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
53 /* Doing classic polling */
54 #define BLK_MQ_POLL_CLASSIC -1
57 * Maximum number of blkcg policies allowed to be registered concurrently.
58 * Defined here to simplify include dependency.
60 #define BLKCG_MAX_POLS 5
62 typedef void (rq_end_io_fn)(struct request *, blk_status_t);
66 typedef __u32 __bitwise req_flags_t;
68 /* elevator knows about this request */
69 #define RQF_SORTED ((__force req_flags_t)(1 << 0))
70 /* drive already may have started this one */
71 #define RQF_STARTED ((__force req_flags_t)(1 << 1))
72 /* may not be passed by ioscheduler */
73 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3))
74 /* request for flush sequence */
75 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4))
76 /* merge of different types, fail separately */
77 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5))
78 /* track inflight for MQ */
79 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6))
80 /* don't call prep for this one */
81 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7))
82 /* set for "ide_preempt" requests and also for requests for which the SCSI
83 "quiesce" state must be ignored. */
84 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8))
85 /* contains copies of user pages */
86 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9))
87 /* vaguely specified driver internal error. Ignored by the block layer */
88 #define RQF_FAILED ((__force req_flags_t)(1 << 10))
89 /* don't warn about errors */
90 #define RQF_QUIET ((__force req_flags_t)(1 << 11))
91 /* elevator private data attached */
92 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12))
93 /* account into disk and partition IO statistics */
94 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13))
95 /* request came from our alloc pool */
96 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14))
97 /* runtime pm request */
98 #define RQF_PM ((__force req_flags_t)(1 << 15))
99 /* on IO scheduler merge hash */
100 #define RQF_HASHED ((__force req_flags_t)(1 << 16))
101 /* track IO completion time */
102 #define RQF_STATS ((__force req_flags_t)(1 << 17))
103 /* Look at ->special_vec for the actual data payload instead of the
105 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18))
106 /* The per-zone write lock is held for this request */
107 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19))
108 /* already slept for hybrid poll */
109 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20))
110 /* ->timeout has been called, don't expire again */
111 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21))
113 /* flags that prevent us from merging requests: */
114 #define RQF_NOMERGE_FLAGS \
115 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD)
118 * Request state for blk-mq.
127 * Try to put the fields that are referenced together in the same cacheline.
129 * If you modify this structure, make sure to update blk_rq_init() and
130 * especially blk_mq_rq_ctx_init() to take care of the added fields.
133 struct request_queue *q;
134 struct blk_mq_ctx *mq_ctx;
135 struct blk_mq_hw_ctx *mq_hctx;
137 unsigned int cmd_flags; /* op and common flags */
138 req_flags_t rq_flags;
143 /* the following two fields are internal, NEVER access directly */
144 unsigned int __data_len; /* total data len */
145 sector_t __sector; /* sector cursor */
150 struct list_head queuelist;
153 * The hash is used inside the scheduler, and killed once the
154 * request reaches the dispatch list. The ipi_list is only used
155 * to queue the request for softirq completion, which is long
156 * after the request has been unhashed (and even removed from
157 * the dispatch list).
160 struct hlist_node hash; /* merge hash */
161 struct list_head ipi_list;
165 * The rb_node is only used inside the io scheduler, requests
166 * are pruned when moved to the dispatch queue. So let the
167 * completion_data share space with the rb_node.
170 struct rb_node rb_node; /* sort/lookup */
171 struct bio_vec special_vec;
172 void *completion_data;
173 int error_count; /* for legacy drivers, don't use */
177 * Three pointers are available for the IO schedulers, if they need
178 * more they have to dynamically allocate it. Flush requests are
179 * never put on the IO scheduler. So let the flush fields share
180 * space with the elevator data.
190 struct list_head list;
191 rq_end_io_fn *saved_end_io;
195 struct gendisk *rq_disk;
196 struct hd_struct *part;
197 /* Time that I/O was submitted to the kernel. */
199 /* Time that I/O was submitted to the device. */
200 u64 io_start_time_ns;
202 #ifdef CONFIG_BLK_WBT
203 unsigned short wbt_flags;
206 * rq sectors used for blk stats. It has the same value
207 * with blk_rq_sectors(rq), except that it never be zeroed
210 unsigned short stats_sectors;
213 * Number of scatter-gather DMA addr+len pairs after
214 * physical address coalescing is performed.
216 unsigned short nr_phys_segments;
218 #if defined(CONFIG_BLK_DEV_INTEGRITY)
219 unsigned short nr_integrity_segments;
222 unsigned short write_hint;
223 unsigned short ioprio;
225 unsigned int extra_len; /* length of alignment and padding */
227 enum mq_rq_state state;
230 unsigned int timeout;
231 unsigned long deadline;
234 struct __call_single_data csd;
239 * completion callback.
