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;
205 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
206 unsigned short throtl_size;
210 * Number of scatter-gather DMA addr+len pairs after
211 * physical address coalescing is performed.
213 unsigned short nr_phys_segments;
215 #if defined(CONFIG_BLK_DEV_INTEGRITY)
216 unsigned short nr_integrity_segments;
219 unsigned short write_hint;
220 unsigned short ioprio;
222 unsigned int extra_len; /* length of alignment and padding */
224 enum mq_rq_state state;
227 unsigned int timeout;
228 unsigned long deadline;
231 struct __call_single_data csd;
236 * completion callback.
238 rq_end_io_fn *end_io;
242 static inline bool blk_op_is_scsi(unsigned int op)
244 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT;
247 static inline bool blk_op_is_private(unsigned int op)
249 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
252 static inline bool blk_rq_is_scsi(struct request *rq)
254 return blk_op_is_scsi(req_op(rq));
257 static inline bool blk_rq_is_private(struct request *rq)
259 return blk_op_is_private(req_op(rq));
262 static inline bool blk_rq_is_passthrough(struct request *rq)
264 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq);
267 static inline bool bio_is_passthrough(struct bio *bio)
269 unsigned op = bio_op(bio);
271 return blk_op_is_scsi(op) || blk_op_is_private(op);
274 static inline unsigned short req_get_ioprio(struct request *req)
279 #include <linux/elevator.h>
281 struct blk_queue_ctx;
283 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio);
286 typedef int (dma_drain_needed_fn)(struct request *);
288 enum blk_eh_timer_return {
289 BLK_EH_DONE, /* drivers has completed the command */
290 BLK_EH_RESET_TIMER, /* reset timer and try again */
293 enum blk_queue_state {
298 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */
299 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */
301 #define BLK_SCSI_MAX_CMDS (256)
302 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
305 * Zoned block device models (zoned limit).
307 enum blk_zoned_model {
308 BLK_ZONED_NONE, /* Regular block device */
309 BLK_ZONED_HA, /* Host-aware zoned block device */
310 BLK_ZONED_HM, /* Host-managed zoned block device */
313 struct queue_limits {
314 unsigned long bounce_pfn;
315 unsigned long seg_boundary_mask;
316 unsigned long virt_boundary_mask;
318 unsigned int max_hw_sectors;
319 unsigned int max_dev_sectors;
320 unsigned int chunk_sectors;
321 unsigned int max_sectors;
322 unsigned int max_segment_size;
323 unsigned int physical_block_size;
324 unsigned int alignment_offset;
327 unsigned int max_discard_sectors;
328 unsigned int max_hw_discard_sectors;
329 unsigned int max_write_same_sectors;
330 unsigned int max_write_zeroes_sectors;
331 unsigned int discard_granularity;
332 unsigned int discard_alignment;
334 unsigned short logical_block_size;
335 unsigned short max_segments;
336 unsigned short max_integrity_segments;
337 unsigned short max_discard_segments;
339 unsigned char misaligned;
340 unsigned char discard_misaligned;
341 unsigned char raid_partial_stripes_expensive;
342 enum blk_zoned_model zoned;
345 #ifdef CONFIG_BLK_DEV_ZONED
348 * Maximum number of zones to report with a single report zones command.
350 #define BLK_ZONED_REPORT_MAX_ZONES 8192U
352 extern unsigned int blkdev_nr_zones(struct block_device *bdev);
353 extern int blkdev_report_zones(struct block_device *bdev,
354 sector_t sector, struct blk_zone *zones,
355 unsigned int *nr_zones);
356 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors,
357 sector_t nr_sectors, gfp_t gfp_mask);
358 extern int blk_revalidate_disk_zones(struct gendisk *disk);
360 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
361 unsigned int cmd, unsigned long arg);
362 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode,
363 unsigned int cmd, unsigned long arg);
365 #else /* CONFIG_BLK_DEV_ZONED */
367 static inline unsigned int blkdev_nr_zones(struct block_device *bdev)
372 static inline int blk_revalidate_disk_zones(struct gendisk *disk)
377 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
378 fmode_t mode, unsigned int cmd,
384 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev,
385 fmode_t mode, unsigned int cmd,
391 #endif /* CONFIG_BLK_DEV_ZONED */
393 struct request_queue {
395 * Together with queue_head for cacheline sharing
397 struct list_head queue_head;
398 struct request *last_merge;
399 struct elevator_queue *elevator;
401 struct blk_queue_stats *stats;
402 struct rq_qos *rq_qos;
404 make_request_fn *make_request_fn;
405 dma_drain_needed_fn *dma_drain_needed;
407 const struct blk_mq_ops *mq_ops;
410 struct blk_mq_ctx __percpu *queue_ctx;
411 unsigned int nr_queues;
413 unsigned int queue_depth;
415 /* hw dispatch queues */
416 struct blk_mq_hw_ctx **queue_hw_ctx;
417 unsigned int nr_hw_queues;
419 struct backing_dev_info *backing_dev_info;
422 * The queue owner gets to use this for whatever they like.
