1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
30 #include <linux/seqlock.h>
31 #include <linux/u64_stats_sync.h>
34 struct scsi_ioctl_command;
37 struct elevator_queue;
43 struct blk_flush_queue;
46 struct blk_queue_stats;
47 struct blk_stat_callback;
49 #define BLKDEV_MIN_RQ 4
50 #define BLKDEV_MAX_RQ 128 /* Default maximum */
52 /* Must be consistent with blk_mq_poll_stats_bkt() */
53 #define BLK_MQ_POLL_STATS_BKTS 16
56 * Maximum number of blkcg policies allowed to be registered concurrently.
57 * Defined here to simplify include dependency.
59 #define BLKCG_MAX_POLS 3
61 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
63 #define BLK_RL_SYNCFULL (1U << 0)
64 #define BLK_RL_ASYNCFULL (1U << 1)
67 struct request_queue *q; /* the queue this rl belongs to */
68 #ifdef CONFIG_BLK_CGROUP
69 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
72 * count[], starved[], and wait[] are indexed by
73 * BLK_RW_SYNC/BLK_RW_ASYNC
78 wait_queue_head_t wait[2];
84 typedef __u32 __bitwise req_flags_t;
86 /* elevator knows about this request */
87 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
88 /* drive already may have started this one */
89 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
90 /* uses tagged queueing */
91 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
92 /* may not be passed by ioscheduler */
93 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
94 /* request for flush sequence */
95 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
96 /* merge of different types, fail separately */
97 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
98 /* track inflight for MQ */
99 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
100 /* don't call prep for this one */
101 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
102 /* set for "ide_preempt" requests and also for requests for which the SCSI
103 "quiesce" state must be ignored. */
104 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
105 /* contains copies of user pages */
106 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
107 /* vaguely specified driver internal error. Ignored by the block layer */
108 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
109 /* don't warn about errors */
110 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
111 /* elevator private data attached */
112 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
113 /* account I/O stat */
114 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
115 /* request came from our alloc pool */
116 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
117 /* runtime pm request */
118 #define RQF_PM ((__force req_flags_t)(1 << 15))
119 /* on IO scheduler merge hash */
120 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
121 /* IO stats tracking on */
122 #define RQF_STATS ((__force req_flags_t)(1 << 17))
123 /* Look at ->special_vec for the actual data payload instead of the
125 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
126 /* The per-zone write lock is held for this request */
127 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
128 /* already slept for hybrid poll */
129 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20))
130 /* ->timeout has been called, don't expire again */
131 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21))
133 /* flags that prevent us from merging requests: */
134 #define RQF_NOMERGE_FLAGS \
135 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
138 * Request state for blk-mq.
147 * Try to put the fields that are referenced together in the same cacheline.
149 * If you modify this structure, make sure to update blk_rq_init() and
150 * especially blk_mq_rq_ctx_init() to take care of the added fields.
153 struct request_queue *q;
154 struct blk_mq_ctx *mq_ctx;
157 unsigned int cmd_flags; /* op and common flags */
158 req_flags_t rq_flags;
162 /* the following two fields are internal, NEVER access directly */
163 unsigned int __data_len; /* total data len */
165 sector_t __sector; /* sector cursor */
170 struct list_head queuelist;
173 * The hash is used inside the scheduler, and killed once the
174 * request reaches the dispatch list. The ipi_list is only used
175 * to queue the request for softirq completion, which is long
176 * after the request has been unhashed (and even removed from
177 * the dispatch list).
180 struct hlist_node hash; /* merge hash */
181 struct list_head ipi_list;
185 * The rb_node is only used inside the io scheduler, requests
186 * are pruned when moved to the dispatch queue. So let the
187 * completion_data share space with the rb_node.
190 struct rb_node rb_node; /* sort/lookup */
191 struct bio_vec special_vec;
192 void *completion_data;
193 int error_count; /* for legacy drivers, don't use */
197 * Three pointers are available for the IO schedulers, if they need
198 * more they have to dynamically allocate it. Flush requests are
199 * never put on the IO scheduler. So let the flush fields share
200 * space with the elevator data.
210 struct list_head list;
211 rq_end_io_fn *saved_end_io;
215 struct gendisk *rq_disk;
216 struct hd_struct *part;
217 /* Time that I/O was submitted to the kernel. */
219 /* Time that I/O was submitted to the device. */
220 u64 io_start_time_ns;
222 #ifdef CONFIG_BLK_WBT
223 unsigned short wbt_flags;
225 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
226 unsigned short throtl_size;
230 * Number of scatter-gather DMA addr+len pairs after
231 * physical address coalescing is performed.
233 unsigned short nr_phys_segments;
235 #if defined(CONFIG_BLK_DEV_INTEGRITY)
236 unsigned short nr_integrity_segments;
239 unsigned short write_hint;
240 unsigned short ioprio;
242 void *special; /* opaque pointer available for LLD use */
244 unsigned int extra_len; /* length of alignment and padding */
246 enum mq_rq_state state;
249 unsigned int timeout;
251 /* access through blk_rq_set_deadline, blk_rq_deadline */
252 unsigned long __deadline;
254 struct list_head timeout_list;
257 struct __call_single_data csd;
262 * completion callback.
