1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_BLKDEV_H
3 #define _LINUX_BLKDEV_H
5 #include <linux/sched.h>
6 #include <linux/sched/clock.h>
10 #include <linux/major.h>
11 #include <linux/genhd.h>
12 #include <linux/list.h>
13 #include <linux/llist.h>
14 #include <linux/timer.h>
15 #include <linux/workqueue.h>
16 #include <linux/pagemap.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/wait.h>
19 #include <linux/mempool.h>
20 #include <linux/pfn.h>
21 #include <linux/bio.h>
22 #include <linux/stringify.h>
23 #include <linux/gfp.h>
24 #include <linux/bsg.h>
25 #include <linux/smp.h>
26 #include <linux/rcupdate.h>
27 #include <linux/percpu-refcount.h>
28 #include <linux/scatterlist.h>
29 #include <linux/blkzoned.h>
32 struct scsi_ioctl_command;
35 struct elevator_queue;
41 struct blk_flush_queue;
44 struct blk_queue_stats;
45 struct blk_stat_callback;
47 #define BLKDEV_MIN_RQ 4
48 #define BLKDEV_MAX_RQ 128 /* Default maximum */
50 /* Must be consistent with blk_mq_poll_stats_bkt() */
51 #define BLK_MQ_POLL_STATS_BKTS 16
54 * Maximum number of blkcg policies allowed to be registered concurrently.
55 * Defined here to simplify include dependency.
57 #define BLKCG_MAX_POLS 5
59 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
61 #define BLK_RL_SYNCFULL (1U << 0)
62 #define BLK_RL_ASYNCFULL (1U << 1)
65 struct request_queue *q; /* the queue this rl belongs to */
66 #ifdef CONFIG_BLK_CGROUP
67 struct blkcg_gq *blkg; /* blkg this request pool belongs to */
70 * count[], starved[], and wait[] are indexed by
71 * BLK_RW_SYNC/BLK_RW_ASYNC
76 wait_queue_head_t wait[2];
82 typedef __u32 __bitwise req_flags_t;
84 /* elevator knows about this request */
85 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
86 /* drive already may have started this one */
87 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
88 /* uses tagged queueing */
89 #define RQF_QUEUED ((__force req_flags_t)(1 << 2))
90 /* may not be passed by ioscheduler */
91 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
92 /* request for flush sequence */
93 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
94 /* merge of different types, fail separately */
95 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
96 /* track inflight for MQ */
97 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
98 /* don't call prep for this one */
99 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
100 /* set for "ide_preempt" requests and also for requests for which the SCSI
101 "quiesce" state must be ignored. */
102 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
103 /* contains copies of user pages */
104 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
105 /* vaguely specified driver internal error. Ignored by the block layer */
106 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
107 /* don't warn about errors */
108 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
109 /* elevator private data attached */
110 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
111 /* account into disk and partition IO statistics */
112 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
113 /* request came from our alloc pool */
114 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
115 /* runtime pm request */
116 #define RQF_PM ((__force req_flags_t)(1 << 15))
117 /* on IO scheduler merge hash */
118 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
119 /* track IO completion time */
120 #define RQF_STATS ((__force req_flags_t)(1 << 17))
121 /* Look at ->special_vec for the actual data payload instead of the
123 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
124 /* The per-zone write lock is held for this request */
125 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
126 /* already slept for hybrid poll */
127 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20))
128 /* ->timeout has been called, don't expire again */
129 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21))
131 /* flags that prevent us from merging requests: */
132 #define RQF_NOMERGE_FLAGS \
133 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
136 * Request state for blk-mq.
145 * Try to put the fields that are referenced together in the same cacheline.
147 * If you modify this structure, make sure to update blk_rq_init() and
148 * especially blk_mq_rq_ctx_init() to take care of the added fields.
151 struct request_queue *q;
152 struct blk_mq_ctx *mq_ctx;
155 unsigned int cmd_flags; /* op and common flags */
156 req_flags_t rq_flags;
160 /* the following two fields are internal, NEVER access directly */
161 unsigned int __data_len; /* total data len */
163 sector_t __sector; /* sector cursor */
168 struct list_head queuelist;
171 * The hash is used inside the scheduler, and killed once the
172 * request reaches the dispatch list. The ipi_list is only used
173 * to queue the request for softirq completion, which is long
174 * after the request has been unhashed (and even removed from
175 * the dispatch list).
178 struct hlist_node hash; /* merge hash */
179 struct list_head ipi_list;
183 * The rb_node is only used inside the io scheduler, requests
184 * are pruned when moved to the dispatch queue. So let the
185 * completion_data share space with the rb_node.
188 struct rb_node rb_node; /* sort/lookup */
189 struct bio_vec special_vec;
190 void *completion_data;
191 int error_count; /* for legacy drivers, don't use */
195 * Three pointers are available for the IO schedulers, if they need
196 * more they have to dynamically allocate it. Flush requests are
197 * never put on the IO scheduler. So let the flush fields share
198 * space with the elevator data.
208 struct list_head list;
209 rq_end_io_fn *saved_end_io;
213 struct gendisk *rq_disk;
214 struct hd_struct *part;
215 /* Time that I/O was submitted to the kernel. */
217 /* Time that I/O was submitted to the device. */
218 u64 io_start_time_ns;
220 #ifdef CONFIG_BLK_WBT
221 unsigned short wbt_flags;
223 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
224 unsigned short throtl_size;
228 * Number of scatter-gather DMA addr+len pairs after
229 * physical address coalescing is performed.