423 * ll_rw_blk doesn't touch it.
428 * various queue flags, see QUEUE_* below
430 unsigned long queue_flags;
432 * Number of contexts that have called blk_set_pm_only(). If this
433 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are
439 * ida allocated id for this queue. Used to index queues from
445 * queue needs bounce pages for pages above this limit
449 spinlock_t queue_lock;
459 struct kobject *mq_kobj;
461 #ifdef CONFIG_BLK_DEV_INTEGRITY
462 struct blk_integrity integrity;
463 #endif /* CONFIG_BLK_DEV_INTEGRITY */
468 unsigned int nr_pending;
474 unsigned long nr_requests; /* Max # of requests */
476 unsigned int dma_drain_size;
477 void *dma_drain_buffer;
478 unsigned int dma_pad_mask;
479 unsigned int dma_alignment;
481 unsigned int rq_timeout;
484 struct blk_stat_callback *poll_cb;
485 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
487 struct timer_list timeout;
488 struct work_struct timeout_work;
490 struct list_head icq_list;
491 #ifdef CONFIG_BLK_CGROUP
492 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
493 struct blkcg_gq *root_blkg;
494 struct list_head blkg_list;
497 struct queue_limits limits;
499 #ifdef CONFIG_BLK_DEV_ZONED
501 * Zoned block device information for request dispatch control.
502 * nr_zones is the total number of zones of the device. This is always
503 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones
504 * bits which indicates if a zone is conventional (bit clear) or
505 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones
506 * bits which indicates if a zone is write locked, that is, if a write
507 * request targeting the zone was dispatched. All three fields are
508 * initialized by the low level device driver (e.g. scsi/sd.c).
509 * Stacking drivers (device mappers) may or may not initialize
512 * Reads of this information must be protected with blk_queue_enter() /
513 * blk_queue_exit(). Modifying this information is only allowed while
514 * no requests are being processed. See also blk_mq_freeze_queue() and
515 * blk_mq_unfreeze_queue().
517 unsigned int nr_zones;
518 unsigned long *seq_zones_bitmap;
519 unsigned long *seq_zones_wlock;
520 #endif /* CONFIG_BLK_DEV_ZONED */
525 unsigned int sg_timeout;
526 unsigned int sg_reserved_size;
528 #ifdef CONFIG_BLK_DEV_IO_TRACE
529 struct blk_trace *blk_trace;
530 struct mutex blk_trace_mutex;
533 * for flush operations
535 struct blk_flush_queue *fq;
537 struct list_head requeue_list;
538 spinlock_t requeue_lock;
539 struct delayed_work requeue_work;
541 struct mutex sysfs_lock;
544 * for reusing dead hctx instance in case of updating
547 struct list_head unused_hctx_list;
548 spinlock_t unused_hctx_lock;
552 #if defined(CONFIG_BLK_DEV_BSG)
553 struct bsg_class_device bsg_dev;
556 #ifdef CONFIG_BLK_DEV_THROTTLING
558 struct throtl_data *td;
560 struct rcu_head rcu_head;
561 wait_queue_head_t mq_freeze_wq;
563 * Protect concurrent access to q_usage_counter by
564 * percpu_ref_kill() and percpu_ref_reinit().