241 rq_end_io_fn *end_io;
245 static inline bool blk_op_is_scsi(unsigned int op)
247 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
250 static inline bool blk_op_is_private(unsigned int op)
252 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
255 static inline bool blk_rq_is_scsi(struct request *rq)
257 return blk_op_is_scsi(req_op(rq));
260 static inline bool blk_rq_is_private(struct request *rq)
262 return blk_op_is_private(req_op(rq));
265 static inline bool blk_rq_is_passthrough(struct request *rq)
267 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
270 static inline bool bio_is_passthrough(struct bio *bio)
272 unsigned op = bio_op(bio);
274 return blk_op_is_scsi(op) || blk_op_is_private(op);
277 static inline unsigned short req_get_ioprio(struct request *req)
282 #include <linux/elevator.h>
284 struct blk_queue_ctx;
286 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
289 typedef int (dma_drain_needed_fn)(struct request *);
291 enum blk_eh_timer_return {
292 BLK_EH_DONE, /* drivers has completed the command */
293 BLK_EH_RESET_TIMER, /* reset timer and try again */
296 enum blk_queue_state {
301 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
302 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
304 #define BLK_SCSI_MAX_CMDS (256)
305 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
308 * Zoned block device models (zoned limit).
310 enum blk_zoned_model {
311 BLK_ZONED_NONE, /* Regular block device */
312 BLK_ZONED_HA, /* Host-aware zoned block device */
313 BLK_ZONED_HM, /* Host-managed zoned block device */
316 struct queue_limits {
317 unsigned long bounce_pfn;
318 unsigned long seg_boundary_mask;
319 unsigned long virt_boundary_mask;
321 unsigned int max_hw_sectors;
322 unsigned int max_dev_sectors;
323 unsigned int chunk_sectors;
324 unsigned int max_sectors;
325 unsigned int max_segment_size;
326 unsigned int physical_block_size;
327 unsigned int alignment_offset;
330 unsigned int max_discard_sectors;
331 unsigned int max_hw_discard_sectors;
332 unsigned int max_write_same_sectors;
333 unsigned int max_write_zeroes_sectors;
334 unsigned int discard_granularity;
335 unsigned int discard_alignment;
337 unsigned short logical_block_size;
338 unsigned short max_segments;
339 unsigned short max_integrity_segments;
340 unsigned short max_discard_segments;
342 unsigned char misaligned;
343 unsigned char discard_misaligned;
344 unsigned char raid_partial_stripes_expensive;
345 enum blk_zoned_model zoned;
348 #ifdef CONFIG_BLK_DEV_ZONED
351 * Maximum number of zones to report with a single report zones command.
353 #define BLK_ZONED_REPORT_MAX_ZONES 8192U
355 extern unsigned int blkdev_nr_zones(struct block_device *bdev);
356 extern int blkdev_report_zones(struct block_device *bdev,
357 sector_t sector, struct blk_zone *zones,
358 unsigned int *nr_zones);
359 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
360 sector_t nr_sectors, gfp_t gfp_mask);
361 extern int blk_revalidate_disk_zones(struct gendisk *disk);
363 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
364 unsigned int cmd, unsigned long arg);
365 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
366 unsigned int cmd, unsigned long arg);
368 #else /* CONFIG_BLK_DEV_ZONED */
370 static inline unsigned int blkdev_nr_zones(struct block_device *bdev)
375 static inline int blk_revalidate_disk_zones(struct gendisk *disk)
380 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
381 fmode_t mode, unsigned int cmd,
387 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
388 fmode_t mode, unsigned int cmd,
394 #endif /* CONFIG_BLK_DEV_ZONED */
396 struct request_queue {
398 * Together with queue_head for cacheline sharing
400 struct list_head queue_head;
401 struct request *last_merge;
402 struct elevator_queue *elevator;
404 struct blk_queue_stats *stats;
405 struct rq_qos *rq_qos;
407 make_request_fn *make_request_fn;
408 dma_drain_needed_fn *dma_drain_needed;
410 const struct blk_mq_ops *mq_ops;
413 struct blk_mq_ctx __percpu *queue_ctx;
414 unsigned int nr_queues;
416 unsigned int queue_depth;
418 /* hw dispatch queues */
419 struct blk_mq_hw_ctx **queue_hw_ctx;
420 unsigned int nr_hw_queues;
422 struct backing_dev_info *backing_dev_info;
425 * The queue owner gets to use this for whatever they like.
426 * ll_rw_blk doesn't touch it.
