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/genhd.h>
7 #include <linux/list.h>
8 #include <linux/llist.h>
9 #include <linux/minmax.h>
10 #include <linux/timer.h>
11 #include <linux/workqueue.h>
12 #include <linux/wait.h>
13 #include <linux/bio.h>
14 #include <linux/gfp.h>
15 #include <linux/rcupdate.h>
16 #include <linux/percpu-refcount.h>
17 #include <linux/blkzoned.h>
18 #include <linux/sbitmap.h>
22 struct elevator_queue;
27 struct blk_flush_queue;
31 struct blk_queue_stats;
32 struct blk_stat_callback;
33 struct blk_keyslot_manager;
35 /* Must be consistent with blk_mq_poll_stats_bkt() */
36 #define BLK_MQ_POLL_STATS_BKTS 16
38 /* Doing classic polling */
39 #define BLK_MQ_POLL_CLASSIC -1
42 * Maximum number of blkcg policies allowed to be registered concurrently.
43 * Defined here to simplify include dependency.
45 #define BLKCG_MAX_POLS 6
47 static inline bool blk_op_is_passthrough(unsigned int op)
50 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
54 * Zoned block device models (zoned limit).
56 * Note: This needs to be ordered from the least to the most severe
57 * restrictions for the inheritance in blk_stack_limits() to work.
59 enum blk_zoned_model {
60 BLK_ZONED_NONE = 0, /* Regular block device */
61 BLK_ZONED_HA, /* Host-aware zoned block device */
62 BLK_ZONED_HM, /* Host-managed zoned block device */
66 * BLK_BOUNCE_NONE: never bounce (default)
67 * BLK_BOUNCE_HIGH: bounce all highmem pages
75 enum blk_bounce bounce;
76 unsigned long seg_boundary_mask;
77 unsigned long virt_boundary_mask;
79 unsigned int max_hw_sectors;
80 unsigned int max_dev_sectors;
81 unsigned int chunk_sectors;
82 unsigned int max_sectors;
83 unsigned int max_segment_size;
84 unsigned int physical_block_size;
85 unsigned int logical_block_size;
86 unsigned int alignment_offset;
89 unsigned int max_discard_sectors;
90 unsigned int max_hw_discard_sectors;
91 unsigned int max_write_same_sectors;
92 unsigned int max_write_zeroes_sectors;
93 unsigned int max_zone_append_sectors;
94 unsigned int discard_granularity;
95 unsigned int discard_alignment;
96 unsigned int zone_write_granularity;
98 unsigned short max_segments;
99 unsigned short max_integrity_segments;
100 unsigned short max_discard_segments;
102 unsigned char misaligned;
103 unsigned char discard_misaligned;
104 unsigned char raid_partial_stripes_expensive;
105 enum blk_zoned_model zoned;
108 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
111 void blk_queue_set_zoned(struct gendisk *disk, enum blk_zoned_model model);
113 #ifdef CONFIG_BLK_DEV_ZONED
115 #define BLK_ALL_ZONES ((unsigned int)-1)
116 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
117 unsigned int nr_zones, report_zones_cb cb, void *data);
118 unsigned int blkdev_nr_zones(struct gendisk *disk);
119 extern int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
120 sector_t sectors, sector_t nr_sectors,
122 int blk_revalidate_disk_zones(struct gendisk *disk,
123 void (*update_driver_data)(struct gendisk *disk));
125 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
126 unsigned int cmd, unsigned long arg);
127 extern int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
128 unsigned int cmd, unsigned long arg);
130 #else /* CONFIG_BLK_DEV_ZONED */
132 static inline unsigned int blkdev_nr_zones(struct gendisk *disk)
137 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
138 fmode_t mode, unsigned int cmd,
144 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
145 fmode_t mode, unsigned int cmd,
151 #endif /* CONFIG_BLK_DEV_ZONED */
153 struct request_queue {
154 struct request *last_merge;
155 struct elevator_queue *elevator;
157 struct percpu_ref q_usage_counter;
159 struct blk_queue_stats *stats;
160 struct rq_qos *rq_qos;
162 const struct blk_mq_ops *mq_ops;
165 struct blk_mq_ctx __percpu *queue_ctx;
167 unsigned int queue_depth;
169 /* hw dispatch queues */
170 struct blk_mq_hw_ctx **queue_hw_ctx;
171 unsigned int nr_hw_queues;
174 * The queue owner gets to use this for whatever they like.