264 rq_end_io_fn *end_io;
268 struct request *next_rq;
270 #ifdef CONFIG_BLK_CGROUP
271 struct request_list *rl; /* rl this rq is alloced from */
275 static inline bool blk_op_is_scsi(unsigned int op)
277 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
280 static inline bool blk_op_is_private(unsigned int op)
282 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
285 static inline bool blk_rq_is_scsi(struct request *rq)
287 return blk_op_is_scsi(req_op(rq));
290 static inline bool blk_rq_is_private(struct request *rq)
292 return blk_op_is_private(req_op(rq));
295 static inline bool blk_rq_is_passthrough(struct request *rq)
297 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
300 static inline bool bio_is_passthrough(struct bio *bio)
302 unsigned op = bio_op(bio);
304 return blk_op_is_scsi(op) || blk_op_is_private(op);
307 static inline unsigned short req_get_ioprio(struct request *req)
312 #include <linux/elevator.h>
314 struct blk_queue_ctx;
316 typedef void (request_fn_proc) (struct request_queue *q);
317 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
318 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
319 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
320 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
323 typedef void (softirq_done_fn)(struct request *);
324 typedef int (dma_drain_needed_fn)(struct request *);
325 typedef int (lld_busy_fn) (struct request_queue *q);
326 typedef int (bsg_job_fn) (struct bsg_job *);
327 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
328 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
330 enum blk_eh_timer_return {
331 BLK_EH_DONE, /* drivers has completed the command */
332 BLK_EH_RESET_TIMER, /* reset timer and try again */
335 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
337 enum blk_queue_state {
342 struct blk_queue_tag {
343 struct request **tag_index; /* map of busy tags */
344 unsigned long *tag_map; /* bit map of free/busy tags */
345 int max_depth; /* what we will send to device */
346 int real_max_depth; /* what the array can hold */
347 atomic_t refcnt; /* map can be shared */
348 int alloc_policy; /* tag allocation policy */
349 int next_tag; /* next tag */
351 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
352 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
354 #define BLK_SCSI_MAX_CMDS (256)
355 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
358 * Zoned block device models (zoned limit).
360 enum blk_zoned_model {
361 BLK_ZONED_NONE, /* Regular block device */
362 BLK_ZONED_HA, /* Host-aware zoned block device */
363 BLK_ZONED_HM, /* Host-managed zoned block device */
366 struct queue_limits {
367 unsigned long bounce_pfn;
368 unsigned long seg_boundary_mask;
369 unsigned long virt_boundary_mask;
371 unsigned int max_hw_sectors;
372 unsigned int max_dev_sectors;
373 unsigned int chunk_sectors;
374 unsigned int max_sectors;
375 unsigned int max_segment_size;
376 unsigned int physical_block_size;
377 unsigned int alignment_offset;
380 unsigned int max_discard_sectors;
381 unsigned int max_hw_discard_sectors;
382 unsigned int max_write_same_sectors;
383 unsigned int max_write_zeroes_sectors;
384 unsigned int discard_granularity;
385 unsigned int discard_alignment;
387 unsigned short logical_block_size;
388 unsigned short max_segments;
389 unsigned short max_integrity_segments;
390 unsigned short max_discard_segments;
392 unsigned char misaligned;
393 unsigned char discard_misaligned;
394 unsigned char cluster;
395 unsigned char raid_partial_stripes_expensive;
396 enum blk_zoned_model zoned;
399 #ifdef CONFIG_BLK_DEV_ZONED
401 struct blk_zone_report_hdr {
402 unsigned int nr_zones;
406 extern int blkdev_report_zones(struct block_device *bdev,
407 sector_t sector, struct blk_zone *zones,
408 unsigned int *nr_zones, gfp_t gfp_mask);
409 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
410 sector_t nr_sectors, gfp_t gfp_mask);
412 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
413 unsigned int cmd, unsigned long arg);
414 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
415 unsigned int cmd, unsigned long arg);
417 #else /* CONFIG_BLK_DEV_ZONED */
419 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
420 fmode_t mode, unsigned int cmd,
426 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
427 fmode_t mode, unsigned int cmd,
433 #endif /* CONFIG_BLK_DEV_ZONED */
435 struct request_queue {
437 * Together with queue_head for cacheline sharing
439 struct list_head queue_head;
440 struct request *last_merge;
441 struct elevator_queue *elevator;
442 int nr_rqs[2]; /* # allocated [a]sync rqs */
443 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
445 struct blk_queue_stats *stats;
446 struct rq_qos *rq_qos;
449 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
450 * is used, root blkg allocates from @q->root_rl and all other
451 * blkgs from their own blkg->rl. Which one to use should be
452 * determined using bio_request_list().
454 struct request_list root_rl;
456 request_fn_proc *request_fn;
457 make_request_fn *make_request_fn;
459 prep_rq_fn *prep_rq_fn;
460 unprep_rq_fn *unprep_rq_fn;
461 softirq_done_fn *softirq_done_fn;
462 rq_timed_out_fn *rq_timed_out_fn;
463 dma_drain_needed_fn *dma_drain_needed;
464 lld_busy_fn *lld_busy_fn;
465 /* Called just after a request is allocated */
466 init_rq_fn *init_rq_fn;
467 /* Called just before a request is freed */
468 exit_rq_fn *exit_rq_fn;
469 /* Called from inside blk_get_request() */
470 void (*initialize_rq_fn)(struct request *rq);
472 const struct blk_mq_ops *mq_ops;
474 unsigned int *mq_map;
477 struct blk_mq_ctx __percpu *queue_ctx;
478 unsigned int nr_queues;
480 unsigned int queue_depth;
482 /* hw dispatch queues */
483 struct blk_mq_hw_ctx **queue_hw_ctx;
484 unsigned int nr_hw_queues;
487 * Dispatch queue sorting
490 struct request *boundary_rq;
493 * Delayed queue handling
495 struct delayed_work delay_work;
497 struct backing_dev_info *backing_dev_info;
500 * The queue owner gets to use this for whatever they like.