431 * various queue flags, see QUEUE_* below
433 unsigned long queue_flags;
435 * Number of contexts that have called blk_set_pm_only(). If this
436 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
442 * ida allocated id for this queue. Used to index queues from
448 * queue needs bounce pages for pages above this limit
452 spinlock_t queue_lock;
462 struct kobject *mq_kobj;
464 #ifdef CONFIG_BLK_DEV_INTEGRITY
465 struct blk_integrity integrity;
466 #endif /* CONFIG_BLK_DEV_INTEGRITY */
471 unsigned int nr_pending;
477 unsigned long nr_requests; /* Max # of requests */
479 unsigned int dma_drain_size;
480 void *dma_drain_buffer;
481 unsigned int dma_pad_mask;
482 unsigned int dma_alignment;
484 unsigned int rq_timeout;
487 struct blk_stat_callback *poll_cb;
488 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
490 struct timer_list timeout;
491 struct work_struct timeout_work;
493 struct list_head icq_list;
494 #ifdef CONFIG_BLK_CGROUP
495 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
496 struct blkcg_gq *root_blkg;
497 struct list_head blkg_list;
500 struct queue_limits limits;
502 #ifdef CONFIG_BLK_DEV_ZONED
504 * Zoned block device information for request dispatch control.
505 * nr_zones is the total number of zones of the device. This is always
506 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
507 * bits which indicates if a zone is conventional (bit clear) or
508 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
509 * bits which indicates if a zone is write locked, that is, if a write
510 * request targeting the zone was dispatched. All three fields are
511 * initialized by the low level device driver (e.g. scsi/sd.c).
512 * Stacking drivers (device mappers) may or may not initialize
515 * Reads of this information must be protected with blk_queue_enter() /
516 * blk_queue_exit(). Modifying this information is only allowed while
517 * no requests are being processed. See also blk_mq_freeze_queue() and
518 * blk_mq_unfreeze_queue().
520 unsigned int nr_zones;
521 unsigned long *seq_zones_bitmap;
522 unsigned long *seq_zones_wlock;
523 #endif /* CONFIG_BLK_DEV_ZONED */
528 unsigned int sg_timeout;
529 unsigned int sg_reserved_size;
531 #ifdef CONFIG_BLK_DEV_IO_TRACE
532 struct blk_trace *blk_trace;
533 struct mutex blk_trace_mutex;
536 * for flush operations
538 struct blk_flush_queue *fq;
540 struct list_head requeue_list;
541 spinlock_t requeue_lock;
542 struct delayed_work requeue_work;
544 struct mutex sysfs_lock;
547 * for reusing dead hctx instance in case of updating
550 struct list_head unused_hctx_list;
551 spinlock_t unused_hctx_lock;
555 #if defined(CONFIG_BLK_DEV_BSG)
556 struct bsg_class_device bsg_dev;
559 #ifdef CONFIG_BLK_DEV_THROTTLING
561 struct throtl_data *td;
563 struct rcu_head rcu_head;
564 wait_queue_head_t mq_freeze_wq;
566 * Protect concurrent access to q_usage_counter by
567 * percpu_ref_kill() and percpu_ref_reinit().
569 struct mutex mq_freeze_lock;
570 struct percpu_ref q_usage_counter;
572 struct blk_mq_tag_set *tag_set;
573 struct list_head tag_set_list;
574 struct bio_set bio_split;
576 #ifdef CONFIG_BLK_DEBUG_FS
577 struct dentry *debugfs_dir;
578 struct dentry *sched_debugfs_dir;
579 struct dentry *rqos_debugfs_dir;
582 bool mq_sysfs_init_done;
586 struct work_struct release_work;
588 #define BLK_MAX_WRITE_HINTS 5
589 u64 write_hints[BLK_MAX_WRITE_HINTS];
592 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */
593 #define QUEUE_FLAG_DYING 1 /* queue being torn down */
594 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */
595 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */
596 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */
597 #define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */
598 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
599 #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */
600 #define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */
601 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */
602 #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */
603 #define QUEUE_FLAG_SECERASE 11 /* supports secure erase */
604 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */
605 #define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */
606 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */
607 #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */
608 #define QUEUE_FLAG_WC 17 /* Write back caching */
609 #define QUEUE_FLAG_FUA 18 /* device supports FUA writes */
610 #define QUEUE_FLAG_DAX 19 /* device supports DAX */
611 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */
612 #define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */
613 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */
614 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */
615 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */
616 #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */
618 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
619 (1 << QUEUE_FLAG_SAME_COMP))
621 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
622 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
623 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
625 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
626 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
627 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
628 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
629 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
630 #define blk_queue_noxmerges(q) \
631 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
632 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
633 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
634 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
635 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
636 #define blk_queue_secure_erase(q) \
637 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
638 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
639 #define blk_queue_scsi_passthrough(q) \
640 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
641 #define blk_queue_pci_p2pdma(q) \
642 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
644 #define blk_noretry_request(rq) \
645 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
646 REQ_FAILFAST_DRIVER))
647 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
648 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
649 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