566 struct mutex mq_freeze_lock;
567 struct percpu_ref q_usage_counter;
569 struct blk_mq_tag_set *tag_set;
570 struct list_head tag_set_list;
571 struct bio_set bio_split;
573 #ifdef CONFIG_BLK_DEBUG_FS
574 struct dentry *debugfs_dir;
575 struct dentry *sched_debugfs_dir;
576 struct dentry *rqos_debugfs_dir;
579 bool mq_sysfs_init_done;
583 struct work_struct release_work;
585 #define BLK_MAX_WRITE_HINTS 5
586 u64 write_hints[BLK_MAX_WRITE_HINTS];
589 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */
590 #define QUEUE_FLAG_DYING 1 /* queue being torn down */
591 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */
592 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */
593 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */
594 #define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */
595 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
596 #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */
597 #define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */
598 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */
599 #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */
600 #define QUEUE_FLAG_SECERASE 11 /* supports secure erase */
601 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */
602 #define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */
603 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */
604 #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */
605 #define QUEUE_FLAG_WC 17 /* Write back caching */
606 #define QUEUE_FLAG_FUA 18 /* device supports FUA writes */
607 #define QUEUE_FLAG_DAX 19 /* device supports DAX */
608 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */
609 #define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */
610 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */
611 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */
612 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */
613 #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */
615 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
616 (1 << QUEUE_FLAG_SAME_COMP))
618 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
619 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
620 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
622 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
623 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
624 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
625 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
626 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
627 #define blk_queue_noxmerges(q) \
628 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
629 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
630 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
631 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
632 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
633 #define blk_queue_secure_erase(q) \
634 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
635 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
636 #define blk_queue_scsi_passthrough(q) \
637 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
638 #define blk_queue_pci_p2pdma(q) \
639 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
641 #define blk_noretry_request(rq) \
642 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
643 REQ_FAILFAST_DRIVER))
644 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
645 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
646 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
648 extern void blk_set_pm_only(struct request_queue *q);
649 extern void blk_clear_pm_only(struct request_queue *q);
651 static inline bool blk_account_rq(struct request *rq)
653 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq);
656 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
658 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ)
660 #define rq_dma_dir(rq) \
661 (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE)
663 #define dma_map_bvec(dev, bv, dir, attrs) \
664 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
667 static inline bool queue_is_mq(struct request_queue *q)
672 static inline enum blk_zoned_model
673 blk_queue_zoned_model(struct request_queue *q)
675 return q->limits.zoned;
678 static inline bool blk_queue_is_zoned(struct request_queue *q)
680 switch (blk_queue_zoned_model(q)) {
689 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
691 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
694 #ifdef CONFIG_BLK_DEV_ZONED
695 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
697 return blk_queue_is_zoned(q) ? q->nr_zones : 0;
700 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
703 if (!blk_queue_is_zoned(q))
705 return sector >> ilog2(q->limits.chunk_sectors);
708 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
711 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap)
713 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap);
715 #else /* CONFIG_BLK_DEV_ZONED */
716 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
720 #endif /* CONFIG_BLK_DEV_ZONED */
722 static inline bool rq_is_sync(struct request *rq)
724 return op_is_sync(rq->cmd_flags);
727 static inline bool rq_mergeable(struct request *rq)
729 if (blk_rq_is_passthrough(rq))
732 if (req_op(rq) == REQ_OP_FLUSH)
735 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
738 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
740 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
746 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b)
748 if (bio_page(a) == bio_page(b) &&
749 bio_offset(a) == bio_offset(b))
755 static inline unsigned int blk_queue_depth(struct request_queue *q)
758 return q->queue_depth;
760 return q->nr_requests;
763 extern unsigned long blk_max_low_pfn, blk_max_pfn;
766 * standard bounce addresses:
768 * BLK_BOUNCE_HIGH : bounce all highmem pages
769 * BLK_BOUNCE_ANY : don't bounce anything
770 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
773 #if BITS_PER_LONG == 32
774 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
776 #define BLK_BOUNCE_HIGH -1ULL
778 #define BLK_BOUNCE_ANY (-1ULL)
779 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24))
782 * default timeout for SG_IO if none specified
784 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
785 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
