1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
32 struct scsi_ioctl_command;
35 struct elevator_queue;
41 struct blk_flush_queue;
44 struct blk_queue_stats;
45 struct blk_stat_callback;
47 #define BLKDEV_MIN_RQ 4
48 #define BLKDEV_MAX_RQ 128 /* Default maximum */
50 /* Must be consistent with blk_mq_poll_stats_bkt() */
51 #define BLK_MQ_POLL_STATS_BKTS 16
54 * Maximum number of blkcg policies allowed to be registered concurrently.
55 * Defined here to simplify include dependency.
57 #define BLKCG_MAX_POLS 5
59 static inline int blk_validate_block_size(unsigned int bsize)
61 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize))
67 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
69 #define BLK_RL_SYNCFULL (1U << 0)
70 #define BLK_RL_ASYNCFULL (1U << 1)
73 struct request_queue *q; /* the queue this rl belongs to */
74 #ifdef CONFIG_BLK_CGROUP
75 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
78 * count[], starved[], and wait[] are indexed by
79 * BLK_RW_SYNC/BLK_RW_ASYNC
84 wait_queue_head_t wait[2];
90 typedef __u32 __bitwise req_flags_t;
92 /* elevator knows about this request */
93 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
94 /* drive already may have started this one */
95 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
96 /* uses tagged queueing */
97 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
98 /* may not be passed by ioscheduler */
99 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
100 /* request for flush sequence */
101 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
102 /* merge of different types, fail separately */
103 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
104 /* track inflight for MQ */
105 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
106 /* don't call prep for this one */
107 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
108 /* set for "ide_preempt" requests and also for requests for which the SCSI
109 "quiesce" state must be ignored. */
110 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
111 /* contains copies of user pages */
112 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
113 /* vaguely specified driver internal error. Ignored by the block layer */
114 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
115 /* don't warn about errors */
116 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
117 /* elevator private data attached */
118 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
119 /* account I/O stat */
120 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
121 /* request came from our alloc pool */
122 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
123 /* runtime pm request */
124 #define RQF_PM ((__force req_flags_t)(1 << 15))
125 /* on IO scheduler merge hash */
126 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
127 /* IO stats tracking on */
128 #define RQF_STATS ((__force req_flags_t)(1 << 17))
129 /* Look at ->special_vec for the actual data payload instead of the
131 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
132 /* The per-zone write lock is held for this request */
133 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
134 /* already slept for hybrid poll */
135 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20))
136 /* ->timeout has been called, don't expire again */
137 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21))
139 /* flags that prevent us from merging requests: */
140 #define RQF_NOMERGE_FLAGS \
141 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
144 * Request state for blk-mq.
153 * Try to put the fields that are referenced together in the same cacheline.
155 * If you modify this structure, make sure to update blk_rq_init() and
156 * especially blk_mq_rq_ctx_init() to take care of the added fields.
159 struct request_queue *q;
160 struct blk_mq_ctx *mq_ctx;
163 unsigned int cmd_flags; /* op and common flags */
164 req_flags_t rq_flags;
168 /* the following two fields are internal, NEVER access directly */
169 unsigned int __data_len; /* total data len */
171 sector_t __sector; /* sector cursor */
176 struct list_head queuelist;
179 * The hash is used inside the scheduler, and killed once the
180 * request reaches the dispatch list. The ipi_list is only used
181 * to queue the request for softirq completion, which is long
182 * after the request has been unhashed (and even removed from
183 * the dispatch list).
186 struct hlist_node hash; /* merge hash */
187 struct list_head ipi_list;
191 * The rb_node is only used inside the io scheduler, requests
192 * are pruned when moved to the dispatch queue. So let the
193 * completion_data share space with the rb_node.
196 struct rb_node rb_node; /* sort/lookup */
197 struct bio_vec special_vec;
198 void *completion_data;
199 int error_count; /* for legacy drivers, don't use */
203 * Three pointers are available for the IO schedulers, if they need
204 * more they have to dynamically allocate it. Flush requests are
205 * never put on the IO scheduler. So let the flush fields share
206 * space with the elevator data.
216 struct list_head list;
217 rq_end_io_fn *saved_end_io;
221 struct gendisk *rq_disk;
222 struct hd_struct *part;
223 /* Time that I/O was submitted to the kernel. */
225 /* Time that I/O was submitted to the device. */
226 u64 io_start_time_ns;
228 #ifdef CONFIG_BLK_WBT
229 unsigned short wbt_flags;
231 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
232 unsigned short throtl_size;
236 * Number of scatter-gather DMA addr+len pairs after
237 * physical address coalescing is performed.
239 unsigned short nr_phys_segments;
241 #if defined(CONFIG_BLK_DEV_INTEGRITY)
242 unsigned short nr_integrity_segments;
245 unsigned short write_hint;
246 unsigned short ioprio;
248 void *special; /* opaque pointer available for LLD use */
250 unsigned int extra_len; /* length of alignment and padding */
252 enum mq_rq_state state;
255 unsigned int timeout;
257 /* access through blk_rq_set_deadline, blk_rq_deadline */
258 unsigned long __deadline;
260 struct list_head timeout_list;
263 struct __call_single_data csd;
268 * completion callback.