231 unsigned short nr_phys_segments;
233 #if defined(CONFIG_BLK_DEV_INTEGRITY)
234 unsigned short nr_integrity_segments;
237 unsigned short write_hint;
238 unsigned short ioprio;
240 void *special; /* opaque pointer available for LLD use */
242 unsigned int extra_len; /* length of alignment and padding */
244 enum mq_rq_state state;
247 unsigned int timeout;
249 /* access through blk_rq_set_deadline, blk_rq_deadline */
250 unsigned long __deadline;
252 struct list_head timeout_list;
255 struct __call_single_data csd;
260 * completion callback.
262 rq_end_io_fn *end_io;
266 struct request *next_rq;
268 #ifdef CONFIG_BLK_CGROUP
269 struct request_list *rl; /* rl this rq is alloced from */
273 static inline bool blk_op_is_scsi(unsigned int op)
275 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
278 static inline bool blk_op_is_private(unsigned int op)
280 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
283 static inline bool blk_rq_is_scsi(struct request *rq)
285 return blk_op_is_scsi(req_op(rq));
288 static inline bool blk_rq_is_private(struct request *rq)
290 return blk_op_is_private(req_op(rq));
293 static inline bool blk_rq_is_passthrough(struct request *rq)
295 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
298 static inline bool bio_is_passthrough(struct bio *bio)
300 unsigned op = bio_op(bio);
302 return blk_op_is_scsi(op) || blk_op_is_private(op);
305 static inline unsigned short req_get_ioprio(struct request *req)
310 #include <linux/elevator.h>
312 struct blk_queue_ctx;
314 typedef void (request_fn_proc) (struct request_queue *q);
315 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
316 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t);
317 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
318 typedef void (unprep_rq_fn) (struct request_queue *, struct request *);
321 typedef void (softirq_done_fn)(struct request *);
322 typedef int (dma_drain_needed_fn)(struct request *);
323 typedef int (lld_busy_fn) (struct request_queue *q);
324 typedef int (bsg_job_fn) (struct bsg_job *);
325 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
326 typedef void (exit_rq_fn)(struct request_queue *, struct request *);
328 enum blk_eh_timer_return {
329 BLK_EH_DONE, /* drivers has completed the command */
330 BLK_EH_RESET_TIMER, /* reset timer and try again */
333 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
335 enum blk_queue_state {
340 struct blk_queue_tag {
341 struct request **tag_index; /* map of busy tags */
342 unsigned long *tag_map; /* bit map of free/busy tags */
343 int max_depth; /* what we will send to device */
344 int real_max_depth; /* what the array can hold */
345 atomic_t refcnt; /* map can be shared */
346 int alloc_policy; /* tag allocation policy */
347 int next_tag; /* next tag */
349 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
350 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
352 #define BLK_SCSI_MAX_CMDS (256)
353 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
356 * Zoned block device models (zoned limit).
358 enum blk_zoned_model {
359 BLK_ZONED_NONE, /* Regular block device */
360 BLK_ZONED_HA, /* Host-aware zoned block device */
361 BLK_ZONED_HM, /* Host-managed zoned block device */
364 struct queue_limits {
365 unsigned long bounce_pfn;
366 unsigned long seg_boundary_mask;
367 unsigned long virt_boundary_mask;
369 unsigned int max_hw_sectors;
370 unsigned int max_dev_sectors;
371 unsigned int chunk_sectors;
372 unsigned int max_sectors;
373 unsigned int max_segment_size;
374 unsigned int physical_block_size;
375 unsigned int alignment_offset;
378 unsigned int max_discard_sectors;
379 unsigned int max_hw_discard_sectors;
380 unsigned int max_write_same_sectors;
381 unsigned int max_write_zeroes_sectors;
382 unsigned int discard_granularity;
383 unsigned int discard_alignment;
385 unsigned short logical_block_size;
386 unsigned short max_segments;
387 unsigned short max_integrity_segments;
388 unsigned short max_discard_segments;
390 unsigned char misaligned;
391 unsigned char discard_misaligned;
392 unsigned char cluster;
393 unsigned char raid_partial_stripes_expensive;
394 enum blk_zoned_model zoned;
397 #ifdef CONFIG_BLK_DEV_ZONED
399 struct blk_zone_report_hdr {
400 unsigned int nr_zones;
404 extern unsigned int blkdev_nr_zones(struct block_device *bdev);
405 extern int blkdev_report_zones(struct block_device *bdev,
406 sector_t sector, struct blk_zone *zones,
407 unsigned int *nr_zones, gfp_t gfp_mask);
408 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
409 sector_t nr_sectors, gfp_t gfp_mask);
411 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
412 unsigned int cmd, unsigned long arg);
413 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
414 unsigned int cmd, unsigned long arg);
416 #else /* CONFIG_BLK_DEV_ZONED */
418 static inline unsigned int blkdev_nr_zones(struct block_device *bdev)
422 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
423 fmode_t mode, unsigned int cmd,
429 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
430 fmode_t mode, unsigned int cmd,
436 #endif /* CONFIG_BLK_DEV_ZONED */
438 struct request_queue {
440 * Together with queue_head for cacheline sharing
442 struct list_head queue_head;
443 struct request *last_merge;
444 struct elevator_queue *elevator;
445 int nr_rqs[2]; /* # allocated [a]sync rqs */
446 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */
448 struct blk_queue_stats *stats;
449 struct rq_qos *rq_qos;
452 * If blkcg is not used, @q->root_rl serves all requests. If blkcg
453 * is used, root blkg allocates from @q->root_rl and all other
454 * blkgs from their own blkg->rl. Which one to use should be
455 * determined using bio_request_list().