175 * ll_rw_blk doesn't touch it.
180 * various queue flags, see QUEUE_* below
182 unsigned long queue_flags;
184 * Number of contexts that have called blk_set_pm_only(). If this
185 * counter is above zero then only RQF_PM requests are processed.
190 * ida allocated id for this queue. Used to index queues from
195 spinlock_t queue_lock;
197 struct gendisk *disk;
207 struct kobject *mq_kobj;
209 #ifdef CONFIG_BLK_DEV_INTEGRITY
210 struct blk_integrity integrity;
211 #endif /* CONFIG_BLK_DEV_INTEGRITY */
215 enum rpm_status rpm_status;
221 unsigned long nr_requests; /* Max # of requests */
223 unsigned int dma_pad_mask;
224 unsigned int dma_alignment;
226 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
227 /* Inline crypto capabilities */
228 struct blk_keyslot_manager *ksm;
231 unsigned int rq_timeout;
234 struct blk_stat_callback *poll_cb;
235 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS];
237 struct timer_list timeout;
238 struct work_struct timeout_work;
240 atomic_t nr_active_requests_shared_tags;
242 struct blk_mq_tags *sched_shared_tags;
244 struct list_head icq_list;
245 #ifdef CONFIG_BLK_CGROUP
246 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
247 struct blkcg_gq *root_blkg;
248 struct list_head blkg_list;
251 struct queue_limits limits;
253 unsigned int required_elevator_features;
255 #ifdef CONFIG_BLK_DEV_ZONED
257 * Zoned block device information for request dispatch control.
258 * nr_zones is the total number of zones of the device. This is always
259 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
260 * bits which indicates if a zone is conventional (bit set) or
261 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
262 * bits which indicates if a zone is write locked, that is, if a write
263 * request targeting the zone was dispatched. All three fields are
264 * initialized by the low level device driver (e.g. scsi/sd.c).
265 * Stacking drivers (device mappers) may or may not initialize
268 * Reads of this information must be protected with blk_queue_enter() /
269 * blk_queue_exit(). Modifying this information is only allowed while
270 * no requests are being processed. See also blk_mq_freeze_queue() and
271 * blk_mq_unfreeze_queue().
273 unsigned int nr_zones;
274 unsigned long *conv_zones_bitmap;
275 unsigned long *seq_zones_wlock;
276 unsigned int max_open_zones;
277 unsigned int max_active_zones;
278 #endif /* CONFIG_BLK_DEV_ZONED */
281 struct mutex debugfs_mutex;
282 #ifdef CONFIG_BLK_DEV_IO_TRACE
283 struct blk_trace __rcu *blk_trace;
286 * for flush operations
288 struct blk_flush_queue *fq;
290 struct list_head requeue_list;
291 spinlock_t requeue_lock;
292 struct delayed_work requeue_work;
294 struct mutex sysfs_lock;
295 struct mutex sysfs_dir_lock;
298 * for reusing dead hctx instance in case of updating
301 struct list_head unused_hctx_list;
302 spinlock_t unused_hctx_lock;
306 #ifdef CONFIG_BLK_DEV_THROTTLING
308 struct throtl_data *td;
310 struct rcu_head rcu_head;
311 wait_queue_head_t mq_freeze_wq;
313 * Protect concurrent access to q_usage_counter by
314 * percpu_ref_kill() and percpu_ref_reinit().