270 rq_end_io_fn *end_io;
274 struct request *next_rq;
276 #ifdef CONFIG_BLK_CGROUP
277 struct request_list *rl; /* rl this rq is alloced from */
281 static inline bool blk_op_is_scsi(unsigned int op)
283 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
286 static inline bool blk_op_is_private(unsigned int op)
288 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
291 static inline bool blk_rq_is_scsi(struct request *rq)
293 return blk_op_is_scsi(req_op(rq));
296 static inline bool blk_rq_is_private(struct request *rq)
298 return blk_op_is_private(req_op(rq));
301 static inline bool blk_rq_is_passthrough(struct request *rq)
303 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
306 static inline bool bio_is_passthrough(struct bio *bio)
308 unsigned op = bio_op(bio);
310 return blk_op_is_scsi(op) || blk_op_is_private(op);
313 static inline unsigned short req_get_ioprio(struct request *req)
318 #include <linux/elevator.h>
320 struct blk_queue_ctx;
322 typedef void (request_fn_proc) (struct request_queue *q);
323 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
324 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
325 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
326 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
329 typedef void (softirq_done_fn)(struct request *);
330 typedef int (dma_drain_needed_fn)(struct request *);
331 typedef int (lld_busy_fn) (struct request_queue *q);
332 typedef int (bsg_job_fn) (struct bsg_job *);
333 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
334 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
336 enum blk_eh_timer_return {
337 BLK_EH_DONE, /* drivers has completed the command */
338 BLK_EH_RESET_TIMER, /* reset timer and try again */
341 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
343 enum blk_queue_state {
348 struct blk_queue_tag {
349 struct request **tag_index; /* map of busy tags */
350 unsigned long *tag_map; /* bit map of free/busy tags */
351 int max_depth; /* what we will send to device */
352 int real_max_depth; /* what the array can hold */
353 atomic_t refcnt; /* map can be shared */
354 int alloc_policy; /* tag allocation policy */
355 int next_tag; /* next tag */
357 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
358 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
360 #define BLK_SCSI_MAX_CMDS (256)
361 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
364 * Zoned block device models (zoned limit).
366 enum blk_zoned_model {
367 BLK_ZONED_NONE, /* Regular block device */
368 BLK_ZONED_HA, /* Host-aware zoned block device */
369 BLK_ZONED_HM, /* Host-managed zoned block device */
372 struct queue_limits {
373 unsigned long bounce_pfn;
374 unsigned long seg_boundary_mask;
375 unsigned long virt_boundary_mask;
377 unsigned int max_hw_sectors;
378 unsigned int max_dev_sectors;
379 unsigned int chunk_sectors;
380 unsigned int max_sectors;
381 unsigned int max_segment_size;
382 unsigned int physical_block_size;
383 unsigned int logical_block_size;
384 unsigned int alignment_offset;
387 unsigned int max_discard_sectors;
388 unsigned int max_hw_discard_sectors;
389 unsigned int max_write_same_sectors;
390 unsigned int max_write_zeroes_sectors;
391 unsigned int discard_granularity;
392 unsigned int discard_alignment;
394 unsigned short max_segments;
395 unsigned short max_integrity_segments;
396 unsigned short max_discard_segments;
398 unsigned char misaligned;
399 unsigned char discard_misaligned;
400 unsigned char cluster;
401 unsigned char raid_partial_stripes_expensive;
402 enum blk_zoned_model zoned;
405 #ifdef CONFIG_BLK_DEV_ZONED
407 struct blk_zone_report_hdr {
408 unsigned int nr_zones;
412 extern int blkdev_report_zones(struct block_device *bdev,
413 sector_t sector, struct blk_zone *zones,
414 unsigned int *nr_zones, gfp_t gfp_mask);
415 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
416 sector_t nr_sectors, gfp_t gfp_mask);
418 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
419 unsigned int cmd, unsigned long arg);
420 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
421 unsigned int cmd, unsigned long arg);
423 #else /* CONFIG_BLK_DEV_ZONED */
425 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
426 fmode_t mode, unsigned int cmd,
432 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
433 fmode_t mode, unsigned int cmd,
439 #endif /* CONFIG_BLK_DEV_ZONED */
441 struct request_queue {
443 * Together with queue_head for cacheline sharing
445 struct list_head queue_head;
446 struct request *last_merge;
447 struct elevator_queue *elevator;
448 int nr_rqs[2]; /* # allocated [a]sync rqs */
449 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
451 struct blk_queue_stats *stats;
452 struct rq_qos *rq_qos;
455 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
456 * is used, root blkg allocates from @q->root_rl and all other
457 * blkgs from their own blkg->rl. Which one to use should be
458 * determined using bio_request_list().
460 struct request_list root_rl;
462 request_fn_proc *request_fn;
463 make_request_fn *make_request_fn;
465 prep_rq_fn *prep_rq_fn;
466 unprep_rq_fn *unprep_rq_fn;
467 softirq_done_fn *softirq_done_fn;
468 rq_timed_out_fn *rq_timed_out_fn;
469 dma_drain_needed_fn *dma_drain_needed;
470 lld_busy_fn *lld_busy_fn;
471 /* Called just after a request is allocated */
472 init_rq_fn *init_rq_fn;
473 /* Called just before a request is freed */
474 exit_rq_fn *exit_rq_fn;
475 /* Called from inside blk_get_request() */
476 void (*initialize_rq_fn)(struct request *rq);
478 const struct blk_mq_ops *mq_ops;
480 unsigned int *mq_map;
483 struct blk_mq_ctx __percpu *queue_ctx;
484 unsigned int nr_queues;
486 unsigned int queue_depth;
488 /* hw dispatch queues */
489 struct blk_mq_hw_ctx **queue_hw_ctx;
490 unsigned int nr_hw_queues;
493 * Dispatch queue sorting
496 struct request *boundary_rq;
499 * Delayed queue handling
501 struct delayed_work delay_work;
503 struct backing_dev_info *backing_dev_info;
506 * The queue owner gets to use this for whatever they like.
507 * ll_rw_blk doesn't touch it.