791 unsigned long offset;
796 struct req_iterator {
797 struct bvec_iter iter;
801 /* This should not be used directly - use rq_for_each_segment */
802 #define for_each_bio(_bio) \
803 for (; _bio; _bio = _bio->bi_next)
804 #define __rq_for_each_bio(_bio, rq) \
806 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
808 #define rq_for_each_segment(bvl, _rq, _iter) \
809 __rq_for_each_bio(_iter.bio, _rq) \
810 bio_for_each_segment(bvl, _iter.bio, _iter.iter)
812 #define rq_for_each_bvec(bvl, _rq, _iter) \
813 __rq_for_each_bio(_iter.bio, _rq) \
814 bio_for_each_bvec(bvl, _iter.bio, _iter.iter)
816 #define rq_iter_last(bvec, _iter) \
817 (_iter.bio->bi_next == NULL && \
818 bio_iter_last(bvec, _iter.iter))
820 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
821 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
823 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
824 extern void rq_flush_dcache_pages(struct request *rq);
826 static inline void rq_flush_dcache_pages(struct request *rq)
831 extern int blk_register_queue(struct gendisk *disk);
832 extern void blk_unregister_queue(struct gendisk *disk);
833 extern blk_qc_t generic_make_request(struct bio *bio);
834 extern blk_qc_t direct_make_request(struct bio *bio);
835 extern void blk_rq_init(struct request_queue *q, struct request *rq);
836 extern void blk_put_request(struct request *);
837 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
838 blk_mq_req_flags_t flags);
839 extern int blk_lld_busy(struct request_queue *q);
840 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
841 struct bio_set *bs, gfp_t gfp_mask,
842 int (*bio_ctr)(struct bio *, struct bio *, void *),
844 extern void blk_rq_unprep_clone(struct request *rq);
845 extern blk_status_t blk_insert_cloned_request(struct request_queue *q,
847 extern int blk_rq_append_bio(struct request *rq, struct bio **bio);
848 extern void blk_queue_split(struct request_queue *, struct bio **);
849 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int);
850 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t,
851 unsigned int, void __user *);
852 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
853 unsigned int, void __user *);
854 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
855 struct scsi_ioctl_command __user *);
857 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
858 extern void blk_queue_exit(struct request_queue *q);
859 extern void blk_sync_queue(struct request_queue *q);
860 extern int blk_rq_map_user(struct request_queue *, struct request *,
861 struct rq_map_data *, void __user *, unsigned long,
863 extern int blk_rq_unmap_user(struct bio *);
864 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
865 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
866 struct rq_map_data *, const struct iov_iter *,
868 extern void blk_execute_rq(struct request_queue *, struct gendisk *,
869 struct request *, int);
870 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
871 struct request *, int, rq_end_io_fn *);
873 /* Helper to convert REQ_OP_XXX to its string format XXX */
874 extern const char *blk_op_str(unsigned int op);
876 int blk_status_to_errno(blk_status_t status);
877 blk_status_t errno_to_blk_status(int errno);
879 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
881 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
883 return bdev->bd_disk->queue; /* this is never NULL */
887 * The basic unit of block I/O is a sector. It is used in a number of contexts
888 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
889 * bytes. Variables of type sector_t represent an offset or size that is a
890 * multiple of 512 bytes. Hence these two constants.
893 #define SECTOR_SHIFT 9
896 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
900 * blk_rq_pos() : the current sector
901 * blk_rq_bytes() : bytes left in the entire request
902 * blk_rq_cur_bytes() : bytes left in the current segment
903 * blk_rq_err_bytes() : bytes left till the next error boundary
904 * blk_rq_sectors() : sectors left in the entire request
905 * blk_rq_cur_sectors() : sectors left in the current segment
907 static inline sector_t blk_rq_pos(const struct request *rq)
912 static inline unsigned int blk_rq_bytes(const struct request *rq)
914 return rq->__data_len;
917 static inline int blk_rq_cur_bytes(const struct request *rq)
919 return rq->bio ? bio_cur_bytes(rq->bio) : 0;
922 extern unsigned int blk_rq_err_bytes(const struct request *rq);
924 static inline unsigned int blk_rq_sectors(const struct request *rq)
926 return blk_rq_bytes(rq) >> SECTOR_SHIFT;
929 static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
931 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
934 #ifdef CONFIG_BLK_DEV_ZONED
935 static inline unsigned int blk_rq_zone_no(struct request *rq)
937 return blk_queue_zone_no(rq->q, blk_rq_pos(rq));
940 static inline unsigned int blk_rq_zone_is_seq(struct request *rq)
942 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq));
944 #endif /* CONFIG_BLK_DEV_ZONED */
947 * Some commands like WRITE SAME have a payload or data transfer size which
948 * is different from the size of the request. Any driver that supports such
949 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to
950 * calculate the data transfer size.
952 static inline unsigned int blk_rq_payload_bytes(struct request *rq)
954 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
955 return rq->special_vec.bv_len;
956 return blk_rq_bytes(rq);
960 * Return the first full biovec in the request. The caller needs to check that
961 * there are any bvecs before calling this helper.
963 static inline struct bio_vec req_bvec(struct request *rq)
965 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
966 return rq->special_vec;
967 return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter);
970 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
973 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
974 return min(q->limits.max_discard_sectors,
975 UINT_MAX >> SECTOR_SHIFT);
977 if (unlikely(op == REQ_OP_WRITE_SAME))
978 return q->limits.max_write_same_sectors;
980 if (unlikely(op == REQ_OP_WRITE_ZEROES))
981 return q->limits.max_write_zeroes_sectors;
983 return q->limits.max_sectors;
987 * Return maximum size of a request at given offset. Only valid for
988 * file system requests.