651 extern void blk_set_pm_only(struct request_queue *q);
652 extern void blk_clear_pm_only(struct request_queue *q);
654 static inline bool blk_account_rq(struct request *rq)
656 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
659 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
661 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
663 #define rq_dma_dir(rq) \
664 (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
666 #define dma_map_bvec(dev, bv, dir, attrs) \
667 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
670 static inline bool queue_is_mq(struct request_queue *q)
675 static inline enum blk_zoned_model
676 blk_queue_zoned_model(struct request_queue *q)
678 return q->limits.zoned;
681 static inline bool blk_queue_is_zoned(struct request_queue *q)
683 switch (blk_queue_zoned_model(q)) {
692 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
694 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
697 #ifdef CONFIG_BLK_DEV_ZONED
698 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
700 return blk_queue_is_zoned(q) ? q->nr_zones : 0;
703 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
706 if (!blk_queue_is_zoned(q))
708 return sector >> ilog2(q->limits.chunk_sectors);
711 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
714 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
716 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
718 #else /* CONFIG_BLK_DEV_ZONED */
719 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
723 #endif /* CONFIG_BLK_DEV_ZONED */
725 static inline bool rq_is_sync(struct request *rq)
727 return op_is_sync(rq->cmd_flags);
730 static inline bool rq_mergeable(struct request *rq)
732 if (blk_rq_is_passthrough(rq))
735 if (req_op(rq) == REQ_OP_FLUSH)
738 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
741 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
743 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
749 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
751 if (bio_page(a) == bio_page(b) &&
752 bio_offset(a) == bio_offset(b))
758 static inline unsigned int blk_queue_depth(struct request_queue *q)
761 return q->queue_depth;
763 return q->nr_requests;
766 extern unsigned long blk_max_low_pfn, blk_max_pfn;
769 * standard bounce addresses:
771 * BLK_BOUNCE_HIGH : bounce all highmem pages
772 * BLK_BOUNCE_ANY : don't bounce anything
773 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
776 #if BITS_PER_LONG == 32
777 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
779 #define BLK_BOUNCE_HIGH -1ULL
781 #define BLK_BOUNCE_ANY (-1ULL)
782 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
785 * default timeout for SG_IO if none specified
787 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
788 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
794 unsigned long offset;
799 struct req_iterator {
800 struct bvec_iter iter;
804 /* This should not be used directly - use rq_for_each_segment */
805 #define for_each_bio(_bio) \
806 for (; _bio; _bio = _bio->bi_next)
807 #define __rq_for_each_bio(_bio, rq) \
809 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
811 #define rq_for_each_segment(bvl, _rq, _iter) \
812 __rq_for_each_bio(_iter.bio, _rq) \
813 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
815 #define rq_for_each_bvec(bvl, _rq, _iter) \
816 __rq_for_each_bio(_iter.bio, _rq) \
817 bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
819 #define rq_iter_last(bvec, _iter) \
820 (_iter.bio->bi_next == NULL && \
821 bio_iter_last(bvec, _iter.iter))
823 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
824 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
826 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
827 extern void rq_flush_dcache_pages(struct request *rq);
829 static inline void rq_flush_dcache_pages(struct request *rq)
834 extern int blk_register_queue(struct gendisk *disk);
835 extern void blk_unregister_queue(struct gendisk *disk);
836 extern blk_qc_t generic_make_request(struct bio *bio);
837 extern blk_qc_t direct_make_request(struct bio *bio);
838 extern void blk_rq_init(struct request_queue *q, struct request *rq);
839 extern void blk_put_request(struct request *);
840 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
841 blk_mq_req_flags_t flags);
842 extern int blk_lld_busy(struct request_queue *q);
843 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
844 struct bio_set *bs, gfp_t gfp_mask,
845 int (*bio_ctr)(struct bio *, struct bio *, void *),
847 extern void blk_rq_unprep_clone(struct request *rq);
848 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
850 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
851 extern void blk_queue_split(struct request_queue *, struct bio **);
852 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
853 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
854 unsigned int, void __user *);
855 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
856 unsigned int, void __user *);
857 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
858 struct scsi_ioctl_command __user *);
860 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
861 extern void blk_queue_exit(struct request_queue *q);
862 extern void blk_sync_queue(struct request_queue *q);
863 extern int blk_rq_map_user(struct request_queue *, struct request *,
864 struct rq_map_data *, void __user *, unsigned long,
866 extern int blk_rq_unmap_user(struct bio *);
867 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
868 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
869 struct rq_map_data *, const struct iov_iter *,
871 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
872 struct request *, int);
873 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
874 struct request *, int, rq_end_io_fn *);
876 /* Helper to convert REQ_OP_XXX to its string format XXX */
877 extern const char *blk_op_str(unsigned int op);
879 int blk_status_to_errno(blk_status_t status);
880 blk_status_t errno_to_blk_status(int errno);
882 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
884 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
886 return bdev->bd_disk->queue; /* this is never NULL */
890 * The basic unit of block I/O is a sector. It is used in a number of contexts
891 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
892 * bytes. Variables of type sector_t represent an offset or size that is a
893 * multiple of 512 bytes. Hence these two constants.