512 * various queue flags, see QUEUE_* below
514 unsigned long queue_flags;
516 * Number of contexts that have called blk_set_pm_only(). If this
517 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
523 * ida allocated id for this queue. Used to index queues from
529 * queue needs bounce pages for pages above this limit
534 * protects queue structures from reentrancy. ->__queue_lock should
535 * _never_ be used directly, it is queue private. always use
538 spinlock_t __queue_lock;
539 spinlock_t *queue_lock;
549 struct kobject mq_kobj;
551 #ifdef CONFIG_BLK_DEV_INTEGRITY
552 struct blk_integrity integrity;
553 #endif /* CONFIG_BLK_DEV_INTEGRITY */
558 unsigned int nr_pending;
564 unsigned long nr_requests; /* Max # of requests */
565 unsigned int nr_congestion_on;
566 unsigned int nr_congestion_off;
567 unsigned int nr_batching;
569 unsigned int dma_drain_size;
570 void *dma_drain_buffer;
571 unsigned int dma_pad_mask;
572 unsigned int dma_alignment;
574 struct blk_queue_tag *queue_tags;
576 unsigned int nr_sorted;
577 unsigned int in_flight[2];
580 * Number of active block driver functions for which blk_drain_queue()
581 * must wait. Must be incremented around functions that unlock the
582 * queue_lock internally, e.g. scsi_request_fn().
584 unsigned int request_fn_active;
586 unsigned int rq_timeout;
589 struct blk_stat_callback *poll_cb;
590 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
592 struct timer_list timeout;
593 struct work_struct timeout_work;
594 struct list_head timeout_list;
596 struct list_head icq_list;
597 #ifdef CONFIG_BLK_CGROUP
598 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
599 struct blkcg_gq *root_blkg;
600 struct list_head blkg_list;
603 struct queue_limits limits;
605 #ifdef CONFIG_BLK_DEV_ZONED
607 * Zoned block device information for request dispatch control.
608 * nr_zones is the total number of zones of the device. This is always
609 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
610 * bits which indicates if a zone is conventional (bit clear) or
611 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
612 * bits which indicates if a zone is write locked, that is, if a write
613 * request targeting the zone was dispatched. All three fields are
614 * initialized by the low level device driver (e.g. scsi/sd.c).
615 * Stacking drivers (device mappers) may or may not initialize
618 * Reads of this information must be protected with blk_queue_enter() /
619 * blk_queue_exit(). Modifying this information is only allowed while
620 * no requests are being processed. See also blk_mq_freeze_queue() and
621 * blk_mq_unfreeze_queue().
623 unsigned int nr_zones;
624 unsigned long *seq_zones_bitmap;
625 unsigned long *seq_zones_wlock;
626 #endif /* CONFIG_BLK_DEV_ZONED */
631 unsigned int sg_timeout;
632 unsigned int sg_reserved_size;
634 #ifdef CONFIG_BLK_DEV_IO_TRACE
635 struct blk_trace __rcu *blk_trace;
636 struct mutex blk_trace_mutex;
639 * for flush operations
641 struct blk_flush_queue *fq;
643 struct list_head requeue_list;
644 spinlock_t requeue_lock;
645 struct delayed_work requeue_work;
647 struct mutex sysfs_lock;
648 struct mutex sysfs_dir_lock;
651 atomic_t mq_freeze_depth;
653 #if defined(CONFIG_BLK_DEV_BSG)
654 bsg_job_fn *bsg_job_fn;
655 struct bsg_class_device bsg_dev;
658 #ifdef CONFIG_BLK_DEV_THROTTLING
660 struct throtl_data *td;
662 struct rcu_head rcu_head;
663 wait_queue_head_t mq_freeze_wq;
664 struct percpu_ref q_usage_counter;
665 struct list_head all_q_node;
667 struct blk_mq_tag_set *tag_set;
668 struct list_head tag_set_list;
669 struct bio_set bio_split;
671 #ifdef CONFIG_BLK_DEBUG_FS
672 struct dentry *debugfs_dir;
673 struct dentry *sched_debugfs_dir;
676 bool mq_sysfs_init_done;
681 struct work_struct release_work;
683 #define BLK_MAX_WRITE_HINTS 5
684 u64 write_hints[BLK_MAX_WRITE_HINTS];
687 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
688 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
689 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
690 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
691 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
692 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
693 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
694 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
695 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
696 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
697 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */
698 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
699 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
700 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
701 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
702 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
703 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
704 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
705 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
706 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
707 #define QUEUE_FLAG_WC 20 /* Write back caching */
708 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
709 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
710 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
711 #define QUEUE_FLAG_STATS 24 /* track rq completion times */
712 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
713 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
714 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
715 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
717 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
718 (1 << QUEUE_FLAG_SAME_COMP) | \
719 (1 << QUEUE_FLAG_ADD_RANDOM))
721 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
722 (1 << QUEUE_FLAG_SAME_COMP) | \
723 (1 << QUEUE_FLAG_POLL))
725 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
726 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
727 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
728 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
730 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
731 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
732 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
733 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
734 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
735 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
736 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
737 #define blk_queue_noxmerges(q) \
738 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
739 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
740 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
741 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
742 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
743 #define blk_queue_secure_erase(q) \
744 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
745 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
746 #define blk_queue_scsi_passthrough(q) \
747 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
749 #define blk_noretry_request(rq) \
750 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
751 REQ_FAILFAST_DRIVER))
752 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
753 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
754 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
755 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
757 extern void blk_set_pm_only(struct request_queue *q);
758 extern void blk_clear_pm_only(struct request_queue *q);
760 static inline int queue_in_flight(struct request_queue *q)
762 return q->in_flight[0] + q->in_flight[1];
765 static inline bool blk_account_rq(struct request *rq)
767 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
770 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
771 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
772 /* rq->queuelist of dequeued request must be list_empty() */
773 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
775 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
777 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
780 * Driver can handle struct request, if it either has an old style
781 * request_fn defined, or is blk-mq based.