316 struct mutex mq_freeze_lock;
318 struct blk_mq_tag_set *tag_set;
319 struct list_head tag_set_list;
320 struct bio_set bio_split;
322 struct dentry *debugfs_dir;
324 #ifdef CONFIG_BLK_DEBUG_FS
325 struct dentry *sched_debugfs_dir;
326 struct dentry *rqos_debugfs_dir;
329 bool mq_sysfs_init_done;
331 #define BLK_MAX_WRITE_HINTS 5
332 u64 write_hints[BLK_MAX_WRITE_HINTS];
335 /* Keep blk_queue_flag_name[] in sync with the definitions below */
336 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */
337 #define QUEUE_FLAG_DYING 1 /* queue being torn down */
338 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */
339 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */
340 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */
341 #define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */
342 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
343 #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */
344 #define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */
345 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */
346 #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */
347 #define QUEUE_FLAG_SECERASE 11 /* supports secure erase */
348 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */
349 #define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */
350 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */
351 #define QUEUE_FLAG_STABLE_WRITES 15 /* don't modify blks until WB is done */
352 #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */
353 #define QUEUE_FLAG_WC 17 /* Write back caching */
354 #define QUEUE_FLAG_FUA 18 /* device supports FUA writes */
355 #define QUEUE_FLAG_DAX 19 /* device supports DAX */
356 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */
357 #define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */
358 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */
359 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */
360 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */
361 #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */
362 #define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */
363 #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */
364 #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */
365 #define QUEUE_FLAG_NOWAIT 29 /* device supports NOWAIT */
367 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
368 (1 << QUEUE_FLAG_SAME_COMP) | \
369 (1 << QUEUE_FLAG_NOWAIT))
371 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
372 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
373 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
375 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
376 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
377 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
378 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
379 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
380 #define blk_queue_noxmerges(q) \
381 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
382 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
383 #define blk_queue_stable_writes(q) \
384 test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
385 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
386 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
387 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags)
388 #define blk_queue_zone_resetall(q) \
389 test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
390 #define blk_queue_secure_erase(q) \
391 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
392 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
393 #define blk_queue_scsi_passthrough(q) \
394 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags)
395 #define blk_queue_pci_p2pdma(q) \
396 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
397 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
398 #define blk_queue_rq_alloc_time(q) \
399 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
401 #define blk_queue_rq_alloc_time(q) false
404 #define blk_noretry_request(rq) \
405 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
406 REQ_FAILFAST_DRIVER))
407 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
408 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
409 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags)
410 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
411 #define blk_queue_nowait(q) test_bit(QUEUE_FLAG_NOWAIT, &(q)->queue_flags)
413 extern void blk_set_pm_only(struct request_queue *q);
414 extern void blk_clear_pm_only(struct request_queue *q);
416 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
418 #define dma_map_bvec(dev, bv, dir, attrs) \
419 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
422 static inline bool queue_is_mq(struct request_queue *q)
428 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
430 return q->rpm_status;
433 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