457 struct request_list root_rl;
459 request_fn_proc *request_fn;
460 make_request_fn *make_request_fn;
462 prep_rq_fn *prep_rq_fn;
463 unprep_rq_fn *unprep_rq_fn;
464 softirq_done_fn *softirq_done_fn;
465 rq_timed_out_fn *rq_timed_out_fn;
466 dma_drain_needed_fn *dma_drain_needed;
467 lld_busy_fn *lld_busy_fn;
468 /* Called just after a request is allocated */
469 init_rq_fn *init_rq_fn;
470 /* Called just before a request is freed */
471 exit_rq_fn *exit_rq_fn;
472 /* Called from inside blk_get_request() */
473 void (*initialize_rq_fn)(struct request *rq);
475 const struct blk_mq_ops *mq_ops;
477 unsigned int *mq_map;
480 struct blk_mq_ctx __percpu *queue_ctx;
481 unsigned int nr_queues;
483 unsigned int queue_depth;
485 /* hw dispatch queues */
486 struct blk_mq_hw_ctx **queue_hw_ctx;
487 unsigned int nr_hw_queues;
490 * Dispatch queue sorting
493 struct request *boundary_rq;
496 * Delayed queue handling
498 struct delayed_work delay_work;
500 struct backing_dev_info *backing_dev_info;
503 * The queue owner gets to use this for whatever they like.
504 * ll_rw_blk doesn't touch it.
509 * various queue flags, see QUEUE_* below
511 unsigned long queue_flags;
513 * Number of contexts that have called blk_set_pm_only(). If this
514 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
520 * ida allocated id for this queue. Used to index queues from
526 * queue needs bounce pages for pages above this limit
531 * protects queue structures from reentrancy. ->__queue_lock should
532 * _never_ be used directly, it is queue private. always use
535 spinlock_t __queue_lock;
536 spinlock_t *queue_lock;
546 struct kobject mq_kobj;
548 #ifdef CONFIG_BLK_DEV_INTEGRITY
549 struct blk_integrity integrity;
550 #endif /* CONFIG_BLK_DEV_INTEGRITY */
555 unsigned int nr_pending;
561 unsigned long nr_requests; /* Max # of requests */
562 unsigned int nr_congestion_on;
563 unsigned int nr_congestion_off;
564 unsigned int nr_batching;
566 unsigned int dma_drain_size;
567 void *dma_drain_buffer;
568 unsigned int dma_pad_mask;
569 unsigned int dma_alignment;
571 struct blk_queue_tag *queue_tags;
573 unsigned int nr_sorted;
574 unsigned int in_flight[2];
577 * Number of active block driver functions for which blk_drain_queue()
578 * must wait. Must be incremented around functions that unlock the
579 * queue_lock internally, e.g. scsi_request_fn().
581 unsigned int request_fn_active;
583 unsigned int rq_timeout;
586 struct blk_stat_callback *poll_cb;
587 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
589 struct timer_list timeout;
590 struct work_struct timeout_work;
591 struct list_head timeout_list;
593 struct list_head icq_list;
594 #ifdef CONFIG_BLK_CGROUP
595 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
596 struct blkcg_gq *root_blkg;
597 struct list_head blkg_list;
600 struct queue_limits limits;
602 #ifdef CONFIG_BLK_DEV_ZONED
604 * Zoned block device information for request dispatch control.
605 * nr_zones is the total number of zones of the device. This is always
606 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
607 * bits which indicates if a zone is conventional (bit clear) or
608 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
609 * bits which indicates if a zone is write locked, that is, if a write
610 * request targeting the zone was dispatched. All three fields are
611 * initialized by the low level device driver (e.g. scsi/sd.c).
612 * Stacking drivers (device mappers) may or may not initialize
615 * Reads of this information must be protected with blk_queue_enter() /
616 * blk_queue_exit(). Modifying this information is only allowed while
617 * no requests are being processed. See also blk_mq_freeze_queue() and
618 * blk_mq_unfreeze_queue().
