1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
32 struct scsi_ioctl_command
;
35 struct elevator_queue
;
41 struct blk_flush_queue
;
44 struct blk_queue_stats
;
45 struct blk_stat_callback
;
47 #define BLKDEV_MIN_RQ 4
48 #define BLKDEV_MAX_RQ 128 /* Default maximum */
50 /* Must be consisitent with blk_mq_poll_stats_bkt() */
51 #define BLK_MQ_POLL_STATS_BKTS 16
54 * Maximum number of blkcg policies allowed to be registered concurrently.
55 * Defined here to simplify include dependency.
57 #define BLKCG_MAX_POLS 3
59 typedef void (rq_end_io_fn
)(struct request
*, blk_status_t
);
61 #define BLK_RL_SYNCFULL (1U << 0)
62 #define BLK_RL_ASYNCFULL (1U << 1)
65 struct request_queue
*q
; /* the queue this rl belongs to */
66 #ifdef CONFIG_BLK_CGROUP
67 struct blkcg_gq
*blkg
; /* blkg this request pool belongs to */
70 * count[], starved[], and wait[] are indexed by
71 * BLK_RW_SYNC/BLK_RW_ASYNC
76 wait_queue_head_t wait
[2];
82 typedef __u32 __bitwise req_flags_t
;
84 /* elevator knows about this request */
85 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
86 /* drive already may have started this one */
87 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
88 /* uses tagged queueing */
89 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
90 /* may not be passed by ioscheduler */
91 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
92 /* request for flush sequence */
93 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
94 /* merge of different types, fail separately */
95 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
96 /* track inflight for MQ */
97 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
98 /* don't call prep for this one */
99 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
100 /* set for "ide_preempt" requests and also for requests for which the SCSI
101 "quiesce" state must be ignored. */
102 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
103 /* contains copies of user pages */
104 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
105 /* vaguely specified driver internal error. Ignored by the block layer */
106 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
107 /* don't warn about errors */
108 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
109 /* elevator private data attached */
110 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
111 /* account I/O stat */
112 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
113 /* request came from our alloc pool */
114 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
115 /* runtime pm request */
116 #define RQF_PM ((__force req_flags_t)(1 << 15))
117 /* on IO scheduler merge hash */
118 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
119 /* IO stats tracking on */
120 #define RQF_STATS ((__force req_flags_t)(1 << 17))
121 /* Look at ->special_vec for the actual data payload instead of the
123 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
125 /* flags that prevent us from merging requests: */
126 #define RQF_NOMERGE_FLAGS \
127 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
130 * Try to put the fields that are referenced together in the same cacheline.
132 * If you modify this structure, make sure to update blk_rq_init() and
133 * especially blk_mq_rq_ctx_init() to take care of the added fields.
136 struct list_head queuelist
;
138 call_single_data_t csd
;
142 struct request_queue
*q
;
143 struct blk_mq_ctx
*mq_ctx
;
146 unsigned int cmd_flags
; /* op and common flags */
147 req_flags_t rq_flags
;
151 unsigned long atomic_flags
;
153 /* the following two fields are internal, NEVER access directly */
154 unsigned int __data_len
; /* total data len */
156 sector_t __sector
; /* sector cursor */
162 * The hash is used inside the scheduler, and killed once the
163 * request reaches the dispatch list. The ipi_list is only used
164 * to queue the request for softirq completion, which is long
165 * after the request has been unhashed (and even removed from
166 * the dispatch list).
169 struct hlist_node hash
; /* merge hash */
170 struct list_head ipi_list
;
174 * The rb_node is only used inside the io scheduler, requests
175 * are pruned when moved to the dispatch queue. So let the
176 * completion_data share space with the rb_node.
179 struct rb_node rb_node
; /* sort/lookup */
180 struct bio_vec special_vec
;
181 void *completion_data
;
182 int error_count
; /* for legacy drivers, don't use */
186 * Three pointers are available for the IO schedulers, if they need
187 * more they have to dynamically allocate it. Flush requests are
188 * never put on the IO scheduler. So let the flush fields share
189 * space with the elevator data.
199 struct list_head list
;
200 rq_end_io_fn
*saved_end_io
;
204 struct gendisk
*rq_disk
;
205 struct hd_struct
*part
;
206 unsigned long start_time
;
207 struct blk_issue_stat issue_stat
;
208 #ifdef CONFIG_BLK_CGROUP
209 struct request_list
*rl
; /* rl this rq is alloced from */
210 unsigned long long start_time_ns
;
211 unsigned long long io_start_time_ns
; /* when passed to hardware */
213 /* Number of scatter-gather DMA addr+len pairs after
214 * physical address coalescing is performed.
216 unsigned short nr_phys_segments
;
217 #if defined(CONFIG_BLK_DEV_INTEGRITY)
218 unsigned short nr_integrity_segments
;
221 unsigned short ioprio
;
223 unsigned int timeout
;
225 void *special
; /* opaque pointer available for LLD use */
227 unsigned int extra_len
; /* length of alignment and padding */
229 unsigned short write_hint
;
231 unsigned long deadline
;
232 struct list_head timeout_list
;
235 * completion callback.
237 rq_end_io_fn
*end_io
;
241 struct request
*next_rq
;
244 static inline bool blk_rq_is_scsi(struct request
*rq
)
246 return req_op(rq
) == REQ_OP_SCSI_IN
|| req_op(rq
) == REQ_OP_SCSI_OUT
;
249 static inline bool blk_rq_is_private(struct request
*rq
)
251 return req_op(rq
) == REQ_OP_DRV_IN
|| req_op(rq
) == REQ_OP_DRV_OUT
;
254 static inline bool blk_rq_is_passthrough(struct request
*rq
)
256 return blk_rq_is_scsi(rq
) || blk_rq_is_private(rq
);
259 static inline unsigned short req_get_ioprio(struct request
*req
)
264 #include <linux/elevator.h>
266 struct blk_queue_ctx
;
268 typedef void (request_fn_proc
) (struct request_queue
*q
);
269 typedef blk_qc_t (make_request_fn
) (struct request_queue
*q
, struct bio
*bio
);
270 typedef bool (poll_q_fn
) (struct request_queue
*q
, blk_qc_t
);
271 typedef int (prep_rq_fn
) (struct request_queue
*, struct request
*);
272 typedef void (unprep_rq_fn
) (struct request_queue
*, struct request
*);
275 typedef void (softirq_done_fn
)(struct request
*);
276 typedef int (dma_drain_needed_fn
)(struct request
*);
277 typedef int (lld_busy_fn
) (struct request_queue
*q
);
278 typedef int (bsg_job_fn
) (struct bsg_job
*);
279 typedef int (init_rq_fn
)(struct request_queue
*, struct request
*, gfp_t
);
280 typedef void (exit_rq_fn
)(struct request_queue
*, struct request
*);
282 enum blk_eh_timer_return
{
288 typedef enum blk_eh_timer_return (rq_timed_out_fn
)(struct request
*);
290 enum blk_queue_state
{
295 struct blk_queue_tag
{
296 struct request
**tag_index
; /* map of busy tags */
297 unsigned long *tag_map
; /* bit map of free/busy tags */
298 int max_depth
; /* what we will send to device */
299 int real_max_depth
; /* what the array can hold */
300 atomic_t refcnt
; /* map can be shared */
301 int alloc_policy
; /* tag allocation policy */
302 int next_tag
; /* next tag */
304 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
305 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
307 #define BLK_SCSI_MAX_CMDS (256)
308 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
311 * Zoned block device models (zoned limit).