990 static inline unsigned int blk_max_size_offset(struct request_queue *q,
993 if (!q->limits.chunk_sectors)
994 return q->limits.max_sectors;
996 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors -
997 (offset & (q->limits.chunk_sectors - 1))));
1000 static inline unsigned int blk_rq_get_max_sectors(struct request *rq,
1003 struct request_queue *q = rq->q;
1005 if (blk_rq_is_passthrough(rq))
1006 return q->limits.max_hw_sectors;
1008 if (!q->limits.chunk_sectors ||
1009 req_op(rq) == REQ_OP_DISCARD ||
1010 req_op(rq) == REQ_OP_SECURE_ERASE)
1011 return blk_queue_get_max_sectors(q, req_op(rq));
1013 return min(blk_max_size_offset(q, offset),
1014 blk_queue_get_max_sectors(q, req_op(rq)));
1017 static inline unsigned int blk_rq_count_bios(struct request *rq)
1019 unsigned int nr_bios = 0;
1022 __rq_for_each_bio(bio, rq)
1028 void blk_steal_bios(struct bio_list *list, struct request *rq);
1031 * Request completion related functions.
1033 * blk_update_request() completes given number of bytes and updates
1034 * the request without completing it.
1036 extern bool blk_update_request(struct request *rq, blk_status_t error,
1037 unsigned int nr_bytes);
1039 extern void __blk_complete_request(struct request *);
1040 extern void blk_abort_request(struct request *);
1043 * Access functions for manipulating queue properties
1045 extern void blk_cleanup_queue(struct request_queue *);
1046 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
1047 extern void blk_queue_bounce_limit(struct request_queue *, u64);
1048 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
1049 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
1050 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
1051 extern void blk_queue_max_discard_segments(struct request_queue *,
1053 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
1054 extern void blk_queue_max_discard_sectors(struct request_queue *q,
1055 unsigned int max_discard_sectors);
1056 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
1057 unsigned int max_write_same_sectors);
1058 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
1059 unsigned int max_write_same_sectors);
1060 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short);
1061 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
1062 extern void blk_queue_alignment_offset(struct request_queue *q,
1063 unsigned int alignment);
1064 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
1065 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
1066 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
1067 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
1068 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1069 extern void blk_set_default_limits(struct queue_limits *lim);
1070 extern void blk_set_stacking_limits(struct queue_limits *lim);
1071 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1073 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev,
1075 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
1077 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
1078 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
1079 extern int blk_queue_dma_drain(struct request_queue *q,
1080 dma_drain_needed_fn *dma_drain_needed,
1081 void *buf, unsigned int size);
1082 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
1083 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
1084 extern void blk_queue_dma_alignment(struct request_queue *, int);
1085 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
1086 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1087 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
1090 * Number of physical segments as sent to the device.
1092 * Normally this is the number of discontiguous data segments sent by the
1093 * submitter. But for data-less command like discard we might have no
1094 * actual data segments submitted, but the driver might have to add it's
1095 * own special payload. In that case we still return 1 here so that this
1096 * special payload will be mapped.
1098 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq)
1100 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1102 return rq->nr_phys_segments;
1106 * Number of discard segments (or ranges) the driver needs to fill in.
1107 * Each discard bio merged into a request is counted as one segment.
1109 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq)
1111 return max_t(unsigned short, rq->nr_phys_segments, 1);
1114 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
1115 extern void blk_dump_rq_flags(struct request *, char *);
1116 extern long nr_blockdev_pages(void);
1118 bool __must_check blk_get_queue(struct request_queue *);
1119 struct request_queue *blk_alloc_queue(gfp_t);
1120 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id);
1121 extern void blk_put_queue(struct request_queue *);
1122 extern void blk_set_queue_dying(struct request_queue *);
1125 * blk_plug permits building a queue of related requests by holding the I/O
1126 * fragments for a short period. This allows merging of sequential requests
1127 * into single larger request. As the requests are moved from a per-task list to
1128 * the device's request_queue in a batch, this results in improved scalability
1129 * as the lock contention for request_queue lock is reduced.