620 unsigned int nr_zones;
621 unsigned long *seq_zones_bitmap;
622 unsigned long *seq_zones_wlock;
623 #endif /* CONFIG_BLK_DEV_ZONED */
628 unsigned int sg_timeout;
629 unsigned int sg_reserved_size;
631 #ifdef CONFIG_BLK_DEV_IO_TRACE
632 struct blk_trace *blk_trace;
633 struct mutex blk_trace_mutex;
636 * for flush operations
638 struct blk_flush_queue *fq;
640 struct list_head requeue_list;
641 spinlock_t requeue_lock;
642 struct delayed_work requeue_work;
644 struct mutex sysfs_lock;
647 atomic_t mq_freeze_depth;
649 #if defined(CONFIG_BLK_DEV_BSG)
650 bsg_job_fn *bsg_job_fn;
651 struct bsg_class_device bsg_dev;
654 #ifdef CONFIG_BLK_DEV_THROTTLING
656 struct throtl_data *td;
658 struct rcu_head rcu_head;
659 wait_queue_head_t mq_freeze_wq;
660 struct percpu_ref q_usage_counter;
661 struct list_head all_q_node;
663 struct blk_mq_tag_set *tag_set;
664 struct list_head tag_set_list;
665 struct bio_set bio_split;
667 #ifdef CONFIG_BLK_DEBUG_FS
668 struct dentry *debugfs_dir;
669 struct dentry *sched_debugfs_dir;
672 bool mq_sysfs_init_done;
677 struct work_struct release_work;
679 #define BLK_MAX_WRITE_HINTS 5
680 u64 write_hints[BLK_MAX_WRITE_HINTS];
683 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */
684 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */
685 #define QUEUE_FLAG_DYING 2 /* queue being torn down */
686 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */
687 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */
688 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */
689 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */
690 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */
691 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */
692 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
693 #define QUEUE_FLAG_IO_STAT 10 /* do disk/partitions IO accounting */
694 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */
695 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */
696 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */
697 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */
698 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */
699 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */
700 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */
701 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/
702 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */
703 #define QUEUE_FLAG_WC 20 /* Write back caching */
704 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */
705 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */
706 #define QUEUE_FLAG_DAX 23 /* device supports DAX */
707 #define QUEUE_FLAG_STATS 24 /* track IO start and completion times */
708 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */
709 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */
710 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */
711 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */
712 #define QUEUE_FLAG_PCI_P2PDMA 29 /* device supports PCI p2p requests */
714 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
715 (1 << QUEUE_FLAG_SAME_COMP) | \
716 (1 << QUEUE_FLAG_ADD_RANDOM))
718 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
719 (1 << QUEUE_FLAG_SAME_COMP) | \
720 (1 << QUEUE_FLAG_POLL))
722 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
723 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
724 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
725 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q);
727 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
728 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
729 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
730 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
731 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
732 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
733 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
734 #define blk_queue_noxmerges(q) \
735 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
736 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
737 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
738 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
739 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
740 #define blk_queue_secure_erase(q) \
741 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
742 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
743 #define blk_queue_scsi_passthrough(q) \
744 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
745 #define blk_queue_pci_p2pdma(q) \
746 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
748 #define blk_noretry_request(rq) \
749 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
750 REQ_FAILFAST_DRIVER))
751 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
752 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
753 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
755 extern void blk_set_pm_only(struct request_queue *q);
756 extern void blk_clear_pm_only(struct request_queue *q);
758 static inline int queue_in_flight(struct request_queue *q)
760 return q->in_flight[0] + q->in_flight[1];
763 static inline bool blk_account_rq(struct request *rq)
765 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
768 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
769 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
770 /* rq->queuelist of dequeued request must be list_empty() */
771 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
773 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
775 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
778 * Driver can handle struct request, if it either has an old style
779 * request_fn defined, or is blk-mq based.
781 static inline bool queue_is_rq_based(struct request_queue *q)
783 return q->request_fn || q->mq_ops;
786 static inline unsigned int blk_queue_cluster(struct request_queue *q)
788 return q->limits.cluster;
791 static inline enum blk_zoned_model
792 blk_queue_zoned_model(struct request_queue *q)
794 return q->limits.zoned;
797 static inline bool blk_queue_is_zoned(struct request_queue *q)
799 switch (blk_queue_zoned_model(q)) {
808 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q)
810 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
813 #ifdef CONFIG_BLK_DEV_ZONED
814 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
816 return blk_queue_is_zoned(q) ? q->nr_zones : 0;
819 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
822 if (!blk_queue_is_zoned(q))
824 return sector >> ilog2(q->limits.chunk_sectors);
827 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
830 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
832 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
834 #else /* CONFIG_BLK_DEV_ZONED */
835 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
839 #endif /* CONFIG_BLK_DEV_ZONED */
841 static inline bool rq_is_sync(struct request *rq)
843 return op_is_sync(rq->cmd_flags);
846 static inline bool blk_rl_full(struct request_list *rl, bool sync)
848 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
850 return rl->flags & flag;
853 static inline void blk_set_rl_full(struct request_list *rl, bool sync)
855 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
860 static inline void blk_clear_rl_full(struct request_list *rl, bool sync)
862 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL;
867 static inline bool rq_mergeable(struct request *rq)
869 if (blk_rq_is_passthrough(rq))
872 if (req_op(rq) == REQ_OP_FLUSH)
875 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
878 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
880 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
886 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