313 enum blk_zoned_model
{
314 BLK_ZONED_NONE
, /* Regular block device */
315 BLK_ZONED_HA
, /* Host-aware zoned block device */
316 BLK_ZONED_HM
, /* Host-managed zoned block device */
319 struct queue_limits
{
320 unsigned long bounce_pfn
;
321 unsigned long seg_boundary_mask
;
322 unsigned long virt_boundary_mask
;
324 unsigned int max_hw_sectors
;
325 unsigned int max_dev_sectors
;
326 unsigned int chunk_sectors
;
327 unsigned int max_sectors
;
328 unsigned int max_segment_size
;
329 unsigned int physical_block_size
;
330 unsigned int alignment_offset
;
333 unsigned int max_discard_sectors
;
334 unsigned int max_hw_discard_sectors
;
335 unsigned int max_write_same_sectors
;
336 unsigned int max_write_zeroes_sectors
;
337 unsigned int discard_granularity
;
338 unsigned int discard_alignment
;
340 unsigned short logical_block_size
;
341 unsigned short max_segments
;
342 unsigned short max_integrity_segments
;
343 unsigned short max_discard_segments
;
345 unsigned char misaligned
;
346 unsigned char discard_misaligned
;
347 unsigned char cluster
;
348 unsigned char raid_partial_stripes_expensive
;
349 enum blk_zoned_model zoned
;
352 #ifdef CONFIG_BLK_DEV_ZONED
354 struct blk_zone_report_hdr
{
355 unsigned int nr_zones
;
359 extern int blkdev_report_zones(struct block_device
*bdev
,
360 sector_t sector
, struct blk_zone
*zones
,
361 unsigned int *nr_zones
, gfp_t gfp_mask
);
362 extern int blkdev_reset_zones(struct block_device
*bdev
, sector_t sectors
,
363 sector_t nr_sectors
, gfp_t gfp_mask
);
365 extern int blkdev_report_zones_ioctl(struct block_device
*bdev
, fmode_t mode
,
366 unsigned int cmd
, unsigned long arg
);
367 extern int blkdev_reset_zones_ioctl(struct block_device
*bdev
, fmode_t mode
,
368 unsigned int cmd
, unsigned long arg
);
370 #else /* CONFIG_BLK_DEV_ZONED */
372 static inline int blkdev_report_zones_ioctl(struct block_device
*bdev
,
373 fmode_t mode
, unsigned int cmd
,
379 static inline int blkdev_reset_zones_ioctl(struct block_device
*bdev
,
380 fmode_t mode
, unsigned int cmd
,
386 #endif /* CONFIG_BLK_DEV_ZONED */
388 struct request_queue
{
390 * Together with queue_head for cacheline sharing
392 struct list_head queue_head
;
393 struct request
*last_merge
;
394 struct elevator_queue
*elevator
;
395 int nr_rqs
[2]; /* # allocated [a]sync rqs */
396 int nr_rqs_elvpriv
; /* # allocated rqs w/ elvpriv */
398 atomic_t shared_hctx_restart
;
400 struct blk_queue_stats
*stats
;
404 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
405 * is used, root blkg allocates from @q->root_rl and all other
406 * blkgs from their own blkg->rl. Which one to use should be
407 * determined using bio_request_list().
409 struct request_list root_rl
;
411 request_fn_proc
*request_fn
;
412 make_request_fn
*make_request_fn
;
414 prep_rq_fn
*prep_rq_fn
;
415 unprep_rq_fn
*unprep_rq_fn
;
416 softirq_done_fn
*softirq_done_fn
;
417 rq_timed_out_fn
*rq_timed_out_fn
;
418 dma_drain_needed_fn
*dma_drain_needed
;
419 lld_busy_fn
*lld_busy_fn
;
420 /* Called just after a request is allocated */
421 init_rq_fn
*init_rq_fn
;
422 /* Called just before a request is freed */
423 exit_rq_fn
*exit_rq_fn
;
424 /* Called from inside blk_get_request() */
425 void (*initialize_rq_fn
)(struct request
*rq
);
427 const struct blk_mq_ops
*mq_ops
;
429 unsigned int *mq_map
;
432 struct blk_mq_ctx __percpu
*queue_ctx
;
433 unsigned int nr_queues
;
435 unsigned int queue_depth
;
437 /* hw dispatch queues */
438 struct blk_mq_hw_ctx
**queue_hw_ctx
;
439 unsigned int nr_hw_queues
;
442 * Dispatch queue sorting
445 struct request
*boundary_rq
;
448 * Delayed queue handling
450 struct delayed_work delay_work
;
452 struct backing_dev_info
*backing_dev_info
;
455 * The queue owner gets to use this for whatever they like.
456 * ll_rw_blk doesn't touch it.
461 * various queue flags, see QUEUE_* below
463 unsigned long queue_flags
;
466 * ida allocated id for this queue. Used to index queues from
472 * queue needs bounce pages for pages above this limit
477 * protects queue structures from reentrancy. ->__queue_lock should
478 * _never_ be used directly, it is queue private. always use
481 spinlock_t __queue_lock
;
482 spinlock_t
*queue_lock
;
492 struct kobject mq_kobj
;
494 #ifdef CONFIG_BLK_DEV_INTEGRITY
495 struct blk_integrity integrity
;
496 #endif /* CONFIG_BLK_DEV_INTEGRITY */
501 unsigned int nr_pending
;
507 unsigned long nr_requests
; /* Max # of requests */
508 unsigned int nr_congestion_on
;
509 unsigned int nr_congestion_off
;
510 unsigned int nr_batching
;
512 unsigned int dma_drain_size
;
513 void *dma_drain_buffer
;
514 unsigned int dma_pad_mask
;
515 unsigned int dma_alignment
;
517 struct blk_queue_tag
*queue_tags
;
518 struct list_head tag_busy_list
;
520 unsigned int nr_sorted
;
521 unsigned int in_flight
[2];
524 * Number of active block driver functions for which blk_drain_queue()
525 * must wait. Must be incremented around functions that unlock the
526 * queue_lock internally, e.g. scsi_request_fn().
528 unsigned int request_fn_active
;
530 unsigned int rq_timeout
;
533 struct blk_stat_callback
*poll_cb
;
534 struct blk_rq_stat poll_stat
[BLK_MQ_POLL_STATS_BKTS
];
536 struct timer_list timeout
;
537 struct work_struct timeout_work
;
538 struct list_head timeout_list
;
540 struct list_head icq_list
;
541 #ifdef CONFIG_BLK_CGROUP
542 DECLARE_BITMAP (blkcg_pols
, BLKCG_MAX_POLS
);
543 struct blkcg_gq
*root_blkg
;
544 struct list_head blkg_list
;
547 struct queue_limits limits
;
552 unsigned int sg_timeout
;
553 unsigned int sg_reserved_size
;
555 #ifdef CONFIG_BLK_DEV_IO_TRACE
556 struct blk_trace
*blk_trace
;
557 struct mutex blk_trace_mutex
;
560 * for flush operations
562 struct blk_flush_queue
*fq
;
564 struct list_head requeue_list
;
565 spinlock_t requeue_lock
;
566 struct delayed_work requeue_work
;
568 struct mutex sysfs_lock
;
571 atomic_t mq_freeze_depth
;
573 #if defined(CONFIG_BLK_DEV_BSG)
574 bsg_job_fn
*bsg_job_fn
;
575 struct bsg_class_device bsg_dev
;
578 #ifdef CONFIG_BLK_DEV_THROTTLING
580 struct throtl_data
*td
;
582 struct rcu_head rcu_head
;
583 wait_queue_head_t mq_freeze_wq
;
584 struct percpu_ref q_usage_counter
;
585 struct list_head all_q_node
;
587 struct blk_mq_tag_set
*tag_set
;
588 struct list_head tag_set_list
;
589 struct bio_set
*bio_split
;
591 #ifdef CONFIG_BLK_DEBUG_FS
592 struct dentry
*debugfs_dir
;
593 struct dentry
*sched_debugfs_dir
;
596 bool mq_sysfs_init_done
;
601 struct work_struct release_work
;
603 #define BLK_MAX_WRITE_HINTS 5
604 u64 write_hints
[BLK_MAX_WRITE_HINTS
];
607 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
608 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
609 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
610 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
611 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
612 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
613 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
614 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
615 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
616 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
617 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
618 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
619 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
620 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
621 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
622 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
623 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
624 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
625 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
626 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
627 #define QUEUE_FLAG_WC 20 /* Write back caching */
628 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
629 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
630 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
631 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
632 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
633 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
634 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
635 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
636 #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */
638 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
639 (1 << QUEUE_FLAG_SAME_COMP) | \
640 (1 << QUEUE_FLAG_ADD_RANDOM))
642 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
643 (1 << QUEUE_FLAG_SAME_COMP) | \
644 (1 << QUEUE_FLAG_POLL))
647 * @q->queue_lock is set while a queue is being initialized. Since we know
648 * that no other threads access the queue object before @q->queue_lock has
649 * been set, it is safe to manipulate queue flags without holding the
650 * queue_lock if @q->queue_lock == NULL. See also blk_alloc_queue_node() and
651 * blk_init_allocated_queue().