439 static inline enum blk_zoned_model
440 blk_queue_zoned_model(struct request_queue *q)
442 if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
443 return q->limits.zoned;
444 return BLK_ZONED_NONE;
447 static inline bool blk_queue_is_zoned(struct request_queue *q)
449 switch (blk_queue_zoned_model(q)) {
458 static inline sector_t blk_queue_zone_sectors(struct request_queue *q)
460 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0;
463 #ifdef CONFIG_BLK_DEV_ZONED
464 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
466 return blk_queue_is_zoned(q) ? q->nr_zones : 0;
469 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
472 if (!blk_queue_is_zoned(q))
474 return sector >> ilog2(q->limits.chunk_sectors);
477 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
480 if (!blk_queue_is_zoned(q))
482 if (!q->conv_zones_bitmap)
484 return !test_bit(blk_queue_zone_no(q, sector), q->conv_zones_bitmap);
487 static inline void blk_queue_max_open_zones(struct request_queue *q,
488 unsigned int max_open_zones)
490 q->max_open_zones = max_open_zones;
493 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
495 return q->max_open_zones;
498 static inline void blk_queue_max_active_zones(struct request_queue *q,
499 unsigned int max_active_zones)
501 q->max_active_zones = max_active_zones;
504 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
506 return q->max_active_zones;
508 #else /* CONFIG_BLK_DEV_ZONED */
509 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
513 static inline bool blk_queue_zone_is_seq(struct request_queue *q,
518 static inline unsigned int blk_queue_zone_no(struct request_queue *q,
523 static inline unsigned int queue_max_open_zones(const struct request_queue *q)
527 static inline unsigned int queue_max_active_zones(const struct request_queue *q)
531 #endif /* CONFIG_BLK_DEV_ZONED */
533 static inline unsigned int blk_queue_depth(struct request_queue *q)
536 return q->queue_depth;
538 return q->nr_requests;
542 * default timeout for SG_IO if none specified
544 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
545 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
547 /* This should not be used directly - use rq_for_each_segment */
548 #define for_each_bio(_bio) \
549 for (; _bio; _bio = _bio->bi_next)
552 extern int blk_register_queue(struct gendisk *disk);
553 extern void blk_unregister_queue(struct gendisk *disk);
554 void submit_bio_noacct(struct bio *bio);
556 extern int blk_lld_busy(struct request_queue *q);
557 extern void blk_queue_split(struct bio **);
558 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
559 extern void blk_queue_exit(struct request_queue *q);
560 extern void blk_sync_queue(struct request_queue *q);
562 /* Helper to convert REQ_OP_XXX to its string format XXX */
563 extern const char *blk_op_str(unsigned int op);
565 int blk_status_to_errno(blk_status_t status);
566 blk_status_t errno_to_blk_status(int errno);
568 /* only poll the hardware once, don't continue until a completion was found */
569 #define BLK_POLL_ONESHOT (1 << 0)
570 /* do not sleep to wait for the expected completion time */
571 #define BLK_POLL_NOSLEEP (1 << 1)
572 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags);
573 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob,
576 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
578 return bdev->bd_queue; /* this is never NULL */
582 * The basic unit of block I/O is a sector. It is used in a number of contexts
583 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
584 * bytes. Variables of type sector_t represent an offset or size that is a
585 * multiple of 512 bytes. Hence these two constants.
588 #define SECTOR_SHIFT 9
591 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
594 #define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
595 #define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
596 #define SECTOR_MASK (PAGE_SECTORS - 1)
598 #ifdef CONFIG_BLK_DEV_ZONED
600 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
601 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
603 static inline unsigned int bio_zone_no(struct bio *bio)
605 return blk_queue_zone_no(bdev_get_queue(bio->bi_bdev),
606 bio->bi_iter.bi_sector);
609 static inline unsigned int bio_zone_is_seq(struct bio *bio)
611 return blk_queue_zone_is_seq(bdev_get_queue(bio->bi_bdev),
612 bio->bi_iter.bi_sector);
614 #endif /* CONFIG_BLK_DEV_ZONED */
616 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q,
619 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
620 return min(q->limits.max_discard_sectors,
621 UINT_MAX >> SECTOR_SHIFT);
623 if (unlikely(op == REQ_OP_WRITE_SAME))
624 return q->limits.max_write_same_sectors;
626 if (unlikely(op == REQ_OP_WRITE_ZEROES))
627 return q->limits.max_write_zeroes_sectors;
629 return q->limits.max_sectors;
633 * Return maximum size of a request at given offset. Only valid for
634 * file system requests.