783 static inline bool queue_is_rq_based(struct request_queue *q)
785 return q->request_fn || q->mq_ops;
788 static inline unsigned int blk_queue_cluster(struct request_queue *q)
790 return q->limits.cluster;
793 static inline enum blk_zoned_model
794 blk_queue_zoned_model(struct request_queue *q)
796 return q->limits.zoned;
799 static inline bool blk_queue_is_zoned(struct request_queue *q)
801 switch (blk_queue_zoned_model(q)) {
810 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
812 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
815 #ifdef CONFIG_BLK_DEV_ZONED
816 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
819 if (!blk_queue_is_zoned(q))
821 return sector >> ilog2(q->limits.chunk_sectors);
824 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
827 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
829 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
831 #endif /* CONFIG_BLK_DEV_ZONED */
833 static inline bool rq_is_sync(struct request *rq)
835 return op_is_sync(rq->cmd_flags);
838 static inline bool blk_rl_full(struct request_list *rl, bool sync)
840 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
842 return rl->flags & flag;
845 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
847 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
852 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
854 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
859 static inline bool rq_mergeable(struct request *rq)
861 if (blk_rq_is_passthrough(rq))
864 if (req_op(rq) == REQ_OP_FLUSH)
867 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
870 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
872 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
878 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
880 if (bio_page(a) == bio_page(b) &&
881 bio_offset(a) == bio_offset(b))
887 static inline unsigned int blk_queue_depth(struct request_queue *q)
890 return q->queue_depth;
892 return q->nr_requests;
896 * q->prep_rq_fn return values
899 BLKPREP_OK, /* serve it */
900 BLKPREP_KILL, /* fatal error, kill, return -EIO */
901 BLKPREP_DEFER, /* leave on queue */
902 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
905 extern unsigned long blk_max_low_pfn, blk_max_pfn;
908 * standard bounce addresses:
910 * BLK_BOUNCE_HIGH : bounce all highmem pages
911 * BLK_BOUNCE_ANY : don't bounce anything
912 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
915 #if BITS_PER_LONG == 32
916 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
918 #define BLK_BOUNCE_HIGH -1ULL
920 #define BLK_BOUNCE_ANY (-1ULL)
921 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
924 * default timeout for SG_IO if none specified
926 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
927 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
933 unsigned long offset;
938 struct req_iterator {
939 struct bvec_iter iter;
943 /* This should not be used directly - use rq_for_each_segment */
944 #define for_each_bio(_bio) \
945 for (; _bio; _bio = _bio->bi_next)
946 #define __rq_for_each_bio(_bio, rq) \
948 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
950 #define rq_for_each_segment(bvl, _rq, _iter) \
951 __rq_for_each_bio(_iter.bio, _rq) \
952 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
954 #define rq_iter_last(bvec, _iter) \
955 (_iter.bio->bi_next == NULL && \
956 bio_iter_last(bvec, _iter.iter))
958 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
959 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
961 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
962 extern void rq_flush_dcache_pages(struct request *rq);
964 static inline void rq_flush_dcache_pages(struct request *rq)
969 extern int blk_register_queue(struct gendisk *disk);
970 extern void blk_unregister_queue(struct gendisk *disk);
971 extern blk_qc_t generic_make_request(struct bio *bio);
972 extern blk_qc_t direct_make_request(struct bio *bio);
973 extern void blk_rq_init(struct request_queue *q, struct request *rq);
974 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
975 extern void blk_put_request(struct request *);
976 extern void __blk_put_request(struct request_queue *, struct request *);
977 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
978 blk_mq_req_flags_t flags);
979 extern void blk_requeue_request(struct request_queue *, struct request *);
980 extern int blk_lld_busy(struct request_queue *q);
981 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
982 struct bio_set *bs, gfp_t gfp_mask,
983 int (*bio_ctr)(struct bio *, struct bio *, void *),
985 extern void blk_rq_unprep_clone(struct request *rq);
986 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
988 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
989 extern void blk_delay_queue(struct request_queue *, unsigned long);
990 extern void blk_queue_split(struct request_queue *, struct bio **);
991 extern void blk_recount_segments(struct request_queue *, struct bio *);
992 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
993 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
994 unsigned int, void __user *);
995 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
996 unsigned int, void __user *);
997 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
998 struct scsi_ioctl_command __user *);
1000 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
1001 extern void blk_queue_exit(struct request_queue *q);
1002 extern void blk_start_queue(struct request_queue *q);
1003 extern void blk_start_queue_async(struct request_queue *q);
1004 extern void blk_stop_queue(struct request_queue *q);
1005 extern void blk_sync_queue(struct request_queue *q);
1006 extern void __blk_stop_queue(struct request_queue *q);
1007 extern void __blk_run_queue(struct request_queue *q);
1008 extern void __blk_run_queue_uncond(struct request_queue *q);
1009 extern void blk_run_queue(struct request_queue *);
1010 extern void blk_run_queue_async(struct request_queue *q);
1011 extern int blk_rq_map_user(struct request_queue *, struct request *,
1012 struct rq_map_data *, void __user *, unsigned long,
1014 extern int blk_rq_unmap_user(struct bio *);
1015 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1016 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1017 struct rq_map_data *, const struct iov_iter *,
1019 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1020 struct request *, int);
1021 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1022 struct request *, int, rq_end_io_fn *);
1024 int blk_status_to_errno(blk_status_t status);
1025 blk_status_t errno_to_blk_status(int errno);
1027 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1029 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1031 return bdev->bd_disk->queue; /* this is never NULL */
1035 * The basic unit of block I/O is a sector. It is used in a number of contexts
1036 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1037 * bytes. Variables of type sector_t represent an offset or size that is a
1038 * multiple of 512 bytes. Hence these two constants.