653 static inline void queue_lockdep_assert_held(struct request_queue
*q
)
656 lockdep_assert_held(q
->queue_lock
);
659 static inline void queue_flag_set_unlocked(unsigned int flag
,
660 struct request_queue
*q
)
662 __set_bit(flag
, &q
->queue_flags
);
665 static inline int queue_flag_test_and_clear(unsigned int flag
,
666 struct request_queue
*q
)
668 queue_lockdep_assert_held(q
);
670 if (test_bit(flag
, &q
->queue_flags
)) {
671 __clear_bit(flag
, &q
->queue_flags
);
678 static inline int queue_flag_test_and_set(unsigned int flag
,
679 struct request_queue
*q
)
681 queue_lockdep_assert_held(q
);
683 if (!test_bit(flag
, &q
->queue_flags
)) {
684 __set_bit(flag
, &q
->queue_flags
);
691 static inline void queue_flag_set(unsigned int flag
, struct request_queue
*q
)
693 queue_lockdep_assert_held(q
);
694 __set_bit(flag
, &q
->queue_flags
);
697 static inline void queue_flag_clear_unlocked(unsigned int flag
,
698 struct request_queue
*q
)
700 __clear_bit(flag
, &q
->queue_flags
);
703 static inline int queue_in_flight(struct request_queue
*q
)
705 return q
->in_flight
[0] + q
->in_flight
[1];
708 static inline void queue_flag_clear(unsigned int flag
, struct request_queue
*q
)
710 queue_lockdep_assert_held(q
);
711 __clear_bit(flag
, &q
->queue_flags
);
714 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
715 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
716 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
717 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
718 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
719 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
720 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
721 #define blk_queue_noxmerges(q) \
722 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
723 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
724 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
725 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
726 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
727 #define blk_queue_secure_erase(q) \
728 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
729 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
730 #define blk_queue_scsi_passthrough(q) \
731 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
733 #define blk_noretry_request(rq) \
734 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
735 REQ_FAILFAST_DRIVER))
736 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
737 #define blk_queue_preempt_only(q) \
738 test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
740 extern int blk_set_preempt_only(struct request_queue
*q
);
741 extern void blk_clear_preempt_only(struct request_queue
*q
);
743 static inline bool blk_account_rq(struct request
*rq
)
745 return (rq
->rq_flags
& RQF_STARTED
) && !blk_rq_is_passthrough(rq
);
748 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
749 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
750 /* rq->queuelist of dequeued request must be list_empty() */
751 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
753 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
755 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
758 * Driver can handle struct request, if it either has an old style
759 * request_fn defined, or is blk-mq based.
761 static inline bool queue_is_rq_based(struct request_queue
*q
)
763 return q
->request_fn
|| q
->mq_ops
;
766 static inline unsigned int blk_queue_cluster(struct request_queue
*q
)
768 return q
->limits
.cluster
;
771 static inline enum blk_zoned_model
772 blk_queue_zoned_model(struct request_queue
*q
)
774 return q
->limits
.zoned
;
777 static inline bool blk_queue_is_zoned(struct request_queue
*q
)
779 switch (blk_queue_zoned_model(q
)) {
788 static inline unsigned int blk_queue_zone_sectors(struct request_queue
*q
)
790 return blk_queue_is_zoned(q
) ? q
->limits
.chunk_sectors
: 0;
793 static inline bool rq_is_sync(struct request
*rq
)
795 return op_is_sync(rq
->cmd_flags
);
798 static inline bool blk_rl_full(struct request_list
*rl
, bool sync
)
800 unsigned int flag
= sync
? BLK_RL_SYNCFULL
: BLK_RL_ASYNCFULL
;
802 return rl
->flags
& flag
;
805 static inline void blk_set_rl_full(struct request_list
*rl
, bool sync
)
807 unsigned int flag
= sync
? BLK_RL_SYNCFULL
: BLK_RL_ASYNCFULL
;
812 static inline void blk_clear_rl_full(struct request_list
*rl
, bool sync
)
814 unsigned int flag
= sync
? BLK_RL_SYNCFULL
: BLK_RL_ASYNCFULL
;
819 static inline bool rq_mergeable(struct request
*rq
)
821 if (blk_rq_is_passthrough(rq
))
824 if (req_op(rq
) == REQ_OP_FLUSH
)
827 if (req_op(rq
) == REQ_OP_WRITE_ZEROES
)
830 if (rq
->cmd_flags
& REQ_NOMERGE_FLAGS
)
832 if (rq
->rq_flags
& RQF_NOMERGE_FLAGS
)
838 static inline bool blk_write_same_mergeable(struct bio
*a
, struct bio
*b
)
840 if (bio_page(a
) == bio_page(b
) &&
841 bio_offset(a
) == bio_offset(b
))
847 static inline unsigned int blk_queue_depth(struct request_queue
*q
)
850 return q
->queue_depth
;
852 return q
->nr_requests
;
856 * q->prep_rq_fn return values
859 BLKPREP_OK
, /* serve it */
860 BLKPREP_KILL
, /* fatal error, kill, return -EIO */
861 BLKPREP_DEFER
, /* leave on queue */
862 BLKPREP_INVALID
, /* invalid command, kill, return -EREMOTEIO */
865 extern unsigned long blk_max_low_pfn
, blk_max_pfn
;
868 * standard bounce addresses:
870 * BLK_BOUNCE_HIGH : bounce all highmem pages
871 * BLK_BOUNCE_ANY : don't bounce anything
872 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
875 #if BITS_PER_LONG == 32
876 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
878 #define BLK_BOUNCE_HIGH -1ULL
880 #define BLK_BOUNCE_ANY (-1ULL)
881 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
884 * default timeout for SG_IO if none specified
886 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
887 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
893 unsigned long offset
;
898 struct req_iterator
{
899 struct bvec_iter iter
;
903 /* This should not be used directly - use rq_for_each_segment */
904 #define for_each_bio(_bio) \
905 for (; _bio; _bio = _bio->bi_next)
906 #define __rq_for_each_bio(_bio, rq) \
908 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
910 #define rq_for_each_segment(bvl, _rq, _iter) \
911 __rq_for_each_bio(_iter.bio, _rq) \
912 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
914 #define rq_iter_last(bvec, _iter) \
915 (_iter.bio->bi_next == NULL && \
916 bio_iter_last(bvec, _iter.iter))
918 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
919 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