636 static inline unsigned int blk_max_size_offset(struct request_queue *q,
638 unsigned int chunk_sectors)
640 if (!chunk_sectors) {
641 if (q->limits.chunk_sectors)
642 chunk_sectors = q->limits.chunk_sectors;
644 return q->limits.max_sectors;
647 if (likely(is_power_of_2(chunk_sectors)))
648 chunk_sectors -= offset & (chunk_sectors - 1);
650 chunk_sectors -= sector_div(offset, chunk_sectors);
652 return min(q->limits.max_sectors, chunk_sectors);
656 * Access functions for manipulating queue properties
658 extern void blk_cleanup_queue(struct request_queue *);
659 void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
660 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
661 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
662 extern void blk_queue_max_segments(struct request_queue *, unsigned short);
663 extern void blk_queue_max_discard_segments(struct request_queue *,
665 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
666 extern void blk_queue_max_discard_sectors(struct request_queue *q,
667 unsigned int max_discard_sectors);
668 extern void blk_queue_max_write_same_sectors(struct request_queue *q,
669 unsigned int max_write_same_sectors);
670 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
671 unsigned int max_write_same_sectors);
672 extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
673 extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
674 unsigned int max_zone_append_sectors);
675 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
676 void blk_queue_zone_write_granularity(struct request_queue *q,
678 extern void blk_queue_alignment_offset(struct request_queue *q,
679 unsigned int alignment);
680 void disk_update_readahead(struct gendisk *disk);
681 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
682 extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
683 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
684 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
685 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
686 extern void blk_set_default_limits(struct queue_limits *lim);
687 extern void blk_set_stacking_limits(struct queue_limits *lim);
688 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
690 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev,
692 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
693 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
694 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
695 extern void blk_queue_dma_alignment(struct request_queue *, int);
696 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
697 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
698 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
701 * Elevator features for blk_queue_required_elevator_features:
703 /* Supports zoned block devices sequential write constraint */
704 #define ELEVATOR_F_ZBD_SEQ_WRITE (1U << 0)
705 /* Supports scheduling on multiple hardware queues */
706 #define ELEVATOR_F_MQ_AWARE (1U << 1)
708 extern void blk_queue_required_elevator_features(struct request_queue *q,
709 unsigned int features);
710 extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
713 bool __must_check blk_get_queue(struct request_queue *);
714 extern void blk_put_queue(struct request_queue *);
715 extern void blk_set_queue_dying(struct request_queue *);
719 * blk_plug permits building a queue of related requests by holding the I/O
720 * fragments for a short period. This allows merging of sequential requests
721 * into single larger request. As the requests are moved from a per-task list to
722 * the device's request_queue in a batch, this results in improved scalability
723 * as the lock contention for request_queue lock is reduced.
725 * It is ok not to disable preemption when adding the request to the plug list
726 * or when attempting a merge, because blk_schedule_flush_list() will only flush
727 * the plug list when the task sleeps by itself. For details, please see
728 * schedule() where blk_schedule_flush_plug() is called.
731 struct list_head mq_list; /* blk-mq requests */
733 /* if ios_left is > 1, we can batch tag/rq allocations */
734 struct request *cached_rq;
735 unsigned short nr_ios;
737 unsigned short rq_count;
739 bool multiple_queues;
742 struct list_head cb_list; /* md requires an unplug callback */
746 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
748 struct list_head list;
749 blk_plug_cb_fn callback;
752 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
753 void *data, int size);
754 extern void blk_start_plug(struct blk_plug *);
755 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short);
756 extern void blk_finish_plug(struct blk_plug *);
757 extern void blk_flush_plug_list(struct blk_plug *, bool);
759 static inline void blk_flush_plug(struct task_struct *tsk)
761 struct blk_plug *plug = tsk->plug;
764 blk_flush_plug_list(plug, false);
767 static inline void blk_schedule_flush_plug(struct task_struct *tsk)
769 struct blk_plug *plug = tsk->plug;
772 blk_flush_plug_list(plug, true);