888 if (bio_page(a) == bio_page(b) &&
889 bio_offset(a) == bio_offset(b))
895 static inline unsigned int blk_queue_depth(struct request_queue *q)
898 return q->queue_depth;
900 return q->nr_requests;
904 * q->prep_rq_fn return values
907 BLKPREP_OK, /* serve it */
908 BLKPREP_KILL, /* fatal error, kill, return -EIO */
909 BLKPREP_DEFER, /* leave on queue */
910 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */
913 extern unsigned long blk_max_low_pfn, blk_max_pfn;
916 * standard bounce addresses:
918 * BLK_BOUNCE_HIGH : bounce all highmem pages
919 * BLK_BOUNCE_ANY : don't bounce anything
920 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
923 #if BITS_PER_LONG == 32
924 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
926 #define BLK_BOUNCE_HIGH -1ULL
928 #define BLK_BOUNCE_ANY (-1ULL)
929 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
932 * default timeout for SG_IO if none specified
934 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
935 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
941 unsigned long offset;
946 struct req_iterator {
947 struct bvec_iter iter;
951 /* This should not be used directly - use rq_for_each_segment */
952 #define for_each_bio(_bio) \
953 for (; _bio; _bio = _bio->bi_next)
954 #define __rq_for_each_bio(_bio, rq) \
956 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
958 #define rq_for_each_segment(bvl, _rq, _iter) \
959 __rq_for_each_bio(_iter.bio, _rq) \
960 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
962 #define rq_iter_last(bvec, _iter) \
963 (_iter.bio->bi_next == NULL && \
964 bio_iter_last(bvec, _iter.iter))
966 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
967 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
969 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
970 extern void rq_flush_dcache_pages(struct request *rq);
972 static inline void rq_flush_dcache_pages(struct request *rq)
977 extern int blk_register_queue(struct gendisk *disk);
978 extern void blk_unregister_queue(struct gendisk *disk);
979 extern blk_qc_t generic_make_request(struct bio *bio);
980 extern blk_qc_t direct_make_request(struct bio *bio);
981 extern void blk_rq_init(struct request_queue *q, struct request *rq);
982 extern void blk_init_request_from_bio(struct request *req, struct bio *bio);
983 extern void blk_put_request(struct request *);
984 extern void __blk_put_request(struct request_queue *, struct request *);
985 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
986 blk_mq_req_flags_t flags);
987 extern void blk_requeue_request(struct request_queue *, struct request *);
988 extern int blk_lld_busy(struct request_queue *q);
989 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
990 struct bio_set *bs, gfp_t gfp_mask,
991 int (*bio_ctr)(struct bio *, struct bio *, void *),
993 extern void blk_rq_unprep_clone(struct request *rq);
994 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
996 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
997 extern void blk_delay_queue(struct request_queue *, unsigned long);
998 extern void blk_queue_split(struct request_queue *, struct bio **);
999 extern void blk_recount_segments(struct request_queue *, struct bio *);
1000 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
1001 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
1002 unsigned int, void __user *);
1003 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
1004 unsigned int, void __user *);
1005 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
1006 struct scsi_ioctl_command __user *);
1008 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
1009 extern void blk_queue_exit(struct request_queue *q);
1010 extern void blk_start_queue(struct request_queue *q);
1011 extern void blk_start_queue_async(struct request_queue *q);
1012 extern void blk_stop_queue(struct request_queue *q);
1013 extern void blk_sync_queue(struct request_queue *q);
1014 extern void __blk_stop_queue(struct request_queue *q);
1015 extern void __blk_run_queue(struct request_queue *q);
1016 extern void __blk_run_queue_uncond(struct request_queue *q);
1017 extern void blk_run_queue(struct request_queue *);
1018 extern void blk_run_queue_async(struct request_queue *q);
1019 extern int blk_rq_map_user(struct request_queue *, struct request *,
1020 struct rq_map_data *, void __user *, unsigned long,
1022 extern int blk_rq_unmap_user(struct bio *);
1023 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
1024 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
1025 struct rq_map_data *, const struct iov_iter *,
1027 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
1028 struct request *, int);
1029 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
1030 struct request *, int, rq_end_io_fn *);
1032 int blk_status_to_errno(blk_status_t status);
1033 blk_status_t errno_to_blk_status(int errno);
1035 bool blk_poll(struct request_queue *q, blk_qc_t cookie);
1037 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1039 return bdev->bd_disk->queue; /* this is never NULL */
1043 * The basic unit of block I/O is a sector. It is used in a number of contexts
1044 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
1045 * bytes. Variables of type sector_t represent an offset or size that is a
1046 * multiple of 512 bytes. Hence these two constants.
1048 #ifndef SECTOR_SHIFT
1049 #define SECTOR_SHIFT 9
1052 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
1056 * blk_rq_pos() : the current sector
1057 * blk_rq_bytes() : bytes left in the entire request
1058 * blk_rq_cur_bytes() : bytes left in the current segment
1059 * blk_rq_err_bytes() : bytes left till the next error boundary
1060 * blk_rq_sectors() : sectors left in the entire request
1061 * blk_rq_cur_sectors() : sectors left in the current segment
1063 static inline sector_t blk_rq_pos(const struct request *rq)
1065 return rq->__sector;
1068 static inline unsigned int blk_rq_bytes(const struct request *rq)
1070 return rq->__data_len;
1073 static inline int blk_rq_cur_bytes(const struct request *rq)
1075 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
1078 extern unsigned int blk_rq_err_bytes(const struct request *rq);
1080 static inline unsigned int blk_rq_sectors(const struct request *rq)
1082 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
1085 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
1087 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
1090 #ifdef CONFIG_BLK_DEV_ZONED
1091 static inline unsigned int blk_rq_zone_no(struct request *rq)
1093 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
1096 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
1098 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
1100 #endif /* CONFIG_BLK_DEV_ZONED */
1103 * Some commands like WRITE SAME have a payload or data transfer size which
1104 * is different from the size of the request. Any driver that supports such
1105 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
1106 * calculate the data transfer size.
1108 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
1110 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1111 return rq->special_vec.bv_len;
1112 return blk_rq_bytes(rq);
1115 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
1118 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
1119 return min(q->limits.max_discard_sectors,
1120 UINT_MAX >> SECTOR_SHIFT);
1122 if (unlikely(op == REQ_OP_WRITE_SAME))
1123 return q->limits.max_write_same_sectors;
1125 if (unlikely(op == REQ_OP_WRITE_ZEROES))
1126 return q->limits.max_write_zeroes_sectors;
1128 return q->limits.max_sectors;
1132 * Return maximum size of a request at given offset. Only valid for
1133 * file system requests.