921 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
922 extern void rq_flush_dcache_pages(struct request
*rq
);
924 static inline void rq_flush_dcache_pages(struct request
*rq
)
929 extern int blk_register_queue(struct gendisk
*disk
);
930 extern void blk_unregister_queue(struct gendisk
*disk
);
931 extern blk_qc_t
generic_make_request(struct bio
*bio
);
932 extern blk_qc_t
direct_make_request(struct bio
*bio
);
933 extern void blk_rq_init(struct request_queue
*q
, struct request
*rq
);
934 extern void blk_init_request_from_bio(struct request
*req
, struct bio
*bio
);
935 extern void blk_put_request(struct request
*);
936 extern void __blk_put_request(struct request_queue
*, struct request
*);
937 extern struct request
*blk_get_request_flags(struct request_queue
*,
939 blk_mq_req_flags_t flags
);
940 extern struct request
*blk_get_request(struct request_queue
*, unsigned int op
,
942 extern void blk_requeue_request(struct request_queue
*, struct request
*);
943 extern int blk_lld_busy(struct request_queue
*q
);
944 extern int blk_rq_prep_clone(struct request
*rq
, struct request
*rq_src
,
945 struct bio_set
*bs
, gfp_t gfp_mask
,
946 int (*bio_ctr
)(struct bio
*, struct bio
*, void *),
948 extern void blk_rq_unprep_clone(struct request
*rq
);
949 extern blk_status_t
blk_insert_cloned_request(struct request_queue
*q
,
951 extern int blk_rq_append_bio(struct request
*rq
, struct bio
*bio
);
952 extern void blk_delay_queue(struct request_queue
*, unsigned long);
953 extern void blk_queue_split(struct request_queue
*, struct bio
**);
954 extern void blk_recount_segments(struct request_queue
*, struct bio
*);
955 extern int scsi_verify_blk_ioctl(struct block_device
*, unsigned int);
956 extern int scsi_cmd_blk_ioctl(struct block_device
*, fmode_t
,
957 unsigned int, void __user
*);
958 extern int scsi_cmd_ioctl(struct request_queue
*, struct gendisk
*, fmode_t
,
959 unsigned int, void __user
*);
960 extern int sg_scsi_ioctl(struct request_queue
*, struct gendisk
*, fmode_t
,
961 struct scsi_ioctl_command __user
*);
963 extern int blk_queue_enter(struct request_queue
*q
, blk_mq_req_flags_t flags
);
964 extern void blk_queue_exit(struct request_queue
*q
);
965 extern void blk_start_queue(struct request_queue
*q
);
966 extern void blk_start_queue_async(struct request_queue
*q
);
967 extern void blk_stop_queue(struct request_queue
*q
);
968 extern void blk_sync_queue(struct request_queue
*q
);
969 extern void __blk_stop_queue(struct request_queue
*q
);
970 extern void __blk_run_queue(struct request_queue
*q
);
971 extern void __blk_run_queue_uncond(struct request_queue
*q
);
972 extern void blk_run_queue(struct request_queue
*);
973 extern void blk_run_queue_async(struct request_queue
*q
);
974 extern int blk_rq_map_user(struct request_queue
*, struct request
*,
975 struct rq_map_data
*, void __user
*, unsigned long,
977 extern int blk_rq_unmap_user(struct bio
*);
978 extern int blk_rq_map_kern(struct request_queue
*, struct request
*, void *, unsigned int, gfp_t
);
979 extern int blk_rq_map_user_iov(struct request_queue
*, struct request
*,
980 struct rq_map_data
*, const struct iov_iter
*,
982 extern void blk_execute_rq(struct request_queue
*, struct gendisk
*,
983 struct request
*, int);
984 extern void blk_execute_rq_nowait(struct request_queue
*, struct gendisk
*,
985 struct request
*, int, rq_end_io_fn
*);
987 int blk_status_to_errno(blk_status_t status
);
988 blk_status_t
errno_to_blk_status(int errno
);
990 bool blk_poll(struct request_queue
*q
, blk_qc_t cookie
);
992 static inline struct request_queue
*bdev_get_queue(struct block_device
*bdev
)
994 return bdev
->bd_disk
->queue
; /* this is never NULL */
998 * blk_rq_pos() : the current sector
999 * blk_rq_bytes() : bytes left in the entire request
1000 * blk_rq_cur_bytes() : bytes left in the current segment
1001 * blk_rq_err_bytes() : bytes left till the next error boundary
1002 * blk_rq_sectors() : sectors left in the entire request
1003 * blk_rq_cur_sectors() : sectors left in the current segment
1005 static inline sector_t
blk_rq_pos(const struct request
*rq
)
1007 return rq
->__sector
;
1010 static inline unsigned int blk_rq_bytes(const struct request
*rq
)
1012 return rq
->__data_len
;
1015 static inline int blk_rq_cur_bytes(const struct request
*rq
)
1017 return rq
->bio
? bio_cur_bytes(rq
->bio
) : 0;
1020 extern unsigned int blk_rq_err_bytes(const struct request
*rq
);
1022 static inline unsigned int blk_rq_sectors(const struct request
*rq
)
1024 return blk_rq_bytes(rq
) >> 9;
1027 static inline unsigned int blk_rq_cur_sectors(const struct request
*rq
)
1029 return blk_rq_cur_bytes(rq
) >> 9;
1033 * Some commands like WRITE SAME have a payload or data transfer size which
1034 * is different from the size of the request. Any driver that supports such
1035 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1036 * calculate the data transfer size.
1038 static inline unsigned int blk_rq_payload_bytes(struct request
*rq
)
1040 if (rq
->rq_flags
& RQF_SPECIAL_PAYLOAD
)
1041 return rq
->special_vec
.bv_len
;
1042 return blk_rq_bytes(rq
);
1045 static inline unsigned int blk_queue_get_max_sectors(struct request_queue
*q
,
1048 if (unlikely(op
== REQ_OP_DISCARD
|| op
== REQ_OP_SECURE_ERASE
))
1049 return min(q
->limits
.max_discard_sectors
, UINT_MAX
>> 9);
1051 if (unlikely(op
== REQ_OP_WRITE_SAME
))
1052 return q
->limits
.max_write_same_sectors
;
1054 if (unlikely(op
== REQ_OP_WRITE_ZEROES
))
1055 return q
->limits
.max_write_zeroes_sectors
;
1057 return q
->limits
.max_sectors
;
1061 * Return maximum size of a request at given offset. Only valid for
1062 * file system requests.
1064 static inline unsigned int blk_max_size_offset(struct request_queue
*q
,
1067 if (!q
->limits
.chunk_sectors
)
1068 return q
->limits
.max_sectors
;
1070 return q
->limits
.chunk_sectors
-
1071 (offset
& (q
->limits
.chunk_sectors
- 1));
1074 static inline unsigned int blk_rq_get_max_sectors(struct request
*rq
,
1077 struct request_queue
*q
= rq
->q
;
1079 if (blk_rq_is_passthrough(rq
))
1080 return q
->limits
.max_hw_sectors
;
1082 if (!q
->limits
.chunk_sectors
||
1083 req_op(rq
) == REQ_OP_DISCARD
||
1084 req_op(rq
) == REQ_OP_SECURE_ERASE
)
1085 return blk_queue_get_max_sectors(q
, req_op(rq
));
1087 return min(blk_max_size_offset(q
, offset
),
1088 blk_queue_get_max_sectors(q
, req_op(rq
)));
1091 static inline unsigned int blk_rq_count_bios(struct request
*rq
)
1093 unsigned int nr_bios
= 0;
1096 __rq_for_each_bio(bio
, rq
)
1103 * Request issue related functions.