1040 #ifndef SECTOR_SHIFT
1041 #define SECTOR_SHIFT 9
1044 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1048 * blk_rq_pos() : the current sector
1049 * blk_rq_bytes() : bytes left in the entire request
1050 * blk_rq_cur_bytes() : bytes left in the current segment
1051 * blk_rq_err_bytes() : bytes left till the next error boundary
1052 * blk_rq_sectors() : sectors left in the entire request
1053 * blk_rq_cur_sectors() : sectors left in the current segment
1055 static inline sector_t blk_rq_pos(const struct request *rq)
1057 return rq->__sector;
1060 static inline unsigned int blk_rq_bytes(const struct request *rq)
1062 return rq->__data_len;
1065 static inline int blk_rq_cur_bytes(const struct request *rq)
1067 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1070 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1072 static inline unsigned int blk_rq_sectors(const struct request *rq)
1074 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1077 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1079 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1082 #ifdef CONFIG_BLK_DEV_ZONED
1083 static inline unsigned int blk_rq_zone_no(struct request *rq)
1085 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1088 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1090 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1092 #endif /* CONFIG_BLK_DEV_ZONED */
1095 * Some commands like WRITE SAME have a payload or data transfer size which
1096 * is different from the size of the request. Any driver that supports such
1097 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1098 * calculate the data transfer size.
1100 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1102 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1103 return rq->special_vec.bv_len;
1104 return blk_rq_bytes(rq);
1107 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1110 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1111 return min(q->limits.max_discard_sectors,
1112 UINT_MAX >> SECTOR_SHIFT);
1114 if (unlikely(op == REQ_OP_WRITE_SAME))
1115 return q->limits.max_write_same_sectors;
1117 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1118 return q->limits.max_write_zeroes_sectors;
1120 return q->limits.max_sectors;
1124 * Return maximum size of a request at given offset. Only valid for
1125 * file system requests.
1127 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1130 if (!q->limits.chunk_sectors)
1131 return q->limits.max_sectors;
1133 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1134 (offset & (q->limits.chunk_sectors - 1))));
1137 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1140 struct request_queue *q = rq->q;
1142 if (blk_rq_is_passthrough(rq))
1143 return q->limits.max_hw_sectors;
1145 if (!q->limits.chunk_sectors ||
1146 req_op(rq) == REQ_OP_DISCARD ||
1147 req_op(rq) == REQ_OP_SECURE_ERASE)
1148 return blk_queue_get_max_sectors(q, req_op(rq));
1150 return min(blk_max_size_offset(q, offset),
1151 blk_queue_get_max_sectors(q, req_op(rq)));
1154 static inline unsigned int blk_rq_count_bios(struct request *rq)
1156 unsigned int nr_bios = 0;
1159 __rq_for_each_bio(bio, rq)
1166 * Request issue related functions.
1168 extern struct request *blk_peek_request(struct request_queue *q);
1169 extern void blk_start_request(struct request *rq);
1170 extern struct request *blk_fetch_request(struct request_queue *q);
1172 void blk_steal_bios(struct bio_list *list, struct request *rq);
1175 * Request completion related functions.
1177 * blk_update_request() completes given number of bytes and updates
1178 * the request without completing it.
1180 * blk_end_request() and friends. __blk_end_request() must be called
1181 * with the request queue spinlock acquired.
1183 * Several drivers define their own end_request and call
1184 * blk_end_request() for parts of the original function.
1185 * This prevents code duplication in drivers.
1187 extern bool blk_update_request(struct request *rq, blk_status_t error,
1188 unsigned int nr_bytes);
1189 extern void blk_finish_request(struct request *rq, blk_status_t error);
1190 extern bool blk_end_request(struct request *rq, blk_status_t error,
1191 unsigned int nr_bytes);
1192 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1193 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1194 unsigned int nr_bytes);
1195 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1196 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1198 extern void blk_complete_request(struct request *);
1199 extern void __blk_complete_request(struct request *);
1200 extern void blk_abort_request(struct request *);
1201 extern void blk_unprep_request(struct request *);
1204 * Access functions for manipulating queue properties
1206 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1207 spinlock_t *lock, int node_id);
1208 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1209 extern int blk_init_allocated_queue(struct request_queue *);
1210 extern void blk_cleanup_queue(struct request_queue *);
1211 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1212 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1213 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1214 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1215 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1216 extern void blk_queue_max_discard_segments(struct request_queue *,
1218 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1219 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1220 unsigned int max_discard_sectors);
1221 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1222 unsigned int max_write_same_sectors);
1223 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1224 unsigned int max_write_same_sectors);
1225 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
1226 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1227 extern void blk_queue_alignment_offset(struct request_queue *q,
1228 unsigned int alignment);
1229 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1230 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1231 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1232 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1233 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1234 extern void blk_set_default_limits(struct queue_limits *lim);
1235 extern void blk_set_stacking_limits(struct queue_limits *lim);
1236 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1238 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1240 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1242 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1243 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1244 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1245 extern int blk_queue_dma_drain(struct request_queue *q,
1246 dma_drain_needed_fn *dma_drain_needed,
1247 void *buf, unsigned int size);
1248 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1249 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1250 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1251 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1252 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1253 extern void blk_queue_dma_alignment(struct request_queue *, int);
1254 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1255 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1256 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1257 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1258 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1259 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1262 * Number of physical segments as sent to the device.