896 #define SECTOR_SHIFT 9
899 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
903 * blk_rq_pos() : the current sector
904 * blk_rq_bytes() : bytes left in the entire request
905 * blk_rq_cur_bytes() : bytes left in the current segment
906 * blk_rq_err_bytes() : bytes left till the next error boundary
907 * blk_rq_sectors() : sectors left in the entire request
908 * blk_rq_cur_sectors() : sectors left in the current segment
909 * blk_rq_stats_sectors() : sectors of the entire request used for stats
911 static inline sector_t blk_rq_pos(const struct request *rq)
916 static inline unsigned int blk_rq_bytes(const struct request *rq)
918 return rq->__data_len;
921 static inline int blk_rq_cur_bytes(const struct request *rq)
923 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
926 extern unsigned int blk_rq_err_bytes(const struct request *rq);
928 static inline unsigned int blk_rq_sectors(const struct request *rq)
930 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
933 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
935 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
938 static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
940 return rq->stats_sectors;
943 #ifdef CONFIG_BLK_DEV_ZONED
944 static inline unsigned int blk_rq_zone_no(struct request *rq)
946 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
949 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
951 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
953 #endif /* CONFIG_BLK_DEV_ZONED */
956 * Some commands like WRITE SAME have a payload or data transfer size which
957 * is different from the size of the request. Any driver that supports such
958 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
959 * calculate the data transfer size.
961 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
963 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
964 return rq->special_vec.bv_len;
965 return blk_rq_bytes(rq);
969 * Return the first full biovec in the request. The caller needs to check that
970 * there are any bvecs before calling this helper.
972 static inline struct bio_vec req_bvec(struct request *rq)
974 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
975 return rq->special_vec;
976 return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
979 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
982 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
983 return min(q->limits.max_discard_sectors,
984 UINT_MAX >> SECTOR_SHIFT);
986 if (unlikely(op == REQ_OP_WRITE_SAME))
987 return q->limits.max_write_same_sectors;
989 if (unlikely(op == REQ_OP_WRITE_ZEROES))
990 return q->limits.max_write_zeroes_sectors;
992 return q->limits.max_sectors;
996 * Return maximum size of a request at given offset. Only valid for
997 * file system requests.
999 static inline unsigned int blk_max_size_offset(struct request_queue *q,
1002 if (!q->limits.chunk_sectors)
1003 return q->limits.max_sectors;
1005 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
1006 (offset & (q->limits.chunk_sectors - 1))));
1009 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1012 struct request_queue *q = rq->q;
1014 if (blk_rq_is_passthrough(rq))
1015 return q->limits.max_hw_sectors;
1017 if (!q->limits.chunk_sectors ||
1018 req_op(rq) == REQ_OP_DISCARD ||
1019 req_op(rq) == REQ_OP_SECURE_ERASE)
1020 return blk_queue_get_max_sectors(q, req_op(rq));
1022 return min(blk_max_size_offset(q, offset),
1023 blk_queue_get_max_sectors(q, req_op(rq)));
1026 static inline unsigned int blk_rq_count_bios(struct request *rq)
1028 unsigned int nr_bios = 0;
1031 __rq_for_each_bio(bio, rq)
1037 void blk_steal_bios(struct bio_list *list, struct request *rq);
1040 * Request completion related functions.
1042 * blk_update_request() completes given number of bytes and updates
1043 * the request without completing it.