1105 extern struct request
*blk_peek_request(struct request_queue
*q
);
1106 extern void blk_start_request(struct request
*rq
);
1107 extern struct request
*blk_fetch_request(struct request_queue
*q
);
1109 void blk_steal_bios(struct bio_list
*list
, struct request
*rq
);
1112 * Request completion related functions.
1114 * blk_update_request() completes given number of bytes and updates
1115 * the request without completing it.
1117 * blk_end_request() and friends. __blk_end_request() must be called
1118 * with the request queue spinlock acquired.
1120 * Several drivers define their own end_request and call
1121 * blk_end_request() for parts of the original function.
1122 * This prevents code duplication in drivers.
1124 extern bool blk_update_request(struct request
*rq
, blk_status_t error
,
1125 unsigned int nr_bytes
);
1126 extern void blk_finish_request(struct request
*rq
, blk_status_t error
);
1127 extern bool blk_end_request(struct request
*rq
, blk_status_t error
,
1128 unsigned int nr_bytes
);
1129 extern void blk_end_request_all(struct request
*rq
, blk_status_t error
);
1130 extern bool __blk_end_request(struct request
*rq
, blk_status_t error
,
1131 unsigned int nr_bytes
);
1132 extern void __blk_end_request_all(struct request
*rq
, blk_status_t error
);
1133 extern bool __blk_end_request_cur(struct request
*rq
, blk_status_t error
);
1135 extern void blk_complete_request(struct request
*);
1136 extern void __blk_complete_request(struct request
*);
1137 extern void blk_abort_request(struct request
*);
1138 extern void blk_unprep_request(struct request
*);
1141 * Access functions for manipulating queue properties
1143 extern struct request_queue
*blk_init_queue_node(request_fn_proc
*rfn
,
1144 spinlock_t
*lock
, int node_id
);
1145 extern struct request_queue
*blk_init_queue(request_fn_proc
*, spinlock_t
*);
1146 extern int blk_init_allocated_queue(struct request_queue
*);
1147 extern void blk_cleanup_queue(struct request_queue
*);
1148 extern void blk_queue_make_request(struct request_queue
*, make_request_fn
*);
1149 extern void blk_queue_bounce_limit(struct request_queue
*, u64
);
1150 extern void blk_queue_max_hw_sectors(struct request_queue
*, unsigned int);
1151 extern void blk_queue_chunk_sectors(struct request_queue
*, unsigned int);
1152 extern void blk_queue_max_segments(struct request_queue
*, unsigned short);
1153 extern void blk_queue_max_discard_segments(struct request_queue
*,
1155 extern void blk_queue_max_segment_size(struct request_queue
*, unsigned int);
1156 extern void blk_queue_max_discard_sectors(struct request_queue
*q
,
1157 unsigned int max_discard_sectors
);
1158 extern void blk_queue_max_write_same_sectors(struct request_queue
*q
,
1159 unsigned int max_write_same_sectors
);
1160 extern void blk_queue_max_write_zeroes_sectors(struct request_queue
*q
,
1161 unsigned int max_write_same_sectors
);
1162 extern void blk_queue_logical_block_size(struct request_queue
*, unsigned short);
1163 extern void blk_queue_physical_block_size(struct request_queue
*, unsigned int);
1164 extern void blk_queue_alignment_offset(struct request_queue
*q
,
1165 unsigned int alignment
);
1166 extern void blk_limits_io_min(struct queue_limits
*limits
, unsigned int min
);
1167 extern void blk_queue_io_min(struct request_queue
*q
, unsigned int min
);
1168 extern void blk_limits_io_opt(struct queue_limits
*limits
, unsigned int opt
);
1169 extern void blk_queue_io_opt(struct request_queue
*q
, unsigned int opt
);
1170 extern void blk_set_queue_depth(struct request_queue
*q
, unsigned int depth
);
1171 extern void blk_set_default_limits(struct queue_limits
*lim
);
1172 extern void blk_set_stacking_limits(struct queue_limits
*lim
);
1173 extern int blk_stack_limits(struct queue_limits
*t
, struct queue_limits
*b
,
1175 extern int bdev_stack_limits(struct queue_limits
*t
, struct block_device
*bdev
,
1177 extern void disk_stack_limits(struct gendisk
*disk
, struct block_device
*bdev
,
1179 extern void blk_queue_stack_limits(struct request_queue
*t
, struct request_queue
*b
);
1180 extern void blk_queue_dma_pad(struct request_queue
*, unsigned int);
1181 extern void blk_queue_update_dma_pad(struct request_queue
*, unsigned int);
1182 extern int blk_queue_dma_drain(struct request_queue
*q
,
1183 dma_drain_needed_fn
*dma_drain_needed
,
1184 void *buf
, unsigned int size
);
1185 extern void blk_queue_lld_busy(struct request_queue
*q
, lld_busy_fn
*fn
);
1186 extern void blk_queue_segment_boundary(struct request_queue
*, unsigned long);
1187 extern void blk_queue_virt_boundary(struct request_queue
*, unsigned long);
1188 extern void blk_queue_prep_rq(struct request_queue
*, prep_rq_fn
*pfn
);
1189 extern void blk_queue_unprep_rq(struct request_queue
*, unprep_rq_fn
*ufn
);
1190 extern void blk_queue_dma_alignment(struct request_queue
*, int);
1191 extern void blk_queue_update_dma_alignment(struct request_queue
*, int);
1192 extern void blk_queue_softirq_done(struct request_queue
*, softirq_done_fn
*);
1193 extern void blk_queue_rq_timed_out(struct request_queue
*, rq_timed_out_fn
*);
1194 extern void blk_queue_rq_timeout(struct request_queue
*, unsigned int);
1195 extern void blk_queue_flush_queueable(struct request_queue
*q
, bool queueable
);
1196 extern void blk_queue_write_cache(struct request_queue
*q
, bool enabled
, bool fua
);
1199 * Number of physical segments as sent to the device.
1201 * Normally this is the number of discontiguous data segments sent by the
1202 * submitter. But for data-less command like discard we might have no
1203 * actual data segments submitted, but the driver might have to add it's
1204 * own special payload. In that case we still return 1 here so that this
1205 * special payload will be mapped.
1207 static inline unsigned short blk_rq_nr_phys_segments(struct request
*rq
)
1209 if (rq
->rq_flags
& RQF_SPECIAL_PAYLOAD
)
1211 return rq
->nr_phys_segments
;
1215 * Number of discard segments (or ranges) the driver needs to fill in.
1216 * Each discard bio merged into a request is counted as one segment.
1218 static inline unsigned short blk_rq_nr_discard_segments(struct request
*rq
)
1220 return max_t(unsigned short, rq
->nr_phys_segments
, 1);
1223 extern int blk_rq_map_sg(struct request_queue
*, struct request
*, struct scatterlist
*);
1224 extern void blk_dump_rq_flags(struct request
*, char *);
1225 extern long nr_blockdev_pages(void);
1227 bool __must_check
blk_get_queue(struct request_queue
*);
1228 struct request_queue
*blk_alloc_queue(gfp_t
);
1229 struct request_queue
*blk_alloc_queue_node(gfp_t
, int);
1230 extern void blk_put_queue(struct request_queue
*);
1231 extern void blk_set_queue_dying(struct request_queue
*);
1234 * block layer runtime pm functions
1237 extern void blk_pm_runtime_init(struct request_queue
*q
, struct device
*dev
);
1238 extern int blk_pre_runtime_suspend(struct request_queue
*q
);
1239 extern void blk_post_runtime_suspend(struct request_queue
*q
, int err
);
1240 extern void blk_pre_runtime_resume(struct request_queue
*q
);
1241 extern void blk_post_runtime_resume(struct request_queue
*q
, int err
);
1242 extern void blk_set_runtime_active(struct request_queue
*q
);
1244 static inline void blk_pm_runtime_init(struct request_queue
*q
,
1245 struct device
*dev
) {}
1246 static inline int blk_pre_runtime_suspend(struct request_queue
*q
)
1250 static inline void blk_post_runtime_suspend(struct request_queue
*q
, int err
) {}
1251 static inline void blk_pre_runtime_resume(struct request_queue
*q
) {}
1252 static inline void blk_post_runtime_resume(struct request_queue
*q
, int err
) {}
1253 static inline void blk_set_runtime_active(struct request_queue
*q
) {}
1257 * blk_plug permits building a queue of related requests by holding the I/O
1258 * fragments for a short period. This allows merging of sequential requests
1259 * into single larger request. As the requests are moved from a per-task list to
1260 * the device's request_queue in a batch, this results in improved scalability
1261 * as the lock contention for request_queue lock is reduced.