775 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
777 struct blk_plug *plug = tsk->plug;
780 (!list_empty(&plug->mq_list) ||
781 !list_empty(&plug->cb_list));
784 int blkdev_issue_flush(struct block_device *bdev);
785 long nr_blockdev_pages(void);
786 #else /* CONFIG_BLOCK */
790 static inline void blk_start_plug_nr_ios(struct blk_plug *plug,
791 unsigned short nr_ios)
795 static inline void blk_start_plug(struct blk_plug *plug)
799 static inline void blk_finish_plug(struct blk_plug *plug)
803 static inline void blk_flush_plug(struct task_struct *task)
807 static inline void blk_schedule_flush_plug(struct task_struct *task)
812 static inline bool blk_needs_flush_plug(struct task_struct *tsk)
817 static inline int blkdev_issue_flush(struct block_device *bdev)
822 static inline long nr_blockdev_pages(void)
826 #endif /* CONFIG_BLOCK */
828 extern void blk_io_schedule(void);
830 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
831 sector_t nr_sects, gfp_t gfp_mask, struct page *page);
833 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */
835 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
836 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags);
837 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
838 sector_t nr_sects, gfp_t gfp_mask, int flags,
841 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
842 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
844 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
845 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
847 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
848 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
850 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
851 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
853 return blkdev_issue_discard(sb->s_bdev,
854 block << (sb->s_blocksize_bits -
856 nr_blocks << (sb->s_blocksize_bits -
860 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
861 sector_t nr_blocks, gfp_t gfp_mask)
863 return blkdev_issue_zeroout(sb->s_bdev,
864 block << (sb->s_blocksize_bits -
866 nr_blocks << (sb->s_blocksize_bits -
871 static inline bool bdev_is_partition(struct block_device *bdev)
873 return bdev->bd_partno;
876 enum blk_default_limits {
877 BLK_MAX_SEGMENTS = 128,
878 BLK_SAFE_MAX_SECTORS = 255,
879 BLK_DEF_MAX_SECTORS = 2560,
880 BLK_MAX_SEGMENT_SIZE = 65536,
881 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
884 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
886 return q->limits.seg_boundary_mask;
889 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
891 return q->limits.virt_boundary_mask;
894 static inline unsigned int queue_max_sectors(const struct request_queue *q)
896 return q->limits.max_sectors;
899 static inline unsigned int queue_max_bytes(struct request_queue *q)
901 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9;
904 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
906 return q->limits.max_hw_sectors;
909 static inline unsigned short queue_max_segments(const struct request_queue *q)
911 return q->limits.max_segments;
914 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
916 return q->limits.max_discard_segments;
919 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
921 return q->limits.max_segment_size;
924 static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
927 const struct queue_limits *l = &q->limits;
929 return min(l->max_zone_append_sectors, l->max_sectors);
932 static inline unsigned queue_logical_block_size(const struct request_queue *q)
936 if (q && q->limits.logical_block_size)
937 retval = q->limits.logical_block_size;
942 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
944 return queue_logical_block_size(bdev_get_queue(bdev));
947 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
949 return q->limits.physical_block_size;
952 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
954 return queue_physical_block_size(bdev_get_queue(bdev));
957 static inline unsigned int queue_io_min(const struct request_queue *q)
959 return q->limits.io_min;
962 static inline int bdev_io_min(struct block_device *bdev)
964 return queue_io_min(bdev_get_queue(bdev));
967 static inline unsigned int queue_io_opt(const struct request_queue *q)
969 return q->limits.io_opt;
972 static inline int bdev_io_opt(struct block_device *bdev)
974 return queue_io_opt(bdev_get_queue(bdev));
977 static inline unsigned int
978 queue_zone_write_granularity(const struct request_queue *q)
980 return q->limits.zone_write_granularity;
983 static inline unsigned int
984 bdev_zone_write_granularity(struct block_device *bdev)
986 return queue_zone_write_granularity(bdev_get_queue(bdev));
989 static inline int queue_alignment_offset(const struct request_queue *q)
991 if (q->limits.misaligned)
994 return q->limits.alignment_offset;
997 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector)
999 unsigned int granularity = max(lim->physical_block_size, lim->io_min);
1000 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT)