1264 * Normally this is the number of discontiguous data segments sent by the
1265 * submitter. But for data-less command like discard we might have no
1266 * actual data segments submitted, but the driver might have to add it's
1267 * own special payload. In that case we still return 1 here so that this
1268 * special payload will be mapped.
1270 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1272 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1274 return rq->nr_phys_segments;
1278 * Number of discard segments (or ranges) the driver needs to fill in.
1279 * Each discard bio merged into a request is counted as one segment.
1281 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1283 return max_t(unsigned short, rq->nr_phys_segments, 1);
1286 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1287 extern void blk_dump_rq_flags(struct request *, char *);
1288 extern long nr_blockdev_pages(void);
1290 bool __must_check blk_get_queue(struct request_queue *);
1291 struct request_queue *blk_alloc_queue(gfp_t);
1292 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1294 extern void blk_put_queue(struct request_queue *);
1295 extern void blk_set_queue_dying(struct request_queue *);
1298 * block layer runtime pm functions
1301 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev);
1302 extern int blk_pre_runtime_suspend(struct request_queue *q);
1303 extern void blk_post_runtime_suspend(struct request_queue *q, int err);
1304 extern void blk_pre_runtime_resume(struct request_queue *q);
1305 extern void blk_post_runtime_resume(struct request_queue *q, int err);
1306 extern void blk_set_runtime_active(struct request_queue *q);
1308 static inline void blk_pm_runtime_init(struct request_queue *q,
1309 struct device *dev) {}
1310 static inline int blk_pre_runtime_suspend(struct request_queue *q)
1314 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {}
1315 static inline void blk_pre_runtime_resume(struct request_queue *q) {}
1316 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {}
1317 static inline void blk_set_runtime_active(struct request_queue *q) {}
1321 * blk_plug permits building a queue of related requests by holding the I/O
1322 * fragments for a short period. This allows merging of sequential requests
1323 * into single larger request. As the requests are moved from a per-task list to
1324 * the device's request_queue in a batch, this results in improved scalability
1325 * as the lock contention for request_queue lock is reduced.
1327 * It is ok not to disable preemption when adding the request to the plug list
1328 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1329 * the plug list when the task sleeps by itself. For details, please see
1330 * schedule() where blk_schedule_flush_plug() is called.
1333 struct list_head list; /* requests */
1334 struct list_head mq_list; /* blk-mq requests */
1335 struct list_head cb_list; /* md requires an unplug callback */
1337 #define BLK_MAX_REQUEST_COUNT 16
1338 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1341 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1342 struct blk_plug_cb {
1343 struct list_head list;
1344 blk_plug_cb_fn callback;
1347 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1348 void *data, int size);
1349 extern void blk_start_plug(struct blk_plug *);
1350 extern void blk_finish_plug(struct blk_plug *);
1351 extern void blk_flush_plug_list(struct blk_plug *, bool);
1353 static inline void blk_flush_plug(struct task_struct *tsk)
1355 struct blk_plug *plug = tsk->plug;
1358 blk_flush_plug_list(plug, false);
1361 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1363 struct blk_plug *plug = tsk->plug;
1366 blk_flush_plug_list(plug, true);
1369 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1371 struct blk_plug *plug = tsk->plug;
1374 (!list_empty(&plug->list) ||
1375 !list_empty(&plug->mq_list) ||
1376 !list_empty(&plug->cb_list));
1382 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1383 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1384 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1385 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1386 extern void blk_queue_free_tags(struct request_queue *);
1387 extern int blk_queue_resize_tags(struct request_queue *, int);
1388 extern struct blk_queue_tag *blk_init_tags(int, int);
1389 extern void blk_free_tags(struct blk_queue_tag *);
1391 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1394 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1396 return bqt->tag_index[tag];
1399 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1400 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1401 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1403 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1405 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1406 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1407 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1408 sector_t nr_sects, gfp_t gfp_mask, int flags,
1411 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1412 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1414 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1415 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1417 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1418 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1420 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1421 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1423 return blkdev_issue_discard(sb->s_bdev,
1424 block << (sb->s_blocksize_bits -
1426 nr_blocks << (sb->s_blocksize_bits -
1430 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1431 sector_t nr_blocks, gfp_t gfp_mask)
1433 return blkdev_issue_zeroout(sb->s_bdev,
1434 block << (sb->s_blocksize_bits -
1436 nr_blocks << (sb->s_blocksize_bits -
1441 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1443 enum blk_default_limits {
1444 BLK_MAX_SEGMENTS = 128,
1445 BLK_SAFE_MAX_SECTORS = 255,
1446 BLK_DEF_MAX_SECTORS = 2560,
1447 BLK_MAX_SEGMENT_SIZE = 65536,
1448 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1451 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1453 return q->limits.seg_boundary_mask;
1456 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1458 return q->limits.virt_boundary_mask;