1263 * It is ok not to disable preemption when adding the request to the plug list
1264 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1265 * the plug list when the task sleeps by itself. For details, please see
1266 * schedule() where blk_schedule_flush_plug() is called.
1269 struct list_head list
; /* requests */
1270 struct list_head mq_list
; /* blk-mq requests */
1271 struct list_head cb_list
; /* md requires an unplug callback */
1273 #define BLK_MAX_REQUEST_COUNT 16
1274 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1277 typedef void (*blk_plug_cb_fn
)(struct blk_plug_cb
*, bool);
1278 struct blk_plug_cb
{
1279 struct list_head list
;
1280 blk_plug_cb_fn callback
;
1283 extern struct blk_plug_cb
*blk_check_plugged(blk_plug_cb_fn unplug
,
1284 void *data
, int size
);
1285 extern void blk_start_plug(struct blk_plug
*);
1286 extern void blk_finish_plug(struct blk_plug
*);
1287 extern void blk_flush_plug_list(struct blk_plug
*, bool);
1289 static inline void blk_flush_plug(struct task_struct
*tsk
)
1291 struct blk_plug
*plug
= tsk
->plug
;
1294 blk_flush_plug_list(plug
, false);
1297 static inline void blk_schedule_flush_plug(struct task_struct
*tsk
)
1299 struct blk_plug
*plug
= tsk
->plug
;
1302 blk_flush_plug_list(plug
, true);
1305 static inline bool blk_needs_flush_plug(struct task_struct
*tsk
)
1307 struct blk_plug
*plug
= tsk
->plug
;
1310 (!list_empty(&plug
->list
) ||
1311 !list_empty(&plug
->mq_list
) ||
1312 !list_empty(&plug
->cb_list
));
1318 extern int blk_queue_start_tag(struct request_queue
*, struct request
*);
1319 extern struct request
*blk_queue_find_tag(struct request_queue
*, int);
1320 extern void blk_queue_end_tag(struct request_queue
*, struct request
*);
1321 extern int blk_queue_init_tags(struct request_queue
*, int, struct blk_queue_tag
*, int);
1322 extern void blk_queue_free_tags(struct request_queue
*);
1323 extern int blk_queue_resize_tags(struct request_queue
*, int);
1324 extern void blk_queue_invalidate_tags(struct request_queue
*);
1325 extern struct blk_queue_tag
*blk_init_tags(int, int);
1326 extern void blk_free_tags(struct blk_queue_tag
*);
1328 static inline struct request
*blk_map_queue_find_tag(struct blk_queue_tag
*bqt
,
1331 if (unlikely(bqt
== NULL
|| tag
>= bqt
->real_max_depth
))
1333 return bqt
->tag_index
[tag
];
1336 extern int blkdev_issue_flush(struct block_device
*, gfp_t
, sector_t
*);
1337 extern int blkdev_issue_write_same(struct block_device
*bdev
, sector_t sector
,
1338 sector_t nr_sects
, gfp_t gfp_mask
, struct page
*page
);
1340 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1342 extern int blkdev_issue_discard(struct block_device
*bdev
, sector_t sector
,
1343 sector_t nr_sects
, gfp_t gfp_mask
, unsigned long flags
);
1344 extern int __blkdev_issue_discard(struct block_device
*bdev
, sector_t sector
,
1345 sector_t nr_sects
, gfp_t gfp_mask
, int flags
,
1348 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1349 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1351 extern int __blkdev_issue_zeroout(struct block_device
*bdev
, sector_t sector
,
1352 sector_t nr_sects
, gfp_t gfp_mask
, struct bio
**biop
,
1354 extern int blkdev_issue_zeroout(struct block_device
*bdev
, sector_t sector
,
1355 sector_t nr_sects
, gfp_t gfp_mask
, unsigned flags
);
1357 static inline int sb_issue_discard(struct super_block
*sb
, sector_t block
,
1358 sector_t nr_blocks
, gfp_t gfp_mask
, unsigned long flags
)
1360 return blkdev_issue_discard(sb
->s_bdev
, block
<< (sb
->s_blocksize_bits
- 9),
1361 nr_blocks
<< (sb
->s_blocksize_bits
- 9),
1364 static inline int sb_issue_zeroout(struct super_block
*sb
, sector_t block
,
1365 sector_t nr_blocks
, gfp_t gfp_mask
)
1367 return blkdev_issue_zeroout(sb
->s_bdev
,
1368 block
<< (sb
->s_blocksize_bits
- 9),
1369 nr_blocks
<< (sb
->s_blocksize_bits
- 9),
1373 extern int blk_verify_command(unsigned char *cmd
, fmode_t mode
);
1375 enum blk_default_limits
{
1376 BLK_MAX_SEGMENTS
= 128,
1377 BLK_SAFE_MAX_SECTORS
= 255,
1378 BLK_DEF_MAX_SECTORS
= 2560,
1379 BLK_MAX_SEGMENT_SIZE
= 65536,
1380 BLK_SEG_BOUNDARY_MASK
= 0xFFFFFFFFUL
,
1383 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
1385 static inline unsigned long queue_segment_boundary(struct request_queue
*q
)
1387 return q
->limits
.seg_boundary_mask
;
1390 static inline unsigned long queue_virt_boundary(struct request_queue
*q
)
1392 return q
->limits
.virt_boundary_mask
;
1395 static inline unsigned int queue_max_sectors(struct request_queue
*q
)
1397 return q
->limits
.max_sectors
;
1400 static inline unsigned int queue_max_hw_sectors(struct request_queue
*q
)
1402 return q
->limits
.max_hw_sectors
;
1405 static inline unsigned short queue_max_segments(struct request_queue
*q
)
1407 return q
->limits
.max_segments
;
1410 static inline unsigned short queue_max_discard_segments(struct request_queue
*q
)
1412 return q
->limits
.max_discard_segments
;
1415 static inline unsigned int queue_max_segment_size(struct request_queue
*q
)
1417 return q
->limits
.max_segment_size
;
1420 static inline unsigned short queue_logical_block_size(struct request_queue
*q
)
1424 if (q
&& q
->limits
.logical_block_size
)
1425 retval
= q
->limits
.logical_block_size
;
1430 static inline unsigned short bdev_logical_block_size(struct block_device
*bdev
)
1432 return queue_logical_block_size(bdev_get_queue(bdev
));
1435 static inline unsigned int queue_physical_block_size(struct request_queue
*q
)
1437 return q
->limits
.physical_block_size
;
1440 static inline unsigned int bdev_physical_block_size(struct block_device
*bdev
)
1442 return queue_physical_block_size(bdev_get_queue(bdev
));
1445 static inline unsigned int queue_io_min(struct request_queue
*q
)
1447 return q
->limits
.io_min
;
1450 static inline int bdev_io_min(struct block_device
*bdev
)
1452 return queue_io_min(bdev_get_queue(bdev
));
1455 static inline unsigned int queue_io_opt(struct request_queue
*q
)
1457 return q
->limits
.io_opt
;
1460 static inline int bdev_io_opt(struct block_device
*bdev
)
1462 return queue_io_opt(bdev_get_queue(bdev
));
1465 static inline int queue_alignment_offset(struct request_queue
*q
)
1467 if (q
->limits
.misaligned
)
1470 return q
->limits
.alignment_offset
;
1473 static inline int queue_limit_alignment_offset(struct queue_limits
*lim
, sector_t sector
)
1475 unsigned int granularity
= max(lim
->physical_block_size
, lim