1131 * It is ok not to disable preemption when adding the request to the plug list
1132 * or when attempting a merge, because blk_schedule_flush_list() will only flush
1133 * the plug list when the task sleeps by itself. For details, please see
1134 * schedule() where blk_schedule_flush_plug() is called.
1137 struct list_head mq_list; /* blk-mq requests */
1138 struct list_head cb_list; /* md requires an unplug callback */
1139 unsigned short rq_count;
1140 bool multiple_queues;
1142 #define BLK_MAX_REQUEST_COUNT 16
1143 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
1146 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1147 struct blk_plug_cb {
1148 struct list_head list;
1149 blk_plug_cb_fn callback;
1152 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1153 void *data, int size);
1154 extern void blk_start_plug(struct blk_plug *);
1155 extern void blk_finish_plug(struct blk_plug *);
1156 extern void blk_flush_plug_list(struct blk_plug *, bool);
1158 static inline void blk_flush_plug(struct task_struct *tsk)
1160 struct blk_plug *plug = tsk->plug;
1163 blk_flush_plug_list(plug, false);
1166 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
1168 struct blk_plug *plug = tsk->plug;
1171 blk_flush_plug_list(plug, true);
1174 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1176 struct blk_plug *plug = tsk->plug;
1179 (!list_empty(&plug->mq_list) ||
1180 !list_empty(&plug->cb_list));
1183 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
1184 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
1185 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
1187 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
1189 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1190 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
1191 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1192 sector_t nr_sects, gfp_t gfp_mask, int flags,
1195 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1196 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1198 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1199 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1201 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1202 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1204 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1205 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1207 return blkdev_issue_discard(sb->s_bdev,
1208 block << (sb->s_blocksize_bits -
1210 nr_blocks << (sb->s_blocksize_bits -
1214 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1215 sector_t nr_blocks, gfp_t gfp_mask)
1217 return blkdev_issue_zeroout(sb->s_bdev,
1218 block << (sb->s_blocksize_bits -
1220 nr_blocks << (sb->s_blocksize_bits -
1225 extern int blk_verify_command(unsigned char *cmd, fmode_t mode);
1227 enum blk_default_limits {
1228 BLK_MAX_SEGMENTS = 128,
1229 BLK_SAFE_MAX_SECTORS = 255,
1230 BLK_DEF_MAX_SECTORS = 2560,
1231 BLK_MAX_SEGMENT_SIZE = 65536,
1232 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1235 static inline unsigned long queue_segment_boundary(struct request_queue *q)
1237 return q->limits.seg_boundary_mask;
1240 static inline unsigned long queue_virt_boundary(struct request_queue *q)
1242 return q->limits.virt_boundary_mask;
1245 static inline unsigned int queue_max_sectors(struct request_queue *q)
1247 return q->limits.max_sectors;
1250 static inline unsigned int queue_max_hw_sectors(struct request_queue *q)
1252 return q->limits.max_hw_sectors;
1255 static inline unsigned short queue_max_segments(struct request_queue *q)
1257 return q->limits.max_segments;
1260 static inline unsigned short queue_max_discard_segments(struct request_queue *q)
1262 return q->limits.max_discard_segments;
1265 static inline unsigned int queue_max_segment_size(struct request_queue *q)
1267 return q->limits.max_segment_size;
1270 static inline unsigned short queue_logical_block_size(struct request_queue *q)
1274 if (q && q->limits.logical_block_size)
1275 retval = q->limits.logical_block_size;
1280 static inline unsigned short bdev_logical_block_size(struct block_device *bdev)
1282 return queue_logical_block_size(bdev_get_queue(bdev));
1285 static inline unsigned int queue_physical_block_size(struct request_queue *q)
1287 return q->limits.physical_block_size;
1290 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1292 return queue_physical_block_size(bdev_get_queue(bdev));
1295 static inline unsigned int queue_io_min(struct request_queue *q)
1297 return q->limits.io_min;
1300 static inline int bdev_io_min(struct block_device *bdev)
1302 return queue_io_min(bdev_get_queue(bdev));
1305 static inline unsigned int queue_io_opt(struct request_queue *q)
1307 return q->limits.io_opt;
1310 static inline int bdev_io_opt(struct block_device *bdev)
1312 return queue_io_opt(bdev_get_queue(bdev));
1315 static inline int queue_alignment_offset(struct request_queue *q)
1317 if (q->limits.misaligned)
1320 return q->limits.alignment_offset;
1323 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
1325 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1326 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1329 return (granularity + lim->alignment_offset - alignment) % granularity;