1135 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1138 if (!q->limits.chunk_sectors)
1139 return q->limits.max_sectors;
1141 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1142 (offset & (q->limits.chunk_sectors - 1))));
1145 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1148 struct request_queue *q = rq->q;
1150 if (blk_rq_is_passthrough(rq))
1151 return q->limits.max_hw_sectors;
1153 if (!q->limits.chunk_sectors ||
1154 req_op(rq) == REQ_OP_DISCARD ||
1155 req_op(rq) == REQ_OP_SECURE_ERASE)
1156 return blk_queue_get_max_sectors(q, req_op(rq));
1158 return min(blk_max_size_offset(q, offset),
1159 blk_queue_get_max_sectors(q, req_op(rq)));
1162 static inline unsigned int blk_rq_count_bios(struct request *rq)
1164 unsigned int nr_bios = 0;
1167 __rq_for_each_bio(bio, rq)
1174 * Request issue related functions.
1176 extern struct request *blk_peek_request(struct request_queue *q);
1177 extern void blk_start_request(struct request *rq);
1178 extern struct request *blk_fetch_request(struct request_queue *q);
1180 void blk_steal_bios(struct bio_list *list, struct request *rq);
1183 * Request completion related functions.
1185 * blk_update_request() completes given number of bytes and updates
1186 * the request without completing it.
1188 * blk_end_request() and friends. __blk_end_request() must be called
1189 * with the request queue spinlock acquired.
1191 * Several drivers define their own end_request and call
1192 * blk_end_request() for parts of the original function.
1193 * This prevents code duplication in drivers.
1195 extern bool blk_update_request(struct request *rq, blk_status_t error,
1196 unsigned int nr_bytes);
1197 extern void blk_finish_request(struct request *rq, blk_status_t error);
1198 extern bool blk_end_request(struct request *rq, blk_status_t error,
1199 unsigned int nr_bytes);
1200 extern void blk_end_request_all(struct request *rq, blk_status_t error);
1201 extern bool __blk_end_request(struct request *rq, blk_status_t error,
1202 unsigned int nr_bytes);
1203 extern void __blk_end_request_all(struct request *rq, blk_status_t error);
1204 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error);
1206 extern void blk_complete_request(struct request *);
1207 extern void __blk_complete_request(struct request *);
1208 extern void blk_abort_request(struct request *);
1209 extern void blk_unprep_request(struct request *);
1212 * Access functions for manipulating queue properties
1214 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
1215 spinlock_t *lock, int node_id);
1216 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
1217 extern int blk_init_allocated_queue(struct request_queue *);
1218 extern void blk_cleanup_queue(struct request_queue *);
1219 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1220 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1221 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1222 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1223 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1224 extern void blk_queue_max_discard_segments(struct request_queue *,
1226 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1227 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1228 unsigned int max_discard_sectors);
1229 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1230 unsigned int max_write_same_sectors);
1231 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1232 unsigned int max_write_same_sectors);
1233 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1234 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1235 extern void blk_queue_alignment_offset(struct request_queue *q,
1236 unsigned int alignment);
1237 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1238 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1239 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1240 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1241 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1242 extern void blk_set_default_limits(struct queue_limits *lim);
1243 extern void blk_set_stacking_limits(struct queue_limits *lim);
1244 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1246 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1248 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1250 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1251 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
1252 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1253 extern int blk_queue_dma_drain(struct request_queue *q,
1254 dma_drain_needed_fn *dma_drain_needed,
1255 void *buf, unsigned int size);
1256 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
1257 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1258 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1259 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
1260 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn);
1261 extern void blk_queue_dma_alignment(struct request_queue *, int);
1262 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1263 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
1264 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
1265 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1266 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable);
1267 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1270 * Number of physical segments as sent to the device.
1272 * Normally this is the number of discontiguous data segments sent by the
1273 * submitter. But for data-less command like discard we might have no
1274 * actual data segments submitted, but the driver might have to add it's
1275 * own special payload. In that case we still return 1 here so that this
1276 * special payload will be mapped.
1278 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1280 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1282 return rq->nr_phys_segments;
1286 * Number of discard segments (or ranges) the driver needs to fill in.
1287 * Each discard bio merged into a request is counted as one segment.
1289 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1291 return max_t(unsigned short, rq->nr_phys_segments, 1);
1294 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1295 extern void blk_dump_rq_flags(struct request *, char *);
1296 extern long nr_blockdev_pages(void);
1298 bool __must_check blk_get_queue(struct request_queue *);
1299 struct request_queue *blk_alloc_queue(gfp_t);
1300 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id,
1302 extern void blk_put_queue(struct request_queue *);
1303 extern void blk_set_queue_dying(struct request_queue *);
1306 * blk_plug permits building a queue of related requests by holding the I/O
1307 * fragments for a short period. This allows merging of sequential requests
1308 * into single larger request. As the requests are moved from a per-task list to
1309 * the device's request_queue in a batch, this results in improved scalability
1310 * as the lock contention for request_queue lock is reduced.
1312 * It is ok not to disable preemption when adding the request to the plug list
1313 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1314 * the plug list when the task sleeps by itself. For details, please see
1315 * schedule() where blk_schedule_flush_plug() is called.