1461 static inline unsigned int queue_max_sectors(struct request_queue *q)
1463 return q->limits.max_sectors;
1466 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1468 return q->limits.max_hw_sectors;
1471 static inline unsigned short queue_max_segments(struct request_queue *q)
1473 return q->limits.max_segments;
1476 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1478 return q->limits.max_discard_segments;
1481 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1483 return q->limits.max_segment_size;
1486 static inline unsigned queue_logical_block_size(struct request_queue *q)
1490 if (q && q->limits.logical_block_size)
1491 retval = q->limits.logical_block_size;
1496 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1498 return queue_logical_block_size(bdev_get_queue(bdev));
1501 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1503 return q->limits.physical_block_size;
1506 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1508 return queue_physical_block_size(bdev_get_queue(bdev));
1511 static inline unsigned int queue_io_min(struct request_queue *q)
1513 return q->limits.io_min;
1516 static inline int bdev_io_min(struct block_device *bdev)
1518 return queue_io_min(bdev_get_queue(bdev));
1521 static inline unsigned int queue_io_opt(struct request_queue *q)
1523 return q->limits.io_opt;
1526 static inline int bdev_io_opt(struct block_device *bdev)
1528 return queue_io_opt(bdev_get_queue(bdev));
1531 static inline int queue_alignment_offset(struct request_queue *q)
1533 if (q->limits.misaligned)
1536 return q->limits.alignment_offset;
1539 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1541 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1542 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1545 return (granularity + lim->alignment_offset - alignment) % granularity;
1548 static inline int bdev_alignment_offset(struct block_device *bdev)
1550 struct request_queue *q = bdev_get_queue(bdev);
1552 if (q->limits.misaligned)
1555 if (bdev != bdev->bd_contains)
1556 return bdev->bd_part->alignment_offset;
1558 return q->limits.alignment_offset;
1561 static inline int queue_discard_alignment(struct request_queue *q)
1563 if (q->limits.discard_misaligned)
1566 return q->limits.discard_alignment;
1569 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1571 unsigned int alignment, granularity, offset;
1573 if (!lim->max_discard_sectors)
1576 /* Why are these in bytes, not sectors? */
1577 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1578 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1582 /* Offset of the partition start in 'granularity' sectors */
1583 offset = sector_div(sector, granularity);
1585 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1586 offset = (granularity + alignment - offset) % granularity;
1588 /* Turn it back into bytes, gaah */
1589 return offset << SECTOR_SHIFT;
1592 static inline int bdev_discard_alignment(struct block_device *bdev)
1594 struct request_queue *q = bdev_get_queue(bdev);
1596 if (bdev != bdev->bd_contains)
1597 return bdev->bd_part->discard_alignment;
1599 return q->limits.discard_alignment;
1602 static inline unsigned int bdev_write_same(struct block_device *bdev)
1604 struct request_queue *q = bdev_get_queue(bdev);
1607 return q->limits.max_write_same_sectors;
1612 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1614 struct request_queue *q = bdev_get_queue(bdev);
1617 return q->limits.max_write_zeroes_sectors;
1622 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1624 struct request_queue *q = bdev_get_queue(bdev);
1627 return blk_queue_zoned_model(q);
1629 return BLK_ZONED_NONE;
1632 static inline bool bdev_is_zoned(struct block_device *bdev)
1634 struct request_queue *q = bdev_get_queue(bdev);
1637 return blk_queue_is_zoned(q);
1642 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1644 struct request_queue *q = bdev_get_queue(bdev);
1647 return blk_queue_zone_sectors(q);
1651 static inline int queue_dma_alignment(struct request_queue *q)
1653 return q ? q->dma_alignment : 511;
1656 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1659 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1660 return !(addr & alignment) && !(len & alignment);
1663 /* assumes size > 256 */
1664 static inline unsigned int blksize_bits(unsigned int size)
1666 unsigned int bits = 8;
1670 } while (size > 256);
1674 static inline unsigned int block_size(struct block_device *bdev)
1676 return bdev->bd_block_size;
1679 static inline bool queue_flush_queueable(struct request_queue *q)
1681 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1684 typedef struct {struct page *v;} Sector;
1686 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1688 static inline void put_dev_sector(Sector p)
1693 static inline bool __bvec_gap_to_prev(struct request_queue *q,
1694 struct bio_vec *bprv, unsigned int offset)
1697 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q));
1701 * Check if adding a bio_vec after bprv with offset would create a gap in
1702 * the SG list. Most drivers don't care about this, but some do.
1704 static inline bool bvec_gap_to_prev(struct request_queue *q,
1705 struct bio_vec *bprv, unsigned int offset)
1707 if (!queue_virt_boundary(q))
1709 return __bvec_gap_to_prev(q, bprv, offset);
1713 * Check if the two bvecs from two bios can be merged to one segment.
1714 * If yes, no need to check gap between the two bios since the 1st bio
1715 * and the 1st bvec in the 2nd bio can be handled in one segment.
1717 static inline bool bios_segs_mergeable(struct request_queue *q,
1718 struct bio *prev, struct bio_vec *prev_last_bv,
1719 struct bio_vec *next_first_bv)
1721 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv))
1723 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv))
1725 if (prev->bi_seg_back_size + next_first_bv->bv_len >
1726 queue_max_segment_size(q))
1731 static inline bool bio_will_gap(struct request_queue *q,
1732 struct request *prev_rq,
1736 if (bio_has_data(prev) && queue_virt_boundary(q)) {
1737 struct bio_vec pb, nb;
1740 * don't merge if the 1st bio starts with non-zero
1741 * offset, otherwise it is quite difficult to respect
1742 * sg gap limit. We work hard to merge a huge number of small
1743 * single bios in case of mkfs.