501 * ll_rw_blk doesn't touch it.
506 * various queue flags, see QUEUE_* below
508 unsigned long queue_flags;
511 * ida allocated id for this queue. Used to index queues from
517 * queue needs bounce pages for pages above this limit
522 * protects queue structures from reentrancy. ->__queue_lock should
523 * _never_ be used directly, it is queue private. always use
526 spinlock_t __queue_lock;
527 spinlock_t *queue_lock;
537 struct kobject mq_kobj;
539 #ifdef CONFIG_BLK_DEV_INTEGRITY
540 struct blk_integrity integrity;
541 #endif /* CONFIG_BLK_DEV_INTEGRITY */
546 unsigned int nr_pending;
552 unsigned long nr_requests; /* Max # of requests */
553 unsigned int nr_congestion_on;
554 unsigned int nr_congestion_off;
555 unsigned int nr_batching;
557 unsigned int dma_drain_size;
558 void *dma_drain_buffer;
559 unsigned int dma_pad_mask;
560 unsigned int dma_alignment;
562 struct blk_queue_tag *queue_tags;
564 unsigned int nr_sorted;
565 unsigned int in_flight[2];
568 * Number of active block driver functions for which blk_drain_queue()
569 * must wait. Must be incremented around functions that unlock the
570 * queue_lock internally, e.g. scsi_request_fn().
572 unsigned int request_fn_active;
574 unsigned int rq_timeout;
577 struct blk_stat_callback *poll_cb;
578 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
580 struct timer_list timeout;
581 struct work_struct timeout_work;
582 struct list_head timeout_list;
584 struct list_head icq_list;
585 #ifdef CONFIG_BLK_CGROUP
586 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
587 struct blkcg_gq *root_blkg;
588 struct list_head blkg_list;
591 struct queue_limits limits;
593 #ifdef CONFIG_BLK_DEV_ZONED
595 * Zoned block device information for request dispatch control.
596 * nr_zones is the total number of zones of the device. This is always
597 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
598 * bits which indicates if a zone is conventional (bit clear) or
599 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
600 * bits which indicates if a zone is write locked, that is, if a write
601 * request targeting the zone was dispatched. All three fields are
602 * initialized by the low level device driver (e.g. scsi/sd.c).
603 * Stacking drivers (device mappers) may or may not initialize
606 * Reads of this information must be protected with blk_queue_enter() /
607 * blk_queue_exit(). Modifying this information is only allowed while
608 * no requests are being processed. See also blk_mq_freeze_queue() and
609 * blk_mq_unfreeze_queue().
611 unsigned int nr_zones;
612 unsigned long *seq_zones_bitmap;
613 unsigned long *seq_zones_wlock;
614 #endif /* CONFIG_BLK_DEV_ZONED */
619 unsigned int sg_timeout;
620 unsigned int sg_reserved_size;
622 #ifdef CONFIG_BLK_DEV_IO_TRACE
623 struct blk_trace *blk_trace;
624 struct mutex blk_trace_mutex;
627 * for flush operations
629 struct blk_flush_queue *fq;
631 struct list_head requeue_list;
632 spinlock_t requeue_lock;
633 struct delayed_work requeue_work;
635 struct mutex sysfs_lock;
638 atomic_t mq_freeze_depth;
640 #if defined(CONFIG_BLK_DEV_BSG)
641 bsg_job_fn *bsg_job_fn;
642 struct bsg_class_device bsg_dev;
645 #ifdef CONFIG_BLK_DEV_THROTTLING
647 struct throtl_data *td;
649 struct rcu_head rcu_head;
650 wait_queue_head_t mq_freeze_wq;
651 struct percpu_ref q_usage_counter;
652 struct list_head all_q_node;
654 struct blk_mq_tag_set *tag_set;
655 struct list_head tag_set_list;
656 struct bio_set bio_split;
658 #ifdef CONFIG_BLK_DEBUG_FS
659 struct dentry *debugfs_dir;
660 struct dentry *sched_debugfs_dir;
663 bool mq_sysfs_init_done;
668 struct work_struct release_work;
670 #define BLK_MAX_WRITE_HINTS 5
671 u64 write_hints[BLK_MAX_WRITE_HINTS];
674 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
675 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
676 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
677 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
678 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
679 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
680 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
681 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
682 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
683 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
684 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
685 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
686 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
687 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
688 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
689 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
690 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
691 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
692 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
693 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
694 #define QUEUE_FLAG_WC 20 /* Write back caching */
695 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
696 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
697 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
698 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
699 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
700 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
701 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
702 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
703 #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */
705 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
706 (1 << QUEUE_FLAG_SAME_COMP) | \
707 (1 << QUEUE_FLAG_ADD_RANDOM))
709 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
710 (1 << QUEUE_FLAG_SAME_COMP) | \
711 (1 << QUEUE_FLAG_POLL))
713 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
714 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
715 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
716 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
718 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
719 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
720 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
721 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
722 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
723 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
724 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
725 #define blk_queue_noxmerges(q) \
726 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
727 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
728 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
729 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
730 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
731 #define blk_queue_secure_erase(q) \
732 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
733 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
734 #define blk_queue_scsi_passthrough(q) \
735 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
737 #define blk_noretry_request(rq) \
738 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
739 REQ_FAILFAST_DRIVER))
740 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
741 #define blk_queue_preempt_only(q) \
742 test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags)
743 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
745 extern int blk_set_preempt_only(struct request_queue *q);
746 extern void blk_clear_preempt_only(struct request_queue *q);
748 static inline int queue_in_flight(struct request_queue *q)
750 return q->in_flight[0] + q->in_flight[1];
753 static inline bool blk_account_rq(struct request *rq)
755 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
758 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
759 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
760 /* rq->queuelist of dequeued request must be list_empty() */
761 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
763 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
765 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
768 * Driver can handle struct request, if it either has an old style
769 * request_fn defined, or is blk-mq based.