1045 extern bool blk_update_request(struct request *rq, blk_status_t error,
1046 unsigned int nr_bytes);
1048 extern void __blk_complete_request(struct request *);
1049 extern void blk_abort_request(struct request *);
1052 * Access functions for manipulating queue properties
1054 extern void blk_cleanup_queue(struct request_queue *);
1055 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1056 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1057 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1058 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1059 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1060 extern void blk_queue_max_discard_segments(struct request_queue *,
1062 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1063 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1064 unsigned int max_discard_sectors);
1065 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1066 unsigned int max_write_same_sectors);
1067 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1068 unsigned int max_write_same_sectors);
1069 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1070 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1071 extern void blk_queue_alignment_offset(struct request_queue *q,
1072 unsigned int alignment);
1073 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1074 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1075 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1076 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1077 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1078 extern void blk_set_default_limits(struct queue_limits *lim);
1079 extern void blk_set_stacking_limits(struct queue_limits *lim);
1080 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1082 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1084 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1086 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1087 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1088 extern int blk_queue_dma_drain(struct request_queue *q,
1089 dma_drain_needed_fn *dma_drain_needed,
1090 void *buf, unsigned int size);
1091 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1092 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1093 extern void blk_queue_dma_alignment(struct request_queue *, int);
1094 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1095 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1096 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1099 * Number of physical segments as sent to the device.
1101 * Normally this is the number of discontiguous data segments sent by the
1102 * submitter. But for data-less command like discard we might have no
1103 * actual data segments submitted, but the driver might have to add it's
1104 * own special payload. In that case we still return 1 here so that this
1105 * special payload will be mapped.
1107 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1109 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1111 return rq->nr_phys_segments;
1115 * Number of discard segments (or ranges) the driver needs to fill in.
1116 * Each discard bio merged into a request is counted as one segment.
1118 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1120 return max_t(unsigned short, rq->nr_phys_segments, 1);
1123 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1124 extern void blk_dump_rq_flags(struct request *, char *);
1125 extern long nr_blockdev_pages(void);
1127 bool __must_check blk_get_queue(struct request_queue *);
1128 struct request_queue *blk_alloc_queue(gfp_t);
1129 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id);
1130 extern void blk_put_queue(struct request_queue *);
1131 extern void blk_set_queue_dying(struct request_queue *);
1134 * blk_plug permits building a queue of related requests by holding the I/O
1135 * fragments for a short period. This allows merging of sequential requests
1136 * into single larger request. As the requests are moved from a per-task list to
1137 * the device's request_queue in a batch, this results in improved scalability
1138 * as the lock contention for request_queue lock is reduced.
1140 * It is ok not to disable preemption when adding the request to the plug list
1141 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1142 * the plug list when the task sleeps by itself. For details, please see
1143 * schedule() where blk_schedule_flush_plug() is called.
1146 struct list_head mq_list; /* blk-mq requests */
1147 struct list_head cb_list; /* md requires an unplug callback */
1148 unsigned short rq_count;
1149 bool multiple_queues;
1151 #define BLK_MAX_REQUEST_COUNT 16
1152 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1155 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1156 struct blk_plug_cb {
1157 struct list_head list;
1158 blk_plug_cb_fn callback;
1161 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1162 void *data, int size);
1163 extern void blk_start_plug(struct blk_plug *);
1164 extern void blk_finish_plug(struct blk_plug *);
1165 extern void blk_flush_plug_list(struct blk_plug *, bool);
1167 static inline void blk_flush_plug(struct task_struct *tsk)
1169 struct blk_plug *plug = tsk->plug;
1172 blk_flush_plug_list(plug, false);
1175 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1177 struct blk_plug *plug = tsk->plug;
1180 blk_flush_plug_list(plug, true);
1183 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1185 struct blk_plug *plug = tsk->plug;
1188 (!list_empty(&plug->mq_list) ||
1189 !list_empty(&plug->cb_list));
1192 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1193 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1194 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1196 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1198 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1199 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1200 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1201 sector_t nr_sects, gfp_t gfp_mask, int flags,
1204 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1205 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1207 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1208 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1210 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1211 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1213 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1214 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1216 return blkdev_issue_discard(sb->s_bdev,
1217 block << (sb->s_blocksize_bits -
1219 nr_blocks << (sb->s_blocksize_bits -
1223 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1224 sector_t nr_blocks, gfp_t gfp_mask)
1226 return blkdev_issue_zeroout(sb->s_bdev,
1227 block << (sb->s_blocksize_bits -
1229 nr_blocks << (sb->s_blocksize_bits -
1234 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1236 enum blk_default_limits {
1237 BLK_MAX_SEGMENTS = 128,
1238 BLK_SAFE_MAX_SECTORS = 255,
1239 BLK_DEF_MAX_SECTORS = 2560,
1240 BLK_MAX_SEGMENT_SIZE = 65536,
1241 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1244 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1246 return q->limits.seg_boundary_mask;
