_flush(struct block_device
*bdev
, gfp_t gfp_mask
,
1978 sector_t
*error_sector
)
1983 #endif /* CONFIG_BLOCK */