771 static inline bool queue_is_rq_based(struct request_queue *q)
773 return q->request_fn || q->mq_ops;
776 static inline unsigned int blk_queue_cluster(struct request_queue *q)
778 return q->limits.cluster;
781 static inline enum blk_zoned_model
782 blk_queue_zoned_model(struct request_queue *q)
784 return q->limits.zoned;
787 static inline bool blk_queue_is_zoned(struct request_queue *q)
789 switch (blk_queue_zoned_model(q)) {
798 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
800 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
803 #ifdef CONFIG_BLK_DEV_ZONED
804 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
807 if (!blk_queue_is_zoned(q))
809 return sector >> ilog2(q->limits.chunk_sectors);
812 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
815 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
817 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
819 #endif /* CONFIG_BLK_DEV_ZONED */
821 static inline bool rq_is_sync(struct request *rq)
823 return op_is_sync(rq->cmd_flags);
826 static inline bool blk_rl_full(struct request_list *rl, bool sync)
828 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
830 return rl->flags & flag;
833 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
835 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
840 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
842 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
847 static inline bool rq_mergeable(struct request *rq)
849 if (blk_rq_is_passthrough(rq))
852 if (req_op(rq) == REQ_OP_FLUSH)
855 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
858 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
860 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
866 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
868 if (bio_page(a) == bio_page(b) &&
869 bio_offset(a) == bio_offset(b))
875 static inline unsigned int blk_queue_depth(struct request_queue *q)
878 return q->queue_depth;
880 return q->nr_requests;
884 * q->prep_rq_fn return values
887 BLKPREP_OK, /* serve it */
888 BLKPREP_KILL, /* fatal error, kill, return -EIO */
889 BLKPREP_DEFER, /* leave on queue */
890 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
893 extern unsigned long blk_max_low_pfn, blk_max_pfn;
896 * standard bounce addresses:
898 * BLK_BOUNCE_HIGH : bounce all highmem pages
899 * BLK_BOUNCE_ANY : don't bounce anything
900 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
903 #if BITS_PER_LONG == 32
904 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
906 #define BLK_BOUNCE_HIGH -1ULL
908 #define BLK_BOUNCE_ANY (-1ULL)
909 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
912 * default timeout for SG_IO if none specified
914 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
915 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
921 unsigned long offset;
926 struct req_iterator {
927 struct bvec_iter iter;
931 /* This should not be used directly - use rq_for_each_segment */
932 #define for_each_bio(_bio) \
933 for (; _bio; _bio = _bio->bi_next)
934 #define __rq_for_each_bio(_bio, rq) \
936 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
938 #define rq_for_each_segment(bvl, _rq, _iter) \
939 __rq_for_each_bio(_iter.bio, _rq) \
940 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
942 #define rq_iter_last(bvec, _iter) \
943 (_iter.bio->bi_next == NULL && \
944 bio_iter_last(bvec, _iter.iter))
946 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
947 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
949 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
950 extern void rq_flush_dcache_pages(struct request *rq);
952 static inline void rq_flush_dcache_pages(struct request *rq)
957 extern int blk_register_queue(struct gendisk *disk);
958 extern void blk_unregister_queue(struct gendisk *disk);
959 extern blk_qc_t generic_make_request(struct bio *bio);
960 extern blk_qc_t direct_make_request(struct bio *bio);
961 extern void blk_rq_init(struct request_queue *q, struct request *rq);
962 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
963 extern void blk_put_request(struct request *);
964 extern void __blk_put_request(struct request_queue *, struct request *);
965 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
966 blk_mq_req_flags_t flags);
967 extern void blk_requeue_request(struct request_queue *, struct request *);
968 extern int blk_lld_busy(struct request_queue *q);
969 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
970 struct bio_set *bs, gfp_t gfp_mask,
971 int (*bio_ctr)(struct bio *, struct bio *, void *),
973 extern void blk_rq_unprep_clone(struct request *rq);
974 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
976 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
977 extern void blk_delay_queue(struct request_queue *, unsigned long);
978 extern void blk_queue_split(struct request_queue *, struct bio **);
979 extern void blk_recount_segments(struct request_queue *, struct bio *);
980 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
981 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
982 unsigned int, void __user *);
983 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
984 unsigned int, void __user *);
985 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
986 struct scsi_ioctl_command __user *);
988 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
989 extern void blk_queue_exit(struct request_queue *q);
990 extern void blk_start_queue(struct request_queue *q);
991 extern void blk_start_queue_async(struct request_queue *q);
992 extern void blk_stop_queue(struct request_queue *q);
993 extern void blk_sync_queue(struct request_queue *q);
994 extern void __blk_stop_queue(struct request_queue *q);
995 extern void __blk_run_queue(struct request_queue *q);
996 extern void __blk_run_queue_uncond(struct request_queue *q);
997 extern void blk_run_queue(struct request_queue *);
998 extern void blk_run_queue_async(struct request_queue *q);
999 extern int blk_rq_map_user(struct request_queue *, struct request *,
1000 struct rq_map_data *, void __user *, unsigned long,
1002 extern int blk_rq_unmap_user(struct bio *);
1003 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1004 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1005 struct rq_map_data *, const struct iov_iter *,
1007 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1008 struct request *, int);
1009 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1010 struct request *, int, rq_end_io_fn *);
1012 int blk_status_to_errno(blk_status_t status);
1013 blk_status_t errno_to_blk_status(int errno);
1015 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1017 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1019 return bdev->bd_disk->queue; /* this is never NULL */
1023 * The basic unit of block I/O is a sector. It is used in a number of contexts
1024 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1025 * bytes. Variables of type sector_t represent an offset or size that is a
1026 * multiple of 512 bytes. Hence these two constants.