->io_min
);
1476 unsigned int alignment
= sector_div(sector
, granularity
>> 9) << 9;
1478 return (granularity
+ lim
->alignment_offset
- alignment
) % granularity
;
1481 static inline int bdev_alignment_offset(struct block_device
*bdev
)
1483 struct request_queue
*q
= bdev_get_queue(bdev
);
1485 if (q
->limits
.misaligned
)
1488 if (bdev
!= bdev
->bd_contains
)
1489 return bdev
->bd_part
->alignment_offset
;
1491 return q
->limits
.alignment_offset
;
1494 static inline int queue_discard_alignment(struct request_queue
*q
)
1496 if (q
->limits
.discard_misaligned
)
1499 return q
->limits
.discard_alignment
;
1502 static inline int queue_limit_discard_alignment(struct queue_limits
*lim
, sector_t sector
)
1504 unsigned int alignment
, granularity
, offset
;
1506 if (!lim
->max_discard_sectors
)
1509 /* Why are these in bytes, not sectors? */
1510 alignment
= lim
->discard_alignment
>> 9;
1511 granularity
= lim
->discard_granularity
>> 9;
1515 /* Offset of the partition start in 'granularity' sectors */
1516 offset
= sector_div(sector
, granularity
);
1518 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1519 offset
= (granularity
+ alignment
- offset
) % granularity
;
1521 /* Turn it back into bytes, gaah */
1525 static inline int bdev_discard_alignment(struct block_device
*bdev
)
1527 struct request_queue
*q
= bdev_get_queue(bdev
);
1529 if (bdev
!= bdev
->bd_contains
)
1530 return bdev
->bd_part
->discard_alignment
;
1532 return q
->limits
.discard_alignment
;
1535 static inline unsigned int bdev_write_same(struct block_device
*bdev
)
1537 struct request_queue
*q
= bdev_get_queue(bdev
);
1540 return q
->limits
.max_write_same_sectors
;
1545 static inline unsigned int bdev_write_zeroes_sectors(struct block_device
*bdev
)
1547 struct request_queue
*q
= bdev_get_queue(bdev
);
1550 return q
->limits
.max_write_zeroes_sectors
;
1555 static inline enum blk_zoned_model
bdev_zoned_model(struct block_device
*bdev
)
1557 struct request_queue
*q
= bdev_get_queue(bdev
);
1560 return blk_queue_zoned_model(q
);
1562 return BLK_ZONED_NONE
;
1565 static inline bool bdev_is_zoned(struct block_device
*bdev
)
1567 struct request_queue
*q
= bdev_get_queue(bdev
);
1570 return blk_queue_is_zoned(q
);
1575 static inline unsigned int bdev_zone_sectors(struct block_device
*bdev
)
1577 struct request_queue
*q
= bdev_get_queue(bdev
);
1580 return blk_queue_zone_sectors(q
);
1585 static inline int queue_dma_alignment(struct request_queue
*q
)
1587 return q
? q
->dma_alignment
: 511;
1590 static inline int blk_rq_aligned(struct request_queue
*q
, unsigned long addr
,
1593 unsigned int alignment
= queue_dma_alignment(q
) | q
->dma_pad_mask
;
1594 return !(addr
& alignment
) && !(len
& alignment
);
1597 /* assumes size > 256 */
1598 static inline unsigned int blksize_bits(unsigned int size
)
1600 unsigned int bits
= 8;
1604 } while (size
> 256);
1608 static inline unsigned int block_size(struct block_device
*bdev
)
1610 return bdev
->bd_block_size
;
1613 static inline bool queue_flush_queueable(struct request_queue
*q
)
1615 return !test_bit(QUEUE_FLAG_FLUSH_NQ
, &q
->queue_flags
);
1618 typedef struct {struct page
*v
;} Sector
;
1620 unsigned char *read_dev_sector(struct block_device
*, sector_t
, Sector
*);
1622 static inline void put_dev_sector(Sector p
)
1627 static inline bool __bvec_gap_to_prev(struct request_queue
*q
,
1628 struct bio_vec
*bprv
, unsigned int offset
)
1631 ((bprv
->bv_offset
+ bprv
->bv_len
) & queue_virt_boundary(q
));
1635 * Check if adding a bio_vec after bprv with offset would create a gap in
1636 * the SG list. Most drivers don't care about this, but some do.
1638 static inline bool bvec_gap_to_prev(struct request_queue
*q
,
1639 struct bio_vec
*bprv
, unsigned int offset
)
1641 if (!queue_virt_boundary(q
))
1643 return __bvec_gap_to_prev(q
, bprv
, offset
);
1647 * Check if the two bvecs from two bios can be merged to one segment.
1648 * If yes, no need to check gap between the two bios since the 1st bio
1649 * and the 1st bvec in the 2nd bio can be handled in one segment.
1651 static inline bool bios_segs_mergeable(struct request_queue
*q
,
1652 struct bio
*prev
, struct bio_vec
*prev_last_bv
,
1653 struct bio_vec
*next_first_bv
)
1655 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv
, next_first_bv
))
1657 if (!BIOVEC_SEG_BOUNDARY(q
, prev_last_bv
, next_first_bv
))
1659 if (prev
->bi_seg_back_size
+ next_first_bv
->bv_len
>
1660 queue_max_segment_size(q
))
1665 static inline bool bio_will_gap(struct request_queue
*q
,
1666 struct request
*prev_rq
,
1670 if (bio_has_data(prev
) && queue_virt_boundary(q
)) {
1671 struct bio_vec pb
, nb
;
1674 * don't merge if the 1st bio starts with non-zero
1675 * offset, otherwise it is quite difficult to respect
1676 * sg gap limit. We work hard to merge a huge number of small
1677 * single bios in case of mkfs.
1680 bio_get_first_bvec(prev_rq
->bio
, &pb
);
1682 bio_get_first_bvec(prev
, &pb
);
1687 * We don't need to worry about the situation that the
1688 * merged segment ends in unaligned virt boundary:
1690 * - if 'pb' ends aligned, the merged segment ends aligned
1691 * - if 'pb' ends unaligned, the next bio must include
1692 * one single bvec of 'nb', otherwise the 'nb' can't
1695 bio_get_last_bvec(prev
, &pb
);
1696 bio_get_first_bvec(next
, &nb
);
1698 if (!bios_segs_mergeable(q
, prev
, &pb
, &nb
))
1699 return __bvec_gap_to_prev(q
, &pb
, nb
.bv_offset
);
1705 static inline bool req_gap_back_merge(struct request
*req
, struct bio
*bio
)
1707 return bio_will_gap(req
->q
, req
, req
->biotail
, bio
);
1710 static inline bool req_gap_front_merge(struct request
*req
, struct bio
*bio
)
1712 return bio_will_gap(req
->q
, NULL
, bio
, req
->bio
);
1715 int kblockd_schedule_work(struct work_struct
*work
);
1716 int kblockd_schedule_work_on(int cpu
, struct work_struct
*work
);
1717 int kblockd_schedule_delayed_work(struct delayed_work
*dwork
, unsigned long delay
);
1718 int kblockd_schedule_delayed_work_on(int cpu
, struct delayed_work
*dwork
, unsigned long delay
);
1719 int kblockd_mod_delayed_work_on(int cpu
, struct delayed_work
*dwork
, unsigned long delay
);
1721 #ifdef CONFIG_BLK_CGROUP
1723 * This should not be using sched_clock(). A real patch is in progress
1724 * to fix this up, until that is in place we need to disable preemption
1725 * around sched_clock() in this function and set_io_start_time_ns().