1318 struct list_head list; /* requests */
1319 struct list_head mq_list; /* blk-mq requests */
1320 struct list_head cb_list; /* md requires an unplug callback */
1322 #define BLK_MAX_REQUEST_COUNT 16
1323 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1326 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1327 struct blk_plug_cb {
1328 struct list_head list;
1329 blk_plug_cb_fn callback;
1332 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1333 void *data, int size);
1334 extern void blk_start_plug(struct blk_plug *);
1335 extern void blk_finish_plug(struct blk_plug *);
1336 extern void blk_flush_plug_list(struct blk_plug *, bool);
1338 static inline void blk_flush_plug(struct task_struct *tsk)
1340 struct blk_plug *plug = tsk->plug;
1343 blk_flush_plug_list(plug, false);
1346 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1348 struct blk_plug *plug = tsk->plug;
1351 blk_flush_plug_list(plug, true);
1354 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1356 struct blk_plug *plug = tsk->plug;
1359 (!list_empty(&plug->list) ||
1360 !list_empty(&plug->mq_list) ||
1361 !list_empty(&plug->cb_list));
1367 extern int blk_queue_start_tag(struct request_queue *, struct request *);
1368 extern struct request *blk_queue_find_tag(struct request_queue *, int);
1369 extern void blk_queue_end_tag(struct request_queue *, struct request *);
1370 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int);
1371 extern void blk_queue_free_tags(struct request_queue *);
1372 extern int blk_queue_resize_tags(struct request_queue *, int);
1373 extern struct blk_queue_tag *blk_init_tags(int, int);
1374 extern void blk_free_tags(struct blk_queue_tag *);
1376 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
1379 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
1381 return bqt->tag_index[tag];
1384 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1385 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1386 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1388 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1390 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1391 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1392 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1393 sector_t nr_sects, gfp_t gfp_mask, int flags,
1396 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1397 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1399 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1400 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1402 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1403 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1405 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1406 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1408 return blkdev_issue_discard(sb->s_bdev,
1409 block << (sb->s_blocksize_bits -
1411 nr_blocks << (sb->s_blocksize_bits -
1415 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1416 sector_t nr_blocks, gfp_t gfp_mask)
1418 return blkdev_issue_zeroout(sb->s_bdev,
1419 block << (sb->s_blocksize_bits -
1421 nr_blocks << (sb->s_blocksize_bits -
1426 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1428 enum blk_default_limits {
1429 BLK_MAX_SEGMENTS = 128,
1430 BLK_SAFE_MAX_SECTORS = 255,
1431 BLK_DEF_MAX_SECTORS = 2560,
1432 BLK_MAX_SEGMENT_SIZE = 65536,
1433 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1436 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1438 return q->limits.seg_boundary_mask;
1441 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1443 return q->limits.virt_boundary_mask;
1446 static inline unsigned int queue_max_sectors(struct request_queue *q)
1448 return q->limits.max_sectors;
1451 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1453 return q->limits.max_hw_sectors;
1456 static inline unsigned short queue_max_segments(struct request_queue *q)
1458 return q->limits.max_segments;
1461 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1463 return q->limits.max_discard_segments;
1466 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1468 return q->limits.max_segment_size;
1471 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1475 if (q && q->limits.logical_block_size)
1476 retval = q->limits.logical_block_size;
1481 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1483 return queue_logical_block_size(bdev_get_queue(bdev));
1486 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1488 return q->limits.physical_block_size;
1491 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1493 return queue_physical_block_size(bdev_get_queue(bdev));
1496 static inline unsigned int queue_io_min(struct request_queue *q)
1498 return q->limits.io_min;
1501 static inline int bdev_io_min(struct block_device *bdev)
1503 return queue_io_min(bdev_get_queue(bdev));
1506 static inline unsigned int queue_io_opt(struct request_queue *q)
1508 return q->limits.io_opt;
1511 static inline int bdev_io_opt(struct block_device *bdev)
1513 return queue_io_opt(bdev_get_queue(bdev));
1516 static inline int queue_alignment_offset(struct request_queue *q)
1518 if (q->limits.misaligned)
1521 return q->limits.alignment_offset;
1524 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1526 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1527 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1530 return (granularity + lim->alignment_offset - alignment) % granularity;
1533 static inline int bdev_alignment_offset(struct block_device *bdev)
1535 struct request_queue *q = bdev_get_queue(bdev);
1537 if (q->limits.misaligned)
1540 if (bdev != bdev->bd_contains)
1541 return bdev->bd_part->alignment_offset;
1543 return q->limits.alignment_offset;
1546 static inline int queue_discard_alignment(struct request_queue *q)
1548 if (q->limits.discard_misaligned)
1551 return q->limits.discard_alignment;
1554 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1556 unsigned int alignment, granularity, offset;
1558 if (!lim->max_discard_sectors)
1561 /* Why are these in bytes, not sectors? */
1562 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1563 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1567 /* Offset of the partition start in 'granularity' sectors */
1568 offset = sector_div(sector, granularity);
1570 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1571 offset = (granularity + alignment - offset) % granularity;
1573 /* Turn it back into bytes, gaah */
1574 return offset << SECTOR_SHIFT;
1577 static inline int bdev_discard_alignment(struct block_device *bdev)
1579 struct request_queue *q = bdev_get_queue(bdev);
1581 if (bdev != bdev->bd_contains)
1582 return bdev->bd_part->discard_alignment;
1584 return q->limits.discard_alignment;
1587 static inline unsigned int bdev_write_same(struct block_device *bdev)
1589 struct request_queue *q = bdev_get_queue(bdev);
1592 return q->limits.max_write_same_sectors;
1597 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1599 struct request_queue *q = bdev_get_queue(bdev);
1602 return q->limits.max_write_zeroes_sectors;
1607 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1609 struct request_queue *q = bdev_get_queue(bdev);
1612 return blk_queue_zoned_model(q);
1614 return BLK_ZONED_NONE;
1617 static inline bool bdev_is_zoned(struct block_device *bdev)
1619 struct request_queue *q = bdev_get_queue(bdev);
1622 return blk_queue_is_zoned(q);
1627 static inline unsigned int bdev_zone_sectors(struct block_device *bdev)
1629 struct request_queue *q = bdev_get_queue(bdev);
1632 return blk_queue_zone_sectors(q);