1028 #ifndef SECTOR_SHIFT
1029 #define SECTOR_SHIFT 9
1032 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1036 * blk_rq_pos() : the current sector
1037 * blk_rq_bytes() : bytes left in the entire request
1038 * blk_rq_cur_bytes() : bytes left in the current segment
1039 * blk_rq_err_bytes() : bytes left till the next error boundary
1040 * blk_rq_sectors() : sectors left in the entire request
1041 * blk_rq_cur_sectors() : sectors left in the current segment
1043 static inline sector_t blk_rq_pos(const struct request *rq)
1045 return rq->__sector;
1048 static inline unsigned int blk_rq_bytes(const struct request *rq)
1050 return rq->__data_len;
1053 static inline int blk_rq_cur_bytes(const struct request *rq)
1055 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1058 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1060 static inline unsigned int blk_rq_sectors(const struct request *rq)
1062 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1065 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1067 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1070 #ifdef CONFIG_BLK_DEV_ZONED
1071 static inline unsigned int blk_rq_zone_no(struct request *rq)
1073 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1076 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1078 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1080 #endif /* CONFIG_BLK_DEV_ZONED */
1083 * Some commands like WRITE SAME have a payload or data transfer size which
1084 * is different from the size of the request. Any driver that supports such
1085 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1086 * calculate the data transfer size.
1088 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1090 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1091 return rq->special_vec.bv_len;
1092 return blk_rq_bytes(rq);
1095 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1098 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1099 return min(q->limits.max_discard_sectors,
1100 UINT_MAX >> SECTOR_SHIFT);
1102 if (unlikely(op == REQ_OP_WRITE_SAME))
1103 return q->limits.max_write_same_sectors;
1105 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1106 return q->limits.max_write_zeroes_sectors;
1108 return q->limits.max_sectors;
1112 * Return maximum size of a request at given offset. Only valid for
1113 * file system requests.
1115 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1118 if (!q->limits.chunk_sectors)
1119 return q->limits.max_sectors;
1121 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1122 (offset & (q->limits.chunk_sectors - 1))));
1125 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1128 struct request_queue *q = rq->q;
1130 if (blk_rq_is_passthrough(rq))
1131 return q->limits.max_hw_sectors;
1133 if (!q->limits.chunk_sectors ||
1134 req_op(rq) == REQ_OP_DISCARD ||
1135 req_op(rq) == REQ_OP_SECURE_ERASE)
1136 return blk_queue_get_max_sectors(q, req_op(rq));
1138 return min(blk_max_size_offset(q, offset),
1139 blk_queue_get_max_sectors(q, req_op(rq)));
1142 static inline unsigned int blk_rq_count_bios(struct request *rq)
1144 unsigned int nr_bios = 0;
1147 __rq_for_each_bio(bio, rq)
1154 * Request issue related functions.
1156 extern struct request *blk_peek_request(struct request_queue *q);
1157 extern void blk_start_request(struct request *rq);
1158 extern struct request *blk_fetch_request(struct request_queue *q);
1160 void blk_steal_bios(struct bio_list *list, struct request *rq);
1163 * Request completion related functions.
1165 * blk_update_request() completes given number of bytes and updates
1166 * the request without completing it.
1168 * blk_end_request() and friends. __blk_end_request() must be called
1169 * with the request queue spinlock acquired.
1171 * Several drivers define their own end_request and call
1172 * blk_end_request() for parts of the original function.
1173 * This prevents code duplication in drivers.
1175 extern bool blk_update_request(struct request *rq, blk_status_t error,
1176 unsigned int nr_bytes);
1177 extern void blk_finish_request(struct request *rq, blk_status_t error);
1178 extern bool blk_end_request(struct request *rq, blk_status_t error,
1179 unsigned int nr_bytes);
1180 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1181 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1182 unsigned int nr_bytes);
1183 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1184 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1186 extern void blk_complete_request(struct request *);
1187 extern void __blk_complete_request(struct request *);
1188 extern void blk_abort_request(struct request *);
1189 extern void blk_unprep_request(struct request *);
1192 * Access functions for manipulating queue properties
1194 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1195 spinlock_t *lock, int node_id);
1196 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1197 extern int blk_init_allocated_queue(struct request_queue *);
1198 extern void blk_cleanup_queue(struct request_queue *);
1199 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1200 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1201 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1202 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1203 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1204 extern void blk_queue_max_discard_segments(struct request_queue *,
1206 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1207 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1208 unsigned int max_discard_sectors);
1209 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1210 unsigned int max_write_same_sectors);
1211 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1212 unsigned int max_write_same_sectors);
1213 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1214 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1215 extern void blk_queue_alignment_offset(struct request_queue *q,
1216 unsigned int alignment);
1217 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1218 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1219 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1220 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1221 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1222 extern void blk_set_default_limits(struct queue_limits *lim);
1223 extern void blk_set_stacking_limits(struct queue_limits *lim);
1224 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1226 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1228 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1230 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1231 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1232 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1233 extern int blk_queue_dma_drain(struct request_queue *q,
1234 dma_drain_needed_fn *dma_drain_needed,
1235 void *buf, unsigned int size);
1236 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1237 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1238 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1239 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1240 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1241 extern void blk_queue_dma_alignment(struct request_queue *, int);
1242 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1243 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1244 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1245 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1246 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1247 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1250 * Number of physical segments as sent to the device.