1727 static inline void set_start_time_ns(struct request
*req
)
1730 req
->start_time_ns
= sched_clock();
1734 static inline void set_io_start_time_ns(struct request
*req
)
1737 req
->io_start_time_ns
= sched_clock();
1741 static inline uint64_t rq_start_time_ns(struct request
*req
)
1743 return req
->start_time_ns
;
1746 static inline uint64_t rq_io_start_time_ns(struct request
*req
)
1748 return req
->io_start_time_ns
;
1751 static inline void set_start_time_ns(struct request
*req
) {}
1752 static inline void set_io_start_time_ns(struct request
*req
) {}
1753 static inline uint64_t rq_start_time_ns(struct request
*req
)
1757 static inline uint64_t rq_io_start_time_ns(struct request
*req
)
1763 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1764 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1765 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1766 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1768 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1770 enum blk_integrity_flags
{
1771 BLK_INTEGRITY_VERIFY
= 1 << 0,
1772 BLK_INTEGRITY_GENERATE
= 1 << 1,
1773 BLK_INTEGRITY_DEVICE_CAPABLE
= 1 << 2,
1774 BLK_INTEGRITY_IP_CHECKSUM
= 1 << 3,
1777 struct blk_integrity_iter
{
1781 unsigned int data_size
;
1782 unsigned short interval
;
1783 const char *disk_name
;
1786 typedef blk_status_t (integrity_processing_fn
) (struct blk_integrity_iter
*);
1788 struct blk_integrity_profile
{
1789 integrity_processing_fn
*generate_fn
;
1790 integrity_processing_fn
*verify_fn
;
1794 extern void blk_integrity_register(struct gendisk
*, struct blk_integrity
*);
1795 extern void blk_integrity_unregister(struct gendisk
*);
1796 extern int blk_integrity_compare(struct gendisk
*, struct gendisk
*);
1797 extern int blk_rq_map_integrity_sg(struct request_queue
*, struct bio
*,
1798 struct scatterlist
*);
1799 extern int blk_rq_count_integrity_sg(struct request_queue
*, struct bio
*);
1800 extern bool blk_integrity_merge_rq(struct request_queue
*, struct request
*,
1802 extern bool blk_integrity_merge_bio(struct request_queue
*, struct request
*,
1805 static inline struct blk_integrity
*blk_get_integrity(struct gendisk
*disk
)
1807 struct blk_integrity
*bi
= &disk
->queue
->integrity
;
1816 struct blk_integrity
*bdev_get_integrity(struct block_device
*bdev
)
1818 return blk_get_integrity(bdev
->bd_disk
);
1821 static inline bool blk_integrity_rq(struct request
*rq
)
1823 return rq
->cmd_flags
& REQ_INTEGRITY
;
1826 static inline void blk_queue_max_integrity_segments(struct request_queue
*q
,
1829 q
->limits
.max_integrity_segments
= segs
;
1832 static inline unsigned short
1833 queue_max_integrity_segments(struct request_queue
*q
)
1835 return q
->limits
.max_integrity_segments
;
1838 static inline bool integrity_req_gap_back_merge(struct request
*req
,
1841 struct bio_integrity_payload
*bip
= bio_integrity(req
->bio
);
1842 struct bio_integrity_payload
*bip_next
= bio_integrity(next
);
1844 return bvec_gap_to_prev(req
->q
, &bip
->bip_vec
[bip
->bip_vcnt
- 1],
1845 bip_next
->bip_vec
[0].bv_offset
);
1848 static inline bool integrity_req_gap_front_merge(struct request
*req
,
1851 struct bio_integrity_payload
*bip
= bio_integrity(bio
);
1852 struct bio_integrity_payload
*bip_next
= bio_integrity(req
->bio
);
1854 return bvec_gap_to_prev(req
->q
, &bip
->bip_vec
[bip
->bip_vcnt
- 1],
1855 bip_next
->bip_vec
[0].bv_offset
);
1858 #else /* CONFIG_BLK_DEV_INTEGRITY */
1861 struct block_device
;
1863 struct blk_integrity
;
1865 static inline int blk_integrity_rq(struct request
*rq
)
1869 static inline int blk_rq_count_integrity_sg(struct request_queue
*q
,
1874 static inline int blk_rq_map_integrity_sg(struct request_queue
*q
,
1876 struct scatterlist
*s
)
1880 static inline struct blk_integrity
*bdev_get_integrity(struct block_device
*b
)
1884 static inline struct blk_integrity
*blk_get_integrity(struct gendisk
*disk
)
1888 static inline int blk_integrity_compare(struct gendisk
*a
, struct gendisk
*b
)
1892 static inline void blk_integrity_register(struct gendisk
*d
,
1893 struct blk_integrity
*b
)
1896 static inline void blk_integrity_unregister(struct gendisk
*d
)
1899 static inline void blk_queue_max_integrity_segments(struct request_queue
*q
,
1903 static inline unsigned short queue_max_integrity_segments(struct request_queue
*q
)
1907 static inline bool blk_integrity_merge_rq(struct request_queue
*rq
,
1913 static inline bool blk_integrity_merge_bio(struct request_queue
*rq
,
1920 static inline bool integrity_req_gap_back_merge(struct request
*req
,
1925 static inline bool integrity_req_gap_front_merge(struct request
*req
,
1931 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1933 struct block_device_operations
{
1934 int (*open
) (struct block_device
*, fmode_t
);
1935 void (*release
) (struct gendisk
*, fmode_t
);
1936 int (*rw_page
)(struct block_device
*, sector_t
, struct page
*, bool);
1937 int (*ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1938 int (*compat_ioctl
) (struct block_device
*, fmode_t
, unsigned, unsigned long);
1939 unsigned int (*check_events
) (struct gendisk
*disk
,
1940 unsigned int clearing
);
1941 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1942 int (*media_changed
) (struct gendisk
*);
1943 void (*unlock_native_capacity
) (struct gendisk
*);
1944 int (*revalidate_disk
) (struct gendisk
*);
1945 int (*getgeo
)(struct block_device
*, struct hd_geometry
*);
1946 /* this callback is with swap_lock and sometimes page table lock held */
1947 void (*swap_slot_free_notify
) (struct block_device
*, unsigned long);
1948 struct module
*owner
;
1949 const struct pr_ops
*pr_ops
;
1952 extern int __blkdev_driver_ioctl(struct block_device
*, fmode_t
, unsigned int,
1954 extern int bdev_read_page(struct block_device
*, sector_t
, struct page
*);
1955 extern int bdev_write_page(struct block_device
*, sector_t
, struct page
*,
1956 struct writeback_control
*);
1957 #else /* CONFIG_BLOCK */
1959 struct block_device
;
1962 * stubs for when the block layer is configured out
1964 #define buffer_heads_over_limit 0
1966 static inline long nr_blockdev_pages(void)
1974 static inline void blk_start_plug(struct blk_plug
*plug
)
1978 static inline void blk_finish_plug(struct blk_plug
*plug
)
1982 static inline void blk_flush_plug(struct task_struct
*task
)
1986 static inline void blk_schedule_flush_plug(struct task_struct
*task
)
1991 static inline bool blk_needs_flush_plug(struct task_struct
*tsk
)
1996 static inline int blkdev_issue_flush(struct block_device
*bdev
, gfp_t gfp_mask
,
1997 sector_t
*error_sector
)
2002 #endif /* CONFIG_BLOCK */