1332 static inline int bdev_alignment_offset(struct block_device *bdev)
1334 struct request_queue *q = bdev_get_queue(bdev);
1336 if (q->limits.misaligned)
1339 if (bdev != bdev->bd_contains)
1340 return bdev->bd_part->alignment_offset;
1342 return q->limits.alignment_offset;
1345 static inline int queue_discard_alignment(struct request_queue *q)
1347 if (q->limits.discard_misaligned)
1350 return q->limits.discard_alignment;
1353 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1355 unsigned int alignment, granularity, offset;
1357 if (!lim->max_discard_sectors)
1360 /* Why are these in bytes, not sectors? */
1361 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1362 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1366 /* Offset of the partition start in 'granularity' sectors */
1367 offset = sector_div(sector, granularity);
1369 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1370 offset = (granularity + alignment - offset) % granularity;
1372 /* Turn it back into bytes, gaah */
1373 return offset << SECTOR_SHIFT;
1376 static inline int bdev_discard_alignment(struct block_device *bdev)
1378 struct request_queue *q = bdev_get_queue(bdev);
1380 if (bdev != bdev->bd_contains)
1381 return bdev->bd_part->discard_alignment;
1383 return q->limits.discard_alignment;
1386 static inline unsigned int bdev_write_same(struct block_device *bdev)
1388 struct request_queue *q = bdev_get_queue(bdev);
1391 return q->limits.max_write_same_sectors;
1396 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1398 struct request_queue *q = bdev_get_queue(bdev);
1401 return q->limits.max_write_zeroes_sectors;
1406 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1408 struct request_queue *q = bdev_get_queue(bdev);
1411 return blk_queue_zoned_model(q);
1413 return BLK_ZONED_NONE;
1416 static inline bool bdev_is_zoned(struct block_device *bdev)
1418 struct request_queue *q = bdev_get_queue(bdev);
1421 return blk_queue_is_zoned(q);
1426 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1428 struct request_queue *q = bdev_get_queue(bdev);
1431 return blk_queue_zone_sectors(q);
1435 static inline int queue_dma_alignment(struct request_queue *q)
1437 return q ? q->dma_alignment : 511;
1440 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1443 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1444 return !(addr & alignment) && !(len & alignment);
1447 /* assumes size > 256 */
1448 static inline unsigned int blksize_bits(unsigned int size)
1450 unsigned int bits = 8;
1454 } while (size > 256);
1458 static inline unsigned int block_size(struct block_device *bdev)
1460 return bdev->bd_block_size;
1463 typedef struct {struct page *v;} Sector;
1465 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
1467 static inline void put_dev_sector(Sector p)
1472 int kblockd_schedule_work(struct work_struct *work);
1473 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
1474 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1476 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1477 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1478 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1479 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1481 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1483 enum blk_integrity_flags {
1484 BLK_INTEGRITY_VERIFY = 1 << 0,
1485 BLK_INTEGRITY_GENERATE = 1 << 1,
1486 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2,
1487 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3,
1490 struct blk_integrity_iter {
1494 unsigned int data_size;
1495 unsigned short interval;
1496 const char *disk_name;
1499 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *);
1501 struct blk_integrity_profile {
1502 integrity_processing_fn *generate_fn;
1503 integrity_processing_fn *verify_fn;
1507 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *);
1508 extern void blk_integrity_unregister(struct gendisk *);
1509 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1510 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *,
1511 struct scatterlist *);
1512 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
1513 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *,
1515 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *,
1518 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1520 struct blk_integrity *bi = &disk->queue->integrity;
1529 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1531 return blk_get_integrity(bdev->bd_disk);
1534 static inline bool blk_integrity_rq(struct request *rq)
1536 return rq->cmd_flags & REQ_INTEGRITY;
1539 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1542 q->limits.max_integrity_segments = segs;
1545 static inline unsigned short
1546 queue_max_integrity_segments(struct request_queue *q)
1548 return q->limits.max_integrity_segments;
1552 * bio_integrity_intervals - Return number of integrity intervals for a bio
1553 * @bi: blk_integrity profile for device
1554 * @sectors: Size of the bio in 512-byte sectors
1556 * Description: The block layer calculates everything in 512 byte
1557 * sectors but integrity metadata is done in terms of the data integrity
1558 * interval size of the storage device. Convert the block layer sectors
1559 * to the appropriate number of integrity intervals.