1252 * Normally this is the number of discontiguous data segments sent by the
1253 * submitter. But for data-less command like discard we might have no
1254 * actual data segments submitted, but the driver might have to add it's
1255 * own special payload. In that case we still return 1 here so that this
1256 * special payload will be mapped.
1258 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1260 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1262 return rq->nr_phys_segments;
1266 * Number of discard segments (or ranges) the driver needs to fill in.
1267 * Each discard bio merged into a request is counted as one segment.
1269 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1271 return max_t(unsigned short, rq->nr_phys_segments, 1);
1274 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1275 extern void blk_dump_rq_flags(struct request *, char *);
1276 extern long nr_blockdev_pages(void);
1278 bool __must_check blk_get_queue(struct request_queue *);
1279 struct request_queue *blk_alloc_queue(gfp_t);
1280 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1282 extern void blk_put_queue(struct request_queue *);
1283 extern void blk_set_queue_dying(struct request_queue *);
1286 * block layer runtime pm functions
1289 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1290 extern int blk_pre_runtime_suspend(struct request_queue *q);
1291 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1292 extern void blk_pre_runtime_resume(struct request_queue *q);
1293 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1294 extern void blk_set_runtime_active(struct request_queue *q);
1296 static inline void blk_pm_runtime_init(struct request_queue *q,
1297 struct device *dev) {}
1298 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1302 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1303 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1304 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1305 static inline void blk_set_runtime_active(struct request_queue *q) {}
1309 * blk_plug permits building a queue of related requests by holding the I/O
1310 * fragments for a short period. This allows merging of sequential requests
1311 * into single larger request. As the requests are moved from a per-task list to
1312 * the device's request_queue in a batch, this results in improved scalability
1313 * as the lock contention for request_queue lock is reduced.
1315 * It is ok not to disable preemption when adding the request to the plug list
1316 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1317 * the plug list when the task sleeps by itself. For details, please see
1318 * schedule() where blk_schedule_flush_plug() is called.
1321 struct list_head list; /* requests */
1322 struct list_head mq_list; /* blk-mq requests */
1323 struct list_head cb_list; /* md requires an unplug callback */
1325 #define BLK_MAX_REQUEST_COUNT 16
1326 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1329 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1330 struct blk_plug_cb {
1331 struct list_head list;
1332 blk_plug_cb_fn callback;
1335 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1336 void *data, int size);
1337 extern void blk_start_plug(struct blk_plug *);
1338 extern void blk_finish_plug(struct blk_plug *);
1339 extern void blk_flush_plug_list(struct blk_plug *, bool);
1341 static inline void blk_flush_plug(struct task_struct *tsk)
1343 struct blk_plug *plug = tsk->plug;
1346 blk_flush_plug_list(plug, false);
1349 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1351 struct blk_plug *plug = tsk->plug;
1354 blk_flush_plug_list(plug, true);
1357 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1359 struct blk_plug *plug = tsk->plug;
1362 (!list_empty(&plug->list) ||
1363 !list_empty(&plug->mq_list) ||
1364 !list_empty(&plug->cb_list));
1370 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1371 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1372 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1373 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1374 extern void blk_queue_free_tags(struct request_queue *);
1375 extern int blk_queue_resize_tags(struct request_queue *, int);
1376 extern struct blk_queue_tag *blk_init_tags(int, int);
1377 extern void blk_free_tags(struct blk_queue_tag *);
1379 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1382 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1384 return bqt->tag_index[tag];
1387 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1388 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1389 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1391 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1393 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1394 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1395 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1396 sector_t nr_sects, gfp_t gfp_mask, int flags,
1399 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1400 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1402 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1403 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1405 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1406 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1408 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1409 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1411 return blkdev_issue_discard(sb->s_bdev,
1412 block << (sb->s_blocksize_bits -
1414 nr_blocks << (sb->s_blocksize_bits -
1418 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1419 sector_t nr_blocks, gfp_t gfp_mask)
1421 return blkdev_issue_zeroout(sb->s_bdev,
1422 block << (sb->s_blocksize_bits -
1424 nr_blocks << (sb->s_blocksize_bits -
1429 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1431 enum blk_default_limits {
1432 BLK_MAX_SEGMENTS = 128,
1433 BLK_SAFE_MAX_SECTORS = 255,
1434 BLK_DEF_MAX_SECTORS = 2560,
1435 BLK_MAX_SEGMENT_SIZE = 65536,
1436 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1439 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1441 return q->limits.seg_boundary_mask;
1444 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1446 return q->limits.virt_boundary_mask;