1746 bio_get_first_bvec(prev_rq->bio, &pb);
1748 bio_get_first_bvec(prev, &pb);
1753 * We don't need to worry about the situation that the
1754 * merged segment ends in unaligned virt boundary:
1756 * - if 'pb' ends aligned, the merged segment ends aligned
1757 * - if 'pb' ends unaligned, the next bio must include
1758 * one single bvec of 'nb', otherwise the 'nb' can't
1761 bio_get_last_bvec(prev, &pb);
1762 bio_get_first_bvec(next, &nb);
1764 if (!bios_segs_mergeable(q, prev, &pb, &nb))
1765 return __bvec_gap_to_prev(q, &pb, nb.bv_offset);
1771 static inline bool req_gap_back_merge(struct request *req, struct bio *bio)
1773 return bio_will_gap(req->q, req, req->biotail, bio);
1776 static inline bool req_gap_front_merge(struct request *req, struct bio *bio)
1778 return bio_will_gap(req->q, NULL, bio, req->bio);
1781 int kblockd_schedule_work(struct work_struct *work);
1782 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1783 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1785 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1786 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1787 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1788 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1790 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1792 enum blk_integrity_flags {
1793 BLK_INTEGRITY_VERIFY = 1 << 0,
1794 BLK_INTEGRITY_GENERATE = 1 << 1,
1795 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1796 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1799 struct blk_integrity_iter {
1803 unsigned int data_size;
1804 unsigned short interval;
1805 const char *disk_name;
1808 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1810 struct blk_integrity_profile {
1811 integrity_processing_fn *generate_fn;
1812 integrity_processing_fn *verify_fn;
1816 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1817 extern void blk_integrity_unregister(struct gendisk *);
1818 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1819 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1820 struct scatterlist *);
1821 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1822 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1824 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1827 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1829 struct blk_integrity *bi = &disk->queue->integrity;
1838 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1840 return blk_get_integrity(bdev->bd_disk);
1843 static inline bool blk_integrity_rq(struct request *rq)
1845 return rq->cmd_flags & REQ_INTEGRITY;
1848 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1851 q->limits.max_integrity_segments = segs;
1854 static inline unsigned short
1855 queue_max_integrity_segments(struct request_queue *q)
1857 return q->limits.max_integrity_segments;
1860 static inline bool integrity_req_gap_back_merge(struct request *req,
1863 struct bio_integrity_payload *bip = bio_integrity(req->bio);
1864 struct bio_integrity_payload *bip_next = bio_integrity(next);
1866 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1867 bip_next->bip_vec[0].bv_offset);
1870 static inline bool integrity_req_gap_front_merge(struct request *req,
1873 struct bio_integrity_payload *bip = bio_integrity(bio);
1874 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
1876 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1],
1877 bip_next->bip_vec[0].bv_offset);
1881 * bio_integrity_intervals - Return number of integrity intervals for a bio
1882 * @bi: blk_integrity profile for device
1883 * @sectors: Size of the bio in 512-byte sectors
1885 * Description: The block layer calculates everything in 512 byte
1886 * sectors but integrity metadata is done in terms of the data integrity
1887 * interval size of the storage device. Convert the block layer sectors
1888 * to the appropriate number of integrity intervals.
1890 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1891 unsigned int sectors)
1893 return sectors >> (bi->interval_exp - 9);
1896 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1897 unsigned int sectors)
1899 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1902 #else /* CONFIG_BLK_DEV_INTEGRITY */
1905 struct block_device;
1907 struct blk_integrity;
1909 static inline int blk_integrity_rq(struct request *rq)
1913 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1918 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1920 struct scatterlist *s)
1924 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1928 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1932 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1936 static inline void blk_integrity_register(struct gendisk *d,
1937 struct blk_integrity *b)
1940 static inline void blk_integrity_unregister(struct gendisk *d)
1943 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1947 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1951 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1957 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1964 static inline bool integrity_req_gap_back_merge(struct request *req,
1969 static inline bool integrity_req_gap_front_merge(struct request *req,
1975 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1976 unsigned int sectors)
1981 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1982 unsigned int sectors)
1987 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1989 struct block_device_operations {
1990 int (*open) (struct block_device *, fmode_t);
1991 void (*release) (struct gendisk *, fmode_t);
1992 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1993 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1994 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1995 unsigned int (*check_events) (struct gendisk *disk,
1996 unsigned int clearing);
1997 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1998 int (*media_changed) (struct gendisk *);
1999 void (*unlock_native_capacity) (struct gendisk *);
2000 int (*revalidate_disk) (struct gendisk *);
2001 int (*getgeo)(struct block_device *, struct hd_geometry *);
2002 /* this callback is with swap_lock and sometimes page table lock held */
2003 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
2004 struct module *owner;
2005 const struct pr_ops *pr_ops;
2008 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
2010 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
2011 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
2012 struct writeback_control *);
2014 #ifdef CONFIG_BLK_DEV_ZONED
2015 bool blk_req_needs_zone_write_lock(struct request *rq);
2016 void __blk_req_zone_write_lock(struct request *rq);
2017 void __blk_req_zone_write_unlock(struct request *rq);
2019 static inline void blk_req_zone_write_lock(struct request *rq)
2021 if (blk_req_needs_zone_write_lock(rq))
2022 __blk_req_zone_write_lock(rq);
2025 static inline void blk_req_zone_write_unlock(struct request *rq)
2027 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
2028 __blk_req_zone_write_unlock(rq);
2031 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2033 return rq->q->seq_zones_wlock &&
2034 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
2037 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2039 if (!blk_req_needs_zone_write_lock(rq))
2041 return !blk_req_zone_is_write_locked(rq);
2044 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
2049 static inline void blk_req_zone_write_lock(struct request *rq)
2053 static inline void blk_req_zone_write_unlock(struct request *rq)
2056 static inline bool blk_req_zone_is_write_locked(struct request *rq)
2061 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
2065 #endif /* CONFIG_BLK_DEV_ZONED */
2067 #else /* CONFIG_BLOCK */
2069 struct block_device;
2072 * stubs for when the block layer is configured out
2074 #define buffer_heads_over_limit 0
2076 static inline long nr_blockdev_pages(void)
2084 static inline void blk_start_plug(struct blk_plug *plug)
2088 static inline void blk_finish_plug(struct blk_plug *plug)
2092 static inline void blk_flush_plug(struct task_struct *task)
2096 static inline void blk_schedule_flush_plug(struct task_struct *task)
2101 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
2106 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
2107 sector_t *error_sector)
2112 #endif /* CONFIG_BLOCK */