1636 static inline int queue_dma_alignment(struct request_queue *q)
1638 return q ? q->dma_alignment : 511;
1641 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1644 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1645 return !(addr & alignment) && !(len & alignment);
1648 /* assumes size > 256 */
1649 static inline unsigned int blksize_bits(unsigned int size)
1651 unsigned int bits = 8;
1655 } while (size > 256);
1659 static inline unsigned int block_size(struct block_device *bdev)
1661 return bdev->bd_block_size;
1664 static inline bool queue_flush_queueable(struct request_queue *q)
1666 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags);
1669 typedef struct {struct page *v;} Sector;
1671 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1673 static inline void put_dev_sector(Sector p)
1678 int kblockd_schedule_work(struct work_struct *work);
1679 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1680 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1682 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1683 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1684 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1685 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1687 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1689 enum blk_integrity_flags {
1690 BLK_INTEGRITY_VERIFY = 1 << 0,
1691 BLK_INTEGRITY_GENERATE = 1 << 1,
1692 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1693 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1696 struct blk_integrity_iter {
1700 unsigned int data_size;
1701 unsigned short interval;
1702 const char *disk_name;
1705 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1707 struct blk_integrity_profile {
1708 integrity_processing_fn *generate_fn;
1709 integrity_processing_fn *verify_fn;
1713 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1714 extern void blk_integrity_unregister(struct gendisk *);
1715 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1716 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1717 struct scatterlist *);
1718 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1719 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1721 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1724 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1726 struct blk_integrity *bi = &disk->queue->integrity;
1735 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1737 return blk_get_integrity(bdev->bd_disk);
1740 static inline bool blk_integrity_rq(struct request *rq)
1742 return rq->cmd_flags & REQ_INTEGRITY;
1745 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1748 q->limits.max_integrity_segments = segs;
1751 static inline unsigned short
1752 queue_max_integrity_segments(struct request_queue *q)
1754 return q->limits.max_integrity_segments;
1758 * bio_integrity_intervals - Return number of integrity intervals for a bio
1759 * @bi: blk_integrity profile for device
1760 * @sectors: Size of the bio in 512-byte sectors
1762 * Description: The block layer calculates everything in 512 byte
1763 * sectors but integrity metadata is done in terms of the data integrity
1764 * interval size of the storage device. Convert the block layer sectors
1765 * to the appropriate number of integrity intervals.
1767 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1768 unsigned int sectors)
1770 return sectors >> (bi->interval_exp - 9);
1773 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1774 unsigned int sectors)
1776 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1779 #else /* CONFIG_BLK_DEV_INTEGRITY */
1782 struct block_device;
1784 struct blk_integrity;
1786 static inline int blk_integrity_rq(struct request *rq)
1790 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1795 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1797 struct scatterlist *s)
1801 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1805 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1809 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1813 static inline void blk_integrity_register(struct gendisk *d,
1814 struct blk_integrity *b)
1817 static inline void blk_integrity_unregister(struct gendisk *d)
1820 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1824 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1828 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1834 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1841 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1842 unsigned int sectors)
1847 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1848 unsigned int sectors)
1853 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1855 struct block_device_operations {
1856 int (*open) (struct block_device *, fmode_t);
1857 void (*release) (struct gendisk *, fmode_t);
1858 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1859 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1860 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1861 unsigned int (*check_events) (struct gendisk *disk,
1862 unsigned int clearing);
1863 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1864 int (*media_changed) (struct gendisk *);
1865 void (*unlock_native_capacity) (struct gendisk *);
1866 int (*revalidate_disk) (struct gendisk *);
1867 int (*getgeo)(struct block_device *, struct hd_geometry *);
1868 /* this callback is with swap_lock and sometimes page table lock held */
1869 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1870 struct module *owner;
1871 const struct pr_ops *pr_ops;
1874 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1876 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1877 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1878 struct writeback_control *);
1880 #ifdef CONFIG_BLK_DEV_ZONED
1881 bool blk_req_needs_zone_write_lock(struct request *rq);
1882 void __blk_req_zone_write_lock(struct request *rq);
1883 void __blk_req_zone_write_unlock(struct request *rq);
1885 static inline void blk_req_zone_write_lock(struct request *rq)
1887 if (blk_req_needs_zone_write_lock(rq))
1888 __blk_req_zone_write_lock(rq);
1891 static inline void blk_req_zone_write_unlock(struct request *rq)
1893 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1894 __blk_req_zone_write_unlock(rq);
1897 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1899 return rq->q->seq_zones_wlock &&
1900 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1903 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1905 if (!blk_req_needs_zone_write_lock(rq))
1907 return !blk_req_zone_is_write_locked(rq);
1910 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1915 static inline void blk_req_zone_write_lock(struct request *rq)
1919 static inline void blk_req_zone_write_unlock(struct request *rq)
1922 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1927 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1931 #endif /* CONFIG_BLK_DEV_ZONED */
1933 #else /* CONFIG_BLOCK */
1935 struct block_device;
1938 * stubs for when the block layer is configured out
1940 #define buffer_heads_over_limit 0
1942 static inline long nr_blockdev_pages(void)
1950 static inline void blk_start_plug(struct blk_plug *plug)
1954 static inline void blk_finish_plug(struct blk_plug *plug)
1958 static inline void blk_flush_plug(struct task_struct *task)
1962 static inline void blk_schedule_flush_plug(struct task_struct *task)
1967 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1972 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1973 sector_t *error_sector)
1978 #endif /* CONFIG_BLOCK */