1449 static inline unsigned int queue_max_sectors(struct request_queue *q)
1451 return q->limits.max_sectors;
1454 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1456 return q->limits.max_hw_sectors;
1459 static inline unsigned short queue_max_segments(struct request_queue *q)
1461 return q->limits.max_segments;
1464 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1466 return q->limits.max_discard_segments;
1469 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1471 return q->limits.max_segment_size;
1474 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1478 if (q && q->limits.logical_block_size)
1479 retval = q->limits.logical_block_size;
1484 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1486 return queue_logical_block_size(bdev_get_queue(bdev));
1489 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1491 return q->limits.physical_block_size;
1494 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1496 return queue_physical_block_size(bdev_get_queue(bdev));
1499 static inline unsigned int queue_io_min(struct request_queue *q)
1501 return q->limits.io_min;
1504 static inline int bdev_io_min(struct block_device *bdev)
1506 return queue_io_min(bdev_get_queue(bdev));
1509 static inline unsigned int queue_io_opt(struct request_queue *q)
1511 return q->limits.io_opt;
1514 static inline int bdev_io_opt(struct block_device *bdev)
1516 return queue_io_opt(bdev_get_queue(bdev));
1519 static inline int queue_alignment_offset(struct request_queue *q)
1521 if (q->limits.misaligned)
1524 return q->limits.alignment_offset;
1527 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1529 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1530 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1533 return (granularity + lim->alignment_offset - alignment) % granularity;
1536 static inline int bdev_alignment_offset(struct block_device *bdev)
1538 struct request_queue *q = bdev_get_queue(bdev);
1540 if (q->limits.misaligned)
1543 if (bdev != bdev->bd_contains)
1544 return bdev->bd_part->alignment_offset;
1546 return q->limits.alignment_offset;
1549 static inline int queue_discard_alignment(struct request_queue *q)
1551 if (q->limits.discard_misaligned)
1554 return q->limits.discard_alignment;
1557 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1559 unsigned int alignment, granularity, offset;
1561 if (!lim->max_discard_sectors)
1564 /* Why are these in bytes, not sectors? */
1565 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1566 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1570 /* Offset of the partition start in 'granularity' sectors */
1571 offset = sector_div(sector, granularity);
1573 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1574 offset = (granularity + alignment - offset) % granularity;
1576 /* Turn it back into bytes, gaah */
1577 return offset << SECTOR_SHIFT;
1580 static inline int bdev_discard_alignment(struct block_device *bdev)
1582 struct request_queue *q = bdev_get_queue(bdev);
1584 if (bdev != bdev->bd_contains)
1585 return bdev->bd_part->discard_alignment;
1587 return q->limits.discard_alignment;
1590 static inline unsigned int bdev_write_same(struct block_device *bdev)
1592 struct request_queue *q = bdev_get_queue(bdev);
1595 return q->limits.max_write_same_sectors;
1600 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1602 struct request_queue *q = bdev_get_queue(bdev);
1605 return q->limits.max_write_zeroes_sectors;
1610 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1612 struct request_queue *q = bdev_get_queue(bdev);
1615 return blk_queue_zoned_model(q);
1617 return BLK_ZONED_NONE;
1620 static inline bool bdev_is_zoned(struct block_device *bdev)
1622 struct request_queue *q = bdev_get_queue(bdev);
1625 return blk_queue_is_zoned(q);
1630 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1632 struct request_queue *q = bdev_get_queue(bdev);
1635 return blk_queue_zone_sectors(q);
1639 static inline int queue_dma_alignment(struct request_queue *q)
1641 return q ? q->dma_alignment : 511;
1644 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1647 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1648 return !(addr & alignment) && !(len & alignment);
1651 /* assumes size > 256 */
1652 static inline unsigned int blksize_bits(unsigned int size)
1654 unsigned int bits = 8;
1658 } while (size > 256);
1662 static inline unsigned int block_size(struct block_device *bdev)
1664 return bdev->bd_block_size;
1667 static inline bool queue_flush_queueable(struct request_queue *q)
1669 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1672 typedef struct {struct page *v;} Sector;
1674 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1676 static inline void put_dev_sector(Sector p)
1681 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1682 struct bio_vec *bprv, unsigned int offset)
1685 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1689 * Check if adding a bio_vec after bprv with offset would create a gap in
1690 * the SG list. Most drivers don't care about this, but some do.
1692 static inline bool bvec_gap_to_prev(struct request_queue *q,
1693 struct bio_vec *bprv, unsigned int offset)
1695 if (!queue_virt_boundary(q))
1697 return __bvec_gap_to_prev(q, bprv, offset);
1701 * Check if the two bvecs from two bios can be merged to one segment.
1702 * If yes, no need to check gap between the two bios since the 1st bio
1703 * and the 1st bvec in the 2nd bio can be handled in one segment.
1705 static inline bool bios_segs_mergeable(struct request_queue *q,
1706 struct bio *prev, struct bio_vec *prev_last_bv,
1707 struct bio_vec *next_first_bv)
1709 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1711 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1713 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1714 queue_max_segment_size(q))
1719 static inline bool bio_will_gap(struct request_queue *q,
1720 struct request *prev_rq,
1724 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1725 struct bio_vec pb, nb;
1728 * don't merge if the 1st bio starts with non-zero
1729 * offset, otherwise it is quite difficult to respect
1730 * sg gap limit. We work hard to merge a huge number of small
1731 * single bios in case of mkfs.