1249 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1251 return q->limits.virt_boundary_mask;
1254 static inline unsigned int queue_max_sectors(struct request_queue *q)
1256 return q->limits.max_sectors;
1259 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1261 return q->limits.max_hw_sectors;
1264 static inline unsigned short queue_max_segments(struct request_queue *q)
1266 return q->limits.max_segments;
1269 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1271 return q->limits.max_discard_segments;
1274 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1276 return q->limits.max_segment_size;
1279 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1283 if (q && q->limits.logical_block_size)
1284 retval = q->limits.logical_block_size;
1289 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1291 return queue_logical_block_size(bdev_get_queue(bdev));
1294 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1296 return q->limits.physical_block_size;
1299 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1301 return queue_physical_block_size(bdev_get_queue(bdev));
1304 static inline unsigned int queue_io_min(struct request_queue *q)
1306 return q->limits.io_min;
1309 static inline int bdev_io_min(struct block_device *bdev)
1311 return queue_io_min(bdev_get_queue(bdev));
1314 static inline unsigned int queue_io_opt(struct request_queue *q)
1316 return q->limits.io_opt;
1319 static inline int bdev_io_opt(struct block_device *bdev)
1321 return queue_io_opt(bdev_get_queue(bdev));
1324 static inline int queue_alignment_offset(struct request_queue *q)
1326 if (q->limits.misaligned)
1329 return q->limits.alignment_offset;
1332 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1334 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1335 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1338 return (granularity + lim->alignment_offset - alignment) % granularity;
1341 static inline int bdev_alignment_offset(struct block_device *bdev)
1343 struct request_queue *q = bdev_get_queue(bdev);
1345 if (q->limits.misaligned)
1348 if (bdev != bdev->bd_contains)
1349 return bdev->bd_part->alignment_offset;
1351 return q->limits.alignment_offset;
1354 static inline int queue_discard_alignment(struct request_queue *q)
1356 if (q->limits.discard_misaligned)
1359 return q->limits.discard_alignment;
1362 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1364 unsigned int alignment, granularity, offset;
1366 if (!lim->max_discard_sectors)
1369 /* Why are these in bytes, not sectors? */
1370 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1371 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1375 /* Offset of the partition start in 'granularity' sectors */
1376 offset = sector_div(sector, granularity);
1378 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1379 offset = (granularity + alignment - offset) % granularity;
1381 /* Turn it back into bytes, gaah */
1382 return offset << SECTOR_SHIFT;
1385 static inline int bdev_discard_alignment(struct block_device *bdev)
1387 struct request_queue *q = bdev_get_queue(bdev);
1389 if (bdev != bdev->bd_contains)
1390 return bdev->bd_part->discard_alignment;
1392 return q->limits.discard_alignment;
1395 static inline unsigned int bdev_write_same(struct block_device *bdev)
1397 struct request_queue *q = bdev_get_queue(bdev);
1400 return q->limits.max_write_same_sectors;
1405 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1407 struct request_queue *q = bdev_get_queue(bdev);
1410 return q->limits.max_write_zeroes_sectors;
1415 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1417 struct request_queue *q = bdev_get_queue(bdev);
1420 return blk_queue_zoned_model(q);
1422 return BLK_ZONED_NONE;
1425 static inline bool bdev_is_zoned(struct block_device *bdev)
1427 struct request_queue *q = bdev_get_queue(bdev);
1430 return blk_queue_is_zoned(q);
1435 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1437 struct request_queue *q = bdev_get_queue(bdev);
1440 return blk_queue_zone_sectors(q);
1444 static inline int queue_dma_alignment(struct request_queue *q)
1446 return q ? q->dma_alignment : 511;
1449 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1452 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1453 return !(addr & alignment) && !(len & alignment);
1456 /* assumes size > 256 */
1457 static inline unsigned int blksize_bits(unsigned int size)
1459 unsigned int bits = 8;
1463 } while (size > 256);
1467 static inline unsigned int block_size(struct block_device *bdev)
1469 return bdev->bd_block_size;
1472 typedef struct {struct page *v;} Sector;
1474 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1476 static inline void put_dev_sector(Sector p)
1481 int kblockd_schedule_work(struct work_struct *work);
1482 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1483 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1485 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1486 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1487 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1488 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1490 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1492 enum blk_integrity_flags {
1493 BLK_INTEGRITY_VERIFY = 1 << 0,
1494 BLK_INTEGRITY_GENERATE = 1 << 1,
1495 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1496 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1499 struct blk_integrity_iter {
1503 unsigned int data_size;
1504 unsigned short interval;
1505 const char *disk_name;
1508 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1510 struct blk_integrity_profile {
1511 integrity_processing_fn *generate_fn;
1512 integrity_processing_fn *verify_fn;
1516 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1517 extern void blk_integrity_unregister(struct gendisk *);
1518 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1519 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1520 struct scatterlist *);
1521 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1522 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1524 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1527 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1529 struct blk_integrity *bi = &disk->queue->integrity;
1538 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1540 return blk_get_integrity(bdev->bd_disk);
1543 static inline bool blk_integrity_rq(struct request *rq)
1545 return rq->cmd_flags & REQ_INTEGRITY;
1548 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1551 q->limits.max_integrity_segments = segs;
1554 static inline unsigned short
1555 queue_max_integrity_segments(struct request_queue *q)
1557 return q->limits.max_integrity_segments;
1561 * bio_integrity_intervals - Return number of integrity intervals for a bio
1562 * @bi: blk_integrity profile for device
1563 * @sectors: Size of the bio in 512-byte sectors
1565 * Description: The block layer calculates everything in 512 byte
1566 * sectors but integrity metadata is done in terms of the data integrity
1567 * interval size of the storage device. Convert the block layer sectors
1568 * to the appropriate number of integrity intervals.