1003 return (granularity + lim->alignment_offset - alignment) % granularity;
1006 static inline int bdev_alignment_offset(struct block_device *bdev)
1008 struct request_queue *q = bdev_get_queue(bdev);
1010 if (q->limits.misaligned)
1012 if (bdev_is_partition(bdev))
1013 return queue_limit_alignment_offset(&q->limits,
1014 bdev->bd_start_sect);
1015 return q->limits.alignment_offset;
1018 static inline int queue_discard_alignment(const struct request_queue *q)
1020 if (q->limits.discard_misaligned)
1023 return q->limits.discard_alignment;
1026 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
1028 unsigned int alignment, granularity, offset;
1030 if (!lim->max_discard_sectors)
1033 /* Why are these in bytes, not sectors? */
1034 alignment = lim->discard_alignment >> SECTOR_SHIFT;
1035 granularity = lim->discard_granularity >> SECTOR_SHIFT;
1039 /* Offset of the partition start in 'granularity' sectors */
1040 offset = sector_div(sector, granularity);
1042 /* And why do we do this modulus *again* in blkdev_issue_discard()? */
1043 offset = (granularity + alignment - offset) % granularity;
1045 /* Turn it back into bytes, gaah */
1046 return offset << SECTOR_SHIFT;
1049 static inline int bdev_discard_alignment(struct block_device *bdev)
1051 struct request_queue *q = bdev_get_queue(bdev);
1053 if (bdev_is_partition(bdev))
1054 return queue_limit_discard_alignment(&q->limits,
1055 bdev->bd_start_sect);
1056 return q->limits.discard_alignment;
1059 static inline unsigned int bdev_write_same(struct block_device *bdev)
1061 struct request_queue *q = bdev_get_queue(bdev);
1064 return q->limits.max_write_same_sectors;
1069 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1071 struct request_queue *q = bdev_get_queue(bdev);
1074 return q->limits.max_write_zeroes_sectors;
1079 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev)
1081 struct request_queue *q = bdev_get_queue(bdev);
1084 return blk_queue_zoned_model(q);
1086 return BLK_ZONED_NONE;
1089 static inline bool bdev_is_zoned(struct block_device *bdev)
1091 struct request_queue *q = bdev_get_queue(bdev);
1094 return blk_queue_is_zoned(q);
1099 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1101 struct request_queue *q = bdev_get_queue(bdev);
1104 return blk_queue_zone_sectors(q);
1108 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
1110 struct request_queue *q = bdev_get_queue(bdev);
1113 return queue_max_open_zones(q);
1117 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
1119 struct request_queue *q = bdev_get_queue(bdev);
1122 return queue_max_active_zones(q);
1126 static inline int queue_dma_alignment(const struct request_queue *q)
1128 return q ? q->dma_alignment : 511;
1131 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1134 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1135 return !(addr & alignment) && !(len & alignment);
1138 /* assumes size > 256 */
1139 static inline unsigned int blksize_bits(unsigned int size)
1141 unsigned int bits = 8;
1145 } while (size > 256);
1149 static inline unsigned int block_size(struct block_device *bdev)
1151 return 1 << bdev->bd_inode->i_blkbits;
1154 int kblockd_schedule_work(struct work_struct *work);
1155 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1157 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1158 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1159 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1160 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1162 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1164 bool blk_ksm_register(struct blk_keyslot_manager *ksm, struct request_queue *q);
1166 void blk_ksm_unregister(struct request_queue *q);
1168 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1170 static inline bool blk_ksm_register(struct blk_keyslot_manager *ksm,
1171 struct request_queue *q)
1176 static inline void blk_ksm_unregister(struct request_queue *q) { }
1178 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1181 struct block_device_operations {
1182 void (*submit_bio)(struct bio *bio);
1183 int (*open) (struct block_device *, fmode_t);
1184 void (*release) (struct gendisk *, fmode_t);
1185 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
1186 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1187 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1188 unsigned int (*check_events) (struct gendisk *disk,
1189 unsigned int clearing);
1190 void (*unlock_native_capacity) (struct gendisk *);
1191 int (*getgeo)(struct block_device *, struct hd_geometry *);
1192 int (*set_read_only)(struct block_device *bdev, bool ro);
1193 /* this callback is with swap_lock and sometimes page table lock held */
1194 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1195 int (*report_zones)(struct gendisk *, sector_t sector,
1196 unsigned int nr_zones, report_zones_cb cb, void *data);
1197 char *(*devnode)(struct gendisk *disk, umode_t *mode);
1198 struct module *owner;
1199 const struct pr_ops *pr_ops;
1202 * Special callback for probing GPT entry at a given sector.