1561 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1562 unsigned int sectors)
1564 return sectors >> (bi->interval_exp - 9);
1567 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1568 unsigned int sectors)
1570 return bio_integrity_intervals(bi, sectors) * bi->tuple_size;
1574 * Return the first bvec that contains integrity data. Only drivers that are
1575 * limited to a single integrity segment should use this helper.
1577 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1579 if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1))
1581 return rq->bio->bi_integrity->bip_vec;
1584 #else /* CONFIG_BLK_DEV_INTEGRITY */
1587 struct block_device;
1589 struct blk_integrity;
1591 static inline int blk_integrity_rq(struct request *rq)
1595 static inline int blk_rq_count_integrity_sg(struct request_queue *q,
1600 static inline int blk_rq_map_integrity_sg(struct request_queue *q,
1602 struct scatterlist *s)
1606 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b)
1610 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1614 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b)
1618 static inline void blk_integrity_register(struct gendisk *d,
1619 struct blk_integrity *b)
1622 static inline void blk_integrity_unregister(struct gendisk *d)
1625 static inline void blk_queue_max_integrity_segments(struct request_queue *q,
1629 static inline unsigned short queue_max_integrity_segments(struct request_queue *q)
1633 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
1639 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
1646 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi,
1647 unsigned int sectors)
1652 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi,
1653 unsigned int sectors)
1658 static inline struct bio_vec *rq_integrity_vec(struct request *rq)
1663 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1665 struct block_device_operations {
1666 int (*open) (struct block_device *, fmode_t);
1667 void (*release) (struct gendisk *, fmode_t);
1668 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1669 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1670 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1671 unsigned int (*check_events) (struct gendisk *disk,
1672 unsigned int clearing);
1673 /* ->media_changed() is DEPRECATED, use ->check_events() instead */
1674 int (*media_changed) (struct gendisk *);
1675 void (*unlock_native_capacity) (struct gendisk *);
1676 int (*revalidate_disk) (struct gendisk *);
1677 int (*getgeo)(struct block_device *, struct hd_geometry *);
1678 /* this callback is with swap_lock and sometimes page table lock held */
1679 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1680 int (*report_zones)(struct gendisk *, sector_t sector,
1681 struct blk_zone *zones, unsigned int *nr_zones);
1682 struct module *owner;
1683 const struct pr_ops *pr_ops;
1686 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1688 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1689 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1690 struct writeback_control *);
1692 #ifdef CONFIG_BLK_DEV_ZONED
1693 bool blk_req_needs_zone_write_lock(struct request *rq);
1694 void __blk_req_zone_write_lock(struct request *rq);
1695 void __blk_req_zone_write_unlock(struct request *rq);
1697 static inline void blk_req_zone_write_lock(struct request *rq)
1699 if (blk_req_needs_zone_write_lock(rq))
1700 __blk_req_zone_write_lock(rq);
1703 static inline void blk_req_zone_write_unlock(struct request *rq)
1705 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED)
1706 __blk_req_zone_write_unlock(rq);
1709 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1711 return rq->q->seq_zones_wlock &&
1712 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock);
1715 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1717 if (!blk_req_needs_zone_write_lock(rq))
1719 return !blk_req_zone_is_write_locked(rq);
1722 static inline bool blk_req_needs_zone_write_lock(struct request *rq)
1727 static inline void blk_req_zone_write_lock(struct request *rq)
1731 static inline void blk_req_zone_write_unlock(struct request *rq)
1734 static inline bool blk_req_zone_is_write_locked(struct request *rq)
1739 static inline bool blk_req_can_dispatch_to_zone(struct request *rq)
1743 #endif /* CONFIG_BLK_DEV_ZONED */
1745 #else /* CONFIG_BLOCK */
1747 struct block_device;
1750 * stubs for when the block layer is configured out
1752 #define buffer_heads_over_limit 0
1754 static inline long nr_blockdev_pages(void)
1762 static inline void blk_start_plug(struct blk_plug *plug)
1766 static inline void blk_finish_plug(struct blk_plug *plug)
1770 static inline void blk_flush_plug(struct task_struct *task)
1774 static inline void blk_schedule_flush_plug(struct task_struct *task)
1779 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
1784 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask,
1785 sector_t *error_sector)
1790 #endif /* CONFIG_BLOCK */
1792 static inline void blk_wake_io_task(struct task_struct *waiter)
1795 * If we're polling, the task itself is doing the completions. For
1796 * that case, we don't need to signal a wakeup, it's enough to just
1797 * mark us as RUNNING.
1799 if (waiter == current)
1800 __set_current_state(TASK_RUNNING);
1802 wake_up_process(waiter);