1570 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1571 unsigned int sectors)
1573 return sectors >> (bi->interval_exp - 9);
1576 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1577 unsigned int sectors)
1579 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1583 * Return the first bvec that contains integrity data. Only drivers that are
1584 * limited to a single integrity segment should use this helper.
1586 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1588 if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1590 return rq->bio->bi_integrity->bip_vec;
1593 #else /* CONFIG_BLK_DEV_INTEGRITY */
1596 struct block_device;
1598 struct blk_integrity;
1600 static inline int blk_integrity_rq(struct request *rq)
1604 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1609 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1611 struct scatterlist *s)
1615 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1619 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1623 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1627 static inline void blk_integrity_register(struct gendisk *d,
1628 struct blk_integrity *b)
1631 static inline void blk_integrity_unregister(struct gendisk *d)
1634 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1638 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1642 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1648 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1655 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1656 unsigned int sectors)
1661 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1662 unsigned int sectors)
1667 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1672 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1674 struct block_device_operations {
1675 int (*open) (struct block_device *, fmode_t);
1676 void (*release) (struct gendisk *, fmode_t);
1677 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1678 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1679 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1680 unsigned int (*check_events) (struct gendisk *disk,
1681 unsigned int clearing);
1682 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1683 int (*media_changed) (struct gendisk *);
1684 void (*unlock_native_capacity) (struct gendisk *);
1685 int (*revalidate_disk) (struct gendisk *);
1686 int (*getgeo)(struct block_device *, struct hd_geometry *);
1687 /* this callback is with swap_lock and sometimes page table lock held */
1688 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1689 int (*report_zones)(struct gendisk *, sector_t sector,
1690 struct blk_zone *zones, unsigned int *nr_zones);
1691 struct module *owner;
1692 const struct pr_ops *pr_ops;
1695 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1697 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1698 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1699 struct writeback_control *);
1701 #ifdef CONFIG_BLK_DEV_ZONED
1702 bool blk_req_needs_zone_write_lock(struct request *rq);
1703 void __blk_req_zone_write_lock(struct request *rq);
1704 void __blk_req_zone_write_unlock(struct request *rq);
1706 static inline void blk_req_zone_write_lock(struct request *rq)
1708 if (blk_req_needs_zone_write_lock(rq))
1709 __blk_req_zone_write_lock(rq);
1712 static inline void blk_req_zone_write_unlock(struct request *rq)
1714 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1715 __blk_req_zone_write_unlock(rq);
1718 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1720 return rq->q->seq_zones_wlock &&
1721 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1724 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1726 if (!blk_req_needs_zone_write_lock(rq))
1728 return !blk_req_zone_is_write_locked(rq);
1731 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1736 static inline void blk_req_zone_write_lock(struct request *rq)
1740 static inline void blk_req_zone_write_unlock(struct request *rq)
1743 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1748 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1752 #endif /* CONFIG_BLK_DEV_ZONED */
1754 #else /* CONFIG_BLOCK */
1756 struct block_device;
1759 * stubs for when the block layer is configured out
1761 #define buffer_heads_over_limit 0
1763 static inline long nr_blockdev_pages(void)
1771 static inline void blk_start_plug(struct blk_plug *plug)
1775 static inline void blk_finish_plug(struct blk_plug *plug)
1779 static inline void blk_flush_plug(struct task_struct *task)
1783 static inline void blk_schedule_flush_plug(struct task_struct *task)
1788 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1793 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1794 sector_t *error_sector)
1799 #endif /* CONFIG_BLOCK */
1801 static inline void blk_wake_io_task(struct task_struct *waiter)
1804 * If we're polling, the task itself is doing the completions. For
1805 * that case, we don't need to signal a wakeup, it's enough to just
1806 * mark us as RUNNING.
1808 if (waiter == current)
1809 __set_current_state(TASK_RUNNING);
1811 wake_up_process(waiter);