1203 * Needed by Android devices, used by GPT scanner and MMC blk
1206 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector);
1209 #ifdef CONFIG_COMPAT
1210 extern int blkdev_compat_ptr_ioctl(struct block_device *, fmode_t,
1211 unsigned int, unsigned long);
1213 #define blkdev_compat_ptr_ioctl NULL
1216 extern int bdev_read_page(struct block_device *, sector_t, struct page *);
1217 extern int bdev_write_page(struct block_device *, sector_t, struct page *,
1218 struct writeback_control *);
1220 static inline void blk_wake_io_task(struct task_struct *waiter)
1223 * If we're polling, the task itself is doing the completions. For
1224 * that case, we don't need to signal a wakeup, it's enough to just
1225 * mark us as RUNNING.
1227 if (waiter == current)
1228 __set_current_state(TASK_RUNNING);
1230 wake_up_process(waiter);
1233 unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors,
1235 void disk_end_io_acct(struct gendisk *disk, unsigned int op,
1236 unsigned long start_time);
1238 unsigned long bio_start_io_acct(struct bio *bio);
1239 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1240 struct block_device *orig_bdev);
1243 * bio_end_io_acct - end I/O accounting for bio based drivers
1244 * @bio: bio to end account for
1245 * @start: start time returned by bio_start_io_acct()
1247 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1249 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1252 int bdev_read_only(struct block_device *bdev);
1253 int set_blocksize(struct block_device *bdev, int size);
1255 const char *bdevname(struct block_device *bdev, char *buffer);
1256 int lookup_bdev(const char *pathname, dev_t *dev);
1258 void blkdev_show(struct seq_file *seqf, off_t offset);
1260 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */
1261 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */
1263 #define BLKDEV_MAJOR_MAX 512
1265 #define BLKDEV_MAJOR_MAX 0
1268 struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
1270 struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder);
1271 int bd_prepare_to_claim(struct block_device *bdev, void *holder);
1272 void bd_abort_claiming(struct block_device *bdev, void *holder);
1273 void blkdev_put(struct block_device *bdev, fmode_t mode);
1275 /* just for blk-cgroup, don't use elsewhere */
1276 struct block_device *blkdev_get_no_open(dev_t dev);
1277 void blkdev_put_no_open(struct block_device *bdev);
1279 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
1280 void bdev_add(struct block_device *bdev, dev_t dev);
1281 struct block_device *I_BDEV(struct inode *inode);
1282 int truncate_bdev_range(struct block_device *bdev, fmode_t mode, loff_t lstart,
1286 void invalidate_bdev(struct block_device *bdev);
1287 int sync_blockdev(struct block_device *bdev);
1289 static inline void invalidate_bdev(struct block_device *bdev)
1292 static inline int sync_blockdev(struct block_device *bdev)
1297 int fsync_bdev(struct block_device *bdev);
1299 int freeze_bdev(struct block_device *bdev);
1300 int thaw_bdev(struct block_device *bdev);
1302 struct io_comp_batch {
1303 struct request *req_list;
1305 void (*complete)(struct io_comp_batch *);
1308 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { }
1310 #define rq_list_add(listptr, rq) do { \
1311 (rq)->rq_next = *(listptr); \
1315 #define rq_list_pop(listptr) \
1317 struct request *__req = NULL; \
1318 if ((listptr) && *(listptr)) { \
1319 __req = *(listptr); \
1320 *(listptr) = __req->rq_next; \
1325 #define rq_list_peek(listptr) \
1327 struct request *__req = NULL; \
1328 if ((listptr) && *(listptr)) \
1329 __req = *(listptr); \
1333 #define rq_list_for_each(listptr, pos) \
1334 for (pos = rq_list_peek((listptr)); pos; pos = rq_list_next(pos)) \
1336 #define rq_list_next(rq) (rq)->rq_next
1337 #define rq_list_empty(list) ((list) == (struct request *) NULL)
1339 #endif /* _LINUX_BLKDEV_H */