1734 bio_get_first_bvec(prev_rq->bio, &pb);
1736 bio_get_first_bvec(prev, &pb);
1741 * We don't need to worry about the situation that the
1742 * merged segment ends in unaligned virt boundary:
1744 * - if 'pb' ends aligned, the merged segment ends aligned
1745 * - if 'pb' ends unaligned, the next bio must include
1746 * one single bvec of 'nb', otherwise the 'nb' can't
1749 bio_get_last_bvec(prev, &pb);
1750 bio_get_first_bvec(next, &nb);
1752 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1753 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1759 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1761 return bio_will_gap(req->q, req, req->biotail, bio);
1764 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1766 return bio_will_gap(req->q, NULL, bio, req->bio);
1769 int kblockd_schedule_work(struct work_struct *work);
1770 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1771 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1773 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1774 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1775 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1776 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1778 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1780 enum blk_integrity_flags {
1781 BLK_INTEGRITY_VERIFY = 1 << 0,
1782 BLK_INTEGRITY_GENERATE = 1 << 1,
1783 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1784 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1787 struct blk_integrity_iter {
1791 unsigned int data_size;
1792 unsigned short interval;
1793 const char *disk_name;
1796 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1798 struct blk_integrity_profile {
1799 integrity_processing_fn *generate_fn;
1800 integrity_processing_fn *verify_fn;
1804 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1805 extern void blk_integrity_unregister(struct gendisk *);
1806 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1807 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1808 struct scatterlist *);
1809 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1810 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1812 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1815 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1817 struct blk_integrity *bi = &disk->queue->integrity;
1826 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1828 return blk_get_integrity(bdev->bd_disk);
1831 static inline bool blk_integrity_rq(struct request *rq)
1833 return rq->cmd_flags & REQ_INTEGRITY;
1836 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1839 q->limits.max_integrity_segments = segs;
1842 static inline unsigned short
1843 queue_max_integrity_segments(struct request_queue *q)
1845 return q->limits.max_integrity_segments;
1848 static inline bool integrity_req_gap_back_merge(struct request *req,
1851 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1852 struct bio_integrity_payload *bip_next = bio_integrity(next);
1854 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1855 bip_next->bip_vec[0].bv_offset);
1858 static inline bool integrity_req_gap_front_merge(struct request *req,
1861 struct bio_integrity_payload *bip = bio_integrity(bio);
1862 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1864 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1865 bip_next->bip_vec[0].bv_offset);
1868 #else /* CONFIG_BLK_DEV_INTEGRITY */
1871 struct block_device;
1873 struct blk_integrity;
1875 static inline int blk_integrity_rq(struct request *rq)
1879 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1884 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1886 struct scatterlist *s)
1890 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1894 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1898 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1902 static inline void blk_integrity_register(struct gendisk *d,
1903 struct blk_integrity *b)
1906 static inline void blk_integrity_unregister(struct gendisk *d)
1909 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1913 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1917 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1923 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1930 static inline bool integrity_req_gap_back_merge(struct request *req,
1935 static inline bool integrity_req_gap_front_merge(struct request *req,
1941 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1943 struct block_device_operations {
1944 int (*open) (struct block_device *, fmode_t);
1945 void (*release) (struct gendisk *, fmode_t);
1946 int (*rw_page)(struct block_device *, sector_t, struct page *, bool);
1947 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1948 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1949 unsigned int (*check_events) (struct gendisk *disk,
1950 unsigned int clearing);
1951 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1952 int (*media_changed) (struct gendisk *);
1953 void (*unlock_native_capacity) (struct gendisk *);
1954 int (*revalidate_disk) (struct gendisk *);
1955 int (*getgeo)(struct block_device *, struct hd_geometry *);
1956 /* this callback is with swap_lock and sometimes page table lock held */
1957 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1958 struct module *owner;
1959 const struct pr_ops *pr_ops;
1962 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1964 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1965 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1966 struct writeback_control *);
1968 #ifdef CONFIG_BLK_DEV_ZONED
1969 bool blk_req_needs_zone_write_lock(struct request *rq);
1970 void __blk_req_zone_write_lock(struct request *rq);
1971 void __blk_req_zone_write_unlock(struct request *rq);
1973 static inline void blk_req_zone_write_lock(struct request *rq)
1975 if (blk_req_needs_zone_write_lock(rq))
1976 __blk_req_zone_write_lock(rq);
1979 static inline void blk_req_zone_write_unlock(struct request *rq)
1981 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1982 __blk_req_zone_write_unlock(rq);
1985 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1987 return rq->q->seq_zones_wlock &&
1988 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1991 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1993 if (!blk_req_needs_zone_write_lock(rq))
1995 return !blk_req_zone_is_write_locked(rq);
1998 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2003 static inline void blk_req_zone_write_lock(struct request *rq)
2007 static inline void blk_req_zone_write_unlock(struct request *rq)
2010 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2015 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2019 #endif /* CONFIG_BLK_DEV_ZONED */
2021 #else /* CONFIG_BLOCK */
2023 struct block_device;
2026 * stubs for when the block layer is configured out
2028 #define buffer_heads_over_limit 0
2030 static inline long nr_blockdev_pages(void)
2038 static inline void blk_start_plug(struct blk_plug *plug)
2042 static inline void blk_finish_plug(struct blk_plug *plug)
2046 static inline void blk_flush_plug(struct task_struct *task)
2050 static inline void blk_schedule_flush_plug(struct task_struct *task)
2055 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2060 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2061 sector_t *error_sector)
2066 #endif /* CONFIG_BLOCK */